Democracy Defended

This page intentionally left blank
Democracy Defended
Is there a public good? A prevalent view in political science is that
democracy is unavoidably chaotic, arbitrary, meaningless, and impossi-
ble. Such skepticism began with Condorcet in the eighteenth century,
and continued most notably with Arrow and Riker in the twentieth cen-
tury. In this powerful book, Gerry Mackie confronts and subdues these
long-standing doubts about democratic governance. Problems of cy-
cling, agenda control, strategic voting, and dimensional manipulation
are not sufciently harmful, frequent, or irremediable, he argues, to be
of normative concern. Mackie also examines every serious empirical il-
lustration of cycling and instability, including Rikers famous argument
that the US Civil War was due to arbitrary dimensional manipulation.
Almost every empirical claim is erroneous, and none is normatively
troubling, Mackie says. This spirited defence of democratic institutions
should prove both provocative and inuential.
ccnnv :acki c is Assistant Professor of Political Science at the Uni-
versity of Notre Dame. He has been Research Fellow, Social and Po-
litical Theory Program, Research School of Social Sciences, Australian
National University; and Junior Research Fellow in Politics, St. Johns
College, University of Oxford.
Contemporary Political Theory
Series Editor
Ian Shapiro
Editorial Board
Russell Hardin Stephen Holmes Jeffrey Isaac
John Keane Elizabeth Kiss Susan Okin
Phillipe Van Parijs Philip Pettit
As the twenty-rst century begins, major new political challenges have arisen at
the same time as some of the most enduring dilemmas of political association
remain unresolved. The collapse of communism and the end of the Cold War
reect a victory for democratic and liberal values, yet in many of the Western
countries that nurtured those values there are severe problems of urban decay,
class and racial conict, and failing political legitimacy. Enduring global injus-
tice and inequality seem compounded by environmental problems, disease, the
oppression of women, racial, ethnic and religious minorities, and the relentless
growth of the worlds population. In such circumstances, the need for creative
thinking about the fundamentals of human political association is manifest. This
new series in contemporary political theory is needed to foster such systematic
normative reection.
The series proceeds in the belief that the time is ripe for a reassertion of the
importance of problem-driven political theory. It is concerned, that is, with works
that are motivated by the impulse to understand, think critically about, and ad-
dress the problems in the world, rather than issues that are thrown up primarily
in academic debate. Books in the series may be interdisciplinary in character,
ranging over issues conventionally dealt with in philosophy, law, history, and the
human sciences. The range of materials and the methods of proceeding should
be dictated by the problem at hand, not the conventional debates or disciplinary
divisions of academia.
Other books in the series
Ian Shapiro and Casiano Hacker-Cord on (eds.)
Democracys Value
Ian Shapiro and Casiano Hacker-Cord on (eds.)
Democracys Edges
Brooke A. Ackerly
Political Theory and Feminist Social Criticism
Clarissa Rile Hayward
De-Facing Power
John Kane
The Politics of Moral Capital
Ayelet Shachar
Multicultural Jurisdictions: Cultural Differences and Womens Rights
John Keane
Global Civil Society?
Rogers M. Smith
Stories of Peoplehood: The Politics and Morals of Political Membership
Democracy Defended
Gerry Mackie
caxniioci uxiviisir\ iiiss
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, So Paulo
Cambridge University Press
The Edinburgh Building, Cambridge cn: :iu, United Kingdom
First published in print format
isbn-13 978-0-521-82708-9 hardback
isbn-13 978-0-521-53431-4 paperback
isbn-13 978-0-511-07110-2 eBook (EBL)
Gerry Mackie 2003
2003
Information on this title: www.cambridge.org/9780521827089
This book is in copyright. Subject to statutory exception and to the provision of
relevant collective licensing agreements, no reproduction of any part may take place
without the written permission of Cambridge University Press.
isbn-10 0-511-07110-8 eBook (EBL)
isbn-10 0-521-82708-6 hardback
isbn-10 0-521-53431-3 paperback
Cambridge University Press has no responsibility for the persistence or accuracy of
uiis for external or third-party internet websites referred to in this book, and does not
guarantee that any content on such websites is, or will remain, accurate or appropriate.
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
isnx-:,
isnx-:c
isnx-:,
isnx-:c
isnx-:,
isnx-:c ,
:
o
,
To Agnes and Ren ee
What sphinx of cement and aluminum bashed open their skulls
and ate up their brains and imagination? (Ginsberg 1956, 17)
Contents
List of gures page xi
List of tables xii
Acknowledgments xiv
1 A long, dark shadow over democratic politics 1
2 The doctrine of democratic irrationalism 23
3 Is democratic voting inaccurate? 44
4 The Arrow general possibility theorem 72
5 Is democracy meaningless? Arrows condition
of unrestricted domain 95
6 Is democracy meaningless? Arrows condition of the
independence of irrelevant alternatives 123
7 Strategic voting and agenda control 158
8 Multidimensional chaos 173
9 Assuming irrational actors: the Powell amendment 197
10 Assuming irrational actors: the Depew amendment 217
11 Unmanipulating the manipulation:
the Wilmot Proviso 241
12 Unmanipulating the manipulation: the election
of Lincoln 258
13 Antebellum politics concluded 281
14 More of Rikers cycles debunked 310
15 Other cycles debunked 335
16 New dimensions 378
ix
x Contents
17 Plebiscitarianism against democracy 409
18 Democracy resplendent 432
Endnotes 444
References 450
Index 468
Figures
8.1 Single-peaked page 174
8.2 Non-single-peaked 174
8.3 Win-sets of median point 177
12.1 Single-peakedness, 1860 271
12.2 Rikers cycle, 1860 272
xi
Tables
1.1 Preference prole of three factions over three
alternatives page 6
1.2 Pairwise-comparison matrix for prole in Table 1.1 7
1.3 Another voter prole 7
1.4 Strong preference rankings over three alternatives 8
1.5 Condorcet paradox of voting 8
1.6 Summary of empirical ndings 18
3.1 Five alternatives, ve procedures, ve winners 45
3.2 Five winners: pairwise comparison matrix and
Borda count 45
3.3 Convergence of voting rules, Danish leaders 53
3.4 Some axiomatic properties of some voting rules 57
3.5 Pairwise comparison matrix to illustrate
YoungKemeny rule 58
3.6 Borda reversal 62
5.1 Probability of Condorcet winner, impartial culture,
strong preference order 96
5.2 Probability of Condorcet winner, increasing
homogeneity, three alternatives 98
5.3 Egomaniacal redistributional instability 99
5.4 Impartiality displaces partiality 101
5.5 An unbalanced cycle 118
5.6 An almost balanced cycle 120
5.7 Another unbalanced cycle 121
6.1 Violation of IIA(A) 128
6.2 Violation of IIA(RM) 129
6.3 Substantively rational to violate IIA(A) 134
6.4 The relevance of irrelevant alternatives 139
6.5 Borda manipulation, initial situation 151
6.6 Borda manipulation, rst step 152
6.7 Borda manipulation, second step 152
7.1 Contrived outcomes 159
xii
List of tables xiii
7.2 Unfair agenda setter 168
9.1 Distribution of votes, 1956 199
9.2 Rikers estimates of factions and preference
rankings, 1956 203
9.3 Pairwise comparison matrix: Riker (1982), Riker (1986) 203
10.1 Bristow and Rayner amendments compared 227
10.2 Votes on 17th Amendment compared 232
10.3 Replacement senators, from 61st to 62nd Senate 234
10.4 Rikers inference of 61st Senate vote on
17th Amendment 235
10.5 Mackies estimates of distribution of preferences
in 61st Congress 235
10.6 Mackies inference of 61st Senate vote on
17th Amendment 236
10.7 Mackies estimates of distribution of preferences
in 62nd Congress 237
10.8 Mackies inference of 62nd Senate vote on
17th Amendment 237
11.1 Datum and warrant, Wilmot proviso 244
12.1 State-level aggregation of rst-place winners,
Upper North 273
Middle America 274
Lower South 275
12.4 Pairwise comparison matrix, 1860 election 278
14.1 Rikers estimates, Agricultural Appropriations, 1958 331
14.2 Pairwise comparison matrix, Agricultural
Appropriations, 1958 332
15.1 Blydenburghs analysis, Revenue Act, 1932 338
15.2 Pairwise comparison matrix, Revenue Act, 1932 342
15.3 Neufeld et al.s account of Muscle Shoals preferences 355
15.4 Pairwise comparison matrix, Neufeld et al.s count 356
15.5 Mackies inferred rankings, Muscle Shoals 358
15.6 Summary of Mackies rankings, Muscle Shoals 360
15.7 Pairwise comparison matrix, before vote switch 360
15.8 Pairwise comparison matrix, after vote switch 361
15.9 Distribution of hypothetical PR voters 364
15.10 Aggregation of preferences by individual not cyclical 364
15.11 Aggregation of preferences by parties cyclical 364
15.12 Iowa Senate preferences, anticorporate farming 369
15.13 Cycle, Danish prime minister 371
Acknowledgments
Thanks for help and encouragement, direct or indirect, to: Tjitske
Akkerman, the late Michael Bacharach, Samuel Bowles, Geoffrey
Brennan, Alaine Chanter, Thomas Christiano, the late James Coleman,
Gary Cox, Dhammika Dharmapala, Keith Dowding, John Dryzek, David
Estlund, James Fearon, Nancy Folbre, Diego Gambetta, Elise Giuliano,
Robert Goodin, Wendy Gordon, Donald Green, Mark Hansen, Russell
Hardin, Gretchen Helmke, Roberta Hoelzle, Stephen Holmes, Ken
Hoover, Eric Humphreys, Jeffery Jenkins, James Johnson, Desmond
King, Peter Kuurild-Klitgaard, Jack Knight, Michael Kochin, David
Laitin, Eerik Lagerspetz, Christian List, Leonard McEwen, Iain McLean,
Janet McLean, David Marsh, Ian Marsh, David Mayhew, Molly
Melching, Brad Moody, Peter Morriss, Tim Mulgan, Michael Munger,
Jack Nagel, Michael Neblo, Avner Offer, Damian OLeary, John Orbell,
Shepley Orr, John Padgett, Philip Pettit, Samuel Popkin, Michel
Regenwetter, Benjamin Reilly, Stuart Romm, Susan Rose-Ackerman,
Donald Saari, Ian Shapiro, Cindy Skach, Priscilla Southwell, Alfred
Stepan, Alex Tabarrok, John Uhr, Robert van der Veen, Federico Varese,
Bruno Verbeek, Stewart Wood, Peyton Young, and Jakub Zielinski. Spe-
cial thanks to my dissertation committee, Jon Elster, chair, Bernard
Manin, and Adam Przeworski, for their inspiration, and especially for
their patience, andto my several ne teachers at the University of Chicago.
None of them is to blame for what I say. The list should be longer, and
I apologize for omissions, which are inadvertent. I learned more about
democracy from my fellow forestry workers in the Hoedads cooperative
than from anybody in academia, and I thank every person who made that
happen.
I also thank various colloquia where some of this material was
presented: American Political Science Association in convention; Social
and Political Theory Program, Brown Bag Seminar, RSSS, Australian
National University; American Politics Workshop, University of
Chicago; Workshop on Deliberative Democracy, University of Chicago;
xiv
Acknowledgments xv
Department of Political Science, Duke University; Department of
Political Science, University of Oregon; Political Economy Seminar,
Nufeld College, University of Oxford (twice); Political Theory Work-
shop, Nufeld College, University of Oxford; Philosophy, Politics, and
Economics Society, St Johns College, University of Oxford; Public
Choice Society in convention; Department of Political Science, Stanford
University; and Department of Political Science, Yale University.
Warm appreciation to these institutions for material and intellectual
support: University Fellowship, Searle Fellowship, and Mellon Fellow-
ship at the University of Chicago; the University of Chicago; the Junior
Research Fellowship in Politics, provided by the Fellows of St. Johns
College, University of Oxford; the people of Australia, whose taxes paid
for my Research Fellowship in the Social and Political Theory Pro-
gram, Research School of Social Sciences, Institute for Advanced Studies,
Australian National University.
Of the many sources for this book, I have made particular use of the
work of William H. Riker and Kenneth J. Arrow, and also: Bo Bjurulf
and Richard G. Niemi; John C. Blydenburgh; James Burnham; Melissa
P. Collie; Robert Cooter and Peter Rappoport; Robert Dahl; Bernard
DeVoto; Dwight L. Dumond; Robin Farquharson; David M. Farrell; Dan
S. Felsenthal and coauthors, Kurt Taylor Gaubatz, William V. Gehrlein,
Leo Goodman and Harry Markowitz; Donald Green and Ian Shapiro;
Bernard Grofman and coauthors, Donald Gross, Melvin J. Hinich and
Michael T. Munger; Herbert Hovenkamp; Keith Krehbiel and Douglas
Rivers; Peter Kurrild-Klitgaard; Eerik Lagerspetz; Samuel Merrill III;
Chaplain W. Morrison; John L. Neufeld and coauthors, Hannu Nurmi,
Keith T. Poole and Howard Rosenthal; Eric Redman; Donald Saari;
Charles Sellers; Amartya Sen; Kenneth Shepsle and Mark S. Bonchek;
Gerald S. Strom; Barry Weingast, and others (omissions are inadver-
tent), adapting data presented in their contributions and working with
their ideas. Where my engagement is critical, I hope it is also con-
structive. I am indebted to the foundation they have provided. Precise
sources are referenced in notes where these debts occur, and in the books
bibliography.
Before graduate school I operated in a competitive political environ-
ment where argument was harsh but friendly. As a result, earlier drafts of
this material were in part too polemical for the academic setting, and I
regret that. My thanks to several people, and especially to one eloquent
reviewer, who convinced me to reform permanently my rhetorical habits.
Further, I want it understood that my criticisms of arguments imply no
personal disrespect for the thinkers who authored them. I agree with
xvi Acknowledgments
Jevons (1871, 275276), that:
If, instead of welcoming inquiry and criticism, the admirers of a great author
accept his writings as authoritative, both in their excellences and in their defects,
the most serious injury is done to truth. In matters of philosophy and science,
authority has ever been the great opponent of truth. A despotic calm is usually
the triumph of error. In the republic of the sciences, sedition and even anarchy
are benecial in the long run to the greatest happiness of the greatest number.
I have tried to avoid errors, but I discover more of my own every time
I revise the manuscript. All scholars err, despite their best efforts. My
purpose in this volume is not the allegation of error for its own sake, but
rather to show that a pattern of errors lies behind the irrationalist view of
democracy.
Portions of Chapter 2 appeared in abbreviated form in Gerry Mackie,
All Men are Liars, in Jon Elster, ed., Deliberative Democracy, Cambridge:
Cambridge University Press. Bits and pieces of this material are included
in a brief essay, Saving Democracy from Political Science, in Robert
Dahl, Ian Shapiro, and Jose Cheibub, eds., The Democracy Sourcebook,
Cambridge, MA: MIT Press, 2003. This volume is a revision and ex-
pansion of my Ph.D thesis, Is Democracy Impossible? A Preface to
Deliberative Democracy, University of Chicago, 2000. Otherwise the
material is unpublished elsewhere.
When I was a small child living in the country outside the small lum-
ber town of Coquille, Oregon, USA, my mother, Agnes I.H. Mackie,
drove me to the library every week, and otherwise always encouraged
my aberrant intellectual inclinations. I remember exactly and vividly how
delighted she was when I read out my rst words. I dedicate the vol-
ume to her memory, and to my mother-in-law, Ren ee Heiman, who has
consistently supported my son Brendan and I through lifes difculties.
The publisher has used its best endeavours to ensure that the URLs for external websites
referred to in this book are correct and active at the time of going to press. However, the
publisher has no responsibility for the websites and can make no guarantee that a site will
remain live or that the content is or will remain appropriate.
1 A long, dark shadow over democratic politics
Democracy and the intellectuals
Democracy is on the march in the world today. By democracy I mean
something like free and equal people associating and communicating in
public spheres, informed by liberal presuppositions, and governed polit-
ically by representative institutions based on wide suffrage and contested
elections. I do not say that democracy is victorious in the world today,
because its reign is fragile in the developing world, is awed in the devel-
oped world (especially in the United States), and is barely emergent on
the international scene. Evaluation should be a comparative enterprise,
however, and most people aware of the alternatives believe that they are
better off under democracy, and democracy is more widely spread now
than it has ever been before.
There were a handful of developing democracies a hundred years ago
(Dahl 1989, 240). Democratic aspirations ared in continental Europe
and areas under its inuence as World War I came to an end, but
Communism and then Fascism smothered the democratic ame. Fas-
cism was discredited as World War II came to an end, and also political
imperialism went into decline, only to be replaced by the realpolitik of the
Cold War. The Communists were glad to extend their tyranny to broad
new territories, and the democracies found it expedient to justify tyran-
nies among their subordinate allies. Meanwhile, Fascism was dismantled
in Mediterranean Europe in the late 1970s, and the democratization of
Spain and Portugal strengthened democratic forces in Latin America in
the 1980s. The fall of the Communist regimes in Eastern Europe in
1989, and then in the Soviet Union, conrmed a trend to democratiza-
tion on a global scale. Most civil wars in Latin America came to an end.
Apartheid was dismantled in South Africa. Authoritarian Marcos fell in
the Philippines, Suharto in Indonesia. The theocracy in Iran came under
democratic pressure. There are no dramatic democratic breakthroughs
in the Arab world, however, or with respect to the IsraeliPalestinian con-
ict. In middle Africa one-party and military regimes are less common,
1
2 Democracy Defended
but corruption, poverty, massacre, and war are as grievous as ever. The
democratic student movement in China was crushed by the Tiananmen
Square massacre in 1989.
I do not know why, but from the beginning academics have tended to
be more disdainful of democracy than are, say, the demos (the people).
Platos hatred for democracy is no secret. In our times, Almost as soon
as representative democracy on a large scale appeared in Europe . . . there
were misgivings about it, especially among intellectuals on both the Left
and the Right (Plamenatz 1973, ix). Victorian England pioneered mass
democracy in Europe, and pioneered in its denunciation: where Plato
opposed democracy on the ground that it produced spiritual anarchy
in individuals, Carlyle, Ruskin, Arnold, Stephen, Maine, and Lecky op-
posed democracy on the ground that it led to social anarchy, according to
Lippincott (1938, 5). The followers of Marx and Lenin damned democ-
racy as a bourgeois sham, and predicted scientic administration and
the withering away of politics in the communist future (see Schwartz
1995). Plamenatz refers to the academic attack on democracy by liber-
als Mosca, Michels, and Pareto, whose debunking of democracy provided
intellectual suckling to fascism. The US had more of a democratic tradi-
tion, personied by Dewey. Deweys most inuential rival was Lippmann,
who argued that the citizenry is ignorant and that experts must rule in
spite of the democratic fallacy (Wiebe 1995). In Europe during the
interwar period Lindsay (1935) and Barker (1951) were virtually alone
as academic defenders of democracy. In the period after World War II,
an exhausted conformism in American culture was accompanied by an
empirical democratic theory that apotheosized the benecial apathy
of the citizenry, and by positivistic animosity to normative theory; Dahl
(e.g., 1956) was nevertheless a milestone in democratic theory. In this
period, although little good was said about democracy, not much bad
was said about it either. The revival of liberal political theory following
Rawls (1971) was kinder to democracy, but was much more liberal than
democratic: for Rawls (1993, 231240), the Supreme Court is the exem-
plar of public reason, not the parliament, not the people. After Habermas
(1984; 1987), an emphasis on the transformation rather than the mere
aggregation of preferences stimulated wider academic interest in democ-
racy (Elster 1986b; 1998). A robust normative democratic theory, pri-
marily but not exclusively on the theme of deliberation, is beginning
to appear.
Although democratization is the main trend in the world today, the
main intellectual trend in American political science is the view that
democracy is chaotic, arbitrary, meaningless, and impossible. This trend
A long, dark shadow over democratic politics 3
originated with economist Kenneth Arrows impossibility theorem, which
was applied to politics by the late William Riker, political scientist at the
University of Rochester. The earlier academic attack on democracy by
Mosca, Michels, and Pareto was revived with fashionable new methods.
Riker had great organizational resources, and used them to promulgate
a particular interpretation of Arrows theorem, to further elaborate a
doctrine he called positive political theory (scientic, rather than
ethical), and to recruit and place his students far and wide.
Riker calls populist any democratic theory which depends on a system-
atic connection between the opinion or will of the citizens and public
policy, and liberalist any democratic theory which requires only that vot-
ing result in the randomremoval of elected ofcials. Riker rejects populist
democracy as infeasible, and offers his liberalist democracy in its place.
What almost everyone means by democracy is what Riker calls populist
democracy; and, I shall argue, Rikers liberalist alternative fails, descrip-
tively and normatively. Thus, I amtempted to label his doctrine antidemo-
cratic. I believe that it is antidemocratic in consequence, whether or not it
is antidemocratic in spirit. But to use such a label throughout this volume
would be tendentious. To call his doctrine antipopulist, though, is to beg
the question in his favor: the word populism has many negative connota-
tions, and I do not mean to defend such things as Peronism, short-sighted
policy, or mob rule. Since Rikers claim is that in the political sphere the
rational individual opinions or desires of citizens cannot be amalgamated
accurately and fairly, it is apt to describe his doctrine as one of democratic
irrationalism. Rikers irrationalist doctrine emphasizes principled failings
of democracy and recommends a constitutionalist libertarianism and the
substitution of economic markets for much of political democracy (Riker
and Weingast 1988).
Displaced by the forces of economic globalization, I came to graduate
school in midlife froma background as a founder and an elected leader of
a large forestry workers cooperative movement, as a lobbyist for forestry
workers with state and federal administrative and legislative agencies, as
a litigant for forestry workers, as an organizer of issue and candidate
electoral campaigns, as policy aide to an elected ofcial at the apex of a
large county government, and as a political journalist. I was quite ab-
bergasted by the irrationalist dogma I encountered in the political science
literature. The elegant models of impossibility and disequilibrium I was
taught bore no relation to my democratic experiences. I am not one of
those who holds that every human life is best fullled in politics, but I
knowthat my life was best fullled in that activity. Although in democratic
politics I had seen plenty of crazy things, some inexplicable, and had been
a hard operator, I had seen nothing that supported the irrationalist mod-
els and interpretations of Riker and his followers; and I had seen more
crazy things happen in the economy than in politics. At that point I did
not know why the models were mistaken, but I did know that if the mod-
els do not t the facts, then it is the models that must go; my political
experiences had made me suspicious of those who belittle empiricism. I
had already struggled against antidemocratic leftist doctrines in my own
mind and in my political environment, and rightist doctrines of the same
consequence aroused my suspicions. I am afraid that younger students,
without the experience and condence that I had, tend to accept the ir-
rationalist models, which are transmitted with professorial authority and
sometimes by means of hasty and mystifying formalisms.
One day in graduate school I was talking with someone who knew a
great deal about China. I asked him what he thought about the student
movement for democracy there. He replied that Arrow and Riker had
shown that democracy is arbitrary and meaningless, and that what China
needed was paternalistic dictatorship by the Communist Party. I was
dumbfounded. The models are wrong! I said. How are they wrong?
he asked. I could not answer him then, but I had learned something
important: not only is positive political theory empirically erroneous,
it can have dangerous consequences. The proposition that democratic
voting is arbitrary and meaningless can be used not only to justify a
constitutional libertarianism such as Rikers, it can also be used to jus-
tify a dictatorship that appeals to the values of stability and order. The
irrationalist doctrine is taught in Americas leading political science de-
partments, law schools, and economics departments. Students absorb
these teachings, and then move on to join the political and economic
elites of the world. I shudder to think of the policies demanded in the
international consultancies and nancial agencies and the national trea-
sury departments of the world by people who were taught the ndings of
Arrow as interpreted and expanded by Rikers school of thought. I worry
that authoritarian movements might nd comfort in Rikers (1982) ir-
rationalist credo, Liberalism against Populism. One purpose of my work
here is to show that Rikers irrationalist doctrine is mistaken, and thereby
to restore democracy as an intellectually respectable method of human
organization.
I have sketched the progress of democracy in the world, an ongoing
academic disdain for democracy, and my motivations for countering the
current version of the academic attack on democracy. Next, I introduce
the problems of voting that inform the irrationalist view. After that, I
provide a sample of quotations from the literature in order to establish
that there is a trend to democratic irrationalism in academic opinion.
Problems of voting: the basics
This section is an introduction to the problems of voting. We start with
majority rule. Majority rule doesnt always report a winner with more
than two alternatives, so we might turn to plurality rule. Plurality rule
might pick a winner that a majority of the voters is against, so we look
for other methods. The Borda method counts the number of times an
alternative beats all other alternatives, but it violates a condition called
the independence of irrelevant alternatives. The Condorcet method says
to pick the alternative that beats all others in pairwise comparison. The
Condorcet method might lead to the paradox of voting, however: no alter-
native wins, called cycling. The Arrow theorem is a generalization of the
paradox of voting. If there is cycling, unfair manipulation of the outcome
by agenda control and by strategic voting is also possible. Different meth-
ods of voting can yield different social outcomes fromthe same individual
preferences.
Ordinary majority rule seems to be the most natural, or commonsen-
sical, way of voting. A majority is made up of more than half the voters.
Often a majority-rule vote is taken over two alternatives; for example, in
a committee a proposal is made to alter the status quo, or often there
are only two candidates in an election. When there are two alternatives,
majority rule will deliver a winner, except when there is a tie. A tie can be
decided by some convention, such as a bias to the status quo, recounting
of the votes, or ipping a coin. Everyone is familiar with ordinary majority
rule.
When there are three or more alternatives there can be problems with
majority rule. If there are three candidates, and none receives a majority,
then there is no winner, and the method is incomplete. Perhaps without
too much thought we might turn to plurality rule as a simple extension of
majority rule: whoever gets the most votes, even if short of a majority, is
the winner. We might not notice the defects of plurality rule because, as it
happens, plurality rule tends to strategically deter more than two serious
candidates from the eld. If there are ve candidates, two of those will
be seen as most likely to win the election, and many voters will cast
their votes so as to decide between the top two rather than waste their
vote on expressing a preference for one of the likely losers. Candidates
interested in winning the election, knowing this tendency among voters,
tend not to enter the race unless they are likely to be contenders. These
are tendencies, not certainties, and I only mention them to explain why
we dont see too many plurality elections with more than a few serious
candidates, and that this may blur the distinction between majority rule
and plurality rule in our minds.
Table 1.1. Preference prole of three
factions over three alternatives
140 235 325
1st A C B
2nd B B C
3rd C A A
There can be a problem with simple plurality rule, however. Suppose
that there are three candidates A, B, and C in an election, and 100 voters.
For simplicity, everyone has strong preferences (denoted by >, meaning
that voters are not indifferent over any alternatives). Faction 1 is made up
of 40 people, and ranks the candidates A > B > C. Faction 2 is made
up of 35 people and ranks the candidates C > B > A. Faction 3 makes
up 25 people and ranks the candidates B > C > A. It will help to display
the preference rankings. With plurality rule, everyone casts a vote for
their rst-ranked alternative. With the prole of voters preferences in
Table 1.1, A would win by plurality rule, even though 60 percent of the
voters are against A. If election were by plurality rule, Factions 2 and 3
might anticipate this outcome and unite their forces on candidate C, who
then would win, showing again the tendency to two candidates under
plurality rule. The tendency is imperfect, or the election might be among
alternatives that dont respond strategically, and in such circumstances it
seems undesirable that A would win the election, as Margaret Thatcher
did in these circumstances.
Borda wrote on the theory of elections in 1784 (see Black 1958;
McLean and Urken 1995). Borda noticed this defect with plurality rule,
and proposed his method of marks, which we shall call the Borda count, to
remedy the defect. Borda thought we should count whether alternatives
are ranked rst, second, third, and so forth. He proposed that if there
were, say, three alternatives, then we would assign two points to each
voters rst-ranked preference, one point to her second-ranked prefer-
ence, and zero points to her third-ranked preference. For the prole in
Table 1.1, Alternative A gets 2 40 + 0 35 + 0 25 = 80 points. Al-
ternative B gets 1 40 + 1 35 + 2 25 = 125 points, and is the Borda
winner. Alternative C gets 0 40 + 2 35 + 1 25 = 95 points. The
full Borda ranking is B > C > A (125 for B > 95 for C > 80 for A). In
a pairwise-comparison matrix, as in Table 1.2, we display the alternatives
by row and by column, and the cell entry is the number of votes the row
entry gets against the column entry. Alternatives dont get votes against
themselves, so those cells are empty. Bordas method counts the number
Table 1.2. Pairwise-comparison matrix for
prole in Table 1.1
A B C Borda
A 40 40 = 80
B 60 65 = 125
C 60 35 = 95
Table 1.3. Another voter prole
151 235 314
1st A C B
2nd B B C
3rd C A A
of times that an alternative beats all other alternatives, and the Borda
score is also the row sum of the entries in the matrix.
Condorcet, another French thinker, wrote on the theory of elections
in 1785 (see also McLean and Hewitt 1994; McLean 1995). Condorcet
proposed as a criterion that the alternative that beats all other alternatives
in pairwise comparison should be the winner. In our example, examining
the italicized cells in the matrix, B >A, B >C, and C >A, or B >C >A.
In this example (and in most practical circumstances) the Condorcet win-
ner and the Borda winner coincide. They need not, however. Condorcet
objected to the Borda method on the ground that it is possible for it to
violate a condition that later came to be called the independence of ir-
relevant alternatives. Assume the prole in Table 1.3. By the Condorcet
method, the social ranking is A > B > C, the same as the ranking of the
faction with the slender majority of 51. Observe, however, that A is the
last choice of 49 voters. The Borda method takes that into account and
reports a social ranking of B > A > C. The dispute is this: Condorcet
insists that in pairwise comparison A beats every other alternative, Borda
insists that B gets more votes over every other alternative than does any
other alternative. The Borda method violates the independence condition
because in deciding the social ranking between two alternatives Xand Y it
takes into account individual rankings of alternatives other than Xand Y,
such as between X and Z and between Y and Z. To comply with the in-
dependence condition, for example for faction 2, we can count that an
individual ranks C > B, that she ranks B > A, that she ranks C > A, but
not that she ranks C > B > A.
Table 1.4. Strong preference
rankings over three alternatives
1. A > B > C 4. C > B > A
2. A > C > B 5. B > C > A
3. C > A > B 6. B > A > C
Table 1.5. Condorcet paradox of voting
Huebert Deuteronomy Louis
1st A B C
2nd B C A
3rd C A B
There is also a problem with the Condorcet method, however, known
as Condorcets paradox of voting. Suppose there are three (or more) al-
ternatives and two (or more) voters. Given three alternatives, there are
six possible strong preference rankings, shown in Table 1.4. Given three
voters, one each with cyclical rankings 1, 3, and 5 (or with 2, 4, and 6),
the result of voting by the Condorcet method over three alternatives is
inconsistent, that is, A beats B, B beats C, and C beats A. Suppose that
the Duckburg Troop of the Junior Woodchucks have misplaced their
Guidebook (which has a section on democratic decision making), and
are deciding on how to spend their treasury over three alternatives, as in
Table 1.5. Huebert and Louis favor Aover B, Huebert and Deuteronomy
favor B over C, and Deuteronomy and Louis favor C over A. The collec-
tive choice cycles over A > B > C > A. Arrows possibility theorem can
be understood as a generalization of Condorcets paradox, applying not
just to simple voting but to any social welfare function that aggregates
individual orderings over alternative social states. The Arrow theorem
requires that the social ranking be transitive, not intransitive as is the cy-
cle. The Borda method would count the cyclical prole in this paradox
example as a tie, AB C (denotes indifference), and thus would not
report an intransitive social ranking, but the Arrow theorem also requires
that a voting rule not violate the independence of irrelevant alternatives
condition, thus disqualifying rules such as the Borda count. Historically,
Arrows theorem is the consequence of noncomparabilist dogma in the
discipline of economics, that it is meaningless to compare one per-
sons welfare to anothers, that interpersonal utility comparisons are
impossible.
Cycling is one problem with Condorcet voting. A second, and related
problem, could be labeled path dependence. What if there were rst a vote
between A and B, which A wins, and second a vote between A and C,
which C wins? It seems that we have voted over all three alternatives and
that we have a winner, C. We neglected, however, to vote between C and
B, which B would win, and which would have disclosed the cycle to us.
Unless we take pairwise votes over all alternatives we might not notice
the cycle, and normally we dont take all pairwise votes. To make things
worse, what if Louis controlled the agenda, and arranged for that order
of voting, A against B, and then the winner against C? Then Louis would
have manipulatively brought it about that his rst-ranked alternative, C,
won, arbitrarily, and voters Huebert and Deuteronomy might even not
have noticed.
A third problem is strategic voting. Suppose again that we have a cy-
cle as above, and an agenda as above, A against B and then the winner
against C. Then Huebert would have an incentive to vote strategically
in the rst round: rather than sincerely voting for A over B, Huebert
strategically votes for B over A. B wins the contest in the rst round, and
beats C in the second round. By voting strategically, Huebert has avoided
the victory of his third-ranked alternative C and brought about the vic-
tory of his second-ranked alternative B. Inaccuracy is a fourth problem.
I showed already that the Borda and Condorcet procedures can select
different social outcomes from the same prole of individuals prefer-
ences. If apparently fair voting rules each select a different public good
from the same voter prole, then arguably the public good is arbitrary.
Inaccuracy, agenda control, and strategic voting also raise the possibility
that a social outcome might tell us nothing about the sincere individual
preferences underlying the outcome. Based on these and further con-
siderations, Rikers hypothesis is that democratic politics is in pervasive
political disequilibrium.
These are the basics. For those new to these topics, be assured that
they will be presented more slowly and in greater detail as we proceed.
A sampling of the literature
Those unfamiliar with the particular intellectual subcultures may doubt
my claim that there is a trend to democratic irrationalism in academic
opinion. To establishmy claim, I offer what I shall refer to inthe remainder
of the volume as a hall of quotations, an unconventional but I hope useful
method of exposition. The people we shall hear from are in economics,
sociology, history, legal theory, political science, and philosophy; they
are anarchists, socialists, liberals, or libertarians; some are my teachers,
colleagues, or friends. We begin with an essay introducing a recent survey
of the state of the political science discipline:
r
The fall of the Weimar Republic and, more broadly, the col-
lapse of many other constitutional democracies with the rise
of fascism and bolshevism in the interwar period alerted the
[political science] discipline to the terrible consequences of
unstable democracies. Later, Arrows impossibility theorem, a
key instance of incisive analytical work on the core problems
of liberal regimes, set forth the theoretical challenge in stark
terms. Instability is an immanent feature of liberal democracy.
Under broad conditions, majority rule leads to the cycling of
coalitions and policy; only nondemocratic practices can allevi-
ate this deep tendency, convoking a tradeoff between stability
and democracy. (Katznelson and Milner 2002, 1718)
r
At its most extreme, Arrovian public choice predicts that liter-
ally anything can happen when votes are taken. At its most cyn-
ical, it reveals that, through agenda manipulation and strate-
gic voting, majoritarian processes can be transformed into the
equivalent of a dictatorship. In a more agnostic mode, it merely
suggests that the outcomes of collective decisions are probably
meaningless because it is impossible to be certain that they are
not simply an artifact of the decision process that has been
used. (Mashaw 1989, 126127)
r
interpersonal comparison of utility has no meaning . . . If we
exclude the possibility of interpersonal comparisons of utility,
then the only methods of passing from individual tastes to so-
cial preferences which will be satisfactory and which will be
dened for a wide range of sets of individual orderings are ei-
ther imposed or dictatorial. (Arrow 1963/1951, 8, 59)
r
This clearly negative result casts doubt on all assertions that
there is a general will, a social contract, a social good,
a will of the people, a peoples government, a peoples
voice, a social benet, and so on and so forth. (Feldman
1980, 191)
r
Aristotle must be turning over in his grave. The theory of
democracy can never be the same . . . what Kenneth Arrow
proved once and for all is that there cannot possibly be
found. . . an ideal voting scheme. The search of the great minds
of recorded history for the perfect democracy, it turns out,
is the search for a chimera, for a logical self-contradiction.
(Samuelson 1977, 935, 938)
r
How can we dene and give expression to the collective wishes
of a community? Arrows argument shows that our intuitive cri-
teria for democratic decision cannot in fact be satised. . . Put
crudely, what Arrow has done is to show that strict democracy
is impossible. (Runciman 1963, 133.)
r
Almost anything we say and/or anyone has ever said about what
society wants or should get is threatened with internal inconsis-
tency. It is as though people have been talking for years about
a thing that cannot, in principle, exist . . . The central result is
broad, sweeping, and negative. Paul Samuelson rates it as one
of the signicant intellectual achievements of this century . . . It
certainly weighed heavily in the decision to award K.J. Arrow
the Nobel prize in economics . . . the cycle is the case and not
the exception. . . the phenomenon is pervasive . . . If the con-
cepts, which help us speak about how we feel whole societies,
polities, and even worlds should behave, do not work at all for
the simple case of a society with a handful of people with just a
few alternatives, then perhaps we apply them at the global level
only because we really do not understand them. . . the concept
of social preference itself must go. (Plott 1976, 512, 514, 517,
525)
r
It is not stating the case too strongly to say that Arrows the-
orem and the research that it inspired wholly undermine the
general applicability or meaning of concepts such as the public
interest and community goals. (Ordeshook 1986, 65)
r
what Arrow showed, with as much rigour as any human sci-
entist could conceivably demand, was that the programme of
an educated citizenry deciding social values . . . did not make
sense. (Tuck 1993, 79)
r
there is no universally workable way for aggregating individ-
ual interests, preferences, or values into collective decisions.
A positive implication of this nding is that no government of
a complex society is likely to be coherently democratic . . . A
normative implication of this lesson is that political theory
cannot be grounded exclusively in democratic procedural val-
ues . . . This is not to say that the democratic, majoritarian
urge is wrong . . . But it is nevertheless conceptually incoher-
ent. (Hardin 1993, 169170)
r
In fact it turns out that majority rule is fatally awed by an
internal inconsistency which ought to disqualify it from con-
sideration in any political community whatsoever . . . the incon-
sistency of the voters paradox infects virtually every method of
social choice which can lay a reasonable claim to being demo-
cratic. . . . There would appear to be no alternative but to em-
brace the doctrine of anarchism and categorically deny any
claim to legitimate authority by one man over another. (Wolff
1970, 59, 63, 72)
r
Arrows contribution provides incontrovertible support for
market process and encouragement for those who seek to con-
strain the range of collective choice to the limited functions of
the minimal state. (Rowley 1993, xiii)
r
One general approach to [the puzzle of why the majority will
should be constitutionally constrained] is to deny that it is at all
puzzling . . . by denying that there exists any meaningful sense
in which any process could even hope to reect any such
thing as the will of the majority, given the well-known the-
orem for which Kenneth Arrow received his Nobel Prize in
Economics . . . At the least . . . the analysis puts the burden of
persuasion on those who assert that legislatures (or executives)
deserve judicial deference as good aggregators of individual
preference. (Tribe 1988, 12)
r
Judicial review is often defended as the only way to escape the
potential tyranny of the majority, but it simultaneously creates
the potential for the tyranny of the judges. The general func-
tion of constitutional theory has been to specify how judicial
reviewcan exist without becoming judicial tyranny. The Arrow
theoremmetaphor suggests that constitutional theory must fail
in that task. (Tushnet 1988, 1617)
r
The idea that there is a social decision that can satisfy ev-
eryone has been annihilated by Kenneth Arrow, who in his
impossibility theorem has demonstrated that no social deci-
sion can amalgamate the diverse preferences of a group in the
way a single individual can amalgamate his own. Thus, theo-
retical economics, in its denial of a communal welfare func-
tion. . . undermines the application of rationality to public de-
cisions . . . WilliamH. Riker has . . . shown that . . . amendments
might be adopted which are not favored by a majority
without this fact ever being known! (Bell 1974, 365,
307308)
r
William Riker is one of the most inuential political scientists
at present writing on the theory and practice of democracy.
(Weale 1984, 369)
r
Rikers later theory of democracy can be viewed as a system-
atic attempt to work out the implications for the theory of
democracy of Arrows general impossibility theoremwithin the
theory of social choice. (Weale 1995, 377)
r
accurate preference aggregation through politics is unlikely to
be accomplished in the light of the conundrums in develop-
ing a social welfare function (Riker 1982; Arrow 1963/1951).
Public choice theory has shown that cycling problems, strate-
gic and manipulative behavior, sheer chance and other factors
make majoritarianism highly unlikely to provide an accurate
aggregation of preferences. (Sunstein 1988, 335)
r
In the light of social choice theory, as argued particularly by
Riker (1982), the democratic process would not converge to a
unique welfare maximum even if one existed. The reasons are
those offered by Arrow (1993/1951): There is no procedure
for aggregating preferences that would guarantee a unique out-
come. Hence, one cannot read voting results as identifying any
unique social preference. (Przeworski 1991, 17)
r
Particularly great attention has been paid to equilibria in the
subeld of rational or public choice. One depressing conclu-
sion has arisen from this work: In politics, unlike in economics
or the natural sciences, virtually no naturally occurring equi-
libria exist. This has distressed a number of workers in the
eld, including its great guru, the late William Riker. For this
nding means by implication that, in politics, almost anything
(theoretically) can happen at almost any time, as equilibria are
disrupted with virtually no advance warning. Two examples of
this process, of fundamental importance to the course of world
history in the twentieth century, can be cited here: the post-
1928 Nazi surge among major parts of the German electorate
an essential condition for the elite decisions that brought Hitler
to power in 1933 and the abrupt and wholly unpredicted col-
lapse of the Soviet Union in 1991 and 1992. (Burnham 1999,
2250)
r
The most inuential social choice theorist after Arrow is
William Riker, who is also founder of the Rochester School of
rational choice theory, which now dominates the pages of the
American political science disciplines leading journals. Riker
radicalized social choice theory to attack any notion of authen-
tic democracy, particularly what he called populism. . . . Not
all social choice theory has this radically anti-democratic polit-
ical cast, but within the discipline of political science the most
inuential strand is indeed that associated with Riker and his
followers. (Dryzek 2000, 3536)
r
the rhetorical convention of discussing the majority makes
no sense. When there exists a modest diversity of preference,
which is, after all, the bare necessity for political controversy,
then cycles are ubiquitous there are too many majorities.
The actual social state chosen by the legislature is determined,
not by some process that yields an alternative presumably bet-
ter than all the rest, but by the order in which the alterna-
tives arise for a vote. The absence of an equilibrium implies
that the person in control of the agenda (e.g., a committee
leader) can bias legislative choice in favor of his or her most
preferred alternative. Thus, there is a fundamental arbitrari-
ness to social choice under majority rule . . . Similarly, strategic
voting, typically secret, is always possible . . . Although strate-
gic voting occurs often, it is hard to discover . . . All of this
shows that the notion of a will of the people has no mean-
ing . . . In modern political science . . . electoral majorities are
seen as evanescent, and the legislator himself as a placeholder
opportunistically building up an ad hoc majority for the next
election. . . Knowing as we do that decisions are often, even
typically manipulated, but being unsure just when manipula-
tion occurs, we are forced to suspect that every outcome is
manipulated. . . Our examples show that this problem actually
arises in practice. (Riker and Weingast 1988, 393396, 399)
r
Much of the discussion of public policy has assumed that polit-
ical solutions can improve on market failures. The model we of-
fer shows that this assumption is not justied. . . political insti-
tutions . . . often lack equilibriumoutcomes . . . political choices
typically entail preference cycles. For our purposes, the lack of
equilibrium implies that there is no basis for unambiguously
claiming that a political solution will improve or fail to im-
prove upon the market failure it sought to correct. (Shepsle
and Weingast 1984, 417, 421)
r
There is, in social life, a tradeoff between social rationality and
the concentration of power. Social organizations that concentrate
power provide for the prospect of social coherence the dic-
tator knows her own mind and can act rationally in pursuit of
whatever it is she prefers . . . Though [social organizations in
which power is dispersed] may appear fairer and more demo-
cratic to the person in the street, they may also be more likely
to be tongue-tied or inconsistent in ordering the alternatives
under consideration. . . Short of actually eliminating one of the
fairness conditions for example, by permitting dictators the
Arrow result does not evaporate . . . It is nearly impossible to
arrange for the making of fair and coherent group choices.
(Shepsle and Bonchek 1997, 6769,166)
r
The various paradoxes of collective decision making seriously
challenge the presumption that legislative changes generally
represent welfare improvements, even in the de gustibus sense
of reecting changes in public taste. Enactments that instead
reect mere cycling, or changes in the agenda setter or in po-
litical tactics, may better be viewed as random and purposeless
from the social welfare perspective. (Shaviro 2000, 68)
r
Arrows theorem casts a very long, dark shadow over demo-
cratic politics . . . All voting systems have some normative blem-
ish and all voting systems can be manipulated. Social choices
in democracy depend on the particular type of majoritarian
voting procedure used by a group, on whether voting is sin-
cere or strategic and on the order in which alternatives are
considered. . . Voting cycles, according to social choice theory,
are endemic to democracy. Social choice theory tells us that
for most policy issues, there is some coalition of actors who
jointly prefer some other outcome. Whenever they have the
power to get this outcome, the social choice may simply re-
ect their power. Stability in politics may well be an arbitrary
feature of an institutional arrangement, with losers attempt-
ing to dislodge winners of their temporary authority . . . Social
choice research shows that policy agreements in a democracy
may simply be the product of agenda manipulation. . . It seems
that we cannot validly infer anything about the preferences of
the society based on the laws produced by a legislature. Nor
can we say anything about the preferences of the society when
a policy is not produced. This has certainly raised fears among
many about the legitimacy of laws in a democracy. (Cain 2001,
111112)
Weale and Dryzek are each commenting on the irrationalist trend
rather than endorsing it. Riker and Weingast are brisk and conclusive
about the supposed incoherence of democracy, Hardin is mournful and
nuanced. Notice that people seize on the disequilibrium results in order
to promote their more favored and demote their less favored institu-
tions. Tribe uses the results to elevate the judicial over the other branches
of government; Tushnet observes that the judiciary is just as tainted.
Rowley, and Shepsle and Weingast, upgrade the market by downgrad-
ing the government; Wolff would abolish government altogether. Arrow
(1997) has recently gone on record that his theorem does not show that
democracy is impossible, since it applies to all aggregations of individu-
als preferences, whether by one branch of government or another, and,
I would make clear, whether by government or market. The irrationalist
doctrines I criticize are not Arrows, they are based on interpretations by
others of Arrows theorem.
Many inuential people suggest that democracy is impossible. The
main purpose of this book is to argue against that view.
Plan of the volume
I hope that I have established both that there is an irrationalist trend,
and that there is a long dark shadow cast over democratic politics. The
proper interpretation of Arrows theorem and related social-choice re-
sults is a serious endeavor that deserves lengthy and detailed scrutiny. It
will take a good deal of spit to displace that ocean of theory. I will argue
that the irrationalist interpretations of social choice theory are based on
unrealistic assumptions, or illustrate logical possibilities rather than em-
pirical probabilities, or emphasize remediable problems, or are outright
mistaken.
This volume proceeds in three stages. First, the theory of democratic
irrationalism is presented and criticized. Second, the empirical examples
used by irrationalists to illustrate and popularize the theory are presented
and criticized one after another. Third, briey, the theory is located in
the larger intellectual and political context. Chapter 1 surveys the practi-
cal advance of democracy, introduces the problems, and establishes that
there is a trend to democratic irrationalism in the academy. Chapter 2
argues that the irrationalist trend has wide inuence in political science,
introduces Rikers distinction between liberalism and populism, and at-
tacks as self-contradictory (among other problems) what I call Rikers
basic argument pattern. Riker repeatedly deploys the basic argument
pattern in order to show that preferences are unknowable and hence
that democracy is arbitrary and meaningless. Chapter 3 presents Rikers
argument that democracy is arbitrary because it is logically possible for
different decision rules to yield different outcomes. I counter that this is
logically possible but empirically improbable. Riker also objects that the
axiomatic approach does not justify any one unique voting rule. I respond
that the axiomatic approach considerably narrows the range of reasonable
voting rules, and that choice from among the reasonable voting rules is
not arbitrary.
Chapters 4 through 6 closely interrogate and denaturalize key assump-
tions of Arrows theorem. Chapter 4 introduces Arrows theorem, the
basis of the claim that democracy is meaningless. The theorem arises as
the consequence of the appearance of the doctrine of noncomparable util-
ity ineconomics. I showthat the cycles that are allegedto make democracy
meaningless are rare. Again the question is not one of logical possibility
but rather one of empirical probability. In Chapter 5, I examine Arrows
condition of universal domain (U). Individual preference orders resemble
one another, enough so as to avoid cycling and related problems most of
the time, which is why we observe so few cycles in the real world. Models
of constant-sum redistribution predict total cycling, but such models ne-
glect behavioral constraints that produce approximately fair outcomes
but for pathological exceptions. The few cycles that do occur should be
trivial, and any which are not trivial can be eliminated by accurate and
fair voting rules. In Chapter 6, I criticize the formal and practical argu-
ments offered in justication of Arrows condition of the independence
of irrelevant alternatives (IIA). Surprisingly, many people who support
the skeptical interpretation of Arrows theorem do so on the basis of a
misunderstanding of the content of its independence condition. I show
that violating the independence condition can be substantively rational,
and argue that the theorems conditions are methodological assumptions
rather than claims with descriptive or normative force. I scrutinize sev-
eral justications of the condition, and conclude that none is sufcient to
justify the repugnant conclusion of Arrows theorem: that social choice is
impossible except by dictatorship.
In Chapter 7, I examine the contention that strategic voting, logrolling,
and agenda control permit the undetected manipulation of outcomes.
These models of manipulation assume, however, the knowability of pref-
erences, demonstrating again the self-contradictory nature of Rikers ba-
sic argument pattern. Further, we see that the possibility of counterma-
nipulation frequently deters attempts at manipulation; and hence that
such manipulation is not frequent, harmful, or irremediable. In Chapter
8, I take up the McKelvey and Schoeld chaos theorems, interpreted
by Riker to mean that there is complete disequilibrium in multidimen-
sional issue spaces. The predictions of the chaos model fail in human
subject experiments, are perhaps impossible to test in natural settings,
and utterly lack realism. Realistic amendments to the model result in
outcomes in the normatively attractive center of preferences. Moreover,
the widespread parliamentary rule permitting a division of the question
upon the motion of any one member practically disposes of any prob-
lem. These two chapters mostly summarize existing developments in the
literature.
It is Rikers dramatic empirical illustrations of political disequilibrium,
more than his theoretical arguments, that are responsible for the wide
Table 1.6. Summary of empirical ndings
#, Cite Subject Mackie Finding
1 APSR 1958 Agricultural
Appropriations
No cycle in sincere preferences (Riker recognizes
strategic votes); best alternative won.
2 LAP APM Agenda
Experiment,
Flying Club
Asymmetric institutions (in this case agenda control
and information control) yield asymmetric
outcomes.
3 LAP APM Powell amendment Riker and others allege cycle in 1956 vote; assume
irrational voters. Votes, debates, and inferences in
1956 and 1957 show that school aid would have
failed with or without Powells desegregation
amendment. No cycle; best alternative won. Adds
to Krehbiel and Rivers 1990.
4 LAP 17th Amendment Eleven errors of fact; assumes irrational voters.
No cycles, not in 1902, not in 1911. 17th
Amendment would have failed with or without a
voting-rights rider. Passed in late 1911 due to
changed composition of the Senate. Conrms
conjecture of Green and Shapiro 1994.
5 LAP Wilmot Proviso Cycle alleged among Mexican war appropriations,
antislavery amendment and status quo. Based on
egregious misreading of Congressional Globe. No
cycle, best alternative won.
6 LAP Lincoln election No cycle. Free soil was primary issue in 1860, and
the further north the more antislavery: latitude
was attitude. Riker 1982, 230, line 2 mistaken:
many Lincoln voters ranked Douglas ahead of
Bell. Complemented by Tabarrok and Lee (1999).
(Douglas was best alternative, not selected due to
antimajoritarian design of electoral college.)
7 LAP Antebellum period Eruptions of slavery issue not due to arbitrary
manipulation of multidimensional issue space.
Dimensions highly constrained (Poole and
Rosenthal 1997). Eruptions related to disruption
of political balance following territorial
acquisitions (Weingast 1998).
8 APSR 1984
APM
Morris at the
Constitutional
Convention
Alleged cycle arises from treating similar
alternatives as identical. If alternatives properly
individuated, then no cycle.
9 APM Lincoln at Freeport Mistaken details; magic bullet interpretation of
Freeport debate now rejected by American
historians. Douglas did face a dilemma, but it was
one forced upon him by the changing preferences
of the Northern and Southern populations, not
by Lincolns discourse.
10 APM The Masters Not examined. Based on ction.
11 APM Pliny Shows that agenda control was defeated by strategic
voting: best alternative won.
Table 1.6. (cont.)
#, Cite Subject Mackie Finding
12 APM Vote trading Not on point. Trade fairly represented opinion?
13 APM Abstention Idiosyncratic.
14 APM Gerrymander Undocumentable.
15 APM Magnuson Many details mistaken, does not fully understand
the parliamentary situation. Magnusons bill
supported by vast majority; an explicit deal was
made to pass Gravels more radical amendment
in the Senate, and to withdraw it in Conference,
in order to gain Nixons attention, and exactly
that happened.
16 APM Reed, Cannon Not on point. Majorities enabled?
17 Blyden
burgh
Internal revenue
1932
Flawed logic; inconsistently applied inferences of
preference orders. No cycle, best alternative won.
18 Bjurulf
and Niemi
Scandinavian
parliaments:
hospital
Half of voters absent; evidence of strategic voting;
if only 2 out of 37 voters were strategic then
equilibrium in sincere preferences, and best
alternative won.
19 Bjurulf
and Niemi
Scandinavian
parliaments:
telephone and
telegraph
Claim that strategic voting led to rejection of best
alternative. Based, however, on unwarranted
assumption that some voters were irrational.
20 Bjurulf
and Niemi
Scandinavian
parliaments:
rie club
Agenda control countered by creative response
such that best alternative prevailed in long run.
21 Neufeld, et al. Muscle Shoals Cycle only apparent and due to bungled strategic
voting that authors recognize; sincere
preferences in equilibrium; best alternative won.
22 Lagerspetz
1997
Finnish electoral
college, 1931
Cycle, and won by non-Borda winner, but perhaps
for extraparliamentary reasons. Institution
poorly designed to deliver popular outcome.
23 Lagerspetz
1997
Finnish electoral
college, 1937
Cycle among same four candidates as 1931, but
won by Borda winner. Poorly designed institution.
24 Lagerspetz
1993
Finnish electoral
college, 1956
Non-Borda, non-Condorcet winner selected by
electoral college. Poorly designed institution.
25 Gross Iowa Corporate
Farming
Great evidence, but given that alternatives are on
one dimension, cycle unlikely in sincere
preferences.
26 Kurrild-
Klitgaard
Danish prime
minister
Cycle among three evenly tied prime ministerial
candidates in eeting poll data. Borda-winner
succeeds in actual election.
Sources: APM = The Art of Political Manipulation (Riker 1986), APSR (1958) = Riker
(1958), APSR (1984) = Riker (1984), Bjurulf and Niemi (1978), Blydenburgh (1971),
Gross (1979), Kurrild-Klitgaard (2001a), Lagerspetz (1993, 1997), LAP = Liberalism
against Populism (Riker 1982), Neufeld, et al. (1994).
popularity of his irrationalist views. That the US Civil War came about
because of a cycle, for example, is an unforgettable lesson. If the illustra-
tions fail, then so does the doctrine of democratic irrationalism. Chapters
9 through 15 tackle the stories of political disequilibrium. Sometimes the
going is tough, but my exposition and commentary is never more dif-
cult than the original material, and is often easier. My examinations are
thorough, which serves several purposes. First, I show that almost all
published and developed cycle claims are mistaken, and my claim would
lack credibility in the absence of thorough argument; remember that I am
challenging the most cherished scriptures of a dominant congregation of
scholars. Second, by example, I show how it is possible to marry an un-
derstanding of historical background creatively to methods of roll-call
analysis so as to generate new insights in political history. Third, several
of the interpretations I develop of historical events are novel and inter-
esting in their own right, for example, why Douglas lost to Lincoln in
1860.
Chapter 9 is about the Powell amendment, a desegregation rider to
a school construction aid bill in 1956. Riker believes that a cycle was
contrived based on his inference that some voters voted strategically. A
manipulation would not have been possible if all voters had voted strategi-
cally, so Rikers nding of a cycle depends on the assumption that some
actors were irrational. Rikers inference of preference rankings is mis-
taken, however. It is quite clear that the events Riker strives to explain are
a consequence of incomplete knowledge of preference rankings among
the actors. Chapter 10 is about Senate deliberations on the 17th Amend-
ment to the US Constitution, which provided for direct election of US
senators. Again Riker alleges a cycle, and again he unwittingly assumes
irrational actors. Rikers interpretation of the 17th Amendment collapses
due to gross errors of fact.
Chapter 11 begins the account of Rikers major case study on the US
Civil War. The Wilmot Proviso in 1846 sought to prohibit slavery in the
vast western territories about to be acquired by the United States. Rikers
assertion of a cycle here depends on an incontrovertibly erroneous read-
ing of the Congressional record. Chapter 12 concerns Rikers allegation
that the election of Lincoln in 1860 and the momentous events that fol-
lowed were the consequence of cyclical preferences among the voters.
Rikers apparently complex argument actually depends entirely on a sin-
gle claim that Bell, the candidate of the Upper South, and not Douglas,
the candidate of the Lower North, was second-ranked by Lincoln voters.
Riker does not warrant the claim, and the claim is contrary to the con-
sensus of historians. Douglas was the candidate favored by the median
voter in the 1860 election, and in Chapter 13 I offer a hypothesis as to
why the awed presidential election system failed to select Douglas. I go
on to criticize Rikers account of the slavery issue from 1800 to 1860,
which he intends to illustrate the possibility of contrived disequilibrium
by means of introduction of new issues and dimensions. Riker contends
that changes at the collective level on the slavery issue were due to arbi-
trary manipulation of multidimensional issue space by superior political
actors. I defend the conventional hypothesis that collective changes were
a consequence of changes in individuals views on the issue.
Rikers remaining cycling claims are debunked in Chapter 14. He
claims that there was a cycle in the US constitutional convention on
the question of the selection of the executive: arbitrary instability formed
the US Constitution. I argue that another interpretation of the record is
more plausible, and with that interpretation of preference rankings there
is no cycle. In 1958 Riker claimed to nd a cycle in US House consid-
eration of an agricultural appropriations measure. Rikers inference of a
cycle incorrectly aggregates both sincere preferences and sophisticated
votes. The nding of a cycle is based on a conceptual error, because with
the aggregation of sincere preferences only there is no cycle. The remain-
ing published and developed claims of cycles from the political science
literature are presented in Chapter 15. Blydenburgh claimed to nd a
cycle in deliberations over the Revenue Act of 1932. His analysis fails
because it is confused about which alternatives are pitted against which;
after the confusions are sorted out, there is no cycle. Bjurulf and Niemi
investigated instability and manipulation in Scandinavian parliaments,
but I show that their several inferences are defective. Neufeld et al. un-
cover an apparent cycle in US Senate deliberations over Muscle Shoals
in the 1920s, but, as they recognize, sincere preferences were in equilib-
rium and the alternative favored by the majority prevailed. Lagerspetz
produces the best evidence and argument on behalf of a cycle claim, but
his cases arise in a poorly designed institution that encourages instability
and unpopular choice. A few minor cycle claims are examined. In sum,
theoretical considerations show that cycles, disequilibrium, and harmful
manipulation are of little practical importance, and almost every pub-
lished and developed example of cycling and manipulation is called into
question.
Chapter 16 returns to the possibility of manipulation by the introduc-
tion of new issues and dimensions. Why dont we see the introduction
of thousands of issues and dimensions as a manipulative political tactic?
The answer is that such introduction is constrained by the consent of the
audience to the claim of relevance by the speaker. Furthermore, delib-
eration in multidimensional issue space can identify a central outcome
such as the intersection of the medians from each dimension. It is not
discussion, only voting, and only voting under the unrealistic assump-
tions of the McKelvey model, that leads to chaos. Disequilibrium is not
a problem of much practical importance, but I note ways in which delib-
eration could further tame disequilibrium should it be a problem. Then
I present two anecdotes from Rikers The Art of Political Manipulation
(1986) that he intends to illustrate the theme of destabilizing introduc-
tion of new dimensions. One is the debate between Lincoln and Douglas
at Freeport. The other is a controversy over the shipment of nerve gas
involving Senator Warren Magnuson. I show that Riker commits errors
of fact that undermine his cases, and argue that it is not arbitrary ma-
nipulations of multidimensional issue space but simply the distribution
of preferences in the respective populations that explains these cases.
In the end, Riker rejects populism (democracy) and accepts
liberalism dened as the mere possibility of rejecting ofcials in
an election, the theme of Chapter 17. One problem, I argue, is that
the objections Riker lodges against populism, if valid, would apply to
his liberalism as well. Further, Rikers liberalism is not the unique al-
ternative to populism (democracy). If democracy were arbitrary and
meaningless then it could be argued that superior individuals should im-
pose the objective good upon the population. I trace Rikers ideas on
democracy to Pareto, who afrmed such a policy of liberal autocracy. The
doctrine of democratic irrationalism can have dangerous consequences.
Chapter 18 argues that theoretical instabilities equally afict private or-
ganizations and the market. The Arrow theorem applies to the market,
there are market analogues to the chaos theorems, and to the Gibbard
Satterthwaite manipulation ndings. Thus, there is no basis to the argu-
ment that democracy should be minimized and the market maximized
because of the ndings of positive political theory. The chapter also an-
swers the claims made in the hall of quotations.
2 The doctrine of democratic irrationalism
Introduction
In this chapter, I claim that the Rikerian legacy is the most inuential
force in the discipline of political science, but that its reign is controver-
sial. I distinguish the doctrine of democratic irrationalism from rational
choice theory in general, and express qualied support for rational choice
theory. In this volume, I do not defend the entirety of democratic princi-
ples, only the one narrow but essential principle of the possibility of the
accurate and fair social amalgamation of individual opinions and wants.
In his Liberalism against Populism, an interpretation of the results of social
choice theory, Riker (1982) makes an apparently powerful case against
the very intelligibility of majoritarian democracy. I introduce his contrast
between liberalism and populism. What everyone else calls democracy,
Riker labels populism, a termwith pejorative connotations. Such populism
is shown by social choice theory to be impossible, he claims. In its stead,
he offers liberalism, which he denes to be the random removal of public
ofcials. This liberalism is the only democracy we can expect after social
choice theory, he says. Finally, it is not widely appreciated that Rikers
central argument against populist democracy is that the preferences of
citizens are unknowable. I begin the volumes analysis by showing that
the most central argument in the irrationalist scheme is self-contradictory
and otherwise mistaken.
Commander Riker and Starship Rochester
I have taken pains to illustrate the general reception of Arrows theorem,
and have provided some indications of Rikers inuence. Now I want
to establish further the inuence of Riker and his Rochester school.
1
At
one point, Riker had published more refereed articles in the American
Political Science Review, the premier journal of the political science pro-
fession, than any other gure (Miller, Tien, and Peebler 1996). Indeed
an editor of that journal wrote to Riker that, there is some danger of
23
turning this journal into the William H. Riker review (Amadae and
Bueno de Mesquita 1999, 281). A New Handbook of Political Science
(Goodin and Klingemann 1996) cites Riker, his student Shepsle, and
coauthor Weingast as three of the eleven most frequently referenced au-
thors discipline-wide (31), and counts Shepsle and Weingast as two of the
ten rst-ranked integrators in the eld of political science, and Riker and
his former student and coauthor Ordeshook as two of the seven second-
ranked integrators in the eld (4041). A review of the elds of social
choice theory, game theory, and positive political theory by sometime
Rochester faculty members Austen-Smith and Banks (1998, 271) fol-
lows Riker in nding that, from a normative perspective . . . any hope of
nding substantive content in the idea of a collective will with respect
to policy choice is slender indeed. Amadae and Bueno de Mesquita
(1999, 269, 271) provide a brief history of the conquest of the political
science discipline by Riker and his followers: The Rochester school of
political science, led by William H. Riker, pioneered the new method of
positive political theory . . . [which] must be acknowledged as a dominant
force in political science. Positive political theory is scientic, and it as-
sumes that humans are both rational and self-interested actors, according
to Amadae and Bueno de Mesquita (270). The movement was born when
in the 1960s the Xerox Corporation, headquartered in Rochester, New
York, richly endowed the University of Rochester to upgrade its social
science departments; Riker was hired to do that job in political science,
and over time brought the department from nowhere to one of the most
successful in the country (279). Among its other contributions, Riker
and the Rochester school used Arrows result to question the efcacy of
democratic government in producing outcomes that are somehow pub-
licly benecial, they say (286). According to the Rochester Department
webpages:
For four decades, since William H. Riker arrived at Rochester in the early 1960s,
the department has helped transform the discipline of political science . . . The
Rochester School of political science has entered the vocabulary of an entire
scholarly discipline . . . The department strives not only for leadership in advanc-
ing rational choice theory but, more broadly, for leadership in advancing the
scientic study of politics . . . We distinguish our programmatic goals from those
of other departments by our strong emphasis on positive theory and general-
ization, and by our historical commitment to (and success in) speaking to the
discipline of political science.
2
Controlling for size, a 2001 study in PS: Political Science & Politics concluded
that Rochester ranked rst in the country in productivity of its PhD alumni, as
measured by publications in leading journals. Rochester is the birthplace of a
distinctive approach to studying politics that emphasizes the development of
The doctrine of democratic irrationalism 25
formal theory and the analysis of quantitative evidence. A cursory examination of
the disciplines leading journals . . . demonstrates how the eld was revolutionized
by this development.
3
Although the Rochester school is not the majority force in the political
science discipline, it is the modal force in it.
Acover story in The NewRepublic, Revenge of the Nerds, was devoted
to Riker and his followers (Cohn 1999). I relate, but do not endorse, some
of its contents. The article is an attack on rational choice theory in general,
which Cohn identies with Riker and his Rochester school; but that is not
quite correct, there is also the Virginia school, the Bloomington school,
the left-liberals formerly centered at the University of Chicago, and indi-
viduals not otherwise classied. Cohn reports that some scholars outside
the Rochester school believe that its members want to diminish the role of
nonmembers in the discipline of political science, and that Rochester has
made considerable progress in realizing its alleged goal. Rational choice
scholarship readily yields publications: all you had to do was come up
with a complication that confounded some existing rational theory and
then derive a new, more complex equation to answer it, and as a result
Rochester school research exploded into the literature and advanced the
careers of its practitioners. James Q. Wilson complained, They dont
read Supreme Court decisions or history. They just sit around and make
models.
The insiders are said to maintain a unied front, to cite and to referee
one anothers papers, to preach that their approach is the only legiti-
mate method in political science, and to establish litmus tests for fac-
ulty hiring, according to Cohns disgruntled and anonymous informants.
Because they are not as broad-minded, they had the advantage, says
one senior scholar at Harvard. Theyd support any candidate who did ra-
tional choice, oppose any non-rational-choice scholars, Cohn relates.
4
Critics call them imperialists, colonizers, and Leninists (in the
organizational sense), and hyperbolically describe them as a cult. Cohn
says that cultists speak with reverence when they discuss their founders,
and continues:
It would only be a slight stretch to compare this reverence with the way ratio-
nal choicers talk about their movements founder, the late William Riker, and
the intellectual compound he built at the University of Rochester. Rochester is
the mother ship, Shepsle says. Its founder . . . was William Riker. Commander
Riker, as we like to call him. And Starship Rochester.
The crew of Starship Rochester respond that these views are unfair cari-
catures. Shepsle remarks that although we were all true believers, there
was no grand imperialistic design. Bueno de Mesquita says that Were
a handful of people, and that, the reason it appears to be this dominant
thrust is because the clarity of the work attracts attention. I will show
that a defect of irrationalist scholarship is that its formality obscures as
much as it claries.
In June 2000, economics students in Paris rebelled against their rigidly
formalist curriculum. Among the reasons they gave were:
1. We wish to escape from imaginary worlds!
2. We oppose the uncontrolled use of mathematics!
3. We are for a pluralism of approaches in economics!
The students opposed the use of mathematics as an end in itself, de-
plored the domination of economics by highly ideological neoclassical
theory, and they rejected what they called its repressive dogmatism.
They favored intellectual engagement with concrete empirical realities,
a pluralism of approaches, and science rather than an obscurantist sci-
entism. The rebellion, which came to be known as the Post-Autistic
Economics Movement (www.paecon.net) spread through France, the
Continent, the United Kingdom, and the United States. Parallel declara-
tions emerged at Cambridge University and at a Kansas City conference.
The French minister of culture appointed a commission to investigate
student concerns, which eventually recommended reforms endorsed by
the students.
In the same spirit, in November 2000 a new perestroika-glasnost
movement erupted in the American political science profession, criti-
cal of the hegemony of narrow rational choice theory in its professional
association and its main journal (Eakin 2000; Jacobsen 2001; Kasza
2001; Miller 2001; all linked at www.paecon.net). Kasza wrote that the
problem is hegemony, not rational choice; that the postmodernist hege-
mony in some literature and history departments may be just as suffocat-
ing as rational-choice orthodoxy is in political science. Kasza continued
that:
William Riker was fond of saying that political science was a sinking ship, and
rational choice theory was the only tugboat that might bring it to port. It is truer
to say that Rikers disciples have acted as pirates out to hijack political science to
a rather barren island. Their piracy is doomed to failure.
Changes in the association and its journal are coming about. Its 2002,
2003, and 2004 presidents are sympathetic to the movements com-
plaints, the perestroika movement is lively and organized, and there is
pressure for contested elections in the association to replace its previous
system of co-optation.
Democracy defended
The theory of democratic irrationalism is a species of the genus ratio-
nal choice theory, but I criticize the species and not the genus. The
theory is one justication for laissez-faire, or perhaps for total anarchism;
if successful, my criticism eliminates that irrationalist justication for
laissez-faire or anarchism, but not other justications. The theory at-
tacks democracy in principle, and I defend it in principle; except inci-
dentally, our discussion here is not about how to improve democracy
in practice. My defense does not pertain to the entirety of questions in-
volved in the justication of democracy, but only to one essential question
about the systematic connection of citizens opinions and desires to public
outcomes.
Green and Shapiro (1994) wrote a sustained critical evaluation of
rational-choice explanations in political science that received wide at-
tention. Their Chapter Six on Legislative Behavior and the Paradox of
Voting (98146) is most relevant to the themes I discuss here. I agree
with nearly all of its observations, which I will not duplicate here. As for
the book as a whole, I agree with Green and Shapiros descriptions of
rational choice scholarship, as far as those descriptions go, but not so
much with their evaluations. I often prefer the work of scholars who are
informed, but not enslaved, by rational choice theory; but I appreciate
any work that shows high creative intelligence. I see the study of poli-
tics as an interpretive enterprise. A plurality of methods contributes to
that enterprise, and the methods of rational choice theory are especially
but not exclusively useful for describing and explaining some collective
actions, strategically interdependent actions, and the character and con-
sequences of voting rules. Although the terms are used informally and
interchangeably, social choice theory is simply the formal description of
economic and political aggregation, and is indispensable even if some of
it is sterile and scholastic. I cannot imagine doing without noncooperative
game theory, which in the right hands yields rich insights into social life,
along with testable, and supported, predictions (for example, see Mackie
1996); although some formal work is irrelevant and some applied work is
unimaginative. Many public choice scholars advocate a rather libertarian
approach, leaving most decision making to the private market while
proposing a signicantly reduced role for government (Block 1998, 983,
emphasis added).
5
Although many rational choice practitioners believe
that their method requires the assumption of egoism, it need not, and
I wholeheartedly reject the assumption that humans are exclusively mo-
tivated by egoistic concerns. I agree with Elsters Davidsonian account
of rational choice theory in the narrow sense, as an initial assumption, in
interpreting individual human action, of the consistency of beliefs and of
desires; and I agree with his claim that there is no alternative to it as a set
of normative prescriptions (Elster 1986a).
Otherwise, rational choice in the broad sense is a catch-all term for
the use of a grab-bag of methods whose only unity is that they are for-
mal models borrowed from the discipline of economics. I do not see
that there can be an objection against formal models as such, used by
scholars who understand their assumptions and their scope, although I
fear that Walt (1999, 8) more likely than not is correct when he con-
cludes that the growing technical complexity of recent formal work has
not been matched by a corresponding increase in insight. Schumpeter,
Arrow, Buchanan, and their followers subsumed democracy to the mar-
ket model and likened the voter to the consumer. If that subsumption is
supposed to be the basis of rational choice political theory, then rational
choice is already dead, because fty years of scholarship have shown that
the analogy of the voter to the consumer is gravely misleading. In ideal-
ized market exchange self-interest has benign social consequences, but
in the idealized democratic forum self-interest has malign social conse-
quences. The connection between individual choice and individual out-
come is direct and obvious for consumers, but comparatively indirect
and obscure for voters. Practically, people tend to self-interest in the
market, and to impartiality in the forum. The Condorcet paradox of vot-
ing predicts radical instability among self-seeking voters, but this is not
observed, and the paradox of participation predicts that almost no one
will vote, because one individuals vote almost never makes a difference,
but this is not observed either.
6
If, as I argue, there is not much intellec-
tual unity to rational choice political science, there is a sociological unity
to it, captured by Amadae and Bueno de Mesquita (1999) and Cohn
(1999). The scientistic aspirations of the Rochester school are dashed
by Green and Shapiros demonstration of methodological pathologies
such as post hoc theory development, vague or slippery predictions, and
selective use of evidence. These pathologies are not necessary implica-
tions of the use of formal models in political science, however.
7
Finally,
we should consider the opportunity costs of rational-choice scholarship.
Rational choice was attractive to me as an alternative to what I considered
exhausted doctrines, such as Parsonsianism, Freudianism, or Marxism.
But if the effect of rational choice on the study of politics is to keep too
many scholars and students away from historical knowledge, contextual
understandings, exposure to multiple cultures, broadminded apprecia-
tion of human motivations, the formulation and testing of competing hy-
potheses, or the facts and feelings of politics, then its costs would exceed
its benets.
Hauptmann (1996) afrms the wide inuence of rational choice in
political science, and that there are many voices in rational choice theory.
She, too, criticizes the analogy from economics to politics (74). Like
many unbelievers she seems to be fascinated and appalled by the brutally
antidemocratic strain in rational choice theory:
On the one hand, rational choice theorists identify democracy with honoring in-
dividual choice, a norm they believe has been overshadowed by pursuing what
to their minds are the dubious goals of securing the common good or increasing
popular participation. On the other hand, they also conclude that the choices
citizens are given are not worth making because they are either too insignif-
icant individually to make any difference or are offered and counted in ways
that end up distorting the very things that were supposed to be honored. . . these
theorists . . . assign such dismal expectations to contemporary democratic systems
that they nd themselves unable to say what is valuable about them. (45)
Hauptmann criticizes the concept of choice in rational choice theory,
by reference to Aristotles usage of prohairesis and an ordinary-language
analysis of choice. She points to election of ofcials by lottery in Athens
as a democracy not based on choice as a value. Hauptmann adequately
identies shortcomings of rational choice theory, but her external cri-
tique is not persuasive, in my opinion. I offer an internal critique of the
irrationalist doctrine.
My target in this volume is not rational choice theory, but an irrational-
ist doctrine contingently associated with rational choice theory. True,
Riker is the godfather of rational choice analysis in political science, and
the University of Rochester, where he taught for more than a quarter
of a century, continues as its intellectual center (Shepsle and Bonchek
1997, viii). The doctrines I identify and the pathologies of scholarship
that Green and Shapiro identify do center on the godfather and perhaps
some of his capos, and they do make up much of the content of rational
choice theory as it is taught to students. But, to infer guilt by association
would be both descriptively and normatively wrong. There are plenty of
rational choice theorists who are indifferent to Rikers irrationalist doc-
trine, and others who have patiently developed theory and evidence that
cumulatively undermine the Rikerian monolith. There are rational choice
investigators who do marvelous empirical research, and who dutifully re-
vise theory so as to reect the data. There are many who engage in con-
structive social choice theory seeking to contribute solutions to human
problems rather than hothousing bugaboo paradoxes.
8
Although there are socialists and left-liberals who do not contest Rikers
irrationalist theory, its primary appeal is to the economic right. Pub-
lic choice theory, in the narrow sense of the term, is the assertion
by the Virginia school that government should be limited in order to
avoid damaging policies enacted by self-seeking interest groups, politi-
cians, and bureaucrats, and the assertion by the Rochester school that
government should be limited because democratic outcomes are arbi-
trary and meaningless. Shapiro (1996, 43) remarks that the implicit
counterfactual in this tradition is that private action is essentially be-
nign, rather than anarchic chaos. For the public choice tradition, says
Shapiro, collective action is only necessary in instances of market fail-
ure, but there is no rational way of organizing collective action. Remark-
ably though, as I shall later show in detail, the Arrow theorem applies
to the market, there are strong analogues to the political chaos theo-
rems in economic theory, and the market, too, is subject to strategic
misrepresentation of preferences. Rather than declaring the market to
be inaccurate and meaningless because of these logical possibilities, and
concluding that it should be severely circumscribed, economists instead
note undramatically that the models prediction does not match observa-
tions, and ask what might be wrong with the models assumptions. In this
volume, I argue against the democratic-irrationalist justication of eco-
nomic libertarianism, but not against other possible justications of the
minimal state.
It might seem curious that irrationalist theory cavils at hypothetical
and unproven deviations from the democratic ideal of equal inuence in
politics, but neglects actual and proven distortions such as the campaign
nance problem in the United States (Drew 2000). Kuttner (1996, 347)
claims that Public Choice theory is almost entirely silent on the dis-
proportionate purchase of inuence by big money. The irrationalists do
not criticize democracy in practice, I surmise, because practical defects
are remediable. The remedy may be difcult to accomplish or involve
an unacceptable tradeoff with other values, but remedies are conceiv-
able. Rikers point is not that democracy is imperfectly approximated in
practice. Riker wants to establish that democracy has defects that are irre-
mediable in principle (also see Weale 1999, 140). Therefore, my response
defends democracy in principle. I know that in practice democracy is a
messy and imperfect business, but the practical shortcomings of democ-
racy are not the topic of this study. Also, if Waldron (1999, 2) is correct,
that we have an idealized picture of judging but a disreputable picture
of legislating, and that we should develop a rosy picture of legislatures
that matche[s], in its normativity, perhaps in its naivete, the picture of
the courts, then a defense of democracy in principle will be welcomed
by some readers.
Finally, my defense does not pertain to the whole of democracy and
liberalism. Rather my attention is conned to a single question of major
importance: the possibility of the accurate and fair amalgamation of opinions
and wants. It is not just crudely populistic democracy, where the majority
could impulsively vote to massacre a minority, that requires a knowable
and systematic connection between citizens opinions and wants on the
one hand and public policy outcomes on the other. A political theory
might construct a liberalism from rst principles, and then suggest that
democratic institutions, in an auxiliary fashion, comparatively would best
approximate recommended liberal outcomes. But even that thinly demo-
cratic scheme requires a systematic connection, rather than one that is
haphazard or accidental. According to Riker, the democratic connection
is random; if so, then any liberal autocracy better than random would be
better than liberal democracy; also, if preferences really are unknowable,
then the liberal theorist could not know such things as that most people
do not want to die and that a few people want arbitrarily to kill, and thus
would lack motivation to formulate liberal principles. Further, it is not
just desire-satisfaction accounts of democratic legitimacy that require a
systematic connection. Any deliberative approach which concedes (as it
must in my view) that reasonable discussion among reasonable people
can terminate in disagreement has to propose a method of practical deci-
sion. When reason is exhausted, if it is not possible accurately and fairly
to amalgamate differing judgments as to the content of the public good,
then any deliberation short of unanimity has no practical result. Nor in
requiring a systematic connection are we committing a fallacy of compo-
sition, improperly ascribing collective desires, beliefs, and preferences to
a collective entity, if indeed that is improper. In wanting to know what
most experts think about smoking, or in wanting to afliate with a group
that settles conicts, not arbitrarily, but according to a voting rule which
systematically connects group outcomes to its members preferences, I
have committed no such fallacy.
Liberalism against Populism
Rikers (1982) volume opens with a discussion of the normative features
of democracy (adapted, with radical changes, from his 1953 book on
democracy in the US). Riker tells us that the democratic ideal is one
of individual self-realization and individual self-respect. The democratic
method is the free and equal participation of each citizen in the political life
of the community. Hitherto democratic theory has assumed that demo-
cratic ends (or ideals) can be attained by democratic means (or methods),
but Riker says that this assumption may or may not be true. Social choice
theory will help us decide whether or not the assumption is true, whether
or not democracy is attainable. The Platonic and Marxist conceptions of
justice were not attainable, he says, because in each case the means could
not reach the end. The same question should be asked of democracy as
of those conceptions of justice: do its means attain its ends?
For Riker, voting is the central act of democracy. The elements of
democracy are participation, liberty, and equality, he says. Participation
is centered on voting, but voting alone is not sufcient for democracy;
only voting surrounded by attendant institutions such as free speech and
political parties that facilitate popular choice is democratic. Participa-
tion has two purposes. The rst purpose is instrumental and negative:
participation restrains oppressive rule by subjecting ofcials to popular
judgment. The second purpose is intrinsic and positive: participation
furthers the individuals self-control both internal discipline and coop-
erative management of the physical and social environment. Self-control
supports the democratic ideal of self-realization and self-respect. Liberty
is instrumentally required to organize participation in government, and
has gone on to become an end in itself. Civil liberties, religious liberties,
and economic liberties are intrinsic ends and also facilitate self-control.
Equality, like liberty and participation, originated as an instrument of vot-
ing, according to Riker, and it too promotes self-control and self-respect.
In a society characterized by democratic justice, people are free (by
reason of democratic liberty) and have the chance (by reason of demo-
cratic equality) to seek self-respect and self-control (through some kind
of democratic participation) (8). The democratic method of voting and
its attendant institutions are meant to attain those ideals. The avoid-
ance of the public good in Rikers normative portrayal of democracy is
no accident.
There are two interpretations of voting, the liberal and the populist,
which Riker implicitly treats as exhaustive and exclusive. Warning: Rikers
usage of the two terms is idiosyncratic, polysemic, and inconsistent at cru-
cial points in his argument. The liberal view is Madisonian, Riker main-
tains at this point in his narrative, and the populist view is Rousseauvian.
Riker eventually concedes that his liberalism would not be endorsed by
Madison, and I will shortly show that his populism would not be en-
dorsed by Rousseau. The liberal view, says Riker, is that the purpose of
voting is to control ofcials, and nothing more. Madison, he writes, held
it necessary that republican government derive all of its powers directly
or indirectly from the great body of the people, and sufcient that it be
administered by ofcials who hold ofce for limited periods (Federalist
No. 39, in Hamilton, Jay, and Madison n.d. (1937)/1787). He adds that
Madison said nothing about the quality of popular decision, whether
such decision be good or bad (but Riker is wrong, as we shall discuss in
Chapter 17: for Madison, the ultimate aimof popular government is the
public good; see his Federalist No. 10, (1937)/1787). All democrats
accept the necessary condition, but only Rikers liberal democrats ac-
cept the sufcient condition, he claims. Popularness, Madisons neces-
sary condition, ensures participation and equality; and limited tenure,
Madisons sufcient condition, ensures liberty. That ofcials can be re-
placed guards against tyranny and protects liberty, Riker says. Another
danger to liberty is that ofcials may be inefcient agents of the people, he
continues. This is where liberalismand populismpart company. Populism
assumes popular competence, that the electorate is right. Liberalism as-
sumes that the electorate can change ofcials only if many people are dis-
satised or hope for a better performance. All liberalism requires is that
ofcials can be rejected by election; however, in order to avoid rejection,
ofcials do act so as to avoid giving offense to future majorities; the union
of those many possible future majorities is often most of the electorate,
hence liberal ofcials are approximately agents of the people, and thus
liberalism satises the democratic ideal of self-control, Riker concludes.
For the populist, according to Riker, liberty and hence self-control are
obtained by embodying the will of the people in the action of ofcials.
Rousseau is the exemplar, or perhaps the demon, of populism. Rousseau
posited a general will that is the objectively correct common interest of the
incorporated citizens, which is computed by consulting the citizens, Riker
claims. What the sovereign people, when speaking for the public inter-
est, want is justied because the sovereign people want it and because it
is their liberty (12). Participation in democratic decision is necessary to
liberty, according to Rikers populism, and the rules thus made must be
respected as right and proper because they embody that liberty. The hur-
ried reader might reject populism as portrayed by Riker because it is so
painfully obvious that democratic decisions can be factually and morally
mistaken. Who could be so silly as to believe that actual democratic de-
cisions by denition precisely carry out some general will and thus that
they are always true or right? The hurried reader would be attracted to
Rikers liberalism which, at this point, requires only that the decisions of
democratic ofcials approximately satisfy the opinions and desires of the
people. The hurried reader would be mistaken.
As is well known, Rousseau distinguished the general will from the will
of all. Chapter 3 of his Social Contract is titled, Whether the general will
can err:
the general will is always rightful and always tends to the public good; but it
does not follow that the deliberations of the people are always equally right. We
always want what is advantageous to us but we do not always discern it . . . There
is often a difference between the will of all and the general will; the general will
studies only the common interest while the will of all studies private interest, and
is indeed no more than the sum of individual desires. (Rousseau 1968/1762, 72)
The general will, true or right, is independent from the will of all, which
can err in attaining the general will. The will of all, actual democratic
decision, might err because it is mistaken about some fact of the mat-
ter, or might err because it aggregates partial interests rather than judg-
ments about the general interest. The younger Riker (1953, 156) knows
this:
The will of the majority is the best guide democrats have to the distinction between
political right and wrong. In the end it is their only guide. But that does not mean
that majorities are infallible or that time will not prove them wrong.
Thus, for Rousseau, and for the younger Riker, actual democratic de-
cisions variably approximate rather than dene the general will or public
good. The democratic public good is dened independently from the
output of actual democratic procedures, these days perhaps as the out-
put of an ideal procedure, or by subjective utilitarianism, or by objective
measures of welfare such as of health, nutrition, longevity, or by appeal to
ethical intuitions, or by secular extension of religious revelation; people
debate and vote among such competing conceptions of the public good.
No one, no democratic theorist and no democratic participant, equates
the entirety of actual democratic output with the democratic public good,
else disagreements about what to do would have no content, nor would
we ever be able to say (holding conditions constant) that a past decision
deserves revision. Rikers populism is an absurd doctrine, one not even
endorsed by its supposed champion Rousseau. Why is Riker tempted into
this error about Rousseau? In his rst chapter he maintains that his liber-
alism approximates the public good, and if it were sufcient for populism
only to approximate the public good, then his distinction between liber-
alism and populism would collapse. If there were no distinction between
the two, then Rikers argument against the populism he wants to attack
would apply as well to the liberalism he wants to defend; and then the
nal conclusion, according to Riker himself, would be that democracy is
wholly indefensible (241).
Add a third source of error in identifying the general will: the basic
problems of aggregating opinions and desires introduced in the previ-
ous chapter. Because different voting systems may yield different out-
comes from the same prole of individual voters preferences, he argues,
democracy is inaccurate. Different methods of aggregating individuals
xed choices may yield different group choices. Furthermore, although
one or more voting rules may be better than other voting rules in a par-
ticular setting, there is no one uniquely justied voting rule across all
settings. Finally, from the outcome of any vote it is not possible to infer
enough about citizens preferences to determine whether that outcome is
a true and fair aggregation of those preferences. Next, Riker continues,
given a xed voting system, then democracy is meaningless: the outcome
of voting is manipulable, and it is not possible to distinguish manip-
ulated from unmanipulated outcomes because of the unknowability of
private intentions underlying public actions. The spirit of the argument
is best conveyed by Condorcets paradox of voting. The Condorcet para-
dox of voting is a special case of the Arrow possibility theorem; the Arrow
theorem shows more generally that, assuming all logically possible indi-
vidual orderings over alternative social states, no method of aggregating
individuals transitive preference orderings guarantees a collective pref-
erence ordering that is transitive. Therefore, even the same method of
aggregating individuals xed choices may yield different group choices,
as the choice arbitrarily or manipulatively cycles from one alternative
to the next. Associated with the possibility of cycling are the possibili-
ties of unfairly manipulating the outcome by means of strategic voting,
agenda control, and introduction of newalternatives or dimensions to the
issue space.
The liberal interpretation of voting, says Riker, accepts that democratic
voting and discussion are inaccurate and meaningless. The only democ-
racy that withstands the scrutiny of modern political science (Riker
and Weingast 1988, 396) is the liberal institution of regular elections that
permits both the rejection of tyrannical rulers and the circulation of
leadership (Riker 1982, 253), preferably supplemented by the liberal
constraints of divided government.
The kind of democracy that thus survives is not, however, popular rule, but rather
an intermittent, sometimes random, even perverse, popular veto. Social choice
theory forces us to recognize that the people cannot rule as a corporate body in
the way that populists suppose. Instead, ofcials rule, and they do not represent
some indenable popular will. (Riker 1982, 244)
One problem with Rikers liberal justication of voting is that his
proposed constraint is no constraint at all and thus ofcials would do as
they please (see Cohen 1986, 30). True, if there is a random chance of
getting a ticket for driving faster than the speed limit, then my speed-
ing is constrained. But if there is a random chance of getting a ticket
whether or not I am driving faster than the speed limit, then I will do as
I please.
In the rst chapter of the volume, Rikers liberalism approximates the
public good (11); but in frank contradiction by the last chapter of the
volume there is no such thing as a knowable public good, the connection
between the democratic electorate and its ofcials is random, and Riker
has to admit that Madison would not endorse his conception (242243).
Rikers liberal interpretation of voting fails as a democratic method to
attain democratic ideals, to use his terms, despite his weak assertions to
the contrary. The younger Riker (1953, 106) supplies the reasons:
responsible government means that public ofcers submit themselves in elections
to the judgement of the people . . . in addition it requires that voters have a clear
sense of their power of selection or at least of their power of choice. How else
can they hold ofcers responsible unless they are aware that voting does some
good, that it is clearly related to their own personal circumstances? On the other
hand, in the ofcials themselves this general responsibility requires a continuing
consciousness of the popular, electoral sanction.
The instrumental and negative purpose of participation, to restrain op-
pressive rule, is thwarted under the liberal interpretation of voting, be-
cause the removal of ofcials is not only imperfect, but wholly random,
and ofcials are thus unrestrained. The intrinsic and positive purpose
of participation, to further the individuals self-control, is thwarted, be-
cause there is no knowable relationship between the individuals vote and
the collective outcome. Liberty is unguarded, because there is no effec-
tive restraint on ofcials. Equality is mocked, because any one decision
can be unfairly manipulated, and hence all decisions could be unfairly
manipulated, if the older Rikers argument is correct.
What if Rikers argument were incorrect? What if the preferences of
citizens were approximately knowable, what if democracy were approxi-
mately accurate, what if democracy were approximately unmanipulated?
Remember that Rikers populism requires that democratic decisions de-
ne, not merely approximate, the public good. If any problems of aggre-
gation were to result in the slightest indeterminacy or error, then populist
democracy fails entirely, Riker insists (291). We need not be cornered by
Rikers denition of populism, however. We are free to adopt a more
standard view, that it is sufcient for the outcomes of actual democratic
procedure to approximate a democratic ideal of the public good. In this
volume I will endeavor to show that, in principle, democratic procedures
adequately approximate the public good.
If the reader is a bit vague about the distinction between populism
and liberalism, some concrete examples may help. A present-day man-
ifestation of populism, according to Riker in 1982, is Marcus Raskin
of the American New Left, and a present-day manifestation of liberal-
ism is William Rusher of the American New Right. Populism is typied
by American progressivism and Franklin Roosevelts New Deal (Riker
1982, 63), and Rikers liberalism would roll back progressive-era and
New Deal reforms and reestablish the minimal state of the Gilded Age
(Riker and Weingast 1988). Riker continues his discussion of liberalism
and populism in his last chapter, and I continue my commentary in the
next to the last chapter of this volume.
Rikers basic argument pattern
Riker knows that he cannot rest his argument only on divergent outcomes
from the same prole and on the mere possibility of cycling, strategic
voting, and agenda control. The problem, despite much repetition to the
contrary in subsequent literature of the Rochester school, is that these
harmful possibilities are not shown to be empirical actualities of nor-
mative consequence. The endpoint of Rikers argument is the skeptical
assertion that populist democracy is impossible because individual pref-
erences are unknowable. Everything else is a step on the way to this
destination. As to accuracy, his nal conclusion is:
Outcomes of voting cannot, in general, be regarded as accurate amalgamations
of voters values. Sometimes they may be accurate, sometimes not; but since
we seldom know which situation exists, we cannot, in general, expect accuracy.
Hence we cannot expect fairness either. (Riker 1982, 236)
And, as to meaninglessness, his nal conclusion is:
Outcomes of any particular method of voting lack meaning because often they are
manipulated amalgamations rather than fair and true amalgamations of voters
judgments and because we can never know for certain whether an amalgamation
has been manipulated. (Riker 1982, 238)
Most of Rikers commentators miss this aspect of his argument. They are
more interested in his discussions of inaccuracy and manipulation. If I
succeed in disposing of the issue of the unknowability of other individu-
als preferences early on, that will simplify the rest of the discussion on
inaccuracy and manipulation in the remaining chapters.
Rikers several lines of attack against the populist interpretation of vot-
ing all ultimately depend on the following pattern of argument:
(1) Because of cycling, strategic voting, agenda control, and multidimen-
sional issue spaces, it is possible that the outcome of any one vote can
be manipulated.
(a) If underlying preferences must be indirectly inferred, then it is
possible in any one instance of voting for there to be undetected
manipulation;
(b) Revealed tastes, actual choices such as votes, are directly
observed, but true tastes, the underlying preferences, must be
indirectly inferred;
(c) Undetectedmanipulationis possible inany one instance of voting.
(2)
(a) If undetected manipulation is possible in any one instance of vot-
ing then it is possible in all instances taken together;
(b) Undetected manipulation is possible in any one instance of voting
(1c);
(c) Therefore, in all instances taken together, underlying preferences
cannot be inferred from votes.
(3)
(a) If underlying preferences of individuals cannot be known, then it
is impossible to aggregate such preferences;
(b) Underlying preferences cannot be known (2c);
(c) Therefore, it is impossible to aggregate underlying preferences;
any such claim is meaningless.
Notice that by substituting communication for references to voting
in (2), the amended argument would prove the meaninglessness of
all communication, including political discussion.
Riker claims not only that undetected manipulation is possible (1c)
but also that it is frequent, but the frequency claim is not essential to his
argument. Moreover, the frequency claim verges on self-contradiction:
if true preferences cannot be known (2c), then it would not be possi-
ble for any observer to estimate the frequency of manipulation. Also, the
claim that actual choices are directly observed but that underlying pref-
erences must be inferred (1b) is heuristic in some investigations but it is
not fundamental. When someone buys a Cadillac, what choice has been
revealed: a means of transportation, a status symbol, a dating ploy, a nos-
talgic memory of the buyers father, a tax dodge, a mistake? When the
only information we have is that someone refrains from buying a feasibly
available Cadillac, what choice has been revealed? As for voting, raising
ones hand might be a sincere vote, might be a strategic vote, might be a
mistake, might be a yawn and a stretch, might be the sign of a follower of
St. John the Baptist, might be a joke, might be an involuntary reex. How
do we knowwhether or not something is a choice? The more fundamental
distinction, it seems to me, is between movement and action (Fay 1996,
9295). Movements are directly observed, but actions (including forbear-
ances) must be interpreted. To infer an agents action from a collection
of movements requires initial attributions of intention and of agent ra-
tionality. In other words, if there really were an insurmountable problem
of interpretation, then not only would the inference of preference from
revealed choice such as voting fail as Riker would have it, so would the
inference of voting from mere movements: we would never know, given
Rikers requirement of certainty, whether particular motions are an act
of voting or something else altogether and thus even the liberal defense
of voting would collapse since humans would be completely opaque to
one another.
Nevertheless, we may for the sake of argument grant Riker his claimthat
preferences are inferred fromchoice (1b). His fatal error is the move from
the claim that undetected manipulation is possible in any one instance to
the claim that in all instances taken together it is therefore impossible to
infer preferences from choices (2a). I will argue that (2a) is defective and
thus that (2c) does not follow; it is false that in all instances taken together,
underlying preferences cannot be inferred from votes. What does my
conclusion do to the claim that it is impossible to aggregate preferences?
Rikers third argument is that if underlying preferences are unknowable,
then it is impossible to aggregate such preferences (3a), or if p then q. I say
that underlying preferences are knowable, not-p. In the present context,
then, I do not show generally that it is possible to aggregate preferences.
I only show here that if it were impossible to aggregate preferences then
the unknowability of preferences would not be the reason for such an
impossibility. It may be impossible to aggregate preferences for some
other reason, for example the claims arising from some interpretations
of the Arrow theorem, a contention disposed of separately in upcoming
chapters.
Knowledge of other minds
Now for the challenge to (2a). If all we know are public votes over alter-
natives, without discussion, in a single, static instance, then what do we
knowabout the underlying preferences behind the actual choices? Strictly
speaking, we do not even know what kind of entities emit the vote; all
we know are some bare rankings, an aggregation rule, and an outcome.
With so little information we could not say that choices might strategically
misrepresent preferences. The best we could say is either that choices are
preferences, or that underlying preference is a meaningless concept.
We could not discover that choices may strategically misrepresent pref-
erences unless we have information from beyond the single instance. It is
obvious to us that choices may misrepresent preferences because we do
not live in the single instance. In the richer information environment, we
know that choices sometimes misrepresent preferences only because we
know that choices sometimes do represent preferences.
Much of ones knowledge, and almost all of ones discursive knowl-
edge, political or not, depends on the testimony of others (Shapin 1994;
Coady 1992). Could that testimony be generally wrong? The skeptic
denies the possibility of knowing anoutside world, or denies the possibility
of knowing other minds, on the argument that since each of our beliefs
(about an outside world or about other minds) taken alone may be false,
they might all be false. The philosopher Donald Davidson replies that it
does not follow from the fact that any one of the bills in my pocket may
have the highest serial number, that all the bills in my pocket have the
highest serial number; nor that since anyone may be elected President,
that everyone may be elected President. Nor could it happen that all our
beliefs might be false (Davidson 1991a, 193). [E]nough in the frame-
work and the fabric of our beliefs must be true to give content to the rest
(Davidson 1991b, 160).
There is no assigning beliefs to a person one by one on the basis of his verbal
behavior, his choices, or other local signs no matter how plain and evident, for
we make sense of particular beliefs only as they cohere with other beliefs, with
preferences, with intentions, hopes, fears, expectations, and the rest . . . Crediting
people with a large degree of consistency cannot be counted mere charity: it is
unavoidable if we are to be in a position to accuse them meaningfully of error and
some degree of irrationality. Global confusion, like universal mistake, is unthink-
able, not because imagination boggles, but because too much confusion leaves
nothing to be confused about and massive error erodes the background of true
belief against which alone failure can be construed. (Davidson 1980, 222)
In interpreting the beliefs of another as intelligible, I must assume that
the objects of his beliefs correspond well enough to the objects of my
own to permit contrast on points where we plainly disagree. Much the
same is true in interpreting anothers desires. In interpreting beliefs of
another, we must, to make sense of exceptions, assume a pressure in the
direction of logical consistency and a pressure in the direction of truth.
In interpreting desires, to make sense of exceptions, we must assume a
pressure in the direction of transitive consistency, and, although less so
than in the case of belief, some core of intelligible similarity in desires, a
pressure in the direction of agreement: with desire as with belief, there
is a presumption (often overridden by other considerations) that similar
causes beget similar evaluations ininterpreter andinterpreted (Davidson
1986, 208). These are known as Davidsons principles of charity in radical
interpretation. Radical interpretation is a thought experiment where we
ask how we would interpret the action of a radically unknown agent, and
charity means that we must initially attribute agent rationality in order to
attain an interpretation.
It may be objected that I misrepresent Riker. He did not say that it is
impossible to knowother minds, only that there are insufcient data from
voting choices to infer underlying preferences. He could say that we know
others beliefs and desires well enough in private life and on the market,
but not when we enter public life and the government. Incentives to mis-
represent are at least as ubiquitous in private life as in public life, but leave
that problem aside. Rikers claim is that if communication is limited to
voting choice, then it is impossible to know underlying preferences. This
may be so for each of our votes taken alone, but from that it does not fol-
low that it is so for all of our votes considered together. A series of votes
on similar issues would begin to generate enough data to allow inference
of underlying preferences, presuming at the individual level logically con-
sistent beliefs, their correspondence to objects, and transitively consistent
desires, that are sufciently similar.
For example, suppose that we observe that a legislator votes for alter-
native C over alternative B in the rst stage of a plurality runoff among
alternatives A, B, and C, and that alternative C wins the most votes in
the rst stage. Then in the runoff between C and A we observe that the
legislator votes for A, and that A is victorious in the runoff. So far, we
have observed that the legislator chooses A > C > B. The next day a
binary vote arises between alternatives B
and C
and the legislator votes

for B
over C
, contradicting yesterdays vote for C over B. At this point

an observer can generate a number of hypotheses. Perhaps the legisla-
tor blundered or is irrational. If we assume that the legislators actions
are rational, then one plausible hypothesis is that his true preference is
B over C and that yesterday he was engaged in strategic (also termed
sophisticated) voting: the legislator and his allies knew that the cham-
bers distribution of preference orders was such that if they voted sincerely
for B in the rst stage B would go into the runoff and defeat A their more
favored alternative, so they strategically voted for Cin the rst stage. Con-
trary to Riker, the possibility of manipulation has not obscured our infer-
ence of the legislators underlying preferences, rather, just the opposite,
our knowledge of the possibility has enabled the inference of preference
ranking. Now suppose that we are able to make a third observation and a
fourth observation: the legislator votes for B
over C
and B
over C
.
That strengthens the hypothesis that the legislators true preference is
B over C.
The foregoing is merely from assuming internal consistency of desires
and understanding the logical properties of voting rules. Considerations
of external consistency may improve the inference of an individuals pref-
erence rankings. Suppose that alternative A is for high social-security
payments, B is for medium social-security payments, and C is for low
social-security payments. Then on the rst days vote we would be able
to hypothesize that voting for C over B was strategic: although there may
actually be occasional contexts where it makes sense to prefer high pay-
ments to low payments to medium payments, generally someone will
prefer high over medium over low. Or suppose that alternative A is aid to
the third world, Bis labor organizing rights, and Cis costly imperialist war
memorials. Fromwhat we knowabout the coherence of ideologies we are
againable to hypothesize immediately that the legislators true preferences
are A > B > C and that his vote for C (imperialist war memorials) over
B (labor organizing rights) was strategic in nature.
The assumption of internal consistency alone permits some inference
of underlying preference. Considerations of external consistency provide
additional principles and data to ease the inference. Finally, public dis-
cussion surrounding the votes provides even more data with which to
triangulate on a reading of others underlying preferences. Suppose
that when the legislator voted for C over B he said out loud in a speech
that his true preference is for B over C and that his vote was strategic in
order to thwart the unfair attempt by the opposition to manipulate the
outcome via agenda control. His announcement itself could have been a
strategic lie, but it is a piece of evidence that gains strength from other
considerations, for example if he later votes for B
over C
, or if he is known
never to lie about such things. In discussion, individuals may sometimes
misrepresent their desires and beliefs, but again enough must be true
to give content to the rest (for a criticism of Rikerian skepticism about
democratic discussion, see Mackie 1998). Sustained public deliberation
over a series of contested issues involves a complex wealth of meanings
that feeds intuitions about the intentions of others. The sum of evidence
from all sources permits one to form judgments about what other people
want and know, judgments that are fallible but reliable enough for human
affairs.
It is the peculiar misfortune of the skeptic that he is always forced to
act as if his conclusions were false. The skeptical philosopher dresses
warmly for cold weather and worries about whats for dinner, even if all
his experiences are just delusions. The skeptical political theorist infers
preferences behind choices in every human situation, even, as it turns
out, in making his case against the possibility of doing so. Riker uses the
very methods of inference I have recited in his case studies that purport
to demonstrate pervasive political disequilibrium and obscurity of pref-
erences. I will argue that Rikers inference of preferences in each of these
cases was clearly mistaken, but that does not change my point against
him on the knowability of preferences. My point is the transcendental
one that he must assume that preferences are knowable in order to at-
tempt his demonstration that they are not. Further, the fact that I am
able to present reasoning and evidence that supports one hypothesized
set of underlying preference rankings and undermines another hypothe-
sized set of preference rankings shows again that inference of underlying
preference from surface manifestations is not only possible but also a
normal and uncontroversial everyday occurrence.
In his analysis of a 1956 Congressional vote over the Powell amend-
ment involving school constructionfunds andschool desegregation, Riker
states:
Since there were only two roll calls on the three alternatives, there is not enough
information to specify complete preference orderings. But there is enough data
from other roll calls in 1956 and 1957 and from the debate to show that some
people probably voted strategically and that there were enough of themto generate
a cycle. (Riker 1982, 153)
Rikers further analysis of the Powell amendment makes an implicit aux-
iliary assumption that elected representatives represent the interests of
their districts. He goes on to condently identify ve natural political
groups(!), almost the exact number of representatives in each group,
and the preferences of each group over three alternatives (Riker 1986,
118122). He seems to have forgotten, among other things, that on his
account there is no such thing as a district interest that could be discov-
ered by electing a representative. It is a delicious irony that his analysis is
forced to assume that Congressional districts have identiable interests. It
was a similar problem, whether countries could be said to have an identi-
able interest, that led Arrow down the path to his theorem which denied
such a possibility (Amadae and Bueno de Mesquita 1999, 274). Yet, in
his attempts to show the empirical relevance of Arrows logically possi-
ble result, Riker is forced to assume the empirical irrelevance of Arrows
result. In another analysis, of the Wilmot Proviso, Riker condently iden-
ties eight factions and proposes an inference of the preferences of each
over three alternatives: There were not enough votes to ascertain pref-
erence orders, but it is easy to guess what they were (Riker 1982, 227).
Finally, from minuscule data in an obscure letter written 1,900 years ago
by Pliny the Younger, Riker is able to identify and estimate the strength
of three factions in the Roman Senate on an issue, involved in a process
of voting that resulted in the socially better outcome, despite rampant
manipulation:
In general, parliamentary situations are like this. Leaders have the kind of
[agenda-setting] power that Pliny exercised, but back-benchers can counter with
strategic voting. So the fox can be outfoxed. And thus a balance can be main-
tained, often resulting, as here, in the selection of the . . . socially better outcome.
(Riker 1986, 85)
There is one unfortunate difference between the skeptical philosopher
and the skeptical political theorist. The philosopher would be ignored if
he recommendedthat humaninstitutions be designedas if his conclusions
were true, but the political theorist might wrongly be heeded.
3 Is democratic voting inaccurate?
Democratic voting as inaccurate
Simple majority voting on binary alternatives is not a problem for social
choice theory, and indeed enjoys several desirable qualities that account
for its paradigmatic appeal (May 1952; Rae 1969; Taylor 1969). Political
issues are not somehow naturally binary, however, and all voting meth-
ods of reducing multiple alternatives to two alternatives are subject to
manipulation. Thus, Riker (1982, 65113) continues, fairness requires
a decision rule that works with more than two alternatives. Any num-
ber of plausibly fair voting methods are available, but the problem is
that, given a xed set of voters preference rankings, it is possible that
the different methods would lead to different outcomes. The method of
counting affects the outcome of counting; thus, voting does not accurately
amalgamate voters values.
For example, if preferences are as in Table 3.1, and if voting is sincere,
then alternative A wins by plurality voting, B by plurality runoff, C by
the Condorcet criterion, D by approval voting, and E by Borda count
(Nurmi 1992, 465). The plurality rule is that the alternative with the
most votes wins. If there were an election among alternatives A through
E, those in the four-voter group would each cast his vote for A, those in
the three-voter group would each cast her vote for B, and those in the
two-voter group would each cast his vote for C: Agets the most votes (but
not necessarily the majority) and thus is the winner under the plurality
rule. The plurality runoff is a two-stage process. If one alternative wins a
majority in the rst stage, that alternative wins the election; otherwise the
top two vote-getters by plurality rule in the rst stage go on to face each
other in the second stage. Here, the two winners in the rst stage would be
A and B. For the second stage, four voters prefer A to B, but seven voters
prefer B to A; thus B is the plurality runoff winner. The Condorcet win-
ner is the alternative that beats all others in pairwise comparison. Refer to
the pairwise comparison matrix in Table 3.2, which shows the number of
votes a row alternative receives against each column alternative: B beats
44
Is democratic voting inaccurate? 45
Table 3.1. Five alternatives, ve
procedures, ve winners
4 voters 3 voters 2 voters
A B C
E C D
D E E
C D B
B A A
Source: Nurmi 1992, 465.
Table 3.2. Five winners: pairwise comparison
matrix and Borda count
A B C D E (BC)
A 4 4 4 4 (16)
B 5 3 3 3 (14)
C 5 6 5 5 (21)
D 5 6 4 2 (17)
E 5 6 4 7 (22)
Note: Votes Favoring Row Alternative over Column
Alternative. The rightmost column is the Borda count
(BC), which is the rowsumof the pairwise comparisons
and alternative faces.
A by ve votes, D beats B by six votes, E beats D by seven votes, and C
beats E by ve votes: C>E>D>B>A, and Cis the Condorcet winner.
In the prole of voter preferences in this example there is a Condorcet
winner and a transitive ranking of collective outcomes; this is not neces-
sarily so for all proles, as will be detailed below. Approval voting per-
mits voters to cast a vote for each alternative that wins their approval.
In this instance, assume that members of the four-voter group approve
of the three alternatives they favor most: A, E, and D, and disapprove
of the remainder; and that members of the other groups approve only of
the two alternatives each favors most: B and C for the three-voter group;
and C and D for the two-voter group. Now we just add up all approval
votes for the result: alternative D wins with six approval votes, more than
the votes received by any other alternative. Finally, if there are ve al-
ternatives as we have in this example, the Borda count assigns a score of
zero to the last-ranked alternative, one to the fourth-ranked alternative,
two to the third-ranked, three to the second-ranked, and four to the
rst-ranked. The scores are summed and the alternative with the highest
score is rst-place winner, and so on. The Borda count is also identical
to the row-sums of pairwise comparison votes, as shown in Table 3.2.
The Borda count assigns 22 votes to E the Borda winner, 21 votes to C,
17 votes to D, 16 votes to A, and 14 votes to B. Thus, from the same
prole of voter preferences, ve different voting procedures, each of them
reasonable and used in real-world contests, yield ve different outcomes.
The Borda count, which assigns one point (or any other equal inter-
val) of difference between each alternative, is the most straightforward
instance of a positional voting method. Plurality voting or negative plu-
rality voting can also be construed as positional methods: plurality voting
assigns one point to a voters top-ranked preference and zero points to
her other preferences, and negative plurality voting assigns a zero vote
to a voters bottom-ranked preference and one vote to each of her other
preferences. An innite variety of positional methods is available; for
example, suppose in our example above that instead of four points we
assign seven points to the top-ranked alternative, and then the rest as in
the Borda count three to the second-ranked, two to the third-ranked,
and so on. That method would yield A as a winner, in place of E the
winner by the Borda count. There exist proles of voter preferences that
permit any outcome by monkeying around with positional weights. An
example that would have delighted Riker: Saari (1995a, 112, 122) nds
a 10-candidate voter prole that permits 84,830,767 different election
outcomes to arise by varying the choice of the positional method!
Even if a method in wide use were justiably fair, say in producing a
Condorcet winner (the choice that beats or ties all other choices in pair-
wise comparison), Rikers (1982, 66114, 234236) favored criterion, we
would not be able to know that, because information on invisible pref-
erence orders is usually not sufciently available from the visible data of
voting choice, he argues. In other words, even if by luck we happened
to possess an accurate voting method, it would not be possible for us to
know that we did. Here Riker falls back, as usual, to his mistaken basic
argument pattern.
Possible but not probable
It is possible that different methods will lead to different outcomes, but is
it probable? Rikers argument depends on demonstration of possibilities
by example, but he concedes that:
The moral and prudential standoff among methods would not in itself occasion
difculty for democratic theory if most of the time most methods led to the
same social choice from a given prole of individual values . . . But it seems a safe
conjecture that, if such a comparison [of commonly discussed methods] were
made, the proportion of social proles from which all the compared methods
produced identical results would indeed be tiny. (Riker 1982, 235)
The evidence defeats Rikers conjecture. We shall review studies that ap-
ply different voting rules to the same prole of individual preferences,
under different assumptions about the distribution of individual pref-
erence orderings. The rst assumption is that individual orderings are
uniformly distributed, the second that such orderings resemble one an-
other, and the third is the use of real data on voter preferences. Many of
these same studies deal with the topic of the frequency of cycles, which
will be considered in the next chapter.
Nurmi (1992) tests the Condorcet efciency (probability of a method
selecting the Condorcet winner when one exists) of widely used voting
methods by computer simulation. The impartial-culture assumption is
widely used in simulations: every logically possible preference order is
considered equally likely; or, given a number of alternatives A, all voters
are assigned with probability
1
A!
to each of the A! possible strict preference
rankings over A(Nurmi 1992, 461). Under the impartial-culture assump-
tion, the Condorcet efciency of commonly used voting rules generally
decreases with an increase in the number of alternatives, and decreases
more slowly with an increase in the number of voters. For example, as-
suming an impartial culture, the plurality runoff with three alternatives
and 5 voters has a Condorcet efciency of 98 percent, with three alter-
natives and 999 voters 96 percent, with seven alternatives and 5 voters
77 percent, and with seven alternatives and 999 voters 67 percent. Dis-
crepancies in choice between any two voting rules also generally increase
with an increase in the number of alternatives, and increase more slowly
with an increase in the number of voters.
Impartial-culture assumes that each possible preference order
among alternatives is equally probable, and thus that preferences have
no relationship to any substantive considerations: one person prefers A,
personal prosperity to B, the torture of kittens, to C, suicidal nuclear
war, but each of the ve remaining permutations of strong preferences is
equally probable, for example, that another person prefers suicidal nu-
clear war to the torture of kittens to personal prosperity, and another
the torture to nuclear war to prosperity, and so on. Even under the
impartial-culture assumption, reasonable voting rules (Borda, plurality
runoff, Hare, Coombs, but not plurality) applied to the same prole of
voters preferences tend to identical outcomes for three, four, or ve alter-
natives and many voters. Plurality rule performs worst of all as numbers
of alternatives and voters increase, as one would expect; and unlike other
rules the Borda count retains high Condorcet efciency even as the
number of alternatives greatly increases.
The estimates of Condorcet efciency and of choice discrepancies
are similar under a bipolar-culture assumption (the rst third of the
electorate has one preference order, the second third is assigned the
impartial-culture assumption, the third has preferences in reverse order
of the rst). However, a slight perturbation of impartial-culture towards
unipolarity (assigning 5 percent of voters to identical preferences and
95 percent to impartiality under a rst condition, or 10 percent to identi-
cal preferences and 90 percent to impartiality under a second condition)
makes a large difference, generally increasing the Condorcet efciency
of voting rules, and reducing the discrepancies between rules as the
number of voters increase, both effects approaching perfection with the
increase from 5 to 10 percent unanimity. For example, with seven alter-
natives and 999 voters, the Condorcet efciency of the plurality runoff is
67 percent under impartial culture, is 96 percent under the 5 percent
unanimity condition, and is 100 percent under the 10 percent una-
nimity condition (Nurmi 1992, 478, interpreted from Figure 7). Under
the 10 percent unanimity condition, and assuming 999 or more voters
and even up to 15 alternatives, there are zero choice discrepancies be-
tween plurality runoff and Hares system, and between six of the pairwise
comparisons among Hares, Coombs, Nansons, and Copelands voting
systems.
Merrill (1988) conducted simulations using both impartial culture and,
in a spatial model, multivariate normal distributions across dimensions.
He tested both the Condorcet efciency and the social utility efciency
of various voting rules under different assumptions. When candidate dis-
persion is equal to voter dispersion, then Black (a Condorcet-consistent
rule which decides cycles by the Borda count thus almost identical to
Condorcet), Coombs, and Borda are of high Condorcet efciency; Hare,
plurality runoff, and approval are intermediate; and plurality is low. If
candidates are less dispersed, more central, than voters, then Condorcet
efciency drops drastically for Hare, plurality runoff, and plurality. The
simulated voters possess Von NeumannMorgenstern utilities, and social
utility of a candidate is the sum of all voter utilities for that candidate,
arguably a more appropriate baseline than Condorcet efciency. Social
utility efciency is highest (in the high 90s) for Borda, Black, approval,
and Coombs procedures, lower for Hare and plurality runoff, and low-
est for plurality (37). Such efciency does not decrease with an increase
in number of candidates for Borda, Black, approval, and Coombs; de-
creases with Hare and plurality runoff, and dramatically decreases with
plurality (36). Social utility, Condorcet (Black), Borda, approval, and
Coombs yield almost identical results. These ndings are strong under
impartial culture, and much stronger under more natural multivariate
normal distributions.
Gehrlein (1995; see also 1997) is able to estimate analytically the re-
lationship between Condorcet efciency and social homogeneity. The
more similar are voters preference orders, the more likely is it that a
Condorcet winner exists. Further, for plurality rule, Borda count, and
plurality runoff, Condorcet efciency increases as voter homogeneity in-
creases (negative plurality vote against your least-favored candidate
is less Condorcet-efcient as homogeneity increases). For lower levels of
homogeneity, Borda count is more Condorcet-efcient than plurality; at
modest levels of homogeneity, plurality, Borda, and plurality runoff rules
display high Condorcet efciency; at higher levels of homogeneity, plu-
rality is more Condorcet-efcient than Borda count; at all levels of homo-
geneity, plurality runoff is more Condorcet-efcient than either plurality
or Borda. The simulations of Merrill and of Nurmi and Gehrleins es-
timates are robust enough to defeat Rikers conjecture, and thus Rikers
dismissal of majoritarian democracy as necessarily inaccurate.
We seldom possess rm data on voters preferences over all alterna-
tives because the plurality methods we often use do not record them.
Chamberlin, Cohen, and Coombs (1984) were able to obtain data on 5
different presidential elections of the American Psychological Association
(APA), each involving 11,00015,000 voters rank-ordering 5 candidates.
The APA is an organization with cleavages: roughly half its members
are academic psychologists and half are nonacademic psychologists, and
further there are the theoretical, methodological, topical, and political di-
visions usual in a social science. We will also discuss the APA study in the
section on cyclical majorities, for now we will look at the results the study
obtains from hypothetical application of different voting rules to these
real-world preferences. The authors use the Condorcet pairwise order-
ing as a baseline (there were no cyclical majorities present), and compare
results that would obtain by applying plurality voting, Borda count, the
Hare method, the Coombs method, approval voting with two votes per
voter, and approval voting with three votes per voter. The various methods
failed to select the Condorcet winner about 20 percent of the time; when
they failed, however, they always picked the second-ranked candidate by
the Condorcet ordering. The various methods also deviated from the
complete Condorcet pairwise majority ordering about 20 percent of the
time. All methods ranked the same two candidates rst or second 85 per-
cent of the time. Some methods were better than others. Plurality rule
deviated the most from the Condorcet winner and from the Condorcet
ordering, while Borda count and approval voting with two votes per voter
were quite close to the Condorcet outcome.
Feld and Grofman (1992) also examine rare data on full-rank order-
ing of multiple candidates from 36 elections held by nongovernmental
associations in Great Britain, which also will be more fully presented
in the discussion of cyclical majorities below. Voters per election ranged
from 9 to 3,422 and candidates ranged from 3 to 29. Each election had
a Condorcet winner. The Borda winner was the Condorcet winner in 34
out of 36 elections. Borda rankings differed from Condorcet rankings by
about 6 percent. Levin and Nalebuff (1995) used some of the same data
to compare nine different voting rules: plurality, single transferable vote,
Borda, Copeland, minmax, KendallWei, power ranking, minimum vi-
olations, and YoungKemeny. All rules obtained similar results except
for plurality; when rankings differed from one another it was because
of a cycle somewhere in the ordering, but even then rules picked the
same winner. When voter preferences are sufciently similar, a variety
of voting systems lead to similar choices, and these choices have desirable
properties, they conclude (4).
Felsenthal, Maoz, and Rapoport (1993) examine a mostly overlapping
set of elections using a different approach. They argue that the Copeland
method should be the normative baseline. The Copeland method selects
the alternative which scores highest on the Copeland index, which for
alternative x is the number of times x beats other alternatives minus the
number of times x loses to other alternatives. The Copeland method gives
the same result as the Condorcet method when there are not cycles, but
provides an order when Condorcet reports cycles. They compared rank-
ings by the Copeland method to ve other voting rules: plurality with
one vote, plurality with several votes, approval voting, the Borda count,
and the repeated alternative vote procedure. Most of the elections un-
der consideration were to ll more than one post. Plurality with one vote
allocates each voter one vote even if there are several slots to ll. Alter-
natively, plurality with several votes allocates each voter one vote per slot
to ll. Approval and Borda we have already encountered. In the repeated
alternative-vote procedure, a variation on the Hare procedure, each voter
ranks at least one and as many of the competing candidates as she wishes.
The votes are counted and if there is a majority winner then she wins. If
not, then the candidate who is ranked rst in the fewest ballots is dropped
fromall the ballots, and the vote recounted. This process is reiterated un-
til there is a majority winner (or a tie, randomly broken). If there is more
than one slot to ll, then the procedure is repeated on remaining can-
didates to ll the second slot, then repeated on remaining candidates
to ll the third slot, and so on. The authors apply the six voting rules
to the proles in each election, and then (Spearman rank-order) corre-
late the rankings generated by the Copeland method against the rankings
generated by each of the ve remaining voting rules. The median corre-
lation between Copeland and plurality with one vote is 0.857, not bad
but nevertheless the poorest performance, and this is not surprising as
we know that plurality performs poorly with more than a few candidates
(there was a median of nine candidates in these elections). The median
correlation between Copeland on the one hand, and plurality with sev-
eral votes, approval voting, Borda count, and repeated alternative vote
on the other hand, was, respectively, 0.953, 0.976, 0.977, and 0.963.
They also identied the candidates who would have won available slots
under each of seven voting rules, and with Copeland as the baseline,
measured the percentage of identical winners selected: 76.0 percent for
plurality with one vote, and for plurality with several votes, approval vot-
ing, Borda count, repeated alternative vote, and single transferable vote,
respectively, 88.8 percent, 89.8 percent, 89.0 percent, 90.7 percent, and
85.6 percent. In other words, the different procedures yielded virtually
identical rankings and virtually identical winners (in elections with only
three candidates the six methods were wholly identical in rankings and
winners). They also measured the methods against other desiderata, such
as whether they pick the Condorcet winner, or avoid the Condorcet loser.
The methods do so from 35 out of 35 elections to 32 out of 35 elections,
and again plurality with one vote is the worst performer. The mildly worse
performance on all indicators by plurality with one vote for more than
three candidates is something of an artifact, I believe. The data are from
elections held under voting procedures which encourage large numbers
of candidates. An actual election held under plurality rule with one vote
would strategically induce a great reduction in the number of candidates,
and plurality is more accurate the lower the number of candidates.
Felsenthal and Machover (1995) explore an expanded list of 92 real
elections that contains the 37 studied in Felsenthal, Maoz, and Rapoport
(1993). They compare Condorcet winners and losers on the one hand to
plurality voting, single transferable vote, and Borda count on the other
hand. Plurality either fails to select Condorcet winners or does select
Condorcet losers in 4.82 percent of instances, single transferable vote in
11.19 percent, and Borda count in 3.51 percent. Some of the elections
studied select one winner, others from two to twelve winners. The Borda
counts fault rate was 9.52 percent for elections with one winner, but was
much more well-behaved for elections with more than one winner. They
also compare plurality, single transferable vote, and Borda by a measure
of Copeland efciency. The Borda count is, again, the best performing
method.
Regenwetter and Grofman (1998) deploy a probabilistic model to in-
fer underlying rankings from votes cast in elections decided by approval
voting. The model was compatible with data in 7 of the 10 elections stud-
ied. The Borda ranking was identical to the approval-vote ranking in all
7 elections. The Condorcet ranking tended to be the same as the Borda
and approval rankings. Regenwetter and Grofman conclude that what I
call the irrationalist interpretation of social choice theory is overstated
(530). They quote Tanguiane (1991) with approval:
Many voting and decision making procedures proved to be efcient for practical
needs although they were poorly justied. The gap between theory and practice
can be explained by the fact that every real situation deals with restrictions ignored
by theory.
They argue that empirically testable cognitive models of social choice
tested against real data should complement the purely theoretical models
of social choice tradition.
Kurrild-Klitgaard (2001b) directly examines Rikers inaccuracy hy-
pothesis with data from the Danish national election surveys. Randomly
sampled respondents are asked to provide thermometer rankings on
various Danish political topics. From the thermometer data, Kurrild-
Klitgaard constructs rankings that would result from the application of
up to seven voting procedures: Condorcet; plurality; Borda; the cumula-
tive method (each voter is assigned the same number of points, say 100,
to assign as she pleases across the alternatives; alternatives are ranked by
total sum of points across voters); Bentham method (straight off the ther-
mometer rankings); the Nash method (same as Bentham, but instead of
adding voters point assignments, multiply them); approval voting assum-
ing top two alternatives are approved; and approval voting assuming top
three alternatives are approved. Rankings by all methods across all topics
are virtually identical, and as usual plurality is the weakest performer.
I shall calculate Spearman rank-order correlation in order to compare
the results: 1.0 means perfect agreement in ranking; 0 means no re-
lation in ranking; and 1.0 means perfect disagreement. Respondents
ranked 11 parties in 1973, 9 in 1994, and 11 in 1998. I take the ap-
proximation of Benthamite utility as the normative baseline. For all three
years rankings by Bentham, Condorcet, and Cumulative were identical;
Borda was identical except for transposition of second and third places in
1994 (Spearman correlation 0.98). Spearman correlations for the Nash
method were 0.99 (1973), 0.98 (1994), and 1.0 (1998); for approval with
three votes from 0.95 to 1.0. Plurality was the worst performer: 0.68 in
1973, 0.78 in 1994, 0.85 in 1998.
Table 3.3. Convergence of voting rules,
Danish leaders
Correlation with Bentham 1973 1994 1998
Condorcet 0.98 1 1
Cumulative 0.99 1 1
Borda 1 1 1
Nash 0.99 0.88 0.96
Plurality 0.72 0.76 0.88
Approval (3) 0.95 0.95 0.96
In 1994 respondents ranked the policies of the 7 parties represented
in Parliament. Bentham, Condorcet, Cumulative, Borda, and approval
with three votes were identical (1.0), Nash was 0.96, and plurality was
0.86. Respondents ranked the leaders by name of 10 parties in 1973, 8 in
1994, and 10 in 1998. In Table 3.3, I present the Spearman correlations
of various voting rules with the Bentham baseline.
Respondents ranked the importance of 28 issues in Danish politics in
1987/1988. The correlation with Bentham was 0.98 for Condorcet, 0.95
for plurality, 0.98 for approval with three votes. Respondents ranked four
important goals in 1994. The ranking by Bentham, Condorcet, plurality,
and approval with two votes was identical. Voters ranked 12 public budget
alternatives in 1990 and 20 such alternatives in 1994. We can reconstruct
a ranking by the Condorcet method, by the plurality method (percent who
say that too little is spent), and a third measure (percent who say too little
is spent minus percent who say too much is spent). Rankings over the
12 alternatives in 1990 were identical across the three methods. Rankings
over the rst 14 of all 20 alternatives in 1994 were identical, but there
was a Condorcet cycle over alternatives 1518.
The Cambridge City Council is one of the few jurisdictions in the
United States to use Hare preference voting, also known as the alternative
vote or instant runoff when used to elect one candidate and as the single
transferable vote when used to elect several candidates from the eld.
Preference voters are required to rank-order candidates. Dave Robinson,
information vice president for Californians for Electoral Reform, a group
advocating proportional representation in American elections, hypo-
thetically applied alternative voting rules to the votes cast in the 1999
Cambridge election, and posted his results to the organizations web-
page (www.fairvoteca.org/learn/cambcomp). There were 18,777 votes,
which selected 9 councilors from a eld of 24 at-large candidates. The
results by preference voting (Droop method) were identical to those by
preference voting (Hare method). Other methods a simulation of con-
ventional single-member-district plurality, single-member district plural-
ity runoff, at-large plurality with one vote, and at-large plurality with nine
votes either were identical or differed by only one candidate from the
preference-voting outcome. One could still favor preference voting on
other grounds, for example, if it created a greater sense of legitimacy,
increased voter turnout, or elicited a more moderate set of candidates.
Finally, in the climax to his volume (1982, 227232), Riker claims to
show that the election of Abraham Lincoln in 1860 was the result of a cy-
cle among the electorate and that different reasonable voting rules would
have yielded different outcomes from the same prole of preferences that
he estimates. I shall show in detail below that the prole he estimates for
the Lincoln election is at its most important point erroneous, and thus
that his claim that the 1860 election would yield different outcomes by
different voting rules is mistaken.
We know that to some extent alternatives are endogenous to voting
rules, such that one voting rule might elicit different alternatives for con-
sideration than another voting rule. When we hypothetically apply dif-
ferent voting rules to a set of preferences actually elicited by a particular
voting rule, we are neglecting to consider the endogeneity of preferences.
A critic could press that it is possible that somehow taking into consider-
ation the strategic interdependence between rules and alternatives would
decrease the convergence of outcomes that we nd when applying voting
rules hypothetically to data on real preferences. We could respond that
it is possible that such consideration would rather increase the conver-
gence among voting rules, and indeed, that is my conjecture. If the critic
pressed harder, we could reply that if endogeneity is a major problem for
the comparisons that demonstrate convergence, then that would count
more heavily against Rikers claim of arbitrary divergence, because the
slim and speculative evidence he produces mostly contrived examples
that show the logical possibility of divergence ignore the endogene-
ity problem as well. If endogeneity bites, then no contrived example is
persuasive.
Even with the implausible assumption of impartial-culture, the reason-
able voting methods converge when there not too many alternatives or
too many voters. Convergence improves when resemblances among in-
dividuals preference orders are admitted, and improve further when real
data are used. Further, it turns out that the impartial-culture assumption
misleads us about what happens as the number of voters increases. Given
certain assumptions, the Bentham method of summed cardinal utilities,
Condorcet pairwise comparison, and the Borda count, almost certainly
produce the same results as the number of (independent) voters increases,
according to Tangian (2000). What are the assumptions? There is more
convergence as (1) the number of voters increases, (2) the distribution of
voters preferences departs from the impartial-culture assumption, and
(3) the less indifference there is among voters about the alternatives. Next,
List and Goodin (2001) extend the Condorcet jury theoremto more than
two alternatives. Suppose along with epistemic democrats that the aim
of democracy is to track the truth about the general will or the public
good. The jury theorem assumes that independent voters are better than
random in selecting the true choice; this amounts to a departure, even
if slight, from the impartial-culture assumption. If so, then commonly
discussed voting rules plurality, Borda, Condorcet, Hare, Coombs
are each good at tracking the truth, and further all are almost equally
good at doing so, again as the number of voters increases (but plural-
ity is not quite as good as the others). We can afford to be relatively
relaxed about [choice of voting rule] from an epistemic point of view
(294), they conclude. Finally, recall that Merrills (1988) simulations
showed nearly identical performance by social utility, Condorcet, and
Borda.
Rikers conjecture is that it is not the case that most of the time most rea-
sonable voting methods lead to mostly the same outcomes. The evidence
at hand is overwhelming that the conjecture fails.
What is arbitrary?
Are some voting rules better than others? Quite denitely so. The rule that
Gerry Mackie decides everything is popular in some quarters but does
not enjoy widespread acclaim. We want voting rules for free and equal
citizens, and we are seeking a rule that is, one way or another, accurate
and fair. Simple majority vote over two alternatives is not aficted with
the perversities identied by social choice theory, and possesses several
properties: it is decisive (an alternative wins, loses, or ties although, as
the BushGore race shows, a tie result may be frustrating); it is anonymous
in that if voters trade names the result is unchanged (that is, it treats all
voters alike and the Mackie rule is thus disqualied); it is neutral in that
if alternatives trade names the result is unchanged (it does not privilege
any alternatives even if some, such as rights, should be privileged); and
it is strongly monotonic (positive responsiveness), that is, if an alternative
X is among the winners and if one voter changes her vote to X then
X becomes the unique winner. Those four properties uniquely identify
majority rule over two alternatives, but only because of the strong mono-
tonicity requirement. The uniqueness claim for majority rule over two
alternatives is much exaggerated in the literature. Several other voting
rules that operate over two or more alternatives (for example, the Borda
count) are decisive, anonymous, neutral, and weakly monotonic (nonneg-
ative responsiveness): if an alternative X is among the winners and if one
voter changes her vote to X then X does not become a loser, it should
never hurt an alternative to get more support. There is certainly a distinc-
tion between positive responsiveness and nonnegative responsiveness, but
not much of one. Positive responsiveness says that if there are a million
voters for Bush and a million for Gore, and one voter switches from Gore
to Bush, then Bush wins but this seems to be a formal rather than a
substantive victory. Nonnegative responsiveness says that if there were a
voting rule such that an increase in votes for Bush would make Bush the
loser, then we would assess a defect in the voting rule intuitively more
compelling than positive responsiveness (see Nurmi 1987, 67). Further,
it is possible, by varying the selection of axioms, to select any one of a
number of voting rules as unique, for example, YoungKemeny is the only
rule that is anonymous, neutral, Pareto, and satises reinforcement and
local independence of irrelevant alternatives (Young 1988); Borda the
only rule to satisfy axioms labeled neutrality, cancellation, faithfulness,
and reinforcement (Young 1974); and so on.
Riker (1982, 99101) proposes some additional criteria for evaluating
voting rules: that the Condorcet criterion rank alternatives by pairwise
comparisons (if such a ranking exists) should be the normative baseline;
reinforcement, such that if two subgroups approve an alternative then so
should the full group voting together; and independence of irrelevant alter-
natives (the social choice between two alternatives would not change if
individual preferences over some third alternative were to vary, which I
shall discuss below with respect to the Arrow theorem). Further desider-
ata might include contraction consistency (if a is the choice from {a, b, c}
then a is the choice from {a, b}), and whether the rule picks the majority
(
1
2
+1) winner.
No rule I list satises Arrows strong independence of irrelevant
alternatives condition, except that majority rule over two alternatives
satises it trivially merely because it is limited to two alternatives. It
trivially satises many of the conditions. Most rules of interest for three
or more alternatives are Condorcet-consistent rules, positional (or scoring)
rules (see Moulin 1988; Young 1988; Saari 1995a), or elimination rules
such as plurality runoff and Hare. Positional rules satisfy decisiveness,
anonymity, neutrality, weak monotonicity, and reinforcement, but not
always the Condorcet criterion. Of the positional rules (including plural-
ity), the Borda count may be most favorably distinct: it is the positional
rule most likely to satisfy the Condorcet criterion, it is the one least
likely to be subject to manipulation by small numbers of voters, it is the
T
a
b
l
e
3
.
4
.
S
o
m
e
a
x
i
o
m
a
t
i
c
p
r
o
p
e
r
t
i
e
s
o
f
s
o
m
e
v
o
t
i
n
g
r
u
l
e
s
C
r
i
t
e
r
i
o
n
W
e
a
k
P
i
c
k
s
5
0
p
e
r
c
e
n
t
+
P
i
c
k
s
C
o
n
d
o
r
c
e
t
C
o
n
t
r
a
c
t
i
o
n
R
u
l
e
D
e
c
i
s
i
v
e
A
n
o
n
y
m
o
u
s
N
e
u
t
r
a
l
M
o
n
o
t
o
n
i
c
W
i
n
n
e
r
W
i
n
n
e
r
C
o
n
s
i
s
t
e
n
c
y
R
e
i
n
f
o
r
c
e
m
e
n
t
M
a
j
o
r
i
t
y
r
u
l
e
:
2
a
l
t
s
Y
Y
Y
Y
Y
Y
Y
Y
P
o
s
i
t
i
o
n
a
l
:
P
l
u
r
a
l
i
t
y
Y
Y
Y
Y
Y
N
N
Y
B
o
r
d
a
Y
Y
Y
Y
N
N
N
Y
A
p
p
r
o
v
a
l
Y
N
Y
Y
N
N
Y
Y
C
o
n
d
o
r
c
e
t
-
c
o
n
s
i
s
t
e
n
t
:
C
o
n
d
o
r
c
e
t
N
Y
Y
Y
Y
Y
N
N
Y
o
u
n
g
K
e
m
e
n
y
N
Y
Y
Y
Y
Y
N
Y
S
c
h
w
a
r
t
z
Y
Y
Y
Y
Y
Y
N
N
E
l
i
m
i
n
a
t
i
o
n
:
P
l
u
r
a
l
r
u
n
o
f
f
Y
Y
Y
N
Y
N
N
N
H
a
r
e
Y
Y
Y
N
Y
N
N
N
N
o
t
e
:
Y
m
e
a
n
s
t
h
e
v
o
t
i
n
g
r
u
l
e
i
n
t
h
e
r
o
w
s
a
t
i
s
e
s
t
h
e
a
x
i
o
m
i
n
t
h
e
c
o
l
u
m
n
,
N
m
e
a
n
s
i
t
d
o
e
s
n
o
t
.
N
m
e
a
n
s
t
h
a
t
f
a
i
l
u
r
e
i
s
l
o
g
i
c
a
l
l
y
p
o
s
s
i
b
l
e
,
b
u
t
n
o
t
n
e
c
e
s
s
a
r
i
l
y
e
m
p
i
r
i
c
a
l
l
y
p
r
o
b
a
b
l
e
.
=
o
n
l
y
i
f
v
o
t
e
r
p
r
e
f
e
r
e
n
c
e
s
d
i
c
h
o
t
o
m
o
u
s
.
S
o
u
r
c
e
s
:
M
o
u
l
i
n
1
9
8
8
,
N
u
r
m
i
1
9
8
7
,
N
u
r
m
i
1
9
9
8
,
R
i
k
e
r
1
9
8
2
,
Y
o
u
n
g
1
9
8
8
.
Table 3.5. Pairwise comparison matrix
to illustrate YoungKemeny rule
A B C
A 33 25
B 27 42
C 35 18
one least likely to select a different winner if an alternative is dropped
(Le Breton and Truchon 1997). The pure Condorcet method satises
anonymity, neutrality, and monotonicity, but does not satisfy decisiveness
or reinforcement if there is a cycle present. Condorcet-consistent rules
generate the same ranking as pairwise comparison where there are no
cycles, and each differs in the way it decides cycles. The Schwartz (1982)
rule counts cycles as ties. Of the Condorcet-consistent rules, the Young
Kemeny rule may be most favorably distinct (Young 1997 dubs it the
maximum likelihood rule; others have named it the YoungKemeny rule;
see Risse 2001 for arguments for its distinctiveness and Saari 2003 for
a reply). The YoungKemeny rule selects the ranking supported by the
most number of votes in all pairwise comparisons within the ranking.
YoungKemeny agrees with Condorcet order when that exists, and, as
does the Borda count, decides a cycle (it may report a cycle as a tie, which
we count as decisive). For example, suppose a prole of voters that gives
rise to the pairwise-comparison matrix in Table 3.5 (Young 1997).
There is a cycle: A > B > C > A. For any election with three alterna-
tives there are six possible preference rankings: A > B > C; C > A > B;
B>C>A; C>B>A; A>C>B; B>A>C. To nd the YoungKemeny
score for one of the six possible rankings, say A > B > C, we sum the
number of pairwise votes for each of A > B, B > C, and A > C (33 +
42 +25 =100). We do this for all six rankings and discover that the rank-
ing B>C>Ahas the highest sum(42 +27 +35 =104) and this ranking
is thus the YoungKemeny ranking. The YoungKemeny result can also
be found by breaking the cycle at its weakest link: in the cycle here
A > B by 33 votes, B > C by 42 votes, and C > A by 35 votes A > B is
the weakest majority in the cycle A > B > C > A, hence we reverse
the weakest majority from A > B to B > A and the noncyclical ranking
B > C > A is our result. The YoungKemeny rule satises anonymity,
neutrality, and reinforcement. It also satises a condition labeled local
independence of irrelevant alternatives, that is, it is not susceptible to ma-
nipulation by the addition or deletion of alternatives either above or below
the given set of alternatives (but is still susceptible to such manipulation
in between, as it were).
Rules that are used in real-world settings include plurality, plurality
runoff, variations on Hare, and, more rarely, approval and Borda count.
The plurality rule was the weakest performer in the various simulations
as the number of alternatives increased, and for good reason. Plurality
rule has the virtue of being extremely simple for the voter and for the or-
ganizer of elections, works adequately if there are fewer alternatives and
larger agreement among voters across the alternatives, but may err when
there are more alternatives and larger disagreement among voters. Such
errors appear now and then in American candidate elections (but such
errors do not appear often, because plurality rule in single-member dis-
tricts tends to strategically elicit no more than two candidates, an effect
not captured by the simulations). In the American context, plurality-rule
situations will be found when there is not a majority winner in a primary
election (the rst stage of a plurality runoff when there are two parties
in the general election) or when there are third or nth parties running
in a general election and no majority winner amongst the candidates. I
managed a campaign in a local nonpartisan primary election among four-
teen candidates that resulted in the forwarding of two candidates to the
general election whose selection made no sense to anybody on the scene
save perhaps to the spouses of the two victors, and the spouses admi-
ration was probably motivated by loyalty alone. In the 1972 Democratic
primary elections McGovern beat Muskie by plurality rule, went on to
win the Democratic nomination, and then was steamrollered in the gen-
eral election against Republican Nixon. Because McGovern was the top
choice for some Democratic primary voters but the bottom choice for
many, Muskie would have beaten McGovern in the primary election
under plurality runoff, Borda count, or the Condorcet criterion (Joslyn
1976, as reported in Mueller 1989, 122), and as more of a centrist prob-
ably would have done better against Nixon. In the 1970 presidential elec-
tion in Chile, leftist Allende won with 36 percent of the vote, the centrist
with 28 percent, and the rightist with 35 percent (Merrill 1988, 4), and
Allendes government and Chilean democracy shortly died at the hands
of a rightist military coup. Colman and Pountney (1978) look at the
266 British members of Parliament in the 1966 general election who
won by a plurality but not a majority of votes cast. They estimate that 15
of the 266 were Condorcet losers, that is, they would have ranked last if
measured by the Condorcet method.
Why does plurality rule tend to be unrepresentative as the number of al-
ternatives increases above two? Because it throws away a lot of information
about voters preferences. Suppose an election by plurality rule among
three candidates: Rapscallion (A), Tweedledum (B), and Tweedledee
(C). Six voters prefer A > C > B, ve voters prefer B > C > A, and four
voters prefer C > B > A. Tweedledum and Tweedledee espouse simi-
lar policies preferred by a strong majority of the population; Rapscallion
espouses policies that motivate a minority at the expense of the major-
ity. Plurality rule selects A, Rapscallion, and ranks the candidates A >
B > C. Now lets see what happens if the candidates face each other in
pairwise contests. Nine out of the fteen voters prefer Tweedledum, B,
to Rapscallion A. Ten of the fteen voters prefer Tweedledee, C, to
Tweedledum, B. The results by Condorcet pairwise comparison (and by
Borda count) are C > B > A, just the opposite of the plurality ordering.
To make things worse, suppose that Rapscallion and his cronies entered
the race only because they thought they would win under plurality, and
would have stayed out otherwise the majority has been thwarted.
A different rule, plurality runoff, forwarding the two top plurality win-
ners to a runoff election, will usually produce the Condorcet winner, and
so is less vulnerable to minority outcomes than pure plurality. Unfor-
tunately, in the Rapscallion example, plurality runoff happens to yield
a ranking of B > A > C. The prospect of that outcome should deter
Rapscallion A from entry, however; if so, then Tweedledee (C) would
beat Tweedledum (B). Plurality runoff, an elimination rule, is a familiar
method, so I will say no more about it.
Imagine that we want to rank the quality of students in a school (exam-
ple adapted from Saari 1995a). If the grading system is binary, such that
for each course a student gains a pass or a fail, the obvious way to rank
students is by the number of pass grades received (resembling majority
rule over two alternatives). Suppose, however, that the grading system
is rather the American one of A, B, C, D, and F, and that each student
accumulates 30 courses over four years. Now, if we rank students only by
the number of As they receive, the outcome of such plurality grading does
not have to be fair, its fairness depends on how grades are distributed. If
there are obvious top choices, one or two students who have attained 30
out of 30 As, then plurality grading picks the winners, but the remainder
of the rankings might not be that fair: someone with one A and 29 Fs
would outrank someone with 30 Bs. If there are not obvious top choices,
then the potential for unfairness extends to the winners: someone with
15 As and 15 Fs beats someone with 14 As and 16 Bs. Using the Borda
count to rank alternatives resembles using the grade point average to rank
students: zero points are assigned for each F, one point for each D, two
points for each C, three points for each B, and four points for each A
(and so on, for more than ve alternatives). If we only want to single out
the top student, then plurality grading and Borda grading (grade point
average) will yield similar results some of the time, but not all of the time.
Notice also that different grading systems have different incentives, just
as different voting systems can elicit different behaviors from candidates;
for example, under plurality grading students would be more likely to
seek easy courses and avoid hard courses (because a B under GPA would
be the same as an F under plurality, it is foolish to risk a B). The plurality
rule discards considerable information; but the Borda count utilizes all
ranking information.
Riker, Arrow, and many others, believe that the Condorcet criterion
(the ranking, if it exists, that results from pairwise comparison of the
alternatives) should be the normative baseline. However, there are strong
arguments in favor of the Borda count as a normative baseline (Le Breton
and Truchon 1997 calculate the Borda-efciency of various voting rules).
I agree with Dummett (1998) that:
with reservations, the Borda count is in principle at once the best tool for reaching
the decision most likely to be correct when the object is to reconcile different judg-
ments about effective means to a common aim, and the most equitable method
of determining a resultant of divergent desires.
Saari (1995a) uses accessible geometric methods that unify and simplify
much of social choice theory; and his work argues, given ordinal data, that
the Borda count is ideally the unique procedure which always respects
the will of the people. His formal work is extended in Saari (2000a;
2000b, among others), and popular expositions, well worth a look, may
be found in Saari (2001a; 2001b). One problem with the Condorcet
criterion is that it is indecisive in the face of cycles (the possibility of a
cycle is a consequence of insisting on the Condorcet criterion). Another
problem with the Condorcet criterion is that it throws away information
about preference rankings, sometimes with consequences that are intu-
itively undesirable. Here is an example. Suppose that 1,001 voters favor
A > B > C and 1,000 voters favor B > C > A. Alternative C is no better
than middle for all voters: it is the last choice of 1,001 voters and the
middle choice of 1,000 voters, so is clearly not a contender. Alternative
A is the rst choice of 1,001 voters but the last choice of 1,000 voters.
Alternative B is middle or better for all voters: the middle choice of
1,001 voters and the rst choice of 1,000 voters. The Condorcet order
awards the contest to A, the candidate ranked worst by almost half the
population, rather than B the obvious Borda winner (adapted from Saari
1995a, 79). Rikers argument for the Condorcet criterion is that when
an alternative opposed by a majority wins, quite clearly the votes of some
people are not being counted the same as other peoples votes (1982,
100). The same complaint can be lodged against the Condorcet criterion
Table 3.6. Borda reversal
3 2 2
Y A B
C Y A
B C Y
A B C
A B C Y (BC) ABCY (BC) ABC
A 2 4 4 (10) (6)
B 5 2 2 (9) (7)
C 3 5 0 (8) (8)
Y 3 5 7 (15)
in this example, however. The fact that almost half the voters rank the
Condorcet winner last is not being counted by the voting rule: B-voters
bottom ranking is not treated equally with A-voters top ranking. Since
Rikers principle that some peoples votes be counted the same as other
peoples votes can be construed to support both the Condorcet criterion
and the Borda count, if not other voting rules, the principle fails to iden-
tify the Condorcet criterion as a uniquely justied and therefore superior
voting rule.
One major practical allegation against the Borda count is that it is
sensitive to manipulation by the adding and dropping of alternatives
(although it is least sensitive of all positional methods, including plurality,
to such manipulation). Here is an example (adapted from Riker 1982,
92). Suppose the voter prole and the resultant pairwise comparison
matrix are displayed in Table 3.6. The Borda ranking for the four alter-
natives is Y >A>B >C. Suppose that alternativeY is removed. Remov-
ing an alternative shouldnt make a difference (contraction consistency).
With Y removed, the Borda ranking for the remaining alternatives is now
exactly reversed: C > B > A. (Condorcet ordering stumbles on this
prole as well: it reports a cycle among either A > Y > C > B > A or
A > C > B > A).
The Condorcet order, however, violates reinforcement. Suppose there
is a club with two divisions, ten members in one and nine members in
another, and there is a choice among three alternatives. In the Western
division, six voters prefer B > A > C, and four voters prefer C > A > B.
Reasonably democratic voting rules, including Condorcet pairwise
comparison, select B as the winner in the Western division and rank
the alternatives B > A > C. In the Eastern division, nine voters fall into
three symmetric groups: three prefer A >B >C; three prefer C >A >B;
and three prefer B > C > A. This is a cyclical prole with an equal num-
ber of voters for each ordering, so a pairwise comparison would report
that A > B > C > A, Borda count a tie, and YoungKemeny a tie. Per-
haps because of the confusion in the Eastern Division, the whole club
of 19 members meets and votes together. The three blocs in the Eastern
Division neatly cancel each other out, so why should they change the
unequivocal result from the Western Division when the club votes as a
group together? Under the Borda count, the outcome for the whole club
is the same as that for the Western Division, B > A > C; adding the tied
cyclical preferences from the Eastern Division changes nothing. By pair-
wise comparison, however, the outcome for the whole club is A > B >
C > A, a cycle (the YoungKemeny rule is tied between A > B > C and
B > C > A). Under Condorcet ordering, what happened to the unequiv-
ocal ranking from the Western division? Its as if by some magical stunt
it has vanished. (Saari 1995a, 5154 shows an easy way to construct
examples like this).
The Borda count and perhaps YoungKemeny may be the most favor-
ably distinct voting rules. How do they fare with respect to truth and fair-
ness of amalgamation? First, truth: suppose that voting is an epistemic ex-
ercise, that there exists a best or correct decision for the decision-making
group that the group seeks to identify by some method of voting. Say
that each voter has the same probability p (
1
2
< p < 1) of being correct.
For two alternatives, simple majority rule is most likely to identify the
correct outcome, according to Condorcets jury theorem. If the voters
are on average better than chance in identifying the most correct of the
two alternatives, and if voters judgments are independent, then the more
voters there are the more likely correct is the aggregate judgment.
1
For three or more alternatives, Young (1988, 1997) has demonstrated
that the YoungKemeny rule is the method with the maximumlikelihood
of identifying the correct ranking of alternatives. The story is a bit differ-
ent if the task is to select the probably best choice rather than to choose the
probably best ranking. Choosing the candidate with the highest probabil-
ity of being best is not necessarily the same as selecting the most highly
ranked alternative from the most probable ranking of alternatives. If p,
the voters probability of being correct, is closely above one-half, then the
Borda count is the method most likely to identify the best choice. If p is
near 1, then the YoungKemeny method is most likely to identify the best
choice; although, as Young points out, if p is near 1, it is still very likely
that the Borda winner is the best candidate even though strictly speaking
not the optimum estimate of the best alternative that YoungKemeny
would provide. Second, fairness: if we have exhausted our epistemic de-
liberations and want to establish a mere compromise among voters then
YoungKemeny (the median of all rankings: Young 1997) and Borda
count (the ranking with the minimum number of disagreements with al-
ternative rankings: Dummett 1998; Risse 2001) remain highly attractive
principles. Therefore, if accuracy of representation or exactitude of fair-
ness are the only considerations, then the Borda count and perhaps the
YoungKemeny rule appear to be the most justied of democratic voting
rules (compare Risse 2001 and Saari 2003).
Riker (1982, 81) dismisses YoungKemeny because, although it is
based on clever and defensible ideas, so are other Condorcet-consistent
rules such as Copelands and Schwartzs. As for Borda count, it seems
especially vulnerable to manipulation by the addition and subtraction of
alternatives, different positional methods lead to different outcomes from
the same prole, and there are few arguments for selecting Borda count
from among the innite number of possible positional rules (since Riker
wrote, the uniqueness properties of especially the Borda count and per-
haps YoungKemeny have become more obvious). Riker (1982, 91) also
mentions a practical difculty for YoungKemeny and Borda count: for
much more than three alternatives voters might nd it difcult to rank all
candidates (I shall not cavil that rational choice theory requires voters to
have complete preference orderings). Where Hare voting has been used,
however, voters have been required to rank-order all candidates, and it
has not been a major practical problem.
Accuracy is one desideratum for a voting rule, and in some circum-
stances simplicity may be another desideratum. Moreover, the Borda
count, perhaps the most accurate rule for aggregation of ordinal data, is
not the simplest rule. Plurality rule is very simple for voters and election
administrators but possibly vulnerable to rapscallions as in the example
above, or even verges on randomness with a half-dozen or more candi-
dates who are more or less evenly matched. Approval voting permits the
voter to approve or disapprove each candidate. Say that the alternatives
are A, B, C, and D a voter can approve only A, or both A and B, or
A, B, and C (the voter can also approve or disapprove of all four, but
then she may not bother with voting). Approval voting cant be manipu-
lated in a three-candidate situation like that of Rapscallion above (Weber
1995), Rapscallion would be deterred. The simulations show generally
that approval voting is more Condorcet-efcient than plurality. Further,
if voter preferences are dichotomous in the sense that voters divide the
candidates into two sets, are indifferent among the candidates in each set,
but rank one set above the other, then voters have no incentive to vote
strategically; similarly, if voter preferences are trichotomous then voters
will also be sincere (Weber 1995). If manipulation by agenda control is
a major concern, then it appears that approval voting is the method that
best resists such manipulation (Nurmi 1987, 192).
In addition to accuracy and simplicity we must also consider howvoters
and alternatives strategically interact with a voting rule.
2
Voting rules are
such that a voter may nd it to her advantage to vote other than her
true ranking so-called strategic voting. An American leftist in 1992
might have preferred as President left-wing Democrat Jesse Jackson to
centrist Democrat Bill Clinton to Republican George Bush. Assume for
the argument that Jackson and Clinton were about even, and that Jackson
would probably lose, and Clinton win, in an election against Bush. If our
leftist and people like her vote for their most favored candidate Jackson
then their least favorite candidate Bush wins, so they end up voting for
Clinton this is strategic voting. It arises under Borda count as well:
calculations are a bit more complicated, but a voter might believe that
the overall distribution of preferences is such that if everyone like her truly
reports her second-ranked candidate then that second-ranked candidate
would come in rst, so voters like her mark their second-ranked candidate
as last in order to assure their rst choice a victory. Strategic voting may
matter because of the possibility that strategic voters might outfox one
another and unintentionally end up selecting an outcome that almost
no one wants. Generally, but not always, the more simple a rule is, the
more vulnerable it is to manipulation. Strategic voting will be discussed
in greater detail in a chapter below.
Elimination rules (Merrill 1988, 13) include plurality runoff, Hare and
Coombs. The Hare procedure is called the alternative vote when used to
select one winner, as in elections to the Australian House of Representa-
tives, and single transferable vote when used to select several winners, as
in elections to the Australian Senate. Voters rank all candidates once (in
Australia, they might optionally select the ranking recommended by one
of the parties). If no candidate receives a majority of rst-place votes, then
the candidate with the lowest number of rst-place votes is eliminated from
all ballots, and the tally is recalculated. The procedure is repeated until
a majority winner is reached. The single transferable vote is more com-
plex, and is implemented in several variations; it is used in multimember
districts in Ireland. The US House of Representatives elects its speaker
by successive elimination, which is structurally identical to Hare if voters
do not change their votes from ballot to ballot: voters cast only their rst-
place votes, and if no majority then the candidate with the lowest number
of votes is eliminated, and then a newballot is taken, until a majority win-
ner is reached. Plurality runoff is equivalent to successive elimination if
there are three candidates, and advocates for the alternative vote in the
US call it the instant runoff. The Coombs method is like Hare, but in-
stead it proceeds by successively eliminating the candidate with the most
last-place votes. Elimination methods especially Hare are the least sus-
ceptible to manipulation by strategic voting, if that is a concern (Nurmi
1987, 192). Bartholdi and Orlin (1991) showformally that for more than
a few voters and candidates it is practically impossible to manipulate the
Hare vote by strategic misrepresentation of preferences. Nor is strategic
misrepresentation observed where Hare is used.
So far we have pretended that alternatives are naturally given, that they
are some exogenous fact that would be the same from voting rule to vot-
ing rule. The alternatives offered, however, can depend on the voting
rule, and here we are in for some surprises. As summarized by Myerson
(1995), Cox (1987; 1990; also see 1997) makes the case that plurality
voting with more than two parties (alternatives) makes for parties widely
scattered across the spectrum, that negative plurality has parties clus-
tering at some point within a wide portion of the spectrum, and that
approval voting, Borda count, and single transferable vote compel par-
ties to cluster in the center at the policies favored by the median voter.
When parties are endogenized, according to Myerson, we nd that plural-
ity rule creates a small number of parties, that Borda count and negative
plurality encourage a large number of parties, and that approval voting,
proportional representation, and single transferable vote can accommo-
date small or large numbers of parties. Plurality voting in single-member
districts tends to encourage two parties, for example, because voters (or
elites persuading them) tend to vote strategically. Why? If a voter be-
lieves or is led to believe that her rst-ranked choice has little chance
of winning and that the race is probably between her second-ranked and
third-ranked choice then shell likely vote for her second-ranked choice so
as to deny victory to her third-ranked choice: this can lead to a so-called
Duvergerian equilibrium with most support going to two candidates and
the tendency to the two-party system. The voter wants to coordinate with
other voters and a focal point in this coordination game is indicated by the
expectation that more people will vote for her second-ranked choice than
for her rst choice (some other kind of focal point would work as well
so long as beliefs were reciprocal among coordinators). In the example
with Rapscallion, Tweedledum and Tweedledee above, strategic voters
could defeat Rapscallions manipulation: ten people prefer Tweedledee to
Tweedledum, but only ve people prefer Tweedledum to Tweedledee;
thus, the ve Tweedledum voters would vote strategically for Tweedledee
in order to thwart Rapscallion, who would be deterred from entering the
race by this expectation. Anon-Duvergerian equilibriumcan emerge only
when the Tweedledee and Tweedledum voters are about equally divided
and there are no other focal points that permit these voters to coordinate
on one candidate against their foe, which is perhaps why the Rapscallion
maneuver is not frequently attempted or achieved.
To speculate, one could praise the Borda count for forcing parties to
the median, yet dread a consequent confusing Trotskyite proliferation of
one, two, three, many centrist parties; one could praise plurality for en-
couraging a clear binary choice between two distinct alternatives (Riker
1982, 88), but deplore the indeterminacy that permits those in society
with the most resources to single out the focal points for each of the
two parties (Myerson and Weber 1993). Further considerations include
variations in voting and party systems, sometimes ne ones, and also
their interaction with other institutional features such as the executive
and the constitutional court and their variations (Myerson 1995). Game-
theoretic models, however, are notorious for delivering large changes in
conclusions from small changes in assumptions (or even generating the
same conclusion from a second model after the rst model is shown to be
in logical error). Theoretical and especially empirical examination of the
ancillary consequences of electoral systems cabinet durability, number
of parties, proportionality of representation, voter turnout incentives, de-
scriptive representation of women and ethnic minorities, incentives for
localismor generalism, among others must also enter into the evaluation
of voting rules.
3
Culture, notably the particular distribution of voter preferences, should
also be considered. Reilly (2001) compares in Papua NewGuinea concil-
iatory politics under Hare-style preferential voting to conictual politics
there after adoption of plurality rule. Papua New Guinea is the most eth-
nically fragmented country in the world, and primary loyalties for many
are to ones own group. Thus, unlike in the US and UK, many candidates
enter the plurality race an average of 21.7 per seat in 1997 and the
winner is typically an arbitrary choice who obtains between 10 percent
and 19 percent of the vote, as low as 6.3 percent. This worsens ethnic
conict, and Reilly recommends preferential voting for many divided so-
cieties, as it requires the successful candidate to campaign across ethnic
groups for second and third preferences. Thus, deep historical under-
standing of single cases as well as comparative empirical analyses must
weigh more strongly in our judgments than isolated models, and wisdom
would help too.
Given that there are many complex interactions and tradeoffs among
many imperfectly measured and differently valued desiderata, it would
not do to be overly dogmatic about choice of voting rule. For cardinal pref-
erences ignoring conceptual, empirical, and strategic misrepresentation
problems I believe that a sumof utilities voting rule which counted each
citizen as one would most accurately and fairly amalgamate preferences
(such summations are not necessarily true or right, however). For ordinal
preferences, an ideal Borda count probably better satises accuracy and
fairness than other rules, although most of the literature endorses the
Condorcet criterion. Moving from the ideal to the practical, the Borda
count may not be so accurate and fair if in reality it is much affected
by strategic manipulation, an open question. Values other than accuracy
and fairness for example, proportionality, moderation of conict, ease
of use, among others might also be thrown into the balance. The voting
rules do not differ that much when aggregating from the same set of real
individual preferences over alternatives; but the different rules may elicit
different alternatives as in Papua New Guinea plurality might elicit di-
visive candidates and preferential vote conciliating candidates. Generally
speaking, pure plurality, widely used in the US and the UK, seems to
be the worst of the commonly used rules unless one puts a high value
on seesawing extremes in government so as to promote innovation in
public policy. Good arguments are made for the variants of Hare, and
for approval voting, as practical voting rules. There are good arguments
for Condorcet as an ideal rule, as a yardstick by which to measure the
accuracy of other rules, and there are better arguments, I believe, for the
Borda count. But as a practical rule the Borda count probably needs more
testing by experience. Good electoral engineers, like their colleagues in
physical engineering, want to test their models in practice and evaluate
them ultimately by real-world consequences.
4
The axiomatic criteria are supercially appealing, but then Riker treats
us to counterexamples (I have not repeated them here) illustrating that
each voting rule violates one or several of the apparently innocuous
criteria and to this we can add the considerations arising fromsimplicity,
fromstrategic responses fromvoters and fromcandidates, and frominsti-
tutions and culture. In a way, the axiom game is rigged: we could happen
on a voting rule that satised a maximum number of important criteria,
but then anyone is free to devise a supercially appealing criterion that
the rule does not satisfy (e.g., does the rule exclude outcomes based on
false beliefs? etc.), and then we are back to Rikers alleged arbitrariness.
Is the messiness indicated by these counterexamples and considerations
sufcient to justify Rikers conclusion that the use of one voting rule over
another is arbitrary? First, Riker himself does not mean that the choice
is arbitrary among all possible voting rules, only among those that are
arguably accurate and fair, those we have discussed such as plurality,
plurality runoff, Hare, Condorcet order, Borda count, approval voting,
and the like, and not among the plainly inaccurate or unfair such as the
Gerry Mackie rule or a rule which said nd out what the majority wants
and then do the opposite. If a social choice rule operates on all logically
possible proles to provide all logically possible nonempty outcomes,
then how many different social choice rules might there be? With ve
voters and four alternatives, there would be 10
235,000,000,000
such social
choice rules, as compared to 10
19
seconds in the universe since the big
bang, according to Kelly (1988, 62). Someone who has encountered one
or two voting rules in his lifetime might be bewildered to discover that
variations on a dozen or so voting rules are used by humans across the
world. Someone who knows that the number of possible social choice
rules is beyond imagination might be impressed to discover that we use
only a dozen or so.
Second, among those rules that are arguably accurate and fair, there
are many logical and empirical considerations that do permit us to say
that one rule would be better than another under specied conditions.
This is the beauty and usefulness of the work in social choice, that its
many ndings permit the cumulative renement of judgment on such
matters. Nor are we at the end of history with respect to designing polit-
ical institutions; conceptual and empirical advances, and yet uninvented
technologies, will cause our evaluations to shift over time. Nurmi (1987)
is the denitive work on comparing voting rules by how they satisfy vari-
ous axiomatic criteria. After mustering far more complex considerations
than I have here, and not mentioning Riker by name, he suggests two
possible responses:
(1) to conclude that no procedure is good enough for all purposes and, hence,
we should revise our ideas of popular choices so that their results are viewed as
nearly random and certainly more or less accidental, or,
(2) to conclude that the differences in performances should be taken into account
in choosing procedures for use in various settings. (191)
Nurmi drolly remarks that he nds the second conclusion more plausi-
ble. After a full survey of the logical possibilities associated with different
voting rules, Nurmi concludes that the more important question is the
comparative frequencies of undesirable faults in practical application.
An instance of the confusion of logical possibility with empirical prob-
ability arose in the early 1990s in a British Labour Party working party
on electoral reform.
5
The group rejected single transferable vote (STV)
ostensibly because of its lack of monotonicity the logical possibility that
increasing the vote for a candidate might cause her to lose. Widely used
plurality runoff is also nonmonotonic, as a logical possibility. The chief
electoral ofcer of Northern Ireland responded that the experience of
the use of STV in Northern Ireland over the past 22 years, involving a
range of election types and sizes, reveals no evidence to support in prac-
tice the lack of monotonicity.
6
Allard (1995) calculated that if STV were
used in the UK, there would be less than one incidence every century of
monotonicity failure.
7
To go fromplurality to STV would mean the cost
of a speculative to slight chance of a nonmonotonic result in exchange
for the benets of a voting rule extremely difcult to manipulate and of
a much more proportionally representative Parliament. Of course, the
Labour Party presently benets greatly, and unfairly, from plurality rule
in the UK, and would lose seats to other parties under STV.
We have seen from simulations, even with uniformly distributed pref-
erence orders, and more so from mildly unipolar distributions, and deci-
sively from empirical data, that the reasonable voting rules deliver similar
results; if there is mild homogeneity among the population, choice of vot-
ing rule need not evoke exquisite anguish. There is no rule that is uniquely
best for all purposes under all constraints; numerous considerations count
for or against any rule; but this does not make choice of a rule arbitrary.
For an analogy, consider the question of whether there is a single method
of human transportation that is uniquely best across all conditions of vari-
ation. Obviously, there is not. We do know that circumstances favoring
crawling or rolling are quite rare, and that teleportation always fails. As to
whether one wants to walk, run, bike, drive one make or model of car or
another, train, bus, plane, and so on, it always depends on purposes and
constraints. Does the mere multiplicity of methods mean that when an
individual chooses one method over another that her choice is arbitrary?
No. Often there are reasons for one choice over another; and often we
might be indifferent among some subset of choices, but theres nothing
at all wrong with considered indifference. And practically, sometimes a
bicycle will have to do when a car would be much better.
Riker himself recommends different methods for different circum-
stances, and gives reasons for his choices. Briey, he recommends for
legislatures the amendment procedure for three alternatives (which yields
the Condorcet winner if one exists, and otherwise the status quo), and
Borda or YoungKemeny for more than three alternatives. For elections
of executives and of legislators presumably in the American case of single-
member legislative districts, he recommends plurality vote so as to en-
courage a two-party system, and approval voting for primary but not for
general elections. The advice is not unreasonable. He goes on, however,
to say that his recommendations are merely a matter of my own taste,
rather than a matter of judgment. He must state that his recommenda-
tions are an arbitrary matter of taste, otherwise he undermines his central
claim that choice among voting rules is arbitrary. He then reverts to the
position that preferences are unknowable, the ultimate bulwark in each
of his lines of argument, Doubtless the results are often fair or true; but,
unfortunately, we almost never know whether they are or not, (1982,
113): the basic argument pattern.
Adecision rule must work with more than two alternatives, but fromthe
same prole of voters preferences it is logically possible for different de-
cision rules to yield different outcomes, according to Riker. It is logically
possible that different rules yield different outcomes, but it is not empir-
ically probable, we have seen. Simulations and voting studies show that,
with more realistic distributions of preference rankings, the reasonable
voting rules yield similar results. Moreover, an axiomatic approach does
not identify a uniquely best decision rule, each voting rule violates one or
another principle froma collection of apparently desirable a priori princi-
ples, Riker continues. The axiomatic approach does not identify one best
rule, but it does narrow the eld to a handful of reasonable voting rules.
That handful of rules yields similar results in real circumstances. If accu-
racy and fairness are the only criteria, then the Borda count is probably
the best. Additional pragmatic criteria support more simple and already
widely used voting rules, such as the plurality-runoff rule and even the
plurality rule. The accumulation of social-choice results does not ob-
scure judgment as to choice of voting rule for the purpose of institutional
design; rather, the social-choice results inform and improve such design
judgments. Next, we shall consider the claim that from the same prole
of preferences even the same voting rule might yield different results.
4 The Arrow general possibility theorem
Democratic voting as meaningless
First, irrationalism claims that voting is arbitrary. Second, irrationalism
claims that voting is meaningless: even if a voting method survives the
rst claim as fair, it is yet meaningless, because: (a) the outcome of vot-
ing is manipulable; and (b) we cannot know that manipulation occurred
since again there is not enough information available from the data of
voting to know the preferences underlying choices expressed in voting.
The second claim of meaninglessness presents and interprets results of
social choice theory. We have already treated (in Chapter 2) the crucial
premise that preferences cannot be known fromchoices. Now, we will be-
gin examination of the premise that voting is manipulable. The premise
of manipulability is derived from the possibility of majority cycling as
shown by Arrow (1963/1951), the possibility of strategic voting as shown
by Gibbard (1973) and Satterthwaite (1975), the possibility of agenda
control as shown by McKelvey (1976) and Schoeld (1978), and nally
the strategic introduction of new issues and dimensions (Riker 1982).
Over the next three chapters, we discuss the Arrow theorem. In this
chapter, we review the origins of the Arrow theorem in the ordinalist rev-
olution in economics, and distinguish social choice as welfare economics
from social choice as voting theory. Next, we present the contents of the
Arrowtheorem, followed by discussion of claims of its empirical relevance
by Arrow and Riker. Then we review all studies found on the question
of the frequency of cycles, and conclude that the incidence of cycles is
rare. Finally, we begin review of justications of the conditions of the
theorem.
The origins of social choice theory
Classical economics emphasized the remarkable coordinating power of
markets, and suggested that a policy of laissez-faire would best ad-
vance the wealth of nations. Bentham and the utilitarians embarked on
72
The Arrow general possibility theorem 73
a wholesale program of practical reform, which overturned traditional
policies, and enacted new policies, many laissez-faire, intended to do the
greatest good for the greatest number. Economists were still bedeviled
by the value paradox why should useful water cost less than useless
diamonds? The marginalist revolution of 18711874 solved that prob-
lem. The price of something is related to its marginal utility, not its
total utility water has great total utility, but in normal circumstances
one more unit has little marginal utility and consumer satisfaction
is maximized when the ratios of marginal utility to price are equal for
each good. Further, goods have a declining marginal utility: after some
point, each additional increment is worth less than the prior increment.
The intersection of utilitarianism with marginal analysis in Marshalls
neoclassical economics yielded a conclusion distasteful to those fond of
laissez-faire: if for each person there is a declining marginal utility of
money, then it would increase overall social welfare if money were taken
from the rich and given to the poor, up to the point equalizing marginal
utility of each person in society. Efciency would best be achieved by
equality.
Utilitarianism had assumed cardinal and interpersonally comparable
utilities, and the utilitarian philosopher proposed pursuit of the great-
est good for the greatest number, that society should maximize the total
sum of utility. Cardinal utility counts it as meaningful to say that I want
a holiday in Andalusia ve times more than I do a holiday in Buffalo;
interpersonal comparability counts it as meaningful to say that Paul likes
playing the guitar more than Matthew likes doing the dishes. Although
interpersonal comparisons of welfare are common in daily life, and in my
view are quite meaningful, they are always open to skeptical attack, and
it was once thought that there are insurmountable difculties in devis-
ing satisfactory formal representations of such comparisons. Meanwhile,
economists found that they could restate the basic propositions of market
economics in terms of ordinal and noncomparable utility. The advan-
tage is that these are less demanding assumptions, and there are not so
many formal and conceptual problems as there are with cardinal utility,
comparable utility, or both. Ordinal utility considers only the order of
ranking of alternatives; in the ordinal framework I cannot say that I like
Andalusia ve times better than I do Buffalo, only that I like Andalusia
better than Buffalo. Further, a notion such as that society should max-
imize the total sum of utility is not possible within an ordinal and non-
comparable framework; for one thing, you cant add what you cant
compare.
The chief ideologist of the ordinalist revolution was Lionel Robbins
(1937/1932), who wanted to establish economics as a Science, and to
distinguish it from Ethics. The purpose of a distinction between facts
and values, is, according to Iris Murdoch (1992, 25):
to segregate value in order to keep it pure and untainted, nor derived from or
mixed with empirical facts. This move however, in time and as interpreted, may
result in a diminished, even perfunctory account of morality, leading (with the
increasing prestige of science) to a marginilisation of the ethical. This originally
well-intentioned segregation then ignores an obvious and important fact of hu-
man existence, the way in which almost all our concepts and activities involve
evaluation.
I agree that it is quite important to distinguish fact from value, but notice
that the claim I just made involves an assertion of fact and an expres-
sion of value. Too often, the discourse which states that value claims are
nonscientic (in the descriptive sense) is twisted into an insinuation that
value claims are unscientic (in the evaluative sense), or merely arbitrary
expressions ungrounded in reason. Robbins tends to do this himself, in
imagining that a committee made up of an economist, Bentham, Buddha,
Lenin, and the head of US Steel would be unable to agree on the ethics of
usury, but that the same committee would be able to agree on the facts of
the economic consequences of anti-usury legislation (1937/1932, 150
151). Values are arbitrary, according to Robbins (150): If we disagree
about ends it is a case of thy blood or mine or live and let live according
to the importance of the difference or the strength of the opponents.
At the same time he does not seem to be aware that his recommen-
dations as to what should count as science are matters of evaluation,
not of fact.
Robbinss second move was to allege that the claims of the reigning
material-welfare school in economics were not scientic but ethical, and
not just ethical, but arbitrary because ethical. The material welfare school
held, according to Robbins, that it is possible to compare the utility or
satisfaction of one person to another person. But such comparisons are
not needed in modern economic theory, he wrote; the comparison is
essentially normative and has no place in pure science.
There is no means of testing the magnitude of As satisfaction as compared with Bs. If
we tested the state of their blood-streams, that would be a test of blood, not satis-
faction. . . There is no way of comparing the satisfactions of different people . . . In
Western democracies we assume for certain purposes that men in similar circum-
stances are capable of equal satisfactions . . . although it may be convenient to
assume this, there is no way of proving that the assumption rests on ascertainable
fact. And, indeed, if the representative of some other civilization were to assure
us that we were wrong, that members of his caste (or race) were capable of ex-
periencing ten times as much satisfaction from given incomes as members of an
inferior caste (or an inferior race), we could not refute him. . . we could not
show that he was wrong in any objective sense, any more than we could show that
we are right. (Robbins 1937/1932, 139140)
Fromthis he concludes that the recommendation by the material-welfare
school for equalization of incomes is unscientic. Notice that Robbinss
doctrine is radical skepticismrather than mere behaviorism: any objective
correlate of satisfaction is prohibited. If we could measure some chemical
in the blood (or these days, study an image of the brains activation), that
would not do; we would be measuring blood, not satisfaction. Robbinss
objection seems to me to be one of postured philosophy rather than of
ordinary science science frequently estimates unobserved variables by
way of indirect measures, without calling into doubt the theoretical use-
fulness of the unobserved entity. I agree with Robbins, though, that in
any case the next step of saying that satisfaction or some other measure
should be equalized or any other policy recommendation, including that
it should not be equalized is a normative question.
Cooter and Rappoport (1984) argue that Robbins and his followers
misconstrued the material-welfare school they superseded. Generally, the
material-welfare school understood that ordinalism is sufcient for mar-
ket economics, and that it may not be possible to compare the satisfactions
of any two individuals. Where they differed was in judging that ordinal-
ism is not sufcient for welfare economics, and in thinking it is possible
to compare the needs of representative persons: If people typically desire
what they need, and if needs are more urgent when people are poor,
then it follows that additional income is more useful to the poor than the
rich (Cooter and Rappoport 1984, 517). In their social-welfare calcula-
tion, goods were evaluated objectively, by whether they contributed to a
persons physical well-being, they distinguished necessities fromcomforts
fromluxuries, and they measured variation among individuals in the sup-
ply of health, food, housing, clothing, and money. The welfare economist
need not be conned to the equalization of satisfaction, a mental state;
the welfare economist could have an objective theory of the human good,
justied in its own right, and not solely because it correlates with the
desire-satisfaction of the typical individual. Sen (1999; see also 1982)
says that the rejection of interpersonal comparisons in welfare economics
was based on interpreting them entirely as comparisons of mental states.
He argues that, even with such mental state comparisons, the case for
unqualied rejection is difcult to sustain (1999, 358). He continues
that such comparisons need not be based only on mental states, but might
directly be based on incomes, or commodity bundles, or resources more
generally, and Sens theoretical and applied work in this area demands
attention.
1
To conclude, is giving food to the hungry better than giving
opera tickets to the bored (Cooter and Rappoport, 1984, 519)? I agree
with Robbins that such is an ethical claim, but disagree that it is arbitrary
because ethical.
As economists completed their ordinalist revolution in the 1930s, they
sought in welfare economics to devise an ordinalist replacement for the
utilitarian formula. They had the Pareto criterion, that x is better than
y if every individual ranks x higher than y, but that criterion is radically
incomplete: a policy change that helps a million people but hurts one is
not a Pareto improvement, and further there is no way to choose among
a multitude of Pareto-superior states. As it happens, voluntary market
exchange satises the Pareto criterion, but collective choice short of una-
nimity does not, and thus any political distribution of endowments other
than the inherited status quo is off limits. Notice that most voting schemes
are ordinalist voters are asked to rank-order alternatives, not to state
intensities on some scale. There was a minor, almost forgotten, current
of analytic consideration of alternative voting rules, involving Condorcet
and Borda, and later Lewis Carroll, Hare, Nanson, among others, and
culminating in Duncan Black. Ordinalist welfare economics intersected
with voting theory, and from Arrows achievement social choice theory
was born.
Social choice theory spans at least three disciplines. Arrows theorem
and much of the discussion of it is motivated by the concerns of wel-
fare economics, that is, what advice should be given to an imaginary
social planner who has the task of providing the greatest social welfare
to a society (the discussion includes those economists who dispute the
legitimacy of such a social-welfare objective). There is also, of course,
a large philosophical literature on theories of justice, each of which as-
pires to provide thorough and coherent arguments about the best way to
organize the basic institutions of a society. The concerns of the justice the-
ories overlap somewhat with the concerns of welfare economics, but also
typically assert standards of the public good (or some similar objective,
such as justice) that are partly or wholly independent of individuals or-
dinary rankings of social states. Welfare economics purports to be merely
descriptive (or at most to provide hypothetical advice), but theories of
justice are frankly normative. Finally, political scientists interested in the
formal and empirical exploration of voting and of other political institu-
tions adapted the ndings of social choice theory to their purposes. Much
of the content of American political science in the last twenty years has
been an elaboration of the analogy of political choice to consumer choice
rst brought to prominence by Arrowand Buchanan in the 1950s. Imme-
diately upon publication of Arrows Social Choice and Individual Values,
Little (1952) objected that Arrows scheme was excessively general in
lumping the economists interest in social welfare with the political theo-
rists interest in voting, and Sen (1982, 158200) later elaborated on the
point. My emphasis in this volume is on voting, and I will mostly ignore
the concerns of welfare economics and of theories of justice, not from
lack of interest, but rather because it would take us too far astray even to
state the many issues at stake.
Immediately, democratic voting is different fromsocial-welfare calcula-
tions in at least one important respect: it is widely although not universally
accepted that citizens should be treated as political equals, in terms of vot-
ing that each relevant person should have an equal vote. The economist
and the philosopher are free to muse, for example, that one person might
be a hundred times better than another in converting lifes experiences
into some form of satisfaction, so that in the pursuit of equality of welfare
such a person deserves fewer resources or perhaps less political inuence
over the distribution of resources. Or one or the other of them also might
propose that the demonstrably competent be granted more than an equal
vote or the overly privileged less than an equal vote. The democratic theo-
rist, however, is entitled to the working assumption of formal equality, one
vote per one person. The economist or the philosopher might consider
that assertion of political equality a defect of democracy as it is presently
understood. That would denitely be a minority view, however, and even
if it were a correct view I cannot imagine in todays circumstances how
the supposed defect would be remedied in practice: if one departure from
formal political equality is granted, that only increases the demand that
another be granted, escalating into a chaos of exceptions only resolved
by a return to formal equality.
Most practical voting systems in use are ordinal: generally, voters are
asked to rank-order two or more alternatives and are not asked to ex-
press by voting intensity of preferences over alternatives, for example,
that Rome is three times as good as Santa Fe and Santa Fe is twice as
good as Gary. A cardinal voting scheme, the argument goes, may be
vulnerable to misrepresentation: it is in each voters interest to exagger-
ate the intensity of her preferences for her favorite choices. Intensity of
preference is expressed in informal discussions and in democratic de-
bate. The fact that Susie hates seafood because she has a life-threatening
anaphylactic reaction to shellsh is enough to overrule a tepid major-
itys preference for the seafood restaurant unless Susie insincerely has
such a story for every occasion. Susies claim also relies on some kind
of comparability: her life-threatening reaction is much more important
than Marks mild distaste for pizza. A minoritys demand to be free from
arbitrary deprivation of life and liberty can persuade a majority to de-
sist, perhaps because the majority is well-motivated or perhaps because
the minority threatens to withdraw social cooperation. One member of
Congress can tell another that she really needs this vote for her district
and would gladly trade votes with others on issues she is nearly indif-
ferent about; there is an element of comparability in that each legislator
is allocated only one vote per question. There are voting schemes and
practices that under one interpretation approximate cardinality but are
not vulnerable to exaggeration: the Borda count (the Borda count need
not be justied, however, as an approximation of cardinality), cumula-
tive voting (the voter is allocated a xed number of points to distribute
across alternatives), and under ordinalist majority rule vote trading across
a series of issues. Because most voting schemes are ordinal in character,
Arrows theorem and its many offspring are relevant to questions about
the comparative desirability of alternative voting schemes and broader
questions of institutional design. Any voting scheme assumes some kind
of comparability: allocating the same voting power to each person as I
have advocated, or weighting votes so as to favor one voter over another,
justied by either welfare or nonwelfare considerations. Arrows theorem
assumes ordinal and noncomparable preferences, and voting tends to be
ordinal but imposes one or another conception of comparability.
Arrow theorem
The Condorcet paradox of voting, recall, arises from a possible distribu-
tion of preference orders among the population such that the aggregate
majority vote is a cycle A > B > C > A. Arrows possibility theorem
can be understood as a generalization of Condorcets paradox, applying
not only to simple majority voting but also to any social-welfare function
that aggregates individual orderings of more than one person over more
than two alternative social states. The theorem shows the joint incon-
sistency of several innocuous-sounding conditions on the social-welfare
function. There are many ways to state, informally and formally, the con-
ditions and the results. A good way to begin is with Arrows (1973) own
informal summary of his theorem:
I stated formally a set of apparently reasonable criteria for social choice and
demonstrated that they were mutually inconsistent . . . The conditions on the so-
cial decision procedure follow: (1) for any possible set of individual preference
orderings, there should be dened a social preference ordering (connected and
transitive) which governs social choices; (2) if everybody prefers alternative A to
alternative B, then society must have the same preference (Pareto optimality);
(3) the social choice made from any set of available alternatives should depend
only onthe orderings of individuals withrespect to those alternatives; (4) the social
decision procedure should not be dictatorial, in the sense that there is one whose
preferences prevail regardless of the preferences of all others . . . The incon-
sistency of these conditions is in fact a generalized form of the paradox of
voting . . .
As in the original Condorcet case of simple majority voting, all that is meant by the
paradox is that it could arise for certain sets of individual preference orderings. If
individual preference orderings were restricted. . . then majority voting and many
other methods would satisfy conditions (2) to (4).
I shall now state some preliminary denitions, and then the assump-
tions of the Arrow theorem, along with some interpretations of the for-
malisms. To begin with, there are alternative social states. A social state,
according to Arrow (1963/1951, 17), is a complete description of the
amount of each consumption commodity, of labor, of productive re-
sources allocated in the economy, and amounts of all collective activities,
ranging from municipal services, to diplomacy, war, and the erection
of statues to famous men. There is an environment X of all alternatives,
and a set S that is a subset of X. Arrows framework is quite general, and
X could be all possible social states and S all feasible social states. In a
specic application such as a presidential election X might be all possi-
ble presidential candidates and S all actual presidential candidates. Each
individual has a preference ordering over all possible social states, and an
individuals preferences need not be egoistically oriented, according to
Arrow. A weak ordering, R, is a generalization of the concept applied to
real numbers of greater than or equal to; a strong ordering, P, is a gen-
eralization of greater than. Strong ordering can be dened in terms of
weak ordering: x P y is dened as x R y and not y R x. In other words, to
say that Italy is better than England is the same as to afrm that Italy is
at least as good as England and to deny that England is at least as good
as Italy. Indifference can also be dened in terms of weak ordering: x I y
is dened to be x R y and y R x. To say that Coke is as good as Pepsi is
the same as saying at the same time that Coke is at least as good as Pepsi
and Pepsi is at least as good as Coke.
Arrow assumes that individuals have consistent preferences over all
possible states of the world. A weak preference ordering is reexive:
x R x means that x is at least as good as itself. A weak ordering is also
connected, or complete: for all x and y, either x R y or y R x. Someone
might be indifferent between Coke and Pepsi, yet she could compare the
two. A person who had no weak preference (or strong preference or in-
difference) between joining the rst space voyage to another inhabited
planet and spending the same number of years with Socrates would have
preferences that were not complete and thus not an ordering. A weak
ordering is also transitive: for all x, y, and z, if x R y and y R z, then
x R z. Arrows individual would never, for example, prefer Tocqueville
to Marx to Mill to Tocqueville. The possibility theorem shows that, if
we accept Arrows conditions, such individual orderings (by denition
reexive, transitive, and complete) cannot be amalgamated into a col-
lective ordering (also by denition reexive, transitive, and complete).
For example, we have seen with the Condorcet paradox that transitive
individual preferences can result in a social preference that is intransitive
and thus not an ordering. Another voting rule could have the defect that
it is incomplete; for example, a voting rule that required unanimity for
every decision would in the abstract, if there were any disagreements,
result in an incomplete social preference, and would be unable to report
social preference or even indifference between some number of alterna-
tives (because of this incompleteness, unanimity rules in practice favor
the status quo). To continue with notation, individuals orderings are
denoted R
1
, . . . , Ri and a collective ordering is denoted R (without any
subscript).
Now I turn to Sens widely used formulations.
r
An element x in S is a best element of S with respect to a binary relation
R if and only if y: (y S: x R y). The set of best elements in S is
called its choice set and is denoted C(S, R) (Sen 1970, 10). To say that
x is a best choice (in the set S with respect to the relation R) means that
for all y, if y belongs to S then x is weakly preferred to y. The choice
set contains the best elements. We might call them the winners of the
contest. If collective preference cycles among the top alternatives, then
there is no best element and the choice set is empty.
r
A collective choice rule is a functional relation f such that for any set
of n individual orderings R
1
, . . . , R
n
(one ordering for each individ-
ual), one and only one social preference relation R is determined, R =
f(R
1
, . . . , R
n
). (Sen 1970, 28). For Sen, the collective choice rule is
the more general case in which the social preference relation need not
be an ordering; it is also made clear that a unique social preference
relation is required. Now follows the Arrow theorem, which assumes
rather a social preference relation that is an ordering. The conditions
are labeled O, U, P, I, and D.
r
A social welfare function (henceforth, SWF) is a collective choice rule f,
the range of which is restricted to the orderings over X. This restriction
is to be called condition Oon f (Sen 1970, 41). The collective ranking
of alternatives generated by the social-welfare function should be as a
collective choice rule unique, and as a social welfare function both
complete and transitive.
r
Condition U (unrestricted domain): The domain of the rule f must
include all logically possible combinations of individual orderings
(Sen 1970, 41). The social-welfare function should accept as input
any and all possible individual preference orderings.
r
Condition P (Pareto principle): For any pair x, y in X, [i: x P
i
y]
x P y (Sen 1970, 41). For example, if every individual prefers Metallica
to AC/DC, then Metallica is preferred to AC/DC in the social prefer-
ence order.
r
Condition I (independence of irrelevant alternatives): Let R and R
be
the social binary relations determined by f corresponding respectively
to two sets of individual preferences, (R
1
, . . . , R
n
) and (R
1
, . . . , R
n
).
If for all pairs of alternatives, x, y in a subset S of X, x R
i
y
x R
i
y, for all i, then C(S, R) and C(S, R
) are the same. Condition I

is the condition most difcult to understand and the most frequently
misunderstood. The social preference over any given pair of alterna-
tives depends only on individuals preferences over the same given pair
of alternatives; and if individuals preferences about some third alter-
native should change, that would not change the social preference over
the given pair of alternatives.
r
Condition D (nondictatorship): There is no individual i such that for
every element in the domain of rule f, x, y X: x P
i
y x P y (Sen
1970, 42). If the social preference over every two alternatives is the same
as one particular individuals preference over every two alternatives,
regardless of the preferences of other individuals, we could call that a
dictatorship.
r
Arrows General Possibility Theorem: There is no social welfare function
(an ordering) that satises the conditions of universal domain, Pareto
principle, independence of irrelevant alternatives, and nondictatorship.
Here is more on the independence condition. Suppose that we have two
individuals, Napoleon who ranks b > a > c > d > e, and Josephine who
ranks a >b >c >d >e. In order to rank a and b, what information does the
social-welfare function that obeys Condition I take into consideration?
Only that Napoleon ranks b over a and that Josephine ranks a over b,
and nothing else. Intuitively, if this were the only information we had,
and if we regard Napoleon and Josephine as equals, we would probably
conclude that given those individual rankings the social choice between
a and b should be a tie. Now suppose that as before Napoleon ranks
b > a > c > d > e, but that Josephine is different and ranks
a > c > d > e > b, for Josephine alternative b has dropped from sec-
ond place to fth place. What information does Condition I permit to
be taken into consideration? Again, only that Napoleon ranks b over a
and that Josephine ranks a over b, and by simple majority rule a would
tie with b. The alternatives deemed irrelevant in this illustration are any
other than a and b. Someone might object: alternative b went fromsecond
to fth in Josephines preferences, it seems that Josephine really hates b,
so it makes more sense now to say that a beats b rather than that a ties b.
Condition I forbids that objection: the social-welfare function accepts
only pairwise information from individual preference orders. My expla-
nation of Condition I is not typical in the literature, which, if it offers any
explanation at all, usually provides an example that makes violation of
Condition I look silly.
A good proof of the theorem can be found in Sen (1970, 4146). Nat-
urally, the proof logically depends on the assumed conditions. The work
the universal domain condition, U, does in the proof is that by allow-
ing any logically possible combination of individual orderings it allows as
one possible instance a cyclical prole of individual orderings such as that
which gives rise to the Condorcet paradox. If the cyclical prole is ex-
cluded for one reason or another, the proof does not go through, as Arrow
himself notes in his informal remarks I quoted above. The work that the
independence of irrelevant alternatives condition, I, does in the proof is
to exclude information other than individuals preferences over pairs. If
we could include information about individuals relative rankings over
more than two alternatives, also the proof would not go through. Condi-
tion Uexcludes Condorcet voting. Condition I excludes Borda and many
other methods. Condition D excludes the only remaining voting rule, the
dictatorship of one.
Just as the Condorcet paradox is used to shock at the elementary level of
study, the Arrow possibility theorem is used to shock at a more advanced
level. There are thousands of articles varying, extending, and elaborating
the theorem. It is often said that the Arrow ndings are robust to several
variations in the assumptions, and that if the spirit of the conditions is
accepted, then there is no magic bullet that puts the problems to rest.
That is not quite right. Sen, the social choice theorist who later won the
Nobel Prize (a criterion cited by the speakers in my hall of quotations),
says that Arrows conditions are not inescapable commandments. The is-
sue is not the absence of rationally defendable social decision procedures,
but rather the importance of disparate conditions that pull in different
directions as we evaluate diverse procedures, he says. We are not at the
edge of a precipice, trying to determine whether it is at all possible for
us to hang on (Sen 1995, 11). The usual attention-getting way of stat-
ing the result is that the only social-welfare function that satises some
simple and apparently fair conditions is a dictatorship. A more boring
but in my view more appropriate way of stating the Arrow result is to
say that if we are required to consider only individuals rankings of pairs
when the collective choice is over more than two alternatives (I ), or if we
must assume that there is no correlation among different individuals
preference orderings (U), then there may be a cycle in the social
choice.
Before we interrogate justications for the several Arrow conditions,
we shall examine the crucial issue of empirical relevance. Throughout
his career Arrow has asserted the empirical relevance of cycling, and he
came to rely on Rikers ndings in support of that belief. In his rst major
work, Social Choice and Individual Values (1963/1951, 3), he introduces
the Condorcet paradox and asserts in a footnote that there were cycles
in recent Congresses over no federal aid to education, federal aid only to
public schools, and federal aid to both public and religious schools, but
he does not develop or defend the assertion. In his 1963 (93) postscript
to the 1951 volume, Arrow cites Riker (1961) as the most complete and
up-to-date summary of the problem of aggregation of individual choices
into collective ones, with particular emphasis on political aspects. Arrow
(1963/1951, 120) repeats: That an intransitive social choice mecha-
nism may as a matter of observed fact produce decisions that are clearly
unsatisfactory has been brought out . . . by Riker . . . Rikers emphasis is
on the possibility that legislative rules may lead to choice of a proposal
opposed by a majority. The work to which Arrow refers (Riker 1961)
was written while Riker was visiting the Center for Advanced Study at
Stanford, where Arrow was in the Economics Department, and Riker ac-
knowledges the generous help and commentary of Arrow on the paper.
The paper is a bibliographic survey, and concludes with a declaration
of the Rikerian doctrine that many observed majorities are merely ap-
parent because of underlying cycles, providing as examples Riker (1958)
on a Congressional appropriations vote (which I show below to be mis-
taken) and a brief mention of Senate deliberations on the 17th Amend-
ment (later developed in Riker 1982, and which I also show below to be
mistaken).
In the same passage, Arrow refers to a few pages in Dahls (1956,
3942) A Preface to Democratic Theory. Dahl maintains that when roughly
equal factions in a group favor mutually exclusive alternatives then demo-
cratic rule may be endangered. If democratic procedures do not provide
a unique outcome, then the factions may pursue their goals by extra-
democratic means, that is, by violence. Deadlock leading to violence
would be avoided only if the groups valued maintenance of democratic
rule more than they did the nondemocratic pursuit of their goals, accord-
ing to Dahl. Indeed, the closer a group approached to an equal division
the less valid the majority principle becomes Dahl believes (1956, 41).
Although social choice theorists call majority rule decisive, that is only
by means of the denitional at that a tie counts as a decision, in the
real world ties are useless when collective action is urgently required.
Dahls concern is illustrated by the confusion and anguish of the 2000
US presidential election; in the end it mattered less whether Gore or Bush
won than that one of them should win (it had to be one of either Gore or
Bush; in the circumstances selection of some third candidate would have
been truly arbitrary). In a footnote, Dahl includes within his concern a
cyclical collective outcome arising from three equally sized groups with
a cyclical prole of individual preferences. In other words, for Dahl, a
cycle over more than two alternatives is a defect of the same type as
a tie between two alternatives. The defect is easily remedied. First, as
Dahl mentions, if the population favors democratic rule generally over
the nondemocratic pursuit of goals in the particular circumstances of tied
outcomes, then democratic stability follows. Second, if ties are a problem
then a tie-breaking procedure should be institutionalized in advance: bias
to the status quo alternative, or perhaps better a random procedure, or
a Republican-majority US Supreme Court, or a neutral constitutional
monarch. Notice, however, that the problem of deadlock is not conned
to tie votes in a majority-rule setting. If the constitution is of the pres-
idential or Madisonian or Rikerian liberal variety, which in the name
of checks and balances allows each of many different minorities a veto
power over changes from the status quo, deadlocks will be more frequent
and thus more dangerous, especially when the status quo is worsened by
unanticipated exogenous factors, than if the constitution is of the par-
liamentary or majoritarian or Rikerian populist variety (Stepan and
Skach 1994).
Arrow (1960) observes that means of minimizing the deadlocks that
arise from the paradoxes of collective choice have evolved in all demo-
cratic systems. These brief notes are interesting, because elsewhere Arrow
seldom engages in empirical reections. He rehearses the defects of plu-
rality rule, including that the rule strategically elicits two alternatives and
thus disguises possible cycles. Plurality runoff, he says, might exclude a
Condorcet winner. Possible ties in US presidential elections are decided
by the election going to the US House of Representatives the theme,
apparently following Dahl, is avoidance of deadlock. Finally, in a recent
volume Arrow (1997, 5) says, That there is nothing unlikely about [the
Condorcet] paradox has been empirically documented by a number of
political scientists beginning with Riker (1958). It was suggested to me
that Arrows theorem was not intended by its author, nor understood
by its audience, to be of empirical relevance; rather it is merely a logical
exercise that illustrates a limit case (which is how it should be under-
stood, I believe). I have shown that Arrow himself was motivated by the
empirical relevance of his theorem; and that no small part of his audience
shares that motivation is demonstrated by the literature of Riker and his
followers. Furthermore, consider Arrows (1963/1951, 21) methodolog-
ical credo:
the present author regards economics as an attempt to discover uniformities in
a certain part of reality and not as the drawing of logical consequences from a
certain set of assumptions regardless of their relevance to actuality. Simplied
theory-building is an absolute necessity for empirical analyses; but it is a means,
not an end.
I read that to mean that the model and its logic is of interest only to the
extent of its empirical applicability.
The later Arrow(1997, 4) asks, regarding social choice theory, Does it
say that democracy is impossible? He answers that, Social choice theory
offers only a limited criticismof democratic procedures (5). He says that
although failure to satisfy the theorems conditions is a legitimate criticism
of a procedure, since the failure is universal the theorem alone offers no
basis for differentially evaluating alternative social choice mechanisms
(8; including, I would add, the market). Instead, In the case of real
social choice procedures, we have to consider the frequency with which
intransitivities [and other violations] occur. This is not the sort of result
I like, but that is the way the world is (8).
Arrow implicitly rejects the claim by Riker and his followers that the
general possibility theorem renders democracy impossible. Arrow explic-
itly accepts the claimby Riker and his followers that empirical cycles have
been robustly demonstrated, and he has always held that the relevance
of his theorem is an empirical question. What is the frequency of social
intransitivities?
[E]mpirical observations of a wide variety of actual collective decision-making
processes indicate that cyclical majorities are very rare. Thus, cycles do not appear
to be a real problem for group decision-making although some paradoxes may
occur which may go undetected. (Feld and Grofman 1986)
Actual observations about majority cycling are scarce, because elections
rarely generate data on pairwise comparison among all alternatives or a
ranking of all alternatives (Gehrlein 1983). In the limited instances where
such data were available fromexperimental subjects or by inference froma
sample of actual legislative situations, the absence of a Condorcet winner
tended to be infrequent, according to the summary by Gehrlein (1983).
Since then, whenever good data were available and analyzed, cycles have
been shown to be infrequent, as we shall now see.
Whats the frequency?
Dobra (1983) solicited real election data from the readers of the journal
Public Choice. This is no random sample: the readers of Public Choice are
motivated to notice and report cycles. The cases reported were mostly
faculty searches and from a small experiment, with small numbers of
voters (median 10, range 4 to 27) and larger numbers of alternatives
(median 5, range 3 to 37). There were three cycles including ties (e.g.,
A > B > C A) and one cycle not including a tie out of the 32 cases.
Although cycles were infrequent, Dobra holds that the infrequency does
not repudiate the work of the disequilibrium theorists (247).
Recall the study by Chamberlin, Cohen, and Coombs (1984) of ve
different presidential elections of the American Psychological Association
(APA) where voters rank-ordered all candidates, permitting hypothetical
comparison of voting rules. The data also permit examination of aggre-
gated preference orders for the presence or absence of cycles. Only about
half the voters in these elections ranked all ve candidates; the authors
generated complete ballots in one condition by lling in remaining pref-
erences randomly (as if voters were indifferent to unranked candidates)
and in another condition by lling in remaining conditions proportion-
ately (as if voters were uninformed about unranked candidates). Either
way, there was a transitive majority ordering of the candidates in all ve
elections. The actual APA preference orders yield zero cycles. With the
same numbers of alternatives and voters as in the APAelection, but rather
an impartial culture, there would be a 24 percent expectation of cycles.
There are no cycles in the APA elections because preference orders are
not randomly distributed as they are under the impartial-culture assump-
tion. Natural preference orders have some minimumof structure, such as
to preclude cycles nearly all of the time. As mentioned above, alternative
voting rules picked the Condorcet winner about 80 percent of the time,
and the second-ranked Condorcet candidate the rest of the time, again
because of minimal similarity among preference orders.
Niemi and Wright (1987) looked at thermometer ratings (respondents
rate 0 to 100 when 0 means very unfavorable and 100 very favorable)
of 14 politicians who were potential or actual candidates for the 1980
American presidential election, from a nationally representative sample
of US voters. If all preference orders within a group are single-peaked (to
be detailed below) and thus unidimensional then no cycle is possible. If
only one of the voters rankings within a group fails the single-peakedness
criterion then a cycle is possible but it is most improbable; and the higher
the proportion of single-peaked preference orders within the group the
less likely is a cycle (Niemi 1969). For three-, four- and ve-candidate
groups, Niemi and Wright found that the observed proportions of unidi-
mensional preference orders in each condition are much higher than what
would be expected by chance under the impartial-culture assumption;
and those observed proportions were such that the probabilities of cy-
cles were 0.04 or less. Further, the ten three-candidate rankings with
the worst unidimensionality were sampled 1,000 times each; there were
29 cycles in these 1,000 samples, and 22 of those cycles involved
the most obscure gure of the 14 candidates (Lucey, John Andersons
vice-presidential running mate). For the least unidimensional four-
candidate rankings, top cycles were found in 154 out of 1,000 samples,
and 137 of those occurred in two of the ten rankings. They also found
absence of a relationship between unidimensionality on the one hand
and on the other hand more distinguishable candidates, voter education,
voter partisanship, judging better-known candidates, longer campaign
exposure, or within-party judgments. Curiously, however, the dimension
used by voters in their data did not appear to be the standard leftright
dimension, but rather perhaps one of likability.
Some very good data from candidate elections in private organiza-
tions in Great Britain were analyzed by Feld and Grofman (1992). These
36 elections, with between 3 and 29 candidates and between 9 and
3,422 voters, were conducted by the single transferable vote procedure,
which requires that voters rank-order candidates. The authors state that
although strategic voting is possible in principle under single-transferable
vote, the necessary calculations are too complex for it to occur in prac-
tice (Bartholdi and Orlin 1991 show that it is NP-complete, that is, too
computationally complex, except in special cases, for voters to be strate-
gic under single transferable vote). From the data of the voters rank-
orderings, pairwise comparisons can be constructed. Every one of the re-
constructed elections had a Condorcet winner. In 34 of the 36 elections
the Condorcet winner and the Borda winner coincided (so the results
speak as well against Rikers claim that democracy is inaccurate). Only
0.5 (one-half of one) percent of the linearly ordered triples in the sample
universe were cyclic; 24 of the 36 elections had no cycles whatsoever, the
largest percentage of cycles in any election was 2.0 percent; and almost
all cycles were among alternatives adjacent in Borda scores (meaning that
they were among close alternatives). Since many observers believe actual
cycles are infrequent, with those asserting otherwise emphasizing the ab-
sence of data to make a condent determination, I agree with Feld and
Grofman that their results are important.
Felsenthal, Maoz, and Rapoport (1993) review a mostly overlapping
set of elections, and make similar ndings. They observe that among elec-
tions with eight or fewer candidates there is only one instance of a (minor)
cycle, and among elections with more than eight candidates there are 13
out of 19 instances of (minor) cycles. Felsenthal and Machover (1995)
consider an expanded set of 92 elections, and replicate the observation:
7.5 percent of elections with eight or fewer candidates contain cycles, but
56.4 percent of elections with more than eight candidates do, and they re-
mark on the sharp jump in occurrence of cycles beyond eight candidates.
I shall propose an explanation for this in the next chapter. Of the 92 elec-
tions, 26 contained cycles; no cycle involved all candidates, and only two
were cycles at the top of the preference order. Of the 26 elections with
cycles, winners were immediately obvious in 19 of the elections, either be-
cause the cycle was not at the top, the number of slots to ll exceeded the
length of the cycle, or indifference relations within cycles appropriately
indicated winners.
Radcliff (1993) reports high unidimensionality derived from studies
of American presidential elections from 1972 to 1984: 77 percent to
85 percent single-peakedness in years with three major candidates in the
primary and general elections (at 75 percent unidimensionality the expec-
tation of a cycle is less than 1 percent, and at 80 percent the expectation is
almost zero, Radcliff 1994); and 50 percent in 1980 when there were ve
major candidates. Radcliff (1993) also nds that most individual voters
rankings are transitive but that individual intransitivities increase with
number of candidates. If intransitive voters are removed from consid-
eration in the ve-candidate case, single-peakedness goes up to 70 per-
cent. Radcliff (1994) uses data fromthe same election studies to examine
the transitivity of collective rankings. There was a Condorcet ordering
in each American presidential election studied (that is, no cycles); the
ordering corresponded to the standard leftright dimension of under-
standing; and each actual election picked the Condorcet winner. Thus:
1972, Nixon >Humphrey >McGovern; 1976, Carter >Ford >Reagan;
1980, Reagan > (Carter Bush) > Anderson > Kennedy; 1984,
Reagan > Hart > Mondale.
Van Deemen and Vergunst (1998) continue the elusive quest for the
empirical cycle. From the Dutch parliamentary election studies of 1982
(13 parties), 1986 (12 parties), 1989 (9 parties), and 1994 (9 parties)
they have survey data on respondents preferences over the alternatives.
If preference rankings were random, as under the impartial-culture as-
sumption, then there would be about a 50 percent chance of cycles. There
are, however, no cycles in their data, not anywhere in the rankings, not in any
of the elections. The authors nd the results surprising: for some reason
or another cycles in large elections are scarce (485). Kurrild-Klitgaard
(2001a) investigates cycles in Danish national election surveys. Respon-
dents ranked 11 parties in 1973, 9 parties in 1994, 11 parties in 1998;
the policies of the 8 parliamentary parties in 1994; 10 party leaders in
1973, 8 in 1994, and 10 in 1998; the importance of 28 issues in Danish
politics in 1987/1988, and of 16 issues in 1994; 4 important goals in
1994; 12 public budget alternatives in 1990, and 20 such alternatives
in 1994. With the number of respondents and often large number of
alternatives, the prediction from simulations based on impartial-culture
assumption would be a large proportion of cycles. For example, with
an impartial culture, many voters and 20 alternatives, the prediction
is a 68 percent incidence of cycles (Gehrlein 1997, 179). There was,
however, only one trivial cycle in this entire set. For the 20 budget alter-
natives in 1994 there was a Condorcet ranking for the rst 14 alterna-
tives, and then a cycle over what would be alternatives 15 through 18,
and then transitive order again over alternatives 19 or 20. Otherwise, no
cycles.
Finally, Regenwetter and Grofman (1998) used a probabilistic method
to estimate rankings from data in seven out of ten real approval-vote
elections, and found only a small chance of a cycle in one of the elec-
tions. We should remember the bias against publishing negative nd-
ings. No doubt many people over the last thirty years have thought that
it would be intellectually and professionally satisfying to demonstrate a
real instance of cycling, yet the positive claims of cycling we have from
the entire political universe can be counted on ones ngers and toes
(and, as we shall see, even these claims collapse under scrutiny). Where
is the pervasive political disequilibrium? Shepsle and Bonchek (1997,
5051) ask:
Is [group intransitivity] merely an arcane logical possibility, a trick foisted on
the unknowing student by professors, philosophers, and textbook writers? Or is
it a profound discovery, the stuff from which important insights about political
philosophy and social life are made. In our opinion, the answer lies much closer
to the latter.
In my opinion, the answer is closer to the former rather than to the latter.
Riker (1965, 52) warned that the Arrow theorem is no mere
mathematical trick without practical signicance and set out to show
that the paradox of voting does occur and is of tremendous importance
in committees and legislatures. At rst, he estimated that whenever on
important issues a proposal loses to the status quo, half the time it is
due to a manipulated cycle. Later, Riker estimated that cycles afict
10 percent of legislative votes (Bell 1974, 308). Still later, Riker (1982,
122123) acknowledged there is a tendency to similarity among prefer-
ence orders that reduces the likelihood of cycles: there is good reason to
believe that debate and discussion do lead to . . . fundamental similarities
of judgment. However, the possibility of manipulation increases the like-
lihood of cycles, he argues. The net result: there are few cycles on unim-
portant issues, but more cycles the more important the issue is to the
manipulators. More precisely:
quite a wide variety of rather mild agreement about the issue dimension guaran-
tees a Condorcet winner . . . not all voters need display the agreement to obtain the
guarantee . . . agreement about dimensions renders uncontrived cyclical outcomes
quite rare . . . intransitivities only occasionally render decision by majoritarian
decisions meaningless . . . at least when the subjects for political decision are not
politically important. When, on the other hand, subjects are politically important
enough to justify the energy and expense of contriving cycles, Arrows result is
of great practical signicance . . . on the very most important subjects, cycles may
render social outcomes meaningless. (Riker 1982, 128)
I will argue in Chapters 9 through 17 that Riker is unable to demonstrate
the existence of a cycle on any issue, minor or major.
Riker later (1990b, 179) granted that, Poole and Rosenthal . . . have
shown with large empirical studies of congressional voting that, in the
absence of grand manipulation, a considerable part of political life is uni-
dimensional. Poole and Rosenthal (1997) dedicate their book to Riker,
their teacher, friend, and colleague. Their spatial analysis of all roll-call
votes in the 1st through 100th Congresses of the United States shows
that about 85 percent of all votes can be accounted for in two dimen-
sions. Moreover, Except for two periods of American history, when race
was prominent on the agenda, whenever voting could be captured by
the spatial model, a one-dimensional model does all the work (227).
The rst and overwhelmingly important dimension is what we popularly
understand as the standard leftright dimension. The second dimension
explains only about 2 percent of the 85 percent captured. The second
dimension varies from Congress to Congress, and varies from public
works to currency to tariffs and other issues; but was most salient as slav-
ery in the period before the Civil War (the 37th Congress in 1850 most
poorly t the spatial model) and as race relations in the civil rights era of
the 1950s and 60s. Testing for third, fourth, and greater dimensions on
the whole does not explain meaningfully more than the two-dimensional
model. Since the mid-1970s, the Congress has become increasingly and
is now almost wholly unidimensional (Poole and Rosenthal 1999). A
unidimensional issue space implies no cycles; and the mostly unidimen-
sional issue space discovered by Poole and Rosenthal implies very few
cycles.
Originally, I suspected that this unidimensionality was somehow a
product of the American two-party system and thus not evidence for
a strong tendency to unidimensionality in politics. The PooleRosenthal
(1999) methods, however, have recently been applied to votes in the
European Parliament (19891997), the British Parliament (1841), the
French National Assembly (19511956), the Czech Parliament (1993
1997), the Polish Parliament (1995), and the United Nations General
Assembly (19461996). The percentage of votes correctly classied by
a single dimension of analysis ranges from 85.9 percent in the UN
(19541969) to 94.2 percent in the Czech Parliament. The Czech Par-
liament is a multiparty system, and the United Nations comprises the
diverse interests of six billion people. It is possible that the apparent
unidimensionality is an artifact of the PooleRosenthal methodology.
Budge (1993) and coworkers examined all party manifestoes or plat-
forms from 1945 to 1981 in 23 democratic countries and applied factor
analysis. They found that one dimension, the standard leftright dimen-
sion, best explained the data. After reporting the ndings of Poole and
Rosenthal and of Budge, Riker (1993, 4) acknowledged that issue spaces
tend to be one-dimensional over time. He responded that second dimen-
sions would be of relevance, presumably with respect to manipulation, in
the short run.
On his own terms, Rikers claim of meaninglessness now stands only
on incidents of grand manipulation. Even if, as Riker maintains, all cycli-
cal manipulations are difcult to detect, he must be able to demonstrate
some instances from the rich universe of politics, especially since cycles
are supposed to be associated with the most important issues on which
we would have the most information. Otherwise, his claim would have
nothing but the glory and the shame of an untestable empirical claim. If
there are only a handful of such incidents, then the meaninglessness claim
fails, and it utterly fails if the handful do not withstand scrutiny. Later
we shall investigate in detail the topics of strategic voting and agenda
manipulation; for now accept that an instance of legislation defeated by
a killer amendment implies the presence of natural cycles or of the ma-
nipulatively contrived cycles that Riker stresses. With respect to strate-
gic voting, Poole and Rosenthal (1997, 147) found very few bother-
some needles in our haystack of 37,000 roll calls. The three instances of
successful killer amendments identied in the political science literature
are the three recited by Riker in Liberalism against Populism, known as
the Wilmot Proviso, the DepewSutherland amendment, and the Powell
amendment, and each does have to do with race, often the second dimen-
sion in American politics, according to Poole and Rosenthal (1997, 162).
With almost fty years of controversy, and strong professional incentives
for unveiling grand manipulations, these are the three that we have from
the universe of American Congressional roll-call votes to support the
proposition that democracy is meaningless. In later chapters, I will show
that Rikers accounts of these three events are mistaken. The jewel in the
crown of Rikers examples of grand manipulation (not related to Congres-
sional roll calls) is his famous allegation of a cycle in the 1860 American
presidential race that resulted in the arbitrary election of Lincoln and con-
sequently the Civil War. As promised, I will later show that his account
of that election is clearly mistaken and that there was no such cycle.
In a review article on the subject, Enelow (1997) writes that, cycling
and majority rule is one of the most heavily researched areas of public
choice (149), yet, citing only Riker, he acknowledges that basically,
the empirical literature testing the theory we have described consists of
a small set of examples (160). If I succeed in showing that Rikers and
others examples are mistaken, then the cycling hypothesis must die from
lack of evidence. Further, it would be shown that the Arrow theorem
cannot be interpreted to conclude that democracy is meaningless.
Justifying the theorems conditions
The denition of a social-welfare function requires that both individual
and social preferences be orderings complete and transitive ordinal
rankings. Condition I goes beyond requiring individual orderings, and in
addition requires that all voting rules proceed by pairwise comparison.
These issues will be discussed in depth in Chapter 6.
Condition P (if all voters prefer Metallica, then society prefers
Metallica) is not as innocent as it rst looks. Thats because the Arrow
theoremis not an engineering guidebook, but rather it is a logical exercise:
it need not be that any of the conditions are actually violated, rather the
transgression occurs if one of them could be violated. Condition P rules
out such social-choice rules as: do whatever the Bible says to do. It may
be that all voters want society to do what the Bible says to do, but the
point is, if the citizens were to change their view about that, the social-
choice rule would still dictate doing what the Bible says to do. Consider,
among other concepts, Rousseaus concept of the general will, inalien-
able rights to life and liberty, the US Constitution, or Rawlsian justice.
Each violates Condition P, to the extent that each is not based on aggre-
gation of individual orderings into a social ordering. It could be that the
unanimous will of all (aggregated votes) is identical to the general will
(the true or right decision), but it could be that everyone is mistakenly
opposed to the general will; and because it could happen, the general will
as a decision rule is in violation of P. It may that everyone agrees that I
have an inalienable right to life and liberty, but it is logically possible that
everyone, including me, would vote to deny that right. Thus, the right is
in violation of P.
Condition D no one person determines the social ordering regardless
of the preferences of other individuals is of direct normative relevance
to democratic theory. The condition is quite thin, however. Other than
Conditions P and D the Arrow theorem is noncommittal about democ-
racy, as the theoremdeclares impossible both democratic voting rules that
give each citizen one vote and undemocratic voting rules that give more
weight to some classes or exclude other classes of persons or establish any
dictatorship of two. Notice also that in the absence of some independent
justication, Arrows nondictatorship condition seems to bring a worry
relating to the interpersonal comparison of welfare into the scheme, con-
trary to its logical positivist foundations. If there is no way of comparing
the welfare of one person to another, then why should we object if one
person gets to decide everything? It could be, and many dictators act as
if it were so, that the satisfactions of the dictator are worth 1,000 times
everyone elses put together.
That leaves Condition U, unrestricted domain, which says that the
domain of the social-choice function includes all logically possible indi-
vidual orderings of the alternative social states. When social preferences
cycle, there is no social choice, according to the denitions. Condition U
contributes to the impossibility result because only cyclical proles of
individual preferences yield cyclical social orderings. If the domain of in-
dividual preferences were limited so as to exclude cyclical proles, then
there would be no impossibility result. Two justications typically are
offered for Condition U. First, that Arrows theorem seeks generality,
and U is the most general assumption. But that is not so. Condition
U requires individuals to have complete and transitive orderings. Obvi-
ously they dont in real life, and a more general condition would be to
permit individuals to have some incomplete and some intransitive pref-
erences. The problem with the more general condition is that it would
detract from the rhetorical force of the Arrow result, since all that would
be shown fromthe more general condition is that incomplete and intransi-
tive individual preferences may aggregate into incomplete and intransitive
social preferences.
Second, it is said that to exclude some preference orderings would
be tyrannical. Condition U has been interpreted to mean that citizens
should be free to prefer any policy option at all and to rank any options
in any way they want, meaning that no institution should have power to
declare certain choices out of bounds at the start (Pildes and Anderson
1990, 2,132). Any domain restriction that would mitigate the Arrowtheo-
rems impossibility result would have harsh consequences for those whose
preference rankings fall outside the domain restriction. We would have
to restrict entry into the community to those having preference order-
ings that do make collective choices possible or if it is already too late
they must somehow be isolated and excluded from the community, or
an impossibility result can again emerge (Mueller 1989, 392393). In
the next chapter, I shall argue that if the question is about the public
interest, then individual preferences are naturally sufciently similar to
one anothers to avoid cycling most of the time. And if the question is
xed-sum redistribution, then destructive self-seeking preferences should
be excluded from public consideration (as taught in kindergarten).
5 Is democracy meaningless? Arrows
condition of unrestricted domain
Introduction
Given theoretically predicted instability, why the empirically observed
stability? There are several types of answers: that stability is an illusion be-
cause we are unable to detect the manipulation that occurs (Riker 1982);
that stability is due to institutional devices (e.g. Shepsle and Weingast
1984); that such institutional devices are themselves pervasively unstable
(Rikers rejoinder 1980a); that stability is due to similarity in preference
rankings among the population, and to preferences for fair distribution;
or is due to some other defect in the models. The counterempirical out-
come of Arrows theorem puts us on notice that one or another of its
conditions must be misconceived.
I begin with similarity among individuals preference rankings, a chal-
lenge to the realism of Arrows condition of unrestricted domain. The
theorems impossibility result is a logical possibility but not an empiri-
cal probability, I shall argue. One kind of similarity in preference rank-
ings is disastrous though: if majority-rule voters divide up a xed good,
and if each is motivated solely by self-interest and not at all by fair-
ness, then we are guaranteed instability. Contrary to theoretical predic-
tion, however, democratic legislators are typically universalistic rather
than factional on distributional questions. This may be due to uncer-
tainty about the future, or a direct concern for fairness, or independently
motivated reciprocity, or public deliberation, or due to some combina-
tion of these devices. Empirical work shows that citizens vote judgments
of general welfare rather than personal welfare. An individual voter al-
most never affects the outcome of an election, hence she is free to ex-
press her disinterested sentiments rather than her interests, and this may
explain the empirical nding. Further, there is empirical evidence that
Americans overestimate the prevalence of self-interest. If there is not suf-
cient fairness in the population to tame Condorcet-voting instability,
I argue, then there are other acceptable voting rules that avoid cycles
altogether.
95
Table 5.1. Probability of Condorcet winner, impartial culture, strong
preference order
Voters Alternatives 3 5 7 9 49 Limit
3 0.944 0.931 0.925 0.922 0.914 0.912
5 0.840 0.800 0.785 0.776 0.752 0.749
7 0.761 0.704 0.682 0.670 0.638 0.631
25 0.475 0.379 0.345 0.327 0.281 0.270
Limit 0.000 0.000 0.000 0.000 0.000 0.000
Note: Table adapted from data in Gehrlein (1983, 174). It is important to distinguish two
forms of the Condorcet Paradox: rst, when there is no Condorcet winner; second, when
there are more than three alternatives, and there is a Condorcet winner, but there is a
majority cycle on alternatives below the Condorcet winner. The Table, and the discussion,
unless otherwise indicated, refer to the rst form of the paradox.
Simulations of homogeneity
In the last chapter, we saw that voting cycles are rare in measures of real-
world preferences. Why might that be so? Simulations suggest that it is
because of natural similarities among individuals preference rankings.
Unrestricted domain is any combination of individual preference orders.
For lack of any more salient and tractable assumption, calculations of
the probability of cycling therefore tend to assume an impartial culture,
just as dened in the discussion of Nurmi (1992) above: each of the
A! orders of alternatives is equally likely. Under the impartial-culture
assumption, the probability of cycling increases slowly as the number of
voters increase, and increases more quickly as the number of alternatives
increase, as seen in Table 5.1.
Simulations assume strong preference orders (no indifference among
alternatives) for computational convenience. If weak preference orders
(voters may be indifferent among some candidates) and impartial cul-
ture are assumed, then the probability of a Condorcet winner changes. If
we permit ties for top place, in the three-voter, three-alternative case the
probability of Condorcet winners increases from 0.944 to 0.995 for weak
preference orders (Van Deemen 1999). If we require a unique majority
choice, then a Condorcet winner among weak preference orders is less
likely than among strong preference orders with smaller numbers of vot-
ers, and a Condorcet winner among weak preference orders is more likely
than among strong preference orders with larger numbers of voters (Jones
et al. 1995). These gyrations are simply due to how the impartial-culture
assumption behaves in the contrasting simulations; what is important, as
Unrestricted domain 97
we shall now see, is that the more similar are preference orders the less
likely are cycles.
Regenwetter, Adams, and Grofman (2000) remark that the empirical
literature nds evidence of virtually no cycles, and that the few cycles ob-
served tend to be over low-ranked alternatives not sharply discriminated
among by voters. Tsetlin, Regenwetter, and Grofman (2001) demon-
strate that for three voters the impartial-culture assumption maximizes
the probability of cycles in samples drawn from an innite population of
individuals with transitive weak preferences. In this setting, any depar-
ture from the impartial-culture assumption, realistic or not, reduces the
probability of cycling. Recall that Tangian (2000) showed that as we de-
part from the impartial-culture assumption, and as the number of voters
increases, then the outcomes selected by summed cardinal utilities, the
Condorcet method, and the Borda method tend to converge. Since there
are no cycles with summed utilities or with the Borda count, that implies
that with a large number of voters cycles are extremely unlikely, according
to Tangian. Similarly, List and Goodin (2001, Appendix 3) show that,
given suitably systematic, however, slight, deviations from the impartial
culture assumption, the probability of a cycle converges either to 0 (more
typically), or to 1 (less typically) as the number of voters increases. The
impartial culture assumption is often used in textbook demonstrations of
the incidence of cycling (e.g., Shepsle and Bonchek 1997, 53, who on the
assumption of impartial culture also claim that the likelihood of cycles
goes up as the number of voters increases, contrary to Tangian), but it is
the assumption most likely to yield cycles.
There are several ways of conceptualizing the similarity of preference
orders. I will begin with the probabilistic. Kuga and Nagatani (1974) de-
velop an antagonism index, a ratio of the number of antagonistic choice
pairs (e.g., x > y and y > x) to the number of pairs of persons in the
population, normalized by the number of persons in the population. Un-
der the impartial-culture assumption the antagonism index would equal
one. When that antagonism index is
2
3
or less, there is no voting paradox
(by analysis), between
2
3
and 1 the probability of paradox increases as
antagonism increases (by simulation). Williamson and Sargents (1967,
811) probabilistic method concludes, where unimodal preferences
exist . . . and a large vote can be expected, the probability of intransitivity
seems to vanish. Bergs Polya-type urn model claims to subsume most
remaining prior measures of homogeneity of preferences, and conrms
prior studies: greater preference similarity among voters, or stronger
mutual inuence among voters, leads to a lesser chance for the para-
dox of occurring (Berg 1985, 379). The model contains a contagion or
Table 5.2. Probability of Condorcet winner, increasing homogeneity, three
alternatives
Homogeneity
parameter #
of voters 0
1
4
1
2
1 2 3 10
3 0.9444 0.9643 0.9750 0.9815 0.9952
5 0.9306 0.9524 0.9665 0.9602 0.9938
7 0.9250 0.9470 0.9626 0.9690
Large 0.91226 0.91919 0.92578 0.93750 0.95493
homogeneity parameter, and as that parameter increases, the probability
of a Condorcet winner increases. Suppose an urn with chips of six dif-
ferent kinds corresponding to the six strong preference orders resulting
from three alternatives. When the homogeneity parameter is zero, vot-
ers are independent of one another, and chips are drawn from the urn
with no replacement (known in the literature as the impartial-culture as-
sumption). When the homogeneity parameter is one, we replace a chip
drawn with one of the same kind (known elsewhere in this literature as
the anonymous-culture assumption). When the homogeneity parame-
ter is two, we replace a chip drawn with two of the same kind, and so
on, reecting the extent to which voters inuence one another. Gehrlein
(1995) further developed estimations of Condorcet efciency and social
homogeneity under the urn model.
In Chapter 3 on the irrationalists claim that democracy is arbitrary,
I argued that natural similarity among individual preference orders re-
duces divergence of outcomes among alternative reasonable voting rules.
I argued that people have similar preference orders because they live in
the same world and they have similar interests in that world; for exam-
ple, most prefer prosperity to torture of kittens to suicidal nuclear war. If
again we move from logical possibility to empirical probability, then the
irrationalists claim that democracy is meaningless also seems to miss the
target. Even if there is no similarity among individuals preference rank-
ings, the probability of a cycle is very low. With a more natural similarity
among preference rankings, such as what we found in our empirical ex-
amination in the last chapter, the probability of a cycle shrinks further. It
is nature that provides the cycle-shrinking domain restriction, not human
institutions. To relax the domain restriction does not require a crushing
uniformity in the population, nor the horrifying picture of preference
police hounding out those with topsy-turvy rankings. Should everyone
have topsy-turvy rankings, then of course there is no mere aggregation
Table 5.3. Egomaniacal redistributional instability
Voter: A B C
Coalition:
1. ABC 200 200 200
2. AB 300 300 0
3. BC 0 301 299
4. AC 299 0 301
5. AB 300 300 0
function that can bring order out of disorder. In that case, however, the
fault lies in the distribution of preference rankings, not in the voting rule.
Condition U was primarily intended as a matter of methodological con-
venience, according to Brennan and Hamlin (2000, 106107). Whatever
the intent, the condition is normative in effect, they continue. The re-
quirement that all and only individual orderings be admitted for social
choice is not neutral, but rather is a particular view on the question. The
condition bites even if everyone were to agree to exclude some kinds of
preferences such as the unfair, the malevolent, or even the evil from
consideration. Such an agreement would violate the condition; the con-
dition is imposed even if all citizens were to reject it. The imposition of
a rule dictating the consideration of evil preferences is not normatively
compelling. Condition U imposes a substantive and questionable ethi-
cal view on members of society, while at the same time failing to reect
the patterns of common interest that might be expected to arise within
genuine societies (107).
Egomaniacal redistributional instability
I said that natural resemblance of preferences from individual to individ-
ual almost eliminates cycling, but in another way similarity of preferences
might force Condorcet instability. Asked what the disagreement was be-
tween himself and King Francis I of France, Charles V replied: My
cousin Francis and I are in perfect accord he wants Milan, and so do
I.
1
If Condorcet majority-rule voters are engaged in division of a xed
good, lets say pie, then instability is inevitable if each insists only on get-
ting the most pie for himself (Sen 1995, 10). In Table 5.3 we begin in stage
one with an equal distribution of 200 each among majority-rule voters A,
B and C. But next in stage two the coalition AB can form, offering each
of A and B more than they would receive under equal distribution. But
in stage three, the coalition BC could offer each of B and C more than
they would receive in stage two. But in stage four, the coalition AC could
offer each of A and C more than they would receive in stage three. But in
stage ve, the coalition AB could return and offer each of A and B more
than they would receive in stage four. And so on.
Say that there are three piemen: Jaime, Lyman, and Simon. Any
majority-rule coalition of two or three is unstable. Condorcet voting can-
not transform the egomaniacal more-for-me preferences of these three
into a fair outcome. Suppose that each of the three piemen has a prefer-
ence in the rst rank of all the pie for himself, and then in the second and
third ranks has random preferences over all the pie going to one of the
two others, such that there is a cyclical prole over the more-for-me
alternatives, and in fourth rank has preferences for fair division of the pie.
There is a cycle over the more-for-me alternatives, each of which is pre-
ferred over fair division. (In contrast to cycling with Condorcet voting,
Borda voting would declare a tie among all-for-Jaime, all-for-Lyman, and
all-for-Simon.) If, however, the rst rank remains the same, but fair di-
vision is moved up to everyones second rank, then fair division becomes
the Condorcet winner, and the more-for-me alternatives cycle in the re-
maining ranks of the social ordering (Borda would rank fair division rst,
and a tie among the all-for-me alternatives). Thus, even if everyones rst
preference is more-for-me, if their second preference is fair division, then
fair division prevails as the social decision. Impartial alternatives drive
out partial alternatives.
To illustrate more explicitly, assume there are ve individuals, labeled 1
through 5. There are six alternatives: v gives the whole world to individual
1, w gives the world to individual 2, x gives the world to individual 3, y
gives the world to individual 4, z gives the world to individual 5, and a
divides the world equally among the ve individuals.
The collective ranking is a > (x > w > v > z > y > x). Fair division
prevails (and would prevail even if fairness were anywhere in the upper
half of the rankings). If preferences for fair division were originally ranked
lowest, and if there were costs to cycling, then perhaps it would develop
that preferences for fair division would come to rank higher. Since there
are many redistributional issues in politics, some commentators are con-
dent that there must be much majority-rule instability, even if it is difcult
to detect empirically. But I think they are making an empirical assump-
tion, controversial and not widely warranted, that political participants
do not at all value fair outcomes. As I have told the story, each voter could
still in a childish way want all the pie for himself, but just so long as each
ranks fair division better than random distributions of the pie then there
is a tendency for the social choice to be fair division. They would also
Table 5.4. Impartiality displaces partiality
Individual: 1 2 3 4 5
Ranking:
First v w x y z
Second a a a a a
Third z v w x y
Fourth y z v w x
Fifth x y z v w
Sixth w x y z v
Pairwise comparison matrix
a v w x y z (BC)
a 4 4 4 4 4 (20)
v 1 1 2 3 4 (11)
w 1 4 1 2 3 (11)
x 1 3 4 1 2 (11)
y 1 2 3 4 1 (11)
z 1 1 2 3 4 (11)
have to rank universalistic outcomes over majority-faction outcomes, but
that would be likely if majority coalitions were unstable and themselves
random in incidence.
In earlier work based on cooperative game theory, Riker (1963) for-
mulated his size principle: that in the majority-rule division of a xed
good, only minimum-winning coalitions would form. If the voting rule
is half the voters plus one, then all distributional measures would be
won only by that margin. The idea is that adding more members to the
coalition would dilute the amount of the good going to each member
of the coalition. If there are 100 voters, dividing the pie among 51 of
them gives each of the 51 more than she would get than dividing the
pie among 52 or among all 100. A problem developed with the theory:
actual legislatures infrequently passed measures by the minimum num-
ber of votes, rather they frequently passed measures by large, almost
unanimous majorities, especially on distributional issues. Empirical stud-
ies concluded that distributive policy making was characteristically asso-
ciated with the development of an extremely large and at times unanimous
coalition that included virtually everyone with a stake in the outcome
(Collie 1988, 430).
2
Congressional porkbarrel spending on projects at the Congressional-
District level resembles sucha constant-sumgame. If there is cycling, then
spending by district should equalize over time, as losers can always create
new winning coalitions; if there is no cycling, then spending by district
should remain stable from year to year. Stratmann (1997) proposes a
measure for this, applies it to federal-funding data, and infers an absence
of cycles. If his nding is correct, it may be due to any number of reasons,
for example institutional constraints in Congress, or perhaps there are
substantive reasons for some districts consistently to obtain more funding
than others, and it is not necessarily due to a similarity among preference
orders. Additionally, Poole and Rosenthals (1997; 1999) data suggest
that although the one dimension of importance in Congress is the struggle
between the poor and the rich, still, redistributive struggles rarely take
a multidimensional form. Brennan and Lomasky (1993, 4445, 83) say
that the instability story strains credulity: one never observes the degree
of volatility in spending or taxation that the cycling model requires.
The empirical tendency was termed the norm of universalism.
Weingast (1979) sought to explain universalism within the inherited co-
operative game-theoretic framework. The set of minimum-winning coali-
tions dominates all coalitions not inthe set. But, Weingast argued, benet-
seeking legislators would be uncertain, as the vote approached, of which
of the many possible minimum-winning coalitions would form and win.
Because of this uncertainty, the chance of ending up in a losing coali-
tion, each legislator beforehand would prefer the universalistic coalition
to any other coalition. The Weingast model assumed that the benets
of projects funded exceeded their costs, which, because of the view that
many government-funded projects are economically unjustiable, occa-
sioned further models by him and others that would explain inefcient
projects.
3
Ferejohn, Fiorina, and McKelvey (1987) modeled the same prob-
lems with noncooperative game theory. In the one-shot game, the cheap-
est projects would be funded, and by a minimum-winning coalition. If
that stage game were indenitely repeated then, as with social-dilemma
models of indenite repetition, any Pareto-optimal outcome is sustained
in equilibrium, from minimum-winning to universalistic. Baron and
Ferejohn (1989) modeled a sequential multilateral bargaining game. Now
the pie rots as the legislators debate how to divide it up. Under a closed
rule, a minimumcoalition wins. Under an open rule, the outcome can be
universalistic, but only if the legislature is quite small in number of voters
(or in blocs) and if the pie rots quickly. Glazer and McMillan (1982)
propose a nifty noncooperative model that yields a universalistic equilib-
rium when legislators want to spend their limited time, in my metaphor,
baking new pies rather than ghting over the division of old ones. Grose-
close and Snyder (1996) propose another model showing that superma-
jority coalitions can be cheaper than minimal-winning coalitions.
There is an alternative hypothesis to explain the observed norm of
universalism, however, which enjoys the advantage of consistency with
a controlled experiment on the topic, and more importantly consis-
tency with legislative discourse on redistributive taxing and spending
tasks.
4
The alternative hypothesis is that a sufcient number of voters are
directly concerned to attain a fair distribution. Miller and Oppenheimer
(1982) conducted experimental tests of Weingasts model of universal-
ism. In controlled experiments, ve-person committees were formed, and
individuals preferences were induced by cash reward. Committees op-
erated by majority rule, any three of the ve members determined the
outcome, after 15 minutes available for discussion; and subjects were
prohibited from redistributing rewards outside the terms of the experi-
ment. Coalitions of three or more could choose among one alternative
that provided an equal distribution to all ve committee members and
other alternatives that would provide various payoffs to the members. In
the rst experimental condition, the value to each member of equal dis-
tribution was $12.40, and, ` a la Weingast, the expected value of belonging
to any one of the decisive three-member coalitions was $8.40. If a three-
member coalition were actually to win on any alternative other than equal
distribution, each of its members would make more than $12.40. Sub-
jects rapidly agreed that equal distribution was the best choice. Typically
some members pointed out that three members could be made better off,
and typically some members responded that an alternative three-member
coalition could be formed leaving two of the three in the original coali-
tion worse off, and then members would settle for the safety of equal
distribution. Subjects were aware of the fragility of equal distribution,
and typically truncated discussion so as to discourage temptation. Four
of ve committees were unanimous for equal distribution, in the fth, the
vote for equal distribution was four against one economics student.
In the second condition, the value of equal distribution was made
$8.50, just above the expectedvalue of belonging to any one of the decisive
three-member coalitions. In four groups out of ve, members chose the
equal distribution. In the fth group, discussion secured unanimous con-
sent to explore coalitionformation, anda three-member coalitionformed.
Weingasts model seems to be supported. Further experiments were run,
however, with the universalistic alternative worth less than the expected
value of $8.40 for belonging to one of the decisive three-person coali-
tions. In the third condition, equal distribution was worth $6.72. Still,
in four out of ve committees the universalistic alternative was chosen.
In the fth committee, three members wanted universalism, and two in-
sisted on coalition formation. In response, the three members who origi-
nally proposed universalism voted for their own three-member coalition,
thereby punishing the two members for advocating an unfair outcome!
In a fourth condition, equal distribution was lowered to $4.20. Again,
four out of ve committees chose equal distribution. The fth commit-
tee devised an alternative universalism: they decided to draw lots among
all alternatives, with an expected value of $7.70 for each member. They
then voted unanimously for the randomly drawn outcome. The third and
fourth conditions do not support the Weingast model. In a fth condition,
equal distribution was worth only $2.10; nally, members agreed that the
equal distribution was not worthwhile. I have gone on at length because
I think it is important to remember the possibility that on distributional
questions many people are motivated directly by fairness.
More than a decade of human-subject experiments, say Fehr and
Fischbacher (2002), demonstrate robustly that many humans possess so-
cial preferences beyond material self-interest, and, they continue, some
economic models that neglect this fact fail in their predictions. Although
the models of economists permit complete heterogeneity of taste with re-
spect to material consumption, they insist that individuals are motivated
only by self-interest and dogmatically exclude heterogeneity in social pref-
erences. The experiments show, however, that roughly 40 to 50 percent of
subjects are motivated by reciprocal fairness. Moreover, the presence of a
critical mass of those motivated by reciprocal fairness changes the incen-
tives of the selsh types, resulting in constraints and interaction patterns
quite different from those that would result if the population were wholly
selsh. A reciprocal individual responds to kind action with kind action,
and to hostile action with hostile action. Such reciprocity is carried out
even if costly; it is not motivated by expectation of future material benet;
it differs from retaliatory behavior in repeated interactions. It also differs
fromaltruism, which is kindness inresponse to either kindness or hostility.
The primary experiment in this paradigm is the ultimatum game. A pair
of subjects are assigned the task of dividing a given sum of money, in
an unrepeated, one-shot game. The rst makes an offer to the second of
from 0 to 100 percent of the sum. The second either accepts or rejects
the offer. If the second accepts, then the second gets the share offered,
and the rst gets the remainder. If the second rejects, then both get noth-
ing. The rational-choice prediction is that the rst will offer the smallest
possible unit of money, say a penny, and the second will accept. The
experiments show, however, that proposals offering less than 20 percent
are rejected with 0.4 to 0.6 probability. Also, it appears that proposers
anticipate rejection of lower offers, which motivates them to make higher
offers. This is shown by comparison to dictator games, where the rst
mover simply decrees the division. The rational-choice prediction is that
rst movers would give nothing; in dictator experiments, however, some
give, but on average less than in the ultimatum game.
5
Henrich et al.
(2001) conducted ultimatum, dictator, and social dilemma games across
fteen nonindustrial cultures, and found that humans consistently devi-
ate from the model of textbook economics. Many appear to care about
fairness and reciprocity and reward cooperators and punish noncoopera-
tors, even when such actions are costly to the individual. Mean offer per
culture in ultimatum games ranges from 0.26 to 0.58.
The Los Angeles County Board of Supervisors, report Cox and Tutt
(1984), formally adopted a rule that nonmandated federal funds be di-
vided equally among the ve supervisorial districts again, the norm
of universalism. Observe that any three supervisors supercially have an
incentive to vote to depart from the norm, but that they do not. One
of the things I did before going to graduate school in political science
was to serve for three years as the directly hired policy aide (aka political
hack) to one of three elected commissioners of a county of about 250,000
people the retail politics of roads, sewers, zoning, garbage, and dogs.
The normof universalismwas common in allocative decisions, with argu-
ments around the edges. One day, however, there was a palace revolution.
We policy aides went into a meeting to discuss allocation of millions of
dollars of CETA make-work funds among the many applicants. The pre-
vious year the norm of universalism was informally followed. In the year
under discussion, a representative advisory committee had ranked pro-
posals. In the meeting, the aides of the other two commissioners junked
the advisory committee proposal, and devised a list of grantees which
excluded any applicant of interest to the third commissioner, my boss.
Moreover, the gang of two continued this exclusive collaboration on all
further issues. Consider that when the norm of universalism is a possi-
bility, everyone is strongly motivated to maintain comity. A permanent
loser, however, has no such motivation and, spared some of the burdens of
governance, has lots of spare time on his hands. There is more to politics
than votes on propositions. My bosss constituencies were extraordinarily
mobilized because of the unfairness of the CETA massacre, he gained
unexpected allies motivated solely by considerations of fairness because
of it, and he triumphed against a well-funded opponent in his next re-
election campaign. It also turned out that one of the gang of two became
subject to a recall election. He survived the recall, but was permanently
damaged politically, and later did not run for reelection. The other of
the gang of two was accused of defrauding the county on a land sale; the
criminal prosecution did not secure a conviction, but the political career
of a gubernatorial aspirant was ended, and later a civil action against him
to recover defrauded value did succeed. New commissioners after that
operated more by the norm of universalism. Perhaps this is an example
of reciprocity as a motivator of equal distribution.
In the famous prisoners dilemma or social dilemma used widely to
model many collective action problems, it is in each individuals interest
not to cooperate, which makes everyone worse off than they would be
if all did cooperate. In terms of material payoffs, the problem may ap-
pear to be a social dilemma, but if there is a critical mass of individuals
independently motivated by reciprocity to punish noncooperators, then
the dilemma is transformed into a coordination game. In a coordination
game there is more than one possible equilibrium, and the question is
whether everyone will coordinate on an equilibrium that makes all or
most better off or on an equilibrium that makes all or most worse off.
In this model, cooperation could fail in one of two ways. First, given a
critical mass of conditional cooperators, promoting cooperation involves
management of peoples beliefs: If people believe that others cooper-
ate to a large extent, cooperation will be higher compared to a situation
where they believe that others rarely cooperate (Fehr and Fischbacher
2002, C16). Second, even if there is a critical mass of conditional cooper-
ators, and people believe there is one, then it still may happen that people
have inherited a welfare-inferior coordination equilibrium that can only
be escaped by coordinated exit (see Mackie 1996). Notice that inherit-
ing a welfare-inferior equilibrium, say a tradition of political corruption,
might cause a false belief that there is not a sufcient mass of conditional
cooperators in the population, when in fact there are.
A nice feature of a model like this is that it might explain both the
presence and the absence of fair cooperation, for example, of democratic
success in Australia, and of democratic failure in Argentina: identical in-
stitutions containing otherwise identical individuals might vary in fair co-
operation just because people vary in their beliefs of what to expect from
one another. This is one reason why some people deplore the rhetoric
of the material-egoist version of rational choice theory: the more hege-
monic is the theory, the more self-fullling it is, by causing people to
believe that the proportion of unconditional and conditional cooperators
in the population is less than it actually is, thereby making people worse
off than they would have been in the absence of the theory. The correla-
tion between self-interest (personal nancial situation) and voting in US
presidential elections is usually 0.08, but after four years of self-interest
rhetoric in President Reagans rst term of ofce, and his call to Ask
yourself, are you better off today than you were four years ago?, the
correlation jumped to 0.36 (Miller and Ratner 1996, 31).
Traditional social choice theory is solely concerned with voting, and
excludes democratic discussion about the content of alternatives and rea-
sons for favoring one alternative over another. Some strains of deliberative
democratic theory seek to escape the social choice-problems of demo-
cratic voting by sole reliance on unanimous agreement attained by demo-
cratic discussion. In democratic institutions, I submit, voting and discus-
sion are complements, not substitutes. A voting rule such as Condorcets
is, or should be, adopted in the rst place due to considerations of fair-
ness. Majority rule would be a hideous institution if it were conned only
to the casting of votes on alternatives without content, or if it were car-
ried out by unfair people, or both. Sen (1982, 21) illustrates. In case A,
person 1 is very rich, and persons 2 and 3 are very poor; 2 and 3 vote
to take some from 1 and give to 2 and 3, reducing inequality. In case B,
person 1 is very poor, and persons 2 and 3 are very rich; again 2 and 3
vote to take some from 1 and give to 2 and 3, increasing inequality. In
the Arrovian noncomparabilist framework, Case A is identical to Case B.
Some people might judge Case A to be morally wrong, but most people
would judge Case B to be morally wrong.
Recall Rousseau (1968/1762, 72): There is often a great difference
between the will of all and the general will; the general will studies only
the general interest while the will of all studies private interest, and is
indeed no more than the sum of individual desires. If we could exclude
egomaniacal preferences from consideration, then cycling would be of
little worry. Is there some way to do this? Robert Goodin and Jon Elster
propose that public discussion helps:
The conceptual impossibility of expressing selsh arguments in debates about
the public good, and the psychological difculty of expressing other-regarding
preferences without ultimately acquiring them, jointly bring it about that public
discussion tends to promote the common good. (Elster 1986b)
The full argument is found in Goodins (1995/1986) landmark essay,
Laundering Preferences. Devices such as vested rights lter the out-
put of collective decisions. Another kind of device is to lter the input
to collective decision, via the exclusion of some individual preferences,
the laundering of them. Acceptable grounds for laundering preferences
include: for one individual, some preferences are more fully considered
than conicting others; some individuals agree amongst themselves to
reciprocal forbearances, such as conditional toleration of the others; indi-
viduals may explicitly have preferences over what kind of preferences they
should have, and the second-order preference should be honored over the
rst-order preferences; individuals may implicitly have such second-order
preferences; and nally, our very choice of aggregating preferences as
a way of making social decisions carries consequences for the kind of
preferences that we can count (140).
Goodin suggests further that an individual might possess multiple pref-
erence orderings, selsh or fair, material or moral, according to context,
and that, in the context of collective decision making, role rationality
dictates the expression of publicly oriented, ethical preferences and the
suppression of privately oriented, egoistic ones. Its pragmatically im-
possible to argue publicly that an alternative which affects others should
be adopted solely because it is good for the arguer. Further, that some-
one is voting in the rst place indicates an ethical context, because vot-
ing is instrumentally irrational according to rational choice theory (to
be discussed in the next section). In addition to these pragmatic mech-
anisms, some or many individuals may possess simply moral motiva-
tions to support impartial outcomes. Elsewhere, Goodin (1992) cautions
against expecting too much from this model of publicity and discursive
defensibility.
If a body of people is such that, on redistributive questions, most in-
dividuals and factions politically insist at all costs on more than their
fair share of the social ofces and product, then there will be instabil-
ity, probably trouble, maybe even violence. Formally, there would be
Condorcet-voting instability, but that is not the main problem. The aw
is in the egomaniacal preferences of the individual citizens, not in the
fact that those unfair preferences are aggregated by majority rule. For
example, Didymus Mutasa, organizational secretary of the ruling party
in Zimbabwe, ZANU-PF, in response to accusations that his government
was withholding food aid from famine areas which did not vote for it in
the election, is reported to have said that, We would be better off with
only six million people [out of a population of 12 million], with our own
people who support the liberation struggle. We dont want all these extra
people.
6
No aggregation procedure can turn unfair individuals into fair
ones.
7
For someone such as Mutasa, public discussion probably wouldnt
do much good either.
Voting and self-interest
The empirical evidence indicates that democratic voters are not mo-
tivated by material self-interest, or only minimally in limited circum-
stances. Kinder and Kiewiet (1979, 1981) were the rst to investigate
sociotropic voting. Generally, Americans vote for incumbent candi-
dates when the economy is good, and for opposition candidates when the
economy is bad. Most earlier work had been done on an aggregate basis,
and thus it was wrongly concluded that citizens vote their pocketbooks: I
vote against the incumbent when Im not doing well. Alternatively, cit-
izens could be sociotropically motivated: I vote against the incumbent
when the country is not doing well. By studying individual rather than
aggregate responses, Kinder and Kiewiet (1979, 524) showed that:
With respect to economic issues, voters appear to choose between congressional
candidates sociotropically. Voters are not egocentric in any narrow sense
they do not vote their own pocketbooks. Rather, their preferences follow a more
collective reckoning.
Many similar empirical investigations were undertaken in reaction to the
rise of the materially self-interested rational actor in the social sciences.
Lewin (1991, 45) reviews the literature on self-interest or public interest
in western politics, and nds that the the results point overwhelmingly
against the self-interest hypothesis. He suggests that it is odd that the
self-interest hypothesis is so tenaciously adhered to, when the great bulk
of evidence goes against it.
8
An even more detailed and wide-ranging
review by Sears and Funk (1991, 47; abridged in Mansbridges excel-
lent 1990 volume on self-interest), including survey data on two dozen
diverse issues, nds that personal self-interest generally has not been of
major importance in explaining the general publics social and political
attitudes. There are a few exceptions, and self-interest is most potent
when the issue provides large, clear, certain, and salient costs or benets
(63). Citrin and Green (1990, 16) conduct a similar survey, and conclude
that the literature reviewed appears to be devastating for the claim that
self-interest, dened narrowly as the pursuit of immediate material bene-
ts, is the central motive underlying American public opinion. A test of
whether individuals are concerned about whether or not their own group
is doing well found another sociotropic effect, that citizens are concerned
about whether economic gains and losses are distributed equitably among
groups (Mutz and Mondak 1997).
There are a great many confusions andequivocations about self-interest
in the rational-actor tradition. Current economic theory has it that indi-
viduals maximize their utility, but the only content that claim has is that
an individual chooses the most highly ranked alternative feasible fromher
consistent ranking of alternatives. Those aims could be egoistic or altruis-
tic, smart or stupid, or apparently random, just as long as their ordering is
consistent. The traditional utilitarian idea that individuals somehowmax-
imize an underlying satisfaction is wholly abandoned. Individual aims, if
they are concerned about the well-being of others, are nevertheless con-
tained within the individuals preference ranking; thus, it is sometimes
said that the theory assumes that individuals are self-interested. But they
are self-interested merely by denition: any alternative that St. Francis
chose, no matter howseless, was self-interested in this tautologous sense
(Riker 1990b). It is more useful to say that current economic theory as-
sumes that people are purposive rather than that they are self-interested.
This is sometimes called the thin theory of rationality (Green and Shapiro
1994, 1719).
A thick theory of rationality goes further, and assumes an underlying
dimension to those choices, often of objective content, in contrast to the
purely subjective value assumed by the thin theory. The paradigm is the
prot-maximizing rm in the competitive market. Economic competi-
tion forces the successful rm to pursue solely prot and to do so with
perfect information. The rationality of the rm is measured by the ob-
jective criterion of prot, and there is a background theory as to why
on average rms should possess this thick rationality. Confusion arises
when we move from rms to ordinary individuals, however. It could be
assumed as a matter of modeling and measuring convenience that on av-
erage ordinary individuals engaged in economic activity are concerned to
maximize material-egoistic wealth. This is a limited theoretical simpli-
cation, however, not a descriptive claim, and should never be permitted to
mislead, especially outside the economic sphere. Casual observation sug-
gests a tendency to selsh wealth maximization. But only a tendency, and
primarily in the economic sphere: many individuals are also motivated by
many other aims, including relations with family and friends, morality,
justice, honor, and self-expression. Unlike rms in perfect competition,
there is no background theory as to why ordinary individuals should exclu-
sively maximize material-egoistic wealth; individuals who do not, are not
irrational. Those who do the C. Montgomery Burns of the world seem
to be more pathological outliers than healthy specimens of humanity, or
so the wisdom of the ages suggests. I hasten to add that both self-interest
and the pursuit of material wealth, in proper proportion and place, are
consistent with if not necessary to the moral life.
The reign of self-interest in popular economics is all the more puzzling
in that the intellectual founders of the economic tradition themselves op-
posed Mandevilles simplistic reduction of all human motivation to self-
interest. Holmess (1990) marvelous essay on the history of self-interest
teaches us that David Hume, Adam Smith, and their fellow travelers be-
lieved both that an ideal of commercial self-interest was better for the
common good than the ideals it displaced of a passion for military glory
and the religious dogma of original sin, and that calculated self-interest
was one human motivation among many. Calculation is contrasted with
destructive impulse and folly, and self-interest is contrasted with varieties
of benevolence on the one side and varieties of malevolence on the other.
Some disinterested motivations are good, and, Holmes teaches, plenty
of disinterested motivations are evil: these days, one thinks of the seless
cruelty of suicide bombers.
Olsons logic of collective action assigns a certain preference ranking to
each individual, together these preferences amount to a social dilemma,
and thus nonrivalrous and nonexcludable public goods tragically are un-
dersupplied, the story goes. Rational individuals, Olson claims, will not
voluntarily contribute to public goods, instead they will free-ride on the
efforts of others, a nding that irks many. Olson and his followers believe
that, since the contribution of any one agent does not affect provision
of the public good, each agent would rst prefer to avoid her own costly
contribution while all others contribute, second prefer mutual coopera-
tion, thirdprefer mutual defection, andlast prefer being the only personto
contribute. The social outcome is mutual defection, even though mutual
cooperation is better for everyone. Pellikaan and van der Veen (2002)
show that Olsons logic wrongly assumes that the objective interest of
rms in the competitive market should apply as well to ordinary individ-
uals. The thick rationality of competitive rms, for whom that ranking is
appropriate, is confounded with the thin rationality of ordinary individ-
uals. They further show that on certain environmental questions Dutch
citizens do not have the preference order leading to a social dilemma, in
fact, many are shown to be unconditional cooperators, willing to con-
tribute even if others do not, just as some moralities would recommend.
We have seen assumptions of self-interest and perfect information jus-
tied within a model of rms in a competitive market unjustiably trans-
planted into the ordinary individual. Next, one variety of public choice
theory goes one step worse: it transplants these same assumptions into the
voter, for whom they are not justied at all. There is no background the-
ory as to why the ordinary individual should be a maximizer of material
self-interest, but there is or should be a background theory of why the
ordinary voter should not be a maximizer of material self-interest. If there
are more than a few voters in an election, then its unlikely that my
vote would make any difference to the outcome, and the more voters
there are the more unlikely it is that I would be decisive. Since voting takes
effort, no instrumentally rational person would bother, and according to
rational choice theory almost no one would vote, which we shall term the
paradox of participation. Goodin and Roberts (1975) may have been the
rst to point out that this unleashes the ethical voter. As in the argument
about public deliberation, suppose that the individual possesses multiple
preference orderings, selsh or fair, material or moral. Since the indi-
viduals vote lacks decisiveness, egoistic preferences are made futile, and
thereby expression of ethical preferences comes to the fore. This is fully
developed by Brennan and Lomasky (1993). In the market, the agent is
decisive; if she chooses A over B, she gets A and forgoes B. When the
voter chooses A over B, however, the outcome is not a function of how
she votes, but of how everyone votes. For each voter, electoral outcome
is detached from electoral choice. The market responds to interests, but
democratic voting invites the expression of disinterested ethical and ide-
ological principles. Thus, the empirical nding of sociotropic voting may
have a theoretical foundation in the account of voting as expressive rather
than instrumental action. As we have learned fromHolmes, however, and
as Goodin, Roberts, Brennan, and Lomasky conrm, the disinterested
need not be the good: justice could be unleashed, but so could jingoism.
All the more important then is sound public deliberation over the content
of political alternatives.
Sens (1995, 15) presidential address to the American Economic As-
sociation deplores a trend among some economists to reduce all socially
motivated action to some kind of cunning attempt to maximize purely pri-
vate gains, and wonders whether the trend is more prevalent in America
than in Europe, citing Tocqueville (1969/1851, Book II, Chapter VIII):
The Americans . . . enjoy explaining almost every act of their lives on the principle
of self-interest properly understood. It gives them pleasure to point out how
enlightened self-love continually leads them to help one another and disposes
them freely to give part of their time and wealth for the good of the state. I
think that in this they often do themselves less than justice, for sometimes in the
United States, as elsewhere, one sees people carried away by the disinterested,
spontaneous impulses natural to man. But the Americans are hardly prepared to
admit that they do give way to emotions of this sort. They prefer to give credit to
their philosophy rather than to themselves.
Miller and Ratner (1996; 1998) update Tocquevilles observation. They
undertook a series of social-psychology experiments which measured
both the actual extent that self-interest guided action, and peoples esti-
mates of the extent of self-interest motivation in the situation. For exam-
ple, they asked subjects whether or not they would be willing to donate
blood for free and whether or not they would be willing to donate blood
for $15, and they also asked subjects to estimate what percentage of their
fellow students would donate blood for free and for $15. Material payoff
only mildly affected willingness to give: 73 percent said they would do-
nate for a price, 63 percent would donate for free. But the same subjects
underestimated by half the number of students willing to donate blood
for free: they thought that 61 percent of the population would donate
blood for a price, and that only 32 percent would donate blood for free.
Similar experiments studied effect of gender on support for abortion cov-
erage, effect of racial status on concern for minority needs, and effect of
payment on willingness to participate in future experiments, and found
the same overestimation of self-interested motivation.
We are enthralled by a myth of self-interest, according to Miller and
Ratner. Public discourse equating rationality with self-interest normalizes
actions consistent with self-interest and pathologizes actions inconsistent
with it. People experience discomfort when they take action inconsistent
with self-interest: one study found that peoples decision about whether
or not to pursue an injury claim was based much more strongly on expec-
tations of nancial gain than on expectations of fair treatment, but that
peoples satisfactions with the claiming process were based far more on
fair treatment than on degree of nancial gain. Thus, people ignore their
own sentiments in favor of the norm that material self-interest should
predominate, and they are led to willingly adopt decision strategies they
know will not maximize their satisfaction (1996, 35). People fear social
isolation when they take actions inconsistent with self-interest. People
justify their actions in terms of self-interest: they explain their charitable
activity, for example, in instrumental terms It gave me something to
do, I liked the other volunteers, It got me out of the house (38).
Miller and Ratner do not indicate awareness of Holmess message that
there are worse things than self-interest. Otherwise, their ndings sug-
gest that a dominant ideology of self-interest does obscure recognition
of motivations of benevolence and fairness in the population. It does
not take saintliness to overcome egomaniacal redistributional instability.
Merely a secondary preference for fair distribution, should one be unable
to grab everything for oneself, would do. Further, among a small num-
ber of people, independently motivated reciprocity would maintain equal
distribution. Empirically, legislators are universalistic on redistributional
questions. Among a larger number of people, the inefcacy of voting
could liberate sentiments of fairness. Empirically, voters are sociotropic.
If there is a worry that there is not enough fairness in the population
to avoid egoistic redistributional instability under the Condorcet voting
rule, then there are good voting rules that avoid cycles.
Cyclebusters
We have seen that cycles are rare. This section will offer two further
claims. First, most cycles are trivial. Second, if there are cycles that are
not trivial, then there exist accurate and fair voting rules that eliminate
such cycles.
The closer alternatives are to one another, the less structure there is be-
tween individuals preference orders. Permit me to explain. Suppose that
the three alternatives under consideration are education appropriations
of a, $110,000, f, $10,000 or g, $0. There are 99 people in our assembly,
and all 99 rank a > f > g, $110,000 > $10,000 > $0; there is just gen-
eral agreement that an education expenditure closer to $110,000 than to
$10,000 is good for everyone in the community. Individual preference or-
derings are perfectly correlated with one another. As we introduce more
ne-grained alternatives the less structure there is between individual
preference orders, but also the less does the absence of structure mat-
ter. Suppose now that the three alternatives under consideration are a,
$110,000, c, $100,000, and e, $90,000. There is unanimity that educa-
tional expenditure should be within the range of $110,000 and $90,000,
but there is not unanimity on a gure within that range: 33 voters rank
c > e > a, 33 rank a > c > e and 33 rank e > c > a. On supercial in-
spection the three different rankings dont seemto resemble one another,
but they are somewhat correlated and can be arrayed along one dimen-
sion; they enjoy the property of single-peakedness which will be eluci-
dated below. The social ordering by Condorcet order and Borda count is
c > e > a, which discloses the resemblance between the individual or-
derings. Now suppose that the three alternatives are b, $100,100, c,
$100,000, and d, $99,900. Everyone thinks that $100,000 is about
right, especially when compared to $10,000, but peoples prefer-
ences over these last three alternatives $100 apart from one an-
other are not well-considered, there is no structure between indi-
vidual preference orders, which happen to be perfectly uncorrelated
for the sake of my illustration: 33 voters rank b > c > d, 33 rank
d > b > c, and 33 rank c > d > b. The collective ordering is a cycle:
b > c > d > b, which discloses the absence of structure between the
individual preference orders.
The closer alternatives are to one another, the less likely is it that indi-
vidual preference orders resemble one another, the more likely is it that a
social cycle results, but the more trivial is the resulting cycle. A cycle over
closely adjacent alternatives is not as troubling as a cycle over remotely
distant alternatives. I want to distinguish between benign cycles and
malign cycles. This distinction would be meaningless in the framework of
ordinal and noncomparable utility, but will be of interest to those whose
common sense is intact. For the noncomparabilist, that a senator prefers
highway expenditures of $15,000,000,000.00 to $15,000,000,000.01 to
$14,999,999,999.99 carries just as much information as that a senator
prefers a policy of preemptive nuclear war to isolationism to charitable
internationalism the only information the Arrovian permits is the order
of preferences. For the Arrovian, a benign cycle among three alternatives
a penny apart from each other on $15 billion in highway expenditures is
indistinguishable from a malign cycle among world war, neutrality, and
charity. I claim, however, that a cycle among three alternatives a penny
apart is of no importance compared to a cycle among war, neutrality, and
charity. If an arbitrary method of choice were applied to select one from
the three highway expenditure alternatives, or even an unfair method such
as one agent with unfair control over order of consideration of alterna-
tives, it would be of no practical concern. If, however, a random method,
or worse, an unfair method were used to select among the alternatives of
war, neutrality, and charity, we would rightly be horried. My hypothesis
is that what empirical cycles there are will much more likely be of the
benign rather than the malign variety.
Nowwe are in a position to propose a possible explanation for the sharp
disjuncture identied by Gehrlein: almost no cycles with eight or fewer
candidates, more cycles with nine or more candidates. Formal theory
does not predict this disjuncture; the likelihood of a cycle should increase
smoothly as we move fromthree to many candidates. Gehrleins data were
mostly from single-transferable vote elections where voters must rank all
candidates, regardless of whether their preferences are fuzzy, indifferent,
or incomplete. There is a rule of seven, plus or minus two, for identi-
cation of items by humans in short-term memory and in unidimen-
sional absolute judgment tasks (e.g., sounds varying only in loudness,
lights varying only in brightness, forms varying only in size; G. Miller
1994/1956; Baddeley 1994; Shiffrin and Nosofsky 1994). Beyond the
capacity limit of about seven, humans commit errors, even on the sim-
plest of tasks, and this is a most robust effect. I suggest that the voters
in Gehrleins study committed such errors when judging nine or more
candidates, and that these individual errors, structureless ranking over
a portion of the range of candidates, sometimes aggregated to cycles.
Humans bypass the information-capacity limit by means of three meth-
ods. First, relative judgment rather than absolute judgment. Rather than
judging brightness on an absolute ten-point scale, instead make relative
judgments, A is brighter than B, B is brighter than C, and so on; then
we can string together the pairwise comparisons. Second, increase the
number of dimensions along which stimuli can differ. We can identify
hundreds of faces, but faces are sorted over multiple dimensions. Third,
make a sequence of several absolute judgments in a row. With respect
to memory, for example, a ten-digit phone number is impractical to re-
member, but chunked into three portions, an area code of three digits, a
prex of three digits, and a sufx of four digits, it is practical to remem-
ber. A fourth method not mentioned by the psychologists cited is the use
of external technologies.
This is uncanny. As a speculative hypothesis, it seems that the capac-
ity limit and the methods of bypassing the limit are both reected by
voting procedures, as are the intuitions that modelers of voting strive
to formalize. Arrow is intuitively certain that compounding of pairwise
comparisons should be the way to rank many alternatives even when
that drives him to dictatorship. Maybe the root of his intuition is a tacit
understanding that absolute judgments cannot distinguish more than
about seven items along a single dimension. Why did the spatial model
of voting, based, so far as I can tell, on no evidence, arise and have such
widespread appeal? Maybe the root of the intuition that underwrites the
spatial model is again a tacit understanding that humans can better sort
multiple alternatives by mapping them to several dimensions (even if
those dimensions are highly correlated with one another). We also seem
to like handling multiple alternatives by chunking them into sequential
procedures: the amendment procedure in legislatures, plurality runoff
elections, federalism. The formal models of voting may be appealing be-
cause they appeal tacitly to the capacity-limit problem, but otherwise be
misleading because they do not expressly recognize and incorporate the
problem.
Here is a new concept. To return to the example of the educa-
tion appropriations, if the alternatives available for consideration were
those between $100,100 and $0, then the collective ordering would be
(b > c > d > b) > e > f > g, a so-called top cycle simply because the cycle
is at the top of the social ranking. The collective choice rejects $0 and
$10,000, but cycles among the closely adjacent alternatives $100,100,
$100,000, $99,900. Does this matter? It violates Condition O, which
requires an ordering over all alternatives. It would matter perhaps if it
resulted in a cyclic deadlock and no action (effectively choosing the last-
ranked $0 as the assemblys votes cycled among the top three alternatives
until the end of eternity), but that could always be remedied by an ar-
bitrary method, and in most instances by a nonarbitrary method, as we
shall see. The arbitrary method would be to dene cycles as ties. Then
the collective ordering would be b c d > e > f > g. Further, if a
single outcome rather than merely formal decisiveness were required,
then we could add the renement of some kind of fair, perhaps random,
tie-breaking institution, and one of b, c, or d would be chosen. This
counting cycles as ties (without the tie-breaking renement) is known
as Schwartzs method, or the method of transitive closure (Sen 1982,
162163, 180183). As a matter of logical possibility, it violates one or
another of Arrows conditions. If the transitive closure is taken on the
set of all possible alternatives X, then the relation satises contraction
consistency, also called Sens alpha (to be explained in Chapter 6 ), and
thus Condition O, but may violate Condition I. If the transitive closure
is taken on the subset S of considered alternatives, then Condition I is
satised, but Condition O may be violated. The Borda count behaves
similarly.
Suppose that on another issue our assembly had noncyclical prefer-
ences at the top of the social ranking, but cyclical preferences further
down the list, for example: u > v > (w > x > y > w) > z. Does this
matter? If the task is to select the best alternative for urgent action, then
it does not. Again, Condition O is violated. Perhaps it would matter if al-
ternatives u and v somehow became unavailable, but then we could turn
to one of our cyclebusting methods. Also, if for some reason our purpose
is to generate a complete ranking we can use a cyclebusting method, for
example, transitive closure: u > v > w x y > z.
Should cycles be a real problem anywhere, there are accurate and fair
voting rules that resolve most of them, as we shall now see. I want to
distinguish between balanced cycles and unbalanced cycles. The canonical
example of cycling has three voters each with exactly the right three rank-
ings that taken together will yield a cycle (A > B > C, C > A > B, and
B >C >A together; or C >B >A, A >C >B, B >A >C together). For
that prole, or any other prole where three numerically equal groups of
voters inhabit each of the sets of three cyclical rankings, which we shall call
a balanced cycle, the Borda count and YoungKemeny rule, correctly in
my view, declare a tie among the alternatives. The mathematician Donald
Saari, a formidable advocate of the Borda count, reports that a class of
fresh-minded fourth-graders patiently explained to him that the notori-
ous three-voter cycle was nothing but a tie (Saari 1995a, 5051). Now
consider the four-voter two-alternative majority-rule case where the Con-
dorcet paradox does not apply. Sometimes, perhaps even frequently for
the sake of argument, the four-voter group will have a tie vote on two
alternatives. Does the possibility of a tie thereby render all their choices
meaningless? What if they have some neutral way of breaking ties tak-
ing turns, ipping a coin, precedent, or ongoing discussions of fairness?
It may be discomting for there to be a tie among two or three alter-
natives, but at least remaining alternatives have been eliminated from
consideration. Once one understands that balanced cycles are merely the
three-or-more-alternative analog to the two-alternative tie, much of the
spookiness of cycling is exorcised. Most empirical cycles would be both
benign and nearly balanced, in my view, and there are remedies for malign
and unbalanced cycles.
Here is an abstract example of an unbalanced cycle (adapted fromDahl
1956, 42). One individual ranks x > y > z, 49 rank z > x > y, and 49
rank y > z > x, and this is a cyclical prole of preferences, x > y > z > x.
Applying the method of transitive closure to this unbalanced prole would
yield x y z. There are 98 out of 99 voters, however, who favor z
Table 5.5. An unbalanced cycle
x y z BC
x 50 1 (51)
y 49 50 (99)
z 98 49 (147)
over x, and can we really say that the social relation between z and x
should be one of indifference? If the one individual who ranks x rst were
an evil agenda-controller among foolishly sincere voters, and she pitted
alternative y against alternative z, then rst y would beat z and next y
would lose to x. Alternative x would be the winner even though 98 of
the voters favor z over x. The Borda count and the YoungKemeny rule,
however, properly identify z as the winner. For someone who believes in
the Arrow theorem, it would make no sense to say that z is better than
x. For someone without such beliefs, it seems unfair for z to lose. The
method of transitive closure would be unsatisfying because the cycle is
unbalanced, z is better than, not as good as, x.
There are several remedies for this unbalanced cycle. If there is pairwise
majority voting, if there is an open agenda such that any alternative can
be proposed in any order, and if the members of the assembly prefer
any choice over eternal cycling, then z will be selected. How? Before
proceeding, an aside. The normative argument against cycling appeals
to the intuition that it is bad for a voting procedure to cycle eternally
among alternatives. That judgment is external to the model, however.
Inside the formal setupno one loses anything fromcycling. The normative
argument illicitly appeals to our real-world experience that indecisiveness
is wasteful we could be doing something better with our time than
cycling, and we also need a decision now. In order to bring the setup in
line with our intuitions, we have to amend our description of preference
orders. Call the costly absence of a decision arising from an eternal cycle
alternative o. Now we have one individual who ranks x > y > z > o, 49
who rank z > x > y > o, and 49 who rank y > z > x > o. To continue, we
will examine three agendas, one beginning with y against z, one beginning
with x against y, and one beginning with x against z. I will show that z is
the nal choice on each agenda.
If the rst-stage contest is between y and z then y wins by one vote
and if the voting is sincere the cycle will go on forever, the last-ranked
choice of all of our voters. If, however, at least one of the 49 who rank
y >z >x >o votes strategically against her rst-ranked y in the rst stage
in order to obtain her second-ranked z in the second stage and overall
avoid her last-ranked o, then z would win and the cycle would be broken.
The incentive of the 49 who rank z rst and z > x > y > o overall is to
vote sincerely in order to obtain z. If in a third stage x is moved against
z, then z beats x 98 to 1. Thus, z is stable. Next, if the rst-stage contest
is between x and y, then x wins the rst stage and loses in the second
stage to z, if voters are sincere in the rst and second stages, and the
third-stage vote is between y and z, and z beats y and the cycle is broken
just as before. Would the 49 who rank z > x > y > o vote strategically
in the rst stage for y over x? No, that would gain them y, their last-
ranked alternative and lose them z, their rst-ranked alternative. Would
the one who ranked x > y > z > o vote strategically in the rst-stage for y
over x? By doing so, this lone individual might gain y, her second-ranked
alternative and avoid x, her last-ranked alternative. Therefore, she might
contemplate voting strategically. If she did so, however, the next choice
would again be between y and z, but we have already seen that z would
win that contest among strategic voters for whom a cycle is costly, so she
would refrain from voting strategically. Finally, if the rst-stage contest
is between x and z, then z beats x by 98 to 1, and again z goes on to beat
y among our strategic voters for whom any choice among x, y, and z is
more important than no choice o. The most strongly favored alternative,
z, is picked from the cycle by strategic voting. The same analysis holds if
this were a barely unbalanced cycle: one who ranks x > y > z > o, two
who rank z > x > y > o, and two who rank y > z > x > o.
The balanced cycle is similar but not identical. Assume three voters
each with orders x > y > z > o, z > x > y > o, and y > z > x > o.
Say that the rst-stage vote is between alternatives y and z. The voter
with ranking z > x > y> o could end the cycle by voting for her third-
ranked alternative y in order to avoid her fourth-ranked alternative o.
Why should she though? She would rather let y beat z and go up against
x. Then the voter with ranking x > y > z > o can end the cycle by
voting for her third-ranked alternative. Why should voter x > y > z > o
be the one to make the sacrice, however? And so on. The three face a
situation resembling an assurance game. The three need to devise and
justify in advance a fair tie-breaking institution, not any more challenging,
practically or normatively, than to devise and to justify such an institution
for conventional two-alternative ties.
Informal recognition of cardinal but uncompared utilities may help
decide ties or cycles arising from ordinal voting. This is the basis of
logrolling. Ozzie prefers A > B, and Sharon prefers B > A. We take
a majority vote over A and B only to disclose a tie. There are other things
in life, however, than A and B. Suppose that Ozzie wants A much more
Table 5.6. An almost balanced cycle
102 100 99
A C B
B A C
C B A
A B C (BC)
A 202 102 (304)
B 99 201 (300)
C 199 100 (299)
than some of those other things, but that Sharon wants some of those
other things more than she wants B. Ozzie could trade some of those
other things to Sharon in exchange for A, thereby breaking the tie.
Now (example adapted from Hovenkamp 1990a) suppose a balanced
cycle, full knowledge of one anothers utilities, and preferences and as-
sociated cardinal but noncomparable utilities as follows: Huey prefers
A(6) > B(5) > C(1), Dewey prefers B(60) > C(40) > A(20), and Louie
prefers C(600) > A(400) > B(200). Suppose the three nephews have
no tie-breaking procedures, and that the value to each of perpetual cy-
cling is zero. Then, Huey will vote for B, B will win, and Huey will get 5
rather than zero; and no cycle. Suppose the nephews do have a fair tie-
breaking procedure. Then the expected value to Huey of voting for A is
1
3
6 +
1
3
5 +
1
3
1 = 4, and the expected value to Huey of
voting for B is 1 5 = 5; hence, Huey votes for B, B wins; and
no cycle.
There are good voting rules that remedy unbalanced cycles. When the
number of voters inhabiting each of the three cyclical rankings are not
exactly equal to one another, when there is an unbalanced cycle, then
the Borda count almost always declares unique winners, not ties. Riker
(1982, 120) is not clear about this property of the Borda count. In dis-
cussing the frequency of cycles he says that the Borda count produces a
tie among cycled alternatives, but he does not note that the Borda count
produces a tie almost always in the rare instance of a perfectly balanced
cycle, an equal number of voters across the cyclical rankings. To fur-
ther illustrate, in Table 5.6 there are 301 voters and three alternatives
distributed across three cyclical rankings, followed by the pairwise com-
parison matrix. By pairwise comparison we have a cycle: A >B >C >A.
The Borda count, however, ranks the alternatives A > B > C, and the
Table 5.7. Another unbalanced cycle
23 17 2 10 8
A B B C C
B C A A B
C A C B A
A B C (BC)
A 33 25 (58)
B 27 42 (69)
C 35 18 (53)
Borda count tells us as well, if we are not dogmatic ordinalists, the so-
cial judgment is that there is not much difference between the alterna-
tives (if the number of voters had been in three equal groups of 100, the
Borda count would have yielded a tie). The YoungKemeny rule has us
break the cycle at the weakest link: the pairwise majority for C > A is
199 votes, less than either the 201 votes for B > C or the 202 votes for
A > B, and thus the YoungKemeny ranking is A > B > C just like the
Borda count.
In Table 5.7 there is a more ragged distribution of preferences
(borrowed from Young 1997), together with the associated pairwise-
comparison matrix, involving ve of the six possible rankings of three
alternatives, that more strongly distinguishes Borda and YoungKemeny
from Condorcet order.
Pairwise comparison delivers a cycle: A > B > C > A. Should we stop
there and either declare that democracy is meaningless or more humbly
declare a tie? Or should we use the ranking information we have to try
and nd a winner with a better method than pairwise comparison? The
Borda order is B > A > C and the YoungKemeny order is B > C > A
(recall that on assumptions Borda picks the most likely winner and
YoungKemeny the most likely ranking).
If cycles are indeed a threat to the very intelligibility of democracy,
then shouldnt we eagerly seek for voting rules that resolve cycles? If that
is our quest, then the accuracy and fairness properties of Borda count
and perhaps of YoungKemeny are distinguished, and thus we should
adopt one of those rules and thereby save the republic from Arrovian
dictatorship. If, in contrast, the prospect of cycles is rare due to mild
homogeneity among voters rankings of alternatives, and mild prefer-
ences for fairness in distribution, then we may relax, knowing that we
could adopt a more accurate voting rule if required by circumstances,
but otherwise condent that simpler voting rules may be adequate for
many purposes. Why would anyone reject the cycle-busting voting rules?
They violate Arrows independence condition, the subject of the next
chapter.
6 Is democracy meaningless? Arrows
condition of the independence
of irrelevant alternatives
1
Introduction
The interpretation that Arrows Condition I, independence of irrelevant
alternatives, prohibits the use of individuals intensities of preference in
the construction of social choices is not precise. Rather, it is the social-
welfare function (as dened in Chapter 4) which demands both individual
and social orderings, and thereby prohibits cardinal utility inputs. Condi-
tion I, as Arrow wrote it, redundantly requires individual orderings, but
goes further and demands that, even given the ordinal data fromindividual
orderings, the social choice over any two alternatives not be inuenced by
individuals preferences involving any third alternatives.
2
This is explicit
in Arrow (1963/1951, 59, emphasis added):
It is required that the social ordering be formed fromindividual orderings and that
the social decision between two alternatives be independent of the desires of in-
dividuals involving any alternatives other than the given two . . . These conditions
taken together serve to exclude interpersonal comparison of social utility either by
some form of direct measurement or by comparison with other alternative social
states.
Arrows is a strong independence condition. Slight weakenings of it allow
the Borda count or the YoungKemeny rule as possible social welfare
functions and further weakenings permit further voting procedures.
Barry and Hardin (1982, 217218) agree that Arrows IIA is a power-
ful condition. Part of its power is that one cannot easily intuit what it
means or why it matters . . . Perhaps because of its subtlety, condition I
is apparently the condition that is most readily taken for granted in the
proof of Arrows and related theorems. Its content is frequently misun-
derstood. Justications of the condition are typically thin and dogmatic,
often no more than an assertion that its appeal is intuitively obvious. My
search for justications of the condition found thicker arguments mostly
by Arrow(1963/1951; 1952; 1967; 1987; 1997), Sen (1970; 1982), some
by Riker (1961; 1965; 1982), and otherwise mostly repetition of points
made by Arrow without further justicatory development.
3
123
The chapter proceeds as follows. First, I explain that historically many
people have misunderstood the content of the independence condi-
tion (IIA(A)), believing it to be another condition (IIA(RM)), one that
does not contribute to the impossibility result. My main point is that
the independence condition can not be defended as intuitively obvious
if sophisticated commentators have trouble grasping even its content.
Second, I show by example that to violate either Arrows independence
condition (IIA(A)) or the contraction-consistency independence condi-
tion (IIA(RM)) can be substantively rational. I point out that Arrow un-
derstands the simplifying assumptions of his model not as ends in them-
selves but as means to empirical analysis; the assumptions themselves have
no descriptive or normative force, even more so when they are contrary
to observation and intuition. Third, Arrovians defend the IIA(A) as for-
bidding the inuence of irrelevant alternatives over the consideration of
relevant alternatives in social choice. I explain that the condition has
nothing to do with forbidding consideration of irrelevant alternatives in
the ordinary sense of the term, but rather requires that social choice be
carried out only by pairwise comparison, thereby delivering the impossi-
bility result. Fourth, I continue discussion of the Arrow theorem as mo-
tivated to avoid interpersonal comparisons of utility, and argue that the
IIA(A) is superuous to that goal. I submit that the Borda and Condorcet
methods are equivalent with respect to comparing or not comparing men-
tal states, and that both are purely ordinal methods. Fifth, I examine
Rikers claims that the IIA(A) serves to forbid undesirable utilitarian vot-
ing rules, probabilistic voting rules, and consideration of irrelevant alter-
natives. I reply that it has not been shown that utilitarian or probabilistic
voting rules are undesirable, but if they are then there are weaker indepen-
dence conditions that forbid those voting rules but still permit other rules
such as the Borda count. Also, I point out that the concern over irrele-
vant alternatives may be of importance in social-welfare applications, but
is of no importance in most voting applications. Sixth, I challenge the
Arrovian view that the Borda count is logically susceptible to manipu-
lation by addition and deletion of alternatives but that the Condorcet
method is not. I argue that both are logically susceptible to such ma-
nipulation, but that the susceptibility is of little practical importance. I
conclude that the frailties of the reasonable voting rules have been much
exaggerated, and that the time has come to move from a destructive
constitutional physics to a constructive constitutional engineering.
The wrong principle is defended
Arrow is an economist, and analogizes political choice to economic
choice. He explicitly considers voting and the market as special cases
Independence of irrelevant alternatives 125
of the more general category of social choice. Arrow borrows his basic
conception from simple consumer economics. He distinguishes all pos-
sible or conceivable alternatives from all feasible or available alternatives.
Suppose that the set Xcontains all possible or conceivable alternatives. S,
a nonempty subset of X, contains all feasible or available alternatives. To
anticipate, S contains the relevant alternatives and X S contains the
irrelevant alternatives.
On any given occasion, the chooser has available to him a subset S of all possible
alternatives [X ], and he is required to choose one out of this set [S]. The set
S is a generalization of the well-known opportunity curve; thus, in the theory of
consumers choice under perfect competition it would be the budget plane. It is
assumed further that the choice is made in this way: Before knowing the set S,
the chooser considers in turn all possible pairs of alternatives, say x and y, and
for each such pair he makes one and only one of three decisions: x is preferred to
y, x is indifferent to y, or y is preferred to x . . . Having this ordering of all possible
alternatives, the chooser is now confronted with a particular opportunity set S.
If there is one alternative in S which is preferred to all others in S, the chooser
selects that one alternative. (Arrow 1963/1951, 12)
This will be a crucial passage.
Many people, including myself for many years, have misunderstood
the content of Arrows independence condition. Indeed, about twenty
years after rst publication of his theorem it was recognized that Arrow
in 1951 at one point seemed narratively to justify a condition that was
not the same as the formally stated condition necessary for his proof.
4
Again on analogy to consumer choice, Arrow(1963/1951) argues that a
social choice from a set of alternatives S, just as for a single individual,
should be independent of alternatives outside S. He illustrates with a
criticism of the Borda count:
For example, suppose . . . an election system. . . whereby each individual lists all
the candidates in order of his preference and then, by a preassigned procedure,
the winning candidate is derived from these lists . . . Suppose that an election is
held, with a certain number of candidates in the eld. . . and then one of the
candidates dies. Surely the social choice should be made by taking each of the in-
dividuals preference lists, blotting out completely the dead candidates name, and
considering only the orderings of the remaining names in going through the pro-
cedure of determining the winner. That is, the choice to be made among the
set S of surviving candidates should be independent of the preferences of indi-
viduals for candidates not in S. To assume otherwise would be to make the result
of the election dependent on the obviously accidental circumstance of whether a
candidate died before or after the date of polling. Therefore, we may require of
our social welfare function that the choice made by society from a given environ-
ment depend only on the orderings of individuals among the alternatives in that
environment. (Arrow 1963/1951, 26)
Arrows surely is too quick. The election has already provided us with a
social ranking. Rather than deleting the dead candidates name fromeach
individuals preference list, why not instead delete the dead candidates
name from the social ranking? Arrows example is thus: two voters rank
the alternatives x > y > z > w, and one voter ranks them z > w > x > y.
The Borda ranking is x > z > y > w, Arrows focus is that x wins (the
Condorcet ranking is x > y > z > w). Now candidate y is deleted. By
the Condorcet method the ranking of remaining alternatives stays the
same, x > z > w, and x would still be the winner. If the Borda method
is reapplied to the remaining three candidates, however, then the Borda
ranking changes to (x z) > w, and both x and z are tied for the win.
Then, certainly, if y is deleted fromthe ranks of the candidates, the systemapplied
to the remaining candidates should yield the same result, especially since, in this
case, y is inferior to x according to the tastes of every individual; but, if y is in
fact deleted, the indicated electoral system would yield a tie between x and z.
Arrow (1963/1951, 27)
There are two problems. First, recall that in Arrows scheme the
chooser, before knowing the set S, forms a ranking over all possible
alternatives in X. Then the chooser encounters S, the set of all feasi-
ble alternatives. The chooser consults the list made from X and selects
the highest ranking alternative or alternatives in S as the choice. Arrow
analogizes social choice to the economic model of individual choice. If
the analogy holds, then social choice should form a ranking over all pos-
sible alternatives X, and consult from the list made from X in order to
decide the winner or winners in S. As this would apply to Arrows story
about the Borda count and the dead candidate, the Borda count would be
carried out on the set X of four candidates and the ranking x > z > y > w
determined. Then y dies. We do not reapply the Borda count to the three
remaining candidates in set S, rather we consult the ranking over four
candidates, x > z > y > w, and delete y from the social ranking, for
x > z > w, and x remains the winner. If we take the Borda count over
X, call that the global Borda rule, and if we take the Borda count over
some subset S, call that the local Borda rule. The general idea of Arrows
scheme suggests that we should apply the global Borda rule to X, but in
this instance Arrow says we should apply the local Borda rule to S. Why
the inconsistency?
The second and much bigger problem is that Arrows example does
not illustrate the IIA condition used in the theorem, but rather a different
condition confusingly labeled independence of irrelevant alternatives
by Radner and Marschak (1954), also called by Sen Condition , and
also called contraction consistency.
IIA(Arrow): Let R
1
, . . . , R
n
and R
1
, . . . , R
n
be two sets of individual orderings
and let C(S ) and C
(S ) be the corresponding social choice functions. If, for all

individuals i and all x and y in a given environment S, xR
i
y if and only if xR
i
y,
then C(S ) and C
(S ) are the same. (Arrow 1963/1951, 27)

IIA(RadnerMarschak): If x is an element of the choice set of S and belongs to
S
1
contained in S, then x is also an element of the choice set of S
1
, i.e., x C(S)
and x S
1
S together imply x C(S
1
).
(Ray 1973, 987, after Radner and Marschak 1954)
The two conditions are logically independent of one another (Ray
1973). In 1951, Arrow(1963/1951, 3233) criticized by example the idea
that summation of normalized von NeumannMorgenstern cardinal util-
ity functions might serve as a social-welfare function. His rst objection
was an example that violated what is here called IIA(RM) or contraction
consistency, and his second objection was an example that violated the
IIA(A). In this passage, Arrow distinguishes the two conditions, but it
seems that he does not in the passage justifying the IIA(A) by appeal to
the example of the dead candidate in the Borda count (1963/1951, 27).
Bordes and Tideman (1991) provide ingenious constructions that would
make Arrowboth accurate and consistent in his justication of the IIA(A),
and their interpretation is more than merely plausible. In the end, I do not
join in Bordes and Tidemans charitable reading, simply because Arrow
(1987, 195) later acknowledged a mixup: Nashs condition (adapted by
Radner and Marschak 1954), he said, refers to variations in the set
of opportunities, mine to variations in the preference orderings . . . The
two uses are easy to confuse (I did myself in Social Choice and Individual
Values at one point).
The later Arrow(1997) explains that the key conditions of the theorem
are Collective Rationality and IIA(A). For a given election the problems
posed by the conditions are hypothetical or counterfactual, he says. First,
what would have happened if we had added a candidate who wouldnt
have won or if we had subtracted a losing candidate? The addition or sub-
traction of such irrelevant candidates could have changed the outcome
of the election, and this would be a violation of the Collective Ratio-
nality Condition, according to Arrow. The Collective Rationality Condi-
tion requires that for any given set of [individual] orderings, the social
choice function is derivable from an [social] ordering (Arrow 1967, 70),
which in turn requires IIA(RM). Second, what would have happened if
voters preferences over noncandidates changed? Change of preference
over irrelevant candidates could have changed the outcome of the elec-
tion, and this would be a violation of the IIA(A) condition, according to
Arrow. What this argues is that the election rule was such that the result
Table 6.1. Violation of IIA(A)
Actual Counterfactual
# voters: 2 2 1 2 2 1
Rank:
1st A B C A C C
2nd B C A B B A
3rd C A B C A B
Condorcet A > B > C > A C > A > B
Borda B > A > C C > A > B
actually obtained might have been different although it should not have
been (1997, 5). Notice that Arrows conditions require that the social
choice not change if individual preferences or availability of alternatives
were to change. A more natural expectation would be that social choice
may or may not change if individual preferences or availability of alterna-
tives were to change. Consider simple majority rule over two alternatives
if one member of the majority changes her vote to the minority position,
that may or may not change the social choice. It seems to me that, in
response to changes in individual preferences or in availability of alter-
natives, any demand that the social choice should always change, or, the
demand by the IIA(A) or IIA(RM), that the social choice should never
change, carries the burden of justication.
I will now illustrate violation of IIA(A). Suppose that there are two
voters who rank A > B > C, two who rank B > C > A, and one who
ranks C > A > B. By the Condorcet method the collective outcome from
the prole is the cycle A > B > C > A and by the global Borda method
the collective outcome is B > A > C. In order to investigate violation
of the IIA(A) we are interested only in the rankings of two alternatives, say
alternatives A and B. Suppose now, counter to the rst supposition, that
the two who rank B > C > A instead rank C > B > A. By the Condorcet
method the collective outcome changes from a cycle to C > A > B,
and by the global Borda method from B > A > C to C > A > B. Focus
on the Borda count. Under the rst supposition, the Borda count yields
B > A. Voters preferences over the pair A and B do not change, but two
voters change from ranking C second to ranking C rst. Then, under the
second supposition, the Borda count yields A > B, a reversal from the
rst supposition. The IIA(A) is violated.
I will nowillustrate violation of IIA(RM), or of contraction consistency.
Suppose again that there are two voters who rank A > B > C, two who
rank B >C >A, and one who ranks C >A >B. Again, by the Condorcet
Table 6.2. Violation of IIA(RM)
Actual Counterfactual
# voters: 2 2 1 2 2 1
Rank:
1st A B C A B
2nd B C A B A
3rd C A B A B
Condorcet A > B > C > A A > B
Borda B > A > C A > B
method the collective outcome fromthe prole is the cycle A>B>C>A
and by the local Borda method the collective outcome is B > A > C. In
order to investigate violation of the IIA(RM) we are interested only in
the rankings of two alternatives, say alternatives A and B. Suppose now,
counter to the rst supposition, that instead of the three alternatives A,
B, and C, there are only two alternatives, A and B. By the Condorcet
method the collective outcome changes from A > B > C > A to A > B,
and by the global Borda method from B > A > C to A > B. Again,
focus on the Borda count. Under the rst supposition, the Borda count
yields B > A. Voters preferences over the pair A and B do not change,
but alternative C is removed from the menu. Thus, under the second
supposition, the Borda count yields A > B. B is an element of {A, B}
{A, B, C}, and B wins among {A, B, C}, but does not win between
{A, B} contraction consistency says that if B wins {A, B, C} then it
should win {A, B}. The IIA(RM) is violated.
In Arrows 1951 presentation (1963/1951), an axiom was presented
that the individual preference relation is complete, and another presented
that it is transitive, and it was further stated that a relation that satis-
ed those axioms was a weak ordering (12). Individual orderings and
social orderings must satisfy the two axioms (19). A social-welfare function
is a rule which for each set of individual orderings of alternatives states a
corresponding social ordering of alternatives (23). A social-choice function
C (S) is the set of alternatives x in S such that, for every y in S, x is
weakly preferred to y. A social welfare function determines a unique
social-choice function, and the social-choice function satises contraction
consistency (Bordes and Tideman 1991, 170). A voting rule that violates
contraction consistency violates the requirements of the Arrow theorem.
Hence, Arrows scheme does require that both IIA(RM) and IIA(A) be
satised. Sen (1993) proved that an impossibility result can be reached
even if the requirement for contraction consistency, IIA(RM), is dropped;
a condition similar to IIA(A) must be retained, however. IIA(A) is the
culprit in the impossibility result.
Arrows possibility theorem shows that social ordering, universal do-
main, Pareto principle, nondictatorship and IIA(A) are inconsistent. Ray
(1973) shows that social ordering, universal domain, Pareto principle,
nondictatorship and IIA(RM) are consistent. All social-welfare functions
satisfy IIA(RM), and at least one, the Borda method, satises the re-
maining conditions, according to Ray. All social-welfare functions satisfy
IIA(RM) if, as Arrow originally suggested, they are taken on X: if we ap-
ply our voting rule to all possible alternatives and consult that ranking to
order any subset S of X, then no contraction of the set of alternatives has
taken place. If it is insisted that the voting rule be applied to X, and then
be reapplied to S, then violations of contraction consistency are possible.
The global Borda count, for example, taken on X, satises IIA(RM),
contraction consistency, but may violate Arrows IIA(A). It satises con-
traction consistency because with the global Borda count there is no con-
traction fromXto S. The global Borda count may violate Arrows IIA(A)
as shown in the example in Table 6.1. The local Borda count, taken on
S, satises Arrows IIA(A), but may violate IIA(RM). To determine the
rankings of A and B, the local Borda count is applied only to A and B, C
does not enter the picture, thus the local Borda count outcome is the same
before and after the change by two voters fromB >C >A to C >B >A
and there is no violation of Arrows IIA(A). The local Borda count may
violate IIA(RM) as shown in the example in Table 6.2, where, contracting
from three alternatives to two alternatives, the outcome changes from
B to A.
There is a wide discourse, illustrated in my hall of quotations, declaring
that democracy is dubious because of the Arrow theorem. The problem
I am getting at here is that many commentators have made this claim
on the mistaken view that the Arrow theorem depends on the IIA(RM)
rather than on the IIA(A) (see, e.g., Riker 1961; 1965). Some of these
commentators spend time justifying IIA(RM), believing they are thereby
justifying the impossibility result. But IIA(RM) is not essential to the im-
possibility result, rather it is IIA(A) that is essential, and IIA(RM) does
not lead to impossibility. Would not such a discovery suggest a revision
in views? Instead, what we see in some commentators (Riker 1982) is not
a revision in view, but rather a new attempt to justify the newly under-
stood IIA(A). The conclusion is driving the premises, the tail is wagging
the dog.
I have made many astounding errors in earlier drafts of this volume
and I fear that, despite my best efforts, astounding mistakes and mis-
understandings remain. Scholars more talented and diligent than I are
bound to err, because the issues under consideration are difcult. My
purpose here is not to dwell on the errors of others. The purpose is to call
into doubt the common assertion that the IIA(A) condition should be
accepted just because it is intuitively obvious. How could the condition
be intuitively obvious if many sophisticated commentators are confused
even about its content, let alone its implications?
The independence conditions are not always
substantively rational
Barry and Hardin (1982, 266) say that Nobody has any immediate views
about the desirability of, say, the independence of irrelevant alternatives,
and we should refuse to be bullied by a priori arguments to the effect
that we would be irrational not to accept it. Arrovians proceed as if
IIA(RM) and IIA(A) were requirements of rationality. One or the other
of the conditions is presented as intuitively obvious, and sometimes an
example is presented that illustrates the absurdity of violating the con-
dition. It is hinted, but never spelled out, that to violate the condition
would be a logical contradiction. I grant that in a set of particular cir-
cumstances, it may well be that a violation of one of the conditions has
absurd consequences. In another set of particular circumstances, how-
ever, it may be acceptable, or perhaps even reason would demand, that
choice violate one of the conditions. A condition may be substantively
applicable in particular circumstances, or it may be useful as a simpli-
cation to assume that one or the other of the conditions applies, or to
assume that it usually applies unless there are special circumstances. But
neither of the conditions is necessary to practical reason in the sense
that it should apply to each and every possible choice regardless of the
particular circumstances. My goal is to show that the conditions are not
requirements of rationality, are not justied by naked appeal to intuition,
and to do so I present examples that illustrate the absurdity of obeying
the condition. If I present plausible counterexamples to the conditions,
then my argumentative goal is achieved.
It should be understood that all Arrow (1952, 49) intends by the word
rational is that an individual is rational if his preferences among can-
didates can be expressed by an ordering; similarly, collective decisions
are made rationally if they are determined by an ordering acceptable to
the entire society. Although the IIA(RM) is implicated in the ordering
assumptions of Arrows theorem, the independence aspect of the IIA(A)
is an additional requirement. To violate these narrow construals of ratio-
nality is not to violate the broader concept of rationality: of having beliefs
and desires, and carrying out plans and actions, for good reasons. Acting
for good reasons is prior to Arrows rational choice model, and in case of
conict it is the model that must go.
Here is anexample showing that violationof IIA(RM), contractioncon-
sistency, may be substantively rational. My example is inspired by Sen
(1993), who argues that the concept of internal consistency of choice,
exemplied by the IIA(RM), is essentially confused, and there is no
way of determining whether a choice function is consistent or not without
referring to something external to choice behavior (such as objectives,
values, or norms). To introduce and motivate his entire scheme, Arrow
(1963/1951, 2) supposes a society that must choose among disarmament,
cold war, or hot war. It is obvious to Arrow that rational behavior on the
part of the community would mean, in analogy to the economists un-
derstanding of individual choice, that the community orders the three
alternatives according to its collective preferences once for all, and then
chooses in any given case that alternative among those actually available
which stands highest on its list (1963/1951, 2). He then uses the
Condorcet paradox of voting to illustrate the possibility that the com-
munity might cycle among the three momentous alternatives. Arrows
scheme requires that choice among more than two alternatives be decom-
posed into pairwise comparisons. The example Arrow chooses, however,
illustrates the folly of insisting on pairwise comparisons over social states.
Suppose that there is an individual who prefers peace so long as it
does not require surrender to the enemy. If she were to face all three of
Arrows alternatives, then she would rank them Cold War > Hot War >
Disarmament. She least prefers Disarmament as that would amount to
surrender to the enemy, but also thinks Cold War is better than Hot War
because there are fewer casualties in Cold War. If she were to face a
choice between Cold War and Hot War, she would choose Cold War. If
she were to face a choice between Hot War and Disarmament, she would
choose Hot War. If she were to face a choice between Cold War and
Disarmament, however, she would choose Disarmament. Why? If Hot
War were off the menu of choice, if Hot War were no longer possible,
then the peace of Disarmament would be preferable to the tension of Cold
War and would not require surrender to the enemy. Her preferences over
a menu of all three alternatives is transitive: Cold War > Hot War >
Disarmament. Her preferences over menus of two alternatives differ,
however, and to chain them yields a cycle: she would prefer Cold War
to Hot War to Disarmament to Cold War. One may not agree with the
order of her rankings, but one would have to agree that her rankings are
substantively rational.
Arrows argument about social choice is by analogy to individual
choice. The Arrovian scheme seems to assume that for an individuals
choice to depend upon the menu of choices is irrational, but I have just
shown by example that it is possible for a rational persons preferences
over alternatives to vary by the menu of alternatives available. If it is
possibly rational for an individuals choices to vary by menu, then, by
analogy, it is possibly rational for a societys choices to vary by menu. A
collective might rationally rank A > B when those two are the only alter-
natives of interest, but rank B > A when alternative C is also available.
Thus to demonstrate formally that a voting rule ranks A > B when A
and B are under consideration, but B > A when A, B, and C are under
consideration is not in itself an objection to the voting rule. One would
have to go beyond formal rationality and further show that the reversal is
substantively irrational in the concrete instance. Suppose that some re-
versals are substantively rational and some substantively irrational. Then,
in the comparative evaluation of voting rules, we would want to know the
probable frequency of substantively irrational reversals for each rule as
conditions vary. Substantively rational reversals would be welcome.
Now for the more important chore, to show that violation of Arrows
IIA(A) may be substantively rational. Suppose that there is to be a recep-
tion and that the caterer will only provide one beverage, either beer or
coffee.
5
The overly rushed organizer of the reception copies a form from
last years event that asks people to rank beer, coffee, water, tea, milk, and
pop and e-mails it out. Attendance is by RSVP only. Five people fromthe
business school will come, and each of themranks beer >coffee >water >
tea > milk > pop. Four people from the law school will come, and each
of them ranks coffee > beer > water > tea > milk > pop. The organizer,
a political scientist indoctrinated in the Arrow theorem, and a believer
in the IIA(A) condition, tallies only preferences over beer and coffee,
the two relevant alternatives. Five want beer rather than coffee and four
want coffee rather than beer: beer wins by majority rule. Beer is the
Condorcet winner, the alternative that beats all others in pairwise
comparison: beer > coffee > water > tea > milk > pop. Beer is also
the Borda winner. It turns out though that the four people from the law
school cancel, and four people from the theology school will attend in-
stead. Their ranking is: coffee > water > tea > milk > pop > beer. The
ranking of the lawyers and the theologians is almost the same, except that
the lawyers rank beer second and the theologians rank beer last. The or-
ganizer looks only at the relevant alternatives, coffee and beer: by pairwise
comparison nothing has changed, beer is still the choice by majority rule.
The Condorcet order remains identical as well. The theologians come to
the reception and are furious. They are teetotalers and would rather have
anything but beer. The organizer loses his job, all because of his dogmatic
belief in the IIA(A) condition. If the political scientist had instead used
Table 6.3. Substantively rational to violate IIA(A)
5 4 4
Business School Law School Theology School
Beer Coffee Coffee
Coffee Beer Water
Water Water Tea
Tea Tea Milk
Milk Milk Pop
Pop Pop Beer
Note: Business School + Law School
Condorcet: Beer > Coffee > Water > Tea > Milk > Pop
Borda: Beer > Coffee > Water > Tea > Milk > Pop
Business School + Theology School
Condorcet: Beer > Coffee > Water > Tea > Milk > Pop
Borda: Coffee > Water > Beer > Tea > Milk > Pop
the Borda count, he would have noticed that the theologians ranked al-
cohol last, and would have provided the Borda-winning beverage, coffee.
The viewpoint of the Arrow theorem is that a social choice is derived
from a social ordering that is aggregated from individual orderings over
social states; only orderings can be observed, and therefore no mea-
surement of utility independent of these orderings has any signicance
(Arrow 1967, 7677), and the IIA(A) condition enforces the ban on
information other than individual orderings. What if there had been a
discussion about the beverage to be served at the reception? Assume that
the theologians have the same mere ordering of preferences. Before the
decision, they explain that they are teetotalers, and would rather have
anything but beer. Alternatively, suppose that they are prohibited from
drinking beer as a matter of their religion, and that they have a right to be
served a beverage at the reception that doesnt offend their beliefs. The
political scientist, our hapless believer in the IIA(A) condition, would
have to reject this information and enact the social choice of beer. Only
orderings can be observed, he would reply to the theologians, and the
only ordering that matters is between coffee and beer. That you would
rather have any beverage but beer is irrelevant.
I have put the argument in terms of the Borda count because my fo-
cus is on voting rules aggregating from ordinal preferences. In principle,
a utilitarian voting rule might be applied to cardinal preferences to ob-
tain a more accurate social ranking (maybe beer would be last), but the
qualitative features of the story would remain the same. Riker (Riker and
Ordeshook 1973, 110111) has an objection to the kind of story I have
told. Adapting his objection to t the current example, what if the or-
derings remained the same, but the participants had cardinal utilities as
follows. Each member of the business school would rank beer as 10, then
the remainder of alternatives as in the Borda count, 4, 3, 2, 1, 0 and
each member of the theology school as in the Borda count, 5, 4, 3, 2,
1, 0. With these cardinal utilities, although the Borda count would still
select coffee, the utilitarian choice would be beer (51) over coffee (40).
My rst response is that in the pure voting exercise all we have are or-
dinal data, and thus we cannot go beyond the Borda count. But if we
had sound cardinal data then why not use it? My second response is that
the Rikerian objection has changed the qualitative features of the story,
now, although the theologians rank beer last, the business schoolers are
crazy for beer. My third response is that if the Borda count generally is
imperfect in approximating cardinal utilities, should that be a goal, then
Condorcet is more imperfect at the task in my original example
Condorcet fails to detect that the theologians rank beer last. In the
Rikerian example, Condorcet picks beer, but, in utilitarian terms, for
the wrong reasons. My fourth response is that with increasing numbers
of voters, Borda tends to the utilitarian outcome (as does Condorcet). To
conclude, in practice, ardor or horror at beer is expressed in discussion,
people vote in a fair-minded way to take account of intense preferences
of others (and maybe entrench some relating to life and liberty as rights
claims), and in more competitive environments may engage in logrolling
(OK, no beer, you theologians, but we get to choose the main dish).
Arrow does not defend his assumptions on logical grounds. As for the
requirement of pairwise comparison for individual choice orderings,
There seems to be no logical necessity for this viewpoint; we could just as well
build up our economic theory on other assumptions as to the structure of choice
functions if the facts seemed to call for it.
18
18
Like Lange, the present author regards economics as an attempt to discover
uniformities in a certain part of reality and not as the drawing of logical conse-
quences froma certain set of assumptions regardless of their relevance to actuality.
Simplied theory-building is an absolute necessity for empirical analysis; but it
is a means, not an end. (Arrow 1963/1953, 21)
What are the facts of the matter?
I suggest that almost everyone would nd the Borda winner norma-
tively superior to the Condorcet winner, in other words, that they would
nd it intuitively obvious that the IIA(RM) and the IIA(A) should be
violated. Controlled experiments among a subject pool of Stanford stu-
dents support this hypothesis (Davies and Shah 2003). Subjects were
shown the preferences of various groups of voters and were asked, Which
alternative should be chosen for the group? In one study, for example,
the Condorcet ranking for one prole of two alternatives and another
prole of three alternatives was X > Y. The Borda winner in the three-
alternative prole, however, was Y, and 38 out of 39 subjects chose Y>X,
in violation of the IIA(RM). In another study, individual pairwise rank-
ings over Xand Ywere identical across two three-alternative proles, and
X > Y was the pairwise majority in each prole. The individual rankings
of Z were varied, however, such that X was the Borda winner in one
prole and Y was the Borda winner in the other prole, and 41 out of
46 subjects chose Y > X in that other prole, in violation of the IIA(A).
These studies yield large effects, in response to minimal manipulations,
are extremely signicant, and the basic result that large majorities choose
in violation of the independence conditions is robust to variations in
experimental design. Further replications across more conditions, re-
searchers, and culture groups are needed, but I expect that the ndings
will stand.
Here Arrow defends pairwise comparison as only a modeling conve-
nience in the case of individual choice. And, his justication for pairwise
comparison in the case of social choice is only by analogy to individual
choice. I have argued that it can be substantively rational to violate either
IIA(RM) or IIA(A). Further, almost all experimental subjects choose as if
it were normatively desirable to violate both conditions. There is no need
to make fetishes of the conditions, they are model-building means, not
normative ends. We should not let the simplications of models mislead
our larger judgments about, say, democracy.
The irrelevance justication is awed
Arrow (1952) says that the IIA(A) condition has always been implicitly
assumed in voting systems, but the claim is mistaken. He offers the ex-
ample of a community deciding between construction of a Stadium and
of a Museum. The community can afford one or the other, but not both,
and the community cannot afford a University at all. Arrow believes that
the choice between the Museum and the Stadium must be independent
of preferences of community members between the feasible Museumand
the infeasible University. The essential argument in favor of this princi-
ple is its direct appeal to intuition (51). It is true, as a matter of practice
rather than of logic, that infeasible or irrelevant alternatives are usually
not placed on ballots. But sometimes they are. It was notorious in the
2000 election that dead Democratic candidate Carnahan remained on
the ballot for US senator from Missouri, and won the election against
Republican Ashcroft. As expected, the Democratic governor appointed
Carnahans wife as his successor. In Great Britain, the Ofcial Monster
Raving Loony Party is an irrelevant alternative that regularly appears on
the ballot.
In his narrative justications Arrow uses an ordinary conception of
irrelevance: voting systems should choose over available alternatives but
not over conceivable alternatives, over feasible alternatives but not over
possible alternatives, over relevant alternatives but not over irrelevant
alternatives. I imagine that many people are against the mischief of ir-
relevance, and also against permitting preferences over irrelevant alter-
natives to inuence wrongly decisions concerning relevant alternatives.
The ordinary irrelevance that Arrovians deplore in narrative justica-
tions is not the irrelevance formally stated in the IIA(A) condition,
however.
What if the community could afford a Museum, a Stadium, or a Uni-
versity, but not any two or all three of these alternatives, and further
could not at all afford a Nuclear Missile? There would have to be a social
choice among the three feasible alternatives. Does the IIA(A) work to
permit consideration of the relevant Museum, Stadium, or College and
forbid consideration of the irrelevant Nuclear Missile? Not at all. Arrows
condition does not partition alternatives into the ordinarily relevant and
the ordinarily irrelevant. The condition applies to all candidates x and y,
lets say in a set S. If there are four relevant alternatives, a, b, c, and d
in S, then the choice among a, b, and c must be independent from pref-
erences involving d. As we consider the three alternatives a, b, and c, the
choice between a and b must also be independent from preferences in-
volving c or d. The choice between a and c must be independent from
preferences involving b or d, the choice between a and d must be inde-
pendent from preferences involving b or c, the choice between b and c
must be independent from preferences involving a or d, and the choice
between b and d must be independent from preferences involving a or c,
even though each of a, b, c, and d is ordinarily relevant. The IIA(A) condi-
tion always boils down to one that requires that the social choice between
any two alternatives x and y not be inuenced by individuals preferences
over any third alternative. The IIA(A) would better be named the pairwise
comparison condition, as it requires that choices among several alternatives
be carried out only with information about choices between pairs. Arrow
(1963/1951, 20) said as much in 1951:
One of the consequences of the assumption of rationality is that the choice to
be made from any set of alternatives can be determined by the choices made
between pairs of alternatives. Suppose, however, that the situation is such that
the chooser is never confronted with choices between pairs of alternatives; in-
stead, the environment may always involve many alternatives . . . we can say that
the choices made from actual environments can be explained as though they
were derived from choices between pairs of alternatives; and, at least concep-
tually, it makes sense to imagine the choices actually being made from pairs of
alternatives.
The IIA(A) means that if someone ranks x > y > z, we count that she
likes x > y, count that she likes y > z, and count that she likes x > z,
but we are not allowed to count that she likes x > y > z. Saari (2001b)
argues that the information lost due to this prohibition is what drives
the impossibility result. We can insist that voting procedures rely only on
pairwise comparisons and end up with Arrows dictatorship result and
with startling interpretations such as those in my hall of quotations, or we
can more sedately interpret the Arrow theorem to mean that procedures
for three or more alternatives require more information than pairwise
comparisons (Saari 1995a, 88).
6
Continue to suppose that any one of the Museum, Stadium, or the
University is a feasible or ordinarily relevant alternative, but not any two
or all three. The choice between the Museum and the Stadium cannot be
affected by peoples preferences between the Museum and the University
or between the Stadiumand the University, according to the IIA(A), even
though the University is a relevant alternative in the ordinary sense of the
term. Voter preferences are distributed as in Table 6.4.
Voters are asked to rank all alternatives. Begin with the actual sce-
nario in Table 6.4. The Condorcet advocate insists that the Stadium
should win, even though almost half the voters rank it last among all
projects. The Borda advocate insists that the Museum should win, it is
the rst choice of almost half the voters and the second choice of the other
half. The presence of the third alternative of the University on the ballot
assists in the decision because it discloses that the 99,000 voters rank
the Stadium last. In choosing between the Museum and the Stadium,
eliminating from consideration the genuinely relevant alternative of the
University ensures that the Stadium wins, and conceals the fact that al-
most half the voters would rather build anything but the Stadium. The
Condorcet rule does not violate the IIA(A), but in this example the Borda
rule does violate the IIA(A). Compare the pair of the Museum and the
Stadium, and inspect the counterfactual scenario in Table 6.4. Sup-
pose the 99,000 voters change their ranking of the university fromsecond
to third. Then the social choice by the Condorcet rule would continue to
be the Stadium, but the social choice by the Borda rule wouldchange from
the Museum to the Stadium the IIA(A) is violated. Next, suppose that
the university is infeasible, is an ordinarily irrelevant alternative. Nothing
in the foregoing analysis changes, except that addition of the ordinarily
Table 6.4. The relevance of irrelevant
alternatives
Actual 99,000 voters 100,000 voters
1st Museum Stadium
2nd University Museum
3rd Stadium University
Act M S C (BC)
M 99 199 (298)
S 100 100 (200)
C 0 99 (99)
Counterfactual 99,000 voters 100,000 voters
1st Museum Stadium
2nd Stadium Museum
3rd University University
C M S C (BC)
M 99 199 (298)
S 100 199 (299)
C 0 0 (0)
irrelevant alternative would have made more information available for
a better decision. The IIA(A) decrees that in all circumstances there is
nothing to be said in favor of any method that considers information
beyond pairwise comparisons. If the IIA(A) were strictly and literally ap-
plied, it would forbid the social-choice process even fromconsidering any
public arguments concerning the alternatives, as that would be informa-
tion beyond the pairwise rankings of voters.
The Arrovian tradition equivocates on relevance. The IIA(A) con-
dition does nothing more than require that in a choice between two al-
ternatives a third alternative should have no inuence. Whether any of
those alternatives are relevant or irrelevant, feasible or infeasible, avail-
able or unavailable, in the ordinary sense of those terms, has nothing to
do with the IIA(A) condition. The choice could be among two ordinarily
irrelevant alternatives, and the IIA(A) would forbid that a third ordinarily
relevant alternative inuence the choice between the two ordinarily irrel-
evant alternatives (then we would have to rename it the independence of
relevant alternatives condition). The choice could be between ordinarily
relevant alternative x and ordinarily irrelevant alternative y, and then the
IIA(A) condition would require that the social choice between x and y
not be inuenced by preferences over some third alternative z, no matter
whether z is ordinarily relevant or irrelevant.
Arrow (1952) says that all actual voting methods respect IIA(A). It is
true that most elections do not consider ordinarily irrelevant alternatives
(and when they do voters mostly ignore them), but it is denitely not true
that all voting methods proceed by pairwise comparison. For example,
suppose there is a natural election among candidates, say there are six.
There are certain qualications for entry, such as residence and age,
and to be eligible a candidate must declare before a certain date. The
election is carried out by Hare preferential voting. This election does not
violate ordinary irrelevance because no ordinarily irrelevant candidates
are considered. It does violate IIA(A), however, because the Hare method
does not proceed by pairwise comparison.
Arrow (1967) contains a remarkable statement:
For example, a city is taking a poll of individual preferences on alternative meth-
ods of transportation (rapid transit, automobile, bus, etc.). Someone suggests
that in evaluating these preferences they also ought to ask individual preferences
for instantaneous transportation by dissolving the individual into molecules in
a ray gun and reforming him elsewhere in the city as desired. There is no pre-
tence that this method is in any way an available alternative. The assumption of
Independence of Irrelevant Alternatives is that such preferences have no bearing
on the choice to be made. It is of course obvious that ordinary political decision-
making methods satisfy this condition. When choosing among candidates for an
elected ofce, all that is asked are the preferences among the actual candidates,
not also preferences among other individuals who are not candidates and who
are not available for ofce.
If the IIA(A) states that nonexistent alternatives should not be listed
on ballots, then there would be no controversy about it. The IIA(A),
of course, states something else entirely, that only pairwise comparisons
should be inputs to social choice. Yes, the IIA(A) agrees with common
sense by excluding the ray gun, but at the cost of excluding all but pairwise
voting in consideration among the feasible alternatives of rapid transit,
bus, automobile, etc. The IIA(A) way overshoots. It is as though someone
in Canberra refuses to leave his room because hes heard that theres a
dangerous snake somewhere in Sydney. We point out to himthat he wont
get bit by walking around Canberra, but he replies that he wishes to get
no closer to the snake.
We must be careful here about confusing the IIA(A) and the IIA(RM).
I dont think that Arrow in the example is thinking of adding the ray
gun to the menu of alternatives (possible violation of IIA(RM)). What he
means, I think, is that preferences over the infeasible ray gun shouldnt
inuence preferences over (any two) feasible alternatives (possible viola-
tion of IIA(A)). The way to avoid that inuence is to decompose all social
choice into pairwise comparisons, and then to string together the pairwise
choices over alternatives of interest, but that remedy carries an immense
price: dictatorship as the only acceptable social-welfare function. It is as
though the obsessive Canberran chooses to starve himself to death rather
than leave his house.
Davies and Shah (1993) dene a new condition of the independence of
unavailable alternatives (IUA): that information be considered only from
rankings of alternatives designated as available, and that information
about alternatives designated as unavailable be ignored. In a between-
subjects experimental condition, the IUA was violated by a majority of
subjects. In a within-subjects condition, and when voter proles were pre-
sented in a pairwise manner, there was a tendency for subjects to respect
the IUA. I suggest that the IUA is compelling in some contexts, but not
in others.
In 1963, Arrow (1963/1951, 110) commented that the austerity im-
posed by the IIA(A) is perhaps stricter than necessary; in many situa-
tions, we do have information on preferences for nonfeasible alternatives.
It can be argued that, when available, this information should be used in
social choice. My business school and theology school example shows
this. Later, Arrow (1997, 5) said of an approach such as Bordas that it is
not willing to take the logical next step, of adding irrelevant alternatives
to the list of candidates just to get extra information. I agree that peo-
ple usually would not advocate adding noncandidates in order to obtain
more information, but this is a practical consideration, not a logical one.
I can conceive of circumstances where people would advocate consid-
eration of irrelevant alternatives just to gain extra information. Suppose
that a forestry workers cooperative is voting to select a site for a new
ofce. The ofce subcommittee has searched diligently and presents to
the membership the only two alternatives available on the market. One is
small but centrally located, the other is remote but large. Discussion sug-
gests that sentiment is stronger for the small ofce, but a straw vote over
small and large indicates a tie between the two. Discussion also reveals
that a large majority would prefer an intermediate alternative if it were
feasible. The situation reects the following preference orders: 25 mem-
bers rank small >intermediate >large; 50 members rank intermediate >
large > small; and 25 members rank intermediate > small > large. The
Condorcet rule and the local Borda rule over the feasible pair yields
50 votes for small and 50 votes for large. Someone suggests voting by
the global Borda count over the feasible small and large alternatives and
the infeasible intermediate alternative. The result is intermediate (175) >
small (75) > large (50). If all along we had used the global Borda count
over the three alternatives, then we would have violated the IIA(A)
with respect to the two feasible alternatives. Global Borda says that
small > large, but if the preferences of the 25 who ranked small >
intermediate > large were to change to small > large > intermediate
then the global Borda result would change to small large, in violation
of the IIA(A). Adding consideration of the infeasible alternative of the
intermediate ofce shows both that an intermediate ofce would be most
favored, and that a small ofce is favored over a large ofce. As a result,
the members instruct the ofce subcommittee to pursue more aggres-
sively intermediate alternatives, and, if none is found, to secure the small
ofce. The example shows that consideration of ordinarily irrelevant or
of third alternatives can be substantively rational, and is even strongly
advisable in some circumstances.
Generally, though, irrelevant alternatives are not added to gain extra
information. Why might that be? One reason, I suppose, is that in all con-
texts relevance is a compelling imperative. A stronger reason, perhaps, is
that in many cases, unlike in my example of the cooperative, there is no
motivation to include alternatives incapable of selection. What happens
if the noncandidate wins? If the election is carried out by plurality rule,
as many are, then the victory of a noncandidate would require a new
election. I recall that once those of a rebellious bent proposed to place
none of the above on American plurality ballots, but the proposal fell
at because of the practical need to have a winner. Adding irrelevant al-
ternatives has the air of frivolity and irresponsibility. Perhaps some would
be motivated to add the noncandidate in order to manipulate the election
and would not vote sincerely, but certainly even if many were to consider
John Stuart Mill the best alternative for mayor few would bother to vote
sincerely for that choice on the ballot. Arrows point is one of practical
observation rather than of logical objection.
Voting rules may be justied independently
of interpersonal comparisons of utility
Arrow(1952, 52) states that to violate the IIA(A) presents an operational
problem: preferences between impossible alternatives make virtually no
sense, for they correspond to no action that an individual could imagine
having to perform. He must be speaking rhetorically, as immediately
before he is speaking of logically possible alternatives or of imaginable al-
ternatives, not of impossible alternatives. He must mean that individuals
are required to have preferences over all possible alternatives X, in
order to decide among feasible alternatives in some subset S, and then
his argument would be that critics of the IIA(A) require people to have
preferences over infeasible alternatives in X S. If he is correct that it
makes no sense to have preferences over infeasible alternatives in the con-
text of applying a voting rule, however, then it would follow as well that
Arrows entire social welfare perspective itself makes no sense, because in
his larger scheme, it is assumed that each individual in the community
has a denite ordering of all conceivable social states (Arrow1963/1951,
17), which includes of course the infeasible, however that is dened. In
his 1963 addendum, Arrow (1963/1951) offers further justication of
the IIA(A). He begins with reference to ordinalism in economics and
expresses the belief that ordinalism is desirable because it is based only
on interpersonally observable behavior. The IIA(A), he says, extends the
requirement of interpersonal observability a step further. One could ob-
serve all preferences over alternatives available to or feasible for society,
but one could not observe preferences over alternatives not available to
or feasible for society, demonstrating the practicality of requiring voting
procedures to respect the IIA(A). Again, this calls into question the entire
Arrovian scheme. Arrow (1967, 61) later directly mentions this inconsis-
tency with respect to the social-welfare perspective: howcan we possibly
know about hypothetical choices if they are not made? His response is
to pass by the issue.
Arrows concern about observability arises from the doctrine of log-
ical positivism, just past its apex of inuence as Arrow wrote in 1951.
Since 1951, logical positivism has fallen into disfavor in all disciplines
outside economics, but inside that discipline its strictures linger to this
day. According to this doctrine, any statement is meaningful if and only
if it is possible to specify a set of observations that would verify it; and
the meaning of the statement is exhausted by the specication of these
observations, write Barry and Hardin (1982, 249).
7
The viewpoint of Arrows Social Choice and Individual Values
(1963/1951, 911, 109111) is the claimthat the interpersonal compar-
ison of utilities has no meaning and, in fact, that there is no meaning rele-
vant to welfare comparisons in the measurability of individual utility (9).
The only meaning that the concepts of utility can be said to have is
their indications of actual behavior, says Arrow, and if such a course
of behavior can be explained by a given utility function it can also be
explained by a second that is a strictly increasing function of the rst.
Von NeumannMorgenstern utilities over alternative probability distri-
butions are no way out, because the behavior explained by a given func-
tion of that type can also be explained by a second that is a positive afne
transformation of the rst.
8
Even if such utility were measurable, says
Arrow, which function out of an innite family should be selected to rep-
resent the individual, and which function should be selected to aggregate
the individual utilities? The selection of the functions requires a de-
nite value judgment not derivable from individual sensations (11). Sen
(1970, 97) points out that the choice of one cardinal utility function or
another is descriptively arbitrary, but in an ethical argument one may
wish to choose some particular scaling in spite of this arbitrariness, on
some other grounds that may be additionally specied. Arrow seems to
go in the other direction: If there is no empirical way of comparing two
states . . . there can be no ethical way of distinguishing them (112). If the
welfare of different individuals is empirically and hence ethically indistin-
guishable, then there is no reason to be more concerned for the poor than
for the rich in social policy, as the rich are no better off than the poor ac-
cording to the doctrine of noncomparability. Perhaps Arrows statement
is not intended to carry outside the context of his criticismof Samuelsons
welfare function: even if the rich and poor were to be indistinguishable in
terms of observed mental states, they are surely empirically and ethically
distinguishable by objective measures.
In 1963 Arrow criticizes Bordas justication of his method. Borda,
says Arrow, gave equal weight to differences between adjacent rankings
and gave equal weight to different voters. The rst raises the problem
of the measurability of utility, the second that of interpersonal compa-
rability of utilities (1963/1951, 94). Borda justied the rst claim with
the argument that if a voter ranks B between A and C, then we have
no information about whether the intensity between A and B is more or
less than the intensity between B and C, according to Arrow. The sec-
ond claim is justied on the grounds of equality of voters, an ethical
claim. Arrow immediately contrasts the Borda method to the Condorcet
method (that a candidate who receives a majority against each other can-
didate should be elected) and praises Condorcet as consistent with the
IIA(A).
We begin with Bordas second claim. Notice that every democratic vot-
ing rule, including Condorcet, gives weight (usually equal) to each voter.
9
If giving equal weight to voters in the Borda count offends the noncom-
parabilist doctrine, then so does the Condorcet method, and indeed any
political aggregation which assigns weights, equal or unequal, to voters. In
1951 Arrow(1963/1951, 46) proved that the method of majority decision
over exactly two alternatives does satisfy the conditions of his theorem, in
other words, is a possible social-welfare function; the impossibility result
pertains to three or more alternatives (when applied pairwise over three
or more alternatives we call this the Condorcet method, which can cycle).
He eludes the interpersonal comparability of utility problem by means of
the following denition:
By the method of majority decision is meant the social welfare function in which
x R y [the social ranking] holds if and only if the number of individuals such that
x R
i
y [individual rankings] is at least as great as the number of individuals such
that y R
i
x.
Either counting the numbers of individuals for, against, or indifferent to
an alternative violates the prohibition on the interpersonal comparison
of utility, and all democratic voting is thereby rendered meaningless, or
the assumption of one vote per individual has nothing to do with inter-
personal comparison of utility. Reading Arrow charitably, it must be that
one vote per individual has nothing to do with comparison of mental
states. If one vote per individual is acceptable because it has nothing to
do with comparison of mental states with respect to the method of major-
ity decision, then it must be acceptable as well with respect to any other
voting method, including Borda. Therefore, the Borda method does not
necessarily rely on interpersonal comparison of utility. One variety of
noncomparabilist might further urge that because of noncomparability,
each voter should be given equal weight.
One argument is Bordas rst claim, as related by Arrow: just because
the individual preference data are ordinal, it is best to assign equal weight
to the differences between adjacent rankings. Neither the Borda count
nor the Condorcet method is a utilitarian method of voting: each is an
operation performed on ordinal rankings.
10
That the Condorcet method
(in the absence of cycles) ranks all alternatives does not necessarily make it
a cardinal voting scheme, and the same goes for the Borda count. Some
friends, and enemies, of the Borda count argue that it approximates a
cardinal voting scheme, that it attempts to squeeze cardinal blood from
the ordinal turnip (Mackay 1980, 73), but the same could be said for
the Condorcet method. Another argument has it that as the number of
voters increases, cardinal variations would tend to cancel each other out,
and the Borda outcome would tend to the cardinal outcome (as would
Condorcet, see Tangian 2000).
The Borda count can be independently justied on purely ordinal
grounds, however. It collects exactly the same information as does the
Condorcet method, as can be seen in the pairwise-comparison matrix,
but instead sums up the number of votes each alternative gets over every other
alternative, and socially ranks by those sums. The alternative that gets
the most votes over every other alternative is the winner. There is no ref-
erence to cardinality or to comparable mental states in this justication
of the Borda count. Arrows denition of the method of majority deci-
sion counts the number of individuals who prefer one or another of a
pair of alternatives, and is phrased such that it does not necessarily have
anything to do with comparable mental states. Counting the number of
times an alternative is preferred by individuals to every other alternative,
as does the Borda count, also does not necessarily have anything to do
with comparable mental states. Objection: but that would make the out-
come depend on the range of alternatives. Reply: just as the outcome
should depend on the range of voters.
Any democratic voting rule Condorcet, Borda, other weights vot-
ers (usually equally). Once that is conceded, the Borda count, just as the
Condorcet method, is otherwise justied within the noncomparabilist
framework. The denition of the social-welfare function requires indi-
vidual and social orderings. The independence aspect of the IIA(A) is
adventitious. Its purpose is to exclude voting rules otherwise compatible
with ordinal input, and thereby deliver the impossibility result.
The IIA(A) does too much
Riker (1982, 101) describes the IIA(A) as requiring that a method of
decision give the same result every time from the same prole of ordinal
preferences. He says that the IIA(A) seems a fundamental requirement
of consistency and fairness to prevent the rigging of elections and the un-
equal treatment of voters, and that it has been disputed. In a footnote
(271), he explains that the main reason for the dispute is that Arrows
original discussion of the condition confused it with contraction consis-
tency. The IIA(A) condition has three desirable consequences, says Riker.
First, it prohibits utilitarian methods of voting. Second, it prohibits ar-
bitrariness in vote counting, such as lotteries or other methods that give
different social choice from the same given prole of individual prefer-
ences. Third, it prohibits, when choosing among alternatives in a set S,
inuences from alternatives in X S. I respond in turn.
First then, we shall consider utilitarian methods of voting. Riker (1982,
118) thinks utilitarian voting gives advantages to persons with ner per-
ception and broader horizons. Would utilitarian voting, which on eth-
ical grounds counted each voter equally, and which accepted intensities
of preference from cardinal rankings, were it practical to carry out and
resistant to misrepresentation of intensity, be undesirable? I think the
main objection to utilitarian voting is its impracticality rather than its
comparable cardinality; if it were a practical voting system, it might be
an excellent one (it might be an excellent voting rule, but it would not
necessarily dene what is true or right). Another advantage of the IIA(A),
according to Riker, is that it forbids the arbitrariness of the Borda count.
As I have shown, the justication of the Borda count need not depend on
comparison of mental states. Furthermore, the Borda count avoids two
problems of utilitarian voting: rst each voter is counted equally, and sec-
ond the weighting of the alternatives is established by the voting scheme
and thus avoids voter misrepresentation of intensity.
Under the Borda count and with more than two alternatives, if an
alternative is lowered in an individual ranking that may lower the social
ranking of the alternative. Is this arbitrary? Riker (1982, 108) provides an
example of how the Borda count violates IIA(A), but the example seems
rather feeble, and arguably supports the opposite view that violation of the
IIA(A) by the Borda count is a desirable attribute. Suppose that Larry ranks
A > B > C, that Moe ranks C > A > B, and that Curly also ranks
C > A > B. Then A and C tie with a Borda score of four points each,
and are both preferred over B which has a Borda score of one point (the
Condorcet outcome from the prole is C > A > B). The IIA(A) says
that the social ranking of A and C should depend only on the individual
rankings of A and C. Thus, if Moe changes his ranking from C > A > B
to C > B > A, then, the IIA(A) decrees, the social ranking of A should
not change. The Borda count after Moes change, however, is four points
for C, three points for A, and two points for B (the Borda outcome is
now C > A > B, and the Condorcet outcome remains C > A > B). The
Condorcet method reports the same outcome for both proles of indi-
vidual preferences: C > A > B. Preferences with respect to only C and A
are the same in each prole, so the Condorcet method here is respecting
the IIA(A) (there is no cycle in either prole). The Borda method distin-
guishes the two proles. Under the rst prole, the Borda method nds
that C is tied with A. Why does this differ from the Condorcet outcome?
Condorcet reports that C is better than A, but the Borda count takes into
account that although C is rst-ranked by two voters, it is last-ranked
by one voter, and that A is rst-ranked by one voter and second-ranked
by two voters. Under the second prole, the Borda method nds that C
is better than A. Why the change in the Borda outcome from the rst
prole to the second prole? In the rst prole, C is tied with A, but the
change from the rst to the second prole is that one voter changes A
from second-ranked to last-ranked. The Borda method responds to this
change, but the Condorcet method does not so respond. The IIA(A) says
that a voting rule should never respond to that sort of change. Practical
evaluation of alternative voting rules must consider actual frequencies of
good and bad violations of the IIA(A).
Second, the IIA(A) also excludes any probabilistic voting rule, accord-
ing to Riker (1982, 118), because fromthe same given prole of individual
preferences it requires that a voting rule always return the same result.
For example, imagine the rule that each of n voters writes down his rst
preference among the alternatives on a ballot and then one ballot is ran-
domly drawn to determine the social choice. Incidentally, such a rule
would be strategy-proof, since each voter has the incentive to report her
true preference, which is why the Gibbard and Satterthwaite theorems
are limited to denite voting rules. Or suppose a rule such that if 90% of
voters prefer x over y, there is a 90% chance that the social choice would
be x over y; such a rule would be prohibited by the IIA(A). The missing
premise in his argument is that Riker does not establish that probabilistic
voting rules are normatively undesirable in all or in any circumstances.
We would have to consider in detail, theoretically and practically, the
consequences of various such rules.
Third, many people believe that judgments on alternatives in X S
are germane to judgments on S itself, that is, they doubt the IIA(A)
(Riker 1982, 129). To charitably reconstruct Rikers argument, there is
no formal method to decide actual relevance and irrelevance, thus, the
only way to guard against the threat of irrelevant alternatives having an
inuence on decisions is by way of the pairwise IIA(A) condition. More
precisely, there is no wholly defensible method to decide on degrees of
irrelevance. In the absence of such a method, Condition [IIA(A)] seems
at least moderately defensible.
If, for whatever reason, one were motivated to satisfy Rikers rst and
second concerns, to disqualify utilitarian or probabilistic voting methods,
then other approaches are available that would do the job yet would also
permit some voting systems; and if there were somewhere a real human
being concerned about wrongful inuences from consideration of ordi-
narily irrelevant alternatives, then there are approaches that address that
concern but avoid the impossibility conclusion.
I begin with Hansson (1973), of whom Riker (1982, 275) is aware.
If feasible and infeasible alternatives can be clearly identied, then one
should simply conne consideration to feasible alternatives (the IUAcon-
dition). For example, if the application of interest is a standard political
election among candidates, then the feasible candidates are those who
stand for the election. Infeasible candidates would not be considered
just because they are infeasible candidates. Then the election would be
independent of infeasible alternatives, not in Arrows sense of avoiding
consideration of third alternatives, but rather in the ordinary sense of
avoiding actually infeasible alternatives. Problem solved. Nor is this ter-
minological trickery: the natural reaction to any proposal to include, for
example, Napoleon and Confucius as extra candidates in the city coun-
cil election is ridicule and rejection. This is a clear and simple solution
to the IIA(A) problem with natural applicability in the political context.
Because it is so simple, I take a moment to reiterate that people sincerely
concerned about consideration of infeasible alternatives have no worries
in this political context.
One of Arrows (1952, 52) main concerns with respect to the
IIA(A) is:
If it is abandoned, a choice among a given set of candidates can be made only if
each individual possesses a list of preferences containing more candidates than
those which are really available. What will be the list of candidates that the indi-
vidual will be asked to order? There is no natural limit except a vague universe
comprising all logically possible candidates. It does not appear correct to make
the choice among a very limited set of possibilities depend on all preferences
among all imaginable possibilities.
Notice that this objection does not apply to the voting context discussed
in the last paragraph. Perhaps it does apply in the social-welfare context,
and perhaps welfare economists have more reason to worry about the
IIA(A) than those in political studies. The worry is that the social-welfare
function would have to consider not only actual alternatives, but also the
analogs of Thomas Jefferson, Socrates, Buffalo Bill, ctional people, an-
imals, ctional animals, androids, and rocks, and it is costly in terms
of resources to gather and process information about preferences
independence is a requirement that we conserve those resources (Kelly
1988, 73). First, the costly resources argument is surprising. I think of
the social welfare function as an idealization, not an exercise actually car-
ried out in full. Decision theory constantly proposes simplifying models
that it would be practically impossible for real agents to carry out, for
example, because of combinatorial explosions. The practical impossibil-
ity of carrying out the calculation is seldom raised as an objection to the
model. Second, it may well be, and I conjecture that it is likely, that the
inuence of irrelevant alternatives would be benecial, neutral, or triv-
ial, especially with large numbers of voters and alternatives. Third, I am
not aware of whether welfare economists have explored the concept of
possibility with more philosophical rigor; if not, perhaps that would help
clarify the problem.
Hansson offers weakened independence conditions. Again, X is the
set of all alternatives, nonempty subset S contains the feasible alterna-
tives, and X S contains the infeasible alternatives. IIA(A) would be
violated if changes in preferences over two alternatives in X S changed
the social choice over alternatives in S, and also would be violated if
changes in preferences over two alternatives, one in S, and one in X S,
changed the social choice over alternatives in S. Hanssons positionalist
independence (PI) condition, like IIA(A), forbids changes in preferences
over two infeasible alternatives in X S to inuence the social choice, but
it permits changes in preferences over two alternatives, one in feasible S
and one in infeasible X S (or both in S), to inuence the social choice
over the alternatives in S. This PI condition would void Arrows impossi-
bility result for the Borda count, the Copeland method, and some other
voting rules.
He also offers a strong positionalist independence (SPI) condition.
How much can you change individual orderings without changing the
social choice between two alternatives x and y? Divide an individuals
rankings of alternatives other than over x and y into ve groups: those
below both x and y; those the same as y; those between x and y; those
the same as x; and those above both x and y. IIA(A) permits that we can
move around in individual orderings alternatives other than x and y as
much as we like, from any group to another, without changing the social
choice. The PI condition allows us to move around alternatives other
than x and y only within each of the ve subgroups. SPI is more generous
than PI; it allows us to move alternatives from above both x and y to
below both x and y, and from below to above, without changing the so-
cial choice. Although stronger, the SPI condition still permits the Borda
count, that is, substituting SPI for IIA(A) in the Arrow theorem shows
that the Borda count is a possible social-welfare function (but not the
Copeland rule, etc.) Much more simply, Saari (1995a, 97) offers an
intensity IIA condition. Suppose that we measure intensity of pref-
erence only by the number of candidates a voter uses to separate a pair,
as in the Borda count. The intensity IIA requires that the ranking of any
two candidates depends only on each voters ranking of those candidates
and the intensity of that ranking. Again, the intensity IIA avoids Arrows
impossibility result. Also, recall that YoungKemeny, a Condorcet-
extension voting rule which resolves cycles, gets by with a local inde-
pendence of irrelevant alternatives condition.
The alternative independence conditions would entirely satisfy Rikers
rst and second concerns to forbid utilitarian and probabilistic voting
rules. The device of considering only feasible alternatives when there is a
clear boundary between the feasible and the infeasible, such as in many
democratic-voting applications, entirely satises Rikers third concern
about avoiding inuence from preferences over infeasible alternatives,
even though it would violate Arrows IIA(A) if more than two feasible
alternatives were considered. When the boundary between the feasible
and the infeasible is less clear, the weakened independence conditions
partially satisfy Rikers concern that decisions be independent of infeasi-
ble alternatives, a concern about which he can say no better than that it
is moderately defensible.
Table 6.5. Borda manipulation,
initial situation
3 2
X Y
Y X
X Y (Borda)
X 3 (3)
Y 2 (2)
Independence is not a practical requirement
Recall that Arrow believes that a positional method such as the Borda
count is defective because it must consider some vague universe of all
possibilities. The Borda method must consider all possible candidates,
not merely actual candidates. The Condorcet method is independent of
irrelevant alternatives, therefore it need not consider all possible candi-
dates, but rather may consider only actual candidates, and thus is practi-
cally advantageous, the story goes. The Arrovian tradition has added that
the Borda method is subject to manipulation by the addition and sub-
traction of candidates, but that the Condorcet method is not. Mueller
(1989, 394) glosses Arrow:
The outcomes under the Borda procedure and similar schemes depend on the
specic (and full) set of issues to be decided. Thus, abandonment of the inde-
pendence axiom raises the importance of the process that selects the issues to be
decided in a way that its acceptance does not.
The supposed contrast is merely academic folklore, however. Notice that
the violation alleged is of the IIA(RM), not the IIA(A). The IIA(A) takes
alternatives considered as given, and varies preferences over alternatives;
the IIA(RM) takes preferences as given, and varies alternatives consid-
ered. Both the Condorcet and the Borda methods depend, although in a
tenuous way, on the set of issues to be decided. And both the Condorcet
and the Borda methods are susceptible to manipulation by addition or
deletion of candidates.
The Borda count is theoretically subject to a special case of manipula-
tion (Dummett 1998) by adding new candidates from outside the given
set a manipulator might be able to change the outcome. Lets begin with a
prole of ve voters and two alternatives, and the accompanying pairwise-
comparison matrix, as in Table 6.5. Obviously, X is the Borda winner. If
rst step
3 2
X Y
Y Z
Z X
X Y Z (Borda)
X 3 3 (6)
Y 2 5 (7)
Z 2 0 (2)
second step
3 2
X Y
Q Z
Y X
Z Q
X Y Z Q (Borda)
X 3 3 5 (11)
Y 2 3 2 (7)
Z 2 2 2 (6)
Q 0 3 3 (6)
the Borda count is the voting rule, then supporters of Y can boost Y to
rst place, however, by introducing an alternative Z that is very similar
to Y but just below it in everyones preference rankings. See Table 6.6.
This little trick changes the Borda order from X > Y to Y > X > Z, a
violation, of course, of contraction consistency. Dummett proposes a x
for this eventuality that shall not detain us here. Another x is strategic
deterrence: the partisans of X can restore X to rst place with an iden-
tical maneuver: they just introduce alternative Q that is similar to X but
just below it in everyones rankings. So we are back again to X > Y, just
where we started. Why would the partisans of Y introduce manipulative
alternative Z, if they anticipated that the partisans of X would respond
with the introduction of countermanipulative alternative Q? The result if
everyone manipulates is the same as the result if no one manipulates, and
if there were transaction costs to adding alternatives then would anyone
bother with this ploy?
Assume, however, that the Borda method is guilty as charged. The
same charges can be laid against the Condorcet method. In order to
carry out a proper Condorcet ranking one has to rank all alternatives fea-
sible and infeasible.
11
Why? Suppose that the two feasible alternatives are
x and y, and that the social preference over the set S of x and y is
x >y and thus that the social choice appears to be x. That is not sufcient
to determine the social choice, however. We also have to examine all infea-
sible alternatives z in X S in order to determine that it is not the case that
the social preferences are z > x and y > z. Because if it happens that for
some alternative z that z >x and y >z, then there is a cycle z >x >y >z.
If there is such a cycle, then the choice set is empty. When we apply the
Condorcet method only to feasible alternatives x and y, the choice be-
tween x and y is x. When we apply the Condorcet method to all alterna-
tives feasible and infeasible, then it may be that there is no social choice
between x and y. The Condorcet method seems to be especially absurd
here: society would prefer relevant x to relevant y despite irrelevant al-
ternative z, but the possibility of irrelevant alternative z prohibits choice
between relevant x and relevant y. If the objection is tenuous against
the Condorcet method, then the objection is tenuous as well against the
Borda method.
How do the Condorcet and Borda methods compare with respect to
the manipulative addition and deletion of alternatives? Someone who
would manipulate by addition or subtraction of alternatives needs a vast
amount of information. She needs to know in advance, before the vote,
the preference rankings of all of the voters. If she is a manipulator by way
of addition of alternatives, which I will show is the more important case,
she also needs to know not only the voters preference rankings over the
relevant alternatives naturally under consideration but also their rank-
ings over some number of irrelevant alternatives in order to be able
to select from the irrelevant alternatives the one or more that would
permit her manipulation to succeed (assuming that the one or more ir-
relevant alternatives she needs exists). Again we stumble across an in-
consistency in Arrovian tradition. On the one hand, change of preference
over an irrelevant alternative may suspiciously change the outcome. On
the other hand, the ordinalist revolution requires that welfare judgments
be based only on interpersonally observable behavior, according to
Arrow (1963/1951, 110), that, ideally, one could observe all prefer-
ences among the available alternatives, but there would be no way to
observe preferences among alternatives not feasible [irrelevant] for so-
ciety. On the one hand, the possibility of manipulation is a threat.
On the other hand, the information required for the manipulation is
not available.
If an aspiring manipulator has the requisite information, then a new
problem will arise in many circumstances. With more than a few voters
and a fewalternatives it is quite complicated to calculate the manipulative
strategy. The calculation of a strategy may take such a long time that it is
not practical to carry out, and in the extreme it may be exponentially hard
to carry out, that is, it would take more time to calculate than there is
in the universe, for example. Bartholdi, Tovey, and Trick (1989) showed
that for more than a fewvoters or alternatives it would be computationally
impractical (NP-hard) to calculate the election outcome by the Dodgson
method or the YoungKemeny method. Bartholdi, Tovey, and Trick
(1992a) showed that a version of the Copeland method is computationally
resistant to manipulation by strategic voting; Bartholdi and Orlin (1991)
showed the same for the single transferable vote. Bartholdi, Tovey,
and Trick (1992b) compared the manipulability of plurality rule to the
Condorcet method with respect to adding candidates, deleting candi-
dates, partitioning candidates, adding voters, deleting voters, and parti-
tioning voters. Although plurality is logically susceptible to the rst three
manipulations, it is computationally resistant to such manipulations;
Condorcet is logically susceptible to the second three manipulations, but
is computationally resistant to them. There is the magnicent and crisp
issue of whether a manipulation is calculable in the lifetime of the uni-
verse that Bartholdi and coworkers consider, and there is also the modest
and fuzzy issue of whether a manipulation is practically calculable by
real humans in real time in real settings, that could only be decided by
empirical investigations.
It is important to distinguish whether the option of manipulation by
addition and deletion of candidates is open only to one actor or is open
to all. If it is open to one unconstrained monopoly actor, then no matter
what reasonable voting rule is in force, she can just delete all candidates
she doesnt like and add the one candidate she does like, which is equiva-
lent to agenda dictatorship. Practically, the addition of alternatives must
be relatively open, and the subtraction of alternatives must be relatively
closed (or made on the basis of criteria selected well in advance under
veil-of-ignorance conditions, otherwise one would face the conundrum
of needing to apply the voting rule in order to decide which candidates to
subtract from consideration by the voting rule). If we posit that an actor
has a manipulative advantage arising from her unconstrained monopoly
power to delete candidates we have not demonstrated that a voting rule
has unfair consequences. We have said nothing different than it would
be unfair to give 99 of the voters one vote each while giving one special
voter 100 votes to cast; we have quietly assumed violation of voter equal-
ity in order to publish loudly a violation of voter equality. Perhaps it is
meant that a trusted and accountable authority would delete a candidate
in order to manipulate by stealth.
Suppose that information and calculability are not problems. Dene
the criterion of manipulation to be changing a winner into a nonwinner
by the addition or deletion of alternatives. Then the Condorcet method
is just as vulnerable to manipulation by deletion of candidates as is the
Borda method. The manipulating Condorcet chairperson simply deletes
all alternatives that in pairwise comparison would defeat her favored al-
ternative. It may be objected that what is meant is that the manipulator
would be able to succeed by stealth under the Borda method but not un-
der the Condorcet method. That smuggles in a newassumption, however:
not only would the manipulator have monopoly control over the deletion
of alternatives, the manipulator would also have a monopoly control over
information about the distribution of preferences and over calculative ca-
pacity. Then all we have shown is that those with unfair power have unfair
inuence. Manipulation by deletion of alternatives does not seem to be
a plausible scenario.
The Condorcet method is as vulnerable to manipulation by addition of
candidates as is the Borda method. If in the absence of manipulation the
social choice would have been x >y, the Condorcet manipulator as with
Borda, any of the voters, not only the chair simply adds an alternative
z that creates a cycle x > y > z > x, and x is no longer the Condorcet
winner. It may be objected that there may not exist an alternative z that
would allow for the manipulation, but the same could be said for the
Borda count, that there may not exist an alternative the addition of which
would change a winner to a nonwinner. We have established that both
Condorcet and Borda are susceptible to manipulation by addition or
deletion of alternatives.
Finally, there is an additional complexity. Manipulation begets coun-
termanipulation. Someone contemplating the strategic addition of an al-
ternative also has to consider the strategic response of other actors. It may
be that strategic addition by all actors would cancel out and the outcome
would be the same as in the absence of strategic action, it may be that
strategic addition by all actors would lead to an outcome intended by no
one, it may be that strategic actions are impractical to calculate. These
issues are not well settled.
Gibbard (1973) and Satterthwaite (1975) showed that voting proce-
dures that satisfy the IIA(A) are immune from strategic voting, but those
that violate the IIA(A) are susceptible to strategic misrepresentation of
preferences. This breathes new life into the IIA(A) and the Arrow theo-
rem. While admitting the attractive qualities of the Borda count, Mueller
reports that its Achilles heel is commonly felt to be its vulnerability to
strategic behavior (120). But the GibbardSatterthwaite theoremshows
that all voting methods of interest share this defect of being susceptible
to manipulation by strategic misrepresentation of preferences, including
Condorcet pairwise comparison among three or more alternatives (but
only if the prole of voters preferences is such as to yield a cycle, which
I have argued is rare). It may be that the Borda count is more practically
and irremediably susceptible to strategic misrepresentation of preferences
than other reasonable voting rules; it may be not (see Chamberlin 1985,
Nitzan 1985, Saari 1990).
Conclusion
The Arrow theorem is a great piece of work. It illustrates an abstract
limit case. It is a logical exercise, it does not describe the real world.
The conditions of the theorem, especially IIA(A), are methodological
assumptions, with no descriptive or normative force of their own. That
a democratic voting rule violates IIA(A) is nothing to be feared, as such
violation has not been established to be normatively undesirable, all the
more so given that insistence on the conditions leaves the dictatorship of
one as the only possible voting rule.
Any democratic voting rule forces a comparability assumption, or re-
quires a nonwelfarist justication of voting weights, and a democratic
voting rule, whether Condorcet or Borda, assigns equal inuence to each
citizen. If there is someone who believes that the idea of cardinal ranking
of alternatives is unintelligible, or judges that it is impractical, she still
might be able to endorse either the Condorcet or Borda methods, which
work on ordinal rankings. If there is someone who fears the threat of
Condorcet cycles, she still might be able to endorse the Borda count as a
method that avoids cycles. If there is someone who worries about inu-
ence from irrelevant alternatives, she can recommend that elections only
consider actual candidates, and when the boundary between feasible and
infeasible is fuzzy she can recommend the Borda count as a method that
satises a slightly weakened independence condition. These, and other
considerations in this chapter, are not a plea for the widespread adop-
tion of the Borda count. Natural similarity among preference rankings
means that in most circumstances the differences in outcomes among
the reasonable voting rules are practically minimal. If there is some-
one concerned about manipulation by strategic misrepresentation of
preferences, she might choose the single transferable vote, which experi-
ence has shown to be practically unmanipulable in this regard. Even our
most inaccurate voting rule, plurality, might be desirable because of its
simplicity. The alleged irrationalities of voting are greatly exaggerated.
It is time for a shift from constitutional physics to constitutional
engineering. Newtons rst law of motion, another great intellectual
achievement, says that an object in motion tends to stay in motion and
an object at rest tends to stay at rest, unless the object is acted upon by an
outside force. An inattentive interpreter of physics may advise engineers
that it is futile to try to make anything go, or try to make anything stop,
because of Newtons rst law. The inattentive interpreter has failed to
notice the qualifying clause: unless the object is acted upon by an outside
force. Arrows theoremsays that there are no social welfare functions that
satisfy his conditions. Yet we observe that there are social decision pro-
cesses (discussion, institutions, voting) that work satisfactorily to translate
what individuals would prefer into democratic social outcomes, and there
are social-welfare functions (voting) that work satisfactorily in practice.
That means that they do not satisfy one or another of the Arrow con-
ditions. That does not make democracy impossible, irrational, arbitrary,
meaningless, and the rest.
There is another important issue of constitutional engineering. It is no
longer enough to show that a voting rule is logically susceptible to ma-
nipulation by addition of alternatives, deletion of alternatives, addition of
voters, deletion of voters, partitioning of voters, agenda control, strate-
gic voting, and so on. It is no longer of interest to interpret such logical
susceptibilities as relevant to global normative judgments about democ-
racy, the trick is just worn out. The era of destructive social choice theory
is past. Rules that are logically susceptible to manipulation may not be
practically susceptible at all, some rules may be more practically suscep-
tible than others, susceptibility may vary with the actual distribution of
preferences in the population, the strong logical susceptibility of one rule
might be more practically remediable than the weak logical susceptibility
of another rule, and so on. The constructive social choice theory of the
future would work on the comparative practical susceptibility of rules to
unfair manipulations, and the identication or invention of institutions
to remedy them, rather than on denunciations of democratic choice.
7 Strategic voting and agenda control
As we have seen, Riker (1982, 122, 128) acknowledges that debate
and discussion lead to similarities of judgment, that a wide variety of
rather mild agreement about the issue dimension guarantees a Condorcet
winner, that uncontrived cyclical outcomes are quite rare, so that
intransitivities only occasionally render decision by majoritarian meth-
ods meaningless. His remaining objection then is that outcomes may
be manipulated by strategic voting, agenda control, and contrived intro-
duction of new alternatives and dimensions. Manipulated outcomes are
meaningless because they are manipulated, and unmanipulated outcomes
are meaningless because they cannot be distinguished from manipulated
ones (Riker 1982, 237). To introduce his concern, consider the follow-
ing exercise.
Suppose that a majority of 60 percent favors Aover B, and a minority of
40 percent favors B over A. The minority faction could contrive a cycle if
it has some idea of the distribution of preferences over original alternatives
A, B, and new issue C. Issue C splits the majority. See Table 7.1. With
C on the scene, A beats B by 60 percent, B beats C by 70 percent, and
C beats A by 70 percent. Then, the minority can propose C against A,
which C wins, then C against B, which B wins, thus, the minority view
favoring B over A prevails (but only if the majority is somehow incapaci-
tated fromcontinuing the cycle by proposing Aagainst B which of course
A would win, ad innitum).
If the majority faction were to have unconstrained monopoly control
over the order of consideration of issues, then the majority could defend
against the contrived cycle by putting B against C, which B would win,
and then B against A, which A would win, vindicating the majority view
(with agenda control, the majority prevents the minority from proposing
C against A). But wait, if the minority were thwarted by majority agenda
control it could vote strategically: rather than voting its true preference B
when Bis posed against C, it could instead vote for C, putting the minority
back at the start of its winning cycle, assuming that the majority lacks its
158
Strategic voting and agenda control 159
Table 7.1. Contrived outcomes
Before contrived cycle % After contrived cycle % After intramajority logroll %
60 A > B 30 A > B > C 30 A > A-100 > B > C
30 C > A > B 30 A+100 > C > A > B
40 B > A 40 B > C > A 60 B > C > A
own strategic capacity, and so on. The majority faction also could resist
the cycle by logrolling in its own ranks, if, say, its rst subfaction (A>B>
C) could compensate its second subfaction (C > A > B) with sufcient
cash (or, more likely, in implicit obligation), when alternative A minus
the cost of compensation is still better than B for the rst subfaction and
alternative A plus the benet of compensation becomes better than alter-
native C for the second subfaction. The contrived cycle in my example is
inspired by Rikers (1982) examples; the logrolling by Tullock (1992).
Strategic voting, logrolling, agenda control, and multidimensional is-
sue spaces are each topics that have inspired vast, challenging, and unset-
tled literatures. There are hundreds of models, considerable variations in
assumptions, and divergent results arising fromalternative assumptions. I
do not propose a comprehensive reviewof these topics. Instead, the focus
will be on looking over ndings relevant to the normative conclusions of
Rikers positive political theory: that multidimensional issue spaces make
the concept of a majoritarian democracy meaningless and that associated
opportunities for manipulation make democracy arbitrary, viz.: Out-
comes of any particular method of voting lack meaning because often
they are manipulated amalgamations rather than fair and true amalga-
mations of voters judgments and because we can never know for certain
whether an amalgamation has in fact been manipulated (1982, 238).
If the issue space is multidimensional, does that mean, in principle, that
there is no determinate public good? How frequent are opportunities for
manipulation? When there is an opportunity, how often are manipula-
tions attempted? When manipulations are attempted, how often do they
lead to unfair and inaccurate outcomes (which I shall call harmful manipu-
lations), and how often do they preserve fair and accurate outcomes? Are
there practical remedies available to deter harmful manipulations, and
if so, how often do they work? This boils down to the crucial question:
howfrequent is irremediable and harmful manipulation? Finally, does the
possibility of manipulation make the inference of preference orders so ob-
scure that we are unable to distinguish manipulated from unmanipulated
outcomes?
The answer to the nal question is necessarily no, due to an internal
contradiction in Rikers argument. In order to succeed at strategic vot-
ing, logrolling, agenda control, or the manipulative introduction of new
alternatives and dimensions, the manipulator needs condent knowledge
of the distribution of preference rankings in the population to be manip-
ulated. This is evident from the foregoing exercise (and such complete
information is assumed in the formal models). To contrive the cycle,
the minority faction would need to know the preference rankings both
among its own members and the members of the majority faction over
the original alternatives, A and B, and also their rankings in relation
to the manipulative new alternative C (a side issue that calls into ques-
tion the practicality of contriving a cycle is if the alternative is new,
how would the manipulator know the populations rankings of it?). The
majority factions attempted manipulation of the sequence of the agenda
also requires complete information about the preference distribution, as
does the minority factions attempted response at strategic voting. As
well, to succeed at the logroll, the majority faction would need to know
the distribution of preferences in its own ranks. Rikers argument is that
the possibility of manipulation makes it impossible to know the prefer-
ences of others (his defense of this premise and its failure is examined in
detail in Chapter 2 on the basic argument pattern), and since the pref-
erences of others cannot be known then it is impossible to distinguish
manipulated outcomes from unmanipulated outcomes. Manipulation is
not possible, however, without knowledge of others preferences. Thus,
we arrive at the contradiction: manipulation is possible only if prefer-
ences are known; but if manipulation is possible, then preferences are
unknown. If p then q; if q then not-p; one or another of these propo-
sitions must go, and it must be the second one: the statement that if
manipulation is possible, then preferences are unknown must be false.
Therefore, the second half of Rikers argument, that harmful manipula-
tions are necessarily impossible to detect, fails. Only the rst half of his
argument remains: that harmful manipulations are frequent, an empirical
assertion.
Strategic voting
Recall that strategic voting (also known as sophisticated voting) comes
about when it is advantageous for one not to vote for ones sincere
preference, as illustrated by the minoritys maneuver in the exercise
above. Another example: in a plurality runoff system, one might vote
against ones rst choice and for ones second choice in the primary elec-
tion as the best candidate to defeat ones third choice in the general
election: dont throw away your vote. Recall that there is a connection
between Arrows theorem and the possibility of strategic voting. Say that
a vote is sincere if it represents the voters sincere preference, and a vote
is strategic if it misrepresents the voters sincere preference. If Deborah
and all remaining voters vote sincerely, we label the outcome the sincere
outcome (in this context, insincerity should be read without pejorative con-
notations). If Deborah votes strategically, and all remaining voters vote
sincerely, we call the outcome the strategic outcome. A voting procedure is
strategy-proof if there is no strategic outcome that Deborah prefers to the
sincere outcome. Arrows possibility theorem showed that any nondicta-
torial social-choice procedure for three or more alternatives must violate
the condition of the independence of irrelevant alternatives (IIA(A)). If
a procedure is strategy-proof then it satises IIA(A), and if a procedure
satises IIA(A) then it is strategy-proof. This is the theorem of Gibbard
and Satterthwaite (as paraphrased by Hinich and Munger 1997, 165):
No voting rule that can predictably choose one outcome from many
alternatives is strategy-proof unless it is dictatorial. (A voting lottery
such that each voter marks a ballot and then one ballot is chosen ran-
domly is strategy-proof, here there is no reason for Deborah to mark
her ballot insincerely, thus the GibbardSatterthwaite theorem is lim-
ited to predictable voting rules.) Therefore, nearly all rules of interest
for three or more alternatives are susceptible to strategic voting. One im-
mediate qualication: again, this susceptibility is a logical possibility, the
theorem does not show that the susceptibility is an empirical probabil-
ity. Riker (1982, 280) notes that if there is enough similarity in voters
preference orders, many voting systems are strategy proof, but does not
develop the point. I conjecture that again it would be a matter of prob-
ability, the more similar are preferences, the less the susceptibility. We
do know for sure that if the distribution of preference orders is such
that they are single-peaked, the GibbardSatterthwaite theorem does not
apply, there is no chance for strategic voting to succeed (Dryzek and
List 2003).
So far, we have assumed that only Deborah is capable of strategic
voting and thereby bringing about the strategic outcome that otherwise
would not have prevailed. What happens if other voters possess strategic
capacity as well? If all voters vote strategically, and if preferences are
separable (that is, if a voters reaction to changing levels of one alternative
is independent of the expected level of any other alternative, Hinich and
Munger 1997, 243), then the outcome is just the same as if all voters
vote sincerely. In these circumstances, sophisticated votes cancel each
other out in a manner of speaking, and our true and fair outcome is thus
restored. The incidence of manipulation could be universal under these
conditions, but the incidence of harmful manipulation would be zero.
This is not too surprising; after all, a model that grants strategic capac-
ity only to one actor violates the democratic ideal of voter equality. The
model gives one actor an unfair advantage and then, in a manner rem-
iniscent of Captain Renaults discovery of gambling in Ricks nightclub
in the movie Casablanca, is shocked to discover that an unfair advantage
leads to an unfair outcome.
How important is strategic voting? Well look at two aspects of the
question: voters selecting among candidates, and members of a commit-
tee selecting among alternatives. I think it is uncontroversial that voters,
for example in American primary elections, engage in strategic voting.
That voters have this capacity constrains the entry of candidates in such a
plurality election (if voters wont waste votes on losers, then it is unlikely
that more than two strong candidates will enter), and thereby reduces
the incidence of strategic voting. Cox (1997, 37148) has shown empir-
ically the incidence in particular circumstances of strategic voting under
a variety of voting systems. Again, potential candidates anticipation of
strategic voting should deter entry and thereby reduce the incidence of
strategic voting among voters. Another sort of constraint on strategic vot-
ing is that it requires a lot of information; undoubtedly, sometimes there
is both enough information and the requisite circumstances for voters to
vote strategically, but that such information is available defeats Rikers
argument that manipulation makes preferences unknowable. Although
hypothetical examples of harmful manipulation are available, I know of
no sustained argument, theoretical or empirical, that the potential for
strategic voting among voters actually leads to irremediable, frequent,
and harmful outcomes.
Now suppose a committee using standard parliamentary amendment
practice (or any other binary agenda procedure), whose members have
complete information including knowledge of each others preferences,
and where members are free to offer amendments. Austen-Smith (1987)
showed that in these circumstances, and whether or not there is a
Condorcet winner, strategic voting and sincere voting are observation-
ally equivalent. If all voters are strategic their votes are the same as if they
were all sincere. That is because those offering proposals would antici-
pate strategic voting and would only offer those proposals that would beat
previous proposals under sincere voting. Under these circumstances, an
empirical survey would disclose zero instances of strategic voting. Nor
would the strategic outcomes differ at all from the sincere outcomes,
and thus the incidence of harmful manipulations would be zero. To draw
out an implication, only if some actors are unfairly constrained from
responding strategically might an unfair decision come about.
Logrolling
Riker briey treats vote trading as another opportunity for strategic vot-
ing with the potential for harmful manipulations, and my rejoinder will
be brief. Recall the example I provided as a normative objection to the
Condorcet criterion: the Red Party of 1,001 voters favor A > B > C
and the Green Party of 1,000 voters favor B > C > A. Pairwise com-
parison gives the contest to A, the candidate ranked last by the Green
Party who make up almost half the population. If we are operating by
pairwise-comparison majority-rule voting this is an ideal opportunity for
vote trading. Those in the Red Party of 1,001 voters prefer A to B,
some fanatically, some just barely; and all those in the Green Party of
1,000 voters rank A as the worst of all the alternatives. A few of those in
the Greens should go to a few individuals in the Reds who least intensely
prefer A to B (the cheapest votes, as it were), and propose a trade: you
Reds who mildly prefer Ato B, vote for Bso we Greens can avoid our last-
ranked candidate, and well owe you for it. When the tables are turned
and there is some Xyou intensely prefer to a Ywhile we only barely prefer
Y to X, well return the favor. Pairwise comparison ignores information
about the relative intensity of preferences, so carrying debits and credits
between issues brings intensity back into the decision system. It would
seem that vote trading makes everyone better off, is welfare-enhancing
(Coleman 1990).
So it would seem, but by varying behavioral assumptions we can craft
one model that shows logrolling to be welfare-enhancing and another
model that shows logrolling to be welfare-reducing. Riker chooses the
assumptions that portray logrolling as welfare-reducing. Since the theo-
retical models cant settle the controversy, we can only rely on empirical
evidence. The empirical evidence is not weighty, but what there is goes
more towards the proposition that vote trading is welfare-enhancing.
1
Additionally, the experimental evidence, according to Muellers (1989,
94) summary, shows that logrolling does not lead to welfare-reducing
outcomes, and generally that the outcomes from committees that use
majority rule tend in practice to be more stable than the theoretical
literature on majority rule leads one to expect.
How might vote trading be welfare-reducing? In ve ways. First, be-
fore vote trading we might have two independent binary votes, one pit-
ting X against Q and another pitting Y against Q. Binary votes are
not subject to cycling. If the issues are joined by vote trading, how-
ever, that yields four alternatives for each voter: XY, QY, XQ, QQ.
Cycling is possible with more than two alternatives, and vote trading
created four alternatives where before there were two, so vote trading
introduces the possibility of cycling where there was none before. As I
always must add, this is a possibility, not a probability. A cycle either
generates costly instability or permits a monopoly agenda-setter to dic-
tate the outcome.
2
Second, if vote traders renege on their commitments
then there is instability as voters drift from one doomed coalition to
another. Third, if to enhance their bargaining power some voters ex-
aggerate their support or opposition on an issue so as to claim more
credits than they are due then that might decrease welfare. Fourth, it
is possible for a majority bloc to vote trade among its own members in
a fashion that benets them but imposes greater costs on the minority
not included in the logroll. Fifth, a linked series of such trades could
end up making everyone in the legislature worse off as each individual
ends up the minority victim of a different externality-imposing majority
(Riker 1982, 161165). In conclusion, vote-trading enormously ex-
pands the potential for strategic voting in legislatures. He acknowledges
that vote trading can enhance welfare even for a majority, but adds
that possibly more frequently in the real world, cases also exist in which
vote-trading makes more people worse off than are made better off.
The argument is insulated by opaque qualiers. Vote trading expands
the potential for strategic voting, but does it actually increase the inci-
dence of strategic voting, and if so, would any harm result? Logrolling
is on net welfare-reducing, possibly; which means that it is also possibly
welfare-enhancing.
The rst and second objections smuggle in an exogenous assumption,
that instability is welfare-reducing. Endogenizing the costs of instabil-
ity changes the incentives for the actors; if cycling or unstable coalitions
shrink the pie then at some point it becomes worthwhile to stick with one
alternative or coalition over another, and if instability doesnt shrink the
pie then instability is not a problem. The rst and second objections alter-
natively rely on the assumption that a monopoly agenda-setter unfairly
imposes an outcome (to be criticized in the next section). The second
and third objections assume that legislators will be happy to cheat and
lie on all occasions and that there are no cheap institutional remedies to
control such naked sociopathy. The fourth objection has the most bite,
and surely there are welfare-reducing logrolls such as perhaps the Smoot-
Hawley tariff or the American military budget. What governmental action
is welfare-enhancing and what welfare-reducing is controversial, how-
ever, and judgment is often colored by ideological prejudices. America
certainly seems greatly undersupplied with public goods compared to
Western Europe. Were the following governmental actions net enhance-
ments or net reductions of welfare: the controversial establishment of
public education in the early nineteenth century, the tariff, rivers and
harbors legislation, freeing the slaves, railroad land grants, the national
park system, the GI bill, the defense establishment, and its spinoffs such
as the internet? The fth objection assumes that voters will myopically
march to disaster. There is a social dilemma aspect to the second through
fth objections. The immediate incentive for each individual actor is to
cheat, lie, and beggar his neighbor such that everyone is cheated, lied to,
and beggared in the end. We observe elsewhere that social dilemmas are
sometimes (but not always) resolved, and there are many mechanisms
by which this may come about (see Ostrom 1990; Lichbach 1996). Per-
haps the least controversial mechanism to suggest in this setting is that
a repeated social-dilemma supergame permits an equilibrium where no
one cheats, lies, or beggars his neighbor. It seems that among legislatures
facing the same kinds of temptations, some succeed at self-regulation and
some fail.
Koford (1982, as related by Stratmann 1997), makes assumptions that
produce a result more congenial to my taste, but I must emphasize that
as far as I know the behavioral realism of his assumptions is not estab-
lished. In Kofords model legislators trade through party leaders who
act as clearing houses of credits and debits. Such leaders are in a po-
sition to prevent cheating and lying and thereby deliver efcient vote
trading. Competition for leadership posts is over who is best at maximiz-
ing welfare from trades. Leaders from different parties form a collusive
duopoly that maximizes joint gain and then struggles over division of
the gain. The majority party colludes with the minority party; otherwise
the minority has nothing better to do than to destabilize the majority.
Leaders only pass bills whose benets exceed the costs, and leaders se-
lect those rank-and-lers who are cheapest to obtain for passage of a
measure.
Stratmann (1997) states that endless cycles are not observed in the
real world of legislatures. Outcomes do not differ much from one ses-
sion to the next, and differences are attributable to the replacement of
legislators. Once legislation is enacted it is usually many years before it
is revisited. Budget allocations do not lurch from one activity to another
from year to year. Cycling dogma would predict that there is usually
a majority that would defeat an incumbent legislator in her home dis-
trict, yet the longevity of incumbents is notorious. As for vote trading,
there is evidence of logrolling and of stable logrolling coalitions in the
US Congress. There is evidence of intense minorities winning over mild
majorities, and there is evidence that those who are least opposed to
the measure are most likely to switch votes, implying the occurrence of
vote trading. Moreover, to date, empirical ndings appear to point to
stable coalitions . . . Further, reciprocity in vote trades has been found,
indicating no widespread reneging . . . So far, the scarce evidence on vote
trading appears to be consistent with the hypothesis that leaders efciently
organize vote trades (Stratmann 1997, 340). The issues are not settled,
but there is no evidence for the Rikerian proposition that logrolling per-
mits manipulations so frequent, irremediable, and harmful as to render
democracy meaningless.
Agenda control
While strategic voting is usually unobservable, agenda control is often
obvious and its incidence can in principle be estimated, Riker continues,
and its incidence is surely very high, for the consequences of agenda
control are apparent in some degree in the content of almost all social
choice (emphasis added, 1982, 169). Riker reckons that there are two
types of agenda manipulation, one exercised by leaders in controlling the
agenda, the other exercised by the capacity of nonleaders to introduce
new alternatives for consideration. This section concentrates on the rst
type, the power of leaders to add, subtract, and order consideration of
alternatives. Although it is identiable and supposedly ubiquitous, Riker
makes no attempt to estimate the frequency of agenda control in any
actual political arena; not even one anecdote of harmful actual agenda
control of the rst type is reported. If the agenda-setter story were valid,
say Brennan and Lomasky (1993, 45), then democratic politicians would
directly appropriate a signicant portion of the public budget. They
would be the richest members of the community, but this is not what we
observe.
Riker sharply qualies his argument. Leaders, he concedes, do not have
unconstrained power. Leaders may nominate candidates, but nonlead-
ers vote on them. Typically, the rank and le can challenge rulings of
the chair. Leaders are elected because of their adherence to local norms
of fairness, and most comply with those norms. Chairs recognize oppo-
nents, accept proposals for consideration that they oppose, and so forth.
Therefore, leaders control is seldom challenged, but there is a subtler
reason for the observation: to challenge the procedure is to challenge
the leadership and thus leaders have some, though constrained control
over the agenda. To draw out Riker, presumably, nonleaders would tol-
erate such control either if it were more welfare-enhancing for the body
than some next-best institutional alternative or if the manipulated issue
was of major importance to the leader but of minor importance to most
other folks. This would not cohere, however, with Rikers other claims
that logrolling is welfare-reducing (1982, 167) or that contrivance of cy-
cles and manipulation by agenda control emerges on the most important
issues (1982, 122). Either agenda control is harmfully ubiquitous, or it
is trivial; it cannot be both.
Assume, however, an unconstrained monopoly agenda-setter, other-
wise we have no bone on which to gnaw. It is just the Captain Renault
demonstration again: one actor is quietly given an unfair advantage and
then loudly discovered to have unfairly distorted the outcome. We are
shocked to discover that someone with dictatorial power attains a dicta-
torial result. Lets consider the one-dimensional case before going on to
the multidimensional case. Suppose that the distribution of preferences
is such that at one end a voter wants the status quo of 0 schools, the me-
dian voter wants 50 schools, at the other end a voter wants 100 schools,
and the agenda-setter prefers 10 schools. The agenda-setter proposes her
preference for 10 schools and this passes since it wins the vote of every-
one who wants 10 or more schools, even though everyone who wants
more than 10 schools is damaged by the agenda-setters unfair exercise
of control. When presented in its simple one-dimensional version the
agenda-control story is quite feeble. We immediately recognize either that
an unconstrained agenda-setter has undemocratic power or that if such
an ofcial were herself constrained by threat of reelection or by appeal
of the ruling of the chair or by some similar device then she would be
constrained to permit the median proposal. The multidimensional ver-
sion is recondite, and so it is not so immediately obvious that the same
objections hold, but they do.
If there is a cycle, and if there is an actor with unconstrained monopoly
agenda power, then the agenda-setter can select the outcome by ordering
the sequence of pairwise voting. Suppose there is a cycle A > B > C > A
and the agenda-setter favors alternative A. Then the agenda-setter pits
B against C which B wins, and then B against A which A wins; each
alternative has been voted on and A is the winner. If the agenda-setter
favors alternative B, then she rst pits Cagainst Awhich Cwins, and then
B against C which B wins; and so on. If the cycles were of what I termed
the balanced type, such that the numbers who rank each alternative rst
are equal, then the agenda-setter has simply broken a tie. If there are
a series of ties of various sorts and the leader settles them always in an
unfair way then she will face the consequences among those who elected
her. If, as is much more likely, the preferences are such so as to be a cy-
cle by pairwise comparison but are determinately ordered by the Borda
count, or by some intuitive apprehension of the intensities of preference,
then the agenda-setter would again have to face the consequences. Sup-
pose that only the agenda-setter ranks C rst and C > A > B, that ten
others rank C second and B > C > A, and that ten others rank C last
and A > B > C. That prole yields the pairwise-comparison matrix
Table 7.2. Unfair agenda setter
A B C (Borda)
A 11 10 (21)
B 10 20 (30)
C 11 1 (12)
in Table 7.2. By pairwise comparison, A > B, B > C, C > A, or
A > B > C > A, a cycle. The unconstrained agenda-setter sets A against
B, which A wins, and then C against A which C wins. The ranking by
Borda count, by YoungKemeny, and by intuitive apprehension of in-
tensities, is B > A > C. The agenda-setter has thwarted the victory of
her least-favored alternative, but the otherwise most popular, B in order
to enact her most-favored alternative, the otherwise unpopular C. As-
sume for the moment that the voters have no counterstrategy against the
agenda-setter. If the agenda-setter were exogenously imposed, then the
voters would have no choice but to acquiesce to her tyranny. If the agenda-
setter faced reelection to ofce by the voters, then she would be deterred
from cycling mischief. The voters do have a counterstrategy, however,
which deters even the unconstrained agenda-setter: what the ten voters
with the ranking A > B > C need to do when the agenda-setter proposes
A against B is to nd at least one among their ranks to strategically
vote for B. The strategic voters choice of their last-ranked alternative
helps B win the contest against A; then when B goes against C, B wins
by 20 votes to 1. Anticipation of voters strategic response would moti-
vate the agenda-setter to refrain from wasting time on a maneuver that
would fail. Harmful outcomes can be concocted if we assume an asym-
metry: that one voter controls the agenda and others do not, that some
voters know others preferences and some do not, or that some voters
have strategic capacity and some do not, and the like. Formally, to make
some actors unequal and then deplore the unequal outcome amounts to
the weeping of crocodile tears. Practically, if there are unjustied asym-
metries they can and should be institutionally remedied. By far the worst
asymmetry in American politics, say, as compared to Europe, arises from
the ineffective regulation of campaign contributions from interested par-
ties. Those with a passionate concern for conserving democratic equal-
ity of voice might lend their talents to the cause of campaign-nance
reform.
Riker (1982, 173181) provides two pieces of empirical evidence for
his hypothesis of the universality of agenda control. One is from Pliny
the Younger, acting as chair in the Roman Senate on a case deciding an
alleged crime. There were three factions in the Senate: one favored A, ac-
quittal, one favored B, banishment, and one favored C, condemnation to
death. If Pliny had offered the customary agenda, guilt against innocence,
then factions B and C would have voted together and defeated A; then in
the choice between B and C banishment would have won. Pliny favored
A and as agenda-setter put the three alternatives simultaneously because
A had the plurality of votes. The C voters, however, responded by strate-
gically voting for B, the choice of the median voter, which won. Agenda
control was foiled by strategic voting. The manipulative attempt failed.
The second is from the contemporary laboratory. Plott and Levine
(1978, as related by Riker 1982, 174181) induced in subjects prefer-
ences over ve alternatives by cash reward. Each experimental group had
the identical distribution of preferences but faced a different agenda se-
quence. Each subject was informed only of her own preferences, not of
others; and subjects were encouraged to discuss the decision but forbid-
den from disclosing information about their payoffs. From pretests Plott
and Levine devised a behavioral formula that predicted how subjects
would choose in such circumstances; they then ran the main test and dis-
covered that their predictions were conrmed, with one exception out of
ve trials. If the agenda-manipulator knows everyones preferences, and
possesses the behavioral formula, and subjects are kept ignorant about
one anothers interests (note the requisite asymmetry), then the agenda-
setter can indeed bring about any outcome. The exception is of major
signicance, however. In one of the ve groups the chair permitted a straw
vote early in the discussion, thereby disclosing preferences, and this led
the subjects away from Plott and Levines manipulative outcome and to
the Condorcet (and Borda) winner. Riker says that he replicated the ex-
periment, and again one out of four groups escaped Rikers manipulation
by recourse to a straw vote. The experiment was successful enough to
show fairly conclusively that conscious manipulation could change out-
comes, Riker comments (emphasis added, 1982, 175). Another way of
describing the implications of the result, however, is that agenda control
is possible only if two conditions apply: the agenda-setter has uncon-
strained power over the sequence of consideration and the agenda-setter
has unconstrained power over the distribution of information about pref-
erences. What has been shown is that birds cant y if you chop off their
wings.
Rikers theory is one of pervasive disequilibrium: An equilibrium of
tastes and values is in theory so rare as to be almost nonexistent. And I
believe it is equally rare in practice (1982, 190). So why dont we observe
pervasive disequilibrium? Individuals . . . are constrained by institutions
that are intended to induce regularity in society. And it is the triumph of
constraints over individual values that generates the stability we observe
(190). Institutions are responsible for stability, and prominent among
those institutions is agenda control, an observable institution of very
high incidence. Rikers colleagues wrote stylized models of agenda con-
trol in the US Congress that attempted to demonstrate this point. The
stylizations, however, did not capture Congressional reality, according
to the summary by Strom (1990, 8391).
3
There were four proposed
institutions of agenda control.
First, it was alleged that Congressional committees enjoy a gatekeeping
power which allows them to prevent legislation from being considered in
full chambers. There are rules, however, which permit the full chamber
to discharge a committee from consideration of a bill. Successful dis-
charge votes are quite rare, but game-theoretic insight suggests that this
would only mean that the discharge rule is an effective deterrent to com-
mittee attempts to defy majority will. Most bills die in committee, but
that reects their original lack of majority support in the full chamber.
The second form of agenda control is the so-called closed rule in the
House, which prohibits amendments on a bill reported by committee,
which would permit the committee to bundle the issues according to its
own predilections. The closed rule is seldom invoked, however, for ex-
ample only once over the six years of the 97th through 99th Congresses.
The Senate does not permit the closed rule, yet there is no difference
in legislative success between House and Senate committees. The third
form of agenda control is the closed rule on bills reported from a confer-
ence committee. By one measure 86 percent of bills were passed with no
conference, however; what happens is that one chamber accedes to the
other chambers version, or amended versions are sent back and forth.
The fourth is that it is costly for members to formulate and propose
alternatives, and thus the specialists serving on the committee have an
advantage on issues within the committees jurisdiction. Evidence for this
proposition is that committee bills have great success on the oor, and
half of oor amendments come from committee members. A more plau-
sible interpretation of this evidence, however, is that committees craft
only legislation that will obtain majority support in the full chamber. It
would thus appear that agenda explanations are not very successful or
fundamental in accounting for the . . . stability of legislative outcomes
(Strom 1990, 91).
Riker believes that in a multidimensional issue space there is no
majority-rule equilibrium and thus that agenda control arbitrarily de-
termines the outcome of the voting process (his interpretation of the
McKelvey and Schoeld chaos theorems, the topic of the next chapter).
He concludes (1982, 237) that empirical evidence of a concern over
agenda content demonstrates the presence of disequilibrium that enables
manipulation and its arbitrary outcomes:
Since we also know from a vast amount of conventional analysis of political in-
stitutions that much political dispute concerns control (and presumably, manip-
ulation) of the agenda, we can be fairly certain that this kind of manipulation is
utterly commonplace.
Riker (1993, v) edited and introduced a volume on agenda formation.
Agendas foreshadow outcomes: the shape of an agenda inuences the
choices made from it, he begins (1993, 1), making agendas seems just
about as signicant as actually passing legislation. Empirical evidence
of concern over agenda content does not demonstrate disequilibrium,
however. There is an alternative hypothesis which is much more plausi-
ble. Political deliberation takes place in real time, which is cruelly nite.
There are multitudes of potential political issues, but at any one session
there is time to consider only a few, we hope the most panoptic and the
most urgent. At the margin of public attention there is sure to be a sharp
contest as to which issues deserve to be on a time-limited agenda. Imag-
ine that I propose the Mackie theorem, which postulates one actor with
an unconstrained monopoly capacity for persuasion. In my model the
persuader can change any other actors preferences, but the other actors
cant change the preferences of the persuader. Therefore, on any issue,
the persuader can determine the collective outcome, proving the arbi-
trariness of democracy. Evidence of the truth of the Mackie theorem is
the fact that people in legislative assemblies devote a great deal of time
to persuasion. Have I made my case?
Summary
Strategic voting, logrolling, and agenda control permit the manipulation
of outcomes, according to Riker, and it is impossible to distinguish ma-
nipulated fromunmanipulated outcomes because of the unknowability of
preferences. The models of manipulation, however, require that manipu-
lators have full knowledge of preferences, in contradiction to Rikers claim
that the possibility of manipulation makes preferences unknowable. As
for strategic voting, if all voters are strategic then their votes are the same
as if they were all sincere. Whether or not logrolling is welfare-enhancing
or welfare-reducing depends on behavioral assumptions; and limited ev-
idence leans towards the conclusion that it is welfare-enhancing. Riker
believes that agenda control is ubiquitous, although he concedes that
the unconstrained monopoly agenda-setter in his models is not found
in reality. An unconstrained monopoly agenda-setter might have unfair
power over outcomes, but agenda-setters are constrained, and further
voters can defeat agenda control with strategic voting. Empirical inci-
dence of agenda control has been much exaggerated. The next chapter
continues the study of opportunities for manipulation, in multidimen-
sional issue spaces.
8 Multidimensional chaos
Chaos in multidimensional issue spaces
Black (1958) showed that if alternatives can be represented as points
along one line and if voters preferences are single-peaked, indicating
resemblance, then a majority-rule equilibrium results. The position of
the median voter on the line will beat any other alternative in majority-
rule voting. This is normatively attractive because a central alternative
prevails. In Figure 8.1 there are ve voters with preferences over alterna-
tives A, B, C, D, and E. Each voters preference curve has only a single
peak. Voter 3 has the median preference C, and C will beat by majority
vote any alternative that it faces. For example, if D is pitted against C,
voters 1 and 2 prefer D > C, voters 3, 4, and 5 prefer C > D, and thus C
wins by majority vote. There are no cycles. The preference orders need
not be so neat as portrayed in the gure; each only needs to be single-
peaked.
Figure 8.2 shows a prole of preferences that is not single-peaked be-
cause one of the voters rankings, in this portrayal #2, has two peaks
(single-peakedness and its lack are impervious to rearrangements of the
labels). Recall that given three alternatives, there are six possible strong-
preference rankings, and that given three voters, one each with cyclical
rankings 1 (A > B > C), 3 (C > A > B), and 5 (B > C > A), or with
rankings 2, 4, and 6 together, the result of majority voting is inconsistent,
that is, A beats B, B beats C, and C beats A. The non-single-peakedness
illustrated in the gure is just a portrayal of a prole with cyclical rankings
1, 3, and 5. Single-peaked preferences are not cyclical, and non-single-
peaked preferences are cyclical.
If the relevant political issue space is in two or more dimensions, how-
ever, and given a number of other usual assumptions, then there is no
equilibrium: the majority-rule outcome can be anywhere in the issue
space, anything can happen, as Riker says frequently. This is Rikers
interpretation of the so-called chaos theorems of McKelvey (1976) and
Schoeld (1978). The chaos theorems are interesting and impressive
173
U
t
i
l
i
t
y
A B C D E
1
2
3
4
5
Figure 8.1 Single-peaked
U
t
i
l
i
t
y
A B C
1
2
3
Figure 8.2 Non-single-peaked
exercises in the abstract. Instructors are fond of demonstrating them
on the blackboard because of their gee-whiz qualities. The theorems,
however, are normatively tangential, their predictions are falsied, and
standing alone they are empirically irrelevant for understanding the world
of politics. The theorems are, of course, logically true, given their ini-
tial premises, but the theorems fail as models. They are not simplifying
Multidimensional chaos 175
abstractions from reality but are rather befuddling negations of reality;
they fail as models because their premises are unrealistic and because they
abstract away fromessentials. Schoeld (1995) offers the metaphor of the
frictionless oscillator, such as a spherical pendulum. In the absence of fric-
tion most initial states result in nonstationary orbits or cycles that would
continue forever in disequilibrium. But with the tiniest epsilon of fric-
tion all initial states lead to a stationary outcome in equilibrium. Whats
missing in the early models is the metaphorical friction that obtains in
real-world circumstances. It is a mistake to argue that the counterfactual
world of no friction somehowreveals a more fundamental truth about the
world of friction. What we have is one set of assumptions with disequi-
librium consequences and another set of assumptions with equilibrium
consequences; the former is not a deeper version of the latter. If there is
to be any hierarchy among the theorems it must be one that privileges
the more realistic assumptions, and those happen to tend towards equi-
librium. Advice based on the counterfactual is absurd; it would be as if I
said I am going to the store to buy some milk and my interlocutor fresh
from his physics lesson warned me that in the absence of friction I would
not be able to proceed because my feet could not grip the ground.
The points I amabout to rehearse are old news to everyone who follows
this literature. Several writers in the Rochester tradition are quite scrupu-
lous about resisting Rikers interpretations and emphasizing the quali-
cations attendant to various results.
1
McKelvey (1986) and Schoeld
(1995) themselves have devoted later work to models of democratic pro-
cesses tending to stable outcomes around the center of the distribution
of preferences. Nevertheless, Rikers interpretations remain strangely in-
uential, perhaps because of his memorable examples, and the chaos
theorems are still widely promulgated as if they revealed a deeper truth:
that apparent representativeness and stability is but an illusion conceal-
ing perpetual arbitrariness and instability lurking underneath the surface
of democratic politics. There is always an intense struggle, beneath
the apparent stability, to induce a genuine disequilibrium of tastes
(Riker 1982, 190).
If the distribution of voters ideal points is pairwise symmetric then
there is a majority-rule equilibrium in two or more dimensions. Imag-
ine ve voters in a two-dimensional issue space, with voter ideal points
distributed in an X-pattern, with four voters at the vertices of the X and
one voter at the intersection of the two lines making up the X. The equi-
librium point is at the ideal point of the voter inhabiting the intersection.
This equilibrium is most fragile, if any one of the ve voters changes
position ever so slightly the equilibrium vanishes. The Plott pairwise-
symmetry condition is sufcient for a majority-rule equilibrium, but it is
not necessary. The more general condition is due to Enelow and Hinich
(Hinich and Munger 1997, 65), that a point is a Condorcet winner if it is
the median in all directions: if every line drawn through the point divides
the ideal points of all voters so that at least half are on either side of the
line, including voters whose ideal points are on the line in both groups.
In one dimension the median is inevitable, but in more than one dimen-
sion the Condorcet winner would almost never exist and would be fragile
if it did.
McKelvey showed that, if the decisions are made by pairwise majority
rule, if the distribution of voters preferences falls in two or more dimen-
sions, if as is almost certainly the case in two or more dimensions there
is no Condorcet winner, and if voters are not strategic, among other as-
sumptions explicit and implicit, then there exists an amendment agenda
that would lead from any one point in the issue space to another and
back again. In other words, with repeated votes the majority-rule voting
outcome is entirely indeterminate. Again, an unconstrained monopoly
agenda-setter could determine the outcome by choice of agenda
sequence.
Normatively tangential
This does not mean in a multidimensional issue space that there is no
normatively attractive point. Say that there is a dimension of concern X
ranging from 0 to 100 and voters with a median preference, 42, along
single dimension X, and that there is a second dimension Y on which the
same voters have a median preference of 53 and a third dimension Zwith a
median of 62. The intersection of these medians in the three-dimensional
space is the point (42, 53, 62). The problem is that with frictionless
pairwise majority-rule voting neither that normatively attractive point nor
any other is in equilibrium; there exist sequences of votes such that every
point could defeat every other. The theoretical instability of majority-
rule voting in multiple dimensions can be seen by imagining three voters,
A, B, and C, in two-dimensional space, each with a bliss point surrounded
by circular indifference contours, as in Figure 8.3. Say that the status
quo happens to be the normatively attractive point, the intersection of
medians. The circular indifference contours intersecting through that
point generate three petals which dene a so-called win-set, denoted
W(X
med
) in the diagram, and all the points in each of the petals is preferred
by two out of the three voters to the status quo that we started from.
Choose any arbitrary point P in W(X
med
), and fromthere three newpetals
form, W(P ). And so on. An agenda-setter can move the outcome through
the issue space, and even outside the Pareto set.
X
B
X
C
X
A
X
med
: Intersection of medians
W

(
X
m
e
d
)
W (X
med)
Pareto set
Project 1 budget (millions of $)
Project 2 budget
(millions of $)
Figure 8.3 Win-sets of intersection of medians
Source: Hinich and Munger (1997, 63)
Suppose we have something so simple as the division of the question
rule, that a motion is in order to divide a bundled proposal into parts,
or a germaneness rule, that an amendment to a proposal must be ger-
mane to the proposal. In many parliamentary rulebooks, any one member
may force division of the question. If issues do inhabit a multidimensional
space then these rules constrain voting to one dimension at a time voting
rst along dimension Xand next along dimension Yand so on and there
is no chaos (Strom 1990, 98113). The sequence of votes will uniquely
select the intersection of medians. This statement requires several qual-
ications. Recall that preferences are separable if the voters preference
along one dimension does not depend on her preference along another
dimension; and are nonseparable if there is such dependence. If prefer-
ences are separable, and if voters are either sincere or strategic, then
dimension by dimension voting will select the intersection of medians as
the equilibrium choice. If preferences are nonseparable, then dimension
by dimension voting will necessarily select that equilibrium, unless voters
are sophisticated. Finally, if, rather than the forward-moving agenda (de-
ned shortly below) we have tacitly assumed, we instead assume the more
realistic backward-moving agenda, then again there is no equilibrium if
voters are sophisticated (Strom 1990, 104). To continue, dimension by
dimension voting is an exemplar of structure-induced equilibrium, the idea
of Rikers students that a number of political institutions function to force
stability upon chaos. Further examples of structure-inducing devices are
the four agenda-control institutions in the US Congress discussed in the
last chapter. One objection to the story about division of the question
and germaneness is that the motions these rules permit are almost never
observed and thus they are irrelevant. If, however, there exist rules that
can induce an equilibrium at the intersection of medians, yet these rules
are rarely invoked, then one alternative explanation would be that there
is rarely a practical disequilibrium for the rules to subdue, and another
alternative explanation would be that the threat of invoking the rule is
sufcient to deter those who would otherwise join dimensions.
Experimental and empirical failures
Riker seeks to avoid the empirical problem of observed stability by
means of his basic argument pattern, to the effect that the possibility
of cycling, agenda control, and multidimensional issue spaces makes it
impossible to know preferences, which was criticized in Chapter 2. If
natural empirical observations are controversial because of the obscurity
or incompleteness of preference data, then we must turn to experimental
settings where preferences are known because they are induced by cash
reward. Typically, a number of human subjects are recruited for an
experimental game in a laboratory setting with nonexperimental factors
controlled to the maximum extent feasible. For example, subjects might
be assigned different ideal points in a two-dimensional space and then
be rewarded with more money the closer the group voting outcome is
to their ideal point; or subjects might be candidates who are rewarded if
their platform is chosen by the majority of articial subjects assigned to
ideal points; subjects might be forbidden fromdisclosing their ideal point
or otherwise communicating so as to x nonexperimental variables.
Subjects will then vote under various experimental conditions, such as a
distribution of preferences with an equilibrium compared to one without
an equilibrium, with an agenda-controller compared to without, with
issue-by-issue voting (division of the question rule) compared to without,
with low payoffs compared to high payoffs, with a low number of voters
compared to a high number of voters, and so on, and in various combina-
tions. Green and Shapiro (1994) devote a useful chapter to Legislative
Behavior and the Paradox of Voting on the cycling question; and their
discussion of the experimental evidence (120146) is especially splendid
and indispensable, although too stern for my tastes. It would be redun-
dant to summarize all their reections, but I do want to borrow a few
points.
If the theories of equilibria and disequilibria in an issue space are true
or useful, and if experiments are sufciently controlled so as to be inter-
nally valid, then when there is an equilibrium or a core the voting process
should select the equilibrium and when there is not an equilibrium or
no core then the voting process should select any outcome in the space.
There are two problems. First, although when there is an equilibrium it
tends to be selected by the voting process, the success rate differs across
different experimental conditions (preference congurations and voting
tasks) in no pattern forecast by theory (Green and Shapiro, 128). In some
experiments the predicted equilibrium is selected less than 50 percent
of the time, in others more than 90 percent of the time; with one team
of experimenters the success rate varied from 33 to 100 percent de-
pending on variations in structure of the game that should not mat-
ter according to theory. Second, theory predicts that in the absence of
an equilibrium different legislatures with identical preference congu-
rations would choose different policies. But, according to Green and
Shapiro (1994, 134):
coreless games do not produce markedly more unstable outcomes than do games
with cores (Fiorina and Plott 1978; Laing and Olmstead 1978), and sometimes
the distinction between the two game forms is hard to detect at all (Laing and
Slotznick 1987, 1991). To be sure, the dispersion of outcomes is typically greater
for noncore games, but the difference is not as striking as the McKelvey result
(1976) and subsequent commentary (Riker 1980[a]) might be taken to imply . . .
If the existence of the core does not have an appreciable effect on the observed
instability in experimental outcomes, then one of the central empirical claims mo-
tivating game-theoretic analyses of legislative behavior the inuence of majority
rule equilibrium on legislative stability receives little support.
McKelvey and Ordeshook (1990, 127) review a decade of experimen-
tal research on the spatial theory of voting, their own work and that of
others, and are more optimistic than Green and Shapiro. Among their
conclusions, however, is that the absence of a core does not imply in-
coherence or chaos patterns to the data warrant explanation. Further
(1990, 138):
With the discovery that Condorcet winners are rare with spatial preference and
that cyclic social preferences can extend across the entire set of feasible alter-
natives (McKelvey 1976, 1979), some scholars were led to the belief that polit-
ical processes are inherently unstable and unpredictable (Riker 1984). Others,
believing that only the possibility of rendering unique predictions is removed
if a games core is empty, and the like, began the development of alternative
equilibrium notions such as the uncovered set and the competitive solution.
Experimental research supports this second view.
Although there are several equilibrium concepts that might account for
the experimental ndings none is clearly superior or widely accepted.
Stroms (1990, 76) textbook remarks that it is currently unclear the
degree to which real legislative decision making is characterized by unsta-
ble chaotic outcomes. There is the additional problem. . . the predicted
chaos is not observed in controlled experiments. Strom summarizes
Fiorina and Plott (1978) who ran three conditions, two with an equilib-
rium and one without. In all conditions outcomes clustered in the center
although were more dispersed in the disequilibrium condition. Fiorina
and Plott did not notice any behavioral differences between equilibrium
and disequilibrium conditions. According to Strom (1990, 74), It al-
most appears as if the subjects were imposing their own equilibrium.
I speculate that although people may have to vote by ordinal pairwise
majority rule, they possess enough knowledge of individual rankings of
alternatives to identify a central point nearby the intersection of medians,
and, if need be, they can use strategic voting to get into that center. Strom
(1990, 75) reports that the experimental results have not been accepted
as sufcient to discredit the theory. There have been few experimental
tests of the theory, he says, and there may yet be auxiliary equilibrium
hypotheses offered that will save it.
The experimental results speak directly to the theory, as the experi-
ments induce distributions of preferences, either in equilibrium or dis-
equilibrium. Strom continues that the apparent lack of agreement be-
tween the theory and natural empirical evidence on legislatures is more
signicant and has more motivated the invention of auxiliary structural
hypotheses to explain stability. I must add that real legislatures may have
nearly unidimensional distributions of preferences that are already in
equilibrium without additional structural constraints. Is observed stabil-
ity in a real legislature due to a unidimensional distribution of preferences
or due to a structural constraint such as the division of the question and
germaneness rules? It cannot be claimed that a structure induces equilib-
riumunless the empirical investigation rst establishes, not assumes, that
the distribution of preferences is in disequilibrium. Further, contrary to
theory, experimental committees in multidimensional space pick stable
central points in the absence of any structure. How then can struc-
ture be offered as an explanation for why a real committee stably chooses
central points? Moreover, there is a bemusing abundance of structural de-
vices, including elements of the constitutional setup, rules of procedure,
and unwritten customs, that are often nonexclusive. Green and Shapiro
(1994, 114120) go so far as to call the devices of structure-induced
equilibrium post hoc accounts of stability. A single empirical legisla-
ture will be structured by dozens of devices. If observed stability in a
legislature is not due to the distribution of preferences, or is not due to
the descriptive failure of the theory, then is it due to one nonexclusive
structural constraint or due to another? Green and Shapiro charge that
there is a tendency to nonfalsiability in the empirical testing of spatial
theory: if the data t the model, then structurally induced equilibriumhas
been demonstrated, but if the data do not t the model then the model
is declared to be stylized and yet untested.
Not only are stability-inducing rules often nonexclusive, worse, Strom
(1990, 105) shows by example that stability-inducing rules are not nec-
essarily additive. The spatial model assumes a forward-moving agenda,
that is, the status quo is the rst issue put to the vote, and in the abstract
this yields outcomes anywhere in the issue space; but legislatures use a
backward-moving agenda, the status quo is the last issue put to the vote,
and in the abstract this yields outcomes not in equilibrium but in the
win-set of the status quo, a region less than the entire issue space. Then
there is the division of the question rule we have already discussed that
yields an outcome in equilibrium. In a legislature of sophisticated voters
with both a backward-moving agenda and a division of the question rule,
however, the outcome is no longer in equilibrium. The conclusion is that
to explain legislative equilibria requires the simultaneous examination
of a whole constellation of procedural rules. This would require an ana-
lytic model of unimaginable complexity, beyond the capacities of our era.
Thus, due to modeling intractability and overdetermination of data, it is
extremely difcult, if not impossible, to test the spatial model in natural
settings.
I think that what Rikers followers call positive political theory enriches
with analytic insight, but will never become a grand-theoretical substi-
tute for, an astute and informed commonsense understanding of politics.
The multidimensional disequilibrium version of spatial theory does not,
so far, survive experimental testing. Even if we ignore experimental fal-
sication, the basic empirical assumption of the enterprise, that the dis-
tribution of preferences is multidimensional rather than unidimensional
is not established. Even if we ignore the evidence that the distribution
of preferences is mostly unidimensional, empirical testing of the theory
in natural empirical settings is extremely difcult. Historically, I think,
much of the motivating passion and prestige of the model is a result of
Rikers charismatic anecdotes of empirical cycling. When it comes time
for empirical illustration of theoretical chaos, the only evidence Strom
(1990, 68) can muster is three of Rikers anecdotes and an article of
Blydenburghs (1971), each of which I will later show to be mistaken.
If those anecdotes do not survive then the spirit of the enterprise is de-
ated. Positive political theory is a worthy academic undertaking, but it
is not entitled to grand normative claims to the effect that democracy is
arbitrary and meaningless or that particular political views are blessed by
political science.
Structured preference orders; alternative voting rules
If the distribution of preferences is single-peaked and unidimensional
then there is no cycling, no disequilibria, no instability. To eliminate
any possibility of cycling, every single preference ordering in the popu-
lation must satisfy the single-peakedness condition (Kramer 1973). For
a period it was believed that this rendered the single-peakedness restric-
tion empirically irrelevant, and that belief lingers: any optimism asso-
ciated with preference restriction theorems was dealt a heavy blow by
Kramer (1973), says Enelow (1997, 155). The simulation by Niemi
(1969), however, which we have already examined, showed, starting from
an impartial culture assumption of equiprobable preference rankings,
that the greater the proportion of preferences that are single-peaked the
more likely is a transitive group ordering, and that the probability of
a Condorcet winner increases with the number of voters. Later, Niemi
(1983) suggested the concept of semi-single-peakedness: a set of pref-
erence curves is semi-single-peaked if they can be arranged so that a
majority of curves slope down to the left away from a given alternative
A and a majority of curves, not necessarily the same ones, slopes down
to the right of that given alternative A. If the set of curves is semi-single-
peaked, then A can beat all other alternatives in pairwise voting; and is
the median voters preference.
The concept of partial-single-peakedness was developed by Feld and
Grofman (1986, 73): small perturbations from the impartial culture as-
sumption will virtually guarantee transitive majority preferences. Here is
their argument. Take the six strong preference orderings over three
alternatives. For any individual preference ordering A > B > C, dene
opposite preference ordering as C > B > A. Dene net preference ordering as
the number of orderings A > B > C minus the number of C > B > A.
Positive net preference orderings are the excess of orderings A > B > C
over orderings C>B>A. The majority decisions of a group are transitive
if and only if, (a) the positive net preferences are single-peaked, or
(b) one positive net preference ordering has a majority of the positive net
preference orderings. Why? There are three different opposite preference
pairings:
1. A> B > C and 4. C > B > A,
2. A> C > B and 5. B > C > A, and
3. C > A> B and 6. B > A> C.
1, 3, and 5 together permit a cycle, so do 2, 4, and 6 together. Opposite
preference orderings cancel each other out, and positive net preference
orderings remain for each of the opposing pairs 1:4, 2:5, 3:6. There is
no cycle if positive net preference orderings are not 1, 3, and 5 together
or 2, 4, and 6 together; in other words, if the positive net preference
is for one of the left side and two of the right side of the opposing pairs
(e.g., 1:4, 2:5, 3:6), or vice versa (e.g., 1:4, 2:5, 3:6), then preferences are
single-peaked, showing (a). (Under impartial culture,
2
3
of orderings of
three alternatives are single-peaked, but for positive net preferences to
be single-peaked, only a mere majority of individual orderings need be
single-peaked). If the left side of each opposing pair (1:4, 2:5, 3:6), or
the right side of each opposing pair (1:4, 2:5, 3:6) has the positive net
preferences, and if one of those orderings has a majority, then it is the
group majority preference, showing (b). The result is extended to more
than three alternatives by reformulating it as a condition on every triple
of alternatives.
Given large numbers the impartial-culture assumption becomes simply
the uniformdistribution of preferences, and then positive net preferences
are zero, Feld and Grofman (1986) continue. Then any small group ep-
silon whose positive net preferences are single-peaked imposes a transi-
tive majority preference ordering identical to its own, so that coherent
ordering prevails over incoherent ordering. Of course, multiple cyclically
opposing small groups would negate that result. But if some larger group
were able to subsume the cyclically conicting coherences of small groups
into a higher-order coherence, stability would prevail again.
Generally, the more similar, even mildly similar, are preference rank-
ings, the less likely is instability. Also, for more dissimilar preference
rankings a larger than 50 percent (supermajority) voting rule can pro-
vide stability. For example, at the extreme, if there were a full cycle, a
consensus rule would be stable at the status quo (although stability at the
status quo may not be normatively desirable). Can we say how big a su-
permajority rule would have to be 51 percent, 67 percent, 95 percent
in order to guarantee an equilibrium given various distributions of pref-
erence rankings? To a certain extent, yes. There are several approaches.
Caplin and Nalebuff (1988; 1991; see also Ma and Weiss 1993) demon-
strate a mean voter theorem. The mean voters preference is unbeatable
under 64 percent majority rule, and no cycles are possible, if the voters
preference rankings are sufciently similar. On their rst go (Caplin and
Nalebuff 1988) they demonstrated stability if the density of voters most
preferred points is a concave function, in other words, if more voters favor
intermediate positions than the average of those favoring extremes. On
their second go, they demonstrated that the 64 percent rule chooses the
mean voters preference for all log-concave densities and for further distri-
butions as well. The large class of distributions with log-concave density
includes the uniform, the normal, the truncated normal, exponential,
and Weibull functions, among others.
The work of McKelvey and Schoeld (1986) as rened by Saari (1997)
relates the dimensionality of the issue space to the size of the supermajor-
ity required for an equilibrium and to the number of voters. Ignore some
untroubling special cases, skip over the distinctions and denitions, and
glance past some stratospheric mathematics. Suppose that there exist
q-rules for n number of voters,
n
2
< q < n, such that
q
n
votes are needed
for a proposal to win, for example, an assembly might require a
2
3
(q = 2,
n =3) vote to change its procedures; and suppose there are k dimensions
in the issue space. Roughly speaking, there is stability if k 2qn. Thus
if there were a hundred voters and a
67
100
rule, then the outcome is stable if
there are less than 2.67100 =34 dimensions in the issue space. In other
words, the maximal dimension of the issue space with stability is about
equal to the number of voters needed to defect from one winning coali-
tion and join another. If 67 cardinals voted for the pope and 33 against,
then 34 of the 67 must defect and join the 33 in the minority in order
to make a new supermajority. Further, for any supermajority rule there
is a large enough number of voters such that there will be stability. This
is one of those half-empty, half-full controversies. One can say that given
a number of voters we can increase the number of dimensions to the
issue space until there is instability, or one can say that given a number
of dimensions to the issue space we can increase the number of voters
until there is stability. We can also dene supermajorities as rules,
1
2
< < 1, such that an proportion of voters is needed to win the
election, for example, 0.67 of the voters. Saari (1997) gives the example
that with a voting rule of as low as 0.5001 there is stability in a hundred
dimensional issue space with around a half million voters.
I must reiterate that all this commotion arises from insistence on Con-
dorcet pairwise majority-rule voting. It is immediate that the Borda count
returns a result at or very near the intersection of medians. If instability
is a major problem, and the evidence is that it is not, then democratic
bodies can resort to voting methods other than the Condorcet Borda,
approval, Hare, YoungKemeny and be spared the grief. If instability
is feared in a constant-sum distribution such as a budget allocation, and
comity is lacking, then institutional designers can propose various fair-
share procedures, perhaps point voting, voting by veto, or probabilistic
majority rule (discussed by Mueller 1996, 160166), or other innovations
that might be devised by theory and recommended by experience in the
future.
Adding back friction
In a two-dimensional issue space with an odd number of voters the
McKelvey agenda-setter can move the outcome from any one point to
another. A few questions: Do we know that the political issue space is
multidimensional? Is the issue space actually continuous, so that ac-
tors would be motivated to choose between alternatives $100,000.01,
$100,000.00 and any proposal in between? An unconstrained monopoly
agenda-setter is an unrealistic assumption and such an institution would
be patently unfair, even in the one-dimensional case as we have seen. Fur-
ther, would an agenda-setter be able to propose a sequence of proposals,
inconsistent with one another and with her past positions, without excit-
ing suspicion or destroying her credibility? Why would the voters, who are
assumed to be sincere, passively let themselves be rolled; is that rational?
What if the number of voters is even? Then there is an equilibrium in
two dimensions, but disequilibrium in three dimensions. Is this a con-
dition with any empirical content, do newspapers report that legislative
observers are worried because there are two dimensions under considera-
tion and an odd number of voters in the chamber? Why do we have to use
ordinal pairwise voting? Of course these questions verge on silliness, but
no more so than the story that democracy is arbitrary and meaningless
because of the McKelvey theorem.
The problem with the chaos theorems is that they are not supported
by experimental and empirical observations of instability. The various
amendments intended to construct models that predict observed stabil-
ity are exceedingly scholastic and are only speculatively connected with
reality. It would go beyond my budget of time and space to present each
of them in a detailed manner, and they are more ably explained by oth-
ers (Miller, Grofman, and Feld 1989; Ordeshook 1986; Schoeld 1995;
Strom 1990, 114126). Because of their tenuous status I shall provide
only the briefest of guides to these ideas.
Kramer (1977) presented a dynamical model of political equilibrium.
Two parties compete in a multidimensional policy space. The parties
compete for voters in an innite series of elections, offering platforms
as points in the space. The voters choose the platform closest to their
respective ideal points; the platform with the most votes wins; followed
by a new election. The trajectory of successive platforms leads to a set
of alternatives closest to being majority winners; these alternatives are
termed equilibria and may be contained in a small region rather than
the entire space as in the McKelvey theorem. Does Kramers elegant
model save us from devastating disequilibrium? Riker (1982, 191) asks.
Unfortunately, I think not because at least three of its assumptions are
extremely unrealistic, he continues. Riker says that the assumption of
two parties is unrealistic and unfair, and that the assumption that par-
ties are motivated by vote maximization is unrealistic. He also complains
that it is unrealistic to assume that dimensions are xed over time. This
objection is founded in Rikers contention that elite political actors can
add dimensions at will, which will be explicated and disputed at a later
point. What is of interest here is Rikers insistence that the criterion for
a models success is its realism. If we accept the realism criterion, then
a generalization follows. If a model is unrealistic, and even more so if a
models predictions are contrary to observations, then the model must be
rejected. The McKelvey theorem is unrealistic and it fails as a predictive
model. Why then is it the foundation of Rikers theory of democracy?
Recall that a point is a Condorcet winner only if every line drawn
through the point divides the ideal points of all voters so that at least
half are on either side of the line, including voters whose ideal points are
on the line in both groups. These lines are termed median lines. There is
inevitably an equilibrium in one dimension, but the distribution of voter
ideal points in multidimensional space will almost never be such that all
median lines intersect at exactly one point. Ideal points will typically be
distributed in a manner, however, such that there will be a small central
region through which all median lines pass; this circular region is termed
the yolk. With an unlikely distribution of ideal points the yolk will be a
single point, that is, the equilibrium point. With other distributions the
yolk may be large, but for many distributions the yolk will be small. The
chaos theorem still holds so far; outcomes can be anywhere inside or
outside the yolk. We shall return to the yolk.
Suppose a point X in multidimensional space. All the alternatives that
are majority-preferred to X we call the win-set, W(X), the petals I men-
tioned earlier.
2
Let Y be an element of W(X), in other words Y is pre-
ferred to X. Now suppose a Z that is an element of W(X) and also an
element of W(Y ). This Y is said to cover X. There is no cycle because Z
beats Y, Y beats X, and Z beats X. Sophisticated voters will not adopt X
over Y if Y is an element of the agenda. If Y covers X it will also be closer
to the center of ideal points than X. This is a chink in the McKelvey
theorem. The McKelvey theorem says, assuming sincere voters, that an
agenda-setter can move from any one point to any other. Now with
sophisticated voters, and if X is covered by Y, an agenda-setter could not
construct an agenda that starts with Y and ends with X. If we add a
backward-moving agenda process the status quo is voted on last to
this brew, then we get the result that sophisticated voters can constrain
an agenda-setter to the points in the win-set of the status quo.
Now depose the agenda-setter and assume that any voter can make
proposals. We need a new concept, the uncovered set. Suppose an issue
space S. Take an alternative X; there are alternatives in S that X does
not cover and these are called the uncovered set of X, or UC(X). Take
an alternative Y; its uncovered set is UC(Y). Take an alternative Z; its
uncovered set is UC(Z). Find the uncovered set for every point in the issue
space S, and then nd the intersection of all those uncovered sets. That
intersection is the uncovered set for issue space S, UC(S). All alternatives
not in the uncovered set must be covered by UC(S). The points inside
UC(S) cover all the points outside UC(S). If B is inside the uncovered set
of the issue space and A is outside the uncovered set of the issue space,
then B covers A. As above, if B covers A, then sophisticated voters will
not select A.
The uncovered set always exists for an issue space. If there is a unique
Condorcet winner then that single point is UC(S). The minimum size of
the uncovered set is a point; otherwise its size depends on the distribution
of ideal points and the shapes of voters indifference curves. It may in-
clude the entire space; on a purely distributional question, the outcome
could be anywhere. If preference curves are circular (Euclidean) then
the uncovered set is contained within a circle centered in the yolk with
a radius four times that of the yolk. Depending on the distribution of
ideal points and the shapes of indifference curves, the uncovered set is
probably a small set near the geometric center of the ideal points. With
sophisticated voting, an open agenda, and a forward-moving agenda pro-
cess, outcomes will be restricted to the uncovered set. The intersection of
medians will be in the uncovered set, and the sophisticated outcome will
be there or nearby. Things are a bit different with the more commonly
seen backward-moving agenda process. With a backward-moving agenda
the only points that can win are in the win-set of the status quo, W(Q).
If the win-set W(Q) and the uncovered set UC(S) have elements in com-
mon, then the outcome will be in both W(Q) and in UC(S), and thus the
outcome will be in the uncovered set, that is, probably in the normatively
attractive center of the distribution of ideal points. If the win-set W(Q)
and the uncovered set UC(S) do not have elements in common, then the
outcome is in W(Q) but not in UC(S), and thus the outcome is not in the
uncovered set. In sum (Strom 1990, 124):
if legislators use sophisticated voting strategies, outcomes will generally be lo-
cated at not too great a distance from the geometric central point of the ideal
points of the legislators. Moreover when legislators have Euclidean preferences,
these outcomes will be conned to a relatively small centrally located set. This
does not mean that the joint median [intersection of medians] will be selected,
but it does imply that signicant deviations from this outcome are not to be
expected.
There are other ways to add friction back in. Tovey (1995), assuming sin-
cere voters, proposes an -core. The incumbent or status quo has some
amount of advantage, > 0, over alternatives. One can interpret this
as friction in moving from one alternative to another; new alternatives
that are extremely close to where we are are just not worth the effort of
moving. For Tovey, the yolk is an asymmetry measure of the distribution
of ideal points; the smaller the yolk the more symmetric the distribu-
tion. Then, if is sufciently large compared to the yolk radius, there
are no intransitivities. Schoeld (1995, 189) proposes a centrally located
set he calls the electoral heart, and argues that for large electorates, the
operation of direct democracy is well behaved. This is related to the ex-
ample above of stability with a voting rule of = 0.5001, a hundred
dimensional issue space, and around half a million voters. Another way
to add friction to models of candidate competition is with probabilistic
voting. Standard models assume that voters ideal points are determinis-
tic, each will vote certainly for the candidate with the platform closest to
her ideal point, and instability obtains in multidimensional space. If in-
stead, for whatever reason, ideal points are construed as probabilistic (an
ideal point is metaphorically the peak of a probability mountain falling
away in all directions), then two candidates competing for votes will con-
verge on an equilibrium (Mueller 1989, 196216). This equilibrium will
be within the Pareto set, and with stronger assumptions in a center of the
distribution: if the probabilities depend on differences in expected util-
ity, then competition drives candidates toward the (weighted) arithmetic
mean of the voters utilities, if the probabilities depend on the ratios of
the utilities then the equilibrium is driven toward the geometric mean
(Mueller 1989, 202).
Hinich and Munger (1994, 1997) combine formal rigor with an ad-
mirable concern for empirical content. Looking for the moment at mass
elections (rather than committees) voters have very little information
about the multitude of political issues, the connections among issues,
and the positions of candidates or parties on each of the issues. Voters
are rationally ignorant, according to Downs (1957); one individuals vote
does not change the outcome of an election and hence voters have no in-
centive to gather political information. I would add to Downs that even if
voting is expressively motivated, the typical voter is still not motivated to
express a detailed position on all possible political issues, but rather more
of a general stance or ideology. One cheers on the football teameven if one
does not understand every jot and tittle of the play. Voters, whether in-
strumental or expressive, are looking for information shortcuts. Ideology
in the neutral sense of the term is such a shortcut. Hinich and Munger
(1997, 191) say that ideology is an internally consistent set of propositions.
I think it would be better to call it a coherent set of propositions. Even a
hard science will contain, despite the most heroic efforts, theoretical ele-
ments that are not consistent with one another and certainly observations
that are not consistent with the theory or with one another. We do not
nd the one consistent system of beliefs, rather we nd that one system
of beliefs is more coherent, has more elements in harmony, than another.
Coherence can be formalized and measured in principle with parallel-
constraint satisfaction models (Thagard 1992), but that is a topic to be
pursued elsewhere.
Issues are connected to one another by three webs, according to Hinich
and Munger (1997, 193). The rst is communication. Just as a science
simplies abundant observations, an ideology simplies abundant issues
so that information can be conveyed with economy. In appealing to a
voter in a mass election a candidates message can be as simple as a single
slogan (see Popkin 1991). The second is commitment. The declaration
of a mere set of issue positions that a candidate promises to enact lacks
credibility. Positions require reasons and explanations; those reasons and
explanations cohere into a system with a few principles and values at
the apex of its hierarchy; to credibly express a commitment to principles
and values requires that they be sincerely applied to issues both old and
new. The third is budget. Spending more on one thing means spending
less on another. Issues are linked by budgetary constraint and by oppor-
tunity cost. If there are two parties, then due to ideology the effective
issue space collapses to one dimension; if a few parties then a few di-
mensions. They work this out as an extension of the spatial theory of
voting. Poole and Rosenthals work showing that the issue space in the
US Congress is mostly unidimensional is fundamental, according to
Hinich and Munger (1997, 196), and is consistent with their model of
ideology.
Poole and Rosenthal are also able to locate each representative and each
senator respectively at a point in the issue space of each two-year House
or Senate session. Remarkably, the correlation of members coordinates
from Congress to Congress is 0.95. They also calculate the annual move-
ment of members in an issue space. Such ideological movement declines
with length of service. There is little movement early in the 200-year
period and even less later on, such that:
Contemporary members of Congress do not adapt their positions during their
careers but simply enter and maintain a xed position until they die, retire or
are defeated. Indeed this stability is so great that, even in their last Congress,
members typically do not alter their liberal/conservative positions. Similarly they
generally dont alter their positions if they are redistricted. The major change in
behavior is that exiting members vote less often. (Poole and Rosenthal 1997, 74)
They suggest that the absence of ideological movement by individual
legislators since World War II is due to increased scrutiny of reputation by
expanding mass media. Further, political change in Congress results from
replacement of legislators, not from ideological changes in individual
legislators. It does not seem to be the inuence of ones fellows in the
caucus, because there is a 0.90 correlation between a members position
as a representative and her position if she is later elected to the Senate. The
few party-switchers, however, make dramatic moves in the issue space,
as if they had undergone ideological conversion. This suggests to me that
a concept of ideology would apply not only to mass elections such as in
Hinich and Mungers concept, but also to committee settings such as
the US Congress. Poole and Rosenthal (1999) also report that the issue
space of multiparty parliaments is mostly unidimensional, which would
not be explained by Hinich and Mungers two-party explanation for the
American data.
The models I have summarized are far more complicated and qualied
than I have been able to indicate. What I have said, although abstruse to
novices, will seem dismayingly casual to devotees. To continue the initial
metaphor, my purpose is merely to sketch some of the more realistic
frictions that have been proposed to account for the observation that the
pendulum always comes to rest. How these models will develop, and
how they will fare comparatively in empirical tests, I dont know. It is
possible that we are not adding frictions, but rather adding epicycles
to explain in Ptolemaic terms the looping motions of planets while we
await a Copernican revolution of a more parsimonious approach to the
theory of political preferences and their collective reconciliation, but I
dont know that either. What is warranted, however, is a rejection of
the McKelvey chaos theorem as a realistic (Rikers criterion) model and
hence a rejection of the normative implications for democracy that Riker
draws from it. Finally, those who remain convinced that ordinal pairwise
voting does have terrible consequences can adopt an alternative voting
rule without the believed problems, such as the Borda count.
Rikers argument resumed
Rikers driving image is the petals of the win-set that emanate from any
given point in a two-dimensional issue space (and their multidimensional
analogs). If the issue space is multidimensional and if we decide by pair-
wise majority rule, then no matter where we are there is always a majority
that prefers somewhere else to where we are and thereby has an incentive
to get there; politics is in pervasive disequilibrium. Riker (1982, 241)
confounds, however, the absence (given certain assumptions) of an equi-
librium with the absence of a public good: The popular will is dened
only as long as the issue dimensions are restricted. Once issue dimen-
sions multiply, the popular will is irresolute. There is, as we have seen, a
normatively attractive point of aggregate subjective welfare, the intersec-
tion of medians; it is just not in majority-rule equilibrium given ordinal
pairwise voting, multiple dimensions, and the unrealistic assumptions of
the basic model of spatial voting.
Riker must reconcile theoretical instability with empirical stability
(1982, 188192). We see only an apparent stability, he says, but it is
really an incremental disequilibrium. The idea of Rikers students that
institutions constrain chaos turns out to be quite helpful to his overall
scheme, as it provides a useful explanation for frequently observed stabil-
ity. Institutions that induce equilibriumare themselves a matter of choice,
however, and in multidimensional space there will always be dissatised
majorities with the incentive to overturn such institutions, Riker argues,
and this conveniently explains revolution as well as stability. Further-
more, we are told that revolution is not infrequent, and the example he
provides is the US Civil War, which he will illustrate was a consequence
of political disequilibrium.
Since revolution occurs and is admissible in a theory of stability, it follows that
some kind of theory of disequilibrium is a priori empirically superior to a theory
of equilibrium. The advantage of theories of disequilibrium is that they both
admit long periods of apparent stability (often indistinguishable fromincremental
change) and episodes of catastrophic revolution. (Riker 1982, 189)
But that is to neglect alternative hypotheses. A major alternative hy-
pothesis to explain change in aggregated preferences over time is change
in individual preferences! At one point, the US government favored
canals, later railroads, and still later highways. Is this due to cycling among
aggregate preferences over xed alternatives? Isolationist sentiments were
strong in the United States before World War II. Did Roosevelt induce
a cycle in order to declare war on Japan, or did preferences in the pop-
ulation change after the attack on Pearl Harbor? The US was friendly
with Iran, then the US was hostile to Iran, and now the US is trying
to be friendly with Iran: is this due to changes in conditions or due to
disequilibrium of tastes in the population? Pick any example you please
of change in a collectives choice, and almost always the Riker hypothesis
of preferences in disequilibrium will be among the least plausible as an
explanation. Holding an individuals preferences xed over time is as-
sumed as a convenience for the usual formal models because they can
handle changes in only one parameter at a time. Just because it is a mod-
eling convenience does not mean that it is a truth about humans, that
their preferences are xed at birth never to change. Even if we stipulate
the ction that all humans have identical underlying desires (Stigler and
Becker 1977), even under that scheme preferences not only differ be-
tween individuals, they also differ across time for an individual because
of differences and changes in beliefs. It is incorrect, and contrary to the
principle of methodological individualism that Riker endorses, to assert
that a territorial population of individuals, largely replaced by births,
deaths, and migration over decades of time, somehow has a xed distri-
bution of preference orders over decades of time, as he tends to do in his
analysis of the slavery issue in nineteenth-century American politics. He
attempts to explain contrasting political decisions forty years apart as due
to disequilibrium arising from unchanging preferences of the population
and the mathematical curiosities of ordinal pairwise voting (his argument
inadvertently appeals to preference change, but he intends it to be an
argument from disequilibrium).
Stability is sometimes real but only when it is imposed by institutions
[and] not the product of preferences and values. If we consider only val-
ues, then disequilibrium seems inherent in majority rule (Riker 1982,
191). Rikers dismal vision is that, In the end. . . institutions are no more
than rules and rules are themselves the product of social decisions. Con-
sequently the rules are also not in equilibrium (Riker 1980a, 444445).
All is chaos. Anything can happen (Riker 1982, 191). Such is the doc-
trine of democratic irrationalism. Consider Rikers proposition that even
the constraining institutions are arbitrary and unstable. The rst of his
errors is to imagine that political passions and political forces are exclu-
sively channeled by the sluice of ordinal pairwise voting. No matter what
Arrows independence condition demands, individuals are motivated by
rankings of preference rather than by pairwise comparisons. If a society
were at the central region of the distribution of preferences in the issue
space, how would a conspiratorial attempt to move to some extreme by
means of pairwise voting maneuvers be received? No doubt by a mo-
bilization of opinion, agitation, organization, contribution, and the like,
even revolution, all forces exogenous to a model of simple majority-rule
voting. Also, a discrete and insular minority may have its vital interests
oppressed, but when the opportunity arises the intensity of its concern
may overwhelm the less-aroused numerical majority. The second error is
to assume that individuals have condent knowledge of all of their own
and of others interests far into the future. Constitutions, rules of proce-
dure, precedents, are often established behind practical veils of ignorance
(imperfect to be sure), and are usually guarded frommyopic disturbances
by requiring supermajorities or other devices for going beyond the rules.
The garden club does not adopt Robertss Rules of Order because some
of its members calculate that those procedures will ensure their victory in
the controversy over the tulips six years hence. On a particular issue one
might be tempted to change the rule in ones favor, provided of course
that others are denied similar opportunities; but ones feasible long-term
interest among free and equal citizens is that no one be allowed to change
the rules in an opportunistic fashion. We need not assume any extraordi-
nary virtue motivating the creation and maintenance of fair institutions;
modest virtue along with some ignorance of future position is motivation
enough.
How frequent is harmful manipulation?
Common sense tells us that strategic voting and agenda control are
possible, but that such manipulations, harmful or not, are rare:
the literature on sophisticated voting exhibits a proclivity toward theoretical pos-
sibilities rather than empirical realities. Theoretical research suggests that oppor-
tunities for strategic manipulation of agendas should occur frequently and that,
to counter the power of the agenda-setter, congressmen are often forced to cast
sophisticated votes. However, agenda manipulation and sophisticated voting are
rarely mentioned in the most detailed accounts of congressional decision making.
Of the half-dozen or so papers which are exceptions to this rule, at least four focus
on the same set of roll-call votes. (Krehbiel and Rivers 1990, 549)
Poole and Rosenthal (1997, 164), commenting on their spatial study of all
roll-call votes over 200 years of the US Congress, conclude that strategic
behavior appears to be a destabilizing force only very rarely. The rare
exceptions that they extract from a review of political science literature
are the votes on the Powell amendment, the Depew amendment, and the
Wilmot Proviso, each as identied and portrayed by Riker in Liberalism
against Populism(1982). I shall showin each of the three cases that Rikers
analysis is awed and hence that there are no well-established cases of
harmful manipulation in the political science literature.
The controversial hypothesis that harmful manipulations are frequent
requires an empirical demonstration. The assertion that they are so fre-
quent as to render democracy meaningless requires a robust empirical
demonstration; as Hume said, it takes far more evidence to establish that
a man has been resurrected from the dead than to establish that a healthy
man has died inexplicably:
When anyone tells me, that he saw a dead man restored to life, I immediately
consider with myself, whether it be more probable, that this person should ei-
ther deceive or be deceived, or that the fact, which he relates, should really have
happened. I weigh the one miracle against the other; and according to the supe-
riority, which I discover, I pronounce my decision, and always reject the greater
miracle. If the falsehood of his testimony would be more miraculous, than the
event which he relates; then, and not till then, can he pretend to command my
belief or opinion. (Hume 1975, 116, sec. X, Part 1)
3
There is no frequency demonstration, fragile or robust, not when Riker
wrote, and none since by any of Rikers followers on this question, which
is why Riker needed the second but self-contradictory half of his ar-
gument (that manipulated outcomes cannot be distinguished from un-
manipulated outcomes): if preferences are unknowable, then empirical
demonstration of frequency or infrequency is obviated. In other words,
the second half of Rikers argument attempted to guarantee that if the
frequency hypothesis were true it could not be empirically tested. It re-
sembles the hypothesis that Satan planted false evidence of fossils in the
earth so as to tempt the faithful to abandon their belief in recent creation:
if the hypothesis were true, it couldnt be tested. Rikers fallback position
would be that, even if possible in principle, nevertheless an empirical test
of frequency would be practically impossible, because data are not suf-
cient to permit testing of a proper random sample from some natural
universe of democratic decision making. Strategic voting is an ineradica-
ble possibility, he says, but the factual question of whether people take
advantage of such possibilities is difcult to answer because we must in-
fer preferences, but considering the real-world examples I have offered,
it does seem likely that strategic voting occurs quite frequently (Riker
1982, 167). Thus, his argument comes to rest on the citation of a few
spectacular examples. These miracles are the glimpse that thereby proves
the existence of a supernatural world otherwise hidden to us.
Green and Shapiro (1994) accuse Riker (108113) and rational choice
scholarship in general (4244) of problems with the selection and in-
terpretation of evidence: a tendency to adduce conrming instances, a
tendency to project evidence from theory, and inattentiveness to com-
peting explanations. First, they say, Rikers theory of the generic insta-
bility of majority rule is consistent with any pattern of empirical ob-
servation of political decisions, since Rikers theory explicitly predicts no
change, incremental change, or revolutionary change (Anything can hap-
pen, Riker 1982, 191). The theory is conveniently unfalsiable. Second,
Rikers method of proof by spectacular example is contrary to method-
ological standards that require unbiased sampling procedures and for-
bid extrapolation from small numbers of instances. Riker claims that
his four stories of manipulation are not isolated examples but typical of
democratic politics (1982, 195). Green and Shapiro (1994, 111) say that
Riker adduces conrming instances of harmful manipulation, but fails to
adduce disconrming instances, such as the strategically ill-conceived
Smith Amendment to the 1964 Civil Rights Act (the Smith amendment,
intended to kill the Civil Rights Act, prohibited discrimination on the ba-
sis of gender). Third, Rikers narrative about the Depew amendment (to
a resolution establishing the direct election of senators which eventu-
ally became the 17th Amendment to the US Constitution), for example,
is poorly established and not strongly argued (the Depew amendment
will be considered in full below). Fourth, the illustrative narratives fail to
consider and reject alternative hypotheses; yes, parliamentary contrivance
occasionally forestalls or defeats the majority will, but is the source of such
outcomes necessarily manipulation of multidimensional issue spaces or is
it more simply just blunder or short-sightedness, or other factors? Green
and Shapiros objections are enough, I believe, to convince anyone but
the faithful that Riker fails to establish his controversial hypothesis about
the frequency of cycles and harmful manipulation.
Green and Shapiros complaints resemble those that Hume brought
against the existence of miracles: even if there is some evidence in sup-
port of the miracle we have to weigh that against the enormous evidence
in support of an overwhelming absence of miracles generally. I should
not believe such a story were it told to me by Cato, was a proverbial saying
in Rome, even during the life of that philosophical patriot. The incred-
ibility of a fact, it was allowed, might invalidate so great an authority
(Hume 1975, 113). I go one step further and show that the evidence
alleged in support of the cycling miracles is deluded: not merely is it
improbable that Vespasians spittle cured a blind man in Alexandria, to
use one of Humes examples (122), but also Ill show that Vespasian was
nowhere near Alexandria at the time alleged and that the man in ques-
tion died blind. Therefore, let us momentarily accord Rikers argument
the most generous charity. Suppose, only for the sake of argument, that
it is theoretically or practically impossible to estimate the frequency of
harmful manipulations, that all we can rely on are some convincing ex-
amples, and that if those examples were established then the failure to
observe instances of harmful manipulation is an illusion. I am not saying
that I will take on all examples from all comers; that is impractical. We
may fairly presume, however, that Rikers examples in Liberalism against
Populism (1982), and those in his refereed articles (1958, 1984), often
repeated in political science training today, represent his best attempt to
establish his case. If his examples fail, then so must his case, on its own
terms.
Summary
Riker originally maintained that the Arrow theorem implied pervasive in-
stability indemocratic politics. Stability rather thaninstability is observed,
however, and Riker responded that we usually do not have enough data
to determine whether or not a cycle has occurred. Then it was demon-
strated that cycles are unlikely with mild similarity among preference
rankings. Riker responded that although natural cycles are rare, politi-
cal actors manipulate the outcome by means of strategic voting, agenda
control, and introduction of new issues and dimensions. Strategic vot-
ing and agenda control tend to cancel each other out, however. Thus,
the McKelvey and Schoeld chaos theorems of total disequilibrium in
multidimensional issue spaces promised to rescue Rikers disequilibrium
hypothesis. Now observed equilibrium is no longer the consequence of
mildly similar preference rankings, but is rather the result of imposed
institutions (themselves in ultimate disequilibrium) that arbitrarily select
an outcome fromamidst the chaos. In this chapter we have seen that there
is a normatively attractive point, the intersection of medians, in multi-
dimensional issue space. We have seen that controlled human-subject
experiments do not support the predictions of the Rikerian chaos in-
terpretation of the McKelvey and Schoeld theorems, and that empirical
investigations in support of such an interpretation would be difcult to
impossible to carry out. Structured preference orders, and supermajority
rules, each reduce the likelihood of instability. Adding realism to the spa-
tial model strategic voting, friction in moving from one alternative to
another, probabilistic voting, or constraints of ideological coherence
tends to return the outcome to the normatively attractive center of
the populations preferences. Riker believes it sufcient for his case that
he demonstrate the possibility of manipulation by way of empirical
illustrations, which we begin to examine in the next chapter.
9 Assuming irrational actors:
the Powell amendment
Introduction
Rikers rst spectacular example is the 1956 school-construction bill in
the US House of Representatives. A bill to fund school construction was
amended to forbid segregation of white and black students in areas re-
ceiving aid. The amended bill failed. Riker offers two versions of this
story; Krehbiel and Rivers (1990) a third; my running commentary chal-
lenges Rikers interpretations and adds more esh to the bones of the
Krehbiel and Rivers model. The majority that wanted school aid was
thwarted by the desegregation amendment, according to Riker. Rikers
rst version argues alternatively that there was a natural cycle or that
there was strategic voting by opponents of school aid that defeated the
majority and hence is an instance of harmful manipulation. The second
version drops the natural-cycle claim, revises the strategic-voting claim,
and presents revised estimates of the distribution of preferences in the
chamber. Since strategic voting can be countered by strategic voting the
question arises as to why those who favored both school aid and deseg-
regation failed to vote strategically against the desegregation amendment
in order to retain school aid. Rikers answer in his second version is that
these legislators were constrained by the irrational preferences of their
constituents.
The same issues came to a vote in 1957, however, and the same leg-
islators who failed to vote strategically against the desegregation amend-
ment in 1956 did vote strategically against the desegregation amend-
ment in 1957, which destroys Rikers second interpretation. Krehbiel and
Rivers offer revised estimates of the distribution of preferences in 1956,
which were incompletely known according to their model, a game with a
Bayesian Nash equilibrium. The Northern Democrats who voted for the
desegregation amendment mistakenly believed that Republicans would
join themin supporting an amended school-aid measure. After votes were
revealed it was plain that there was not a majority for school aid, with or
without the desegregation amendment. I add to Krehbiel and Riverss
197
already convincing exposition additional evidence in support from the
Congressional record and newspaper accounts. Particularly, their model
would predict that, with complete information about the distribution of
preferences, as in 1957, supporters of school aid would vote strategically
against the desegregation amendment. I show that this is exactly what
happened in 1957, and that the actors openly declared that their votes
were strategic.
The Powell amendment
It is 1956, the Republican Eisenhower is President and facing reelection,
the Democrats control Congress, and the postwar baby boom is well un-
derway. Two years before, the US Supreme Court, in Brown v. Board
of Education (litigated by the National Association for the Advancement
of Colored People, or the NAACP) had declared school segregation un-
constitutional, but neither the local and state governments responsible
for segregation nor the federal government responsible to the Constitu-
tion had taken any steps of any kind to implement the decision. In the
United States education is traditionally a state and local concern, in pol-
icy and in funding, but as the election approaches it looks like it will
be politically popular to offer a one-time ve-year program of federal
aid for the construction of new schools. The Eisenhower administra-
tion proposes a bill for federal school aid in 1956, and a Democratic-
amended version of the legislation eventually arrives on the oor of the
House of Representatives. Adam Clayton Powell, the African-American
Member of Congress from Harlem, offers an amendment to the school-
funding bill. The amendment proposes to deny the new school aid to
any school district that doesnt conform to recent Supreme Court de-
cisions, in other words, a school district that accepted federal funds
would be required to desegregate. The Powell amendment comes to a
vote, and passes 229 to 197. The school-aid bill as amended by Powell
comes to a nal vote and is defeated, receiving 199 afrmative votes
and 227 negative votes. Riker argues that voters who were both against
school aid and against the desegregation mandated by Powells amend-
ment voted strategically for Powells amendment in order to defeat the
entire school-aid bill. The manipulators would be betting that south-
ern members of Congress who wanted school aid would not be able to
vote for a bill mandating desegregation. The upshot is that Powell and
his liberal allies who prefer school aid to no school aid are thwarted,
that they could have had their way if they hadnt passed the Powell
amendment. A cycle has been contrived. A harmful manipulation has
succeeded.
The Powell amendment 199
Table 9.1. Distribution of votes, 1956
Original
Yea Nay Total
Yea 132 97 229
Powell Nay 67 130 197
Total 199 227 426
There are three alternatives under consideration:
O: The Original school-aid bill.
P: The school-aid bill amended by Powell so as to require desegregation.
Q: The status Quo of no school funding.
The recorded votes are as displayed in Table 9.1. Fromthe recorded votes
it is obvious that P >O, the Powell amendment beats the original school-
aid proposal (by 229 to 197, Congressional Record Roll-Call Vote No. 90,
Congressional Quarterly Vote No. 46). Also it is obvious that Q > P, the
status quo of no funding beats the school-aid bill as amended by Powell
(by 227 to 199, Congressional Record Roll-Call Vote No. 92, Congressional
Quarterly Vote No. 48). That leaves O against Q, the original proposal
against the status quo, to account for. The 199 representatives who voted
yea on nal passage of the Powell-amended school-funding bill, P > Q,
could order the original bill, O, in one of three ways:
r
P >O>Qand O>P >Q. Notice that such voters prefer federal school
aid (with or without the desegregation amendment) over the status quo.
r
P > Q > O. These are voters who rank school aid with desegregation
over no school aid over school aid with segregation.
Next, nothing in the ideological circumstances of 1956 rendered
[P > Q > O] likely . . . so I conclude that the 199 who voted for -
nal passage [P] had either [P > O > Q or O > P > Q] and thus all
preferred [O > Q] (Riker 1982, 154). Some Southern Democrats,
he argues, must have preferred school aid without desegregation over
the status quo of no school aid over school aid with desegregation
(O > Q > P). Further, according to Riker, 18 Democrats, including
15 from southern and border states, who voted to defeat school aid with
the Powell amendment in 1956, voted for a school construction proposal
lacking the Powell amendment in 1957. The reader should be alert that
we take data sometimes from a 1956 vote and sometimes from a 1957
vote. Altogether then, the 199 who voted for nal passage of the Powell-
amended school-aid bill must have preferred O>Q, and there are at least
another 18 others whom the 1957 evidence indicates ranked O > Q, so
that 199 + 18 = 217 > 209, O > Q. A cycle has been demonstrated:
from the recorded votes, P > O and Q > P, and from inference based on
the 1957 vote O > Q, in all we have Q > P > O > Q, a cycle.
Does an alleged cycle show once again that the aggregation of pref-
erences is incoherent? No. The Borda count for the three alternatives
assuming Rikers counts above is Q > P > O, the status quo is ranked
rst, the Powell-amended school aid is ranked second, and the origi-
nal school aid-bill is ranked last. The YoungKemeny method reports
the same ranking. Moreover, the Borda ranking survives even if we help
Rikers cycle assertion along by supposing that everyone, 426 > 0, favors
O > Q. If there is a cycle, then the problem is not with the preference
rankings, the problem is with pairwise voting procedure. I have claimed,
however, that cycles are rare so that we dont have to worry too much
about the defects of pairwise voting. I shall now show that Rikers calcu-
lations are erroneous, and that there was not a cycle as he imagined.
Riker does not give a citation to the 1957 vote on school construc-
tion, but the only roll-call vote on the issue in 1957 took place on July 25
(Congressional Record Vote No. 154, Congressional Quarterly Vote No. 56).
After some parliamentary maneuvers that I shall recount later, an oppor-
tunity arose to vote on school aid without a desegregation amendment, a
pure contest between O
and Q. The vote was 208 > 203 (actual votes

if we count pairs, announced votes and Congressional Quarterly-polled
votes would be 215 > 209) for Q over O
. Federal aid to schools was

killed in 1957 and thereafter dropped off the national agenda.
Unfortunately, Rikers use of the 1957 data is miscalculated. He states
that 18 Democrats, including 15 from southern and border states who
voted to defeat school aid with the Powell amendment in 1956, voted in
1957 for the school construction proposal lacking a Powell-type amend-
ment. This is mistaken. There are only 14 Democrats who satisfy these
constraints. Also, 7 of those were from Alabama, and for local reasons in
1957 7 out of 8 voting Alabama representatives voted for federal school
aid, eventhough 31 out of 33 representatives inthe immediately neighbor-
ing states of Florida, Georgia, Mississippi, and Tennessee were against
federal school aid.
1
The fact that 7 out of the 14 Democrats are the
anomalous Alabamans who were most likely O > P > Q casts doubt on
the story that a sufcient number of southern Democrats naturally had
the preference ranking O > Q > P.
Further, recall that Riker dismisses the preference order P > Q > O as
unlikely. Consider, however, that ideologically there might be some Re-
publicans who wouldupholdcivil rights evenif encumberedby school aid,
P, but otherwise would not support increased federal spending, Q > O.
If we check the 199 who voted for P > Q in 1956, we discover that
among them there are indeed 13 Republicans who voted Q>O
in 1957:
these 13 are unequivocally P > Q > O
. Therefore, Rikers calculation of

199 + 18 = 217 > 209, O > Q, must be corrected to 199 + 14 13 =
200 < 226, Q > O.
What happens if we look for all voters, Democratic and Republican,
who voted to defeat school aid with the Powell amendment in 1956 and
voted in 1957 for the school construction proposal lacking a Powell-type
amendment? We nd that there are 24 who satisfy these constraints, in-
cluding the 14 Democrats already identied.
2
We should also look for
all voters, Democratic and Republican, who voted to pass school aid
with the Powell amendment in 1956 and voted in 1957 against the
school construction proposal lacking a Powell-type amendment. There
are 18 who satisfy these constraints, including the 13 Republicans already
identied.
3
With these new gures, the calculation would be 199 + 24
18 = 205 < 221, again Q > O. From the 1956 recorded votes we have
P > O, Q > P, from consistent use of the 1957 data we have Q > O,
and thus we have Q > P > O which is not a cycle. The assertion of a
natural cycle depends on inconsistent data selection and therefore fails.
The status quo is preferred to the Powell amendment, is preferred to
federal aid to schools, just as history played out in 1956, 1957, and in
long run.
Riker (1982, 156) alleges alternatively that there was probably strategic
voting sufcient to contrive a cycle on the vote; that opponents of school
aid voted strategically and thus succeeded in bringing about its defeat
(Riker 1986 is the stronger version of his argument for the presence of
strategic voting, so I postpone exposition and criticism of his strategic-
voting argument for the moment). The problem with the contrived-cycle
argument on the assumptions given in Riker (1982) is that if the op-
ponents of school aid had voted strategically, then so could have the
proponents of school aid. Remember that if everyone votes strategically
the outcome is the same as if everyone votes sincerely. For the contrived-
cycle argument to succeed, some voters would have to vote strategically
and others vote sincerely. So, why would the proponents of school aid fail
to vote strategically? This problem is not addressed, let alone stated, in
Riker (1982).
If the reader think me unfair, allow me to point out that my conclusion
that there was no natural cycle in 1956 is supported by Denzau, Riker, and
Shepsle (1985)and Riker (1986). In their second version of the Powell
amendment story the assertion of a natural cycle in 1956 is dropped,
quietly and without explanation. In the second version, the cycle (and
consequent harmful manipulation) is alleged to be contrived by strate-
gic voting. Since the new argument is conned to strategic voting, their
major explanatory challenge becomes to account for why the majority of
the proponents of school aid voted in a sincere and thus self-defeating
fashion.
In 1982 the assertion was that the 199 who voted for the Powell amend-
ment had either the preference ranking P >O>Qor O>P >Q. In 1986
the assertion was that those 199 are divided as follows. The 132 who voted
both for the Powell amendment and for the nal passage of the Powell-
amended school-aid bill (we shall label groups with two letters, the rst
their yea or nay votes on the Powell amendment, the second their yea
and nay votes for passage of the amended school aid bill, hence these 132
are labeled YY ) had the preference ranking P > O > Q, they preferred
school-aid-and-desegregation to school-aid-and-segregation to the status
quo of no-school-aid-and-segregation. Call the 132 in YY the Powellites.
The 67 who voted against the Powell amendment but for the Powell-
amended school-aid bill (NY ) rank O > P > Q, the unamended school-
aid bill over the Powell-amended bill over the status quo of no school
aid. Call the 67 in NY the School-Aiders. In 1982 Rikers assertion was
that segregationist southern Democrats ranked O > Q > P, that their
rst-best was school aid with no Powell amendment, their second-best
the status-quo of no school aid, and their worst the desegregating Powell
amendment.
In Riker (1986), the 130 who voted against both the Powell amendment
and against nal passage of the Powell-amended school bill (NN ) are said
to rank O > Q > P. Call the 130 in NN the Southerners. That leaves the
99 Republicans who voted for the Powell amendment but against the
Powell-amended school-aid bill (YN). In 1982 Riker asserts that some
of them rank Q > O > P and some rank Q > P > O. In 1986 Riker
divides them in half: 48 Segregationist Republicans rank Q > O > P, no
school aid over some school aid, over school aid with desegregation; and
49 Desegregationist Republicans rank Q > P > O they prefer no school aid,
but if there is school aid it should be with a desegregationist amendment
rather than without one.
According to the 1986 version of the distribution of preferences, there
is no natural cycle: O > Q, 329 to 97; Q > P, 227 to 199; and O > P,
245 to 181; O > Q > P. The cycle is contrived by the 48 Segregationist
Republicans. They are said to sincerely prefer Q>O>P. If they vote ac-
cording to their sincere preferences, the outcome is their second-ranked
preference, a school-aid bill without a Powell amendment. If these Seg-
regationist Republicans instead vote strategically as if they were Desegre-
gationist Republicans, that is, Q > P > O, then a cycle is contrived: O >
Q, 329 to 97; Q > P, 227 to 199; and P > O, 229 to 197; O > Q >
P > O. We know that if there is a cycle, then the pairwise outcome
Table 9.2. Rikers estimates of factions and preference rankings, 1956
Recorded Votes How Many Nickname Ranking
YY 132 Powellites P > O > Q
NY 67 School-Aiders O > P > Q
NN 130 Southerners O > Q > P
YN 48 Segregationist Republicans Q > O > P
YN 49 Desegregationist Republicans Q > P > O
Table 9.3. Pairwise comparison matrix: Riker (1982), Riker (1986)
P O Q
1982 1982 1982
1986 Sincere 1986 Sincere 1986 Sincere
1986 Strategic 1986 Strategic
P 229 199
1982 48 = 199
1986 Sincere 181
1986 Strategic + 48 =
229
O 199 + 18 =
1982 197 217
1986 Sincere + 48 = 18 + 130 =
1986 Strategic 245 329
48 =
197
Q 227 209
1982 + 18 130 =
1986 Sincere 227 97
Note: Boldface type in data indicates the estimates made by Riker in (1982), italic
type indicates sincere preferences estimated by Riker (1986), and underlined type
indicates sophisticated preferences estimated by Riker (1986).
depends on the sequence of the agenda. The sequence of voting sup-
posedly foreseen by the manipulating Segregationist Republicans was,
rst, P against O, the Powell amendment against the original bill, P >
O, 229 to 197, P wins; and the second was thus Q against P, the sta-
tus quo against the Powell-amended school-aid bill, Q > P, 227 to
199, and Q wins. By voting strategically, the Segregationist Republicans
have avoided their second-ranked outcome, O, an unamended school-
aid bill, and achieved their rst-ranked outcome Q, the status quo of no
school aid.
If the Powellites had anticipated that the Segregationist Republicans
would vote strategically, however, then they could themselves have voted
strategically and thereby have avoided their bottom-ranked outcome, Q,
the status quo, and instead have achieved, O, school aid, their second-
ranked outcome. To achieve their second-ranked outcome rather than
their third-ranked outcome the Powellites, anticipating the strategic re-
sponse of the Segregationist Republicans, should not introduce the Powell
amendment or should vote against any similar amendment introduced by
the opponents of school aid. Then the contest is between O and Q, and
by Rikers estimate O > Q, 329 to 97. So why did the Powellites vote
sincerely and thus against their own interests?
In fact there was major public controversy among nonsouthern
Democrats about the Powell amendment. The NAACP supported, and
was indeed behind, the measure. The liberal National Education Asso-
ciation opposed the desegregation amendment, fearing that it would kill
the school-aid bill. The trade union federation, the AFL-CIO, originally
supported and later opposed the Powell amendment. Mrs. Roosevelt op-
posed the amendment, as did former President Truman, gures with
records of early and strong support for African-American civil rights. In
an open letter, Truman stated exactly the Riker hypothesis:
The Powell amendment raises some very difcult questions. I have no doubt that
it was put forward in good faith to protect the rights of our citizens. However,
it has been seized upon by the House Republican leadership, which has always
been opposed to Federal aid to education, as a means of defeating Federal aid and
gaining political advantage at the same time. I think it would be most unfortunate
if the Congress should fall into the trap which the Republican leadership has thus
set. That is what would happen if the House were to adopt the Powell amendment.
The result would be that no Federal legislation would be passed at all, and the
losers would be our children of every race and creed in every State in the Union.
(Riker 1986, 126)
Thus it is absolutely certain that the NAACP, AdamClayton Powell, and
those we have labeled the Powellites, were aware of this estimate of the
situation. So, why did they ignore advice such as Trumans and act in an
apparently self-defeating manner?
Rikers second version of the Powell amendment story solves this prob-
lem by positing that the Powellites were constrained by their electoral in-
terests to vote sincerely. They would be unable to explain a strategic vote
against a desegregation amendment to their constituents who favored
desegregation. Denzau, Riker, and Shepsle (1985) take this homely little
solution and drape it in yards and yards of formalism, almost as if they be-
lieve along with Keats that beauty is truth and that is all you need to know.
Curiously, 44 of the 46 references intheir paper onthe Powell amendment
are to work by rational-choice scholars, yet there is not a single reference
in support of the historical facts of the case. They acknowledge that the
Powellites image-enhancing activity, which discourages strategic behav-
ior, comes ultimately at the expense of constituent welfare (1,132). They
do not explain why the Powellites constituents are irrational. Although
Riker seems not to have realized the problem at the time of writing, his
rst version of the story implied that some legislators are irrational and
that some are not. His second version of the story saves legislator ra-
tionality with the assumption that the voters in Powellite districts are
uninformed and irrational, although voters in Segregationist Republican
districts are informed and rational. According to Riker, Powell himself
was motivated by expressive rather than instrumental concerns: Powells
motive was, undoubtedly, to force some representatives to stand up for the
symbolism he created. . . His amendment said to the Democratic leader-
ship, We blacks must be treated with dignity (1986, 128). Powell, and
more so the NAACP behind the amendment, had an enviable record
of instrumental accomplishment, however. Their successes against formal
segregation are well known, and I am sure it is safe to say that they would
be quite unhappy if all this activity turned out to be merely an expres-
sive exercise. That the Powell amendment was not merely expressive but
rather also instrumentally motivated is shown, for example, by the fact
the NAACP did not seriously push for a school-desegregation amend-
ment in Congress until after they had won Brown v. Board of Education
in 1954. It is extraordinary that a rational-choice explanation of a political
event ends up requiring the assumption that a large number of actors are
not instrumentally rational. It suggests that a failed hypothesis is being
saved by any means necessary.
There is an insurmountable problem for Rikers second version of
the story. The failure of the 1956 school-aid bill was front-page news
and became a major issue in the 1956 Presidential and Congressional
election campaigns: Republicans blamed Democrats for the failure, and
Democrats blamed Republicans. If ever in 1956 the Powellites had second
thoughts about their alleged irrationality, however, all they had to do was
reintroduce O, school aid without a desegregation rider. Rikers rst ver-
sion of the story nds that O>Q, and his second version nds that O>Q
by an overwhelming margin, 329 to 97, and we can add on Rikers behalf
that the 1956 outcome of Q > O would be evidence of grossly harmful
defeat by manipulation of the supposed Condorcet winner O. As I said
though, in 1957 the Northern Democrats brought about an opportunity
to vote on school aid without a desegregation amendment, a pure contest
between Qand O
, and Qbeat O
by 215 >209. There were distributional

differences between O in 1956 and O
in 1957, but they only strengthen

my argument. O in 1956 was a Democrat-favored distributional formula
that Republicans claimed caused them to vote against P on nal passage,
but O
in 1957 was the Republican-crafted distributional formula, so

the Republicans excuse was no longer available (the Republicans need-
based formula favored the South because of the Souths lower incomes).
To reiterate, O
lacked a desegregation amendment, and possessed the

Republican-sponsored but South-leaning distributional formula, but it
was the Southern Democrats and the Republicans who killed O
. Riker
(1986) predicts O > Q by a margin of 232 votes, but as revealed in 1957,
Q > O
by six votes, so the Riker prediction is off by 238 votes out of a possi-
ble 426 and is in the wrong direction. Riker appeals to the 1957 data in his
attempt to demonstrate a natural cycle in his rst version of the Powell
amendment (Riker 1982), so clearly he is aware of the nature of the 1957
data, nor could he claim that it is illegitimate for another researcher to
look to the 1957 deliberations for more data points.
The failed prediction means that something must be seriously wrong
with Rikers (1986) second estimates of preference rankings, but what?
The major problem, it turns out, is with the Republicans. He estimates
that the 97 Republicans who voted YN, yea for the Powell amendment
but nay against passage of Powell-amended school-aid, are 48 Q>O>P
and 49 Q > P > O. Krehbiel and Rivers (1990) estimate that these
97 Republicans are either P > Q > O or Q > P > O, in other words, that
the Republicans must have ranked O the unamended school-aid bill last.
Why? First, the Powell-amended school-aid would be more distribution-
ally advantageous to the Republicans, for the simple reason that money
denied to segregationist school districts in the south would ow north to
Republican districts, so most Republicans would rank P > O. Second,
and more importantly, the Republicans were the party who fought the
Civil War over slavery and consistent with that tradition in the 1950s
tended to support civil rights legislation, so few would rank O > P.
Krehbiel and Rivers compare votes on the 1956 school-aid bill with
votes on the unencumbered Civil Rights Act of 1956 and nd that only
10.4 percent of Southerners (NN on the Powell amendment and on
nal passage of the Powell-amended school-aid bill) supported the Civil
Rights Act of 1956, 94.5 percent of Powellites (YY) supportedcivil rights,
90.2 percent of School-Aiders (NY), and 81.5 percent of Republicans
(YN). Thus, 79 of the 97 YN Republicans voted for an unrelated civil
rights measure. According to Rikers estimation of the natural distribution
of preferences, P > O by 245 to 181, so that to reverse this ranking for a
contrived cycle requires that there exist at least 33 Segregationist Repub-
licans who would vote strategically (24533 = 212 < 214 = 181 + 33,
P <O), but the unrelated vote on the unencumbered civil rights measure
suggests that at best there were only 18 such Segregationist Republicans.
Finally, the Republicans are the party of lower taxes and less government,
so we are uncertain about their ranking of Qand P, even though they rank
both above O.
There is also a problem with the (NN ) Southerners who Riker has
ranked as O > Q > P. Clearly, they rank desegregation, P, last, but how
do they rank O and Q? They might like federal school aid out of dis-
tributional advantage (O), or they might dislike it altogether because of
their suspicion of federal involvement in state issues, an issue that they
had already fought and lost a war about (Q). So, we conclude that we
are uncertain about their ranking of O and Q, even though they rank
both above P. There is a nal problem with the Northern Democrats,
the 132 Powellites (YY ), and the 67 School-Aiders (NY ). Riker says
that the Powellites rank P > O > Q and that the School-Aiders rank
O > P > Q. In the oor debates, however, the School-Aiders state that
their ranking is P > O > Q but that they are voting O > P > Q for
strategic purposes to obtain the otherwise unavailable O. Further, dur-
ing the debate, Powellites explicitly acknowledged the strategic logic of
liberal opponents of the Powell amendment [the School-Aiders]. Yet the
Powell amendment passed three times (Krehbiel and Rivers 1990, 564).
So we shall consider that the 199 Northern Democrats Powellites and
School-Aiders who voted for nal passage of Powell-amended school-
aid ranked P > O > Q, but that 97 of the 199 expressed O > P > Q in a
strategic maneuver that failed due to mistaken estimate of the distribution
of preferences.
With these new assessments of the preference rankings, including the
uncertainty about the preferences over P and Q among the Republicans
and the uncertainty about preferences over O and Q among the South-
erners, Krehbiel and Rivers (1990) write a game of incomplete infor-
mation with a Bayesian Nash equilibrium. There are four players. Both
the Powellites and the School-Aiders rank P > O > Q, but we allow
cardinal utilities because the players are making probability estimates.
The Powellites very much prefer the desegregation amendment above
the two remaining alternatives but also prefer school aid to no school aid,
P O > Q. The School-Aiders prefer school aid with desegregation
to school aid without desegregation, but very much prefer school aid
of any kind over the status quo, P > O Q. The Southerners are
O > Q > P with probability s and Q > O > P with probability 1 s. The
Republicans are P > Q > O with probability r and are Q > P > O with
probability 1 r. There are three blocs, the 199 Northern Democrats
(Powellite and School-Aider), the 130 Southerners and the 97 Repub-
licans, and the fusion of any two blocs makes for a winning majority.
The vote on any bill, P or O, against the status quo, Q, comes last in
any agenda, and there is never a reason to vote strategically on the last
vote. The Northern Democrats thus become pivotal; if they adopt O,
school aid with segregation, then the Southerners will vote between O
and Q (the Southerners rank P last). If the Northern Democrats adopt
P, school aid with desegregation, then the Republicans will vote between
P and Q (the Republicans rank O last). A Northern Democrat legislator
strategically votes for O if he thinks that s, the probability that the South-
erners are of the type O > Q > P (rather than the type Q > O > P) is
larger than the probability r that the Republicans are of the type P > Q >
O (rather than the type Q > P > O). A Northern Democrat votes for P if
she thinks that the probability s is smaller than the probability r. To put it
another way, a Northern Democrat votes against the Powell amendment
if he thinks it more likely that hell get votes from the Southerners for
an unamended school-aid bill than that hell get votes from the Repub-
licans for Powell-amended school aid. A Northern Democrat votes for
the Powell amendment if she thinks shes more likely to get votes from
Republicans for Powell-amended school aid than she is to get votes from
the Southerners for unamended school aid.
According to the model, a Northern Democrat for whom the Powell
amendment is much more important than the remaining two alternatives
(P O > Q) will play boldly and take a greater risk of losing by in-
sisting on the Powell amendment, and a Northern Democrat who much
prefers federal school aid to the status quo (P > O Q) will play cau-
tiously and vote against the Powell amendment. Krehbiel and Rivers
(1990) present data showing that Northern Democrats from states that
lose redistributionally (state gets back less federal aid than the income
tax it pays in) tend to vote for the Powell amendment, and Northern
Democrats from states that win redistributionally tend to vote against
the Powell amendment. They present further data showing that, control-
ling for redistribution and demographic variables, Democratic legislators
are more likely to vote for the Powell amendment the more African-
Americans there are in their district.
Only after the vote is it apparent that the Northern Democrats could
not have won by adopting the Powell amendment and counting on Re-
publican support; it is revealed that the Republicans are of the type Q >
P > O. The 1956 vote reveals that 97 Republicans prefer Q > P and 97
Republicans prefer P > O. We already know that 81.5 percent of these
Republicans voted on record in favor of civil rights on another bill in 1956
(P > O), and the ranking inference Q > O
is further conrmed by the

clean vote in 1957 on school aid without desegregation versus the status
quo; by my calculations, 86.3 percent of those 97 Congressional Districts
voting YN in 1956 voted for Q > O
in 1957 (even when including 13

replaced legislators and legislators with unknown 1957 rankings in the
total). From the larger context it also becomes apparent that for emer-
gently understood distributional reasons most Southerners will not sup-
port federal school aid under any circumstances, and thus the Northern
Democrats could not have won by rejecting the Powell amendment and
counting on Southern support either. From an account a few days after
the vote:
The position of the Southern Democrats is perfectly simple. They are opposed
to any legislation which might be used, directly or indirectly, to ease them into
obeying the Supreme Courts decision against racial segregation in the public
schools. They used every parliamentary device in the book to keep this school-
aid bill frompassing for this reason: they knewthat once this bill was on the books,
it would have been easy next year for some Northern Senator or Congressman
to tag onto the appropriation bill for the Department of Health, Education and
Welfare an amendment stating that none of the funds in the bill could be allocated
to states that had not carried out the Supreme Courts integration decisions. Once
this were done, the Southerners would have been disarmed, for they could not
libuster an appropriation bill without cutting off all funds for the Department
of Health, Education and Welfare, including Social Security funds, and nobody
is prepared to do that. (Reston 1956)
Their political friends wondered why Powell and the NAACP were
so insistent on their amendment when it seemed likely that the Senate
would libuster any aid bill that carried a desegregation rider. The African
Americans may have been one step ahead on this question. Powell con-
dently dismissed the libuster problem in Northern Democratic forums,
and eventually the Southerners came to understand the basis of his con-
dence. Powells insistence on his amendment would reveal a majority in
the Congress willing to enforce Brown v. Board of Education. Once this ob-
jective was accomplished, federal school aid could be pursued on its own
merits. During the 1956 debate Representative Roosevelt (D-CA), from
the context clearly privy to Democratic strategy, stated that the Powellites
would stand up and ght for civil rights, even at the cost of temporarily
delaying the necessary money to nish some of our schoolrooms
(Congressional Record, July 3, 1956: 11761).
I suggested that if Rikers version of the story were correct, then the
Powellites could quickly come back, propose O, win O, as Roosevelt
seemed to have in mind, and be the heroes in the upcoming Presidential
and Congressional elections. After the distribution of preferences was
revealed by the nal vote, however, it became clear to the Democrats that
they should not bother to raise the issue again; there were not enough
votes to win school aid with or without a Powell amendment. How did
the Democrats let it get to the oor in the rst place? Originally the
preference orders of other legislators were not fully known. The School-
Aiders expected that enough Southerners would support federal school
aid without desegregation (incorrect), and the Powellites expected that
enough Republicans would both support civil rights (correct) and support
school aid (incorrect). Federal aid for local school construction was a
big election issue for Republican President Eisenhower. Incidentally, it
is disturbing that Rikers narratives fail to mention that federal school
aid was Eisenhowers proposal (it was at rst called the Hobby bill after
the Republican secretary of health, education, and welfare), for example,
The story I have to tell involves a bill for federal aid to education, a bill
sponsored by the Democratic leadership of the House of Representatives
in 1956. (It was clearly not a bill put forward by the administration, which
was of course Republican) (Riker 1986, 117). Failure of school aid was
described as a major defeat for Eisenhower (Morris 1957a).
The silence of the President during this weeks debate is extremely interesting.
Ever since his rst inaugural address, he has talked about the urgent need for
prompt action in this eld, but with his own party divided and wavering on
what to do, he did not send a single word to Congress during this weeks debate,
though he had repeatedly and publicly urged the Congress to adopt his views on
foreign-aid legislation the week before. (Reston 1956)
The Northern Democrats expected that Republican Eisenhower would
deliver his own troops and were surprised when he did not. The failure
of the school bill was little expected by most students of Congress until
the sudden House debacle this week (Lewis 1956).
Brief investigation of secondary historical literature shows that the
NAACP strategy was not stubborn and foolish as in the Rikerian ver-
sions of the story. As Powells biographer (Hamilton 1991, 224) put it,
the NAACP and other civil rights organizations faced a constant dilemma
in the 1940s and 1950s: How far should one go in insisting that seg-
regation and discrimination be totally eliminated? Should the benets
projected under less than full elimination of segregation be accepted as
the best that could be obtained? When was the latter a sellout of ones
principles? If they were too reasonable in pursuing their desegregation
objectives then nothing would change, and if they were too unreasonable
then they would lose the allies they needed to succeed. From 1950 on,
the NAACP and Powell routinely proposed, but selectively lobbied for,
desegregation amendments to legislation. After winning Brown v. Board of
Education in 1954 the NAACP sought to have the decision implemented.
The Eisenhower administration was repeatedly evasive on implementa-
tion and suggested that the remedy was with the Court or with Congress.
It was at that point that the NAACP decided to demand Congressional
support for the desegregation of schools with the Powell amendment,
which they backed with extensive lobbying. Pro-segregationists claimed
that the Powell amendment would kill the school aid bill, and School-
Aider liberals tended to believe this claim. There was quite a contro-
versy among northern liberals. The NAACP politely yet persistently re-
quired that its friends rally to the Powell amendment. What motivated
their stand? According to Hamilton (1991, 234), Mitchell, the NAACP
strategist, wanted a public recorded vote of who was willing to stand
up for civil rights, and if the Powell amendment had been withdrawn
then there would have been no such record. I would add that if, as was
likely, the NAACP won a costly-to-attain majority on desegregation in
the Congress, then from that point forward everyone would be on notice
that school desegregation was nally onto the national agenda. One rep-
resentative said, during the 1956 debate, This is the day each Member
pays a price for his civil rights vote. He has to decide whether the indi-
vidual welfare of 15 million Negroes is more important than a vague risk
of losing a $400 million federal plum (Brownson (R-IN), Congressional
Record, July 3, 1956: 11766). I surmise that the NAACP expected Eisen-
hower Republicans to help pass the school-aid bill in 1956 and were as
surprised as everyone else by the outcome. One NAACP board mem-
ber who publicly supported the Powell amendment circulated a caustic
memorandum to other members of the board. He wrote that We have
been made to look like political suckers and amateurs because passing
the Amendment was exactly what the enemies of Federal aid for schools
wanted (Hamilton 1991, 235). Nevertheless, the NAACP did demon-
strate a costly majority with the Powell amendment in 1956, the rst civil
rights bill since Reconstruction was passed in September 1957, a few
weeks later President Eisenhower felt compelled to order federal troops
to Little Rock, Arkansas to enforce a federal court school desegregation
order, and the great civil rights revolution was underway. In the larger
context, the Powell amendment in 1956 was instrumentally rational for
the NAACP. It would be desirable to examine NAACP documents for
details of their strategy in 1956 and 1957, but that is not possible in this
project. I do not have any direct information about the NAACPs stand
in 1957, but the adroitly rational nature of its strategy can be inferred
from the remainder of the school-aid story.
If the incomplete-information outcome in 1956 was a tragedy, the
complete-information outcome in 1957 was a farce. The Krehbiel
and Rivers model would predict that in circumstances of more com-
plete information most of the Northern Democrats (those we labeled
Powellites and School-Aiders above) should strategically vote against any
desegregation amendment, and that prediction is exactly conrmed by
1957s events. The alternative hypothesis of Denzau, Riker, and Shepsle
(1985), that the Powellites were constrained by irrational constituents, is
thus utterly demolished (it is distressing that these authors do not men-
tion the 1957 vote directly contrary to their hypothesis even though at
least one of themis aware of it). The Republicans rejected the Democratic
formula for the level and distribution of funds in 1956, and so a smidgen
of uncertainty remained for the Northern Democrats: would enough Re-
publicans support their own Administrations school-aid proposal? After
all, three weeks before the 1957 vote Vice President Nixon addressed
the National Education Association, and rather confusingly stated that
although President Eisenhower strongly urged adoption of the school-aid
bill there was virtually no chance that it would pass (Fine 1957). It is not
apparent when reading the record of the 1957 debate forward innocently,
but reading it backwards, with knowledge of the denouement, it is plain
that the Northern Democrats, hoping for success but expecting failure
on school aid, are full of mischief and mockery towards their Republican
colleagues.
Eisenhower won the 1956 election and on January 28, 1957 repro-
posed school-aid legislation. The legislation came to the oor of the
House on July 23, 1957. The 1957 bill was a 5050 compromise be-
tween Democratic and Republican positions on level and distribution of
funding. Representative James Roosevelt (D-CA), liberal light and son of
President and Mrs. Roosevelt, seemed to be central to the formulation
and declaration of Northern Democratic strategy. Shortly after the bill
arrived in the chamber, Roosevelt, who had spoken and voted for the
Powell amendment in 1956, took the oor, indicated that Representative
Powell was absent from the House, and declared that, We believe that
this year there should be a clear-cut determination of the fundamental
issue of aid to school construction, and that it should not be clouded by
another civil-rights debate . . . We hope that no one on either side of the
aisle will force the committee to consider any amendment which would
distract in any way from the fundamental purpose of the bill (Congres-
sional Record, July 23, 1957: 12482). Representative Diggs (D-MI), one of
the three African-American members of the House, conrmed the po-
sition of Roosevelt, and announced that he would vote present on any
Powell-type amendment. He said, This is not in any way to be construed
as a retreat from our advocacy and support of the principle enunciated
by the 1954 Supreme Court decision. It is rather to be interpreted as a
strategic withdrawal from using the present proposed school-construction
measure as a vehicle to supplement that decision. . . [Americans] can ex-
amine the record after the vote on this bill and determine who is for
and who is against this proposal without any complications (emphasis
added, ibid., 12483).
On the second day of debate, one Republican after another came to the
oor to denounce federal aid on the grounds of economy. In response,
Democrats gleefully quoted the 1956 Republican platforms endorsement
of school construction aid and quizzed equivocal Republicans as to the
Presidents position on the measure before the House. Udall (D-AZ),
who sat on the committee of jurisdiction, pressed, Is the President back
of the Committee bill or not? Because if he is not, then the work of the
committee is a shambles, and none of us knowwhere we are. We are adrift
here (Congressional Record, July 25, 1957: 12723). McGovern (D-SD,
12722) poured on some delicious political rhetoric from the sidelines:
I have often wondered how it is possible for a man in public life to be so popular
with such a variety of people as is this man from Abilene [Eisenhower, aka Ike].
I think, however, after listening to the explanations of the Presidents position on
Federal assistance for the public schools that I know why everybody likes Ike. It
is simply this: Ike, himself, likes everybody so well that he embraces with equal
good humor all possible sides of issues on which there are sides to embrace.
Those who favor Federal aid to education. . . are sure that Ike agrees with them.
Those who are opposed to Federal aid are equally sure that the President is
opposed, or at least lukewarm, about bringing the Federal government into this
eld. . . Ike . . . no sooner signals with the left-turn indicator than we notice that
the right-turn indicator is also blinking. Just about the time we wonder whether
the Presidential car is swinging right or left, the brakes go on and we are left dead
center in the middle of the road. Little wonder that even sophisticated Washington
reporters get into trouble when they try to follow the Presidential car too closely.
Congressional Quarterly Almanac (1957, 592) devoted a boxed aside to
the vexing question of Eisenhowers true position on the subject, and
in his rst news conference after the 1957 defeat of the school-aid bill
Eisenhower professed that he had never heard that Democrats were
willing to compromise on the legislation.
Representative Powell was conveniently on the Riviera during the de-
bate (reportedly, Congressional Record, July 25, 1957: 12735), and it
was obvious that no other Democrat would be offering a desegregation
amendment. On the third day of debate, Stuyvesant Wainwright (R-NY),
who had voted for the Powell amendment and against nal passage of
school aid in 1956 (YN ), moved a Powell-type amendment (Congres-
sional Record, July 25, 1957: 12482). The Democrats fell in line, and one
after another of the committee Democrats declared that the Wainwright
proposal was a killer amendment and demanded a strategic vote for its de-
feat. The Democrats added some scarcely believable justications, that
the situation was different in 1957 than in 1956, that the Wainwright
amendment was decient in craftsmanship, and so on, and were taunted
with charges of inconsistency and hypocrisy by some of the Republicans.
Wainwright, incidentally, said that although he was very much against the
proposal now on the oor he would vote in favor of a bill that contained
the original Eisenhower formula for level and distribution of aid, and he
explained that he had been sincere in voting for the Powell amendment
but against nal passage in 1956 (12599). This is worth noting because
conservative Republican Wainwrights YNvote is cited as prime evidence
of strategic voting in 1956 (the point related but not endorsed by
Krehbiel and Rivers 1990, 564). Allegedly, Wainwright was a hypocrite in
1956 for promising, in response to a query from Powell, that he would
vote for the Powell-amended school-aid bill to demonstrate his respect
for the civil-rights cause. Actually, Wainwright in 1956 promised to vote
for a Powell-amended school-aid bill only if it contained the Republican
distributional formula; the nal 1956 bill did not contain the Republican
formula, and so he voted against it (Wainwrights 1956 statement is a
bit muddy in isolation but if one reads the surrounding debate it is clear
that his promise was contingent on the nal bill containing the Repub-
lican aid formula, Congressional Record, July 3, 1956: 11758). In 1957,
Wainwright proved he was true to his word; when a clean proposal con-
taining the Republican distributional formula came to the oor he voted
in its favor.
Then something very peculiar happened. The killer Wainwright
amendment came to a voice vote and the Chairman announced that the
nays won the WainwrightPowell amendment was defeated, temporar-
ily. A teller vote (a fast counting of votes without recording names) was
demanded, and the outcome was reversed: there were 136 yeas for the
Wainwright amendment and 105 nays against it. Notice that only 241
out of 426 representatives recorded votes on the amendment. We know
from a news report (Morris 1957a) that only about a dozen Democrats
voted for the Wainwright amendment, that the remainder voting yea
for the Wainwright amendment were Republicans, and that Southern
Democrats ed to the cloakrooms, strategically absenting themselves
from the vote so the killer amendment would pass (if the Southern
Democrats had remained and voted sincerely against it, the Wainwright
amendment would have failed).
Then Ayres (R-OH, and a YN voter in 1956) offered a substitute bill
containing the original Eisenhower formula for level and distribution of
funding, which would also have the effect of striking the Wainwright
Powell amendment, and he promised that if his substitute were adopted
no new Wainwright amendment would be offered by Republicans.
The Northern Democrats swiftly rallied to the Ayres amendment and
promised to support it, even though they had gone from 5050 compro-
mise to 1090 compromise on the contents of the bill. For a moment
there was a sliver of hope for the legislation, because the bill under con-
sideration favored the south distributionally and lacked a desegregation
clause. Then Smith (D-VA, and an NY voter in 1956) moved to report
the bill from the Committee of the Whole to the House with a recom-
mendation that the enacting clause of the legislation be stricken. The
motion to strike takes precedence, usually only stated formally so as to
obtain the right to speak, but, Smith continued, he really meant it on
this occasion. This was a motion to kill the legislation with no further
deliberation, and it succeeded, Q > O
. A yea vote means to kill the bill

and a nay vote means to support the original Republican proposal with
no desegregation amendment (Congressional Record Roll-Call Vote 154:
Congressional Quarterly Almanac Vote 56). The recorded votes were 208
yea (for Q) and 203 nay (for O
). Wainwright kept his promise to support

a clean bill. Of the 208 votes to kill school aid, 97 were from Democrats
and 111 were from Republicans. Of the 203 votes to save school aid,
126 were from Democrats and 77 were from Republicans. Bills usually
dont come to the oor unless they are expected to pass, so why did this
one come to the oor? Representative Martin (R-MA), the only one of
the three Republican leaders to vote to save school aid explained, One
group wanted to bring it out because this was the time to kill it. The other
wanted to put the Republican party on the spot (Morris 1957b).
Calvert and Fenno (1994) offer a brilliant analysis of a 1986 US Sen-
ate Resolution providing for television coverage of Senate proceedings.
The complete-information model of Austen-Smith (1987) showed that
strategic voting and sincere voting are observationally equivalent, in other
words that strategic voting is unidentiable. The Calvert and Fenno
model allows for probabilistic beliefs about the agenda sequence and
about the distribution of preferences in the chamber. If legislators learn
more about the distribution of preferences in the period between mak-
ing proposals and voting on them, then there is a chance for identiable
sophisticated voting. They claim to identify 22 sophisticated votes on
an amendment to the television-coverage resolution, but an attempt to
thwart the majority through agenda manipulation failed. They argue that
senators possess a clear consciousness of the possibility of agenda ma-
nipulation and sophisticated voting, even if it is infrequently observed in
actual practice. Other than a general statement of faith that cycles are
theoretically pervasive on distributional issues, the authors do not claim
that harmful manipulation is frequent. Indeed it is instructive that their
effort to demonstrate an instance of identiable sophisticated voting was
forced to rely on an example that lacked a harmful outcome.
Rikers awed interpretation of the Powell amendment, by the way,
is one of the major examples intended to convince us that democracy
is meaningless: considering the real-world examples I have offered, it
does seem likely that strategic voting occurs quite frequently. If it does,
I conclude that the meaning of social choice is quite obscure (Riker
1982, 167). The examples are all we have, and if the examples fail then
the conclusion fails. It turns out that Riker is only able to demonstrate
an instance of harmful manipulation by mistakenly assuming irrational
actors. He makes the same error in his study of the Depew amendment,
the topic of the next chapter.
10 Assuming irrational actors: the Depew
amendment
Introduction
Riker wants to demonstrate, with a story about the Senate deliberations
on the 17th Amendment to the US Constitution, that manipulation by
introduction of new alternatives is typical of democratic politics. The
17th Amendment would establish the direct election of US senators.
Riker claims that in 1902 Chauncey Depew, a senator from New York,
added a voting-rights rider to the direct-election proposal that introduced
a disequilibrium. Depew is alleged to have contrived a cycle that pre-
vented action on the direct election question for nine years, until 1911.
I provide historical background on the election of senators controversy.
I offer an alternative analysis, which shows that there was no cycle in
1911 and hence no cycle in 1902. The 17th Amendment failed to pass
the 61st Senate early in 1911 due to lack of sufcient votes for either
the version with another voting-rights rider or the version without such a
rider, not due to a cycle. The 17th Amendment passed the 62nd Senate
later in 1911 due to the election of new senators who supported the
Amendment. Altogether, I identify 11 errors of fact and interpretation in
Rikers account.
Rikers tale
Although there may be equilibria in politics, they are extremely fragile
to manipulation by the introduction of new alternatives, because of the
McKelveySchoeld ndings about the indeterminacy of outcomes in
multidimensional issue space, according to Riker (1982, 192). His expli-
cation of the US Senate vote in 1911 on the topic of the direct election
of senators, he says, is intended to illustrate that fragility. The US Con-
stitution originally provided that the legislatures of each state appoint
two senators as the representatives of the state in the US Senate. The
17th Amendment to the Constitution provides that senators are elected
by the people of their state, just as representatives are elected by the peo-
ple of each Congressional district. The controversy over direct election
217
of senators was originally one along a single dimension with an equilib-
rium point, but clever parliamentary tacticians invoked at least two other
dimensions, one of party loyalty, the other of racism, in order to destroy
the equilibrium and generate cycles to their advantage, he says.
Beginning about 1890 a coalition of forces emerged demanding a con-
stitutional amendment providing for the direct election of senators. The
US House of Representatives passed several resolutions to that effect,
each of which quietly died in Senate committee, because senators selected
by state legislatures were reluctant to change the rule to one of election
by the people. But, says Riker, by 1902 political pressure was such that
the Senate could no longer ignore the issue. A Republican senator from
New York, Chauncey Depew, disrupted the equilibrium outcome in fa-
vor of direct election by proposing in committee an amendment to the
direct-election proposal that would establish a uniform national standard
for voting rights.
Depews motive, says Riker, was to divide the proponents of re-
form. The Peoples Party, various progressives, and Western Republicans
wanted direct election due to populist sentiments among their constituen-
cies. Self-proclaimed white-supremacist Democratic senators in the south
wanted direct election, because of white populism and because the sys-
tem of primary nominations in the one-party south was yet another way
to disenfranchise AfricanAmericans. Depews proposed voting-rights
amendment in committee threatened those voting rules and practices
in the south designed to deny the vote to AfricanAmericans. The South
would not buy direct election of senators at that price, so Depew had
single-handedly blocked the Constitutional Amendment.
The issue did not get to the Senate oor in 1902, Riker tells us, but it
did nally in February 1911 (numbered and italicized statements mark
erroneous readings of the historical record), and (1) This time the pro-
gressive proponents of reform protected their Southern allies against the DePew
maneuver by adding a proviso to guarantee white supremacy (Riker 1982,
194). (2) The opponents of direct election moved to delete this sentence (1982,
192). (3) The motion to delete was known as the Sutherland amend-
ment, and it was a negatively stated version of the DePew amendment (1982,
192). Thus, there were three alternatives:
r
S, for South, the proposal to provide for the direct election of US
Senators as reported from committee to the oor of the Senate.
r
N, for North or for No Change (to be made clear shortly), proposal S
as amended by Sutherland; white supremacists claimed that N was a
threat to their voting practices and peculiar way of life.
r
Q, for the Status Quo, no proposal to provide for the direct election of
Senators.
The Depew amendment 219
Because it was a proposal to amend the US Constitution, a successful
resolution would require a two-thirds favorable vote. The vote on the
Sutherland amendment was between N and S, and N won on a majority
vote of 50 yeas to 36 nays (N> S). The vote on the amended resolution,
between N and Q, obtained 54 yeas and 34 nays, failing to attain the
requisite two-thirds supermajority and so the resolution failed (Q > N).
Further, all 54 who voted for Nover Qmust presumably have favored S
over Q, Riker contends. Also, at least ten Southern Democrats who voted
for Q over N presumably favored S over Q. Together that makes 64 out
of a total of 86 or 88 senators favoring S over Q, over two-thirds and
satisfying the supermajority requirement, and thus the collective choice
if it had come to a vote would have been S > Q. We have N > S, Q > N,
S > Q, or N > S > Q > N, a cycle. Hence a small minority of 24 won by
generating a cycle arising from the introduction of a new alternative in a
newdimension, showing the empirical relevance of McKelveySchoeld.
He adds in a footnote, (4) But in June 1911, the Democrats had a clear
majority and defeated the Sutherland amendment, so the constitutional amend-
ment passed. [(5)] Thus, the cycle was broken, but it had lasted more than
ten years (1982, 287).
A later version of the tale, after introducing the Condorcet paradox of
voting, continues by informing the student that:
When tastes are circular (and if there are three or more alternatives and two or
more voters, then almost always tastes are at least potentially circular), then the
outcome depends as much on the procedure of amalgamation as on the tastes
of the participants. Hence it is always possible to manipulate the outcome by
manipulating the agenda. (emphasis added, Riker 1986, 1011)
The hurried or untutored reader would conclude from this that manipu-
lation by agenda control is possible in all circumstances. As we have seen,
and as Riker himself conceded except for major issues, cycles and the
attendant possibility of agenda control are naturally rare. Note the wob-
bling from almost always and at least potentially to always. Similarly, in the
second version of the tale, we nowlearn that it was conceivable unlikely
but conceivable that a proposal for direct election without the Depew
amendment would have passed in 1902 (1986, 14), in other words, a cy-
cle in 1902 was unlikely. In 1902, the majority Republicans in the Senate
would favor voting rights for AfricanAmericans:
Hence, there was, possibly, a cycle: The DePew amendment beat the Constitu-
tional amendment which perhaps beat the status quo, which in turn beat the
DePew amendment . . . It may seem to some readers that this is a rather fanciful
tale to tell about a motion that never even got to the oor of the Senate. But we
can be pretty sure it is just about true because the Constitutional amendment did
get to the oor nine years later. (emphasis added, Riker 1986, 15)
We glide fromunlikely to true quite quickly! He continues with a repetition
of the tale of 1911; the argument is that since there was a cycle in 1911
there was probably one in 1902. In 1911 we have 50 to 36, N > S, 54 to
34, N < Q as in the rst telling. In the second telling we have the 54 (or
perhaps 53 of them) who voted for N > Q favoring S > Q, and this time
8 (rather than 10) Southern Democrats who would have voted for
S >Qif offered the choice, generating the same cycle as in the rst telling.
Riker adds a new inference of preference orders for four groups in
1911. If a senator voted yea for the Sutherland amendment (N > S)
and yea for nal passage (N > Q) we call him a YY voter, and so forth.
(6) The 20 Northern and Border Democrats and 8 Western Republicans
who voted against Sutherland but for nal passage (NY) must have or-
dered S >N>Q, according to Riker. (7) The YY voters, 25 Republicans
and one confused Southern Democrat who voted for both the
Sutherland amendment (N > S) and for nal passage (N > Q) could
have ordered either N > S > Q or N > Q > S. (8) These 25 Republicans
were exactly those who were being manipulated (Riker 1986, 16). (9) They
wanted the Constitutional Amendment for direct election of senators, but
they were constrained by their identication as Republicans to vote for the
Sutherland amendment. (10) Thus, they must have ordered N > S > Q
(and not N > Q > S). Riker needs for these 25 Republicans to be N >
S > Q; if about eight or more of them were N > Q > S, then he would
have no cycle and his larger argument would collapse. Finally, (11) A
few months later Democrats, now with an absolute majority, could prevent the
DePewSutherland maneuver. Since over half of Republicans were also in favor
of direct election, it was then easy to pass the Seventeenth Amendment, even
without the protective clause (Riker 1986, 17).
The rst thing to notice is that, exactly as with his rst version of the
Powell amendment, Riker is unaware that his supposedly rational choice
account assumes irrational voters. Why? He labors to convince us that
25 Republicans were N > S > Q; if that is so then he gets his cycle.
Recall that there is no incentive to vote strategically at the last stage of
an agenda sequence. There are two possible last stages, S against Q or
N against Q. The rst stage prior to either of these is a vote between N
and S. Riker tells us that if it had come to a vote S would have beaten Q
under the two-thirds rule, and as for the other possibility we know that
the actual vote was Q > N. In the rst stage, the 25 Republicans vote be-
tween N and S. If they vote for their rst-ranked choice, N, then N > S,
and the contest becomes one between N and Q, which Q wins, but this is
the 25 Republicans last-ranked choice. If, however, at least 8 out of the
25 Republicans would have voted for their second-ranked choice S in the
rst-stage contest between Nand S, then S would have won; S then would
have beaten Q in the last stage, and the 25 Republicans would have at-
tained their second-ranked choice and avoided their last-ranked choice. If
Rikers story were true, thenthe 25 Republicans votedirrationally for their
last-ranked choice over their second-ranked choice. Clearly, something
must be wrong, again.
Historical background
What is wrong is an egregious misreading of the historical record. Haynes
(1938) is the main source on the subject. Senatorial selection by state leg-
islatures was tainted with abuses, and dissatisfaction steadily increased.
Even after Congress regulated the process in 1866, nearly half the states
experienced serious deadlocks in choosing a senator; the multiplicity of
candidates sometimes made the outcome no better than chance; bribery,
riot, assault, and martial law were not unknown; among other abuses
(95). Earlier in the nineteenth century sporadic proposals for direct elec-
tion were made, and they began to increase in the 1870s. By 1893, the
House passed a joint resolution with the requisite two-thirds approval,
calling for direct election. AConstitutional amendment requires approval
by two-thirds of the House, two-thirds of the Senate, and then three-
fourths of the state legislatures. If Congress fails to act, two-thirds of
the state legislatures can call for a Constitutional convention; however,
such a convention would not be constrained to consider a single topic,
raising serious apprehensions about utilizing this alternative. The House
passed similar resolutions in 1894, 1898, 1900, the margin of approval
ever increasing to unanimity in 1902 (there were abstentions). In 1896,
the Senate Committee on Privileges and Elections favorably reported
a direct-election amendment, but it never came to a vote. More than
two-thirds of state legislatures passed memorials and petitions urging
Congress to propose an amendment. There were no proposals from the
House in the 58th, 59th, and 60th Congresses (19031908). As Byrd
(1988, 398) explained, After 1902, the House stopped trying, for nine-
teen resolutions submitted over three decades had quietly submerged in
the still waters of the Senate Committee on Privileges and Elections.
Riker maintains that it was Depews maneuver in 1902 that halted Con-
gressional consideration of direct elections. The standard story, however,
is that there were not enough votes in the Senate for the proposed amend-
ment to the Constitution. According to Haynes (1938, 98), it would
have been useless to force a new joint resolution through the House
as long as opposition within the Senate was presenting an impassable
barrier.
Thwarted in Washington, the movement found devices at the state
level. A partys state convention might nominate a candidate, although
this is far from direct election. In the 1890s, many states turned to the di-
rect primary election, and the practice in the one-party South became for
the legislature to endorse, often unanimously, the winner of the primary
election even if it had been hardfought. In the West, where there were two
parties, Oregon developed a system that after 1909 was rapidly adapted
by many states. Those running for the state legislature were permitted to
include endorsement of either Statement No. 1, pledging the legislator to
always vote for the peoples choice for US senator regardless of individual
preference, or Statement No. 2, stating that he was free to disregard the
peoples choice. Further, a petition was circulated among Oregon citizens
whereby each pledged not to vote for any candidate who did not endorse
Statement No. 1. The incentives for politicians are clear. In 1909, the peo-
ples choice for senator was Chamberlain, a Democrat, and the people
also elected a majority of Republicans in each house of the state legisla-
ture The pledges held and Democrat Chamberlain was selected by the
Republican legislature, and thus the Oregon system passed the acid test.
By the end of 1910, 14 out of 30 newly elected US senators were desig-
nated by popular vote, not to mention the incumbents previously selected
by popular designation. Out of 46 states, 29 had already instituted the
direct Senatorial primary election most with Oregon-type constraints on
legislatures, and another 8 allowed for expression of popular preference
in primaries (Hoebeke 1995, 150). Meanwhile, fresh Senatorial election
scandals arose in nonreformed states.
One short of the requisite number of state legislatures had applied
to Congress for a Constitutional convention, which may have motivated
some diehards in the Senate to act on an amendment so as to prevent a
convention, but this is speculation, as the dangerous call for a convention
was likely a bluff (see Caplan 1988, 6165). Haynes, a professional po-
litical scientist whose major works were on state selection of senators and
two volumes on the history and practice of the US Senate, believes that
the most important factor in explaining the emergence of the issue in the
Senate was the increasing number of senators who had been designated
by popular vote. Not one of themcould antagonize the proposed amend-
ment without seeming to stultify himself and to affront his constituency
(1938, 107). A similar psychology aficted the older Republicans who
opposed reform. They had been selected by state legislatures, and how
could they repudiate the procedure that had placed them at the pinnacle
of the nations esteem (there is some testiness on the point from them in
the record, e.g., Congressional Record, January 24, 1911: 1339)? Borah,
the most tenacious champion of reform, had been defeated by the Idaho
legislature in 1903, but in 1907 was designated by popular election and
became a US senator. I have great affection for the bridge which car-
ried me over he said (Haynes 1938, 108). Haynes does not say that
reform was thwarted by Depew in 1902, rather he refers to opposition in
the Senate from 1902 to 1911 (106) and change by replacement in the
Senate (107). Indeed, no author except Riker considers, or even men-
tions, any 1902 DePew maneuver as a factor in deterring Senate vote
on the issue until 1911 (Byrd 1988; Caplan 1988; Haynes 1938; Hoebeke
1995; King and Ellis 1996; Kyvig 1996; Zywicki 1994).
To be sure, the old guard in the Senate did work to thwart the re-
formers, not that hard a job given their seniority and the supermajority
requirement, but as the old guard was replaced by senators pledged to the
reform a breakthrough was bound to happen. Bristow (R-KS) recently
become a senator by virtue of designation by popular election, offered a
reform resolution in December 1909, which never got out of committee,
and his discharge motion was obstructed. But early in the next session,
January 1911, Borah favorably reported out of the Judiciary Commit-
tee the earlier Bristow proposal with a new committee amendment. The
committee amendment, however, originating from Rayner (D-MD), was
of great consequence. The effect of Bristows proposal was only to change
the method of selection from state legislature to state popular election.
Rayners committee amendment added a provision prohibiting federal
regulation of Senate elections, in other words, to put in the Constitution
a provision forbidding regulation of the Souths white-supremacist voting
practices with respect to choosing senators, and if that were successful a
movement to remove federal right of control over elections to the House
would likely follow.
At the end of the Civil War the victorious North enacted the 13th
Amendment to the Constitution, forbidding slavery; the 14th Amend-
ment, which guaranteed due process of law to all persons in the United
States, among other things; and the 15th Amendment, which provided
that The right of citizens of the United States to vote shall not be de-
nied or abridged by the United States or by any State on account of
race, color or previous condition of servitude. The Civil War was fol-
lowed by a decade of Reconstruction, which involved military and polit-
ical rule of the South by Republican forces. The South resisted northern
efforts at reform and eventually drove AfricanAmericans from partici-
pation in government by terrorist methods. Gradually, the vicious forces
in the South regained their power, and cut down northern institutions
such as active government and public education, and around the turn
of the century the movement culminated in laws requiring strict segre-
gation of the races, and imposing literacy tests, poll taxes, and all-white
primary elections designed to disenfranchise AfricanAmericans, setting
in place the segregationist South that endured until the 1960s. This was
the venomous atmosphere in which the Senatorial debate on voting rights
took place.
In 1890, as the white supremacists began to seek legal dress for their
segregationist crusade, the Republicans in the Senate sought to establish
federal regulation of elections in order to enforce the 14th and especially
the 15th Amendments, and although they had a majority in the Sen-
ate they were thwarted by the most impassioned libuster in Senate
history (Hoebeke 1995, 163). The southerners labeled this the Force
Bill, and advertised that it foretold a return to hated northern military
occupation of the south. Meanwhile, the Republican Party was weakened
nationally by corruption scandals and the national Democratic Party re-
gained some strength. The rst House Joint Resolution calling for direct
election of senators in 1894 contained a provision supported by the ma-
jority Democrats in the House and opposed by its Republican minority
(Voteview 53rd House, Roll-Call #293, July 20, 1894).
1
The provision
denied to the federal government the authority to regulate election of
senators, at best an undermining and at worst an implicit repeal of the
14th and 15th Amendments. Only some reform proposals contained this
extraneous provision: widespread public opinion favored direct election
but only the minority of white supremacists sought to insert race into the
controversy.
To remind the reader of the rancor and the righteousness associated
with that extraneous issue of Negro voting rights in the South, I excerpt
from the debate on the 17th Amendment remarks by Bacon (D-GA):
there is no question concerning public affairs . . . in which the people of the South
are so vitally interested. . . as their determination to preserve white supremacy,
andit is no use to mince words about it . . . we have fought the battle . . . through the
darkest night through which a people ever passed. We have rescued our civilization
by the sacrices and the trials which we then endured. We have not only rescued
the civilization of the South, but we have rescued this entire Nation from the
destruction of civilization which would have undoubtedly ensued if a Haiti had
been made of the South. With civilization and order overthrown in the South,
the deadly poison would have extended to the entire country.
(Congressional Record, February 27, 1911: 3532)
Percy (D-MS) explained that in his county there were 5,000 white people
and 44,000 negroes, and then boasted that:
the framers of the constitution of 1890 in Mississippi met to frame a constitu-
tion under which a white mans government could be maintained. . . There was
no concealment about the object sought to be attained. It was to obstruct the
exercise of negro suffrage to the point where it would not be a menace to the
government. That object was avowed. . . Restrained by the Federal Constitu-
tion from discriminating against the negro race, the convention discriminates
against its characteristics and the offenses to which its members are prone.
(Congressional Record, February 27, 1911: 3541)
The young Depew was an elected Republican ofcial during the Civil
War, and his response to the southern maneuver was:
This resolution virtually repeals the fourteenth and the fteenth amendments
to the Constitution. It validates by constitutional amendment laws under which
citizens of the United States, constituting in the aggregate more than one-tenth
of the electorate, are to be permanently deprived of the right of suffrage . . . when
it comes to deliberately voting to undo the results of the Civil War, when it comes
by constitutional amendment to permanently taking from 10,000,000 people the
rewards of education and intelligence, that reward being in a free government the
right to vote, I can not assent to or be silent upon the proposition.
(Congressional Record, January 24, 1911: 1336)
Although Depew was opposed to popular Senate elections on any ac-
count, and was not above trying to defeat the proposal by offering an
amendment unacceptable to the South, there was no affectation in his
desire to retain the federal [right to] control of elections (Hoebeke 1995,
163).
Riker has it exactly backwards. It was not that the movement for di-
rect election was sailing merrily along until 1902 when Depew cleverly
introduced an amendment designed to split the supporters of direct elec-
tion which sank the ship of reform. Rather, it was the segregationists
beginning in 1894 who sought parasitically to attach their nationally un-
popular cause to the nationally popular cause of direct election. Many of
the northeastern Republican senators were politically formed by the Civil
War. They sincerely opposed direct election from a sense of traditionalist
elitism, and since they fought a bloody war over the matter they sincerely
opposed mistreatment of AfricanAmericans in the South. Naturally,
they would vote against both proposals, whether bundled together or not.
The Republicans in1902 bundledinanextreme northernpositiononvot-
ing rights as a counterweight to Democratic attempts to bundle in an ex-
treme southern position on the same question. Riker does not realize that
there was another alternative available in 1902, or in any other year, and
that would be a joint resolution calling for direct election that would also
preserve the status quo with respect to federal right to regulate elections.
As all commentators except Riker observe, the main problemfor the cause
of direct election in the Senate had been traditionalist opposition by men
who had been favored by selection by state legislatures. Riker presents
no evidence, repeat, none, and there is no evidence in the other sec-
ondary literature nor in the debates of 1911, for the linked propositions
that direct election could have passed the Senate in 1902 and thus that
there was a cycle for ten years (his fth incorrect claim) we shall see that
there were not enough votes for either formof direct-election amendment
(N or S) even as late as February 1911. In 1899, well before the alleged
Depewmaneuver of 1902, a special committee of the Pennsylvania leg-
islature concluded that the US Senate would never act on direct election
until confronted with applications for a constitutional convention from
two-thirds of the states and set about organizing the states to that end
(Caplan 1988, 63). Rikers only support for 1902 is the assertion of a
cycle in 1911. He is aware that in 1911 there was much more support
for reform (1982, 194) than in 1902, but does not notify the reader that
this fact impairs his claim for a cycle in 1902.
Riker is also misinformed as to the content and nature of the var-
ious amendments pertaining to voting rights. Bristows was a neutral,
status-quo proposal with respect to the federal regulation of elections.
That is what went into the Senate Judiciary Committee. What came
out was a proposal to abandon the federal right to regulate the elec-
tion of senators (see Table 10.1). The Bristow proposal was neutral, the
Rayner addition from the Judiciary Committee departed from the sta-
tus quo towards the extreme southern position, and the former Depew
amendment departed from the status quo towards the extreme northern
position. The old Depew amendment was offered but ignored in 1911
deliberations:
The qualications of citizens entitled to vote for United States Senators and
Representatives in Congress shall be uniform in all the States, and Congress shall
have power to enforce this article by appropriate legislation and to provide for
the registration of citizens entitled to vote, the conduct of such elections, and the
certication of the result. (Riker 1986, 14)
Not until the Voting Rights Act of 1964 were measures similar to this
enacted and enforced.
Howdid the Rayner addition come about? Surely after some backroom
negotiations the Bristow proposal ended up in a Judiciary subcommittee
of three members chaired by Borah. Rayner (D-MD) was the swing vote
and insisted upon removal of federal right to regulate elections (Kyvig
1996, 212). The price for Borah to get direct-election to the oor was
the addition. This was no mystery. This amending doubtless served to
get for the resolution a favorable report from the committee, but in the
Senate it at once raised a new issue that proved most divisive, according
to Haynes (1938, 109). It was not that opponent Depewstrategically split
Table 10.1. Bristow and Rayner amendments compared
Original Constitution Proposed amendment
Bristow (N) Paragraph 1, Section 3, Article I.
The Senate of the United States
shall be composed of two
Senators from each state chosen
by the legislature thereof for six
years; and each Senator shall
have one vote.
The Senate of the United
States shall be composed of
two Senators from each state,
elected by the people thereof
for six years; and each
Senator shall have one vote.
The electors in each state
shall have the qualications
requisite for electors of the
most numerous branch of
the State legislatures.
Rayner addition
to Bristow (S)
Paragraph 1, Section 4, Article I.
The times, places, and manner
of holding elections for Senators
and Representatives shall be
prescribed in each state by the
legislature thereof; but the
Congress may at any time by
law make or alter such
regulations, except as to the
places of choosing Senators.
The times, places, and manner
of holding elections for
Senators shall be as
prescribed in each State by
the legislature thereof.
Source: Congressional Record, January 21, 1911: 1218.
the ranks of direct-election supporters, rather it was the extreme southern
wing of the supporters promising to bolt the coalition unless they had their
controversial proviso. Carter (R-MT) complained on the Senate oor:
It was manifestly used as a oat to bring the main amendment out of the commit-
tee room. Those who accepted that mode of transportation had more zeal than
knowledge, for if the oat does not serve as a sinker in either branch of Congress it
will surely prove a deadly weight in more than one-fourth of the state legislatures.
(Congressional Record, January 21, 1911: 1218)
Nor, according to Carter, was there any mystery about the import of the
Rayner amendment:
The occasion demands plain speech and forbids evasion. . . The adoption of the
amendment would give substantial though limited national sanction to the dis-
franchisement of the Negroes in the Southern States. In their disfranchisement
we now passively acquiesce, but with this supine attitude some Senators are not
content; they ask us to actually strip Congress of the power to question election
methods and actions in so far as the election of United States Senators may be
concerned. (Ibid.)
Rikers claim (1) that the progressive proponents of direct election pro-
tected their southern allies against the Depew maneuver by adding a
proviso to guarantee white supremacy is incorrect: Rayner, the author
of the proviso, was a border Democrat, not a progressive Republican.
The huge majority of progressive proponents of direct election, such as
Carter just quoted above, detested and opposed the Rayner amendment,
as evidenced by their votes for the Sutherland amendment to void Rayner.
What was the Sutherland amendment? Sutherland (R-UT) was a pro-
gressive proponent of direct election who voted both times for nal pas-
sage of a direct-election resolution. His motion was simply to strike the
pro-southern Rayner addition and thereby return to the neutral Bristow
proposal. Thus, Rikers claim (2) that it was the opponents of direct elec-
tion who moved to delete the Rayner amendment is as false as it could
be, as it was the proponents of direct election who did so. Further, Riker
erroneously claims (3) that the Sutherland amendment was a negatively
stated version of the Depew amendment. This is not so. Look at the ac-
tual language of the amendments displayed in Table 10.1. The Depew
amendment was the more pro-northern position on federal regulation
of elections, the Sutherland amendment (or Bristow proposal) was the
neutral position that left the status quo undisturbed; and the Rayner
amendment was the more pro-southern position. Both the Depew
amendment and the Sutherland amendment were proposed on the Senate
oor; they were not equivalent, the rst was ignored in 1911, and the sec-
ond was passed. This is manifest in the record, and not just by implication.
Rayner (D-MD) warned that the Depew amendment goes much fur-
ther than the Sutherland amendment, and it goes much further than
the force bill attempted to go (Congressional Record January 20, 1911:
1163).
The Sutherland amendment passed by a healthy majority, 50 yeas to
36 nays (N > S), but then the restored Bristow proposal failed to get
the requisite two-thirds majority in the 61st Congress. This is suppos-
edly the great manipulative delay due to a contrived cycle. Notice, how-
ever, that unlike usual legislation that Congress may repeal at any time
this was a resolution to propose a Constitutional amendment for the
approval of three-fourths of the state legislatures. If a proposal failed
among the state legislatures it would be unlikely that the issue would
be considered again soon; if the proposal succeeded among the state
legislatures and became part of the Constitution it would be nearly im-
possible to amend. Thus if a senator believed that the composition of the
Senate were changing in favor of his position, he would be better off to
delay the vote to a newly elected Congress. Rikers great manipulative
obstruction lasted a scant six weeks, when the issue was revived again
in the 62nd Congress. In the new Congress the Democratic-controlled
House passed the Rayner version of the resolution; a motion in the House
like the Sutherland amendment to strike the southern demand for un-
regulated state election of senators failed, receiving 190 Democrat nays
and 120 Republican yeas, a pure party-line vote (Voteview 62nd House,
Roll-Call #4, April 13, 1911). The House resolution was referred to the
Senate Judiciary Committee which reported it out in half an hour. In
the 62nd Congress it was Bristow who proposed the oor amendment
to strip the resolution of the Rayner provision. The Bristow amendment
in the 62nd Congress was identical to the Sutherland amendment in the
61st Congress.
In the 61st Congress, Bristowhad voted against the Sutherland amend-
ment to restore Bristows original proposal (S > N) and for nal pas-
sage of the direct-election resolution (N > Q). Yet in the 62nd Congress
Bristow proposed the same amendment he had voted against in the last
Congress. This is an important clue. Senators from the Lower South,
smelling defeat of their cause in the air, attempted to bully Bristow for
his inconsistency. Bristows explanation was simple: in the 61st Congress
he had voted strategically. He is worth quoting at length, because the
entire strategic situation is disclosed without any doubt that my interpre-
tation is distorted.
When this [current, 62nd] Congress met I went over the membership of the
Senate as it is now composed and ascertained that 10 of the Senators who had
voted against the joint resolution [N>Q] when the roll was called on February 26
[in the 61st Congress] were not now Members of the Senate, and from inquiry I
was convinced that more than a sufcient number of the 10 new Senators would
have voted for the joint resolution if they had been Members of the Senate to
have carried it . . . the joint resolution would have been adopted instead of having
been rejected. . . I believe . . . it can get more votes in this body in that form [N ]
than in any other form in which it can be presented. . . that it will command more
support from the people of the United States in the form we voted on it [in the
61st Congress, N > Q]. (Congressional Record, June 12, 1911: 1904)
I voted against the Sutherland amendment [for S > N ] in the last session of
Congress because I was advised by the Senator in charge of the joint resolution
[Borah] and by other Senators whose judgment I had condence in [LaFollette?]
that it would be safer with that amendment defeated than with that amendment
adopted. . . I believed at the time I voted against the Sutherland amendment that
the resolution would be stronger if that amendment was defeated. I think nowthat
I was mistaken that it was stronger with the Sutherland amendment incorporated
in it than it would be if not so amended, and it received more votes than it would
otherwise have received. I believe further, it is stronger before the people now, in
that form, than it would otherwise have been.
(emphasis added, Congressional Record, June 12, 1911: 1905)
Bristow also explained that if his neutral proposal on the question of fed-
eral regulation of state election were referred to the states that would occa-
sion controversy only in the (Lower) South, but that if the pro-southern
proposal were adopted that would cause controversy and loss of sup-
port in the much larger remainder of the nation (Congressional Record,
June 12, 1911: 1906). In short, Bristow was saying that he strategically
voted against the Sutherland amendment in the 61st Congress because
he thought that otherwise direct election would not get enough votes
from southern senators. After all, Rayner (D-MD), who had prevailed
on Borah in the Judiciary Committee to include the southern proviso,
had thundered on the oor upon the introduction of the Sutherland
amendment: Suppose that every Southern State is in favor of the joint
resolution the way we reported it, and that every Southern State is against
it the way the Senator from Utah [Sutherland] proposes to amend it
(Congressional Record, January 20, 1911: 1163). But after votes were
recorded on N > Q Bristow realized that Rayner was all thunder and
no lightning: it was not the South but only the Lower South that had
blustered on the issue, and when N came to a vote, the Upper South sup-
ported it over Q. Next, Bristow said, he polled the replacement senators
coming into the 62nd Congress and that further strengthened his belief
that victory was possible for N > Q and so this time around he would
vote sincerely for N.
Bristow was correct. In the 62nd Congress, the Bristow (Sutherland)
amendment passed, 45 to 44; a southern proposal to substitute a less
extreme pro-southern provision failed; and the direct-election resolu-
tion (N) received more than two-thirds approval, 64 yeas and 24 nays.
The amended joint resolution was sent back to the House, which would
not concur. The Senate voted (along party lines) to insist on its amend-
ment and to ask for a conference. The conference committee met sixteen
times without success. Eventually an exhausted House concurred in the
Senate version. In the House the northern Democrats had been support-
ing the southern Democrats against the Republicans on state regulation
of elections; in the end the northern Democrats joined the Republicans
to defeat a southern-Democratic motion to amend the Senate version
(Voteview 62nd House Roll-Call # 139, May 13, 1912). Then the House
voted for the Senate version (N > Q), 238 votes for and 39 votes against,
the scattered nays from Virginia and parts of the Lower South (Voteview
62nd House Roll-Call # 131, May 13, 1912). The state legislatures re-
sponded swiftly, by May 31, 1913 more than three-fourths had voted
their approval and the 17th Amendment to the constitution was adopted.
Only Utah voted against ratication. As the three-fourths threshold
was crossed it was mostly states in the Lower South who had yet failed
to act: Alabama, Delaware, Florida, Georgia, Kentucky, Maryland,
Mississippi, Rhode Island, South Carolina, and Virginia.
Alternative interpretation
The fact that two different Congresses recorded votes on the same two is-
sues gives us an abundance of data points fromwhich to infer preferences
(see Table 10.2).
For those senators participating in both the 61st and 62nd Congresses,
we nd consistency that is almost perfect, but with one major and reveal-
ing exception, we shall see. Senators similar to Depew(who did not return
to the 62nd Congress), 15 Republicans mostly fromthe Northeast, voted
for the Sutherland and Bristow proposals to retain federal right to reg-
ulate elections (N > S), but both times against referring the proposed
amendment to the states (YNYN, Q > N > S). Warren (R-WY), the
only inhabitant of cell YNYY, is a telling case. Right before the vote in
the 62nd Congress he explained that he believed that a popularly elected
Senate would be unwise, but that he must bow to the opinion of the
people in favor of the reform, a pitifully incoherent position probably
inspired by the dramatic electoral defeats of anti-reform Republicans all
around him (see Hoebeke 1995, 184). Most of the various progressives
Border, Midwestern, and Western Republicans 21 of them, voted for
the Sutherland and Bristowamendments and both times for referring the
proposed amendment to the states (YYYY; N > S > Q or N > Q > S).
Democrats from the Upper South, 20 of them, voted against the Suther-
land and Bristow amendments but nevertheless for the amended referral
of direct election to the states (NYNY, S > N > Q). The Insurgent
Republican leaders, LaFollette and Borah, voted with the Upper South
Democrats. Eight Lower South Democrats voted against everything, and
said that they would have voted for direct election if they would have
obtained their Jim Crow proposal (NNNN, S > Q > N).
The Insurgent Republicans were a distinct minority faction in the
Congress from 1909 to 1916. A typical insurgent had migrated to a
western state as a youth, had been a small-town lawyer, had identi-
ed with postbellum Republicanism, but had moved from right to left
with political maturity along with his populist rural constituency. The
insurgents wanted a progressive income tax and an inheritance tax, con-
trol or defeat of the corporate interests, and the extension of popular
government. At the end of the 61st Congress, the established insur-
gent core in the Senate contained Borah (NYNY ), Bristow (NYYY),
Clapp (NYYY ), Crawford (YYYY ), Cummins (NYYY), and LaFollette
(NYNY ); Gronna (NYYY ) was new but core; and Bourne (NYYY ),
Table 10.2. Votes on 17th Amendment compared
62nd YN YY NY NN
Congress:
61st Congress:
YN 15 mostly Warren (YNYY, R-WY)
northeastern
Republicans
YY 21 Border midwestern, and
western Republicans
NY 6 insurgent-Republican 20 Upper South
rank-and-le Democrats 2
insurgent-
Republican
leaders
NN 8 Lower South
Democrats
Key:
Y = Recorded vote, Yea
Y = Paired vote, Yea
0 = Not voting
YYYY means:
Y = Yea vote on 61st Senate Roll-Call #244 (N > S )
Y = Yea vote on 61st Senate Roll-Call #248 (N > Q)
Y =
Nay vote on 62nd Senate Roll-Call # 25 (S > N)

Y = Yea vote on 62nd Senate Roll-Call #26 (N > Q)
etc.
YNYN: Brandegee (R-CT), Burnham (R-NH), Crane (R-MA),

Dillingham (R-VT), Gallinger
(R-NH), Heyburn (R-ID), Lorimer (R-IL), Oliver (R-PA), Page (R-VT), Penrose (R-PA), Richardson
(R-DE), Root (R-NY), Smoot (R-VT), Wetmore (R-RI).
YNY0: Lodge (R-MA).
YYYY: Bradley (R-KY), Briggs (R-NJ), Burton (R-OH), Clark (R-WY), Clarke (D-AR), Cullom
(R-IL), Curtis (R-KS), Dixon (R-MT), DuPont (R-DE), Gamble (R-SD), Guggenheim (R-CO),
Jones (R-WA), McCumber (R-ND), Nelson (R-MN), Nixon (R-NV), Perkins (R-CA), Smith (R-MI),
Stephenson (R-WI), Sutherland (R-UT).
Y0YY: Crawford (R-SD). YY0Y: Frye (R-ME).
NYYY: Bourne (R-OR), Bristow (R-KS), Brown (R-NE), Clapp (R-MN), Cummins (R-IA),
Gronna (R-ND).
NYNY: Bailey (D-TX), Borah (R-ID), Chamberlain (D-OR), Culberson (D-TX), Davis (D-AR), Gore
(D-OK), LaFollette (R-WI), Martin (D-VA), Newlands (D-NV), Overman (NYNY, D-NC), Owen
(D-OK), Paynter (D-KY), Rayner (D-MD), Shively (D-IN), Simmons (D-NC), Smith (D-SC), Smith
(D-MD), Stone (D-MO), Swanson (D-VA), Taylor (D-TN), Thornton (D-LA), Watson (D-WV).
NNNN: Bacon (D-GA), Bankhead (D-AL), Fletcher (D-FL), Foster (D-AL), Johnston (D-AL), Percy
(D-MS), Terrell (NNNN, D-GA). NN0N: Tillman (D-SC).
The motion to adopt the Bristow amendment (N > S, 62nd Senate Roll-Call # 24) tied at 4444 and
the Vice President broke the tie in its favor, making 4544 for passage. Lower Southerners unsuccessfully
objected to the Vice Presidents tie-breaking vote. Lower Southerners also immediately moved a com-
promise that would have voided the Bristow amendment (S
> N, 62nd Senate Roll-Call # 25), which

failed 4346, and it is this vote that is displayed in the table. Adjusting for polarity, the only changes
between the two votes were: Dillingham (R-VT) paired yes for the northern position N on the rst vote
and was recorded in favor of the northern position N on the second vote; Gronna (R-ND) voted against
the northern position on the rst vote, but voted in favor of the northern position on the second vote.
Source: Constructed from votes reported in Voteview.
Brown (NYYY ), and Dixon (YYYY ) were in the insurgent penumbra
(Holt 1967, 3).
The exception to consistency is six Insurgent Republicans who fol-
lowed faction leaders LaFollette and Borah in the 61st Congress and
voted against the Sutherland amendment but for nal passage, but who
in the 62nd Congress departed from LaFollette and Borah (NYNY) to
vote for the Bristow amendment and for nal passage (NYYY). One of
these was Bristow himself, and we have heard his explanation. The other
ve were Bourne, Brown, Clapp, Cummins, and Gronna. These six voted
strategically against the Sutherland amendment in the 61st Congress be-
cause they believed they needed votes from the South for the success of
the direct election cause. After the vote was revealed, however, as Bristow
declared, it was plain that they had enough votes from the Upper South
for success and would be better off voting sincerely to ensure the victory
of Non the next occasion. Most of these six, since they voted strategically,
must have had the preference order N> S > Q. What about the 21 other
progressive Republicans who voted YYYY? Were they N > S > Q or
N > Q > S? Riker gets his cycle by inferring, presumably, that none
of them was N > Q > S. But if they were N > S > Q as Riker wants
them to be, then they should have joined with their fellow progres-
sive midwestern and western Republicans certainly the Bristow six,
and probably Borah and LaFollette and voted strategically against the
Sutherland amendment (NY), especially given that their fellow progres-
sive Republicans who voted strategically failed in their objective to defeat
N in order to avoid Q. Many of the 21 must have been N > Q > S. This
follows from Bristows statement that direct-election would enjoy more
support with N than with S, reiterated by Bristow when a pro-southern
senator (Percy, D-MS) wheedled that together the Democrats and the
progressive Republicans could pass S > Q: The Senator [Percy] is un-
der the impression, Bristow replied, that a majority [50 percent rather
than
2
3
] can pass the resolution. . . the Senators mathematics are badly
wrong (Congressional Record, June 12, 1911: 1905). Riker does not un-
derstand that the preference order that he thinks unlikely, N > Q > S,
makes sense: Bristow himself came close to changing from N> S > Q to
N > Q > S. Why does N > Q > S make sense? If one wants direct elec-
tion, and if one believes that N would attract the votes of three-fourths
of the state legislatures and be adopted, but that S would fail to win the
votes of three-fourths of the state legislatures and fail to be adopted, then
one would rank the issues in the Senate N>Q>S. Ones position would
be: dont send a direct-election amendment to the states unless it is one
that will succeed. Finally, we know that two senators stated on the record
that their ranking was N> Q> S (Curtis, R-KS, and Jones, R-WA, cited
in Haynes 1938, 111).
Table 10.3. Replacement senators, from 61st to 62nd Senate
62nd Senate YN YY NY NN
61st Senate
YN NY, Aldrich CT, Bulkeley (R) to 5 Republican seats
(Y0, R) to McLean (R); to Democratic, 1
Lippitt (R) MI, Burrows (R) Republican stays
to Townsend (R) Republican
YY IA, Young (R) to 4 Republican seats
Kenyon (R) to Democratic, 1
Republican stays
Republican
NY
NN FL, Taliaferro (D) MS, Money
to Bryan (D) (NN, D)
Williams (D)
Key:
YY means:
Y = Yea vote on 61st Senate Roll Call #248
Y = Yea vote on 62nd Senate Roll Call #26
etc.
YN to NY: CA, Flint (R) to Works (R); ME, Hale (R) to Johnson (D); NJ, Kean (R) to
Martine (D); NY, Depew (R) to OGorman (D); VT, Dick (R) to Pomerene (D); WV,
Scott (R) to Chilton (D).
YY to NY: IN, Beveridge (R) to Kern (D); MO, Warner (R) to Reed (D); NE, Burkett (R)
to Hitchcock (D); MT, Carter (R) to Myers (D); WA Piles (R) to Poindexter (R).
Source: Constructed from votes reported in Voteview.
It remains to examine the replacement senators (see Table 10.3).
There were 17 incoming senators. The Republicans lost nine seats, the
Democrats lost none. Three seats retained by the Republicans went from
opposing direct election to supporting direct election of senators. Five
seats lost by the Republicans to the Democrats went fromopposing direct
election to supporting direct election. One seat retained by the Democrats
went from opposing direct election to supporting direct election. Alto-
gether then, nine seats converted to direct election; six seats retained a
commitment to direct election; two seats remained against direct election;
and no seats converted from support of direct election to opposition.
There is a trend towards support of direct election. At the same time, the
supporters of the BristowSutherland amendments for neutrality on the
question of federal regulation of elections declined by a whopping 11. An
enactment of a direct-election amendment had become inevitable, but
the Republicans may have calculated that the 62nd Congress was their
last chance to attain N over S, which may explain their tenacious and
Table 10.4. Rikers inference of 61st
Senate vote on 17th Amendment
N S Q
N

50 54
S 36

64
Q

34 24
Notes:
= winner (it takes >

2
3
to beat Q)
italic = winner by majority rule
Table 10.5. Mackies estimates of distribution of
preferences in 61st Congress
Vote # Label Ranking
YN 24 Northeast Republicans QNS
YY 25 Sincere progressive Republicans NQS
NY 8 Strategic progressive Republicans NSQ
NY 20 Upper South Democrats SNQ
NN 9 Lower South Democrats SQN
triumphant refusal over almost two years to compromise with the House
on the BristowSutherland amendment.
The nal task is to examine whether Rikers assertion of a cycle in
the 61st Congress is correct. He assumes, without mention or justica-
tion, that all voters are sincere. His data and inferences would yield the
pairwise-comparison matrix in Table 10.4. We see that N > S, S > Q,
Q > N; a cycle, N > S > Q > N.
I infer that all voters were sincere except for the Bristow six and per-
haps Borah and LaFollette among the progressive Republicans. Most
of these eight voted strategically S > N, against their sincere preference
N>S, because they mistakenly believed that was the only way they would
get enough votes to avoid their least-favored outcome Q and gain their
second-ranked outcome S. As I have argued, many of the remaining pro-
gressive Republicans were N > Q > S; otherwise they would have voted
strategically with their colleagues. Those who voted YN unambiguously
ranked Q>N>S, and 25 of those who voted NYunambiguously ranked
S > N > Q. If all we knew were the two roll-calls, those who voted NN
could have been Q > S > N, but they said they were S > Q > N and
there is no evident motive for misrepresentation. If they were capable
Table 10.6. Mackies inference of 61st
N S Q
N

57 53
S 29 37
Q

33

49
Notes:

2
3
to beat Q)
of strategic voting, those who voted NY would have been N > Q > S.
Up to eight of those who voted NY were of the type N > S > Q voting
strategically for S > N. Then we simply tally these preferences into a
pairwise-comparison matrix (see Table 10.6).
If the Senate operated by simple majority rule on the question, then
the outcome would have been N > S, N > Q, and Q > S, for an over-
all noncyclical ranking of N > Q > S. Under the rule for referring a
Constitutional amendment to the state legislatures, a two-thirds vote to
overcome the status quo, and if all would have voted sincerely in the 61st
Congress, then the outcome would have been N > S, Q > N, Q > S,
for an overall ranking of Q > N > S. My conclusion, novel to the litera-
ture reviewed, is that there were not enough votes in the 61st Congress
to pass a Constitutional amendment under any circumstances, imply-
ing that it was the replacement of senators in the 62nd Congress that
was crucial to success of the 17th Amendment. In the 61st Congress S
could not have beaten Q, and N was strongly favored over S, but there
were not more than two-thirds who favored N over Q. Rikers belief that
Republican manipulation rather than simple lack of votes caused the de-
feat of direct election in the 61st Congress may just be an uncritical
transmission of Democratic electioneering propaganda on the issue. If
my assertion that many of the 25 sincere progressive Republicans were
N > Q > S is controversial, consider a sensitivity analysis. In order to
avoid Rikers cycle, I need that only ve in that category have the ranking
N > Q > S, and the rest could rank N > S > Q as Riker would have
them. I have already established that N > Q > S is a reasonable pref-
erence order, and shown that it has support in the record. There was
no cycle in 1911, and since the claim of a cycle in 1911 is Rikers only
evidence for asserting a cycle in 1902, his claim (5) fails, that Depew had
contrived a cycle in 1902 that delayed consideration of direct election for
ten years.
Table 10.7. Mackies estimates of distribution of preferences
in 62nd Congress
Vote # Label Ranking
YN 16 Northeast Republicans QNS
YY 22 Strong progressive Republicans NQS
NY 16 Weak PRs & new N.Democrats NSQ
NY 25 Other Democrats SNQ
NN 8 Lower South Democrats SQN
Table 10.8. Mackies inference of 62nd
N S Q
N

47

71
S 40 47
Q 16

30
Notes:

2
3
to beat Q)
How did the replacement of senators from the 61st Congress to the
62nd change the distribution of preference rankings? Here my inferences
are less condent since I do not have two votes on the same issue for the
replacement senators; but the conclusionof no cycle inthe 62ndCongress
is robust to reasonable variation in the estimates of preference rankings.
I assume that all voters voted sincerely (there is no reason or evidence to
assume otherwise) and that where there is a doubt a new senator voted
like his geographic neighbors. That yields the estimates in Table 10.7, and
the labels must change somewhat to reect lack of strategic voting and
the big Democratic gains in the North. That translates into the pairwise
matrix in Table 10.8. That gives us under the actual voting rule N > S,
N > Q, Q > S, or the noncyclical ranking N > Q > S. I assume that
all 22 Strong Progressive Republicans are N > Q > S; and if 10 out of
those 22 have the ranking I assume the conclusion stands. If 11 or more
of the 22 instead rank N > S > Q, then the collective ranking would be
the noncyclical N> S > Q. In neither case is there a cycle. Of course, we
know that N won in the actual vote.
My analysis does not concentrate on the ultimate causal factors behind
the passage of the 17th Amendment. I favor the viewthat it was one step in
the relentless march of democratization, that the Amendment came about
because thats what the people wanted. A second, complementary, factor
was the deadlocks, corruption, and disorder associated with selection by
state legislatures. In an excellent paper, King and Ellis (1996) show that
before adoption of the 17th Amendment in 1913 it was easier relative to
the House delegation to elect a Republican senator, but that after 1913 it
was easier to elect a Democratic senator relative to the House delegation.
The explanation offered is that aggregation by direct election mobilizes
different interests than does aggregation by indirect election through the
state legislature. Of course, the Democrats did more zealously seek the
amendment than did Republicans, and that is the third factor disclosed
by King and Ellis. Zywicki (1994) offers a rent-seeking explanation of the
reform that it allowed special interests such as workers and farmers to
better work their will on the Senate but the data he provides in support
are not conventionally signicant.
Nowwe are in a position to further elucidate Rikers errors. He claimed
(6) that the eight western Republicans who voted NYin the 61st Congress
ranked S > N > Q. However, I have shown that six of these Insur-
gent Republicans including Bristow voted strategically for S > N in the
61st Congress and sincerely for N > S in the 62nd Congress, and thus
that they ranked N > S > Q. Borah, the carrier of the resolution on
the oor was constrained to S by the commitment he made in committee
to get the bill to the oor, and on the oor said he would even support
proposals to strip the Congress of the right to regulate House elections
(Hoebeke 1995, 165), although this is suspicious since not one other
Republican in the Congress held this position. LaFollette, the leader of
the Insurgent Republicans, I surmise was also constrained to S by an
early commitment he made as a factional leader to southern Democrats
to get the bill onto the oor; Stephenson, LaFollettes ideological sidekick
from Wisconsin, never once voted for S. To continue the surmise, in the
62nd Congress LaFollette would keep his own commitment but plausi-
bly claim that his troops were out of control given the contribution of his
partys obstruction of direct-election legislation to heavy Republican elec-
tion losses. That the Insurgent Republicans were strategizing together on
the measure is suggested by Gronnas (R-ND) instant conversion in the
62nd Congress from voting against the Bristow amendment one minute
and in the next minute casting a vote equivalent to favoring the Bristow
amendment with the intended effect of providing its margin of victory.
Riker claims (7) that southerner Clarke (D-AR) was possibly confused
because he voted YY in the 61st Congress against the southern position
(he voted N > S, N > Q). Clarke was quite unusual, but he was not con-
fused, for he again voted YY in the 62nd Congress. Riker claims (8) that
it was the group that I label the 25 sincere progressive Republicans who
were being manipulated by the so-called DepewSutherland maneuver
(9) because they were constrained by their identication as Republicans.
But, as we have seen, there was no such identity constraint on the Bristow
six, the strategic progressive Republicans, nor on prominent progressive
Republicans Borah and LaFollette. So why are some Republicans irra-
tionally constrained by identity and some not? To throw an unjustied
identity constraint into a rational choice model again suggests ad hoc
hypothesis rescue. And how can it be said that the 25 sincere progressive
Republicans were harmfully manipulated, given that they won their top-
ranked alternative in the next Congress? Riker is misled because he does
not detect that it is the Bristow Republicans making the (failed) strate-
gic maneuver and because he does not understand that N > Q > S is a
feasible preference order. That is Rikers tenth incorrect claim, and the
one that gives him his cycle, to assume that all 25 sincere progressive
Republicans presumably had the preference order N > S > Q. I have
shown that there is reason to believe that many did not.
Riker also claims (4) that in the 62nd Congress in June 1911 the
Democrats had a clear majority and were able to defeat the Suther-
land amendment thereby permitting the Constitutional amendment to
pass out of the Senate. This is at wrong. The Republicans were still
the majority party in the 62nd Senate. Although the Senate did not
vote on the Sutherland amendment in the 62nd Congress, they did vote
on the identical Bristow amendment, and they did not defeat Bristow
Sutherland as Riker claims; instead they passed it. Alternatively, Riker
claims (11) A few months later Democrats, now with an absolute ma-
jority, could prevent the DepewSutherland maneuver. Since over half of
Republicans were also in favor of direct election, it was then easy to pass
the Seventeenth Amendment, even without the protective clause (Riker
1986, 17). Again, this is quite wrong. The reader by now will under-
stand that the Depew amendment was not seriously considered in 1911,
and that the BristowSutherland maneuver succeeded and made pos-
sible the passage of the proposed constitutional amendment out of the
Senate. And how did the clause protecting the southerners disappear
from the resolution except for something like the BristowSutherland
amendment?
Riker concludes that the case of the 17th Amendment, is not, I believe,
an isolated example of manipulation, but a typical instance (Riker 1982,
195). His account of the Powell amendment, intended to show harmful
manipulation arising from strategic voting, failed because it assumed ir-
rational actors. His account of the Depew amendment, intended to show
harmful manipulation arising from the contrived introduction of new
dimensions, failed again because it assumed irrational actors, and also
because of egregious misreadings of the historical record. With the next
two cases Riker will attempt to demonstrate a case of grand manipulation.
Beginning from the Wilmot Proviso in 1846 and culminating in the elec-
tion of Lincoln in 1860, he argues, the US Civil War arose in response to
arbitrary manipulation of political disequilibrium. If not argumentatively
then rhetorically, these two cases sustain Rikers hypothesis of pervasive
disequilibrium. The reader is led to believe not only that cycles happen,
not only that democratic politics is arbitrary and meaningless, but also
that the most cataclysmic event in American history is an instance of
such disequilibrium. The lesson is powerful and unforgettable. It is also
wrong.
the Wilmot Proviso
Introduction
I begin with a summary of Rikers two claims of disequilibrium relating
to the politics of the Wilmot Proviso. In 1846 President Polk requested
an emergency appropriation in effect to commence the acquisition of the
northern half of Mexico. The Wilmot Proviso was an amendment moved
by northern Democrats in the House of Representatives to prohibit slav-
ery in the lands to be acquired fromMexico. Rikers rst claimof disequi-
librium is that initially the least-desired alternative of no appropriations
and no antislavery proviso prevailed in the House. His second claim of
disequilibrium is that there was a cycle among legislators preferences
among the status quo, the acquisition appropriation, and the appropri-
ation proposal with the antislavery amendment. Then I present a more
far-ranging and critical review of Rikers second claim and then of his
rst claim. The problem with Rikers second claim is that he misreads
the vote in the Congressional record. A corrected reading shows that
there was no cycle. The problem with Rikers rst claim is that it lacks
context. A more thorough investigation discloses that the appropriation
initially failed not due to some profound disequilibrium, but rather due
to mischiefs and blunders. Finally, I present my own interpretation of the
votes in question. By means of analysis of roll-call votes I am able to of-
fer a fresh interpretation of the Wilmot Proviso controversy in Congress,
including an explanation of how the Proviso was defeated.
Rikers story
Northern Democrats were annoyed by the southern tilt of the Democratic
President Polk, says Riker, and opportunistically reintroduced the slavery
issue into national politics. The motives of the northern Democrats are
evidenced by Polks diary: this remarkably astute observer, intimately ac-
quainted with events and personalities honestly assessed the slavery issue
as opportunism on both sides (Riker 1982, 224). Wilmot did not care
241
about slavery, he only wanted to protect free labor fromslave competition,
and the Wilmot Proviso was devised to protect the northern Democrats
against Whig agitation on the slavery issue, according to Riker. The
Mexican war was underway, and Polk sought a special two million dollar
appropriation in order to bribe the Mexican army to accept a war settle-
ment favorable to American interests. Wilmot moved as an amendment
to Polks appropriation a provision prohibiting slavery in any territory
acquired from Mexico. Almost all northerners voted for the Proviso and
almost all southerners voted against it, whether Whig or Democrat. The
Whigs, by the way, were the more commercial party, and the Democrats
were the more agrarian party, and both parties were bisectional. The
outcome involved disequilibrium in two ways, says Riker.
There are three alternatives under consideration:
OA, the Original Appropriation proposed by Polk
WP, the original appropriation as amended by the Wilmot
Proviso
SQ, the Status Quo, no appropriation.
Riker says that Polk almost certainly ordered OA > WP > SQ, and that
the Senate had agreed to OA in secret session so that it must have pre-
ferred either OA > WP > SQ or OA > SQ > WP. The House did indeed
pass WP, which it must have preferred to the alternatives. Riker recounts
some estimated rankings of factions in the House that are useless to reit-
erate here because, as we shall see, they are mistaken. What he is trying
to establish at this point is that SQ was the least-desired alternative in the
House, which I agree is correct. We can again strengthen Rikers argu-
ment by noting that just as OAwas introduced but before WP was offered,
several motions with the effect of immediately killing OA failed (Congres-
sional Globe, August 8, 1846: 1212; VoteviewRoll-Call Votes, 29th House,
#450, #451, #452). The House passed WP, and WP went to the Senate
where it was libustered by an antiwar northern Whig who top-ranked
SQ, and SQ prevailed due to the libuster. Thus the least-desired alter-
native, SQ, won, clearly an outcome in disequilibrium (Riker 1982,
225).
There was also a second disequilibrium within the House itself, ac-
cording to Riker. First, by revealed vote a majority in the House ranked
WP >OA. The Wilmot amendment to Polks appropriation carried. This
conclusion is correct (but Rikers warrant is not, we shall see). Second,
OA > SQ, by inference, says Riker: its reasonable that the Democrats, a
majority in the House, would support their President Polks request. This
conclusion is correct. As I said, in the August 8, 1846 deliberations lead-
ing up to the Wilmot Proviso several motions to table OA failed; and in
a clean vote on OA
without any Wilmot Proviso on March 3, 1847,

The Wilmot Proviso 243
OA
> SQ by a recorded vote of 115 for, all but two of those from
Democrats, to 82 against including only ten Democrats (Congressional
Globe, March 3, 1847: 573; Voteview, 29th House Roll-Call #637). Third,
by inference, says Riker, SQ > WP: a coalition of rst, all southerners,
Democratic or Whig, would oppose WP because it contained the pro-
hibition of slavery, and second, the northern Whigs would oppose WP
because it contained an appropriation for a war that the Whig Party op-
posed. We have WP > OA, OA > SQ, and SQ > WP, or WP > OA >
SQ > WP, a cycle.
Review
Riker further estimates the ranking of the three alternatives by each of
eight natural factions in the House, which taken together again conrm
the cycle, he says. The ranking estimates, however, are pure fantasy, be-
cause Rikers inference that SQ > WP is directly contradicted by the
revealed votes in the record. Riker wrongly believes that the vote he re-
ports of 79 yeas and 93 nays was on adopting the amendment to the original
appropriation, WP >OA(Display 91, The Vote on the Motion to Lay
on the Table the Motion to Engross the Wilmot Proviso, Riker 1982:
226). In fact, the vote that he refers to was on passing the amended appro-
priation, WP > SQ. There is no doubt that Rikers estimates of rankings
of the eight natural factions is based on mistakenly reading WP > OA
for WP > SQ. His listing of the factions rankings (227) clearly builds
from his Display 91 of the misinterpreted vote. The Display asserts that
93 legislators ranked WP > OA and 79 ranked OA > WP, when in fact
93 legislators ranked WP > SQ and 79 ranked SQ > WP. The estimates
assert, for example, that seven northern administration Democrats rank
OA > WP and 51 northern Free Soil Democrats rank WP > OA, when
in fact the vote he references would have his seven northern Democrats
voting SQ > WP and his 51 northern Democrats voting WP > SQ.
If it is shown by ve revealed votes that WP > SQ, then, I submit,
a weakly warranted inference that SQ > WP must be mistaken. If we
correct Rikers error we have WP > OA, OA > SQ, and WP > SQ, or
WP > OA > SQ, and no cycle whatsoever.
Yes, the great cycle that initiated the disequilibrium that culminated in
the Civil War is all based on a simple misreading of the record. The vote he
reports is not on tabling engrossment of the Wilmot Proviso (WP > OA)
but is on tabling engrossment of the Wilmot-amended appropriation
(WP >SQ). Riker (1982, 290) cites to Congressional Globe, 29thCongress,
1st Session, 1218. He says (225) that the Wilmot Proviso was voted on
eight times on August 8, 1846, but the largest and crucial vote was on a
Table 11.1. Datum and warrant, Wilmot Proviso
Riker R. Warrant Mackie M. Warrant
WP > OA Revealed vote WP > OA Revealed vote
Motion to Lay on the Two voice votes, August 8,
Table the Motion to 1846; one recorded vote,
Engross the Wilmot February 15, 1847.
Proviso, August 8, 1846
(Recorded vote March 3,

1847 for OA > WP)
OA > SQ Inference OA > SQ Revealed vote

All Democrats, 60 percent Three recorded votes,
of the House, supported August 8, 1846; one
the administration recorded vote,
on the war March 3, 1847.
SQ > WP Inference WP > SQ Revealed vote
All southerners, D or Three recorded
W, opposed the WP, votes, August 8,
and northern Whigs 1846; two recorded
opposed the war votes, February 15, 1847.
Full Rank WP > OA > SQ > WP WP > OA > SQ
Notes:
Rikers claim of WP > OA is correct, but his warrant is not. He erroneously believes that
the vote he cites is on WP > OA, when in fact it is on WP > SQ.
In the end seven northern Democrats changed their votes from yes to no on the Wilmot
Proviso (not mentioned in Rikers story), for sound strategic reasons emergent on March 3,
1847.
motion to lay on the table a motion to engross, which was defeated by 79
to 93. There is not any motion to engross nor any vote resembling 79 yeas
to 93 nays on page 1218 of the record. There is a vote to table engross-
ment on page 1217 of the record, with 78 yeas and 93 nays (Voteview 29th
House, Roll-Call #456). Quite clearly this vote is on the whole bill, as are
three further votes with the same effect of endorsing WP > SQ on page
1218 (one unrecorded and two recorded, Congressional Globe, August
8, 1846: 1218; Voteview, 29th House, Roll-Calls #457 and #458). The
whole controversy replayed again on February 15, 1847, when WP >SQ
by a vote of 115 to 106 (Congressional Globe, 425; Voteview, 29th House,
Roll-Call #582). Altogether that makes ve revealed votes showing that
the House of Representatives ranked WP > SQ.
Because I will be doubted, it is probably best to review the parliamen-
tary action, beginning at page 1217. We are in the Committee of the
Whole House. McKay, who is carrying Polks OA (House Resolution
534A), moves a perfected version. Ingersoll moves a substitute for OA
(perhaps to confuse procedurally newcomer Wilmot). Wilmot moves WP,
his amendment to OA. A point of order is made against WP that it is
not germane. The chair overrules the point of order, and his motion
is appealed. The decision of the chair is sustained by a teller vote (no
recorded names) of 92 to 37; this test vote suggests a majority for WP
over OA. Wick moves to amend WP with an extension of the Missouri
Compromise: slavery permitted in the southern portion of the new ter-
ritory but prohibited in the northern portion. The Wick amendment to
Wilmots amendment fails, with 54 yeas and 89 nays. Then we get to a
point of the main action, a vote on Wilmots amendment (WP > OA)
which passes by teller vote, 83 yeas and 64 nays, so the amendment was
adopted. Perry offers an amendment to the bill that is rejected. Con-
stable offers an amendment that is rejected. McHenry adds a technical
amendment that was adopted with 36 yeas and 0 nays. Hunt, Bell, and
Ashmun offer amendments that are rejected. The Ingersoll substitute for
OA is in order. Wilmot moves to amend the Ingersoll substitute with his
Proviso, and Wilmots amendment to Ingersoll passes by teller vote of
77 yeas and 58 nays. The Ingersoll substitute is defeated by voice vote.
The committee, on motion, rose and reported the message, together
with the bill to the House. The bill was read a rst and second time by its
title (emphasis added).
The US House of Representatives operates by more time-saving rules
when it meets as the Committee of the Whole. The Committee of the
Whole works out the wrinkles on an accelerated basis, and then reports its
result to the members, meeting as the House of Representatives. The of-
cial House functions by more cumbersome and time-consuming rules,
and here major controversies may be replayed. The rules of the House
also require that a bill be read three times before a vote. This bill was most
unusually introduced on behalf of President Polk earlier that day, on the
next to the last day of the Congressional session. The next order of busi-
ness after the amendments are completed, then, is a motion to engross
(record) the bill. Tibbatts moves to lay the whole subject on the table (em-
phasis added), and if this motion succeeds the bill is killed (effectively WP
against SQ). There was a roll-call vote with 78 yeas and 94 nays. This
is the vote that Riker misinterprets as a vote to amend the McKay bill
(WP against OA). In fact, it is a vote about the Wilmot-amended McKay
bill (WP against SQ), and the House refused to lay the bill on the table
(emphasis added; WP >SQ). Nowthat the motion to table has failed, the
motion to engross is in order, and a roll-call vote is taken on the motion
to engross and it passes 85 yeas and 79 nays (CR page 1219, indicating
that WP > SQ). Thus, the motion to engross is adopted and the bill goes
to third reading. Now, the question being on the passage of the bill
(emphasis added), the vote on the bill is by division of the House (names
not recorded) with 87 yeas and 64 nays (WP > SQ). Next, Brinkerhoff
moves to reconsider the vote by which the bill was passed, and this fails on
a recorded vote of 70 yeas and 83 nays (WP > SQ). (An immediate mo-
tion to reconsider is offered on controversial measures because otherwise
the vote remains open to a reconsideration motion and mischief can thus
follow; also, this is probably why, to save time, they took an unrecorded
vote on passage, knowing that a recorded vote on reconsideration would
be coming next). Altogether there were two unrecorded votes showing
WP > OA (strictly speaking, the second of these was WP as an amend-
ment to the Ingersoll substitute which also contained the two million
dollar appropriation). There were one unrecorded and three recorded
votes showing WP > SQ. Riker mistakenly believes that all six of these
votes are about WP > OA, which is nonsensical because the last vote of
any bill passing out of the House on its way to the Senate must be against
the status quo.
Theoretically, any reader should be able to detect the nonsensical er-
ror embodied in Rikers claim that SQ > WP even without going back
to check the references to the records of Congress, yet for almost twenty
years many intelligent people have repeated this story without reporting
the error. I feel that it is my reluctant duty to report a problem with
public-choice style of explanation. This style of explanation is often not
immediately intuitive yet is gilded with an abstract formalism that sug-
gests that something important and believable is being said. I am not the
rst to suggest that there is no necessary relationship between formalism
and profundity, and that it is just as possible that such models obscure as
that they reveal.
Now, return to the rst claim of disequilibrium. The story was told in
a great hurry, but we need to rerun it again in slow motion and keep our
eyes closely on the cards. Rikers account reads as if the House added
the Wilmot Proviso to Polks appropriation proposal (OA) at the very last
minute, as if the House were up to something hasty and duplicitous that
foolishly brought about its bottom-ranked alternative, SQ, no appropri-
ation. The standard history is quite different.
1
The 1844 Democratic Party convention could not obtain the requisite
two-thirds majority for presumptive candidate former President Martin
Van Buren of New York (because he had offended the South by oppos-
ing the annexation of Texas) nor other leading gures, and James Knox
Polk of Tennessee emerged as an unexpected compromise gure, the rst
dark-horse presidential nominee in American history. During his cam-
paign Polk forthrightly promised the annexation of Texas and the max-
imum claim against the British over the Oregon Territory (the famous
slogan Fifty-Four Forty or Fight! originated after the campaign). Polk
was the youngest, and most effective, President to that time in the his-
tory of the republic, but died three months after nishing only a single
four-year term. The annexation of Texas with a contrived boundary claim
that doubled its size sparked war with Mexico, and Polk manufactured
provocations so as to make the war of territorial conquest appear to be
defensive in nature. Land-hungry expansionist Democrats, especially in
the South, supported the war; Whigs opposed it from sincere republican
principle, economy, commercial motives, and partisan advantage; and
northern abolitionists suspected that the land grab was engineered by
the southern slave conspiracy to enlarge southern political and economic
power. Radical northern suspicions were not entirely imaginary, for ex-
ample, the Charleston Courier, in South Carolina, the state that started the
Civil War fourteen years later, declared: Every battle fought in Mexico,
and every dollar spent there, but insures the acquisition of territory which
must widen the eld of southern enterprise and power for the future. And the
nal result will be to adjust the whole balance in the [US], so as to give
us control over the operations of the Government in all time to come
(Jay 1849, 182).
Along his forceful way Polk alienated a number of factions in the north-
ern wing of the Democratic Party who began to accuse the administration
of a southern tilt. Polk promoted the annexation of Texas, precipitating
war with Mexico, but when it came to the Oregon Territory he com-
promised with Great Britain on what is now the present boundary at
49 degrees of latitude, giving up what is nowsouthwestern Canada. Since
the United States couldnt afford two major wars, by committing to war
with Mexico, Polk made it impossible to pressure credibly the British in
the present-day Northwest, which annoyed Democrats in the Midwest
(what is called today the Midwest was in 1846 called the Northwest,
e.g., Northwestern University in Chicago, but to reduce confusion I shall
anachronistically label the area the Midwest) who wanted opportuni-
ties in the northwestern direction. A midwestern Democratic senator
complained:
Texas and Oregon were born in the same instant, nursed and cradled in the same
cradle [the Democratic Party Convention] . . . There was not a moments hesita-
tion, until Texas was admitted; but the moment she was admitted the peculiar
[an allusion to the souths peculiar institution, slavery] friends of Texas turned
and were doing all they could to strangle Oregon. (quoted in Morrison 1967, 12)
When the Oregon treaty with Great Britain came to the Senate for rati-
cation (requiring a two-thirds vote), 13 out of 16 northern Democrats
deserted Polk and voted against it. The treaty was saved by the unanimous
vote of the Whigs in its favor, who supported it because they disfavored
war and expansion, and by 15 out of 16 southern Democrats who wanted
to secure the northern ank so as to attack what was then upper Mexico
to the south (Voteview, 29th Senate, Roll-Call #114, June 18, 1846).
Since the southern slave interests also had annexationist designs on the
remainder of Mexico, Central America, and the Caribbean, in contem-
poraries eyes the Mexican War had unknown implications for the future
character of the republic.
To further sweeten the British, Polk fought for and won lower tariffs
on imported goods, a traditional Democratic goal, but one which en-
dangered some northern Democrats from industrial areas, particularly
Pennsylvania; and some northerners believed that Polk had obtained the
acquiescence of the antislavery British to the annexationof Texas as a slave
state by going easy in Oregon negotiations and by promising themto lower
tariffs on British goods. Polk snubbed Van Buren on federal patronage
appointments and also seemed otherwise to support the competing fac-
tion in Van Burens New York Democratic Party. Polk managed to make
the Mexican War an accomplished fact, and New England Democrats
resented having to vote for Polks war appropriations. The Whigs, who
opposed the war on principle, also at times felt compelled to provide votes
from their own ranks to pass war authorizations and appropriations (e.g.,
HR145, to prosecute war against Mexico, passed 174 to 14 in the House,
Voteview, 29th House, Roll-Call # 72, May 12, 1846; and passed 40 to
2 in the Senate, Voteview, 29th Senate, Roll-Call #218, May 13, 1846).
This is not as peculiar as it seems on rst glance. Just as, for example,
with the Vietnam War, antiwar legislators needed to support the valiant
troops in the eld, which included Whig generals and the sons of leading
Whig Congressmen. The Whigs forerunners, the Federalist Party, not
only opposed but actively obstructed Madisons prosecution of the War
of 1812, and the popular contempt that followed was the nal death blow
to the ailing Federalist coalition; a lesson the Whigs kept clearly in mind.
Similarly, the Whigs wanted to avoid blame should the war they opposed
end in defeat, as they initially expected it might. This is where Rikers
inference of SQ > WP went wrong; yes, the Whigs opposed the war, but
no, they would not vote to end the war if their votes were decisive; they
spoke of a right to oppose the President but a duty to support the country.
Further, a bill to favor a popular midwestern cause by lowering the sale
price of public lands was kept alive while Polk still needed the votes of
midwestern Democrats to win his tariff reduction bill, but once Polk won
the tariff reduction, he had the lands bill killed. Finally, the midwestern
Democrats also would be the major beneciaries of the usual rivers and
harbors bill (aka pork) because improvements in public transportation
increased their markets to the east, to the south, and thence overseas, but
Polk successfully vetoed that measure.
Riker accepts Polks own account of matters without comparing that
to the accounts of other participants, which is not the way to do history.
The point is not that Polk was right or wrong, but that Polks was one
viewpoint among many. Yes, Polk thought that the gures on either side
of the slavery agitation were insincere opportunists, but it is evident from
his diary that Polk was richly endowed with that human propensity that
sees ones own actions as motivated by lofty principle and the actions
of ones opponents as motivated by mundane interests. For Polk, to op-
pose his views was to be insincere. Further, Polks biographer (Sellers
1966, 487) comments that for reasons of age and background he was
not remotely equipped to understand the emotions men brought to the
emerging slavery controversy. By the end of the Congressional session
in August, 1846, Polk had pursued traditional party goals so effectively
that those Democratic factions that needed partial or ineffective accom-
plishment of one or another of the partys goals were damaged. A more
incompetent president would have been less resented. Polk was not a lik-
able man either. In the words of the strange yet perfectly accurate song,
James K. Polk, by the rock-music group They Might be Giants, the
Napoleon of the stump was austere, severe, he held few people dear.
Polks biographer describes the sentiment in Congress as a spreading
feeling that the President was congenitally disingenuous and manipula-
tive in his methods (Sellers 1966, 478). Churchill Cambreleng, Polks
former Congressional colleague, complained:
Heavenforgive me for having any handinlaying the foundations of this blundering
administration. Tyler was bad enough, but he had this advantage there was no
mock mystery nor genuine duplicity in his conduct if he betrayed his friends
he was an honest knave, without any hypocritical cant about the sabbath &c &c.
(quoted inSellers 1966, 478)
So it was not that Polk was the one honest man in a sea of manipulators,
as in Rikers account, but rather that all concerned acted from composite
motivations of principle and interest.
The two houses had resolved to adjourn the Congressional session on
Monday, August 10 at noon. Polk had freshly betrayed his friends in the
nal week of the session, and then he sprung a new surprise. One pre-
text for the war against Mexico was that it owed large debts to American
interests that it was unable to repay. The prowar faction thought that
the Mexicans could repay by ceding land. No Mexican government
could cede land without the political support of its army though, and
Polk wanted two million dollars to bribe the unpaid Mexican army as a
down-payment on full settlement of the war on American terms. On
August 4 he secretly asked this of the Senate, responsible for foreign af-
fairs, which gave its approval, but Whigs in the Senate demanded that
he also consult the House, responsible for appropriations. Polk wanted
secrecy both to preserve his advantage with Mexico and to avoid pub-
lic controversy at home since the request would involve the rst formal
admission that territorial conquest rather than defense was the object of
the war (informally that was plain, since Polk had already seized roughly
what is now the southwestern United States Colorado, New Mexico,
Utah, Arizona, Nevada, and California his main objective). On August 6
a bill to admit the Oregon Territory was in the House, and Thompson,
a protariff Democrat from Pennsylvania, moved an amendment to pro-
hibit slavery in the territory, using the identical language that founder
and southern Democrat Thomas Jefferson had written to prohibit slav-
ery in territories in the Ordinance of 1787. Thompsons amendment
passed resoundingly and even received a few southern votes (but was not
acted upon by the Senate in that session). On Saturday, August 8, on
a miserably hot Washington afternoon, a special message arrived in the
House from the President. It was Polks request for two million dollars
for extraordinary expenses relating to the Mexican War, and in his mes-
sage Polk mentioned that the funds would be payment for concessions
on adjustment of a boundary between the two republics. The cat was
out of the bag. It quite naturally occurred to the disgruntled factions of
northern Democrats that they could offer an antislavery amendment on
Polks measure just as had been done with Oregon two days before, and
thereby have their revenge on Polk. Wilmot had been an administration
loyalist, but was sure to suffer at home for being the only Pennsylvanian
to vote for Polks tariff reduction, with nothing in exchange for his loyalty.
When the bill came to the oor that evening, in an atmosphere of heat,
disorder, and drunkenness, the northern Whigs signaled the play to the
northern Democrats. The northern Whigs said they could not support
the appropriation unless the bill were amended to prohibit slavery in the
acquired territory, and invited the Democrats to offer such an amend-
ment. Wilmot offered his amendment, and it was passed by voice votes.
Then the amended appropriation passed the House on several recorded
votes, as we have already reviewed. The House bill was brought to the
oor of the Senate on the morning of Monday, August 10 about 40 min-
utes before scheduled adjournment at noon. An amendment was moved
to strike the Wilmot Proviso, which some say might have succeeded, and
then the original Polk proposal could have been hurried back to the House
for its concurrence. But Senator John Davis, Whig of Massachusetts, and
one of the two senators who had voted against the rst bill to authorize the
Mexican War, began an unhurried discussion of the issues. His plan may
have been to delay the Senate to the last minute so that it would have
no choice but to accept the House bill as is, with the Wilmot Proviso
(several contemporaries believed that the Senate would have passed
WP > SQ anyway), or his plan may have been the defeatist goal SQ
(either way Davis was reviled by all responsible opinion). The clock in
the Senate was eight minutes slower than the clock in the House, how-
ever, so when Davis would have been ready to yield the oor back for a
vote in the Senate, the House had already adjourned, which technically
ended the session for the Senate as well.
The status quo, the least-favored alternative among Polk, the House,
and the Senate, had prevailed. This was not the product of some profound
political disequilibrium, however. It came about as a result of Polks delib-
erately befuddling last-minute tactics, antimajoritarian libuster, and be-
cause of an exhausted blunder arising from unsynchronized timepieces.
Alternative interpretation
Now for an alternative interpretation of the voting on the Wilmot Proviso
and Mexican War appropriations. We have four sets of data.
First, OA > SQ, as shown by four votes at two different times. Votes
to table OA failed on August 8, 1846 (Voteview Roll-Calls, 29th House,
#450, #451, #452) showing that the House favored OA > SQ. Again
on March 3, 1847 a vote was recorded on whether to adopt Polks three
million dollar appropriation without the Wilmot Proviso and the motion
passed (OA
> SQ, Voteview, 29th House, Roll-Call #637).

Second, WP > SQ, as shown by ve votes at two different times. The
vote on August 8, 1846 on whether to adopt the Wilmot-amended war
appropriations passed (WP >SQ, Congressional Globe, 1st Session: 1217
1218; Voteview, 29th House, Roll-Call #456, motion to table defeated
with 79 yeas and 93 nays, also Roll-Calls #457 and #458). In the 2nd
Session of the 29th Congress the whole issue replayed on February 15,
1847 and again Wilmot-amended war appropriations beat the status quo
on two recorded votes (WP > SQ, Congressional Globe 425; Voteview 29th
House, Roll-Calls #581 and #582).
Third, WP > OA, as shown by three votes at two different times. The
Wilmot amendment to Polks war appropriations was adopted by voice
votes on August 8, 1846 (WP > OA, 92 yeas to 37 nays, and 83 yeas
to 64 nays, Congressional Globe, 29th Congress, 1st Session: 1217). That
was in the 1st Session, and in the 2nd Session replay on February 15,
1847 a roll-call vote on a Wilmot Proviso to Polks new request for war
appropriations was recorded (WP >OA
, Voteview, 29th House, Roll-Call

#580, 115 yeas to 106 nays).
Fourth, OA > WP, as shown only on the last vote. On March 3, 1847
a motion to adopt the Wilmot amendment to Polks war appropriations
failed (OA
> WP, Voteview, 29th House, Roll-Call #635).

From the rst, second, and third sets of data it is apparent that the
House of Representatives ranked the three alternatives WP > OA > SQ,
and thus that there was no cycle. There is an explanation for the reversal
in the fourth set of data, that we will get to shortly. Riker was able to obtain
his cycle on the weakly warranted inference that SQ>WP. His argument
was that because the northern Whigs opposed the war they would never
support a war appropriation even with an antislavery amendment, but in
fact they voted this way three times, once in August 1846 and twice in
February 1847.
There are sufcient data to recover most legislators preference rank-
ings. That would be like killing a mosquito with a ame-thrower, however,
and it is enough to look at the behavior of natural blocs of voters. There
were more northerners than southerners in the House of Representa-
tives, and more Democrats than Whigs. In the 29th Congress there were
87 northern Democrats, 63 southern Democrats, 59 northern Whigs, and
25 southern Whigs, according to Voteview. It is clear both fromvoting pat-
terns and ideologies that the southern Democrats ranked OA>SQ>WP,
and that the southern Whigs ranked SQ > OA > WP. We know from the
rst and third collections of votes and from ideology that the northern
Whigs ranked SQ>OA and WP >OA in both 1846 and 1847; and from
the second collection of votes that they ranked WP > SQ in both 1846
and 1847; thus, they were WP > SQ > OA. The northern Democrats
were split into three camps, presented in increasing order of their oppo-
sition to expansion of slavery into the new territories. The rst camp,
about a dozen in number, were geographically proximate to the south-
ern Democrats and had the same ranking as they, OA > SQ > WP, and
hereinafter I shall exclude this camp from the northern Democrats and
include them in the southern Democrats. The second, loyalist, camp
were the prowar yet antislavery northern supporters of Polk, who ranked
WP > OA > SQ, and made up the bulk of the northern Democrats. The
third, antiwar and antislavery, camp, about a dozen in number, had the
same ranking as the northern Whigs, WP > SQ > OA.
The Whigs opposed the war, and voted against it whenever possible,
but they were always careful not to obstruct the war, for the reasons men-
tioned above. Opposition to the war was a party position, and different
factions had different reasons to oppose the war; the northern Whigs op-
posed it in large part because they opposed the expansion of slavery into
new territories. Therefore, if the northern Whigs were to contribute to
the decisive outcome, they would opt for WP > SQ because WP would
thwart both the Democratic President and the southern slavery interests,
but SQ would undesirably position them as defeatists. Their antislavery
constituents understood their vote WP > OA, but demanded to know
why they voted for the Wilmot-amended war appropriations, WP > SQ.
A contemporary explained:
It may however, be asked with what propriety they could vote for an appropriation
even with the proviso, which they themselves contended was to be used for the
purposes of bribery and corruption. To this question they gave a far more satisfac-
tory answer, than they ever returned to the question why they [had earlier] voted
for a war which they denounced as iniquitous. Mr. Stewart of Pennsylvania, thus
ably vindicated the policy and duty of voting for the appropriation with the proviso:
As a friend of peace, present and prospective, I am in favor of this proviso. The
object of this war being the acquisition of southern territory, as long as there is
a hope of accomplishing this object, there will be no peace. Put an end to this
hope; and you at once put an end to the war, by defeating its object. The moment
the President nds this proviso accompanying this grant of money, he will be for
making peace, and so will all the South. They want no restricted territory. If the
restriction is imposed, and the territory acquired is to be free, from that moment
the President would pay Mexico to keep her territory, rather than bring it in on
such conditions . . . impose this restriction and the war will be promptly ended.
Jay (1849, 186)
Stewart was ungenerous about Polks motives, which were genuinely na-
tionalistic rather than sectional, but Stewart reected the beliefs at the
time of intensely antislavery northerners about the purposes of the war.
I hypothesize, therefore, that in August, 1846, the northern Democrats
expected that the northern Whigs would support WP > SQ and also, of
course, that the northern Whigs would vote WP > OA. The northern
Democrats could also be sure that if the vote were between OA and
SQ, then the northern and southern Democrats together could ensure
OA > SQ against the Whigs. That put the northern Democrats in the
pivotal position between WP and OA so that by voting their sincere po-
sition WP > OA, although they would repel southern Democrats, they
would attract enough northern Whigs to carry the war appropriation they
needed with the bonus of an antislavery amendment; they could win their
rst-ranked alternative. There is no incentive to vote strategically in the
last stage of an agenda sequence when an alternative faces the status
quo (WP against Q, or OA against Q); therefore, all votes were sincere
in August 1846 (and in the vote with identical issues and outcomes in
February 1847). Davis had libustered in the Senate reportedly on the
belief that WP would not have passed there, but no less an authority than
Lewis Cass (D-MI), chair of the Military Affairs Committee and leader
of the prowar forces in the Senate, believed that WP might have passed
if it had come to a vote in the Senate, and at that time said he was dis-
appointed that he was denied the chance to vote in its favor (Morrison,
1967, 28). Wilmot also said that public opinion and informed judgment
considered that WP would have passed in the Senate on August 10, 1846
(Congressional Globe, February 8, 1847: 352). Incidentally, this is contrary
to Rikers unwarranted assertion that the Senate must have top-ranked
OA (in secret session on August 4 they had voted OA > SQ, but WP was
not on the table then). Certainly OA and perhaps WP could have been
enacted by both chambers, but thanks to the blunder of the clocks, things
didnt turn out exactly as planned.
On two separate occasions the House adopted the Wilmot amendment
(WP > OA), once by two voice votes on August 8, 1846 and again by
recorded vote on February 15, 1847, 115 yeas to 106 nays. From the
recorded vote it is apparent that both the antislavery northern Democrats
(WP > OA > SQ) and the antislavery and antiwar northern Democrats
(WP > SQ > OA) voted to adopt the Wilmot amendment (WP > OA
).
But on March 3, 1847, the third occasion, the House defeated the Wilmot
amendment (OA
> WP), 97 yeas and 102 nays. This would only be

possible due to strategic abstentions or to vote changes between February
and March. We have a very clean comparison between the WP against OA
votes on February 15, 1847 and on March 3, 1847. Contemporaries al-
leged strategic abstention in March, but analysis of the two votes shows
nothing unusual about the patterns of those not voting: about ten voters
who voted yes in February did not vote in March and about ten voters who
voted no in February did not vote in March. We can identify exactly the
seven legislators who changed their votes from yea in February to nay
in March and who thereby changed the outcome (the March outcome
was 97 < 102; if the seven had not changed from February it would have
been 97 +7 =104 >95 =102 7). They were Edsall (NJ), Foster (PA),
Garvin (PA), Henley (IN), Russell (NY), J. Thompson (PA), Woodworth
(NY); all were northern Democrats of the type WP > OA > SQ.
To explain the reversal we have to backtrack to the commencement of
the 2nd Session of the 29th Congress in December, 1846.
2
The more
powerful incentives of intraparty unity suppressed the divisive sectional
issue. Wilmots Proviso attracted almost no public attention upon its
introduction and was ignored in the 1846 elections. Polk, a slave-owner
but no slavery ideologue, thought that the Wilmot Proviso was unfair
to the South and more practically that it would depress needed south-
ern political support for annexations of the territory about to be gained.
Although the administrations newspaper seemed to acquiesce to the
Wilmot Proviso at the end of the 1st Session, by the beginning of the
2nd Session Polk had decided not to seek the bribery appropriation un-
less he could attain it without the antislavery amendment. Polk called in
Wilmot, persuaded Wilmot that Polk was sincere about not extending
slavery, beseeched Wilmot not to offer his Proviso, and Wilmot agreed to
introduce his measure as a joint resolution apart from the appropriations
question. Wilmot later said that Polk had offered him an ambassadorship
in exchange for his cooperation (Morrison 1967, 188). The President also
called in Cass (the prowar leader who had indicated support and success
in the Senate for the Proviso) and two other leading senators to seek their
cooperation, but they were not encouraging, or perhaps Cass was play-
ing hard to get. The more intensely antislavery northern Democrats, led
by Preston King (D-NY), quietly sought to revive the issue in the new
session, but their brethren, like their President, Polk, wanted to avoid
exploitation by the Whigs, and preferred to accomplish annexation rst
and antislavery second as separate issues.
Root, a Whig fromOhio, asked whether the northern Democrats could
trust Polk. They had trusted him on Oregon, but he betrayed them by
making a deal with the British that limited northern expansion in order
to enable southern expansion. Would they now trust him to forge a treaty
with Mexico? Polk could write a treaty with the Mexicans that permitted
slavery in the conquered territories, present it as an accomplished fact to
the more compliant Senate (with sole power to ratify treaties), and the
House would have nothing to say about the matter (Congressional Globe,
December 26, 1846: 88). The Whig taunts lashed raw nerves among the
northern Democrats. On January 4, 1847, Democrat King deed the ad-
ministration and sought to reintroduce the Wilmot Proviso (WP > SQ)
from the oor of the House, but failed with 88 northern yeas and
89 southern nays (Congressional Globe, 105; Voteview, 29th House, #105,
mislabeled in Voteview as a vote on Expenses of China Relations). King
gave a speech on the oor the next day, which triggered an avalanche
of controversy. Then the Oregon bill with its antislavery proviso came
up again in the House, and intensely proslavery southerners sought to
amend it to the effect of extending the line of the Missouri Compromise
(36 degrees, 30 minutes) to the Pacic. The southern maneuver was fresh
evidence of southern intentions for the new territory and breached the
understanding among Democrats that annexation would be undertaken
prior to and separate from the slavery issue. Northerners in the House
united to defeat the southern maneuver on Oregon. As the session pro-
ceeded Wilmot came to the conclusion that the Polk forces would prevent
consideration of the promised joint resolution on slavery in the territories,
so that he had no alternative but to propose a proviso again when Polks
request for a bribery appropriation came to the oor. After a week of pas-
sionate and excited debate, on February 15, 1847 a Wilmot-amended
appropriation passed the House despite all of Polks carrots and sticks.
But Polk had more success in the Senate; the authoritative Cass converted
to Polks position. Cass considered the expansion issue to be more im-
portant than the slavery issue, wanted to avoid division in the Democratic
Party because of his presidential ambitions (he would be the Democratic
nominee in 1848), and if Polk had offered neophyte Wilmot an ambas-
sadorship who knows what he offered to the more prominent Cass. Dur-
ing Senate deliberations of Polks request on March 1, 1847, a northern
Whig proposed the Wilmot Proviso (WP > OA; Voteview, 29th Senate,
Roll-Call #356). All southern Whigs and all southern Democrats voted
against the antislavery amendment; all but two northern Whigs voted for
the antislavery amendment, but the northern Democrats divided seven
to seven, and we can speculate that Polk must have provided powerful
private inducements as well. Missouri is sometimes confusing as it was a
minor slave state but is usually also counted as a northern state; its two
senators voted against the Wilmot Proviso. The remaining ve votes were
from Cass of Michigan, one senator from Illinois, two from Indiana, and
one from New York.
Now the Senate measure immediately came to the House on March 3,
the last day of the 29th Congress. It would be nine long months before
the 30th Congress would meet and apparently Polk had provided secret
information that the bribery appropriation was now urgently needed to
attain peace, a goal no one could oppose. At this point the Houses rank-
ing was WP > OA > SQ and the Senates expressed ranking was OA >
WP > SQ. The House could either concur with the Senates OA and at-
tain its second-ranked alternative or by refusing to concur it could choose
its least desired alternative, SQ, by inaction. Polk was busy with persua-
sion in the House as well; an antislavery contemporary says that the
whole inuence of the Government, and all the appliances of party disci-
pline, were now put in place to induce the House to concur with the Sen-
ate (Jay, 1849, 191). So on March 3, 1847, seven northern Democrats
changed from yea to nay on the Wilmot Proviso and it was killed, and
then the House concurred with the Senate and Polks appropriation was
attained. Antislavery forces charged that the seven were compensated,
and there is surely some truth to this, but we must also notice that, given
the urgency of decision and the Senates insistence on OA, it would be
rational for sufcient Northern Democrats of the type WP >OA>SQto
vote strategically against WP in order that their second-ranked outcome,
OA, should prevail rather than their last-ranked outcome, SQ. One of
the seven, Wentworth of New York, printed an explanation in the record
(Congressional Globe, 29th Congress, 2nd Session: Appendix, 438439).
He is explicit that his vote is strategic: I have changed no principle but
have embraced expediency. He fears that:
My motives in so doing will be misrepresented and my conduct denounced. My
vote will be ascribed to corrupt considerations. Those who, in my place, would
not have hesitated to ask payment in advance for their vote, will be the rst,
and most eager to calumniate me . . . I neither seek, nor want, ofce. I am no
supplicant for Executive favors.
Wentworths explanation is eloquent and apparently sincere. He says he is
motivated by a desire for a swift and successful end to the war, which has
claimed several thousand American lives. He says that when he voted for
the Wilmot Proviso two weeks previously he gave notice that he would
support whatever measure came from the Senate. He says that in the
future he will vote for proposals to prohibit slavery in the territories, but
that in this instance he must vote strategically, and that he will succeed
in defending his vote to his antislavery constituents.
Rikers rst story about arbitrary manipulation of a cycle initiating the
events that led to the Civil War fails. His second story, about the election
of Lincoln in 1860 is the subject of the next chapter.
the election of Lincoln
1. Introduction
By way of background, Rikers overarching hypothesis is that the slavery
dimension of concern was suppressed by the Democratic Party manipu-
lative elite with the Missouri Compromise of 1820. The main dimension
of contention between the Democrats and the Whigs, both bisectional
parties, was economic, broadly speaking the Democratic coalition was
agrarian, and the Whig coalition was commercial in orientation. Another
manipulative elite, the northern wing of the Whigs, the weaker party in
this period, sought to nd an issue that would split the Democrats and
thereby allow the northern Whigs to organize a newly dominant coali-
tion. The Wilmot Proviso in 1846 was their rst effort to contrive a cycle,
and the election of Lincoln in 1860 was their last and most successful ef-
fort at contriving a cycle. There is much that is wrong about this story,
but that is for the next chapter.
In this chapter we rst examine Rikers analysis of the 1860 presiden-
tial election. Riker estimates the preferences among the population over
the four candidates. These estimates show both that there was a cycle
among the top three candidates and that different hypothetical voting
rules would yield different outcomes. This is the perfect illustration of
Rikers contentions that democracy is meaningless and arbitrary and that
manipulation is probable on grand issues. I relate the histories and the
ideologies of the four parties in the runup to the 1860 election. The
ideologies of the four parties suggest a unidimensional and noncyclical
distribution of preferences, with the question of slavery in the territories
being the primary dimension of dispute in the election. County-level,
state-level, and region-level aggregates further support the hypothesis
of noncyclicity. Rikers demonstration of meaningless and arbitrary out-
comes depends on an unwarranted estimate that Bell, the candidate of
the Upper South who received 2 percent of the vote in the North, was
ranked second by more than 63 percent of Lincoln voters. If, as was
far more likely, Douglas, the candidate of the Lower North, was ranked
258
The election of Lincoln 259
second by more than 37 percent of Lincoln voters, then most voting meth-
ods considered yield the same ranking, and all yield the same winner:
Douglas.
2. Rikers analysis of the 1860 election
There were four major candidates in the American presidential election
of 1860. They were Abraham Lincoln, the nominee of the six-year-old
northern political alliance known as the Republican Party; Stephen A.
Douglas, the candidate of the northern wing of the Democratic Party;
John Bell, from the recently formed Constitutional Unionist Party, an
attempt at a conservative centrist coalition with its base in the up-
per south; and John C. Breckinridge, Vice President in the outgoing
Buchanan administration and candidate of the renegade southern wing of
the Democratic Party. The election of Lincoln as President in 1860 was
a replay of the disequilibrium of the Wilmot Proviso, says Riker (1982,
228). Lincoln won by a plurality of about 40 percent, he continues, and
so one must suspect a cycle. Riker displays 15 of the possible 24 rankings
possible for strong preferences over four candidates and then estimates
the total number of voters across the nation for each of the 15 likely rank-
ings (each ranking total is the sum of regional subtotals that he does not
display). From the voting data we know only the rst-ranked choices of
the voters, so, as Riker explains, his estimation of the rankings over the
remaining three choices in each of the 15 cases is intended as an informed
historical judgment (he calls themguesses, but presumably they are not
arbitrary or he would not have bothered to present them). Fair enough.
The estimation of full rankings then permits the calculation of hypo-
thetical outcomes by alternative voting rules. The results seem to con-
rmRikers thesis that democracy is arbitrary and meaningless. Different
voting rules lead to different outcomes, and pairwise comparison (the
Condorcet criterion) discloses the presence of a cycle. Here are Rikers
results:
r
Plurality: Lincoln > Douglas > Breckinridge > Bell
r
Pairwise Comparison: (Douglas > Lincoln > Bell > Douglas) >
Breckinridge
r
Borda Count: Douglas > Bell > Lincoln > Breckinridge
r
Approval Voting (two votes): Bell >Lincoln >Douglas >Breckinridge
r
Approval Voting (three votes): Douglas > Bell > Lincoln >
Breckinridge.
Riker concludes that with ve methods of voting Douglas wins twice, Bell
once, Lincoln once, and they are in a cycle and hence tie once. Clearly,
if my guesses are even roughly right, there was complete disequilibrium
in 1860 (1982, 229).
Rikers demonstrations of disequilibria in the Wilmot Proviso and in
the election of 1860 are the primary evidence for his inuential theory
of political disequilibrium and are widely accepted and repeated in the
political science discipline today. We have seen that he was in error about
the Wilmot Proviso and now we shall see that he was in error about the
election of 1860 as well.
His Display 92, Possible Preference Orders in 1860, by Candidate
of First Choice and by Region, which lists his estimates of full pref-
erence rankings for the entire electorate, is daunting to the reader. Ex-
cept for a few words about the regional breakdown of each ranking in
the display itself, there is no textual justication for the ranking esti-
mates. They are presented as authoritative, and one could only check
them by deep immersion in the history of the period. Fortunately, we
need not sweat through all 15 rankings to detect the error, because it
occurs in the rst 2 rankings he lists. These two categories are the es-
timated full rankings of all those whom we know ranked Lincoln rst.
Moreover, the two rankings contain 40 percent of the voting popula-
tion of the entire country, so if there is a major error among them then
we need go no further. Rikers rst group is 450,000 voters who prefer
Lincoln > Douglas > Bell > Breckinridge and they are made up of one-
fourth of New England, mid-Atlantic, and midwest Lincoln voters; and
all southern Lincoln voters. The second group is 1,414,000 voters who
prefer Lincoln > Bell > Douglas > Breckinridge, and they are made
up of three-fourths of New England, mid-Atlantic, and midwest Lincoln
voters; and all border and western Lincoln voters. The problem is with
these two estimates. The large majority of Lincoln voters did not rank
Bell second, as Riker claims; they ranked Douglas second, and once we
understand this error Rikers demonstration collapses.
For ease of exposition, henceforth ignore Rikers inclusion of southern,
border, and western Lincoln voters in the two rankings under consider-
ation, as they represent about 5 percent of his total in the two rankings
and dont change anything in either his analysis or mine. Reconstruct-
ing Rikers data, he has it that about 1,794,000 voters in the free north
(excluding the northern slave state Missouri, border states, and Oregon
and California in the west) voted for Lincoln, and he is correct enough
on that. His further claim is that of those, one-fourth or about 450,000
ranked Douglas second, and three-fourths or about 1,346,000 ranked
Bell second. I claim that it is just the reverse, that most, lets say three-
fourths or more, of Lincoln voters ranked Douglas second, and few, say
one-fourth or less, rank Bell second.
3. The four parties
Who were the four parties and what did they stand for? To understand,
we need to go back to the Missouri Compromise of 1820 which admit-
ted Missouri as a slave state but divided the remainder of the immense
Louisiana Territory at parallel 36 degrees and 30 minutes, with slavery
prohibited north of that boundary.
1
The gaining of the huge Oregon
and Mexican territories in the late 1840s reawakened dormant sectional
tensions over the question of slavery in the new possessions. Controversy
intensied over four years fromthe introduction of the Wilmot Proviso in
1846, which proposed to prohibit slavery in the new territories acquired
fromMexico, to the adoption of the Compromise of 1850, which one his-
torian (Potter 1976, 90) quipped would better be named the Armistice
of 1850. Among other elements of the Compromise of 1850, a law was
enacted to force northern governments to return runaway slaves to their
owners in the south, California was admitted as a free state, the Wilmot
Proviso was squelched, and in its place for the territories that are roughly
now Nevada, Utah, Arizona, and New Mexico the so-called doctrine of
popular sovereignty was enacted: states organized from those territories
would be free or slave as their citizens might decide. The Missouri Com-
promise remained in force for the Louisiana Purchase. Then in 1854,
for interesting reasons that cant detain us here, leading northern Demo-
cratic Senator Douglas won from Congress the KansasNebraska Act,
which organized the Kansas and Nebraska territories on the principle of
popular sovereignty thereby abrogating the Missouri Compromise. This
ignited anger in the North and split the northern Democrats.
Meanwhile, in 1848, the Whigs, who had been weakened by their op-
position to the hugely successful Mexican War, selected war hero and
slave-owner Zachary Taylor (who did not know he was a Whig until
tapped) as their presidential candidate. At the same time, former Pres-
ident Van Buren left the Democratic Party to become the candidate of
the new Free Soil Party, which opposed slavery in the territories, but did
not support abolition of slavery where it was found in the slave states
nor necessarily equal treatment for AfricanAmericans. Taylor won the
election, and the few Free Soil votes did not affect the outcome. Taylor
was a nationalist, not a sectionalist, and he took New Yorker William H.
Seward as a principal adviser, a powerful and talented man but a leading
opponent of slavery. Taylor, who died in midterm(Vice President Millard
Fillmore took his place), weakened the Whig Party in the South. South-
ern Whigs felt that he betrayed them on regional issues, for example,
by admitting California as a free state; the southern Democrats proved
more zealous than the southern Whigs in pursuit of sectional interests;
and popular northern resistance to the new Fugitive Slave Law was an
incentive for northern Whigs to become more vociferously antislavery in
their outlook, offending their southern colleagues. The Whig convention
of 1852 deadlocked along sectional lines, and nally selected the north-
erners choice, Wineld Scott, like Taylor a southerner and a war hero.
The southern Whigs suspected that, like Taylor, Scott would betray them;
Scott lost in the south, some southern Whigs migrated to the Democrats,
and the Whig Party became all but dead in the South. The Democrats
ran Franklin Pierce, a northerner who endorsed the Compromise of 1850
and denounced the abolitionists who opposed that compromise. As a re-
sult of Whig weakness in the south, Democrats were victorious in the
Presidential and Congressional elections of 1852.
The northern population grew rapidly between 1820 and 1860 due
to high natural increase and to high immigration.
2
New immigrants and
their immediate descendants were 19 percent of the northern white male
labor force in 1820 and 46 percent in 1860. The high rates of population
growth meant that labor markets were frequently glutted. Immigration
to the North exploded further after the Irish famine of 18451847 and
the crushed revolutions of 1848 in Europe, and climaxed in 1854. Much
of society beneted from rapid economic growth, but native-born white
artisans and their shopkeepers, about 25 percent of the northern popu-
lation, experienced a hidden depression; from 1848 to 1855 they suf-
fered a 25 percent to 50 percent decline in real wages. There was ination
in the same period, and a long northern recession from 1853 to 1855 as
well. Due to urbanization and consequent epidemic disease, northern life
expectancies from 1790 to 1850 declined by 50 percent. The new immi-
grants tended to be Catholic and vote Democratic against the Whig estab-
lishment. Natives of native-born parents were 62 percent of the northern
Democratic vote in 1852, but 39 percent in 1860; naturalized males made
up 10 percent of the northern presidential vote in 1840, 14 percent in
1852, and 25 percent in 1860. Native northern workers responded polit-
ically, but could nd no succor with the Northern Democrats, always the
party of immigrants, nor with the Northern Whigs, the party of the bosses
who got rich off immigrants. The nativists were without a political home.
The 32nd Congress passed the KansasNebraska Act abrogating the
Missouri Compromise. In the same Congress, the Democrats, who had
always favoreda cheap landpolicy, abruptly turnedagainst it, quite simply
because southern Democrats realized that cheap land was increasing the
population and prosperity of the north and west and thereby threatening
traditional southern control over the federal government. The abolition-
ists had unsuccessfully alleged a Great Slave Conspiracy from around
1845, but the Democrats betrayal on KansasNebraska and cheap land
in 1854 suddenly made the charge plausible among northern workers,
who now feared an invasion of the already distressed northern labor mar-
kets by slave labor. This set the stage for a nativist political upsurge, and
the merger of nativist and antislavery concerns.
In 1852 the southern wing of the Whig Party was decimated over the
national partys apparent betrayal of its interests; andin1854 the northern
wing of the Democratic Party began to bleed away. Northern Democrats
suffered in the elections of 1854 after the treacheries of KansasNebraska
and defeated cheap land. Of the 91 incumbent free-state Democrats in
the House of Representatives, 66 went down to defeat; of the 44 who
had voted for KansasNebraska only seven survived the election. The
Democratic Party, already a bit tilted to the South, became even more
southern as unhappy northerners exited. In the next Congress, south-
ern Democrats outnumbered northern Democrats by two to one. The
KansasNebraska Act added bolting antislavery northern Democrats to
the bubbling soup of factions adrift abolitionists, Free Soilers, north-
ern Whigs, nativists. In 1854 and 1855 the nativist American Party (aka
the Know-Nothings) seemed to be the force that would replace the
Whigs as the party opposing the Democrats. They swept local and state
ofces in Massachusetts in 1854. They were the dominant opposition
in the key northern states; the Free Soilers were the lead opposition in
only a few. All that changed in 1856. The Free Soilers succeeded in split-
ting the Americans into proslavery and antislavery wings, and harvested
from the antislavery faction. At the same time immigration dropped by
half from 1854 to 1856, ination ended, and real wages increased. The
various anti-Nebraska groupings slowly coalesced locally and nation-
ally into the new Republican Party. After a protracted stalemate in 1856,
the nascent Republican coalition elected a nativist the Speaker of the
House on a plurality vote, and by 1860 the northern Americans were
fully absorbed into the Republicans, who became the victorious opposi-
tion. Republican policy and practice made subterranean appeals against
the twin despotisms of slavery and papery, but repudiated explicit nativist
proposals (a party that proposes that a signicant portion of those who
have the vote shall have it taken away cannot long be successful). Fogel
(1992, 229230) emphasizes that the salience of slavery over nativism in
1860 was quite contingent. If the Panic of 1857 had been extended, if
immigration rates had returned to 1854 levels, the Republicans would
have been unable to keep the lid on the two-fths of its membership who
were former Americans, and by alienating the immigrant vote may not
have been able to attain a majority in the electoral college in 1860.
In 1856, the Republicans ran their rst presidential candidate, John
C. Fremont, another politically inexperienced and rather frivolous war
hero. He had a Catholic father and so was not tainted by vote-losing na-
tivism and had no record in public life and so was not tainted by zealous
denunciations of slavery as were the powers behind the scenes in the Re-
publican Party. The Democratic convention, which required a two-thirds
majority, was deadlocked until Douglas, still hot from KansasNebraska,
withdrewin favor of Buchanan, in expectation of consideration at the next
opening for the presidency. Northern Whigs drifted to the Americans or
the Republicans. The leftover southern Whigs rallied to the American
Party in the south; the Southern Americans were proslavery and also ap-
pealed to southern interests by opposing the foreigners and cheap land
that expanded northern population and power. They nominated Millard
Fillmore, former Whig President and supporter of the Compromise of
1850. Fillmores practical hope would be to deny an electoral college
majority to Buchanan, which would throw the decision to the House
of Representatives where his centrist candidacy might naturally prevail.
The race was between Buchanan and Fremont in the north, and between
Buchanan and Fillmore in the south, and of course Buchanan beneted
froma split opposition. Also, those centrist voters in the north who feared
disunion should Republican Fremont win voted for Buchanan from the
only bisectional party. Buchanan won. Fremont, however, swept the up-
per north and did well although losing in the lower northern states of
Illinois, Indiana, Pennsylvania, New Jersey, and also California. Fremont
was a weak candidate, and Republicans were encouraged since all they
would have to do is add Pennsylvania and either Illinois or Indiana with
a stronger candidate and they might win in 1860. If a united Democratic
Party carried the South, however, and Oregon and California, all it would
need for victory is New York alone or Pennsylvania alone, or Indiana and
Illinois together with one more state. The 1860 election would be won
by whoever did better in the lower north.
Two days after Buchanan was inaugurated in 1857 the Supreme Court
issued its decision in the Dred Scott case, arguably the point at which the
Civil War became inevitable. Dred Scott was a Missouri slave who sued
for freedom on the grounds that his owners had for a time kept him in
territories north of the boundary of the Missouri Compromise. Five of
the judges were southern Democrats; two of the four northerners on the
court were Democrats and one of those, Grier, was a champion of slavery
who had southern relatives. Buchanan had secretly intervened with Grier
to urge him to vote with the southern justices, and Buchanan blurted
out a hint of his foreknowledge of the decision in his inaugural address.
Chief Justice Taney for the majority denied citizenship to blacks including
Dred Scott, found the Missouri Compromise of 1820 unconstitutional,
and ruled that Congress lacked the authority to exclude slavery from the
territories. In other words, the referee threw the game for the South. Re-
publicans were apoplectic. Its nakedly partisan nature and the multiple
and divided justices opinions accompanying the decision weakened its
authority, however, and boosted the following of the Republican Party.
Northerners implausibly but forcefully construed the portion of the deci-
sion relating to slavery in the territories as mere obiter dictum, a comment
in passing without force, on the argument that since the court ruled that
Scott was not a citizen the decision stopped at that point. The South,
however, was now able to festoon its positions on slavery with all the or-
naments of constitutionalism and legality; the Dred Scott case increased
its sense of legitimacy and of grievance, and augmented its rhetorical bar-
gaining power. The Supreme Court decision did not execute itself, and
southerners began to agitate for a federal slave code that would enact the
doctrines of the Dred Scott decision.
The decision was intended to deny the newborn Republican Party its
reason for being, but it also frayed the doctrine of popular sovereignty
which was all that held the Democratic Party together. Taneys opinion
found that the territorial legislatures lacked the authority to prohibit slav-
ery, and if effective would void popular sovereignty, but the opinions of
other justices were silent on that question and thus perhaps it was mere
dictum. If the issue arose directly to the Supreme Court, however, would
the same justices who denied Congress the authority to regulate slavery in
the territories then nd that the territorial legislatures had the authority to
prohibit slavery? Douglas attempted to rescue popular sovereignty from
Dred Scott with the argument that although slave-owners had a Consti-
tutional right to their property in a territory, the territory need not pro-
tect that right with enforcement or legislation. Douglass construal riled
southern Democrats and gave them more reason to demand a federal
slave code to prevent such dodges. The doctrine of popular sovereignty
also suffered in the controversies over the adoption of a state constitution
in Kansas and its admission as a state to the union. Northern and south-
ern immigrants to Kansas agitated at length and to the verge of civil war
(200 died in guerilla skirmishes) on the slavery question resulting in two
different conventions and constitutions for the state, one slave and one
free. The free staters outnumbered the slave staters in Kansas by about
two to one, but Buchanan demanded that Democrats in Congress sup-
port the slave Lecompton constitution. Douglas thought that to endorse
the fraudulent slave constitution would damage the doctrine of popular
sovereignty, kill himin his upcoming 1858 reelection bid for the Senate in
Illinois and would destroy any chance of attracting northern votes to the
Democratic Party in the 1860 Presidential election; and so he deed the
administration, but in the end Lecompton was accepted by the Senate.
Even with all-out party pressure, however, Buchanan could not carry
his Democratic majority in the House because of northern defections,
and Kansas was not admitted. In 1858, more northern voters exited the
Democratic Party over the spectacle of bleeding Kansas. The free state
Democrats went from53 souls in the House to 32, and 12 of those 32 sur-
vivors were those who voted against Buchanan on Lecompton. At this
point the House Democrats consisted of 69 southerners, 19 northerners
loyal to the administration, and the 12 northerners who voted against the
south (called the Lecompton Democrats).
The Congressional party was now by a small majority southern. The
rules of the Democratic convention provided for representation fromeach
state according to its electoral-college vote (each state gets to vote its
number of representatives and senators), and since the north was more
populated than the south, the convention party was by a small majority
northern. Throughout his career Douglas, the leading political gure of
the 1850s and a determined compromiser, had positioned himself at the
center of the national party; geographically, the median voter would live
in the Lower North (that is where Douglas did best in the 1860 election).
He had stood aside for Buchanan in 1856, and in 1860 he was the heir
presumptive. The convention met on April 23, 1860, unfortunately in
Charleston, South Carolina (a state that yet today ies the secessionist
ag over its state capitol), in the heart of plantation slavery. Douglas was
the median candidate of the convention party, but also convention rules
required a two-thirds majority vote for nomination. The majority of the
Congressional party were southern, their position was endorsed by the
Supreme Court, and by the Democratic President, and thus they felt en-
titled to having their views adopted. They wanted a federal slave code,
which Douglas would not give them, because that would lose the elec-
tion in the North, and hence in the country. The convention was chaotic
and bitter. Douglas could not win two-thirds for the nomination, but
he did win on the platform by majority vote. Winning on the platform
meant that Douglas kept out the call for a federal slave code, the inclusion
of which would ensure defeat everywhere in the North, but its exclusion
caused Democrats fromthe Lower South to bolt the convention. Douglas
may have welcomed their exit, betting that their temporary absence would
allowhimto attain two-thirds of the remaining delegates, but the conven-
tion chair, a southern sympathizer, unexpectedly ruled that two-thirds of
all delegates, present and absent, would be required (Douglas might not
have had two-thirds of the present delegates either). After ten days and
59 deadlocked ballots on the nomination, the convention adjourned for
six weeks, to resume in Baltimore. The bolters determined to reconvene
in Richmond, ve weeks hence.
The Constitutional Union Party convened on May 9. This group re-
united the remainders of the Whig and American Parties in the Upper
South moderates from Maryland to Missouri and also attracted
some older Whig conservatives from the North uncomfortable with the
Republicans. Its organizers were politicians of eminence and distinc-
tion, respected, intellectual, and moderate in conviction. This was the
good-government party (supercially resembling the Liberal Democrats
in present-day United Kingdom) that would oppose the secessionists
in the South, just as Douglas forces in the Democratic Party opposed
the radical Republicans in the North. The organizers hoped to displace
the Republican Party as the anti-Democratic Party, and originally be-
lieved that if they met rst in 1860 and nominated leading moderates the
Republicans would be forced to ratify their ticket. Their more modest ex-
pectation was that the Republicans would nominate Seward, leaving the
Lower North open to their moderate appeal. The party failed to attract
more than symbolic support in the north, however, and its ambitions
never came to fruit. They nominated John Bell of Tennessee, aged 64,
and Edward Everett of Massachusetts, 67, respectable, but not charis-
matic, candidates. As in 1856, this factions most realistic hope was to
deny any other candidate a majority in the electoral college and thus
emerge as the brokered victors in a centrist compromise in the House of
Representatives.
The Republicans met next in Chicago on May 16. Seward was the dom-
inant candidate going into the convention. Seward was talking moderate
in 1860, but his record of vigorous denunciation of slavery was indelible.
Party leaders feared he could not carry the Lower North, where the elec-
tion of 1856 was lost, and which was now threatened both by Douglas
and by the Constitutional Unionists. Lincoln was a barely noticed can-
didate going into the convention, but his managers cleverly worked to
make him everyones second choice. Lincolns prior political career was
one of well-planned moderation, and he was at the center of opinion in
the party. The choice of Lincoln rather than Seward doomed Bell and
would doom the Democrats unless they reunited.
The Democrats reconvened in Baltimore on June 18. The southern
delegates expected to control the convention and dump Douglas, but
Douglas was the more clever politician. The disappointed southerners
retired to their rump convention in Richmond, called for a federal slave
code, and nominated John C. Breckinridge, the failed Buchanans Vice
President. There seemed to be at least four schools of thought, not nec-
essarily exclusive, among the bolting southerners. One was a myopic sec-
tionalist passion that paid no heed to consequences. The second was to
bully the tough Douglas into concessions, which failed. The third, among
the more moderate bolters, was a rather muddled hope that the multi-
plication of candidates would deny Lincoln a majority in the electoral
college, which would throw the election to the House of Representa-
tives evenly split between parties under the requisite voting rule, and
eventually the stalemate would go the Senate with a strong Democratic
majority where the Republicans would lose. The fourth, among the more
extreme bolters, was the frank desire for secession, which would have been
much harder to justify if the candidate of a united Democratic Party were
elected. Many of the leaders did expect that splitting the Democratic
Party would elect Lincoln and hasten disunion. A contemporary jour-
nalist reported on the Charleston convention that, the seceding States
came to the convention with a deliberate purpose to break up the con-
vention if they failed to get, as they knew they would fail to get, their
extreme ultimatum, and their ultimate design is to break up the Union
by breaking up the Democratic Party (quoted in Nevins 1950, 227).
Douglas was comfortable with popular sovereignty as a compromise
because he expected that slavery would be impractical in the territories.
As a practical politician, he would let nature rather than policy settle the
issue. The southern demand for a federal slave code had little practical
importance, and on the Senate oor Douglas complained:
There is no necessity for legislation; no grievances to be remedied; no evil to be
avoided; no action is necessary; and yet the peace of the country, the integrity of
the Democratic party are to be threatened by abstract resolutions . . . The people
will ask what all this is for; what it means; why is it so important to have a
vote . . . Why? There must be some purpose. (quoted in Nevins 1950, 180181)
After the Baltimore convention, when it appeared that Lincoln might win
against the divided opposition, Jefferson Davis reportedly approached
Douglas with a proposal to organize a fusion campaign against Lincoln,
promising that Bell and Breckinridge would withdraw as candidates if
Douglas would as well. Douglas reportedly replied that, the election
should never go the House, before it shall go into the House, I will throw
it to Lincoln (quoted in Nevins 1950, 285). It seems that Douglas feared
disorder should the election go to the House. He reportedly believed that
the leadership of the southern bolters, counting on either Lincoln to win
or better for the election to go the House, had formulated a plan for
a seizure of power in Washington in the weeks after the election. The
Buchananites already held the federal ofces in the capital. Breckinridge
would carry the far south, and if he could win Virginia and Maryland,
adjacent to the District of Columbia, military forces fromthose two states
could back a Buchananite takeover (Nevins 1950, 295). Thus, Douglas
campaigned hard in Virginia, Maryland, and North Carolina, not for
himself, but for Bell and against disunion, so as to thwart a coup.
Douglas declared forthrightly that the union should be maintained by
military force. A month before the election, when it was plain to Douglas
that Lincoln would win, even though quite ill, he bravely went into the
Lower South and ercely campaigned against disunion; he was in Mobile,
Alabama on election day. Douglas died of illness in 1861 at the age of 48
while visiting his base in Chicago.
The platforms of the four parties manifest a monotonic north to south
gradation of position on the free soil question, the primary dimension of
dispute in the election. Lincoln and the Republicans declared all territo-
ries closed to slavery; Douglas and the northern Democrats held to the
doctrine of popular sovereignty; Bell and the Constitutional Unionists
implicitly considered the territories open to slavery but a matter for polit-
ical negotiation; and Breckinridge and the southern Democrats declared
all territories open to slavery on implicit threat of secession.
3
The Republican platform maintained inviolate the rights of the states,
especially the right of each state to order and control its own domestic
institutions; in other words, it guaranteed slavery in the slave states. The
Republicans rejected the new dogma that the Constitution permits the
extension of slavery into any and all territories as a dangerous political
heresy; in other words, they held that the Dred Scott decision was limited
in effect. They declared that the normal condition of all territory of the
United States is freedom; slavery would be prohibited in all territories of
the United States. They explicitly afrmed that the union of the states
must and shall be preserved, and denounced threats of disunion. They
said that the controversies over the Lecompton constitution in Kansas
revealed the Democratic doctrine of popular sovereignty to be a deception
and a fraud.
The Democratic Party (Douglas) afrmed its Cincinnati platform of
1856, the doctrine of popular sovereignty. Popular sovereignty contained
a useful ambiguity. For northerners it meant that a territorial legisla-
ture could prohibit slavery from the beginning, for the southerners it
meant that the territory was open to slave-owners until it might apply to
Congress for admission as a free or slave state (allowing slavery a toehold,
as it would be expensive for a free state to compensate slave-owners for
loss of property). The Douglas platform directly mentioned this differ-
ence of opinion, and pledged to abide by the decisions of the Supreme
Court on the question. Douglas had earlier suggested that territorial
governments might decline to protect slave property, but the platform
again pledged faithful enforcement of whatever doctrine was decided
by the Supreme Court. They also endorsed execution of the Fugitive
Slave Law.
The Constitutional Unionists said that platforms are misleading and di-
visive, and theirs was brief and basic. Their brief platform is a plati-
tudinous appeal to no political principle other than the Constitution
of the Country, the Union of the States, and the Enforcement of the
Laws (emphasis omitted, Morison 1971, 1127). Because of this, and
their poor showing in the election, many commentators neglect to locate
them geographically and ideologically. Who they were and what they
did not say is more important than the little they did say in the plat-
form document.
4
The convention did not seek to establish uniformity of
opinion on the free soil question, leaving that to every individuals judg-
ment. The party would scrupulously avoid mention of the tiresome topic
of slavery; although its state afliates might take particular stands. Where
Douglas sought an ambiguous formulation on the free soil question,
the Constitutional Unionists sought none. Their platform pledged them,
however, to the just rights of the people and of the States reestablished,
and the Government again placed in that condition of justice, fraternity,
and equality which, under the example of the Constitution of our fathers,
has solemnly bound every citizen of the United States (emphasis added).
Notice that the pledge of continued union is conditioned on the reestab-
lishment of states rights and equality between the sections as found in
the original constitutional compromise. It is the original constitutional
compact that commands their loyalty, not present-day misinterpretations
or usurpations.
Outside the platform, the Constitutional Unionists denounced
Douglass doctrine of popular sovereignty, and believed that Congress
must protect the property rights of slave-owners in the territories. They
also denounced the Breckinridge Democrats for threatening secession;
but did not deny the right to secede should compromise fail. The practi-
cal difference between the two southern factions was that Breckinridges
Democrats considered the election of a Republican president sufcient
to justify secession, while Bells Constitutional Unionists thought that
resistance would only be justied by an overt act against southern inter-
ests by a northern administration. Presumably, Rikers inference that Bell
was the second choice of most Lincoln voters was based on the thought
that the Constitutional Unionists contained many former Whigs, and
that Whiggish Lincoln voters in the North would go for Whiggish Con-
stitutional Unionists rather than for Douglas from their ancient enemy
the Democratic Party. But most northern voters did not care to vote
for the party of the Lower South: the Constitutional Unionists received
2 percent of the vote in the free states, 13 percent of their vote nation-
wide. Rikers position that three-fourths of Lincoln voters in the North
Lincoln Douglas Bell Breck
U. North L. North U. South L. South
Figure 12.1 Single-peakedness, 1860
ranked Bell second is not only undefended, it goes against the clear and
convincing evidence.
The breakaway Democrats (Breckinridge) declared that all territories are
open to slavery, and that neither Congress nor a territorial legislature had
constitutional authority to legislate otherwise. They said that the federal
government had the duty to protect persons and property wherever they
were threatened, that is, that the federal government should coercively
suppress state, local, and territorial governments that legislate against
slavery or fail to enforce a proposed federal slave code not directly men-
tioned in the platform. Territories that form a state constitution and peti-
tion for statehood ought to be admitted whether they propose to prohibit
or permit slavery.
4. Unidimensional preferences in the 1860 election
I maintain that voter preferences in the 1860 election were for the most
part single-peaked and unidimensional, as in Figure 12.1. Riker main-
tains that the preferences of Lincoln voters were not single-peaked be-
cause they ranked Bell second and Douglas third, as in Figure 12.2. Riker
does not have direct data for his assertion that Lincoln voters ranked Bell
second; it is an inference, and one not defended in his text. I have labored
Lincoln Douglas Bell Breck
U. North L. North U. South L. South
Figure 12.2 Rikers cycle, 1860
in the preceding section to showthat it is ideologically and geographically
implausible that the majority of Lincoln voters ranked Bell second. Ag-
gregate data will further support my position. Anyone can accuse an
inference from an aggregate to its individual members of being an in-
stance of the poorly named ecological fallacy. If the aggregate of rst-place
votes shows A rst, B second, and C third, that does not necessarily mean
that individuals who rank A rst rank B second. All A-voters could rank
C second; from the aggregate data alone we are not entitled to a certain
inference that A-voters rank B second. If, however, we have independent
evidence that A-voters rank B second, then aggregate data bolster the
conclusion that A-voters rank B second. The ecological fallacy is not a
fallacy, but rather a logical possibility. It is a possibility and not an in-
evitability. The independent evidence I have just adduced on ideology
supports my proposed inference from aggregate rankings to individual
rankings. Why would a voter in the north choose Lincoln? High among
the reasons must be Lincolns appeal to northern interests including his
position on Free Soil; otherwise one of the other candidates would do. If
such a voter were denied the choice of Lincoln, who would his second
choice be? Why not Douglas, who is the next most adjacent on issues of
northern interest? And if Douglas were not available where next would the
typical northern voter turn? He would turn to Bell, the candidate of the
Upper South, of course, not to Breckinridge the candidate of the Deep
South. A zealous critic could claim that this is not enough; but please
Table 12.1. State-level aggregation of
rst-place winners, Upper North
Upper North L > D > Bl > Br expected
Connecticut L > D > Br > Bl
Illinois L > D > Bl > Br
Indiana L > D > Br > Bl
Iowa L > D > Bl > Br
Maine L > D > Br > Bl
Massachusetts L > Bl > D > Br
Michigan L > D > Br > Bl
Minnesota L > D > Br > Bl
New Hampshire L > D > Br > Bl
New York L > D
Ohio L > D > Bl > Br
Pennsylvania L > Br > D > Bl
Rhode Island L > D
Vermont L > D > Br > Bl
Wisconsin L > D > Br > Bl
notice that my hypothesis that Douglas the candidate of the Lower
North ranked second among Lincoln voters in the North rather than Bell
the candidate of the Upper South is supported by some evidence, while
Rikers opposite hypothesis is supported by none. Riker is attempting to
support his controversial and apparently counter-empirical doctrine of
political disequilibrium; and the burden of proof is on one who would
defend his extraordinary claim.
A look at a map of the county-level winners in the 1860 presidential
elections (McPherson 1993, 128) shows that latitude is attitude. The
Upper North voted for Lincoln (more free soil), the Lower North voted
for Douglas (less free soil), the Upper South voted for Bell (less slave
soil), and the Lower South voted for Breckinridge (more slave soil). Na-
tionally, Douglas and Bell did less well than Lincoln and Breckinridge
and thus the eastwest swathes of Douglas and Bell in the center are
each thin and those of Lincoln and Breckinridge at the extremes are each
thick. The strongly unidimensional (north to south) map of rst-place
county winners supports the hypothesis that preferences were generally
single-peaked.
We can easily check whether aggregates at the state level show Douglas
or Bell second in the North and whether aggregate preferences are gen-
erally single-peaked (see Table 12.1). Italics indicate an outcome con-
trary to the expectation of single-peakedness. Obviously, Douglas nished
second everywhere in the North except for in Massachusetts and
rst-place winners, Middle America
West L > D > Bl > Br expected
California L > D > Br > Bl
Oregon L > Br > D > Bl
Lower North D > L > Bl > Br expected
Missouri D > Bl > Br > L
New Jersey D > L
Upper South Bl > Br > D > L expected
Delaware Br > Bl > L > D
Kentucky Bl > Br > D > L
Maryland Br > Bl > D > L
Tennessee Bl > Br > D
Virginia Bl > Br > D > L
Pennsylvania. Bell nished second in Massachusetts, but that was be-
cause the Vice Presidential nominee on the Constitutional Union ticket
was Edward Everett of Massachusetts who got a favorite-son vote in his
home state. Incidentally, this suggests that to insist on the ecological
fallacy is overly rigorous: in Massachusetts, Bell was second in terms
of rst-place votes, and was probably second-ranked by Lincoln vot-
ers as well. In Pennsylvania, in a last-minute attempt to deny Lincoln
a majority in the electoral college and salvage local campaigns, Douglas
and Breckinridge forces formed an anti-Lincoln fusion candidacy with
Breckinridge as the nominal candidate, and even then stubborn voters
continued to vote for Douglas and for Bell. New York had a fusion of
Douglas and Bell forces with electoral-college votes pledged two-thirds
to Douglas and one-third to Bell. Rhode Island had a fusion ticket un-
der Douglas. My ideological prediction that Bell would nish third is
not supported: he nishes third in three states and fourth in eight states.
This is the candidate that Riker would have us believe is ranked second by
northern Lincoln voters. Breckinridge often did better than Bell because
he enjoyed organizational support fromthe outgoing Buchanan machine,
including many of the northern Democratic legislators and most of the
local patronage appointees. In this region, 55 percent of voters favored
Lincoln, 35 percent Douglas, 8 percent Breckinridge, and 2 percent Bell,
L > D > Br > Bl.
The two free western states should followthe northern pattern. Lincoln
nishes rst in both. In California, Bell nishes fourth. In Oregon
rst-place winners, Lower South
Lower South Br > Bl > D > L expected
Alabama Br > Bl > D
Arkansas Br > Bl > D
Florida Br > Bl > D
Georgia Br > Bl > D
Louisiana Br > Bl > D
Mississippi Br > Bl > D
North Carolina Br > Bl > D
South Carolina Br
Texas Br > Bl
Breckinridge nishes second, but this is because his Vice Presidential
candidate, Lane, was a senator from Oregon. For the Lower North on
ideological grounds I predict that voters wouldrst support Douglas, then
Lincoln the other northern candidate, then Bell from the Upper South,
then Breckinridge from the Lower South. Douglas is rst in Missouri,
but against the schematic expectation Lincoln is last in this northern
slave state. In New Jersey there was an anti-Lincoln fusion candidacy of
Douglas, Bell, and Breckinridge, under the ag of Douglas. Not showing
at the state-level of aggregation, but connecting Missouri to New Jersey
from west to east, are Douglass strengths in southern Illinois, southern
Indiana, southern Ohio, and the fusion candidacy under Breckinridges
name in Pennsylvania. In the Lower North region, limited to two states,
Douglas obtained 42 percent of the vote, Lincoln 27 percent, Bell
21 percent, and Breckinridge 11 percent. For the Upper South the ideo-
logical prediction is that Bell would be rst, followed by the other south-
ern candidate, Breckinridge, then Douglas, then Lincoln. Expectations
for the Upper South are satised in Kentucky, Tennessee, and Virginia,
and are imperfectly satised in the states of Delaware and Maryland.
Overall in the upper south region, Bell received 45 percent of the vote,
Breckinridge 43 percent, Douglas 11 percent, and Lincoln 2 percent.
South Carolina, true to form, did not let its citizens vote for President,
a choice reserved for the political elite. Texas had an anti-Breckinridge
fusion slate under the banner of Bell. The Lower South satises expec-
tations. Overall in the Lower South, Breckinridge got 55 percent of the
vote, Bell 38 percent, Douglas 7 percent, and Lincoln was not on the
ballot. The largest exception among all regions to the ideological single-
peakedness prediction places Bell fourth in the North rather than second
where Riker needs him to be.
Finally, lets look at the section-level aggregate of votes. Riker does not
provide a source for his election data, nor does he explain his calculations,
nor is his arithmetic consistently reliable. Thus, it is not possible to infer
what Rikers source document was, and the numbers I use (derived from
Burnhams regional totals, 1955, 246256) are quite close (1 or 2 percent)
to what can be reconstructed from Riker but are not identical. There
were about 3,450,000 voters in the North (excluding the western states
California and Oregon but including the slave state Missouri). Of these,
53 percent voted for Republican Lincoln, 36 percent voted for northern
Democrat Douglas, 8 percent voted for southern Democrat Breckinridge
and 4 percent voted for Constitutional Unionist Bell: L > D > Br > Bl.
After my rst analysis of the issue, Tabarrok and Spector (1999) ap-
peared, an intriguing, useful, and a more technically sophisticated anal-
ysis of the 1860 election. Tabarrok and Spector had Rikers rankings
over the four candidates. They also queried 100 Civil War historians as
to their views on the rankings, and 15 responded with opinions about
the percentage of voters of various rankings over Lincoln, Douglas, Bell,
and Breckinridge. For voters who ranked Lincoln rst, the historians
median opinion was that 60 percent ranked Douglas second and 40 per-
cent ranked Bell second. Of the 15 historian respondents, 12 had a view
contrary to Rikers; each of those 12 believed that 50 percent or less of
Lincoln voters ranked Bell second. Two of the three respondents who
seem to agree with Rikers estimate (that Bell ranked second among
more than 63 percent of Lincoln voters) are also those who make the
most extreme estimate in either direction (90 percent and above for
Bell). Tabarrok and Spector constructed a median historians prole,
and compared it to Rikers. As I said, I believe that many commenta-
tors, a number of historians included, fail to locate the Constitutional
Unionists ideologically and geographically, and my estimate of their sup-
port among Lincoln voters would not exceed 10 percent; however, noth-
ing in the analysis here depends on my apparently extreme view on the
question. The plane of possible outcomes in Tabarrok and Spectors
positional-vote space, using Rikers prole, is large, and permits a large
variety of rankings by way of varying the weights of a positional voting
rule from plurality to Borda count to antiplurality and everything in
between. The positional vote plane using the median historians prole
is small, and Douglas wins . . . under any positional voting system which
gives signicant weight to second- or second- and third-ranked prefer-
ences (278); Lincoln wins by plurality count, which gives no weight to
any preferences except for the rst-ranked. Condorcet pairwise compar-
ison is cyclic with Rikers prole, but with the median historians prole
Douglas beats Bell by 15.1 percent, Bell beats Lincoln by 0.033 percent
(three-hundredths of 1 percent), and Lincoln beats Breckinridge by
25.4 percent.
The ideological data, aggregate data at the county, state, and sectional
levels, and the opinion survey of antebellum historians, all support the
hypotheses that Douglas and not Bell was second-ranked among Lincoln
voters, and that preferences were for the most part single-peaked across
the country.
What does this revision do to Rikers demonstration that democracy is
arbitrary because different voting methods result in different outcomes
and that democracy is meaningless because cycles will be contrived on
major issues such as the future of slavery? The rst thing we must do is
to correct Rikers original claims. He claims that the Borda count ranks
D > Bl > L > Br. This is not correct, the Borda count from his gures
indicates D > L > Bl > Br. He also states that the actual method
and outcome was the plurality ranking L > D > Br > Bl. That is not
quite correct, as the actual outcome was determined by the electoral
college where the ranking was L (majority) > Br > Bl > D. Customarily,
the electors vote on the instructions of their states and if they deadlock,
then the top three candidates in the electoral college go to the House
of Representatives where each state delegation casts one vote, and if the
House deadlocks then the top two Vice Presidential candidates go to the
Senate where each senator has one vote. It is important to know that this
arrangement, and not national-level plurality rule, was the voting scheme
that motivated the candidates decisions. That gives us a corrected version
of Rikers ndings:
r
Electoral College: Lincoln (majority) >Breckinridge >Bell >Douglas;
r
Plurality: Lincoln > Douglas > Breckinridge > Bell;
r
Pairwise comparison: (Douglas > Lincoln > Bell > Douglas) >
Breckinridge;
r
Borda count: Douglas > Lincoln > Bell > Breckinridge;
r
Approval voting (two votes): Bell >Lincoln >Douglas >Breckinridge;
r
Approval voting (three votes): Douglas > Bell > Lincoln >
Breckinridge.
Riker has one-fourth of Lincoln voters ranking Douglas second and three-
fourths of Lincoln voters ranking Bell second. Bell got 2 percent of the
vote in the states ranking Lincoln rst, but lets be generous and adopt
the median historians view that Bell was the second choice of 40 percent
of Lincoln voters and that Douglas was the second choice of 60 percent of
Lincoln voters. What happens then to the arbitrarily different outcomes
and the meaningless cycle? They disappear! A pairwise-comparison ma-
trix is displayed in Table 12.4. Rikers gures are in roman type. The cells
showing how many preferred Douglas to Bell, and how many preferred
Table 12.4. Pairwise comparison matrix, 1860 election
(thousands)
L D Bl Br (Borda)
Lincoln 2165 2542 2968 (7675)
Douglas 2516 2265 3658 (8439)
718 (8797 )
+1076 =
2623
Bell 2139 2416 3090 (7645)
+718 (7287 )
1076 =
2058
Breck 1713 1023 1591 (4327)
Note: Rikers estimates in Roman type, Mackies revision in italic
type.
Bell to Douglas, are revised to reect my proposed revision of the pro-
portions of Lincoln voters second-ranking Douglas and Bell. My revised
totals are in italic.
The consequences are straightforward. By pairwise comparison, we
have D > L, L > Bl, L > Br, D > Bl, D > Br, Bl > Br; and that reduces
to D > L > Bl > Br, which is not a cycle. The Borda count is still
D > L > Bl > Br. Approval voting with two votes changes from what
it was to D > L > Bl > Br. Approval voting with three votes remains
D > Bl > L > Br. Here is a summary of the nally corrected outcomes:
r
Pairwise comparison: Douglas > Lincoln > Bell > Breckinridge;
r
Borda count: Douglas > Lincoln > Bell > Breckinridge;
r
Approval voting (two votes): Douglas >Lincoln >Bell >Breckinridge;
r
Breckinridge.
With plurality runoff, Lincoln and Douglas would go to the runoff
and Douglas would win (a runoff between rst-round losers Bell
and Breckinridge has Bell as the winner, yielding the overall ranking
D > L > Bl > Bk). All ve of these voting methods select Douglas
as the winner. Except for approval voting with three votes (in this
example equivalent to antiplurality), the rules identify the same ranking
D>L>Bl >Bk. That antiplurality is an exception is not a surprise, since
antiplurality is an even more inaccurate voting rule than is the plurality
rule. Antiplurality means that everyone votes against her least-favored
candidate, and it is inaccurate because all other ranking information is
ignored.
Its not perfectly proper for me to mix one nding from the median
historians prole, that Lincoln voters ranked Douglas second, with the
remainder of Rikers prole. From the data in Tabarrok and Spector
(1999), I have borrowed in most cases and calculated in a few cases, how
the same voting rules would rank the entirety of the median historians
prole:
r
Pairwise comparison: Douglas > Bell > Lincoln > Breckinridge;
r
Borda count: Douglas > Bell > Lincoln > Breckinridge;
r
Approval voting (two votes): Douglas >Bell >Lincoln >Breckinridge;
r
Breckinridge.
Again, with plurality runoff, Lincoln and Douglas would go to the
runoff and Douglas would win (a runoff between rst-round losers Bell
and Breckinridge has Bell as the winner, yielding the overall ranking
D> L > Bl > Bk). All ve of these voting methods select Douglas as the
winner. The median historians prole yields rankings that are a bit more
similar than from Rikers prole. The remaining big difference between
the two proles is that the median historians ranks Bell just barely above
Lincoln. It is plausible that Bell as a more centrist candidate, attractive to
the Lower South, the Upper South, and a bit in the Lower North, might
do about as well as Lincoln as a less centrist candidate.
The champion of Riker may still have doubts and, all evidence to the
contrary, still insist that many Lincoln voters ranked Bell second. How
sensitive is Rikers assertion of a cycle? He says that 25 percent of Lincoln
voters ranked Douglas second and 75 percent ranked Bell second. Rikers
cycle assertion is fragile: if merely 30 percent of Lincoln voters rank
Douglas second and 70 percent Bell, then Rikers cycle vanishes. How
sensitive is the assertion of different outcomes fromdifferent voting rules?
If the second ranking of Lincoln voters is 33 percent for Douglas and
67 percent for Bell, then that is enough for pairwise comparison, Borda
count, approval voting with two votes, antiplurality, and plurality runoff
to pick Douglas as the winner. If 37 percent of Lincoln voters ranked
Douglas second and 63 percent ranked Bell second, then that is enough
for pairwise comparison, Borda count, approval voting with two votes,
and plurality runoff to converge on the same ranking: D > L > Bl > Br.
So Rikers case depends on the assertion that more than 63 percent of
Lincoln voters in the north ranked Bell, the candidate who had obtained
2 percent of the votes in the North, second. That is too fragile to carry
a case for pervasive political disequilibrium, a hypothesis which, as we
have seen, is otherwise unsupported.
Rikers primary argument in Liberalism against Populism (1982) is
the skeptical assertion that democracy is impossible because individual
preferences are unknowable. Because preferences are unknowable we
cannot ascertain whether or not outcomes of voting are accurate or fair;
hence democracy is inaccurate. Further, because preferences are un-
knowable, there is a possibility of manipulation, andwe cannot distinguish
manipulated from unmanipulated outcomes; hence, democracy is mean-
ingless. I criticized Rikers basic argument pattern as self-contradictory
in the second chapter. The unknowability of preferences also explains, in
Rikers scheme, why disequilibrium is pervasive yet difcult to identify.
In order to be persuaded to abandon the concept of the public good and
the idea of democracy as in some sense the expression of the peoples will,
most people would require that it be robustly demonstrated that manip-
ulation of outcomes be frequent, harmful, and irremediable. There is no
such demonstration. Because of his belief in the unknowability of pref-
erences, Riker also believes that the relevance of political disequilibrium
can only be demonstrated by means of artful anecdotes. We have ex-
amined the four major illustrations of disequilibrium in Liberalism against
Populismand found each to be in error. All of his examples fail, and hence
so does his overall case.
13 Antebellum politics concluded
Introduction
Pairwise comparison, the Borda count, approval voting with two votes,
negative plurality, and plurality runoff voting rules all choose Douglas as
the most preferred candidate; and all but the negative plurality voting rule
arrive at the same ranking: Douglas > Lincoln > Bell > Breckinridge.
Yet Douglas was the last-ranked candidate in the electoral college (though
second-ranked in the popular vote), and Lincoln won a majority in the
electoral college (and was rst-ranked but with a 40 percent plurality in
the popular tally). How is it that the actual election rule selected Lincoln
rather than Douglas? Does the fact that Lincoln rather than Douglas
was chosen support Rikers claim that democracy is arbitrary because
different voting rules result in different outcomes from the same prole
of preferences? Why was the 1860 election so peculiar?
Lincoln beat Douglas because with plurality rule and four major candi-
dates many voters preferences for Douglas over Lincoln were not counted
by the voting rule and because Lincoln absorbed Douglass votes in the
electoral college. Advocates of democracy have no obligation to defend
the presidential election system, I argue, which was clumsily designed
with antimajoritarian intent. Whether or not the election of Douglas,
the median voters candidate, would have prevented secession and war
is impossible to say; I suggest that the South may have seceded anyway.
Pure plurality rule usually elicits no more than two major candidates, but
the actual presidential election rule was more complex: rst, votes are
aggregated by states, then, generally, states cast votes for single winners
in the electoral college, and if the electoral college were to deadlock, its
top three candidates would go to the House where the South had better
strength, and if the House were to deadlock, then the Vice President be-
comes President if he had a majority in the electoral college, otherwise,
the top two Vice Presidential votegetters in the electoral college would
go to the Senate, controlled by the Democrats. The prospect of the elec-
tion going to Congress elicited four major candidacies, rather than two
281
as under pure plurality. I offer a novel hypothesis: that an unintended
consequence of the misunderstood strategic situation with four major
candidates was that sophisticated voting tragically made it appear that
the population was more polarized than it was in reality.
Then we undertake close examination of Rikers theory of manipula-
tion in multidimensional issue space, which he illustrates with the history
of the slavery issue in American politics from 1800 to 1860. Riker rejects
change in individual preferences, the standard hypothesis, as an explana-
tion for change in collective choices. Change in collective choices is due to
variations in the manipulative success of more potent actors in arbitrarily
imposing an outcome on the chaos of multidimensional disequilibrium,
he believes. The losing commercialist coalition repeatedly sought an issue
that would divide the winning agrarian coalition. With the slavery issue
they episodically came close to success in 1819 and 1846, and nally suc-
ceeded in 1860. An alternative hypothesis for the episodic nature of the
issue of slavery in the territories, that it was related to the minority South
maintaining its veto power in the US Senate, is more convincing. I nd
errors in Rikers historical narrative of the period. I present evidence in
support of the standard hypothesis that changes in collective attitudes and
actions on the slavery issue track changes in individual preferences, and
argue that the standard hypothesis better ts the evidence than Rikers
hypothesis of arbitrary manipulation.
Why did Lincoln beat Douglas?
Lincoln won an electoral college majority due to the distortions of two
institutions. First, it is well known that it is possible for a two-tier system
like the electoral college, in which a district casts a vote on behalf of the
majority (or plurality) winner in the district, to have perverse outcomes.
Suppose there are 100 districts and 100 voters in each district. Candi-
date Jerry Garcia wins 51 votes in each of 51 districts, but 0 votes in the
remaining 49 districts. Garcia is the majority winner in the electoral
college, even though he has received only 2,601 out of 10,000 popular
votes. If rival candidate Leonard Cohen wins the remaining 49 votes in
the 51 districts that each give Garcia 51 votes and Cohen wins all
100 votes in each of the remaining 49 districts, then Cohen with a total of
7,399 out of 10,000 popular votes loses to Garcia in the electoral college.
The candidate with 2,601 votes beats the candidate with 7,399 votes.
Second, we already know that plurality rule is potentially the least ac-
curate of the commonly used voting rules. Suppose there are two factions
in the population: those who favor electric blues and those who favor pop
music. The electric blues voters are united on candidate Jerry Garcia,
Antebellum politics concluded 283
who polls 27 percent of the vote. The pop voters are divided, 26 per-
cent rst-rank Britney Spears, 24 percent rst-rank Celine Dion, and
23 percent rst-rank Janet Jackson. The candidate of the electric-blues
voters wins, even though he is not the rst choice of 73 percent of the
population. Now combine the two institutions. In each of 51 districts
Garcia obtains 27 votes and is the plurality winner, in the remaining
49 districts Garcia obtains no votes. Garcia wins the election with 1,377
out of 10,000 votes. The candidate rst-ranked by 14 percent of the vot-
ers is the victor. Majority rule over two candidates uses all information
from voters and picks an obvious winner. Majority rule over more than
two candidates does not use all information, does pick an obvious winner,
but is often not decisive (e.g., when there are three candidates and none
gets a majority). Plurality rule over several candidates throws away large
amounts of ranking information. If there are four candidates and each
voter has complete preferences over the four, then ranking information is
potentially available for six pairwise comparisons (A:B, A:C, A:D, B:C,
B:D, C:D), and for 24 relative-rankings among four alternatives, A, B,
C, and D, but plurality rule gathers only each voters rst-ranking and
throws away all other information. To illustrate, return to Garcias vic-
tory with 1,377 votes. It is logically possible that each of the remaining
8,623 voters ranks Garcia last out of the four candidates; the candidate
last-ranked by 86 percent of the voters wins the election.
We rarely observe such distortions, however, for two reasons. First, the
examples provided are illustrative extremes, and for many distributions
of preferences the distortion would not come about. For district voting,
it is possible although unlikely that one candidate would have only 51
out of 100 votes in 51 out of 100 districts and 0 out of 100 votes in
49 out of 100 districts. For plurality voting, there is sometimes a clear
majority winner even when there are multiple candidates. Second, and
more important than the distribution of preferences, potential candidates
strategically interact with the voting rules. For district voting, the party
behind Leonard Cohen might choose a similar candidate, but one a bit
less polarizing than Cohen, say that they choose Bob Dylan who is a
lyricist like Cohen but a little more electric. Dylan will get fewer votes in
the 49 districts that were 100 percent for Cohen, but more votes in the
51 districts that were 51 percent for Garcia. The CohenDylan party only
needs to pull four votes fromthe Garcia voters and it can afford to sacrice
49 49 =2,401 properly distributed votes in its quest. Alternatively, the
Cohen party can stick with Cohen, but pull its electioneering resources
away from the 49 districts where it is the sure winner and towards the
51 districts where it is the near winner. Anticipating the strategic response
of the CohenDylan party, the Garcia party, if it is motivated to win
the election, would have to move far from Garcia and nd a candidate
with appeal across more districts, in short, a candidate with majority
appeal. For plurality voting in a single-member district, as we have seen,
voters who do not want to waste their votes will vote for one among
the two candidates they believe most likely to win, and weak candidates,
anticipating those strategic votes, are deterred from entering. Thus, in
the contest between electric blues and pop music, the pop voters will
steer their resources and their votes probably to Britney Spears, the top-
ranked among them, in order to defeat the icky Garcia. Those who back
Celine Dion and Janet Jackson strategically vote for Spears in order to
defeat Garcia. Only a few need to do so and, in the simple world we
assume, anticipation of such strategic voting deters Dion and Jackson
from entering the contest. Finally, if either voting by district aggregation
or plurality voting are a problem, there are easy institutional remedies,
for example, go from district voting to direct popular voting and go to
assured plurality runoffs. The electoral college in particular was intended
to be an antidemocratic institution, and democrats have no obligation to
defend its outcomes.
The electoral college ranking was Lincoln (majority) > Breckin-
ridge > Bell > Douglas and the popular vote was Lincoln (plurality) >
Douglas >Breckinridge >Bell. There are two questions to answer at the
outset. First, why is Douglas out of order, why is he second in the popu-
lar vote and last in the electoral-college vote? Second, how does Lincoln
move froma popular-vote plurality to an electoral-college majority? States
tended to follow a winner-take-all tradition; a state would cast all of its
electoral votes for the majority (or plurality) winner of popular votes in
that state. Lincoln won majorities in 15 out of the 18 states contributing
to his victory in the electoral college. The 15 states where Lincoln had
a majority contained 80 percent of Douglass nationwide popular vote
and 91 percent of Douglass free-state popular vote. The 18 states that
gave Lincoln all of their electoral votes contained 84 percent of Douglass
nationwide popular vote. Lincoln was boosted froma popular plurality to
an electoral-college majority by absorbing the votes that went to Douglas
in the free states. That is how Douglas, the candidate second-ranked in
the popular vote nished last in the electoral college vote. Next, why did
Lincoln win the plurality vote, when pairwise comparison, Borda count,
and the others selected Douglas? We have a discrepancy between the
plurality results (Lincoln > Douglas) > (Breckinridge > Bell), and the
results by most other reasonable voting rules (Douglas > Lincoln) >
(Bell > Breckinridge). For strategic reasons that I shall set forth shortly,
the national four-party race for the most part collapsed into two sectional
two-party races. In the North the two parties with most organizational
force were Lincolns Republicans and the Douglas Democrats; northern
voters tendednot to waste their votes onBreckinridge or Bell. Inthe South
the two parties with most organizational force were Bells Constitutional
Unionists and the Breckinridge Democrats; southern voters tended not
to waste their votes on Douglas, and Lincoln was not even on the bal-
lot. Most southerners would prefer Douglas to Lincoln, but that is not
the comparison they faced or reported preferences on. The information
that southerners preferred Douglas to Lincoln was not counted under
the plurality rule, which picked Lincoln over Douglas nationwide. The
other voting rules would have counted that information about southern
preferences, and thus would have picked Douglas over Lincoln nation-
wide. Most northerners would prefer Bell to Breckinridge, but that in-
formation was not counted under the plurality rule as well, which picked
Breckinridge over Bell nationwide. The other voting rules would have
counted that information about northern preferences and thus would
have picked Bell over Breckinridge nationwide. Plurality rule tends to
strategically elicit two major candidates between whom it can accurately
choose; thus, we are not alert to its inaccuracy in deciding among four
major candidates.
We also must remember that candidates are not exogenous to the voting
system. One voting system will encourage candidates A, B, C, D, E,
and F to enter the contest; another voting system will instead encourage
candidates C
and D
to enter, and another might inadvertently reward

mostly candidates A
and F
. Although most of the time most reasonable

voting rules will pick the same winner, sometimes they will not, and when
they do not this is due either to candidates positions being very close to
one another or is due to the interaction of potential candidates with the
rules of the voting system. The voting system in place in 1860 elicited
the candidacies of Lincoln, Douglas, Bell, and Breckinridge. A different
voting system may have elicited a different set of candidates. The voting
system that political actors faced in 1860 was not only a plurality race
in each state and then each state casting its vote in the electoral college.
There was also the provision that in event of no majority in the electoral
college, the top three candidates would go to the House where each state
would have one vote and if the House deadlocked that the top two Vice
Presidential candidates would effectively go to the Senate for decision.
Suppose that the system would have been different, that only a plurality
would have been required in the electoral college and no election would
be forwarded to the House. Then, speculatively, the smallest party, the
Constitutional Unionists (roughly speaking, former southern Whigs and
southern Americans) would not have formed or would have abandoned
their candidacy since they would no longer have a hope of being the
centrist brokers in a contest decided in the House. That would leave
the Republicans (formerly the northern Whigs and northern Americans,
among others), the northern Democrats (Douglas), and the southern
Democrats (Breckinridge). Continuing the speculation, the Republicans
would be motivated to pick a candidate at least as centrist as Lincoln so
as to appeal not only to the Upper and Lower North, but also to their
former Whig brethren the Constitutional Unionist voters in the Upper
South. The southern Democrats would not be tempted by the fantasy of
winning in the House or the Senate, and the Democrats further would be
motivated to beat the Republicans to the center and choose a candidate
like Douglas. The contest would then be fought over the center.
The purpose is only to illustrate that candidates are endogenous to the
voting system, not to argue that this would be the most probable out-
come in that counterfactual world. The point is that if candidates are
endogenous, then it may not be informative to observe that the hypothet-
ical application of voting system B to the actual candidates and rankings
endogenous to voting system A would result in a different winner. If the
hypothetical voting system B had been in place, then also an array of
candidates and rankings different from those endogenous to A would
have emerged. Thus, to establish that democracy is arbitrary it would not
be enough for Riker to show that Lincoln won the actual election and
that with the candidates and rankings endogenous to the actual election
that Douglas (or another) would have won one or another hypotheti-
cal election each with a different voting rule. Because of the possibility
of endogenous differences in the candidate menus he would also have
to establish that the voters underlying policy preferences that motivate
themto choose among candidates aggregate to arbitrarily different policy
outcomes from one reasonable voting rule to the next.
This can be illustrated by explication of a problem with my speculative
counterfactual, its unstated assumption that the Lower South did not
want secession. It assumes that the voting rule requires winning only a
plurality in the electoral college. In another counterfactual world if the
leaders of the Lower South want secession then just as in the actual case
they would be determined to split the Democratic Party. In the seces-
sionist counterfactual world, if a centersouth Democratic Party were to
have won the election, then the leaders of the Lower South would have
lacked the legitimacy to rally a critical mass to the cause of secession.
Imagine yet a third voting rule in 1860, direct popular plurality elec-
tion. Again, if there were secessionists they would be motivated to split
the Democratic Party for the same reason, to prevent a centrist victory.
The Democratic Partys unfortunate two-thirds rule made it such that the
secessionists could attain their aims regardless of Douglass response. If
they had succeeded in driving Douglas fromthe ticket then their southern
sectionalist minority platform would have lost the national election and
the northern sectionalist election victory would have been a justication
for secession. If, as did happen, they failed to drive Douglas from the
ticket, the most southern one-third of the party could claim the cloak of
legitimacy in splitting the party, and by splitting the Democrats deliver
the election to Lincoln again justifying secession.
I think the main problem was that many leaders in the Lower South
wanted secession in 1860 and probably would have had secession and
war under most likely voting rules. After all, provinces or regions occa-
sionally attempt secession from authoritarian regimes that lack any rea-
sonable voting rules.
1
With secession from nondemocratic regimes it is
clearly not the voting rule but rather the desire to secede that is the main
causal agent. In 1856 the southern Democrats favored Douglas at the
convention and approved of the noninterventionist Cincinnati platform.
In 1860 the same candidate and the same platform aroused their de-
structive wrath. At the conclusion of the Baltimore convention in 1860,
Douglas warned, Secession from the Democratic Party means seces-
sion from the federal Union (Johanssen 1973, 772). He also said there
that There is a mature plan throughout the Southern States to break
up the Union (Johanssen 1973, 790).
2
During the campaign Douglas
told an audience in Virginia that in case of secession, the President
of the United States, whoever he may be, should treat all attempts to
break up the Union, by resistance to its laws, as Old Hickory treated
the Nulliers in 1832 [i.e., credible military threat] (Johannsen 1973,
789). What would have happened if Douglas had won (and had avoided
death from illness in 1861)? It is impossible to say, of course. Perhaps he
would have been less provocative to the South and more able to nd a
workable compromise than Lincoln. As the leading northern Democrat
in the Senate, Douglas feverishly sought peace, and he chastised both
sections in that pursuit. He told a friend, however, Better a million of
men should fall on the battle eld than that this govt should lose one
single State! (Johannsen 1973, 813). After the fall of Fort Sumter to the
South Carolinians, he said, If I were President, Id convert or hang them
all within forty-eight hours, and, We must ght for our country and
forget all differences (Johannsen 1973, 860). Secession and war may
have followed even if Douglas had won.
A number of historians have devoted their scholarly lives to the topic
of the causation of the Civil War. I do not claim any original contribution
to that effort. My purpose is to borrow from standard secondary sources
in order to demonstrate the inaccuracies and implausibilities of Rikers
secondary account. As a student of politics who has an understanding of
voting rules, however, I would like to offer a modest hypothesis for further
investigation not a strong theory about the peculiar 1860 election. My
hypothesis is consistent with the historical sensibility that appreciates that
a concatenation of blunders, uncertainties, passions, and ambiguities can
forge a causal path to a destination wanted by almost no one (I must add
that because some social consequences are partly unintended does not
mean that all social consequences are wholly unintended). Lets begin by
comparing the elections of 1856 and 1860.
The South Americans and the remnants of the Whigs much the same
forces as those who would form the Constitutional Unionists (Bell) in
1860 separately nominated Millard Fillmore as their candidate in 1856.
In 1856, rst President Franklin Pierce and second Stephen Douglas
as manager in Congress were favored by southerners at the Democratic
convention because of their work to pass the KansasNebraska Act in
1854, but were opposed by northerners for the same reason. On the
seventeenth ballot, James Buchanan, conveniently absent as ambassador
abroad during the KansasNebraska deliberations, emerged as the com-
promise. The Democrats 1856 Cincinnati platform took the noninter-
ventionist popular-sovereignty position on the question of slavery in the
territories; Fillmore had a version of popular sovereignty as well. The
Republicans and the North Americans nominated John C. Fremont; they
were, of course, against slavery in the territories. Potter (1976, 259265),
whom I have been following, says that the three-way race reduced almost
to two separate elections, one between Democrat Buchanan and Republi-
can Fremont in the free states and another between Democrat Buchanan
and South-American and southern-Whig Fillmore in the slave states.
Potter observes that Buchanan might be expected to have suffered the
disadvantage in that his positions had to span both sections while his op-
ponents Fremont and Fillmore could appeal exclusively to northern or
southern voters respectively. Nevertheless, Potter observes that Buchanan
had the advantage because of what we would call in political science
a classic case of strategic voting: a great many northern citizens be-
came convinced that the election of Fremont meant disunion, and that
the candidate to vote for was the one who could beat Fremont. Since
Buchanan appeared to have the better chance, this tactical factor, as much
as anything else, brought about his victory and Fillmores ruin (263).
Buchanan won most of the slave states and some of the free states, the
rst President not to win both sections since 1828. Much is made of the
fact that Republican Lincoln received only 40 percent of the popular vote
in 1860, but this is a hangover from southern propaganda, as Democrat
Buchanan was also a plurality winner in 1856 with only 45 percent of the
popular vote. The South controlled the country for most of the period
from 1800 to 1860 in part because it was the majority faction in the ma-
jority party. Together, Fillmore, a prosouthern candidate, and Buchanan,
a southern-leaning candidate, polled 67 percent of the vote in 1856.
This was an era without opinion polls and without the communication
and transportation connections we enjoy today. People of one section
were able to generalize about people of the other section only on the ba-
sis of events so vivid as to merit national reportage. One of the most salient
pieces of evidence for contemporaries concerning the attitudes both of
ones own and of other regions were the election returns, which possess
the illusion of objective authority. Although there was an awareness of the
phenomenon of strategic voting, I would venture that most did not con-
sider howstrategic voting, plurality rule, and the electoral college garbled
the information to be found in election outcomes. The Democratic Party
scored its biggest victory ever in 1852 after the collapse of the Whigs,
Pierce gaining the presidency with the electoral college votes of 27 out
of 31 states and Democrats taking 63 percent of the House. Pierces to-
tals were bloated, however, by the defection of southern Whigs to the
Democrats; and he did not attain a popular majority in the North. Per-
haps imagining that their convincing victory was due to the popularity of
their policies rather than to collapse of the Whigs and the vagaries of the
electoral college, the Democrats mistakenly believed that they could ab-
rogate the Compromise of 1850 in order to set up Kansas for admission
as a slave state to balance Californias admission as a free state four years
before, and were deeply surprised by the consequent northern uproar.
After the KansasNebraska Act in 1854 they suffered their worst defeat,
falling to 18 percent of the House (Weingast 1998, 181).
Duvergers law is a probabilistic generalization that plurality elections
in single-member districts tend to the production of a two-party system.
As we have seen, this is due to strategic candidates anticipating strategic
voters. The US presidency is supercially a plurality election in a single-
member district, and this helps explain the tendency to two parties in
American politics. The mechanism is at work in local and state elections
as well, wherever there are plurality elections in single-member districts,
such as for governorships, most Congressional races, and many state leg-
islative races. Political forces in the US bifurcate at the local level, but
then have incentives to federate in order to win statewide ofces, and
further incentives to federate nationally in order to win the presidency,
helping perhaps to explain the crazy-quilt pattern within American po-
litical parties. Armed with Duvergers law lets reexamine the election
of 1856. There were three parties. The theory proposes that voters will
concentrate on the two parties believed most likely to win. Local and
state elections motivate party organizations to the complementary labor
of mobilizing for the partys presidential candidate as well. In the south the
two parties most likely to win in local and state races were the bisectional
Democrats and the sectional ex-Whig Fillmorites. The Democrats had
organizational force inmore of the South thandidthe ex-Whigs, ex-Whigs
won races where they were strong but overall Buchanan won the South.
In the North the two parties most likely to win in local and state races
were the bisectional Democrats and the sectional Republicans. Suppose
a voter is a moderate motivated to support union and peace over disunion
and war. If he is a voter in the South, he can vote for either Buchanan or
Fillmore. If he is a voter in the North, he must vote for Buchanan even if
he otherwise wants Fremonts policies, because the election of Fremont
might trigger southern secession. Enough voters in the North supported
Buchanan to provide him an electoral-college majority in that section.
When the southern-dominated Democratic Party took ofce in 1857
it could imagine that its fortunes had recovered with the election of
Buchanan, and southern forces could further imagine that the 67 per-
cent of the vote for Buchanan and Fillmore indicated a reconsidered na-
tional inclination to conciliate southern interests. Just as they took ofce,
moreover, the Supreme Court with the Dred Scott decision declared that
the southern position was constitutionally legitimate and thus that the
Republican Partys goals were politically impossible. In 1852 the Demo-
cratic Party, insensible to its true weakness, won with defecting southern
Whigs; in 1856 weaker yet they won again with defecting northern Whigs.
Consider the surprise of some then, at the election of Lincoln in 1860
and the plausibility of conspiracy theories in either section.
Look at the 1860 election with Duvergers law in mind. Now there is
no longer a bisectional Democratic Party and there are four parties in
total. If Duvergers law was at work at the national level with respect to
the Presidential race then there should be only two parties. But the Presi-
dential election is only supercially a plurality contest in a single-member
district. Remember, there is provision for the election go to the House if
there is no majority in the electoral college. In the House each state gets
one vote (more advantageous to the South than the electoral college), and
an election there could go either way. There were 34 state delegations in
the House, thus 18 states were needed for a majority. The Republicans
controlled 15 of the state delegations; the Democrats controlled 14 and
in each of 4 more states needed only to convert one South-Americanist
Representative; Tennessee had a majority of Americans and was unlikely
to support the Republicans. If the House deadlocked, the election would
go to the Senate where the Democrats had 38 votes, the Republicans 26,
and the Americans two (Fogel 1989, 382). The prospect of becoming
centrist brokers in a House election is part of what motivated the Upper
South candidacies of Fillmore in 1856 and Bell in 1860. In 1856 south-
ern forces were split between Buchanan and Fillmore, yet Buchanan and
his southern-leaning policies prevailed. What difference would it make in
1860 if a fourth party were added to the mix, some actors in the northern
and southern wings of the Democratic Party may have reasoned. When
there were four candidates in 1824, the election went to the House, and
fourth-ranked Clay from the Upper South was pivotal to the House elec-
tion. In 1836, the Democrats ran Martin Van Buren and Whig factions
ran three candidates. The Whigs gured that their three candidates would
appeal to three regions, and that they could decide on a winner in the
House, but instead Van Buren won a majority in the electoral college.
More parties make it more likely that there wont be a majority in the
electoral college; well muddle through and work things out in the House,
and with great luck we might even win in the Senate. The actors did not
anticipate that adding a fourth party amidst expectations of an election
possibly going to the House fundamentally altered the strategic situation.
Four parties and Duvergers law operating at the local and state level,
as we have seen, turned the four-way race into two sectional two-way
races.
In the 1856 race in the North between Democrat Buchanan and
Republican Fremont, a vote for Democrat Buchanan was a vote for mod-
eration. It was entirely different inthe four-way 1860 race. Extreme north-
ern voters would vote for Lincoln and extreme southern voters would vote
for Breckinridge. Heres how a moderate northern voter might reason,
however, to the extent he believed the electoral college would deadlock
and the election go to the House: I prefer the moderate unionist policies
such as those of Douglas, but if the southerners believe that we will sup-
port a moderate northern position then they will be motivated to support
an extreme southern position so that the compromise in the House is
drawn in their direction. Since the South would choose its more extreme
candidate the North should choose its more extreme candidate so that in
the House the compromise is nearer to our true position. The southern
voter reasons reciprocally: I prefer the moderate unionist policies such
as those of Bell, but if the northerners believe that we will support a
moderate southern position they will be motivated to support an extreme
northern position so that the compromise in the House is drawn in their
direction. Since the North would choose its more extreme candidate the
South should choose its more extreme candidate so that in the House the
compromise is nearer our true position. I suggest that for some voters in
each section the equilibrium was to vote for the more extreme candidate
in order to accomplish moderate ends. The North always had more elec-
toral votes than the South, but this had never made a difference before.
The unintended effect of the strategic voting for the more extreme can-
didates contributed to Lincolns margin of victory in the crucial Lower
North. The problem was exacerbated by the fact that plurality voting
among four major candidates did not register the information that south-
erners preferred Douglas to Lincoln and that northerners preferred Bell
to Breckinridge. Finally, the electoral college gave a majority to Lincoln
with only a plurality of the popular vote.
I do not claim that the collapsing-middle hypothesis is strongly sup-
ported, but if it did have any strength, consider the implications. I argued
that people gathered information on the attitudes of others from vivid
events such as the election returns. Suppose that a majority of people
in each section are moderate unionists (The New York Times before the
election said that the moderate position of popular sovereignty was sup-
ported by nine-tenths of the people, Morison 1971, 1117), but a majority
of people in each section vote for the more extreme candidate for strate-
gic reasons. A northerner reading the election returns is shocked he
is a moderate, but obviously the southerners are extremists; the south-
erner knows that he is a moderate but that the northerners are extremists.
Moreover, the moderate comes to believe that the people in his own sec-
tion are more extreme than they really are and that he is behind the
times. We have surprise, suspicion, feelings of betrayal, and quickening
polarization of attitudes, consistent with historical depictions of the out-
come. The hypothesis is not inconsistent with the proposition that the
leaders of the Lower South wanted secession. If the Democratic Party
would have united on Douglas in 1860 there was a chance they could
have reprised their Presidential victory of 1856. Some split the party be-
cause they wanted to rule or ruin, and some split the party because they
expected to muddle through, and it was the latter who were betrayed by
unintended effects. The election dynamic also might have made it easier
for the rule-or-ruiners in the South to mobilize the undecided in their
section.
Rikers theory of dimensional manipulation
Riker (1982, 197) says that the Arrowtheoremleads us to understand the
hitherto impenetrable mystery, the motive force for the perpetual ux in
politics. Riker criticizes as reductionist theories that attempt to explain
changes in preferences; it is the task of political science to explain political
change as a function not of preference change but of institutions variably
constraining pervasive disequilibrium. In multidimensional issue spaces
there is almost always an alternative policy majority-preferred to any pre-
vious winner; hence there is continuing and intense dissatisfaction for a
majority of political participants. The majority of dissatised losers has an
incentive to introduce destabilizing new issues (Arrow) and new dimen-
sions (McKelvey). Each voter sees politics in two or three dimensions,
different voters see different dimensions so that in total there may be
many dimensions, perhaps more dimensions than there are voters. Lead-
ers competing for position create and advocate new dimensions. Just as
Procter and Gamble offers new brands of soap on the consumer market,
so does the politician offer new issues and dimensions to the citizenry.
The economic and the political market each accepts or rejects the alterna-
tives offered. Market needs are well dened, however, and political needs
are ill dened. Therefore, politicians must try out alternatives more or
less randomly (210). The political market is more like organic nature
than it is like the economic market. Thus, the rise and fall of issues is a
process of natural selection, in which politicians, like genes, seek to sur-
vive and ourish (210). The mechanism of selection of political issues is
apparently institutions or constitutional structures (211). Politics, like
biological evolution, has no purpose.
Rikers account of the evolution of the slavery issue up to the Civil War
is an illustration of his theory of manipulation and the natural selection
of issues. The survey he hopes will establish the existence of manipulative
agenda control on a grand scale. The political losers, he argues, success-
fully and luckily introduced a new issue or dimension so as to generate
cycles or disequilibrium that they resolved with the Civil War thereby
xing their coalition as supreme in American politics for several gener-
ations. I accept part of Rikers story. From 1800 to 1860 there was an
intersectional coalition of agrarian expansionists Jeffersonian Republi-
canism and then Jacksonian Democracy that frequently controlled the
federal government. There was also a less successful commercial coali-
tion, rst organized as the Federalists, later organized as the intersectional
Whig Party, and nally organized as the Republicans, which constrained
the agrarians but less frequently controlled the federal government. The
commercial party desired high tariffs to protect industry, the agrarians
low tariffs to obtain cheaper imports and promote exports; the commer-
cial party wanted internal improvements such as canals and railroads, the
agrarian party did not directly benet from commercial development;
and so on. The multitude of issues imperfectly dividing the coalitions,
and their evolution over time, are more intriguingly complex than this
caricature, but we cant afford to go into detail here (see Ashworth 1995,
366492 for a thick discussion up to 1850). I also accept that from 1800
to 1860 politics was often organized along two dimensions, a usually
stronger dimension of economic conict and a usually uncorrelated and
weaker dimension of sectional conict.
Rikers argument is that the losing commercialists sought with one or
another issue to split the winning agrarians. They tried the slavery is-
sue several times and nally hit the jackpot with it in 1860. They used
their victory in the Civil War to establish a new institutional equilibrium
favorable to their coalition and unfavorable to the losers of that war.
In Rikers theory (1982, 200) it is the skill, energy, and resources of
leaders, rather than the preferences of the population, that account for
political change. For purposes of the theory it is as if the distribution
of preferences were xed from 1800 to 1860; the engine of change is
differences in politicians abilities to manipulate the agenda inside mul-
tidimensional issue space. Why does Riker assign causality to political
leaders? Because of the episodic nature of the crises over slavery. Slavery
existed for 200 years, and was settled by the constitutional convention
for 30 years when the rst crisis, the Missouri controversy, arose in 1819.
No economic factors account for the rise and fall of the slavery issue in
the two years of the Missouri controversy, underlying sentiments were
the same before and after, by elimination that leaves political leadership
as the explanatory variable, according to Riker. The slavery issue sub-
sided and lay quiescent until the controversy over the gag rule in the
1830s and then ripened into the episodic crises of the Wilmot Proviso
in 1846, the KansasNebraska Act of 1854, and the election of Lincoln
in 1860.
There is an alternative explanation, however. Weingast (1998), Rikers
quondam coauthor, explains that national crisis erupted whenever the
sectional equilibrium in the Senate was threatened. For purposes of ap-
portionment of representation in the House and the electoral college the
South enjoyed a
3
5
vote for every slave owned, it had an effective sectional
veto over selection of the President, it was the majority faction in the
majority party, it was overrepresented on the Supreme Court, and it also
enjoyed blocking power in the Senate. The US Senate is made up of two
senators fromevery state, andfrom1792 to 1858 the number of either free
or slave states almost never exceeded one another by more than one; as the
country expanded there was a deliberate effort especially by the South to
maintain an equal sectional balance in the Senate (Weingast 1998, 154).
Antislavery measures often passed the majoritarian House only to die at
the hands of the southern veto in the antimajoritarian Senate (Weingast
1998, 168). In 1819 it was proposed to admit Missouri as a slave state,
raising the question of whether the remainder of the vast but unsettled
Louisiana Territory would be open to slavery. Northerners opposed ad-
mitting Missouri as a slave state for one reason because it was well north
of the existing slave states; southerners opposed requiring Missouri to be
a free state in part because it was settled by southerners. The eventual
compromise, passed with many southern and fewer northern votes, was
to admit slave Missouri as an exception above that latitude but other-
wise close the remainder of the Louisiana Territory north of 36 degrees
30 minutes to slavery; also Maine was brought in as a free state to balance
Missouri in the Senate.
The next sectional crisis arose over the Wilmot Proviso starting in 1846,
which emerged just as the United States was almost doubling its terri-
tory with the Mexican and Oregon acquisitions. California was admitted
as a free state in 1850, giving the North an edge of one in the Senate,
and under old expectations the South would be due the admission of
a slave state. Blocked by the Missouri Compromise, the South had no
practical opportunity for the organization of a new slave state, however.
That is one reason why the Democrats at the height of their power in
1854 passed the KansasNebraska Act. The plan was to encourage ad-
mission of Kansas as a slave state (and Nebraska as a free state many
years later), but Kansas was north of 36 degrees 30 minutes and could
not be organized as a slave state unless the Missouri Compromise was
repealed; thus, another crisis. Minnesota and Oregon were admitted as
free states in 1858 and 1859 respectively, making for an imbalance of 18
free states and 15 slave states in the Senate. By 1860 the South had lost
not only the Senate, but the presidency, and then came the Civil War. Yes,
there is a political explanation for the succession of episodic slavery crises,
but it is the threat to sectional equilibrium in the Senate, not, as Riker
would have, it the arbitrarily variable activation of dimensions by political
operators.
It is not exactly true that slavery and other sectional issues were quies-
cent except for periodic crises. Even in 1776 South Carolina threatened to
secede over the question of federal taxation of slaves. Under the Articles
of Confederation there was a controversy over navigation rights on the
Mississippi that agitated the southerners; the Ordinance of 1787 forbade
slavery in the Northwest Territory. In the Federal convention of 1787,
delegates Gouverneur Morris and Rufus King excoriated slavery. At the
rst Congress in 1790, Benjamin Franklin forwarded a Quaker petition
beseeching federal action for the ultimate abolition of slavery (Stephens,
in Hesseltine 1962, 124). In 1798 the Virginia and Kentucky Resolu-
tions in response to the Alien and Sedition Acts were veiled threats of
southern secession. In 1801 the Federalist Party was obsessed with the
thought that Jefferson would not have won the presidency without the
arrangement permitting the slave states
3
5
of a persons worth of repre-
sentation for every slave owned. In 1804 the House voted to ban slavery
in the Orleans Territory (the measure lost in the Senate). Federalist lead-
ers based mostly in the northeast threatened secession and a separate
peace themselves during the War of 1812, destroying themselves as a
party in the process. There were other political controversies over slavery
in the west in 1784, 1785, 1789, 1802, 1804, and 1812. Then there was
the Missouri controversy in 1819 and 1820. After the Missouri Com-
promise, sectional hostility was not so much reduced as partly diverted
into other channels, primarily of an economic nature. Meanwhile, the
problem of slavery continued to be a troublesome element in American
politics, always rumbling below the surface and sometimes erupting into
controversy on the oors of Congress (Fehrenbacher 1995, xiii). In 1832
South Carolina called a state convention and unilaterally nullied high
federal tariffs, but President Jackson credibly threatened military action,
and the crisis was defused by compromise (Divine et al. 1999, 304306).
Abolition became an organized and mobilized force in the 1830s. From
1836 to 1844 Congress enforced a gag rule to prevent the acceptance
of citizens petitions against slavery. Northerners delayed admission of
slave Texas for almost a decade (18361845). Then came the Wilmot
Proviso and all that followed. (Except where noted otherwise, incidents
are collated from Moore 1967, 132; Riker 1982, 215; Weingast 1998,
168). This is merely a panoramic sampling; close-ups would show fre-
quent tensions over sectional issues including slavery. At the same time
even in the feverish 1850s slavery was by no means the only political
issue; people were concerned with other federal issues, and were much
more concerned with state and local issues, and of course supremely con-
cerned simply with leading their lives. The slavery conict was always on
a smolder though and would are, not arbitrarily, but when there was
reason for it to are.
Rikers account of the Missouri controversy marshals support for his
hypothesis of activation of dimensions by political leaders, but appears
not to be completely supported by his sources. Consonant with his theme
of arbitrary manipulation, Riker (1982, 217) tells us that:
The immediate origin of the Missouri issue can fairly be attributed to the pettiest
of politics. The motion to amend the Missouri bill was offered by James Tallmadge
of New York. He was about to become a candidate for the state senate from
New York City . . . [Democrat] Tallmadge had a special reason for raising the
slavery issue . . . In NewYork City there were a signicant number of black voters,
concentrated in marginal wards and typically Federalist in loyalty. Contemporary
writers as well as recent historians have asserted that, as a result of his amendment,
Tallmadge stood to gain black votes.
Riker cites Moore (1967, 3637) in support of his fair attribution. The al-
legation is a routine one against politicians concerned with racial equality
in the United States. During debate about the 17th Amendment in 1911
southerners belligerently charged that northern senators concerned over
negro voting rights in the South, and particularly New Yorker Depew,
were motivated by greed for negro votes in their states; some senators
were even known to have received negroes in their Senate ofces, they
charged. Gary Miller (1997, 182) notes that the same insistent deroga-
tion of the motives of those who challenged the received racial pattern
was made against Rufus King during the Missouri controversy in 1820,
and, In mid-twentieth century America, white proponents of civil rights
would be regularly accused of . . . taking the stands that they did in order
to win the Negro vote.
It takes historical excavation even to determine that in 1818 Tallmadge
was not a candidate to succeed himself in the US House of Representa-
tives. Tallmadge was ill (Moore 1967, 40) and also his son had just died,
and that may have been why he left the US Congress. In 1819 he was
nominated by the Clinton faction of the Democrats for the New York
State Senate. Moore (1967, 37) reports that it was believed his stand on
Missouri would help him in the state Senate election, but Moore does
not in any way suggest that this prospect is what motivated Tallmadges
Missouri stand. In 1817, Tallmadge had worked for the nal emancipa-
tion act in New York State, to free all slaves ten years hence (Freehling
1990, 144). Moreover, Tallmadge lost the state senate election and later
abandoned the Clinton faction for the rival Tammany faction that had
defeated him in 1819. Moore (1967, 36) tells us that Tallmadge was a
man of great ability who did not rise as far as he could have because he
was too much of an opportunist, changing sides too easily (that a politi-
cian would fail because he is too much the opportunist is an impossible
thesis in Rikers scheme). But does Moore (1967, 3839), Rikers source
for the charge, believe that Tallmadge was an opportunist on the slavery
question?
Though more of a politician than a statesman, Tallmadge did have deep convic-
tions and seems to have been motivated primarily by humanitarian and patriotic
considerations in opposing the extension of slavery to new states. He retired from
Congress soon after the long contest over Missouri began but continued to take
a lively interest in it and wrote letters to his friend Congressman John W. Taylor,
urging him to keep up the ght. In his correspondence with Taylor he spoke of
the contestants as being cupidity on one side and principle and suffering
human nature on the other. While doubtful that his amendment to the Missouri
bill would ever pass, he expressed a hope that it may have produced moral effects
which will eventually redeem our beloved country from Disgrace and Danger.
These sentiments were shared heartily by his family. His wife memorized portions
of a speech on the subject, and his parents were proud of him Especily, as his
religious mother wrote, because they thought he was on the Side of truth and
justice.
Riker claims that Tallmadge was an opportunist on the slavery issue, but
his sole citation is to a source which says he was not.
Tallmadges failed quest in early 1819 was in later 1819 taken up
by New Jersey Quakers and by philanthropists (the most entertainingly
named of the reform groups, I think, was the New Jersey Society for the
Suppression of Vice and Immorality and for the Encouragement of Virtue
and Good Morals, Moore 1967, 70), especially one Elias Boudinot who
also championed the Indians, and then was also taken up by prominent
New Yorker and Federalist Rufus King, and by others, and eventually
aroused the political North to public meetings, petitions, mandates
slavery in the territories became a major national issue. Now there is an
odd turn in Rikers argument. It was not Tallmadges allegedly oppor-
tunistic ploy, nor the organizational efforts of Boudinot, King, and their
ilk, but rather the interest of the large public (Riker 1982, 217) that
explains the event. He rejects a preference-change argument on the pre-
vious page (216): no great change of the American heart occurred in
18191820. But his argument about a large public interest must explain
why the interest was dormant before 1819 and after 1820.
What he means by the interest of a large public relates to the death of
the Federalist Party by 1819. The Federalists were commercialist and also
aristocratic; the electorate, expanded both by extension of the franchise
and settlement of the frontier, tended to be agrarian and wanted agrarian
policies from the national government; and the Federalist leaders had
arguably committed treason with respect to the War of 1812. Although
the Federalist Party was dead, its constituencies for high tariffs, internal
improvements, and commercial development still remained, even if some
of themhad been forced to nd political refuge in an unnatural home, the
Jeffersonian-Republicans. Thus, the commercialist losers were trying out
a new issue, slavery, in the hope of splitting the winning agrarian coali-
tion, the motive for the agitation was to nd a new and disequilibrating
program, a new agenda whereon dissident [Jeffersonian-] Republicans
and old Federalists could combine to win (Riker 1982, 218). They tried
out slavery in 1820, they tried out anti-Masonic agitation after 1830,
they tried out nativism after 1850, and they tried again with slavery after
nativism failed, the story goes. [T]he slavery issue was embraced with
enthusiasm in 1819 and repeatedly thereafter until it generated cyclical
majorities and civil war (Riker 1982, 219). The commercialists failed
the rst time they tried because the agrarian coalition squelched the slav-
ery issue with the Missouri Compromise; the issue was buried for a time
because the agrarians won and the commercialists lost. Rikers argument
is hard to grip. The motor force is at one point creation of new dimen-
sions by the Yankee elite, at another point it is the activation of latent
dimensions by the same elite, and at yet another it is already manifest
political preferences among the masses.
Conventional explanations, Riker says, for the rise of antislavery agita-
tion in the 1830s include the political secularization of religious enthusi-
asm, an explosion of philanthropic associations and reform movements,
antislavery among them. Among the various orid causes, only slavery
became a national political issue, so the argument frommoral zeal is with-
out force, according to Riker. A second explanation is that slavery was
in decline all over the world and would have arisen as an issue regard-
less of political institutions. A third explanation is that protable cotton
plantationism expanded mightily in the 1830s and that the expanding
territory, population, and wealth of the slave power increasingly threat-
ened free labor in the North. Riker rejects the global-decline and the
slave-power-expansion explanations as well because they do not account
for the Missouri agitation of 18191820. His alternative theory is that in
1820 and in the mid-1830s the commercialist faction promoted slavery
on the national agenda so as to try and split the agrarians. The com-
mercialist elite avoided the issue from 1820 to 1829 because they were
concentrated on the election and reelection of John Quincy Adams as
President (J.Q. Adams was Monroes Secretary of State, won the pres-
idency in the House in 1824, and lost to Jackson in 1828). After that,
the commercialists were distracted by the curious anti-Masonic political
movement. The northern commercial elite was also drawn into the Whig
coalition, a motley crew originally united only by its opposition to the
policies of President Jackson, and including southern elites who opposed
the economic policies, and crude populism and (white) majoritarianism
of Jackson; the northern Whigs were restrained on slavery from the need
to keep their southern allies in the coalition.
Slavery was always an evil but only sometimes a political issue, says
Riker. It appeared again on the national agenda around 1835 because it
was in the interest of the commercial elite to put it on the agenda. They
encouraged petitions from antislavery groups concerning the abolition of
the slave trade or of slavery in the District of Columbia (one place where
Congress had constitutional jurisdiction). Their motive, Riker says, was
to obstruct Congressional business. The Democrats responded with a gag
rule that effectively forbade the receipt of such petitions, in order to get
on with business and also to please Southern Democrats who wanted to
delegitimate antislavery sentiments. The leader against the gag rule was
former President John Quincy Adams, who had taken up a career as a gad-
y in the House of Representatives. J.Q. Adams, says Riker (1982, 222),
was the chief generator of the slavery issue that ultimately destroyed the
Jacksonian coalition. Thus, we are back to the familiar thesis of arbitrary
manipulation of multidimensional issue space with victory to the forces
possessing a superior will to power. When the Whigs won a majority in the
House they attempted to continue the gag rule in order to mollify their
southern wing, and moved to censure Adams. Adamss powerful rhetoric
forced a tabling of the motion to censure. Another northern antislavery
Whig, Giddings, was censured as well, but resigned and was triumphantly
reelected by his district. Then in 1844 the tiny antislavery Liberty third
party elded a Presidential candidate, who may have cost the Whigs the
election in New York by a few thousand votes (the Whigs won the Presi-
dential contest in New York four years before and four years after), and
thus for the presidency as well. Thereafter, northern politicians, Whig
and Democratic, had to worry about the antislavery issue, says Riker.
We are into extremely complicated territory here and in this space I
can do no more than sketch some issues. Rikers continual impugnment
of the motives of politicians becomes offensive. In one moment he tells us
that he does not mean to suggest that the great antislavery politicians J.Q.
Adams and Joshua Giddings were insincere opportunists. In the next mo-
ment he states that Adams never mentioned slavery as President and that
Giddings did not support black suffrage in Ohio in 1858, as if to take back
what he said. Adams and Giddings were not abolitionist cranks grinding
out fanatical pamphlets in some puritan village but were supremely prac-
tical politicians on the national stage. A similar tension is played out time
and again in politics: a practical politician who must actually satisfy the
many constraints on implementation will advocate the moderate version
of a cause whose followers take more extreme postures because they lack
the responsibility of governance. There is nothing insincere about prag-
matic moderation. Riker reconciles the tension by saying that Adams and
Giddings were Whigs rst and abolitionists second. This is inaccurate,
as Adams, allegedly the mastermind of the commercialist coalition, was
never fond of either party division or of the Whig Party; and Adams was
never an abolitionist in the strict sense of the term. And Adams must
have been quite the political genius in 1835 to foresee that, following his
death in 1844, although the slavery issue would destroy his Whig Party
in 1850 his Yankee grandsons would come roaring back with a sectional
Republican victory 25 years later in 1860. The Whig coalition certainly
never believed that Adamss or Giddingss agitations promoted the partys
political fortunes. And who initiated the gag rule controversy? Historian
Lee Benson (in Stampp 1974, 21) argues that southern leaders cultivated
the controversy in order to split the Whigs!
James H. Hammond [of South Carolina] accidentally stumbled on the issue of
trying to prevent Congress from receiving or routinely disposing of antislavery
petitions. Accidentally discovered by Hammond. . . what came to be known as
the Gag Rule issue was consciously used to heighten sectional antagonisms
and Southern national consciousness. If a small group of Northern abolitionists
had not existed, Southern nationalists, sectionalists and provincials would have
to create them as, in fact, they did.
Thus, if we did not have the luxury of reading history backwards, follow-
ing the unsurpassed victories of the Democratic Party in the year 1852 we
would have had to congratulate the southern agrarians for their brilliant
success in splitting and mastering the Whig commercialist coalition.
Also, I believe, Riker has his causation backwards. Did men such as
Adams and Giddings instrumentally take up issues in order to advance the
Whig coalition they belonged to, or did they instrumentally belong to the
Whig coalition in order to advance the issues they believed in? Adams had
aristocratic disdain for party spirit and had belonged to several. Giddings
was motivated to enter politics from antislavery conviction, and I imag-
ine that he would join whatever coalition would best advance his political
goals including prominently antislavery. Giddings abandoned the Whigs
for the Republicans. Further, I do not deny that politicians gleefully seek
to divide the opposition and that this was an important part of antebellum
politics. The disagreement is about the causality. For Riker political out-
comes are a function of arbitrary manipulation in multidimensional issue
space, no better than random. I think that, barring antidemocratic insti-
tutions, political outcomes are a function of the populations preferences,
and that collective change in a proper democracy tracks changes in the
opinions and the wishes of the political public. Yes, political gures such
as John Quincy Adams or Abraham Lincoln help form public opinion.
But they operate against constraints; they must begin from unordered
beliefs and desires in the population; they must craft a coherent, attrac-
tive, and useful vision; and the ultimate constraint on their ideological
leadership is its popularity. Lincoln, for example, succeeded because he
was luckily in the right places at the right times, because he was diligent,
talented, and moderate, but also because of his gift for wise and apt argu-
ment. His argumentation helped people form their intellectual and emo-
tional responses to difcult problems better than did the argumentation of
others.
Rikers story is that the commercialists kept at it with slavery until they
succeeded at generating a cyclical majority. The most compelling and
memorable part of his volume are the two case studies demonstrating
cycles with the Wilmot Proviso in 1846 and the Lincoln election in 1860
resulting in Civil War. We have seen, however, that the case studies are
mistaken. Without those two cases the multidimensional-disequilibrium
argument is like a joke without a punch line. It doesnt work. But there
are a few issues left to examine. Concluding his vignette on the Wilmot
Proviso, Riker (1982, 227) declares, the issue of slavery, rst broached
in 1819 as a means to break up the [agrarian] Jeffersonian intersectional
coalition, actually did so in 1846. He also explains why the commer-
cialists rst succeeded in 1846. The northern Whig commercialists had
often agitated on slavery but the northern Democrats who had resisted
until that point succumbed in 1846 because they were newly afraid of
losing elections on the slavery issue. Notice that Rikers is inadvertently a
preference-change argument here, not an argument from multidimen-
sional disequilibrium. Weingasts (1998) argument, that the sectional
balance in the Senate was threatened by ambiguity in the status of the
territories, is the more parsimonious explanation for the eruption of the
issue in 1820, 1846, and 1854. Finally, the complete disequilibrium of
1860, explains Riker (1982, 229), is how the slavery issue destroyed the
Jeffersonian-Jacksonian coalition, but we know now that his analysis of
1860 is wrong.
Three elements of Rikers doctrine are: that majorities are arbitrary and
meaningless (1982), that the Civil War was brought on by the Yankee elite
which had sought its opportunity for sixty years (1982, 213232), and
that the US Supreme Court should protect constitutionally guaranteed
property rights threatened by the arbitrary and meaningless nature of
majorities (Riker and Weingast 1988). These three elements happen to
be echoed in Confederate ideology. Edward Pollard, one of the Souths
leading apologists, wrote during the war in 1863 that it was a mistake
for northerners to consider that the Constitution resulted in the estab-
lishment of a grand consolidated government to be under the absolute
control of the numerical majority (Pollard, in Hesseltine, 1962, 44).
Pollard said that the Missouri Compromise of 1820:
afforded early and conclusive evidence of the secret disposition of all parties in the
North. . .The issue of the controversy was not only important to the slave interest,
but afforded a new development of the Northern political ideas of consolidation
and the absolutismof numerical majorities. The North had acted on the Missouri
matter as if the South had no rights guaranteed in the bond of the Union, and
as though the question at issue was one merely of numerical strength. . .The
majority must govern was the decantatum on the lips of every demagogue, and
passed into a favorite phrase of Northern politics.
The Richmond, Virginia Semi-Weekly Examiner complained in 1860 that
in the Northern States the popular power has no check but the popular
reason and will . . . There is but one defense . . . it is the ability and the will
of the minority to resist the action of the ruling majority (in Stampp
1974, 142143). The writer explains that the conservative inuence of
domestic slavery and other class distinctions serve as a desirable check
on the popular will in the South.
Hesseltine (1962, 38) says that, In the viewof Alexander H. Stephens,
Vice-President of the Confederacy, the Southerners were not only exer-
cising their constitutional rights, but were defending the constitution of
the Founding Fathers from the victorious conspirators who had seized
the apparatus of the federal government. Stephens (in Hesseltine 1962,
125) himself says that the usurpers had awaited their moment for sixty
years:
This Party, as we have seen, [in 1798] assumed the popular name of Federal, as it
assumed the popular name of Republican in 1860. [The Alien and Sedition Acts]
of 1798 came near stirring up a civil war, and would most probably have resulted
in such a catastrophe if the Party so organized with such principles and objects
had not been utterly overthrown and driven from power by the advocates of our
true Federal system of Government, under the lead of Mr. Jefferson, in 1800. It
was after this complete defeat on these other questions that the Centralists rallied
upon this question of the status of the black race in the States, where it continued
to exist, as the most promising one for them to agitate and unite the people of
the Northern States upon, for the accomplishment of their sinister objects of
National Centralization or Consolidation.
Finally, the Supreme Courts disastrous Dred Scott decision interpreted
the Constitution to read that the property right in slaves should be pro-
tected against the depredations of majorities. Riker and Weingast (1988,
373) call in the Virginia Law Review for a return to the jurisprudence
of the nineteenth and early twentieth century, in which the Supreme
Court would interpret the Constitution to read that property rights
should be protected against the depredations of majorities. I am not sug-
gesting that Riker and Weingast wish to reinstitute slavery. I am mis-
chievously pointing out that they neglect to address an awkward implica-
tion of their argument.
Preference developments from 1800 to 1860
The more conventional and the more parsimonious category of explana-
tion for political changes in the period under study is developments in
the populations preferences. We have seen the evidence for the disequi-
librium hypothesis. What is the evidence for the preference-development
hypothesis?
The slavery issue was plucked by a canny cabal from the chaos of mul-
tidimensional issue space in order to split and subordinate their enemies,
the story goes. If the abolition of slavery is just a matter of arbitrary manip-
ulation, however, then we should see among similar societies variations in
the presence of the institution of slavery. In some democracies it happened
that slavery was the issue used by one faction to subordinate another, but
in other countries it happened that other issues served that purpose; such
that in some countries slavery was abolished and in some countries it
was not. My suggested prediction from Rikers premises fails, however.
Among the developed democracies slavery is extremely rare in practice
and is universally and strictly prohibited by law. This suggests that there
is something to the global-development hypothesis that Riker rejects. In
1820 all the Western Hemisphere but for Canada, the northern United
States, Haiti, and the Dominican Republic was slave territory. Slavery was
abolished in the four Central American countries and Chile in 1823, in
Mexico in 1829, in all British possessions in 1834, in Paraguay in 1843, in
all French possessions in 1848, in Bolivia in 1851, in Colombia in 1852,
in Ecuador and Venezuela in 1854, in Peru in 1855. By the time of the US
Civil War only Cuba, Puerto Rico, and Brazil continued to condone slav-
ery. The day before Lincolns inauguration the Tsar of Russia freed the
serfs (see Nevins 1950, 132170 on the global decline of slavery). While
the struggle to end slavery was often associated with violence, it was only
in the United States that slaveowners resorted to full-scale warfare to halt
the abolitionist trend (Fogel 1989, 205). The southern ideologists re-
jected the global-development hypothesis as well. In his bill of particulars
against the North, apologist Pollard (in Hesseltine 1962, 55), discusses
international opinion just after loss of sectional equilibrium in the Senate
and unfair tariffs and before the horror of John Browns Harpers Ferry
raid:
It was said that the South was an inferior part of the country; that she was a
spotted and degraded section; that the national fame abroad was compromised
by the association of the South in the Union; and that a New England traveler
in Europe blushed to confess himself an American because half of the nation of
that name were slaveholders.
The plaint of the wounded bully is ill concealed.
Poole and Rosenthal (1997) set out to measure the dimensionality of
all roll-call votes in the US Congress. As we have seen, they found that
for all such votes, 83 percent are explained by one dimension and another
2 percent are explained by a second dimension. The rst dimension usu-
ally captures party loyalty and the second dimension usually differentiates
the members by region within each party (4651). Before the rise of the
second, Democrat and Whig, party system in the mid-1830s the second
dimension frequently involved public works. From 1837 through 1850
the second dimension frequently involved slavery or public lands. The
Congress of 1851 and 1852, after the Compromise of 1850, is spatially
chaotic, completely disorganized in dimensional terms. By the Congress
of 1853 and 1854 slavery had become the rst dimension of politics and
remained so into the Civil War. Furthermore, the major vehicle for re-
alignment in the antebellum period was replacement of legislators, not
changes in legislators positions (90), which I interpret to mean that the
change was responsive to the electorates wishes. Independently from
Poole and Rosenthal and their roll-call data, historian Fogel (1989, 321)
identies the Congress of 18511852 as a period of transition. The po-
litical agenda of the prior decades was obsolete and a new agenda was
yet to emerge. That Congress was split into a multiplicity of factional
groupings who only occasionally coalesced neither party nor section
was inuential. The primary political issues in 18511854 were local,
responses to the turmoil of mass immigration, urbanization, and early
industrialization in the North.
Poole and Rosenthal also trace the history of roll-call votes on slav-
ery issues (1997, 91100). In the rst 14 Congresses there were only
43 roll-call votes on the issue and only weak party or sectional patterns.
Almost three-fourths of the roll-call votes on slavery between 1817 and
1831 took place in the 15th (18171818) and 16th (18191820) Con-
gresses. Historians call this the Era of Good Feelings; controversy over
foreign wars had died away, the Jeffersonian-Republicans had adopted
some of the Federalists economic policies, because of the demise of the
Federalists there was no partisan divide, and President James Monroe
was reelected in 1820 with only one symbolic vote opposed in the elec-
toral college. The Missouri controversy played out in 1819 and 1820,
and slavery votes t well into one dimension (indeed is the only issue
with a high degree of t), but otherwise the spatial model is quite weak
in this period, the 17th Congress (18211822) being the worst-tting
to the model in American history. Poole and Rosenthal state that the
collapse of the party system during the Era of Good Feelings did not
occur because slavery was the new, destabilizing dimension (1997, 95).
The collapse of the rst, Republican and Federalist, party system came
about because the foreign-policy and economic issues that structured it
were no longer of importance. Slavery rose as a Congressional issue with
the Missouri controversy but fell with the Missouri Compromise. The
second, Democrat and Whig, party system arose along an economic di-
mension but slavery never completely vanished as an issue, according to
Poole and Rosenthal. The bulk of slavery votes came after 1835, they
say, and slavery voting fell increasingly along a second dimension. By
1853, this [rst] economic dimension collapsed and was replaced by the
slavery dimension (95). I suggest rather that after 1853 the slavery di-
mension merged with the economic dimension. What attraction did the
Democratic Party hold for the northern agrarians? Originally, they were
drawn to the Jacksonian coalition because it promised greater democratic
equality and liberty (for white males) than did the aristocratic commer-
cialists (see Fogel 1989, 316319). Democrats advocated territorial ex-
pansion and the cheap sale of federal land to promote those values, and
Whigs opposed territorial expansion and advocated that federal land be
sold at high prices in order not to depress the price of the land they owned
or inate the price of the labor they employed. During the Congress of
18531854 the Democrats abandoned their dening cheap-land policy
because their southern wing had concluded that the policy politically ad-
vantaged the North. In the same Congress the Democrats abrogated the
boundary of the Missouri Compromise, and thereafter slave settlers and
free settlers clashed in the territory of Kansas. Meanwhile, over the ob-
jections of some of its formerly Whig members, the Republican Party
took up the cause of cheap land. Northern agrarians were drawn to
the new party that promoted their democratic equality and liberty, the
Republicans.
Poole and Rosenthals ndings suggest that tensions over slavery in-
creased over a long period, that the realignment within Congress was sud-
den, and the realignment was initiated well before the Republican Party
became a force in politics (1997, 99). I interpret Poole and Rosenthals
data to mean that there was not a common multidimensional issue space
from1800 to 1860 andthus that variations inpolitical outcome from1800
to 1860 were not due to the arbitrary manipulation of that space by politi-
cal conspirators. Poole and Rosenthal showclearly that there was a strong
party dimension and a weak sectional dimension and that by 1853 the
sectional dimension became strong and in its wake restructured the par-
ties along sectional lines. It was not, however, the Slave Power Conspiracy
nor the Black Republicans (phrases of the respective sectional conspiracy
theorists) who brought about the realignment, rather it was changes in the
preferences of the political population. By the 1850s, slavery was not a
newissue but a very old one that had become more intense in both the North
and the South (emphasis added, 91). Weingast (1998, 163) too holds
that northern opposition to slavery increased over the antebellum period.
According to Potter (1976, 3841), during the colonial period there was
little difference between the sections concerning the morality of slavery,
although slavery was far more prevalent in the South. The ideals of the
Revolutionary War inspired both North and South to condemn slavery
as evil; emancipation and colonization societies were as common in the
Upper South as they were in the North. But as the protable cotton econ-
omy expanded and slavery with it, the economic and ideological center of
the South shifted from Virginia to South Carolina. By 1832 the southern
antislavery movement had vanished and an ideology of slavery as a pos-
itive good had arisen. Meanwhile, slavery had been gradually abolished
in the North. Sectional opinion on slavery further polarized in the 1840s
and 1850s.
Weingast (1998) criticizes Rikers model for failing to explain the sta-
bility of the second two-party system even with slavery always lurking in
the background and for failing to explain why collapse came in the 1850s.
The crisis of 18191820 and the crisis of 18461850 were each resolved,
but the crisis of the 1850s was not. Weingast explains the change as a
product of four factors. First, the slave states needed to expand in num-
ber in order to preserve sectional veto in the Senate but in the 1850s they
ran into constraints; due to the increasing price of slaves expansion was
not economically feasible in available new territories and politically slave
settlers were hemmed in by the Missouri Compromise. Second, immi-
gration to the North changed the size of the population and the nature
of its economy and society. In 1800 the sections were roughly equal with
two million souls each; in 1860 the North had a population of 20 million
compared to 7 million whites and 4 million blacks in the South. It became
more feasible to organize an all-northern party. Third, early in the pe-
riod settlers in what we now call the Midwest were mostly self-sufcient
and what commerce there was went south along the Mississippi River.
Later in the period, Midwest inhabitants produced more for the market
and shipped their goods east by water or by rail. The second and third
factors meant that the pivotal voter became more oriented to the North
than to the South; I call this a matter of preference development. Fourth,
Americans at the time did not fully understand the political implications
of the second and third factors. We know now that political space had
collapsed and that the Democrats were about to collapse with it, but
in 1853 and 1854 the Democrats believed they were at their strongest.
They passed the KansasNebraska Act only to drive northerners out of
the Democratic Party and into the anti-Nebraska groupings that became
the Republican Party.
Finally, Americans early on had dealt with slavery by means of geo-
graphical separation. There were free states and there were slave states,
and most of life revolved around local or, at best, state concerns. The fed-
eralist institutions of the Constitution reinforced the separation. When
the destabilizing issue of slavery in the territories arose, the natural solu-
tion was geographical separation as well. Slavery shall be allowed south
of this line and shall be forbidden north of this line. Later on, increas-
ing population, communication, and transportation brought the sections
into increasing contact. Notice that trouble began when geographical
separation failed. The KansasNebraska Act violated the boundary of
separation the northerners thought they had. The subsequent mixture
of free settlers and slave settlers in Kansas resulting from the Demo-
cratic doctrine of popular sovereignty proved a political and a psycholog-
ical nightmare it drove John Brown mad. Most people are familiar
with Sewards (in Stampp 1974, 105) remark about an irrepress-
ible conict, but few know the ne causal analysis that prefaces his
remark:
Hitherto, the two systems have existed in different States, but side by side in
the American Union. This has happened because the Union is a confederation
of States. But in another aspect the United States constitutes only one nation.
Increase of population, which is lling the States out of their very borders, to-
gether with a new and expanded network of railroads and other avenues, and
an internal commerce which daily becomes more intimate, is rapidly bringing
the States into a higher and more perfect social unity or consolidation. Thus,
these antagonistic systems are continually coming into closer contact, and col-
lision results. Shall I tell you what this collision means? They who think it is
accidental, unnecessary, the work of interested or fanatical agitators, and there-
fore ephemeral, mistake the case altogether. It is an irrepressible conict between
opposing and enduring forces, and it means that the United States must and will
sooner or later become either entirely a slave-holding nationor entirely a free-labor
nation.
The northern public wanted both to oppose slavery and support the Con-
stitution and the Union, according to Potter (1976, 4650), and they
managed those contradictory goals by keeping them contextually sepa-
rated. They accepted the constitutional obligation that left the question
to each state and they abolished slavery in their own northern states. They
had no personal responsibility for the actions of the southern states, and
they believed that slavery would eventually die out. Anything that tended
to expose the incompatibility of the values of antislavery and Union by
bringing themto the same level and forcing themto confront one another
in the same context was . . . extremely threatening to the northern mind.
This was why the abolitionists incurred so much hostility (Potter 1976,
47). But where the abolitionists failed, the acquisition in 1846 of vast new
territories subject to federal jurisdiction forced the values of antislavery
and Union repeatedly into the same context, and the contradiction be-
tween them could no longer be managed by techniques of geographical
or emotional compartmentalization.
Riker promised to explain to us the perpetual ux of politics, but his ev-
idence was faulty. Nor can he explain to us stable continuity in politics
slavery was a tumultuous issue in the antebellum period, but after the
Civil War slavery did not return as an institution nor has it returned to
any country that has effectively abolished it. The great political and moral
drama of the Civil War was not a consequence of arbitrary manipulation
of the agenda in a multidimensional issue space. As Seward said, They
who think it is accidental, unnecessary, the work of interested or fanat-
ical agitators, and therefore ephemeral, mistake the case altogether (in
Stampp 1974, 105).
14 More of Rikers cycles debunked
Introduction
We shall return to the problem of manipulation by introduction of new
issues and dimensions. Meanwhile, we shall examine the remaining pub-
lished and developed anecdotes of cycling I have been able to nd in
the political science literature, beginning with Rikers in this chapter, fol-
lowed by others claims in the next chapter. In the rst case, Riker detects
a cycle in the deliberations of the Convention that crafted the US Con-
stitution. The question was how best to select the executive of the new
regime. Riker believes there was a cycle among three alternatives for
the national legislature to select the executive by joint ballot of the two
chambers, the same but by separate ballot of the two chambers, and se-
lection by electors in the states and hence that the nal outcome was
arbitrarily decided by the more intense will to win of the faction that
favored selection by electors. I show that Rikers purported cycle arises
from a failure to distinguish among similar but not identical alternatives.
I argue that it is a more plausible interpretation of the record to distin-
guish among similar alternatives, and if this is done, the reversal, tie, and
cycle alleged by Riker vanish. The Convention, supposedly deadlocked in
cyclic indeterminacy on the question of the selection of the executive, ap-
pointed a committee to resolve the question overwhelmingly dominated
by supporters of selection by electors. Appointment of the committee is
consistent with my equilibriuminterpretation of events, but is an anomaly
inconsistent with Rikers disequilibrium interpretation.
Rikers rst published attempt to demonstrate an Arrovian cycle
was in 1958, a study of a sequence of votes on agricultural appro-
priations in the US House of Representatives. Riker believes he has
demonstrated a collective ranking intransitive in dollar appropriations
(250 > 200 > 225 > 142 > 100) that indicates the presence of a cycle.
Riker obtains his cycle, however, by counting 13 or more voters who
voted strategically for 200 > 225 million dollars. Riker is explicit that the
sincere ranking of these 13 or more voters was 225 > 200. If the sincere
310
More of Rikers cycles debunked 311
rankings of all voters are aggregated then the outcome is transitive (250 >
225 > 200 > 142 > 100) and there is no cycle. In the absence of
strategic voting, the agenda sequence dictated by House rules would
have prevented attainment of the collectively highest-ranked alternative
of 250 million dollars; but with strategic voting, the House selected
its highest-ranked alternative. Rikers nding of a cycle is conceptually
mistaken.
Federal Convention, 1787
Riker (1984), in his presidential address to the American Political Science
Association, believes he has found a cycle in the records of the Federal
Convention of 1787. He provides much more detail on this case than is
his wont; consequently, there is less need for me to add historical context.
Also, I shall neglect many details of his essay and focus on his dramatic
central claim of a cycle that changed the content of the US Constitution.
One of the more difcult issues at the Convention was the question
of what method should be used to select presidents of the new republic.
The Convention began its deliberations from the so-called Virginia plan,
which among other things recommended the election of the President by
the national legislature. Those behind the Virginia plan and ultimately the
nal Constitution were nationalists who believed that the prior Confed-
eration was an ineffectively loose alliance of states ruled by irresponsible
state legislatures, a threat to commerce and property within, and endan-
gered by foreign machinations without. Thus, the original proposal was
that the executive be selected by the national legislature rather than by
the states. This original proposal did not cohere, however, with another
prominent opinion among the nationalists, that the powerful newgovern-
ment be self-constrained by checks and balances among its branches. If
the President were selected by the national legislature, then he would not
be independent, rather the President would be the creature of the legis-
lature, and intrigue over appointment would be one of the undesirable
consequences (see Madison, II 34).
1
Those who were not so strongly
nationalist wanted the states to have some role in the selection of the ex-
ecutive, and a minority perhaps of a more populist bent preferred that the
President be elected by the people. Whether by the national legislature,
by state governments, or by the people, there remained the controversy
of how to weight the interests of the small states against the large states,
and the free states against the slave states, in the selection of an executive.
Just prior to rst serious consideration of the question of the executive
the convention had reached the Great Compromise, providing for equal
representation of states in the Senate and representation proportional to
population in the House of Representatives. One issue was which agents
should select the executive, a second was how to count the votes of those
agents, and a third complication was the question of length of term for
the executive and eligibility for reelection. There was a desire to reward
Washingtons ability, but a fear of eventual monarchy or worse arising
from the strong executive power. Finally, the three dimensions of con-
cern were not independent of one another, an adjustment in one might
require an adjustment in another.
Election of the executive by the national legislature was initially a fea-
ture of both the Virginia (large states) plan and the New Jersey (small
states) plan. That mode of selection was approved on rst test, June 2,
eight votes to two, Pennsylvania and Maryland opposed (Roll Call #12,
II 81). On July 17, a motion for election by the people failed, only
Pennsylvania in the afrmative (Roll Call #165, II 32); a motion for elec-
tion by electors failed, only Maryland and Delaware in the afrmative
(Roll Call #166, II 32); and a motion for election by the national legisla-
ture passedunanimously (Roll Call #167, II 32). The rst contest of inter-
est to our more formal analysis was a motiononThursday, July 19 to strike
that the President be selected by the national legislature and insert that
he be selected by electors appointed by legislatures of the states, and that
states less than 100,000 in population be allocated one elector, those be-
tween 100,000 and 300,000 two electors, and those above 300,000 three
electors. The portion providing for electors (Roll Call #182, II, 5059)
passed with six votes for (CT, NJ, PA, DE, MD, VA), three votes against
(NC, SC, GA), one state divided (MA), and two absences (NHand NY);
note that Rhode Island did not participate at all in the convention. The
portion providing that those electors be appointed by the state legislatures
(Roll Call #183) passed with eight afrmative votes, Virginia and South
Carolina opposing. The portion allocating different numbers of electors
to the states was agreed to the following day (Roll Call #191, II 6070).
Label the alternative of selection of the executive by electors appointed
by the state legislatures on July 19 as B. Label selection of the executive
by the national legislature, originally proposed by the Virginia Plan, as
A. On July 19 the convention voted for B > A. A six-year term of ofce
was also agreed to, and a proposal to forbid reelection failed.
On Monday, July 23, Houstoun (GA) moved and Spaight (NC) sec-
onded to reconsider the question of the mode of appointment of the
executive (Roll Call #208, II 95). Notice that Houstouns Georgia and
Spaights North Carolina, along with South Carolina (Butler of SC com-
plained about the distance to send electors on July 19, II 59), opposed
election by electors on July 19: Houston urged the extreme inconve-
niency & the considerable expense, of drawing together men from all the
States for the single purpose of electing the Chief Magistrate (II 95). At
this point it was assumed that electors would meet in the national capi-
tal, and that the capital would be New York or Philadelphia (McDonald
1985, 245). Houstouns July 23 motion to reconsider passed with seven
votes for (NH, MA, CT, DE, NC, SC, GA) and three votes against (PA,
MD, VA). On the same day constitutional questions other than those
pertaining to the executive were unanimously referred to a ve-member
Committee on Detail. On Tuesday, July 24 Houstoun moved to strike
by electors appointed for that purpose by the Legislatures of the States
and to insert by the national Legislature (Roll Call #215, II 98) and the
motion succeeded, with seven votes for (NH, MA, NJ, DE, NC, SC, GA)
and four votes against (CT, PA, MD, VA). It is not clear in the record, but
it seems to me that Houstouns motion must have repealed by implication
the scheme allocating different numbers of electors to different states; if
there were no electors then the scheme of allocating electors would be nu-
gatory. Houstoun dwelt chiey on the improbability, that capable men
would undertake the service of Electors from the more distant States
and Spaight again seconded Houstons motion (II, 99). Label the al-
ternative of selection of the executive by electors appointed by the state
legislatures on July 24 as B
. On July 24, the convention voted for A >B
.
This was a reversal of the vote of July 19. Whose votes changed? New
Hampshire was not yet at the Convention on July 19, but on July 24
voted for election of the President by the national legislature. Mas-
sachusetts was divided on July 19, but on July 24 voted for election by
national legislature. The New Jersey and Delaware delegations changed
from supporting election by electors on July 19 to supporting election by
national legislature on July 24. Riker (1984, 1011) guesses that New
Jersey and Delaware were inconsistent due to shifting absences among
the members of their delegations.
We might also hypothesize that the delegates to the Convention were
rational. If they were rational, then two further possibilities come to mind.
First, that when they voted on July 19 many delegates failed to consider
the question of the expense and difculty of convening electors, but that
by July 24 they realized that the expense and difculty was too much. If B
is cheap election by electors as mistakenly understood on July 19 and B
is expensive election of electors as correctly understood on July 24, then

delegates were consistent in preferring B > A > B
. Second, New Jersey

and Delaware were members of the small-state bloc. Excluding Virginia,
the most populous state, and Delaware, the least, the large states (MA,
PA, VA, NC, SC, GA) were on average only about 10 percent larger
in population, but were much larger in territory and thus more likely
to gain population, than the small states (NH, CT, NY, NJ, DE, MD)
(McDonald 1985, 218). The distinction had little to do with population;
rather, the large states had claim to western lands and the small states
did not (Farrand 1913, 82). One might speculate that some members
of the New Jersey and Delaware delegations (although their small-state
allies Connecticut and Maryland favored election by electors both on
July 19 and July 24) observed that Pennsylvania and Virginia, two of the
largest states, strongly wanted election by electors, and that by holding
out New Jersey and Delaware imagined that they might extract favorable
concessions from Pennsylvania and Virginia on the allocation of electors
among states. The two possibilities are not necessarily exclusive. I do not
know of direct evidence for the second possibility in the record, but later
I shall describe indications that some small states sometimes held out on
voting for election by electors in order to extract concessions from the
large states.
With an executive elected by electors the delegates had approved of a
six-year term open to reelection. After returning to election by national
legislature, on July 26 they changed to a single seven-year term for the
executive (Roll Call #224), so as to prevent intrigue in reappointment.
Finally, the entire scheme for the executive including several elements
not discussed here was approved on Thursday, July 26 by a vote of six
for, three against, and one state divided (Roll Call #225, II 118); New
Jersey remained in the afrmative but Delaware went to the negative on
the entire scheme. If we label the entire scheme as A
, then the convention

preference was A
> B
just as the convention preference merely for elec-

tion by national legislature was A > B
. For our purposes A
, the entire
executive scheme, is equivalent to A, the portion of the scheme directing
that the executive be chosen by the national legislature, hence A
A
(one might argue that A
> A, but neither of these inferences are essen-

tial to my interpretation). The convention then adjourned until Monday,
August 6, so that the Committee on Detail could do its work.
Although on July 26 the Convention returned to election of the exec-
utive by the national legislature, debate on the question raised problems
with the national-legislature alternative and suggested solutions to those
problems which inuenced the ultimate outcome in the Convention.
Madisons objections to the national-legislature alternative were quite
cogent, although his suggested alternative of election by the people was
politically infeasible. McDonald (1985, 246) is of the opinion that one
of Madisons arguments, that the equivalent of selection of the executive
by the national legislature in Germany and in Poland was much inu-
enced by foreign interference (II 110), affrighted a number of delegates
into moving toward a decentralized election. One problem with elec-
tion by electors is that the delegates believed that electors would mostly
vote for candidates from their own states, with divisive and unstable re-
sults. Williamson suggested that electors vote for three candidates, and
G. Morris immediately suggested that they be required to vote for two,
at least one not from their own state (II 113114). This idea was used in
the nal compromise.
The next Convention debate on the question of the executive is a month
later, on Friday, August 24. It was an inconclusive and confusing day; im-
portant motions on the topic failed, some on tie votes. Rutledge moved to
amend election of the executive by the national legislature to require elec-
tion by the joint ballot of the House and the Senate, and this succeeded
seven to four, the larger states tending for and the smaller states tending
against (Roll Call #356, II 399). If the two chambers balloted separately
then that would increase the bargaining power of the small states because
they would enjoy equal representation in the Senate; a joint ballot would
decrease small-state bargaining power. It was not only an issue of large
state against small state, however. The primary concern was to avoid
the deadlock that might arise from separate balloting: New Hampshire
voted against its small-state interest and for joint balloting for this
explicit reason (II 402) and was joined by three other small states. Label
A
as amended by joint ballot as C; then C > A
. Next, the small states

moved that in the joint ballot each state be entitled to one vote; this failed
with ve afrmative votes, four from small states, and six negative votes,
ve from large states (Roll Call #357, II 399). A motion to eliminate
selection of the executive by the national legislature in favor of election
by the people failed, with Pennsylvania and Delaware in the afrmative,
and the nine remaining states in the negative (NY was absent, Roll Call
#355, II 399). The problemwith direct election for the delegates is that it
would favor large states over small states and free states over slave states;
in contrast, a scheme of election by electors is amenable to allocating
electors among states in some manner that balances the interests among
the different states. Another motion to eliminate selection of the executive
by the national legislature in favor of election by electors chosen by the
people of the states failed with ve votes in the afrmative (CT, NJ, PA,
DE, VA) and six votes in the negative (NH, MA, MD, NC, SC, GA; Roll
Call #359, II 399). Note that New Jersey and Delaware are back in the
election by electors camp. A motion to postpone consideration of mode
of selection and term of ofce of the executive failed on unrecorded vote,
and a motion to send those issues to committee made up of one mem-
ber from each state also failed on a tie vote (Roll Call #360). Finally,
a motion was offered asking for support of election by electors as an
abstract question (Roll Call #361, II 404), that is, some undetermined
form of choice by electors as opposed to selection by national legislature.
This motion failed on a tie vote with four states in the afrmative (NJ,
PA, DE, VA), four states in the negative (NH, NC, SC, GA), two states
divided (CT, MD), and one absent (MA). We have already labeled na-
tional legislative election with joint ballot as C; now, label election by
electors in the abstract as B
. The Convention vote was tied and thus

B
C. Finally on August 24, the delegates decided to postpone further

deliberation on questions of the executive.
It is here that Riker detects his cycle. He labels national legislative
election without joint ballot as A, election by electors as B, and legislative
election with joint ballot as C. Reverting to my labeling, national legisla-
tive election (without joint ballot) beat election by electors in the states,
A>B
on July 24 and A
>B
on July 26. Election by electors in the states

tied with national legislative election (with joint ballot) on August 24,
B
C. Earlier on August 24 national legislative election (with joint

ballot) beat national legislative election (without joint ballot), C > A
. By
Rikers labeling we have A > B C > A, a cycle. According to Riker,
there was a separationist faction in the Convention, loosely led by
G. Morris (ever the opportunist and an exceptionally adroit parliament
man, Riker 1984, 12), which favored the separation of powers rather
than a more unied form of government. Since there was a cycle, he
argues, any result could have obtained in the abstract, and the outcome
was decided by the superior will to power (intense will to win, Riker
1984, 14) of the leaders of the separationist faction. Morris masterfully
exploited the cyclic deadlock in order to maneuver the question into a
Committee on Remaining Matters (discussed below) dominated by sep-
arationists, Riker claims. Had Rutledge not brought up the joint ballot
[#356, C>A
], this cycle would not have been revealed indeed it would

not have existed. Legislative election would probably have survived. . . the
signicance of the cycle that Morris revealed was that it gave him another
chance in the committee [on Remaining Matters] (Riker 1984, 13). In
the absence of a cycle, or in the absence of Morriss clever exploitation
of the cycle, the nal Constitution would have had the executive selected
by the national legislature, Riker claims.
By my more strict labeling of the alternatives we have B
C, C > A
,
and A
> B
, for a noncyclic or transitive ordering of B
C > A
> B
.
Is it reasonable to strictly distinguish alternatives? In order to generate
his cycle Riker distinguishes between two similar but not identical al-
ternatives, national legislative election with joint ballot (C) and national
legislative election without joint ballot (A); thus, it must be legitimate
to distinguish among similar but not identical alternatives. Riker (1984,
14) is aware that all the continuing alternatives were changed in gross
or subtle ways throughout the event. Is my interpretation reasonable in
substance? My proposed transitive ordering states that the convention
was indifferent between election by electors in the abstract (with details
of which agents select electors by what voting rules and how to pay for
distant electors left undened) and national legislative election with joint
ballot, that both those alternatives were preferred to national legislative
election without joint ballot, and that national legislative election without
joint ballot was preferred to the concrete scheme of election by electors
appointed by the state legislatures. This is reasonable; someone might be
indifferent between the prospect of a trip to an unnamed destination in
France and a trip to London, prefer both to a trip to Oslo, and prefer
Oslo to Rouen in France.
The proposal to support election by electors in the abstract failed on
a tie vote of four for, four against, two divided states, and one absent
(Roll Call #361, II 399). The four states against were New Hampshire,
North Carolina, South Carolina, and Georgia; and none of those four
had ever supported anything other than selection of the executive by the
national legislature (Roll Calls #12, #182, #215, #225, #359, #360,
#361). The two divided states were Connecticut and Maryland, and this
is signicant. Up until Roll Call #361, Connecticut had usually sup-
ported election by electors, whether appointed by state legislatures or
elected by the people (Roll Calls #182, #215, #359 by electors; but
#225 by national legislature). Maryland had always supported election
by electors appointed by state legislatures but never election by electors
elected by the people (Roll Calls #12, #182, #215, #225, #359). Con-
necticut and Maryland were also small states and earlier in the day had
lost on a ve to six vote the motion that each state would have a single
vote if selection of the executive was by joint ballot in the national legis-
lature. Three of the four supporters of selection by national legislature,
New Hampshire, North Carolina, and South Carolina, had opposed the
small-states motion. There was nothing that the supporters of selection
by national legislature could offer Connecticut and Maryland, but the
two states were obviously ripe for some kind of compromise with sup-
porters of selection by electors. Michaelsens (1987, 65) monograph on
the creation of the presidency states that for the proponents of election
by electors, The primary consideration was to enlist the backing of the
smaller states which favored selection by the legislature as fairer to them
than a popular election. Massachusetts, a large state, had in the past
voted thrice for election by national legislature (#11, #167, #215), was
absent for one vote by national legislature (#225), voted against electors
(#166, #359), was divided on electors (#182), and was absent on electors
in the abstract (#361). This absence or abstention is peculiar, and sug-
gests political motivation, since just minutes earlier Massachusetts voted
against electors selected by the people (#359), and voted on all 11 of the
remaining votes that day.
On Friday, August 24 the Convention postponed questions on the ex-
ecutive to the following day, but the pace of closing deliberations on other
matters kept it off the agenda. On Friday, August 31 the hurried delegates
consigned all postponed issues (including questions on the executive) to
a Committee on Remaining Matters (of eleven members, one from each
participating state; II 481, not II 463 as incorrectly cited by Riker 1984,
13). The diligent Committee reported out some minor items on Saturday,
but spent the rest of the weekend working out compromises. On Tuesday,
September 4, the Committee reported their proposal on executive ques-
tions (II 493504). As for selection of the executive they recommended
that each state appoint in such manner as its legislature may direct a num-
ber of electors equal to the number of senators and representatives the
state is entitled to, and that electors would meet in each state and forward
their votes to the president of the Senate for counting, the person with the
majority of votes would be declared the winner, and there were further
provisions, among them mechanisms for dealing with plurality outcomes
and tie votes. This settled the vexing question of whether electors should
be selected by the people, by the state legislature, or by the state executive,
through leaving it to each state to decide the issue in its own way. The
exasperating controversy over allocating electors among small states and
large states was settled by adaptation of the scheme already devised in the
Great Compromise, cutting the issue down the middle; and since repre-
sentation in the House was apportioned according to the Compromise
counting slaves as three-fths of a person the conict between free and
slave states was reconciled.
2
The objection that had rst scuttled election
by electors on July 24, the expense of sending electors to the capital, was
settled by having the electors meet in their respective states. The device of
having the electors meet in their respective states simultaneously resolved
the problems of executive independence, avoidance of cabal and intrigue
in the electoral college, and the high cost of travel, opening the door to
a solution (Anderson 1993, 137). G. Morris details the rationale of the
committee at II 500.
Label the Committee on Remaining Matters September 4 alternative
of an electoral college as B
. On September 5, Rutledge (SC) moved

that the Committees report be postponed so that the convention could
adopt the earlier plan of selection of the executive by joint ballot of the
national legislature, already labeled as C (Roll Call #445, II 511). The
motion failed, supported only by North Carolina and South Carolina,
thus B
> C. To continue the former analogy, if the delegates were in-

different between a trip to an unnamed destination in France and a trip
to London, they far preferred a trip to Paris over either. Notice that
Connecticut and Maryland, each divided on August 24, had come back
to the cause of election by electors, and Georgia for the rst time declined
to vote against election by electors. A quite involved debate followed on
issues pertaining to the executive that are not the focus of our study.
The founders believed that the electoral college would nominate presi-
dential candidates and that the House of Representatives, each state with
one vote (the committee proposal was the Senate, but that was amended
on the oor in order to reduce the power of the Senate), would select
from the nominees. They believed that the large states would have an ad-
vantage in nomination and the small states an advantage in selection. The
small states believed (as did their opponents) that they had obtained such
an advantage by this arrangement that they made concessions elsewhere
in the negotiations (Jillson 1979, 395). This supports my hypothesis that
Connecticut and Maryland, each divided on election of electors in the
abstract on August 24, were available for a compromise on selection of
executive that would advance small state interests. The Committees re-
port was completed on the morning of September 5. The remainder of
September 5 and all of September 6 was devoted to debating and amend-
ing the Committees proposal on the executive, which was perfected by
the close of business on the 6th. The full report of the Committee on
Remaining Matters on all postponed issues, as amended in deliberations,
was adopted on Saturday, September 8, and was sent to a Committee on
Style. Except for a few details, the Constitution was nished.
The small states and those who favored federal over national govern-
ment exacted maximum concessions from large states and nationalists
in the construction of the executive. Large states wanted an executive
elected on the basis of population. What they got was an electoral col-
lege half-way between representation by population and representation
by state; but nal decision would be by the House with each state as-
signed one vote (delegates believed that the electoral college would not
produce majorities and that the election would always go to the House).
The nationalists wanted an executive independent from the national leg-
islature. What they got was nomination by an electoral college indepen-
dent from the national legislature, but, selection (they believed) by one
chamber of the national legislature. I learned about small-state inuence
from McDonald (1985, 252). The small states had agreed to the Great
Compromise on representation, including that money measures would
originate in the House where large states would be favored. On August 8,
the small states (instigated by NJ, DE, MD) reneged on the Compro-
mise, moved to strike the requirement that money measures originate in
the House, and they won by a vote of 7 to 4, because not all large state
delegates considered the money requirement expedient (Roll Call #254,
II 215). Some large-state delegates were quite passionate about the re-
quirement, however, and brought it up again on August 9, August 13,
August 15, and August 21. The question was sent to the Committee on
Remaining Matters. That Committee completed its report on September
5, and recommended a provision that money measures originate in the
House but be amendable in the Senate (a similar measure had died by a
vote of 4 to 7 on August 13, Roll Call #289, II 267). The rst recorded
vote on September 5, just after the Committee had completed its report,
was on Morriss motion to postpone consideration of the provision that
money measures originate in the House, and the postponement passed
9 to 2 (Roll Call #445, II 509510). Morris said that there had been
a deal, and that he did not want to vote for the money provision until
he saw that the bargain had been consummated. Madison grumpily ex-
plained in a footnote to his ongoing record of the Convention (II 514)
that Mason, Gerry, and some other large state delegates supported the
electoral college with its tilt to the small states in exchange for enough
small states support for the provision that money measures originate in
the House. Sherman of Connecticut, advocate of small-state interests,
emphasized in convention debate that the Committees proposal was a
compromise between the large states and the small states, the large ef-
fectively nominating and the small effectively electing the executive (II
512513). Indeed, the successes of the small states led three of the more
ideological nationalists, Randolph, Mason, and Gerry, to decline signing
on to the Constitution, and Madison was not happy with the outcome
(McDonald 1985, 252).
Jillson (1979) is another analysis of convention votes on the selection
of the executive. Jillson collects all votes of any kind on the executive
department into two batches; one from the beginning of the convention
to before September 4, the other from September 4 forward. He then
does factor analysis to support his claim that there were stable voting
groups on the executive question. For the early and middle parts of the
convention, there was a more cohesive peripheral coalition of ve states,
in the far north (NH, MA) and in the far south (NC, SC, GA), in favor of
election by national legislature, and a coalition of six states in the center
(CT, NJ, PA, DE, MD, VA) in favor of election by electors. The central
coalition was less cohesive than the peripheral coalition and was divided
up into a large state faction (PA, VA, weakly MA), a small-state fac-
tion (NJ, DE, MD), and Connecticut voting independently of the other
small states. Jillsons analysis suggests that distance to travel was a pri-
mary motivation behind opposition to election by electors. Further, recall
that it was New Jersey and Delaware who rst supported and then
opposed election by electors, and that I speculated that they did so out of
a small-state concern; in Jillsons data they show up as two of the three
members of the small-state faction on the executive question. For the
late part of the convention, Jillson claims that a coalition of mostly small
(four) and mostly northern (ve) states were at the heart of the successful
compromise (NH, MA, CT, NJ, DE, GA). That coalition was opposed
by three factions, the large states (PA, VA), two southern states (MD,
SC), and North Carolina independent of the other southern states.
Jillson provides useful context on the dynamics of the deliberations as
well. Michaelsen (1987) provides background on prior state and confed-
eration experiences with the executive, and this bit of refreshing historical
detail is at least as illuminating as Rikers formalisms.
Rikers cycle claim confronts an anomaly. Why did the convention,
supposedly in cyclic deadlock over the issue, appoint a Committee on Re-
maining Matters so strongly disposed to election by electors, 7 to 10 out
of the 11 by Rikers count? In a footnote, Riker (1984, 13) speculates that
the convention voted individually rather than by states in the appointment
of committees so that the large delegation from Pennsylvania favoring
election by electors determined the convention outcome. Why would a
convention full of skilled and experienced politicians commit such a blun-
der, however? Further, the Convention had earlier voted for a committee
of one member from each state, which forged the Great Compromise on
representation, whereby the small states gained equal representation in
the Senate against the interests of large-state Pennsylvania. Pennsylvania
opposed establishment of that representation committee, which was sup-
ported by the ten other states (I 516). If Pennsylvania could stack such
committees, why did it vote against establishment of the representation
committee, and why didnt it stack the committee and thereby defeat the
small states on representation? The Brearley Committee on Remaining
Matters was also made up of one member from each state (II 481). How
can one claimthat the Brearley Committee is stacked if its proposal on se-
lection of the executive survives a challenge on the oor by a vote of 9 to 2?
A simpler hypothesis is that the composition of the Brearley Committee
[on Remaining Matters] suggests that the decision to abandon legislative
election had already been made (Anderson 1993, 136). This simpler
hypothesis coheres with the observation that Connecticut and Maryland,
each divided on election by electors in the abstract, but having favored
more concrete proposals for election by electors in the past, voted for the
Committee on Remaining Matters compromise proposal containing a
particular scheme of election by electors that satised small-state inter-
ests. Rikers cycling hypothesis does not explain the voting behavior of
Connecticut and Maryland.
Riker also holds that it was the political genius of G. Morris and his
separationist camp that defeated selection of the executive by the national
legislature and enacted the method of selection ultimately adopted by
the Convention. The claim is doubtful. One authority on the Convention
writes, Inall these debates over the executive, while there was the greatest
diversity of opinion, lines of division do not seem to have been clearly
drawn. Members expressed simply their individual and personal points
of view (Farrand 1913, 118). Riker believes that G. Morris, McClurg,
and Houston collaborated on July 17 (II 3235) to stigmatize selection by
national executive by moving and arguing that such an executive should
serve during good behavior, that is, for life, in order to make himproperly
independent of legislative inuence. One problem with Rikers story is
that it appears that G. Morris seemed to genuinely believe that lifetime
tenure is best. In debate on the motion on July 17, Morris, expressed
great pleasure in hearing it. This was the way to get good Government . . .
He was indifferent howthe Executive should be chosen, provided he held
his place by this tenure (II 33). Is that stigmatization? G. Morris also
advocated that senators be appointed for life, and by the executive, so
that the Senate too would be independent from the popularly elected
House and from the people, so as protect wealth and property from the
masses (I 512513). Morris advocated that judges be appointed by the
executive, with the advice and consent of the Senate (II 44). He also
called for a strong and energetic executive; if elected by the people rather
than by the national legislature then for two-year terms but with perpet-
ual re-eligibility, and he believed that long tenure was likely; he was also
initially against a provision for impeachment of the executive (II 5255).
Delegate Luther Martin, reporting to the Maryland Legislature after the
convention, complained that there was a party at the convention that
wanted lifetime tenure for the executive, that given the powers of the of-
ce re-eligibility amounts to lifetime tenure, and that the life-tenure party
succeeded in the last days of the convention in winning re-eligibility for
election; Martin seems to be speaking about G. Morris (III 216). Another
problem is that it was Houston, on July 24 (II 99101), who moved to
scrap election by electors, because of difculty of attendance by elec-
tors from distant states, and return to election by national legislature.
That motion succeeded, thus Houston acted to defeat the position Riker
claims he conspired to advance. Madison stated explicitly in his journal
(II 3334), in order to protect his friend and Virginia ally McClurgs rep-
utation against charges of supporting monarchy, that McClurgs support
for life-tenure was a strategic exercise, consistent with Rikers claim, that
is, assuming that Madisons journal entry was itself sincere, rather than
a politically motivated reinterpretation of the facts.
Riker believes that the Brearley Committee and its proposed compro-
mise on the executive was the creature of G. Morris. The motion to form
the Brearley Committee was made by Sherman of Connecticut (II 481),
however, an advocate of the small-state interest, who was also elected
to serve on that Committee (recall my hypothesis that Connecticut and
Maryland were holding out on the vote to support election by electors in
the abstract). We know nothing for certain about how this compromise
was made, but I infer that G. Morris put it together (1984, 14), says
Riker. He cites indirect and speculative evidence for this view. There is,
however, direct documentary evidence that Brearley Committee member
Pierce Butler of South Carolina claimed to have devised the compromise
on selection (III 302). Finally, if there were a separationist camp led by
Morris at the Convention, it would seemnot that they cleverly succeeded,
as Riker would have it, but that they failed, in that many delegates believed
that the mode adopted would have one chamber of the legislature select-
ing the President. Perhaps Morris was clever in correctly predicting that
the adopted electoral college scheme would pick majority winners and
thus avoid selection by the legislature, but many other delegates believed,
wrongly as it turned out, that nal selection would be by the legislature
(Mason, II 500; Pinckney, II 501; Rutledge, II 511; Sherman, II 51213;
Randolph, II 513; Wilson II 522; Williamson, II 524; Hamilton, II 525).
By carefully distinguishing alternatives we can identify a noncyclic or-
dering throughout the event. On Rikers account the convention voted
for both B > A and A > B; on my account they voted for B > A > B
.
On Rikers account there is a cycle among A, B, and C; on mine
there is not. In total, my account identies the following decisions:
B > A, A > B
, A
> B
, C > A
, B
C, and B
> C. I assume
that A
A. There is no direct vote between B
and B
, but B
> C
by eight votes to two although B
C by four votes to four, thus I

think its fair to say that B
> B
. That yields B
> B
C > A
, and
B > A
A > B
; from that data I think it is plausible to assume

that C > B; and nally, I propose the total transitive ordering of
B
> B
C > B > A
A > B
. Alternative B is rst, tied for sec-

ond, fourth, and seventh in the ranking. What Riker labels as C is just
another version of election by national legislature (but with joint ballot);
we could label it A
, and if we did then alternative A would be tied for

second and tied with itself for fourth in the ranking. If we neglect to dis-
tinguish among different versions of similar proposals, then we will nd
that the convention ranked B > B A > B > A A > B, etc., and thus
that politics is pervaded by gross irrationality. If we distinguish among
alternatives as carefully as did those who proposed and voted on them,
then we nd order and reason.
Jon Elster suggests I am caught in a contradiction, but I believe he is
mistaken. We have already labeled selection of the executive by electors
appointed by state legislatures as B
, selection by electors in the abstract

as B
, and selection of executive by joint ballot of the national legislature

as C. Further, label selection of the executive by electors appointed by
the people as D, and label other possible schemes of selecting electors as
E, F, G, . . . , Roll Call #361 was a tie between election of the executive
by the national legislature, and election of the executive by electors in the
abstract: C B
. Elster claims that B
is equivalent to (B
or D), and
thus that #361 shows that C (B
or D). My suggested ordering of

B
C>A
>B
implies C>B
. Roll Call #359 showedthat C>D. Next,

according to Elster, C > B
and C > D are equivalent to C > (B
or D),
contradicting the equivalence he proposes from Roll Call #361. Such a
contradiction would be fatal to my hypothesis that there was a transitive
ordering of B
C > A
> B
.
The rst problem is that C B
is not equivalent to C (B
or D).
My claim is not one of pure logic. When G. Morris moved electors in the
abstract against the status quo (B
versus C), he and the delegates were

aware that the convention had already expressed C > B
(C > A
, Roll
Call # 356; A
> B
, Roll Call #215 and others), and C > D (Roll Call

#359), so he and they could not have intended B
to be equivalent to
(B
or D), we presume that he was not wasting his time by introducing

alternatives that had been clearly defeated. What he must have intended
was that B
contain the most satisfactory compromise over (1) method

of selection of electors and (2) method of allocating number of electors
per state, and an already rejected requirement that electors must be cho-
sen by state legislatures (B
) could not have been most satisfactory, nor

an already defeated requirement that electors must be selected by the
people (D). B
does contain all remaining methods of selecting electors,

E, F, G, . . . , In the end, the concrete proposal of the electoral college,
B
, included the provision that each state is free to decide how to choose
electors: by the legislature, by the people, or by some other method.
Logically, those who voted for B
on this basis also should have voted

for B
, but psychologically the concrete proposal was more attractive

than the abstract proposal that contained it as a possibility. Incidentally,
the method of state option is not equivalent to (B
or D), because B
re-
quires that electors must be exclusively chosen by state legislatures, D
that electors must be exclusively chosen by the people, and state option
permits any method to be chosen. The second problemis that C>B
and
C > D are not equivalent to (C > (B
or D)). (C > B
and C > D) do im-

ply that (C > (B
or D)), but not vice versa, because (C > B
and D> C)
also imply that (C >(B
or D)). Would it do to say that C >B
and C >D
imply that (C > (B
and D))? No, that would not be right either, because

although we know that C is preferred to B
and C is preferred to D, we do
not know anything about whether C is preferred to B
and D together (B
and D do seem to be mutually exclusive, however). It is better to say that

((C > B
) and (C > D)). Returning to the rst problem, my construal

is that B
did not contain B
and did not contain D, and thus there can

be no inference that (C B
) or that (C D). There is no contradiction

between (C > B
) and (C > D) in the rst problem, and the absence

of an indifference relationship between C, on the one hand, and B
or
D, on the other hand, in the second problem. My suggested ordering,
B
C > A
> B
, therefore stands.
In Rikers world it was not the arguments of Morris and the Commit-
tee on Remaining Matters (II 500) as to the reconciliation of inter-
ests most likely to promote the public good that persuaded the con-
vention to adopt by an overwhelming majority the method for selecting
the executive. Rather, it was because of the separationists rhetorical
and heresthetical skill and persistence, because of the cycle generated by
Rutledges unwise motion, and because of the clever appeal to diverse in-
terests put together in the proposal for an electoral college (Riker 1984,
14). In Rikers world, politicians act cynically rather than realistically,
slyly rather than capably, cleverly rather than wisely; politicians make
maneuvers rather than proposals, and they manipulate rather than
arrange. Sherman accepted the entire rhetorical stance of the separa-
tionists (9), rather than was persuaded of the correctness of their views.
This style of interpretation purports to be mere positive description and
thus value-neutral, but it plainly contains a contestably misanthropic view
of human motivation and interaction that drives the results. What does
Riker mean when he calls Morris, the architect of election by electors,
an opportunist? If Morris is an opportunist, does that mean that he
places his own private good above everyone elses, or merely the good of
his delegation above that of the conventions, and, if either, why would
respectively the delegation or the convention entertain his counsel or re-
spect his judgment? Would he cheat, steal, and murder if he could get
away with it? Why not just call him effective? Is everyone an opportunist,
or is it only politicians? Among politicians, is Roosevelt an opportunist
of a same or of a different kind than Stalin? Are Hubert Humphrey and
Joseph McCarthy both opportunists?
In order to preserve the useful meanings of contrasts, such as those
between opportunism and principle, rhetoric and argument, deception
and sincerity, both poles of the contrast must apply in the world. It could
not be that every actor is at all times cynical, sly, clever, maneuvering,
manipulative, rhetorical, and deceptive; otherwise the labels are drained
of content. At least one important exception must be the scholar who
claims universal opportunism, for otherwise we should not believe him.
My view is that since individuals vary in honesty, trustworthiness, and
public spirit in the face of identical material incentives, individual char-
acter is at least as important as political institutions. An alternative view
is that individuals are equally corrupt in the raw but variably ethical
depending on the institutional constraints they face or depending on the
sphere of activity in which they interact. Proponents of this alternative
view might argue, for example, that we can rely on the claims of those
operating within the institutions of scholarship, but not on the claims
of those operating within the institutions of politics. Is it that somehow
politics is uniquely repugnant among human activities? There is a fertile
abundance of incident upon which the posture of world-weary pessimism
concerning the ethics of democratic politicians can draw for inspiration.
That these are so widely known to the public is simply because politics
is the publics business, however. I think it is an illusion that humans
are more venal in politics than they are in scholarship, family, enter-
prise, or religion. Indeed, it has been my experience that even though
politicians are as ethically variable as nonpoliticians, on average they are
more ethical than the people they represent (which must almost neces-
sarily be so, since an important criterion of electoral selection is faithful
representation); corrupt politicians correlate with corrupt populations.
Many of those who are so sure all politicians are dishonest are just imag-
ining what their own performance would be if they were lucky enough
to gain ofce. The pessimistic view of politics is not only debatable as
description, it is associated with the party of tradition inside the arena of
politics, while the optimistic view is associated with the party of change.
In the American context, hostility towards politics and politicians is a
leading theme in the ideology of the right, with antecedents in the ten-
sions over desegregation, the defeat of the South in the Civil War, and
earlier still in the content of southern subcultures identied by Fischer
(1989), shaped by place and time of emigration fromGreat Britain. There
is not some extrapolitical science apart from ongoing political debate
that settles the contest in favor of one partisan attitude or the other.
There was plenty of hard politics at the convention, but alternatives
were not cannily inventedby clever conspirators as inRikers account. Go-
ing into the convention, eight states elected their governors in the legisla-
tures (PA, NJ, VA, MD, DE, NC, SC, GA) and ve states elected the gov-
ernor by popular election (NH, MA, RI, CT, NY; Anderson 1993, 134);
Massachusetts in the convention voted against popular election of an ex-
ecutive, and this was because its delegate Gerry had turned against that
method because the people had not reelected his ally Governor Bowdoin
after the governor had moved against the Shaysite rebels (Michaelsen
1987, 70). The thesis of Michaelsens (1987, 79) monograph is that the
institution of the presidency came about as a consequence of American
experience at the State and Confederation levels, and through the work
of hard-headed, practical, experienced politicians. The delegates be-
gan from these experiences, and used their political skills to harmonize
these contrasting experiences with the practical requirement of devising
a weighting of state interests sufcient to win ratication of the proposed
Constitution. Contrary to Riker, Rutledge was neither obtuse nor provin-
cial on August 24 when he moved (Roll Call #356) that election of the
executive in the national legislature be by joint ballot. With the exception
of tiny Georgia, the seven remaining states that elected their governors in
the legislature did so by joint ballot (even unicameral PA, which balloted
jointly with its executive council, Anderson 1993, 134); and selection by
joint ballot was in the early Pinckney draft of a constitution (Michaelsen
1987, 60). Separate ballot would give rise to dangerous deadlocks and
intrigues among the two chambers and presidential hopefuls. Yes, joint
ballot for President happened to favor the large states, and they voted
for it, but two small states (NH, DE) joined them to make for the ma-
jority against separate ballot. Joint ballot was the best policy regardless
of state interest, and the convention voted for the best policy. The next
motion, by the small states to insert each State having one vote, failed
on a ve to six vote, and several following motions failed on close votes.
Rather than forcing close votes, the delegates shortly appointed a com-
mittee to work out a compromise designed to enjoy broad support. The
delegates brought with them not only their experiences in state and con-
federation governments, among the leading members some had drafted
state constitutions, some had made a systematic comparative study of
state constitutional conventions and documents, and some had studied
the political theorists on choice of institutions. An independent execu-
tive and an electoral college seem to be an almost miraculously ingenious
compromise among interests apparently impossible to reconcile. Yet the
same ideas can be found nine years earlier. In 1778 a proposal for a
Massachusetts state constitution failed; and twelve of the towns that voted
against it issued a document called the Essex Result, which called ex-
plicitly for checks and balances among three branches of government,
and for a single executive elected annually at county conventions by elec-
tors previously chosen by the people (Michaelsen 1987, 17). Only by
ignoring historical context can we imagine that Rikers herestheticians
invent alternatives from thin air that bamboozle their opponents.
I do not want to make too much of my claimthat aggregate preferences
were consistent over time on the question of selection of the executive.
Many alternative modes of selection were discussed, but none won wide
support, and delegates came to nd the subject confusing and tedious.
Gerry said, We seem to be entirely at a loss on this head, and proposed
to send the issue to committee, Perhaps they will be able to hit on some-
thing that may unite the various opinions which have been thrown out
(II 103). Madison added that There are objections agst. every mode
that has been, or perhaps can be proposed (II 109). The delegates were
not merely aggregating preferences, they were mutually persuading, indi-
vidually and collectively deliberating, forming and changing preferences
apart and together in the course of the convention. If I had made only a
preference-change argument, however, the reader would have suspected
that I had nothing against the correctness of Rikers cycling claim on its
own terms. Preferences were formed in the course of the Convention; that
was one of its purposes. Early in the deliberation, on July 20, with re-
spect to impeachment of the executive, Mr. Govr. Morriss opinion had
been changed by arguments used in the discussion (II 68). Late in the
deliberation, on September 4, with respect to appointment of the exec-
utive, Wilson said that the subject had greatly divided the convention,
and that He had never made up an opinion on it entirely to his own
satisfaction (II 501). Or, as Franklin commenced his closing speech at
the Convention (II 641642):
I confess that there are several parts of this constitution which I do not at present
approve, but I am not sure I shall never approve them: For having lived long,
I have experienced many instances of being obliged by better information or
fuller consideration, to change opinions even on important subjects, which I
once thought right, but found to be otherwise. It is therefore that the older I
grow, the more apt I am to doubt my own judgment, and to pay more respect to
the judgment of others.
In such an atmosphere one should not expect individual or collective
preferences to be consistent between, say, July 24 and August 24 as re-
quired for Rikers mistaken cycle. To a lesser extent, preferences might
be inconsistent at a xed time as well, simply because individuals and
the collective body have not completed the task of consistently ordering
desires and beliefs. The task of ordering is a major part of the content
of the deliberation, and in this case was not complete until the consti-
tutional draft was completed. If we criticized a scholar on the grounds
that consecutive drafts of an article were inconsistent with one another
and with the nal document, the argument would have no bite. It is to
the nal version that we should look for internal consistency, and given
the inevitable conicts among goods, only to whether it is more coher-
ent than feasible alternatives. Thus I consent, Sir, to this Constitution,
Franklin continued (II 643), because I expect no better, and because I
am not sure, that it is not the best. The opinions I have had of its errors,
I sacrice to the public good.
Agricultural appropriations, 1958
Rikers rst published attempt to demonstrate a cycle and thereby the
relevance to politics of the Arrow theorem was in 1958, a study of a se-
quence of votes on agricultural appropriations on the oor of the House
of Representatives in 1952. Riker does not repeat this story in Liberalism
against Populism (1982) nor in The Art of Political Manipulation (1986).
Although Riker dropped the agricultural appropriations story, his follow-
ers continue to repeat it, for example Strom (1990, 2829) introduces
cycling and its empirical relevance with a version of the story.
A proposal from the Subcommittee on Agriculture of the Commit-
tee on Appropriations for the budget of the Soil Conservation Service
came before the Committee of the Whole in the US House of Repre-
sentatives. The subcommittees original proposal was for $250 million.
From the record, the Korean War was on and there were calls for econ-
omy. The minority Republicans, as usual, were for economy, some for
a smaller reduction in the appropriation and some for a larger reduc-
tion. The American Farm Bureau Federation, its members the supposed
beneciaries, opposed the large appropriation as wasteful and socialis-
tic. Urban Democrats seemed to act as if they felt that rural Democrats
were getting more than their due in appropriations. Jacob Javits (R-NY),
not on the subcommittee, moved an amendment that as later corrected
proposed reducing the appropriation to about $142 million (142 against
250). Next, H. Carl Andersen (R-MN), ranking minority member on the
subcommittee, moved a substitute for the Javits amendment, of $200 mil-
lion (200 against 142). Then Whitten (D-MS), chair of the subcommittee
and thus advocate of the original proposal, moved to amend Andersens
substitute to $225 million (225 against 200). Finally, OToole (D-NY)
moved to amend Javitss amendment to $100 million (100 against 142).
This is the maximum number of alternatives permitted by House rules.
According to House procedures, rst the OToole amendment to
Javitss amendment (100 against 142) would be voted on, then the
Whitten amendment to the Andersen substitute (225 against 200). Then
the perfected amendment goes against the perfected substitute. The
OToole amendment to the Javits amendment was defeated by voice vote,
thus Javits (142) > OToole (100). The Whitten amendment to the
Andersen substitute lost on a division of 74 to 139, thus Andersen (200) >
Whitten (225). Next, Javits, the perfected amendment, went against
Andersen, the perfected substitute. Riker completely omits this fact: on
rst count, Andersen beat Javits, on a division of 132 yeas and 131 nays,
thus Andersen (200) > Javits (142). On second count, Andersen lost
to Javits, on a teller vote of 126 yeas and 131 nays, thus Javits (142) >
Andersen (200). That left the vote between the Javits amendment and the
original subcommittee proposal, and the Javits amendment was defeated,
35 yeas and 220 nays, thus Original (250) > Javits (142).
The votes we have are apparently consistent. Javits > OToole,
Andersen > Whitten, Javits > Andersen, Original > Javits, reduces to
Original > Javits > Andersen > Whitten > OToole. In million dollar
terms, the collective preference was apparently the peculiar ranking
250 > 142 > 200 > 225 > 100, which is intransitive in dollar amount.
According to Riker (1958, 358):
From the fact that all amendments failed one might infer that a majority fa-
vored the original proposal. Nevertheless, one awkward fact casts doubt on this
inference: although the largest amount stayed in the bill the third largest amount
(Andersen) beat the second largest amount (Whitten). From this fact one may
reasonably suspect an intransitivity here, for if the largest amount were really fa-
vored over all others, and the amount was the dominant criterion, then logically
the second largest sum should have defeated the third largest.
Riker is not as surprised that $142 million, the fourth largest amount,
beat the third largest amount and the second largest amount as well. He
explains that On the crucial vote, although the Andersen substitute was
formally pitted against the Javits amendment, the members clearly as-
sumed that the substitute lay against the original (Riker 1958, 359). In
other words, when voting on Andersen (200) against Javits (142), mem-
bers who favored the Original (250) believed that if Javits (142) won,
then on the next scheduled vote Javits (142) would lose to the Original
(250). Hence, voters who favored the Original (250) voted strategically
for Javits (142) over Andersen (200); but the sincere collective preference
was Andersen (200) >Javits (142). Recognizing this instance of strategic
voting revises the apparent collective preference to 250 > 200 > 225 >
142 > 100; the fourth largest amount is now in order, but still the third
largest amount beats the second largest amount. We are entitled to ask,
though, given the nal outcome in favor of the Original (250), whether
some of those who voted for Andersen (200) against Whitten (225) were
also strategic voters. Perhaps voters who most favored the Original (250)
would sincerely prefer Whitten (225) to Andersen (200), but voted strate-
gically against Whitten (225) on the expectation that the sequence of
votes would lead to victory for the Original (250), as it did. If there were
such strategic voters on Andersen (200) against Whitten (225), then the
Table 14.1. Rikers estimates, Agricultural
Appropriations, 1958
Number Faction Name Preference Order
108 OriginalWhitten 250 > 225 > 200 > 142 > 100
13 OriginalAndersen 250 > 200 > 225 > 142 > 100
30 Javits 142 > 100 > 200 > 225 > 250
91 Andersen 200 > 225 > 250 > 142 > 100
10 Whitten 225 > 250 > 200 > 142 > 100
5 OToole 100 > 142 > 200 > 225 > 250
sincere collective preference was Whitten (225) > Andersen (200). If we
grant that strategic voting succeeded on both votes, then we obtain the
sincere preference rankings of Javits > OToole, Whitten > Andersen,
Andersen > Javits, Original > Javits, which reduces to Original >
Whitten > Andersen > Javits > OToole. In million dollar terms this is
250 > 225 > 200 > 142 > 100, which is perfectly transitive if dol-
lar amount is the standard of measure, and thus Rikers alleged cycle
vanishes.
Why then does Riker believe there is a cycle? He estimates the num-
bers and preference orders of different factions by means of several plau-
sible inferences that I shall not repeat here. I report his results (1958,
361), in Table 14.1. Each faction is named after its most favored al-
ternative, and then Riker posits further rankings such that alternatives
closer to the most favored alternative are more favored, except for the
OriginalAndersen faction whose ranking is jumbled in dollar amount.
The OToole faction, for example, ranks the lowest appropriation rst, the
second-lowest second, andso on, andthe highest appropriationlast. Riker
divides the Original faction into two camps. The rst, OriginalWhitten,
or the Original faction proper, ranks the highest appropriation rst, the
second-highest Whitten second, and so on, and the lowest appropriation
last. Whence the second, OriginalAndersen camp, who perversely rank
the third-lowest amount Andersen above the second-lowest amount, the
very paradox we are attempting to explain? According to Riker, these
13 or more voters sincerely rank the alternatives the same way as do
the OriginalWhitten voters, 250 > 225 > 200 > 142 > 100, but voted
strategically for Andersen (200) and against Whitten (225) in the con-
test between the two. The OriginalAndersen voters followed a highly
contrived strategy in voting . . . Assuming that the original paragraph
[250] could defeat the Andersen substitute [200] but not the Whitten
amendment [225], they voted rst against the latter and then, safely, for
Table 14.2. Pairwise comparison matrix, Agricultural
Appropriations, 1958
Original Javits Andersen Whitten OToole
250 142 200 225 100 (BC

)
Original 222 131 136 222 (711)
Javits 35 35 35 252 (357)
Andersen 126 222 139 13 222 (696)
= 126
Whitten 121 222 118 + 13 222 (696)
= 131
OToole 35 5 35 35 (110)
Notes: The winner of a pairwise comparison is indicated by italic.
This is the Borda count for corrected estimates, that is, it excludes the strategic
and includes the sincere preferences of the 13 OriginalAndersen voters.
the former (Riker 1958, 359). In order to determine collective rank-
ings, however, we cannot mix sincere and strategic individual rankings.
It is by counting the OriginalAndersen voters strategic ranking rather
than their sincere ranking that Riker generates his illusory cycle. His es-
timates of the numbers in factions and their preference rankings yields
the rst-listed number in each cell of the pairwise-comparison matrix in
Table 14.2.
Using Rikers estimates that mistakenly include the strategic rankings of
the OriginalAndersen faction does indeed yield the collective preference
order of 250 > 200 > 225 > 142 > 100 that Riker suspects of intransi-
tivity because it is not transitive in dollar amount of appropriation. It is a
matter of taste whether or not we would call that a proper cycle, but there
is no need to debate the question since Rikers proposed collective prefer-
ence order is mistaken. The collective ranking is of Andersen (200) over
Whitten (225) according to Rikers estimates that include the strate-
gic ranking of the OriginalAndersen faction. If we counted instead
the OriginalAndersen factions sincere ranking, then we would sub-
tract 13 votes in the cell counting how many votes Andersen got
over Whitten and we would add 13 votes in the cell counting how
many votes Whitten got over Andersen. This reverses the outcome be-
tween the two alternatives, and now Whitten (225) has a majority over
Andersen (200). With that reversal we obtain the pairs 250 > 142,
250 > 200, 250 > 225, 250 > 100, 142 > 100, 200 > 142, 200 >
100, 225 > 142, 225 > 200, 225 > 100, each of which ranks the larger
amount over the smaller amount, and indeed the pairs reduce to the
collective ranking of 250 > 225 > 200 > 142 > 100, which is transitive
in dollar amount. Rikers nding of a collective ranking that is intransitive
in dollar amount is merely an artifact of improperly counting the strategic
ranking of the 13 OriginalAndersen voters rather than their sincere
ranking.
The Borda count for the corrected rankings, which by use of infor-
mation about relative rankings will give us an approximation of voters
intensity of preference, shows $250 million in rst place (711), followed
closely by a tie between $225 million (696) and $200 million (696); then
there is a huge gap to $142 million (357) and $100 million (110). There
is a possible instance of strategic abstention and logrolling that Riker does
not report. Recall that on rst count, Andersen (200) beat Javits (142)
by 132 to 131, a vote that Riker fails to mention in his article, and that
on second count, Andersen (200) lost to Javits (142) by 126 to 131. The
number of no votes is the same, 131 on each vote, but the number of
yea votes declined from 132 to 126. There may have been simply errors
between the two counts, or shifting attendance, or perhaps six yea voters
strategically abstained on the second count in exchange for consideration
from the Democratic and agricultural coalition on other appropriations
issues. If we redo the pairwise-comparison matrix and add the 6 voters
in the appropriate cells (for the reader who wishes to check, this involves
adding 6 voters to the 91 voters in the Andersen faction), then we obtain
a consistent but peculiar ranking: 200 > 250 > 225 > 142 > 100. The
third-place amount has jumped to rst. This is extremely weak, however,
200 > 250 by only one vote and 200 > 225 by only one vote, so not too
much should be made about it. From the Borda count we know that the
three top alternatives 250, 225, 200 are almost tied and that the pres-
ence or absence of a few members might change the outcome. Adding
the six voters, the Borda count becomes even closer among the three top
alternatives but $250 million is still in rst place: 250 (723), 225 (714),
200 (720), 142 (357), 100 (110). One hypothesis about logrolling is that
it permits an outcome that incorporates the intensity of preference in-
formation that pairwise voting throws away. By pairwise voting, the rst
count including the six voters has Andersen (200) as the winner, even
though the Original (250) would be the Borda winner; then, perhaps
there was logrolling, which had the result of selecting the Borda winner.
This is too weak to be a demonstration of the logrolling hypothesis, but
it is consistent with it.
Since Rikers initial premises are mistaken, there is no purpose in ex-
amining the entire remainder of his argument, but there are a few points
of interest. Even with the benet of his initial error, he still must make
several more or less arbitrary assumptions (Riker 1958, 362) to ob-
tain his cycle. On the one hand, it seems likely that the House took
collectively irrational action due to cycling to the tune of $25 million
(1958, 358). On the other hand, there was quite possibly no cycle:
yet, this detailed investigation was undertaken simply because there was reason
to suspect an irrationality. For the sake, therefore, of reconstructing a possible
intransitive ordering of preferences, we shall assume [the reading of preferences
that results in a cycle]. (Riker 1958, 360)
Thus, in the further development of his argument, although there was no
persuasive reading of legislators preferences, the reading that generated a
cycle was arbitrarily selected for demonstration purposes. Finally, if Riker
had succeeded in demonstrating a cycle in this case study, it would have
been an insignicant nding. If the alleged cycle were among alternatives
distant in Borda ranking, say among 250, 225, and 100, and 100 won,
then we would have had cause for concern. The cycle that Riker alleged
would have been among the three top alternatives (250, 225, 200), how-
ever, any one of which was preferred to the fourth (142) and fth (100)
alternatives, and all such a cycle would indicate, as we saw in the dis-
cussions of the Borda count, is indifference or near indifference among
neighboring alternatives.
Riker (1958, 356) claims that collectively irrational action due to cy-
cling is probably fairly frequent. This case study is an example chosen
almost at random (Riker 1958, 357). If cycles are so frequent that one
can nd them by grabbing the Congressional Record and picking a page
at random, then, one must ask, why publish this case based on extraor-
dinarily weak (not to mention mistaken) data? Why not instead choose
almost at random cases that contain plentiful roll-call votes, nd the
strongest ones of those, and then publish strong demonstrations based
on strong data? Could it be that the reason is that cycles are rare and
irrelevant to politics?
15 Other cycles debunked
Introduction
In this chapter we review all remaining published and developed cycle
claims that I could nd in the literature, as well as some undeveloped cy-
cle claims. Blydenburgh (1971), inuenced by Arrow and Riker, sought
to demonstrate a cycle in deliberations on the Revenue Act of 1932 in
the US House of Representatives. The vote was among a sales tax, an
income tax, and an excise tax. Blydenburghs rst argument is that a ma-
jority was against each alternative. A majority voted for the excise tax,
however; and Blydenburghs inference that a majority nevertheless had
preferences against the excise tax is in error arising from confusion about
which alternative is pitted against which. The second argument makes
two assumptions in order to obtain complete inferred preferences from
incomplete revealed preferences. The rst assumption is arbitrary and
weakly warranted, however, and further, Blydenburgh, without explana-
tion, inconsistently applies the second assumption; if the rst assumption
is dropped, or if the second assumption is consistently applied, then he
has no cycle. His third argument again errs due to confusion about which
alternative is pitted against which; and ultimately reduces to the failed sec-
ond argument. Blydenburghs erroneous analysis is frequently cited by
partisans of the irrationalist doctrine.
Bjurulf and Niemi (1978) explore Rikerian doctrine in the records of
the Scandinavian parliaments. They claim to nd three cycles. The rst
concerns the construction of a hospital. In order to show a cycle they
must go beyond expressed votes and infer some of the individual prefer-
ence rankings. They reject a more plausible inference about one groups
ranking (that makes for no cycle), on the basis that it has no support
in the record, only to advance a less plausible inference that also has no
support in the record (and makes for the cycle); if their less plausible in-
ference is incorrect for any 2 of the 37 members of the group in question,
then there is no cycle. Furthermore, about half the chamber was absent
for these votes, making inference of the full chambers rankings quite
335
speculative. The second concerns appropriations for Swedish telephone
and telegraph during the Depression. There is a cycle if we assume that
one of the factions was strategically irrational, but there is no evidence to
warrant that assumption. In the absence of conicting evidence, it is bet-
ter to construe actors as rational, and if we do then there is no cycle. The
third concerns appropriations for the rie club. The Social Democrats
did vote strategically so as to thwart the majority will; but the majority
found a creative device to restore the most favored alternative. Bjurulf and
Niemi (1978) generalize that manipulations are frequent in Scandinavian
politics, but as these are presumably their best cases, the generalization
is doubtful.
Neufeld, Hausman, and Rapoport (1994) nd a cycle over three alter-
natives concerning Muscle Shoals in the USSenate in 1925, a project later
realized by Franklin Roosevelt as the Tennessee Valley Authority. Their
demonstration is unique, they say, because they rely solely on recorded
votes, not on inferences. One inference that they, and we, are able to
draw, however, is that a key group of actors was voting strategically. Once
we recover the sincere preferences of those strategic voters and adjust
chamber totals accordingly, we see that sincere preferences are in equi-
librium, and that the Condorcet winner prevailed in the end. Shepsle
and Bonchek (1997) conclude that there is no such thing as the pub-
lic interest on the basis of Rikers cycling examples and on two novel
examples, one based on phantom evidence, and one a cycle that specu-
latively could have occurred in a possible world but did not occur in our
actual world.
There are credible claims of occasional cycles, and of nonpopular
choices, in the Finnish electoral college (Lagerspetz 1993, 1997). I argue
that this is an exceptional case: the institution is one peculiarly suited to
antimajoritarian mischief, and the two strong cycle claims are froma time
when Finland was a highly polarized society, during the Great Depres-
sion and within memory of civil war and terror. It is possible that there
may have been cycles among the preferences of disciplined parties, but
not among the preferences of the population they represent. There is also
strong evidence of an apparent cycle among preferences of Iowa senators
over anticorporate farming legislation (Gross 1979), but I suspect that a
cycle is unlikely in sincere preferences. Kurrild-Klitgaard (2001a) nds
a eeting cycle in one of many public opinion polls concerning persons
who might ll the unelected Danish prime minister post in 1994. The
momentary cycle is due to the closeness of the election among the forces
who would select the prime minister; in the end, the most favored per-
son lled the post. I also review some what-if cycles, those based in part
on observed votes and in part on frankly hypothetical preferences. And,
Other cycles debunked 337
I dismiss coffee-break cycles: my coauthor in Barcelona knows someone
in Auckland whose student found a cycle in Nepal, or was it Manitoba?
I conclude that many of the erroneous cycle claims are due to confusion
about sincere and strategic voting.
1932 Revenue Act
Blydenburgh (1971) thinks he uncovers a cycle in the choice of type of tax
by the US House of Representatives in the Revenue Act of 1932. Riker
and Weingast (1988) cite the study and its conclusions with approval, as
one of three examples offered in support of their disequilibrium hypoth-
esis. Blydenburgh introduces the topics of Arrovian cycling and Rikerian
manipulation, and presents the paper as a test of the hypothesis that the
use of the closed rule on revenue bills in the House functioned to force
an arbitrary equilibrium on a topic otherwise destined to disequilibrium
because of redistributive heterogeneity. Blydenburgh searched revenue
debates from 1932 to 1954, and found two cases where there was both
an open rule and sufcient roll-call votes to test for the existence of a
Condorcet paradox. Generally, it is probably not true that a paradox of
voting would have occurred on all major revenue bills introduced under
the closed rule (Blydenburgh 1971, 71), and the second case of con-
sideration under an open rule in 1938 only came dangerously close to
a cycle on Blydenburghs analysis, which is to say, it was not a cycle.
That leaves the rst case where Blydenburghs position is that a cycle
probably occurred.
It was 1932, a bill was needed to raise a billion newdollars for Hoovers
antidepression programs, and public opinion then believed that it was
imperative to balance the budget. The Ways and Means Committee re-
ported a bill to the House that contained a new manufacturers sales tax.
First, on the House oor, a motion to delete the new sales tax passed
by a roll-call vote of 236 to 160 (Q > S ). Second, a motion to impose
an income tax failed by a roll-call vote of 178 to 211 (Q > I ). Third, a
motion to impose an excise tax passed by a roll-call vote of 204 to 187
(E > Q). On nal passage the entire bill passed by a vote of 327 to 64.
The three roll-call votes allow Blydenburgh to reconstruct a portion of
legislators expressed preferences.
The rst claim is that there is a negative majority against each alterna-
tive. Commenting on the second case, Blydenburgh (1971, 70) explains
that, A negative majority is a necessary but not sufcient condition of
a paradox of voting. That would mean, of course, that demonstrating
a negative majority would not demonstrate a cycle. I confess that I do
not understand what Blydenburgh is up to with his negative-majority
Table 15.1. Blydenburghs analysis, Revenue Act 1932
Group Number in Vote
Revealed Inferred Corrected

label group S, I, E preference preference inference
A 85 NYY (I, E) > S I > E > S I > E > S
B 77 NYN I > (S, E) I > E > S I > E > S
C 38 NNY E > (I, S) E > S > I E > S > I
D 30 NNN Against all Indifferent Indifferent
E 9 YYY For all S > I > E +Indifferent
F 7 YYN (I, S) > E S > I > E S > I > E
G 69 YNY (E, S) > I S > E > I S > E > I
H 71 YNN S > (E, I) S > (E, I) S > E > I
Note:
Those who abstained on all votes are not counted. Additionally, those who partially ab-
stained are as follows: 3 NNA, 1 NAN, 1 NAA, 2 NAY, 1 YAN, 2 YAA.
analysis, hence, to be safe, I shall quote his argument in full, and then try
to puzzle it out.
It can be seen from the total vote on the amendments that a majority of 236 op-
posed the sales tax, 211 opposed the income tax, but only 187 opposed the excise
tax. Evidence was found in sources outside the voting record that at least seven
congressmen in Group A can be identied with the [pro-income-tax] soak the
rich coalition (one might thus characterize the whole voting group). Apparently
income tax was the alternative these seven most preferred. Their complete pref-
erence ordering is I > E > S. Further investigation showed that ve congressmen
in Group G were members of the Ways and Means Committee and supporters of
the bill. The sales tax was apparently the most preferred alternative of these ve
congressmen, and, thus, their complete preference ordering is S > E > I. Adding
these twelve (seven from Group A and ve from Group G) to the 187 against E
on the third roll call produces a negative majority of 199. Therefore, there was
a negative majority against each alternative and a paradox of voting among the
three amendments. (6566)
Blydenburgh said in discussion of the second case that a negative majority
is necessary but not sufcient for a paradox of voting. So why in the rst
case does he say that a negative majority means a paradox of voting?
Next, in analyzing the legislative cases in this volume I have found it
essential constantly to be clear about which alternatives are being voted
on. There are never votes on lone alternatives, votes are always on pairs
of alternatives; when not explicit the vote is often against the status quo,
Q. The three roll-call votes under consideration were not direct contests
among S, I, and E. The rst vote was between the sales tax and the status
quo, and Q > S. The second vote was between the income tax and the
status quo, and Q > I. The third vote was between the excise tax and the
status quo and E > Q. That gives us collective expressed preferences of
Q > S, Q > I, and E > Q, which implies both E > Q > S and hence
E > S, and E > Q > I and hence E > I. With these data we do not know
the collective preference between the sales tax S and the income tax I,
but we do know that the excise tax is preferred to both, E > (I, S).
What happens when E goes against Q? We have the recorded vote, E
beats Q. What would happen if E went against I ? Group C and Group G
voted for E over I, for a total of 107 votes. Group B and Group F voted
for I over E, for a total of 84 votes. The members of Group D voted
against all taxes; thus, in a contest between E and I they should abstain.
The members of Group E voted for all taxes, so add nine votes to E and
nine votes to I. On present assumptions we do not know how Group
A and Group H ranked alternatives E and I, and together they make up
156 votes. That leaves 116 for E, 93 for I and 156 undetermined. In order
to turn Efromthe winner against I to the loser against I, we would need to
demonstrate that at least 90 of the 156 undetermined votes in Groups A
and G were I >E (156 =90 +66; 93 +90 =183 >182 =116 +66) yet
Blydenburgh claims only 7 fromGroup Afor I >E; 7 falls well short of 90.
What would happen if E went against S? Group A and Group C voted
for E over S, for a total of 123 votes. Group F and Group H voted for S
over E, for a total of 78 votes. The members of Group D voted against
all taxes; thus, in a contest between E and S they should abstain. The
members of Group E voted for all taxes, so add nine votes to E and nine
votes to S. We do not know how Group B and Group G ranked S and E,
and together they make up 146 votes. That leaves 132 for E, 87 for S, and
146 undetermined. In order to turn E from the winner against S to the
loser against S, we would need to demonstrate that at least 96 of the 146
undetermined votes in Group Band Group Gwere for S >E(146 =96 +
50; 87 + 96 = 183 > 182 = 132 + 50), yet Blydenburgh claims only 5
from Group G for S > E; 5 is well short of 96. To keep the exposition
simple, I neglected to include the ten voters who abstained on one or two
of the three votes in the calculations, but this does not affect the substance
of my argument.
Is it that the vote is supposed to be between alternative E and some
alternative not-E? Alternative E won by a vote of 204 to 187, so say that
there were 204 votes for E and 187 votes for not-E. Blydenburgh has
7 voters from Group A who rank I > E > S and thus prefer I to E, so
subtract 7 from the total of 204 for E and add 7 to the 187 who are
not-E. Blydenburgh has ve voters from Group G who rank S > (E I )
and thus prefer S to E, so subtract another ve from the total for E and
add ve to the total for not-E. Complete the adding and subtracting and
there are 192 total votes for E and 199 total votes for not-E, perhaps
this is Blydenburghs negative majority of 199. There is no alternative
not-Ecomposed of alternatives I and Sfor voters to choose, however. This
is pairwise voting. We are already sure that E beats Q. If there were such
a composite alternative not-E, then notice that the preferences of the
12 voters have been fallaciously construed. True, the seven voters in
Group A prefer I to E and thus not-E to E, but since they are I > E > S
they also rank E over S and thus E over not-E. Their seven votes for E >
not-E are necessarily cancelled by their seven votes for not-E > E. That
gets us back to 199 for E, 192 for not-E, and no cycle. Further, those in
Group Gprefer S to E and thus not-E to E, but since they are S >(E I )
they might rank Eover I andthus Eover not-E. If so, their ve votes for E>
not-E would necessarily be cancelled by their ve votes for not-E > E.
Some such reasoning as this accounts, I think, for Blydenburghs rabbit
out of a hat feat of turning a revealed majority for E into an inferred
majority against E.
Next, Blydenburgh introduces what he calls stronger assumptions in
order to complete the inference of preference orders and alternatively
demonstrate a cycle. The rst assumption, which we shall see has an
odd consequence, is that in the sequence of voting voters would vote
for their most-preferred alternative and if that failed then on the next
vote vote their next most-preferred, and so on. The second assumption
is that the excise tax is more like the sales tax than the income tax, so it is
unlikely that the income tax would come between the sales tax and the
excise tax in individual preference orderings (66). He claims that these
assumptions yield the inferred preference orders in the column with that
label in Table 15.1. Summing up pairwise contests given Blydenburghs
inferred individual preference orders, the sales tax S defeats the income
tax I by 194 to 162. The excise tax E beats the sales tax S by a vote of 200
to 156. We have E > S, and S > I, or E > S > I. The remaining question,
according to Blydenburgh, is about the contest between the income tax
I and the excise tax E.
He has 396 voters in total. The 30 voters of Group D voted against all
taxes and he calls them indifferent. Also, three of the voters abstained on
both the income tax and the excise tax. Thus a majority for a vote be-
tween the excise tax and the income tax would be 182 >(
39633
2
) =181.5.
Summing from his inferred individual preference orders, there are 180
votes for the income tax, two short of the requisite majority of 182, and
112 votes for the excise tax (including, from the partially abstaining
voters, two votes E > Q, two votes Q > E and three votes Q > I ).
Then, he says we are unable to order the 71 voters in Group H whose
revealed preferences are S > (E, I ). If only four of those in Group H
were indifferent between E and I then the income tax would have a ma-
jority with 180 votes (112 + 71 = 183 4 = 179 < 180), according
to Blydenburgh. If all those in Group H strictly ranked the two alterna-
tives, but merely 2 of the 71 preferred the income tax to the excise tax,
then again the income tax would win (180 + 2 = 182 > 181 = 112 +
69), according to Blydenburgh. If the income tax beats the excise tax,
I > E, then the individual preference conguration gives rise to the cycle
E > S > I > E, he concludes.
Now we shall examine his inference of preference orders. First, the
30 in Group D who voted against all taxes he terms indifferent, but I
would like to call them negatively indifferent. Next, I mentioned that the
rst assumption had an odd consequence. I would call the nine voters
in Group E who voted for all taxes positively indifferent, but Blydenburgh,
applying his rst assumption, determines that their preference order is the
same as the sequence of votes, S > I > E. Someone is asked in sequence
whether she would buy vanilla, chocolate, or strawberry ice cream, she
answers yes in each case, and the investigator concludes that she prefers
vanilla to chocolate to strawberry. Can Blydenburghs rst assumption
provide a correct inference? I think not.
Further, his second assumption is inconsistently applied, with no ex-
planation. The second assumption is that the income tax will not be
between the sales tax and the excise tax in individual preference rank-
ings. Notice immediately that this is contrary to his odd inference that
the positively indifferent voters in Group E ranked S >I >E: I should not
be between S and E. There are far bigger problems, however. The second
assumption transforms Group As revealed preference of (I, E) > S into
an inferred preference of I >E >S. The assumption is also applied to the
incomplete revealed preferences of Group B and Group C respectively
in order to obtain complete inferred preferences. The revealed prefer-
ence of Group H is S > (E, I ). Applying the second assumption would
yield the complete inferred preference order of S > E > I. Blydenburgh
says, however, with no explanation for the inconsistency, that we cant
completely order Group H, meaning that the second assumption applies
except to Group H. Why? It happens that if we applied the second as-
sumption to Group H there would be no cycle; recall, he needs a few in
Group H to be indifferent or for a few to be S > I > E in order to get his
cycle. He only needs a few votes though, the reader might say, but no. We
do not know how the 85 voters in Group A ranked E and I; the second
assumption infers that they ranked I over E, but some or many could
have ranked E over I. If we consistently apply the second assumption to
both Group A and Group H, then when Blydenburgh says that a few in
Group H could have preferred I to E the proper reply is, yes, but a few
Table 15.2. Pairwise-comparison matrix,
Revenue Act 1932
I E S Q
I 178 171 178
E 187 209 204
S 194 156 160
Q 211 187 236
in Group A could have preferred E to I; one arbitrary exception cancels
out the other.
If we correct the 9 voters of Group E from S > I > E to S I E,
then for Blydenburgh to make his cycle he would need not 4 but rather
13 voters in Group H to be indifferent between I and E, or he would
need not 2 but rather 11 voters in Group H to prefer I > E. Or, if we
apply the second assumption consistently to both Group Aand Group H,
then individual preferences do not aggregate into a cycle (by two votes,
to be sure). In both cases the collective ranking is E > S > I. If we make
both corrections (I have ignored the partially abstaining voters for ease
of exposition) then E > S, 209 votes to 156; S > I, 194 votes to 171; and
E > I, 187 votes to 178. Again, E > S > I and there is more robustly no
cycle.
Part of the confusion is due to the fact that there were actually four al-
ternatives, not three: the sales tax, the income tax, the excise tax, and the
status quo of no tax. The pairwise-comparison matrix Table 15.2 contains
my corrected inferences of the rankings among S, I, and E as well as the
revealed votes between each of those and Q. The pairwise rankings are
E > I, E > S, E > Q, S > I, Q > I, Q > S, and that reduces to
E >Q>I >S, which is consistent with the commonsense narrative about
this vote, that is, the excise tax beat the status quo and was adopted but
the income tax and the sales tax were defeated. Its like breathing fresh
air again.
If there were a cycle, then the last alternative voted on would have
won arbitrarily, he says. The last alternative voted on was E, which won.
Blydenburgh (1971, 67) proposes that, If the excise tax the amendment
that actually was adopted had been introduced rst, it would have been
defeated by the 180 voters who preferred the income tax plus the 28 vot-
ers who were opposed to any tax. Had the sales tax been introduced last,
it would have been adopted. What if the excise tax had been introduced
rst? Then the contest would be between E and Q, and we know E would
win that. If the contest were between E and S, then E would win 200 to
156 on Blydenburghs inferred rankings. If the contest were between E
and I, and if his inferred preferences are correct, then I would win, but
I have argued that there is no cycle because on my corrections E > I,
187 to 178. If the contest were between E and the impossible composite
not-E, then, we have seen, E would win. Now Blydenburgh makes a third
argument: 182 votes are required for I to have a majority over E; I has
180 votes against E, to the 180 we add the 28 voters (Group D) who
were against all taxes and obtain a majority for I over E. Actually, Group
D possessed 30 voters who voted against all taxes, but this is a side issue.
The members of Group Dvoted against all taxes, they voted against Eand
they voted against I; thus, if the contest were between E and I they should
vote for neither and abstain. If the members of Group D abstain, then
on the remainder of Blydenburghs count there are 180 voters for I > E,
112 votes for E > I, the ranking of the 71 voters in Group H is unde-
termined, and we are squarely back at the issues already encountered in
discussion of his second argument.
What would have happened if the sales tax were voted on last? If the
vote were between S and E, then E would win by 200 votes to 156 on
Blydenburghs inferred count. And we know that E would beat Q. If the
vote were between S and I, then S would win 194 votes to 162 on Bly-
denburghs inferred rankings. But S must go against Q, and then Q wins.
Next, Blydenburgh (1971, 67) says that if the income tax had been voted
on last, it probably could not have passed because of the opposition
of Group H. Because there is a cycle, the argument goes, any alter-
native voted on last will win. But if the income tax were voted on last,
it would not win, he says. Does this mean that there is a cycle, but only
between S and E? Howwould that be? I do not understand his argument.
Finally, Blydenburgh uses the same methodologies to examine votes on
the Revenue Act of 1938. The tale always gets better in the telling, and
Blydenburgh (1971) is sometimes cited as having demonstrated cycles in
both 1932 and 1938 (e.g., Shepsle and Bonchek 1997, 61). Blydenburgh
does not nd a cycle in 1938, and the reader may be relieved to learn that
there is no need to reconstruct his second case.
Anyone who would read Blydenburgh (1971) attentively, with pencil
and paper at hand to work through the details of his argument, would
quickly discover the heap of conceptual confusions I have related. Yet, so
far as I know, there are no other reports that Blydenburghs cycle nding
is erroneous. Rather, the nding is cited approvingly as central evidence
for the claim that democracy is meaningless (Riker and Weingast 1988;
Shepsle and Bonchek 1997). This raises an important problem about
the Rochester approach. Riker (1965) promised us that mathematics
would introduce precision and eliminate ambiguity frompolitical science.
Bueno de Mesquita (quoted in Cohn 1999) told us that the Rochester
school is dominant because the clarity of its work attracts attention. What
we have seen, however, over and over again, is obscurity, not clarity, and
confusion, not insight; an obscurity and confusion, moreover, disguised
fromthe eyes of the world by supercially impressive formalisms. Stephen
Walt reminds us of the costs of formalization, that precision comes at a
price. He says there is a tendency for formal theorists to present their
ideas in an overly complex and impenetrable manner, and he criticizes
the style in which formal methods are deployed to lend a quasi-scientic
patina to otherwise simple (and sometimes mistaken, I would add) ideas.
According to Walt (1999, 21):
the larger the audience that can understand and evaluate a theory, the more likely
is it that errors will be exposed and corrected and the better a theory has to be in
order to retain approval. By contrast, an incorrect theory that is presented in an
opaque or impenetrable way may survive simply because potential critics cannot
gure out what the argument is.
The same malady notoriously aficts certain narrative theories; consider,
for example, the work of Talcott Parsons.
A roll-call analysis is a comparatively simple exercise. The one we have
here was botched from inception in 1971, yet as of 1997 it still ranks
high in the Rochester canon. If gross error in a simple paper high in
the canon goes unnoticed for 26 years, then what are we to think about
the more formal treatments of more complicated problems since in the
same tradition?
Bjurulf and Niemi on Scandinavian parliaments
Bjurulf and Niemi (1978) intend their paper to be an empirical explo-
ration of Rochester doctrine. A case is being made, they write four years
before the publication of Liberalism against Populism by Riker, Niemis
colleague at Rochester, that voting systems are manipulable by strategic
voting, agenda setting, and vote trading, although there is little hard evi-
dence on the extent of manipulation in actual settings, they continue. The
lack of evidence is due largely to the fact, as we have heard before, that the
usual voting rules typically do not record voters rankings over all alterna-
tives. Another reason for the lack of evidence, they add, is that legislators
naturally desire not to publicize that they have voted strategically or that
they have attempted to control the agenda. In the cases I have analyzed
in this volume, however, legislators are frequently at pains to publicize
the fact that their votes are strategic. In the 1957 replay of the Powell
amendment, one Democrat after another stood up to say that his vote
was strategic. In debate on the 17th Amendment to the US Constitu-
tion, Bristow freely declared the strategic nature of his voting. In the nal
defeat of the Wilmot Proviso, one of the seven who switched votes pub-
lished an explanation to his constituents concerning his strategic response
to the situation forced upon him. This unsystematic sample shows that
legislators sometimes naturally desire to publicize the strategic nature of
their votes, surely so that constituents can appreciate their true positions.
To continue the exposition, Bjurulf and Niemis cases lead them to the
conclusion that, at least in the Swedish context, manipulation occurs not
by way of an agenda-controllers ordering of consideration of alternatives,
nor by introduction of confounding new issues or dimensions, but rather
by way of strategic voting emerging in response to unmanipulated se-
quences of alternatives. They report on three cases, and also report more
generally on strategic voting in Scandinavian legislatures. I was unable to
inspect the parliamentary record they rely on, lacking both access to the
documents and ability to read Swedish.
The rst case took place in Chamber I of the Swedish Parliament in
1931. There were three alternatives under consideration:
r
Big: Build the Karolinska hospital as planned.
r
Small: Build the rst section of the hospital, but not additional sections.
r
Nothing: Do nothing, remain at the status quo.
Big was the choice recommended to the chamber by the committee of
jurisdiction. The rst chamber vote, according to Bjurulf and Niemi, was
between Small and Nothing, and Nothing won the rst vote. The second
vote was between Big and Nothing, and Big won the second vote and
thus the whole contest.
r
Nothing > Small, 4641, with 63 abstaining or absent.
r
Big > Nothing, 5416, with 80 abstaining or absent.
There was not a vote between Big and Small, but Bjurulf and Niemi
infer the collective ranking Small > Big by 47 to 37 from reconstruc-
tion of the preference orderings of legislators present and voting (7).
Combining the latter inference with the two recorded votes they obtain
Big > Nothing > Small > Big, a cycle.
Their reconstruction is as follows. First, 10 legislators voted Small >
Nothing and Nothing > Big, and apparently these 10 ranked Small >
Nothing > Big. This is a reasonable preference order, in my view; such
voters are willing to pay for a small hospital, but would rather have nothing
than pay for a big one. Next, 14 legislators voted Small > Nothing, but
abstained on Big versus Nothing. Bjurulf and Niemi interpret the absten-
tions as indifference between Big and Nothing. Three legislators voted
Nothing > Small but were absent for the vote between Big and Noth-
ing. Next, 17 legislators voted Small > Nothing and for Big > Nothing.
Bjurulf and Niemi present evidence from the debate that legislators of
this type expressed a preference for Small > Big for an overall ranking
of Small > Big > Nothing. Six legislators voted Nothing > Small and
Nothing > Big. Bjurulf and Niemi infer that these six also rank Small >
Big and thus that their overall ranking is Nothing > Small > Big: Since
these legislators favored no expenditure at all, we conclude that they
preferred a smaller expenditure . . . to a larger one (8). This is the nat-
ural presumption I believe, but its also logically possible that these
six ranked Nothing > Big > Small. The latter ranking might make
sense, say, if there were a large private hospital these legislators con-
stituents used that would be destroyed by competition from any size
of public hospital: they would prefer the status quo of a private hos-
pital, but if a public hospital must be built then better make it a big one.
Nothing > Big > Small is a plausible inference, but in the absence of ev-
idence it is not as plausible as Bjurulf and Niemis inference, Nothing >
Small > Big. Finally, 37 legislators voted both Big > Nothing and
Nothing > Small, and Bjurulf and Niemi infer that they ranked Big >
Nothing > Small. I believe that this inference is mistaken. As with the
6 legislators just discussed, in the absence of other evidence, it is more
plausible that the 37 ranked Big > Small > Nothing and were voting
strategically on the rst vote between Nothing and Small. They sincerely
preferred Small to Nothing, but since they also preferred Big to Small,
on the rst vote they strategically voted for Nothing over Small on the be-
lief that on the second vote Big would more probably beat Nothing than
would Big beat Small. It is less plausible that the 37 sincerely preferred
Big > Nothing > Small. Bjurulf and Niemi acknowledge in a footnote
that the 37 could have been voting strategically and if so that the demon-
stration of a cycle would fail. Their response is that there is no evidence
in the record to support the inference that any of the 37 were voting
strategically. But neither is there evidence in the record, else they would
have reported it, for their less plausible inference of Big > Nothing >
Small. For the 37 we have no direct voting evidence, rather we must
infer whether they rank Big > Small or Small > Big. Why doesnt the
presumption that Bjurulf and Niemi applied to the 6 legislators, that the
second-ranked alternative should be intermediate in quantity to the rst-
and third-ranked alternatives, apply to the 37 legislators? If the presump-
tion is consistently applied then the 37 must have sincerely ranked Big >
Nothing > Small, and must have been voting strategically.
Summing up Bjurulf and Niemis reconstructions, a total of 43 legis-
lators favored Small > Big and 41 favored Big > Small, and thus the col-
lective preference would be Small > Big; this, together with the recorded
votes of Big > Nothing and Nothing > Small, is the inference that gives
Bjurulf and Niemi their cycle. If, however, merely 2 out of the 37 whom
they infer ranked Big >Nothing >Small were, as I argue, strategic voters
who sincerely ranked Big > Small > Nothing, then the collective prefer-
ence would be Big >Small which yields the noncyclical collective ranking
Big >Nothing >Small. Notice a peculiarity about these votes: there were
about as many abstentions and absences as there were recorded votes: on
the rst pair 87 voted and 63 did not, and on the second pair 70 voted and
80 did not. I have a hunch that the actual collective ranking for all legis-
lators including the abstainers was actually Big > Small > Nothing, but
this cannot be shown. Depending on the context, either there was con-
siderable indifference among the three alternatives, or the full-chamber
outcome was so clearly for Big that many legislators did not bother to
attend the oor and those who bothered to come and vote against Big
wanted to take a position of favoring economy in government. Without
an explanation for the large number of nonvoters, no inference fromthese
data is strongly warranted (Bjurulf and Niemis remaining cases report
comparatively few abstentions and absences).
In their second case, Bjurulf and Niemi do not claim that there was
a true cycle, but they do claim that one faction by strategic voting ma-
nipulated its favored outcome from second place unfairly to rst place.
Strategic voting can be defeated by strategic voting, however; thus, as
we saw with Rikers accounts of the Powell amendment and of the
17th Amendment, either the claim is incorrect or we must explain why
some voters were irrational and voted in a sincere and self-defeating fash-
ion. The vote was in the Swedish Parliament, in 1934, over how much to
expand the Swedish telephone and telegraph company. There were three
alternatives:
r
Twelve: Spend 12.35 million Swedish crowns on expansion
r
Eleven: Spend 11.35 million Swedish crowns on expansion
r
Ten: Spend 10.35 million Swedish crowns on expansion.
The Social Democrats, the minority government, had campaigned
in 1932 on increased state expenditures. Their 11 members on the
24-member committee of jurisdiction in Chamber I favored alternative
12, and the Social Democrats must have ranked 12 > 11 > 10. The 7
Conservatives on the committee, whose party had campaigned in 1932
on small state expenditures, proposed 10, and they must have ranked
10 > 11 > 12. Those present of the 6 Farmers Party and Liberal Party
members of the committee proposed alternative 11 as a compromise.
Bjurulf and Niemi suggest that 11 must have been the median preference.
They acknowledge that the Liberals and especially the Farmers prefer-
ences between 12 and 10 are uncertain, in fact they themselves may have
been unsure about their preference between these two alternatives since
in voting they sometimes favored one and sometimes the other (9, emphasis
added). FromBjurulf and Niemis reportage I infer that, going fromleft to
right in Chamber I there were 61 Social Democrats, 19 Farmers, 17
Liberals, and 46 Conservatives; presumably the center of the Farmers
Party was the median position in Chamber I. In Chamber I, the rst
vote was between 11 and 10. The Farmers and the Liberals voted for
their compromise proposal 11, the Conservatives sincerely voted for 10,
and most of the Social Democrats as well voted strategically for 10, and
thus 10 beat 11. All the evidence supports the inference that the Social
Democrats vote for 10 was strategic; we can surmise that they believed
that it was more likely that Chamber I wouldvote for 12 >10 thanit would
vote for 12 >11. The next vote in Chamber I then was between 12 and 10,
but 10 beat 12 by 58 votes to 56 votes; thus, the Social Democrats strate-
gic vote had been a blunder. If they had voted sincerely they would have
won their second-ranked alternative, 11, rather than their last-ranked
alternative, 10.
Meanwhile, in Chamber II alternative 10 beat 11 by a rise vote (the
Social Democrats there also apparently voting strategically), and then 12
beat 10 by a rise vote. I calculate that Chamber II contained, from left to
right, 5 Communists, 104 Social Democrats, 36 Farmers, 24 Liberals,
and 58 Conservatives. The median voter was presumably in the far left
of the Farmers Party. Chamber II was to the left of Chamber I. If the
Communists and Social Democrats were united and disciplined on an
issue (as they were on this one) they needed to pull only 5 votes from any
of the 118 legislators in the remaining parties (and Bjurulf and Niemi note
that the parties were not always cohesive, 21). In Chamber I alternative
10 had won and in Chamber II alternative 12 had won, and apparently
the Swedish rule was that differences between Chambers were settled by
a joint vote of both Chambers. The results of the joint vote were that
12 > 10 by 67 to 56 in Chamber I, by 117 to 71 in Chamber II, and
by 184 to 127 in the decisive summed votes of both Chambers. Why the
reversal in Chamber I? The bulk of the Farmers in Chamber I voted for
10 >12 on the rst round but for 12 >10 on the second round. The vote
of the Farmers in Chamber II over 12 versus 10 when Chamber II voted
alone is unrecorded, but on the later joint vote with Chamber I most of
the Farmers in Chamber II voted for 12 > 10.
The collective ranking in Chamber I when it voted alone might have
been 11 > 10 > 12, because 11 might have been the position of the
median voter and we know that 10 > 12 on the second vote of the
rst round. After the curious switch of the Farmers on the later joint
vote the collective ranking in Chamber I may or may not have become
11 > 12 > 10. In Chamber II we know that the collective rankings
were 12 > 10 and 11 > 10, but we have no idea how Chamber II
ranked 12 and 11; thus, the collective ranking in Chamber II was either
12 > 11 > 10 or 11 > 12 > 10. In both Chambers the Social Democrats
voted strategically for 10 > 11. The Conservatives in both Chambers
failed to vote strategically for 11 >10 in response to the Social Democrats
and, the argument goes, thereby ended up with their third-ranked alter-
native rather than their second-ranked alternative. Bjurulf and Niemis
proposition that the Social Democrats manipulation succeeded assumes
that the Conservatives were irrational. The Conservatives in Chamber
I were clearly not irrational, however: by voting sincerely the Conserva-
tives won their rst-ranked alternative in Chamber I, while the bungled
strategic vote of the Social Democrats left the Social Democrats with their
last-ranked alternative there.
Thus, Bjurulf and Niemi must believe that it was the Conservatives
in Chamber II who were irrational for failing to vote strategically for
11 > 10. If the Conservatives in Chamber II believed, however, that the
sincere collective ranking in Chamber II was 12 > 11 > 10 then those
Conservatives would have had no reason to vote strategically for 11 >10,
indeed they would have looked foolish for doing so. Only if those Con-
servatives believed that the Chamber II ranking was 11 > 12 > 10 would
they have acted irrationally by voting sincerely. We dont have enough
data to know whether Chamber II ranked 12 >11 >10 or 11 >12 >10,
and the presumption of rationality commands that in the absence of other
information we accept that the Conservatives in Chamber II believed that
the collective ranking there was 12 >11 >10. In support of the view that
the Conservatives in Chamber II rationally believed that the collective
ranking in Chamber II was 12 > 11, consider that in the second-round
contest between 12 and 10 in Chamber II 24 Farmers and Liberals voted
for alternative 12 and 23 Farmers and Liberals voted for alternative 10,
and 13 Farmers and Liberals abstained or were absent; to win a collective
vote for 12 >11 the 104 united Communists and Social Democrats would
have needed only 5 out of the 60 Farmer and Liberal votes. On another
point, the Social Democrats were the largest (43 percent of Chamber I
and 46 percent of Chamber II) and the governing party. There is no rea-
son to believe that their proposal of 12.35 million Swedish crowns repre-
sented the median position of their party. It is more likely that they would
offer a gure they believed to be of median appeal to the joint Chambers,
attracting all Communists, all Social Democrats, and the requisite hand-
ful of votes from the Farmers and Liberals. Finally, the three alternatives
may have been so close to one another as to have been a matter of some
indifference to the participants. The Conservatives took the position of
supporting a slightly smaller expansion, the Farmers and Liberals initially
took an intermediate position that distinguished them from the Social
Democrats on the left and the Conservatives on the right, but, as Bjurulf
and Niemi observed, the same Farmers and Liberals sometimes voted for
12 and sometimes for 10. We are more likely to witness inconsistencies
that suggest the possibilities of a cycle or a successful manipulation when
alternatives are very close to one another (and the normative implica-
tions of instability are trivial) rather than when alternatives are far apart
from one another. In conclusion, Bjurulf and Niemi fail to demonstrate a
harmful manipulation by the Social Democrats in the vote over expansion
of the Swedish telephone and telegraph company.
Riker writes of strategic legislators as omniscient Machiavellian
schemers. Green and Shapiro (1994, 111) remind us that they can be
bumbling oafs as well, as the Social Democrats were on the telephone
and telegraph vote in Chamber I above and on the riemans vote below.
The third case, again in the Swedish Parliament, concerns the 1927 ap-
propriation for the voluntary riemans association. The minority com-
mittee report from the Conservatives asked for 500,000 Swedish crowns
in order to include riemen, or rather boys, aged 12 to 15 in the activity, a
novel proposal. The Liberal government and the relevant committee re-
port recommended the conventional appropriation of 470,000 Swedish
crowns. The Social Democrats offered a minority report sincerely rec-
ommending no appropriation for the riemans association. Bjurulf and
Niemi argue that that the collective preference of the joint chambers was
470,000 > 0 > 500,000, but that the last-ranked alternative, 500,000,
perversely prevailed in the end. Social Democrats, knowing that the
Liberals were uncomfortable with the Conservatives 500,000 because
the Liberals disapproved of involvement of boys in the association, strate-
gically abstained on the rst vote, between 500,000 and 0, in Chamber I.
If the Social Democrats had acted sincerely the outcome would have
been 0, but just enough of them abstained so that 500,000 won by one
vote. The next vote then, was between 500,000 and 470,000. The Social
Democrats strategically abstained again, forcing the outcome to 500,000
(their sincere action would have been to vote for 470,000). In the absence
of strategic abstention by the Social Democrats the outcome in Chamber I
would have been 470,000. Apparently, the Social Democrats believed that
if the choice were between 500,000 and 0 then most Liberals would vote
for 0, assisting the Social Democrats to win their rst-ranked alternative.
This was quite a blunder, as the Liberals ultimately responded in a fash-
ion so as to attain the Liberals objective and thwart that of the Social
Democrats.
Meanwhile, in Chamber II the votes were 0 > 500,000 and then
0 > 470,000, apparently with no strategic voting by any agents. Again,
because the two Chambers disagreed, the issue went to a joint vote be-
tween 500,000 and 0. In the joint vote, however, most Liberals (f ) in
Chamber I switched from voting 0 > 500,000 to voting 500,000 > 0.
The Liberals (f ) in Chamber I were decisive: if in the joint vote they had
voted as they had when the issue rst arose in Chamber I, then the joint
collective outcome would have been collectively second-ranked 0 rather
than collectively last-ranked 500,000. What was the true position of these
Liberals and why did they switch? Bjurulf and Niemi (12) are condent,
citing debate evidence, that these Liberals sincerely ranked 0 > 500,000.
Further:
The explanation for this switch is very probably some sort of deal made with the
Conservatives and the Farmers prior to the joint vote. We base this conclusion
on the fact that the next year, 1928, the sum appropriated for the voluntary rie-
mans association was 440,000 Swedish crowns, with no dissenting alternatives
fromthe Conservatives. The [1928] appropriation contained nothing for riemen
between 12 and 15 years of age. Thus, the situation seems to be have been that
the Liberals, faced with the alternative of no appropriations or one that included
appropriation for the very young, would support the [1927] appropriation with
the understanding that the Conservatives and Farmers would not bring up this
proposal again the following year. (16)
The best way to make sense of these events is to recognize that a fourth al-
ternative had been introduced: 500,000
this year (30,000 above the con-

ventional appropriation) so that the Conservatives and Farmers would
not disappoint their constituents and so that the Social Democrats would
be punished for their chicanery, on the understanding that 440,000
would be offered in the following year (30,000 below the conventional
appropriation), which I denote with a prime mark. True, the Liberals
(f ) ranked 470,000 > 0 > 500,000, and the abstention of the Social
Democrats denied the Liberals their rst-ranked choice in the short
run. But the other parties countered the Social Democrats by crafting
500,000
, which won majority support in the joint vote. The Liberals

(f ) must have ranked 470,000 > 500,000
> 0 > 500,000. Strategic ac-

tion was countered by strategic action such that the nal outcome was
what it would have been in the absence of strategic action; attempted
manipulation was neither irremediable nor harmful.
Bjurulf and Niemi report on a more general survey of Scandinavian
parliaments. They looked primarily in years when there were minor-
ity governments, because majority governments would rarely generate
a record of strategic voting among more than two alternatives. They un-
covered further cases that resemble the three cases we have analyzed in
detail, and suggest that if analyzed in detail the further cases would fur-
ther demonstrate cycles, harmful manipulation, and so on. Since their
three cases, which it is fair to assume are their best examples, are weak,
their generalization is weak. In Sweden, they say, strategic manipulation
frequently resembled that in the second case: the Social Democrats vote
strategically for the Conservative alternative in order to eliminate the cen-
trist alternative offered by the Farmers and Liberals, according to Bjurulf
and Niemi. In 19571958, the Social Democrats, who were larger than
any combination of two of the three bourgeois parties, attempted this
more often than not whenever there were more than two alternatives to
vote on. Bjurulf and Niemi point out that nevertheless the Social Demo-
cratic alternative had to be centrist enough to defeat the Conservative
alternative. There was reportedly much less strategic voting in the other
Swedish period examined, 19251938. In Finland during periods of mi-
nority government, the situation was much the same as in Sweden. A
variation in Finland was for the Agrarians and Finnish Peoples Party
to vote strategically for the Communist alternative so as to eliminate the
more centrist Social Democratic alternative. Bjurulf and Niemi claimthat
there were true cycles in Finnish politics, but none that were taken ad-
vantage of strategically. Even in periods of minority government strategic
voting was infrequent, in 90 percent or more of roll calls there were only
two alternatives to vote on. There were almost no strategic votes in
Norway and Denmark as these countries used the successive proce-
dure in which alternatives are voted on one at a time until some alterna-
tive receives a majority vote. This procedure is especially vulnerable to
unpredictable voting-order effects which motivates legislators, say Bjurulf
and Niemi, to limit to two the alternatives brought to the oor.
Does the analysis of Bjurulf and Niemi support the proposition that
democracy is meaningless? Their rst case attempted to show a natural
cycle, but relied on the assumption that 37 of the voters had an extraor-
dinary preference ranking. I showed that if only 2 out of those 37 were
voters with an ordinary preference ranking but voting strategically then
there was no cycle. Their second case purported to show a case of harm-
ful strategic voting, but violated the presumption of rationality: without
evidence they assumed that the Conservatives in Chamber II had beliefs
which made their actions irrational, when it was possible to assign an
equally plausible alternative belief that preserved the rationality of those
Conservatives. My interpretation of their second case illustrates bungled
strategic voting. The third case is an instance of strategic voting being
countered by creative strategic activity that restored the centrist outcome.
My viewis that we should see either no strategic voting, bungled strategic
voting, or strategic voting successfully countered by strategic voting, in
any case such that generally the centrist outcome prevails. Bjurulf and
Niemis general survey shows that even with minority governments votes
involving more than two alternatives and associated possibilities for strate-
gic voting are rare. Furthermore, frequently harmful manipulation is not
established because their general survey relies on methods and insights
shown to be mistaken in their three fully worked out cases. In sum, I
dispute Bjurulf and Niemis conclusion that (harmful?) manipulation is
a frequent occurrence.
Neufeld, Hausman, and Rapoport on Muscle Shoals
Neufeld, Hausman, and Rapoport (1994) review and summarize the
standard cycling story, citing Riker and his followers, among others.
Cycles are difcult to identify, again because it is rare for there to be
recorded votes over all alternatives. Indeed, the literature on cyclical ma-
jorities has failed to uncover a single clear example of cyclical voting (427,
emphasis added), even though Rikers analyses of the Powell amendment
and the 17th Amendment appear to be consistent with the paradox
(426). Neufeld et al. are an example of how Rikers already mistaken
stories become even more garbled in transmission. They cite as primary
evidence for a cycle in the Powell amendment votes the fact that in the
following year an education bill without the Powell amendment passed,
although alas the evidence is too weak for a certain judgment, they con-
tinue. Recall that I showed that the unamended education bill actually
failed in the following year, and cited that as important evidence against
Rikers cycle argument! The 17th Amendment failed, they write, because
although the proposal for the direct election of senators contained a pro-
vision prohibiting federal supervision of elections, an amendment by the
House eliminated the prohibition. I showed that the 17th Amendment
earlier would have failed in the Senate with or without the prohibition
provision; and later it was the House that demanded the prohibition and
the Senate which successfully resisted it. Enough of my points against
Neufeld et al. Their complaint against Riker is that his cases depend
not only on recorded votes but also on uncertain inferences. In contrast,
they believe they have nally identied a denitive and signicant ex-
ample of cyclical voting (423) based only on recorded votes. It had to do
with Muscle Shoals, an obstacle to navigation on the Tennessee River in
Alabama, one of the most important issues in the US Congress in the
1920s, and which eventually evolved into the Tennessee Valley Authority
in 1933 under Roosevelt. The votes were in a major political body, the US
Senate. And by luck there were three pairwise votes recording expressed
preferences over three alternatives. Finally, one of the participants stated
that the body was caught in a voting circle among the three alternatives.
What more could one ask?
If Muscle Shoals were tamed, an obstacle to navigation would be re-
moved, huge amounts of hydroelectricity would be generated, and ni-
trates could be manufactured for agricultural and military uses. Among
the questions were how to proceed, whether to use public enterprise or
private enterprise, and if the latter then which private enterprise. Senator
Underwoods bill was crafted in consultation with the Republican admin-
istration, and it appeared to have the solid backing of President Coolidge
(429), and would lease the facilities to private enterprise. Senator
Norris (R-NE), a left-liberal insurgent Republican and advocate of pub-
licly owned power utilities, denounced the Underwood plan as a giveaway
to private interests, which would result in another scandal like Teapot
Dome, already a huge embarrassment to the Coolidge administration. In
response:
Republican senators came from meetings with the President convinced that Un-
derwoods measure would not pass and that the best disposition of Muscle Shoals
was for it to be sent to a commission that would make recommendations to
Congress after a years study. (Neufeld et al. 1994, 430)
The administrations senators crafted a proposal to create such a com-
mission, introduced by Senator Jones. Thus matters stood as the Senate
recessed for Christmas.
The votes occurred over one week in January 1925. Norriss proposal
(N) was rst. Underwood offered his proposal (U) as a substitute, and
the Senate voted for U > N, private development over public develop-
ment, by 48 to 37 (the rst vote). Several days of debate followed. Then,
according to Neufeld et al., the Senate voted for J > U, a study commis-
sion over private development, by 46 to 33 (the second vote). Norris then
reintroduced his proposal (with a cosmetic change to dodge the Senate
rule against reintroduction of proposals) again, and Norris beat Jones,
public power beat a study commission, N > J by 40 to 39 (the third
vote). Thus, U > N > J > U: what we seem to have here is an unequiv-
ocally demonstrated cycle. Underwood sought then to reintroduce his
proposal, which Norris moved was out of order, declaring that we are
in a circle with three points in it, such that one alternative would beat
the next and we would go around the circle again, and we would be just
where we started (431). Underwoods proposal was reintroduced the
next day and passed by a vote of 46 to 33 (U > N, the fourth vote). Then
Joness proposal was reintroduced, but there was something new in the
air, as Norris promised not to reintroduce his proposal if Joness won.
Indeed, U beat J, 43 to 38 (the fth vote). Yesterday Norris beat Jones
and Jones beat Underwood, but today the combined forces of Norris and
Jones arent enough to beat Underwood. This should make us suspicious
Table 15.3. Neufeld et al.s account of Muscle
Shoals preferences
Group number and label Preferences #
1. Anti-Ford Republicans J > N > U 3
2. Jones group J > U > N 13
3. Norris group N > J > U 24
4. Southern Democrats N > U > J 4
5. Underwood group U > J > N 17
6. Southern Democrats U > N > J 9
about the purported cycle. Underwood went on to beat the status quo by
a vote of 50 to 30 (the sixth vote).
There was no natural cycle because, as Neufeld et al. themselves point
out, many of those voting for Jones truly favored Underwood, but at
the prompting of the Coolidge administration introduced and voted
for the Jones compromise on the belief that Underwood could not win
in the Senate. Voting then revealed that Underwood could indeed win,
and those Jones voters then shifted to Underwood. There is evidence
that the shift came at the direction of the Coolidge White House. The
Republican oor leader shortly before the fth vote [U>N, 46 to 33] met
with the President who was, by this time, reported as favoring Under-
woods position (433). Neufeld et al. do not distinguish between natural
cycles and apparent cycles contrived by strategic voting, believing that
for prior attempts to demonstrate empirical cycles, such as Rikers, it
is the attempt at manipulation (sophisticated voting by individuals) that
creates the effect of cyclical majorities (426), which is an incorrect read-
ing of Riker, who sometimes wrongly believes that he has uncovered a
natural cycle. Muscle Shoals appeared to be a cycle, as so often in these
case studies, only because of bungled strategic voting; sincere preferences
were in equilibrium. Despite both apparent cycling and strategic voting
the nal outcome of the Muscle Shoals votes was in the center of the
Senates opinion. Neufeld et al. make no claim that harmful manipulation
occurred.
The apparent cycle is borderline as well, because if only one of the vot-
ers had switched from N to J the expressed collective ranking would have
gone from the cyclical J > U > N > J to the noncyclical J > U > N.
Neufeld et al. tally the preferences of the 71 senators who voted on each
of the rst three votes, Uagainst N, J against U, and Nagainst J. They as-
sign senators to the six groups displayed in Table 15.3. For interpretations
of each of the groups, see the original article. Notice that the southern
Neufeld et al.s count
J N U BC
J 33 40 73
N 37 31 68
U 30 39 69
Democrats in Groups 4 and 6 wanted private or public development of
the Tennessee River and did not want the delayed action of a Jones com-
mission. Perhaps the Republicans who devised the Jones compromise
on the mistaken belief that Underwood couldnt win underestimated the
strengthof SouthernDemocratic support for this huge project insouthern
territory. Neufeld et al. observe that it was a shift of voters fromGroup 2 to
Group 5 that voided the cycle in the end. Prior to that shift, pairwise out-
comes, and associated Borda counts, were as shown in Table 15.4. J > U
by 40 to 30, U > N by 39 to 31, and N > J by 37 to 33, for the apparent
cycle J >U>N>J. The cycle is weak even on its own terms. The Borda
count gives us the ranking of J >U>N. The YoungKemeny rule, which
tells us to break the cycle at the weakest link, would strike N > J by 37
to 33, and thus would also provide the ranking J > U > N. Neufeld et al.
do not do a precise count, but, as I shall show in detail below, when we
substitute for the strategic preferences expressed in the tables above the
sincere preferences revealed by the exodus from Group 2 to Group 5, the
true collective ranking becomes U > J > N, and it was U that passed out
of the Senate: no cycle, no harmful manipulation.
Neufeld et al. rely only on the votes of the 71 senators who voted on all
three pairwise comparisons on the rst, second, and third votes (Uagainst
N, J against U, N against J ). It is also possible to infer the preference
orders of senators who voted in only two of those three votes. If a senator
voted U > N in the rst vote, missed the second vote, and voted N > J,
we can make the fallible inference that he ranked U > N > J. The fourth
and fth votes, after many Group 2 voters migrated to Group 5, only
record preferences expressed over U against N and J against U. Again, it
is possible to infer preference orders. For senators who voted for only two
out of the rst three votes, and for senators who voted on both the fourth
and fth contests, my fallible inference rule permits the 1,2,3 inference
to borrow from the 4,5 rankings, or the 4,5 inference to borrow from
the 1,2,3 rankings. Then, we can compare preference orders during the
rst, second, and third votes to preference orders during the fourth and
fth votes, and identify senators whose preferences changed fromone day
to the next, and make further inferences about the changes. I report this
exercise inTable 15.5. The columnlabeled1,2,3 reports inferredrankings
arising from the rst, second, and third votes. The column labeled 4,5
reports inferred rankings arising from the fourth and fth votes. The
columnlabeled6 reports the nal vote betweenUnderwoodandthe status
quo. The summary rule of inference is to use all pairwise comparisons
to infer ranking of the three alternatives, unless there is a contradiction.
An asterisk indicates that a ranking was borrowed from 1,2,3, by 4,5 or
vice versa. Adouble asterisk means that a ranking was not borrowed from
one column to another. If an entry is underlined in the column labeled
4,5 that means the senators ranking changed from that of 1,2,3, and
the change from one ranking group to another is reported in the column
labeled comments. An underlined entry in the column labeled 6 indicates
a vote inconsistent with earlier expressed preferences. No inferences are
possible about senators who missed four, ve, or six out of the six votes,
nor do I count Greene.
The rankings from Table 15.5 are summarized in Table 15.6. The col-
umn labeled N, H & R reports Neufeld et al.s rankings from actual votes
1, 2, and 3. The column labeled M 1,2,3 reports the actual votes and my
inferences over votes 1, 2, and 3, and the column labeled 4,5 reports my
inferences over votes 4 and 5. There were 96 senators, and 88 of themcast
at least one vote out of the six. I obtain expressed or inferred preferences
for 80 or 81 senators, as compared to the 71 that Neufeld et al. obtain by
relying only on recorded votes. One of the senators, McKinley, expressed
a cyclical preference in votes 1, 2, and 3. When the administration sena-
tors in the Jones group were voting strategically during votes 1, 2, and 3,
my 80 senators made up a collective ranking of J > U > N J, a weak
cycle. If we add the votes of cyclical McKinley, however, then we obtain
the collective preference ranking J > U > N, which is not a cycle! The
collective rankings are obtained from the pairwise-comparison matrix in
Table 15.7.
Between votes 1,2,3 and votes 4,5, 17 senators changed expressed
preference rankings. The most important are the ten who changed from
Group 2 to Group 5; these were mostly the administration Republicans
who were freed to vote for their most preferred outcome, Underwood.
Further, two senators switched from 1 to 2, and another two switched
from 2 to 1, canceling each other out. The remainder are: one from 6
to 4, one (McKinley) from 7 to 5, and one from 4 to 3. That yields a
pairwise-comparison matrix, shown in Table 15.7, for sincere preferences
that again conrms the complete absence of any cycle. The Condorcet,
pairwise-comparison order is U > J > N, as is the Borda count. It was U
that won in the Senate. Hence, Neufeld et al. have failed to demonstrate
Table 15.5. Mackies inferred rankings, Muscle Shoals
Votes
Senators 1,2,3 4,5 6 Comments
Ashurst NJU (3) NJU (3)
QU
Ball JUN (2) UJN (5)
UQ 2 TO 5
Bayard UJN (5) UJN (5) UQ
Bingham JUN (2)
UJN (5)
UQ 2 TO 5
Borah NJU (3) NJU (3) QU
Brookhart NJU (3) NJU (3) QU
Broussard 5/6 ABSENT
Bruce UNJ (6) UNJ (6) UQ
Bursum JUN (2) UJN (5)
UQ 2 TO 5
Butler UJN (5) UJN (5) UQ
Cameron JUN (2) UJN (5)
UQ 2 TO 5
Capper NJU (3) NJU (3) QU
Carraway 5/6 ABSENT
Copeland NJU (3) NJU (3) ABSENT
Couzens JNU (1) JNU (1) QU
Cummins JNU (1) JUN (2)
QU 1 to 2
Curtis UJN (5) UJN (5) UQ
Dale UJN (5) UJN (5) UQ
Dial UNJ (6) UNJ (6) UQ
Dill NJU (3) NJU (3) QU
Edge UJN (5) UJN (5) UQ
Edwards 5/6 ABSENT
Elkins 5/6 ABSENT
Ernst JNU (1)
JUN (2)
UQ 1 to 2
Fernald UJN (5) UJN (5) UQ
Ferris NJU (3) NJU (3) QU
Fess UJN (5) UJN (5) UQ
Fletcher UNJ (6) NUJ (4)
UQ 6 to 4
Frazier 5/6 ABSENT
George UNJ (6) UNJ (6) UQ
Gerry UJN (5) UJN (5) UQ
Glass 5/6 ABSENT
Gooding NJU (3) NJU (3) QU
Greene UJ UN, UJ UQ Incomplete
Hale UJN (5) UJN (5) UQ
Harreld JNU (1) JNU (1) QU
Harris NUJ (4) NUJ (4) UQ
Harrison UNJ (6) UNJ (6) UQ
Hein UNJ (6) UNJ (6) UQ
Howell NJU (3)
NJU (3)
QU
JohnsonCA NJU (3) NJU (3) QU
Johnson-MN 6/6 ABSENT
Jones-NM NJU (3)
NJU (3) QU
Jones-WA JUN (2) JNU (1)
QU 2 to 1
Table 15.5. (cont.)
Votes
Kendrick NUJ (4) NUJ (4) UQ
Keyes UJN (5) UJN (5) UQ
King UJN (5) UJN (5) UQ
(A)
Ladd UNJ (6) UNJ (6) UQ
LaFollette NJU (3) NJU (3) QU
Lenroot 6/6 ABSENT
McCormick JUN (2)
UJN (5)
UQ 2 to 5
McKellar NJU (3) NJU (3) QU
McKinley NUJN (cycle) UJN (5)
? UQ ? to 5
McLean UJN (5) UJN (5) UQ
McNary NJU (3) NJU (3) QU
Mayeld NJU (3) NJU (3) QU
Means JUN (2) UJN (5)
UQ 2 to 5
Metcalf UJN (5) UJN (5) UQ
Moses JUN (2)
UJN (5)
UQ 2 to 5
Neely NJU (3) NJU (3) QU
Norbeck 4/6 ABSENT
Norris NJU (3) NJU (3) QU
Oddie JUN (2)
UJN (5)
UQ 2 to 5
Overman NJU (3) NJU (3)
QU
Owen UNJ (6)
UNJ (6)
UQ
Pepper JUN (2) JUN (2) UQ
Phipps JUN (2) UJN (5)
UQ 2 to 5
Pittman UNJ (6)
UNJ (6)
UQ
Ralston NJU (3)
NJU (3)
QU
Ransdell NJU (3) NJU (3)
QU
Reed-MO 6/6 ABSENT
Reed-PA JUN (2) JUN (2) UQ
Robinson 5/6 ABSENT
Sheppard NJU (3) NJU (3) QU
Shields UJN (5) UJN (5) UQ
Shipstead NJU (3) NJU (3) QU
Shortridge JUN (2) JUN (2) UQ
Simmons NJU (3) NJU (3) QU
Smith NJU (3) NJU (3) QU
Smoot JUN (2) JNU (1)
UQ 2 to 1
Spencer UJN (5)
UJN (5)
UQ
Staneld 6/6 ABSENT
Stanley UNJ (6) UNJ (6) UQ
Stephens 6/6 ABSENT
Sterling JUN (2) JUN (2) UQ
Swanson NUJ (4) NUJ (4) QU ?
Trammell NUJ (4) NJU (3)
UQ 4 to 3
Underwood UJN (5) UJN (5) UQ
Table 15.5. (cont.)
Votes
Wadsworth JUN (2) JUN (2) UQ
Walsh-MA NJU (3) NJU (3) QU
Walsh-MT NJU (3) NJU (3) QU
Warren UJN (5) UJN (5) UQ
Watson JUN (2)
UJN (5)
UQ 2 to 5
Weller JUN (2)
JUN (2)
UQ
Wheeler 6/6 ABSENT
Willis UJN (5) UJN (5) UQ
Notes:
Rules of inference: use all rankings, unless they are inconsistent.
Allow changes between 1,2,3 and 4,5, otherwise assume same rankings among two.
Italics in column 4,5 means inconsistent with 1,2,3,
= ranking borrowed from 1,2,3, by 4,5 or vice versa
= ranking not borrowed from other column

Table 15.6. Summary of Mackies rankings, Muscle Shoals
Group number and label Preferences N,H & R M 1,2,3 M 4,5
1. Anti-Ford Republicans J > N > U 3 4 4
2. Jones group J > U > N 13 18 8
3. Norris group N > J > U 24 26 27
4. Southern Democrats N > U > J 4 4 4
5. Underwood group U > J > N 17 18 29
6. Southern Democrats U > N > J 9 10 9
7. McKinley N > U > J > N 1
Table 15.7. Pairwise comparison matrix, before
vote switch
J N U BC
J 40 +1 48 88 + 1
N 40 34 + 1 74 + 1
U 32 + 1 46 78 + 1
after vote switch
J N U BC
J 41 39 (80)
N 40 35 (75)
U 42 46 (86)
a denitive and signicant example of cyclical voting. What they have
shown is another case of bungled strategic voting.
Shepsle and Boncheks cycles
The cycles mentioned by Shepsle and Bonchek (1997) do not satisfy my
criteria of consideration, that the case both be published and developed.
Their textbook deserves examination, however, since it is based on a
Harvard core course in politics. Shepsle and Bonchek state that Riker is
their inspiration (viii). The rst observation in their volume is about
politicians:
Their sins are routinely depicted; their persons are often held in contempt; and
their actions are regularly alleged to border on the venal, the immoral, and the
disgusting. In nearly every culture politicians are taken as scoundrels of one sort
or another . . . necessary evils at best, but scoundrels nonetheless. (5)
Shepsle and Bonchek declare that the study of politics used to be based on
description and judgment, but has moved on to explanation and analysis.
In their hands it has become scientic (7).
Part II of their volume, on group choice, is a repackaging of Rikers doc-
trine of democratic irrationalism. One empirical example offered in sup-
port of the argument is Rikers (1982) and Denzau, Riker, and Shepsles
(1985) account of the Powell amendment, but students are not informed,
even by citation, of Krehbiel and Riverss (1990) criticismof that account.
Other examples offered in support of the hypothesis are Rikers account
of the slavery issue, and of the Magnuson amendment (which I debunk
in an upcoming chapter), and generally the student is referred to Rikers
The Art of Political Manipulation (1986, also upcoming) which contains
more popular expositions of his cycle claims.
Three examples are offered to establish the existence and importance
of cycles on redistributive questions (Shepsle and Bonchek 1997, 60
62). One paragraph alleges that there was a cycle over a wealth tax, a
land tax, and no tax in the 1861 US House of Representatives, which led
to adoption of an income tax which was preferred over the three cyclic
alternatives. For details of the case we are referred to Alt (1983), but
Alt (1983) contains no mention of the case, let alone of any details. A
second paragraph tells us that scholars have identied majority prefer-
ence cycles in Depression-era revenue bills. The scholars, we learn in
the footnote, are Blydenburgh (1971), who thought he found one cycle,
and, as we have just seen, examination of the Blydenburgh article would
show that the one claim is confused and invalid. Blydenburgh looked for
cycles on revenue bills brought to the oor under an open rule. Shepsles
intellectual program is that the institutions of Congress function to over-
come what would otherwise be pervasive disequilibrium; the open rule on
the oor is in disequilibrium, but the closed rule (no oor amendments
allowed) arbitrarily imposes an equilibriumby forcing an up or down vote
on the committees proposal. Here he argues that Blydenburgh (1971) is
evidence that members of Congress agree in advance to impose insti-
tutional restrictions on one anothers legislative rights . . . in order to avoid
preference cycles (Shepsle and Bonchek 1997, 61, emphasis added). The
argument is purely functionalist, however, as there is no evidence that
anyone intentionally designed a Congressional institution for the pur-
pose of avoiding preference cycles. A third paragraph argues that in the
Tax Reform Act of 1986 it was possible for there to have been a third
alternative that would have created a cycle, but such an alternative was
not offered for consideration. Such is the state of the evidence.
In conclusion, the student is told that the literature on social choice is
quite sophisticated and covers, in an entirely more analytical style, much
of the same ground as the more qualitative work on democratic politi-
cal philosophy (Shepsle and Bonchek 1997, 80). The triing evidence
we have examined is offered in support of the philosophical conclu-
sion that there is no such thing as the public interest: A public has
no identiable interest if its preferences are either incoherent or overly
idiosyncratic . . . the public interest is a normative ideal that cannot be
given concreteness in most real settings (193).
Finnish electoral college
Suppose that Fredonia chooses its prime minister by the following
method. Each citizen writes down one name on a slip of paper, the papers
are thrown into a hat, one is blindly drawn, and the person named on
the slip is declared the winner. Just because the method was used some-
where does not mean that the democratic theorist is obliged to defend
it. Although the Fredonian method is somehow fair, it is not popular,
in that it does not pick the candidate most-favored, in some sense. As a
democratic theorist I would recommend a method that is both fair and
popular, among other desiderata. Earlier, I argued that the most popular
candidate in the 1860 American presidential election was Douglas, and
that the democratic theorist need not defend the clumsy and antimajori-
tarian American electoral college.
From 1925 to 1988, Finland, a semipresidential regime, selected its
president by means of a method which often resulted in mischief and
occasionally in probably nonpopular outcomes. The faults are not due to
a fatal problem in principle with democracy, but are due to the specic
institutional design and perhaps due to the preference proles, I sub-
mit. First, there are usually about six serious parties, and on the basis
of proportional representation (dHondt), 300 electors from those par-
ties are elected to an electoral college. Second, each party may or may
not nominate a presidential candidate in advance of the election. Third,
the election proceeds by modied plurality runoff: at least 151 votes at
any stage wins the election, in the third stage the top two vote-getters
from the second stage face one another; however, parties may withdraw
or nominate new candidates on either the second stage or the rst stage
of the sequence. Fourth, vote is by secret ballot, but individuals never-
theless are sanctioned to follow the party line, preventing counterstrategy
during rollcall, and reducing the effective number of electors to a small
number. Fifth, the electors have one month between their election by the
people and the meeting of the electoral college to collect information
and devise strategies, preference information is revealed in rst-stage
and second-stage voting, and during the college there are negotiation
breaks between stages, allowing for further information collection and
strategy formulation. Sixth, the president leads on foreign policy, has
the formal power to nominate the cabinet, and informal power to nomi-
nate individual ministers; thus, there may be bargaining over policy, and
over party and individual portfolios, in the course of the election. Sev-
enth, Finland had a civil war, red terror, and white terror in 1918 that
killed 1 percent of the population, and this polarized politics in the 1920s
and 30s; in addition to the socialist-nonsocialist cleavage, there was an
uncorrelated cleavage on foreign policy towards the neighboring Soviet
Union (500 times the population of Finland; Finland seceded from the
Russian Empire in 1917), and cleavages between democrats and non-
democrats on right or left, between city and country, and between Finn
and Swede.
Eighth, if preferences aggregated by party are cyclical, it is possi-
ble for those same preferences to be noncyclical if aggregated by indi-
vidual, the more accurate measure. Proportional representation of par-
ties can amplify the differences between parties and mute cross-cutting
Table 15.9. Distribution of hypothetical PR voters
# of Voters: Rank 7 5 2 1
1st A C B B
2nd B A C A
3rd C B A C
Table 15.10. Aggregation of
preferences by individual not cyclical
A B C (BC)
A 12 8 (20)
B 3 10 (13)
C 7 5 (12)
Table 15.11. Aggregation of
preferences by parties cyclical
A B C
A 12 7
B 3 10
C 8 5
preferences of minority members within parties. Each party might pur-
sue the preferences of its median voter, but not of its members not at
the median. The process of aggregation by party discards information
about citizens preferences, and generally the more information is dis-
carded the more the chance of a nonpopular outcome. Suppose that
there is a population divided into three parties, of A-voters, B-voters,
and C-voters. Preferences are as in Table 15.9. This population clearly
favors A > B > C, by both Condorcet and Borda count, as shown
in the pairwise-comparison matrix in Table 15.10. Now suppose that
each party acts on the preference of its median voter. In the exam-
ple, there is unanimity in parties A and C. Party B decides by major-
ity vote that its position shall be the
2
3
majority position of B > C > A.
Although the populations preferences are noncyclical, preferences ag-
gregated at the party level are cyclical, A > B > C > A, as summarized
in Table 15.11. Most of Lagerspetzs estimates are based on the united
preferences of disciplined parties, not on the actual preferences of the
range of members of each party, and of course they are not based on
the range of preferences in the population. As there is some evidence of
cycles in the Finnish electoral college in 1931 and 1937, but weak or no
evidence in settings elsewhere, I suggest that the Finnish electoral college
is an institutional outlier.
The presidential term of ofce is six years, with exceptions in extraor-
dinary circumstances. There were elections in 1925, 1931, 1937, 1950,
1956, 1962, 1968, 1978, 1982, and 1988; in 1994, a newelection method
was adopted, popular vote in a two-stage plurality runoff, the top two
votegetters in the rst stage going to the second stage. Under the old
method, there was a rst-stage majority winner in the electoral college
in 1950, 1962, 1968, and 1978; and the second-stage winner was unam-
biguous in 1982 and 1988. There are problems alleged concerning the
rst three elections, 1925, 1931, 1937, and the 1956 election. Lagerspetz
(1997) claims that there were at least two and maybe three cycles in 1931,
1937, and 1956. I do not read Finnish and cannot access detailed Finnish
political history. Lagerspetz (1993, 1997) cites plentiful source material
on the controversial elections, and the two articles seem to be based on
strong historical research and strategic insight.
In 1925, the Condorcet and Borda ranking of the four leading candi-
dates, working from Lagerspetzs (1997) estimates of electors rankings,
was Relander > Ryti > Suolahti > Tanner. Relander was the nal choice
of the electoral college, but it took strategic voting to get him through
the maze of the modied plurality runoff, according to Lagerspetz. In
1931 there were four major candidates: Kallio (KA), Stahlberg (ST),
Svinhufvud (SV), and Tanner (TA). Lagerspetz (1997, 60) says that this
is the clearest example of cyclical preferences on an important political
issue described in contemporary literature, and if his preference esti-
mates are right, then the claim is correct. From his estimate of prefer-
ences, I calculate that the Condorcet pairwise-majority ranking of candi-
dates was (ST >
167
KA >
211
SV >
151
ST) >
210
TA. The Borda ranking
was KA (554) > ST (526) > SV (450) > TA (270). Svinhufvud (SV),
third-ranked by the Borda count, was selected by the electoral college,
however. What happened? The Social Democratic candidate, Tanner,
was certain to lose, and the Social Democrats, 90 of the 300 electors, de-
cided to seek a victory for their second choice, Stahlberg, the candi-
date of the republican-liberal Progressive Party. They resolved to vote
for Stahlberg on the second round, but also to allocate some votes to
their third-ranked Kallio from the Agrarians, believing (correctly) that
Stahlberg would do better against Kallio than against Svinhufvud. How-
ever, they made their plans public, and this invited a counterstrategy from
the Agrarians, and on the rst round some conservative Agrarians voted
for their second-ranked Svinhufvud in order to avoid their third-ranked
Stahlberg. The Social Democrats withdrew from their plan and in the
second round cast no votes for Kallio, but some Agrarians did vote for
Svinhufvud. That left Svinhufvud and Stahlberg in the third round, and
the Agrarian Party exercised superior discipline over their ranks to elect
Svinhufvud, according to Lagerspetz, by a vote of 151 to 149. It may be,
however, that Svinhufvud was selected by the electors for reasons other
than ordinary preference. The Whites whom Svinhufvud had led in the
civil war were agitating for a rightist coup with someone like Svinhufvud
as charismatic fascist leader. Their actual coup attempt failed in 1932 be-
cause of President Svinhufvuds forthright opposition, and thus Finland
remained democratic. Otherwise, the election of 1931 raises some nor-
mative concern, as two candidates were more favored than the candidate
selected, but there is a remedy: avoid institutional designs such as the
Finnish electoral college, as Finland eventually did. A minimal reform
to the 19251988 scheme would be to ensure organizational and infor-
mational symmetry if each party is equally well-informed or poorly
informed, equally well-organized or unorganized, then strategic voting
wont succeed. Going to direct election in a two-stage runoff helps in that
regard, and also avoids the party-aggregation effect. A more ambitious
reform would be to go to direct election by single transferable vote. That
would far reduce strategic errors, widen and deepen the consideration of
preferences below the rst, and more likely yield centrist outcomes.
The same candidates were in play in 1937. Party preferences are plain.
1
The collective ranking was probably (KA>
189
SV>
162
ST>
158
KA) >
205
TA, a cycle. The Borda count is Kallio (536) > Stahlberg (501) >
Svinhufvud (478) > Tanner (285). This time the Social Democrats were
determined to prevent reelection of their last-ranked Svinhufvud, and re-
solved, informing only allies, to vote for their second-ranked Stahlberg
on the rst round, and failing that, their third-ranked Kallio on the sec-
ond round. They carried through and on the rst round, Stahlberg got
150 votes, one short of victory, and with votes fromthe Social Democrats
Kallio won a majority against both Svinhufvud and Stahlberg in the sec-
ond and nal round.
If there were the same cycle in 1931 and 1937, but choice of different
winners, then this would be a nice illustration of Rikers hypothesis, al-
though limited to the particular institutional setting. We can take some
comfort from the fact that the Borda winner won the 1937 election, but
that could have been an accident as the Borda winner did not win in
1931 or 1956. Agrarians were median on economic policy and at some
point after World War II were median on foreign policy. This large centrist
party was included in 54 out of 64 parliamentary coalitions through 1988,
compared to 33 for the Social Democrats, and 26 for the Conservatives.
In the 1937 presidential election Social Democrats cooperated in elect-
ing the Agrarians Kallio, initiating the most frequent coalition pattern,
Red-Earth, over the following 60 years.
In 1956, six parties were elected to the electoral college, each with
a nominated candidate. Parties were allocated electors as follows 7,
20, 56, 57, 72, and 88 no party had a majority. The three main can-
didates were Tuomioja for the Conservatives (TU), Fagerholm for the
Social Democrats (FA), and Kekkonen for the Agrarians (KE). From
Lagerspetzs estimates of party preferences, Tuomioja would be the bare
Condorcet winner, TU >
151
KE >
151
> FA, the strong Borda winner,
TU(364) >FA(226) >KE(207), and the standard plurality runoff win-
ner against Kekkonen by a vote of 151 to 149; thats including votes but
ignoring candidates of the three smaller parties. After the rst ballot, can-
didates from the three smaller parties were dropped, and the incumbent
but aged President Paasikivi (PA) substituted for Tuomioja. Paasikivi
was a bit stronger, by Condorcet, PA >
156
KE >
151
> FA, by Borda
PA (384) > KE (295) > FA (221), and by standard plurality runoff
against Kekkonen by a vote of 156 to 144. Paasikivi was in ill health,
however (he would die in 1956), otherwise he would have been a sure
winner, according to Lagerspetz. Kekkonen, the candidate most likely to
continue Paasikivis foreign policy, won the election.
Kekkonens victory was probably due to informational and organiza-
tional asymmetry: the Communists voted strategically but the Social
Democrats failed to countervote strategically. On the second ballot,
the Communists split their vote between their most-favored candidate,
Kekkonen, and their least-favored candidate, Fagerholm, so that on the
third and nal ballot Kekkonen would face weaker Fagerholm rather
than stronger Paasikivi. Kekkonens Agrarians carefully estimated all
300 votes, and proposed the maneuver to the Communists, who car-
ried it out with discipline and secrecy, according to Lagerspetz. If the
Social Democrats had countered with some votes away from their own
Fagerholm to Paasikivi on the second round, then the Agrarian-
Communist maneuver would have failed and CondorcetBorda winner
Paasikivi would have won. Lagerspetz suggests that the most likely ex-
planation has to do with internal politics of the Social Democratic Party.
The large Tanner faction in the party probably favored Paasikivi over
Fagerholm and his faction. Another SDP leader, Penna Tervo, may have
persuaded the party to precommit to backing Fagerholm to the end os-
tensibly as a matter of party unity or perhaps as a way of forcing other
parties to coordinate on its candidate. Further, says Lagerspetz, Tervo
was on bad terms with both Tanner and Fagerholm, and may have favored
Kekkonen over the candidates of his own party. Thus, it is believed that
Tervo may have gotten his way by tricking his party to vote for its most-
favored candidate, resulting in the victory of its least-favored candidate!
The contest between Paasikivi and Kekkonen was close, as was the
contest between Kekkonen and Fagerholm. Lagerspetz estimates that
the 72 Social Democratic electors ranked FA > PA > KE. Suppose
that just seven of those Social Democrats instead favored Kekkonen
from the centrist Agrarians over Paasikivi from the Conservatives.
Then the Condorcet collective preference of the body would have been
KE >
151
PA >
226
FA (it would take 45 such reversals to make Kekkonen
the Borda winner, however). Another way of looking at the outcome is
as a step towards consolidation of a Red-Earth coalition. If the Agrarians
(Kekkonen) and the Social Democrats (Fagerholm, Tanner, Tervo) had
enough in common to coordinate on domestic policies, then together they
would outweigh the various conservatives backing Tuomioja or Paasikivi.
Also, perhaps coordinating on the Agrarian candidate with his promise of
continuing Paasikivis realistic attitude towards the Soviet Union fortu-
itously would be more attractive to other parties than coordinating on the
candidate from the Social-Democratic Party which generally was hostile
to the Soviet Union.
Kekkonen was reelected four times and served 25 years as President,
suggesting that somehow the policy choice was not an unstable one. In
1966 he allied his centrist Agrarians with an SDPwhich had become more
moderate about the Soviet Union and with the Communists to form a
centerleft coalition that brought neocorporatist consensus to Finnish
politics. At the same time social cleavages between working class and
middle class, and Finn and Swede, became less salient, and the Commu-
nist vote slowly receded. Perhaps after its civil war andbefore WorldWar II
Finland was less stable, with each interest pulling as hard as it could in its
own direction, and perhaps wartime solidarity and growing experience
after World War II led to moderation and more stability notice that in
this account stability is a function of variable preferences rather than of
the constant of the aggregation rule.
Finnish democracy has a remarkable history. Over almost a century, a
democracy of a few million people survived civil war and terror, a large
Communist movement, Fascist agitation, the Depression, great-power
pressures from the Nazis on one side and the Soviets on the other, years
of ferocious warfare in the 1940s, and abandonment to the Soviet sphere
of inuence after the war, to become one of the most free and prosperous
countries in the world. In foreign policy it managed to avoid being swal-
lowed by giants, and in domestic policy it achieved a strong economy and
Table 15.12. Iowa Senate preferences,
anticorporate farming
Number Preference orders
of legislators A > B > C > D
5 OCRN
2

OCNR
1

ORCN
4 CORN
2

CNOR
6 RCNO
7 NRCO
Note:
= not single-peaked
welfare state. Could it be that these policies were not chosen, but rather
happened by random spin of the wheel?
Corn in Iowa
Gross (1979) presents the best evidence for any cycle claim. He has from
27 senators their stated rankings over four alternatives, he has actual
votes from the 27 over three of the possible six pairs, he has explanations
from each strategic voter about why her vote differed from her stated
preferences, and he shows a cycle. The Iowa House passed a bill banning
corporate farming in the state and sent it to the Senate (O). The Senate
Agricultural Committee weakened the bill by dropping the corporate
farming prohibition but including a restriction of vertical integration in
the livestock feedlot industry (C). A coalition of senators who actually
preferred no bill proposed an amendment to C which required only that
corporate farms report annually (R). Favoring no legislation on the topic
is labeled N. The distribution of sincere preferences, as stated by senators,
is displayed in Table 15.12.
By pairwise majority voting this aggregates to C > N > O > R > C, a
cycle. The Borda count yields 52 for C, 43 for R, 34 for O, and 33 for
N. If voting had been sincere, then on the rst vote R > C, 14 to 13, on
the second vote O > R, 14 to 13, and on the third vote, N > O, 15 to 12.
The actual vote was R > C, 14 to 13, R > O, 14 to 13, and R > N, 23 to
4. There was strategic voting on the rst vote which did not change the
outcome. There was strategic voting on the second vote which did change
the outcome: the one N>R>C>Ovoter strategically cast a vote O>R,
believing that if O won it would be defeated by his or her rst-ranked N,
and two of the C > O > R > N voters, future conference committee
members, voted R > O, believing that otherwise N would lose to O and
that Senate selection of R would result in a conference compromise near
their most-favored C. R won the third vote and went to conference. Gross
does not report further events.
The issue runs along a single dimension, from stronger to weaker reg-
ulation of corporate farming: O > C > R > N. Thus, it is curious that
collective preferences are cyclic under pairwise majority voting, and sug-
gesting that the reportedly sincere preferences are actually a mixture of
sincere and strategic preferences. Some of the individual rankings are not
single-peaked. Two senators reported O > C > N > R. I can make sense
of that ranking: perhaps they viewed alternative R as so insulting that
they would rank it below no action, or perhaps they wanted to deny N-
dominant senators the opportunity of passing a phony corporate-farming
bill. Two senators reported C > N > O > R. Again R is out of place,
and again the same two explanations might apply. One senator ranked
O > R > C > N: strong regulation over merely symbolic regulation over
weak regulation over no regulation. I cant make sense of that ranking,
although that could be because I lack enough imagination. If that sena-
tors ranking is straightened out, it becomes O > C > R > N, and then
the collective cycle disappears. If all four non-single-peaked voters are
made single-peaked by forcing R to where it would be on one dimension,
then the collective ranking becomes C > R > N > O. Grosss article is
about how anticipated conference committee action can affect voting
during the initial stages of the legislative process (1979, 79). It is possi-
ble, and I think likely, that some of the sincere preferences reported by
the senators were sincere with respect to their own chamber, but strate-
gic with respect to interaction with the House; and the House might have
passed a more extreme measure than it actually favored, anticipating that
a more conservative Senate would moderate the measure. Senators were
also strategic with respect to action in a future legislative session with an
altered conguration of members. All seven of those who stated that they
ranked N rst, for example, voted for R and against N on the last vote.
Gross explains that these seven votes can be seen as acceptance . . . of a
minimal bill in order to avoid stronger legislation in the future (1979,
92). I conclude that a cycle in sincere preferences is unlikely.
Ghosts in Denmark
The 1994 general election for the Danish Parliament was expected to
be very close, among the ideological camps, and among the gures most
likely to become prime minister, according to Kurrild-Klitgaard (2001a).
Table 15.13. Cycle, Danish prime minister
Hans Uffe Poul (Borda)
Hans 39 42 (81)
Uffe 38 47 (85)
Poul 47 45 (92)
Polls were regularly taken by the Danish polling company GfK, including
how respondents ranked the likely prime ministerial contenders pair-
wise against one another. The poll of May 14, 1994 found that Hans
Engel was favored over Uffe Ellemann-Jensen, 39 percent to 38 per-
cent (23 percent dont know); that Uffe Ellemann-Jensen was favored
over Poul Nyrup Rasmussen, 47 percent to 45 percent (8 percent dont
know); and Poul Nyrup Rasmussen was favored over Hans Engell by
47 percent to 42 percent (11 percent dont know). There is a cycle:
Hans > Uffe > Poul > Hans. Of course the Condorcet order, but also
the Borda count, remains the same if we were to exclude nonvoters and
renormalize pairwise votes.
There are several problems with the supposed cycle. First, the citi-
zens of Denmark do not elect a prime minister, rather, they elect parties,
which form coalitions, and the governing coalition selects the prime min-
ister. There were no cycles over the nine parties in the 1994 election,
according to the Danish National Election Survey (Kurrild-Klitgaard
2001a). Pouls Social-Democratic Party was ranked rst, Uffes Liberal
Party was ranked second, and Hanss Conservative Party was ranked
third in that survey. Second, a number of polls were taken through-
out 1994 by the polling company GfK, and only the one we have un-
der discussion showed a cycle. The cyclic poll was in May, the election
was in September. Third, the cycle is statistically fragile, the compar-
ison between Hans and Uffe is almost equal, as is that between Uffe
and Poul. Kurrild-Klitgaard says consideration of sample error cuts both
ways, as there were two other of the polls that would become cycles
if we made small changes in the responses. Fourth, if this is a cycle it
is a balanced one, and arises because the three candidates are almost
tied four months before the election. Just as there is nothing norma-
tively devastating about ties between two leading candidates, the same
goes for a balanced cycle among three candidates. We would worry if
there was an unbalanced cycle that made it possible for an unpopular
candidate to win. Fifth, data from the Danish Election Survey for 1994
on evaluation of party leaders show a transitive ranking of nine party
leaders, and the rst three are: Hans > Uffe > Poul (Kurrild-Klitgaard
2001b). The source does not report when in 1994 the respondents were
polled, but does mention that the race was quite close. Presumably, rank-
ings were volatile in a close race over several months of polling. Sixth,
the Borda count, which allows us to extract an approximation of in-
tensities out of the May GfK information, would rank the candidates:
Poul > Uffe > Hans. This is identical to the ranking of the parties of
the three candidates in the Danish National Election Survey, the Social
Democrats in rst place. Seventh, the Social Democratic Party led in the
actual election and formed a minority government returning Poul the
incumbent as prime minister. Eighth, in related work Kurrild-Klitgaard
(2001b) shows an almost total absence of cycles in twenty years of Danish
National Election Survey data.
These data probably deserved publication, because even the echo of a
rumor of a ghost of a cycle is rare. Evidence such as this is not sufcient,
however, to justify a nding that democracy is arbitrary and meaningless.
What-if cycles and coffee-break cycles
A number of cycle reports in the literature are what we might call what-
if ? cycles. The researcher observes some preference rankings over several
alternatives, and remarks that if unobserved rankings were in a certain
pattern, then there would be a cycle. The point is to illustrate a possibil-
ity rather than make an empirical claim, but such articles are later cited
incorrectly as empirical evidence of cycling. The earliest entry in this
genre is Farquharsons (1969, 5253) account, in his Theory of Voting,
of a 1955 US Senate vote. According to Farquharson, Senator Al Gore,
Sr. presented a proposal for road funding, with a DavisBacon provi-
sion for protection of local wage standards, B. Gores side later moved to
strip the wage-standards requirement, in order to draw more votes from
southern Democrats, C, and this motion succeeded, C > B. Suppose
that A is the status quo of no roads bill. A motion to kill Gores amended
bill by sending it back to committee failed, 50 to 39 against, and thus
C >A. Farquharson says that preference orders were, plausibly: north-
ern Democrats B > C > A, southern Democrats C > A > B,
Republicans A > B > C, one of the two cyclical proles. Accord-
ing to his suggested prole, on the rst vote, C > B, because only the
northern Democrats had an incentive to vote strategically, then C > A.
Farquharsons cycle requires that Republicans most favored no spending
(A), next favored spending at high wage standards (B), and least favored
spending with low unregulated wages (C).
True, as mentioned in Farquharsons account, Davis and Bacon
who originally sponsored the wage-standards legislation in 1931 were
Republicans, but that was at the onset of the Great Depression. In 1947,
Republicans passed the antilabor Taft-Hartley Act over Trumans veto,
and in 1959 the antilabor Landrum-Grifn Act. Thus, it is less plausi-
ble that in 1955 most Republicans preferences were antispending and
prolabor A > B > C and more plausible that they were the antispending
and antilabor A > C > B, and therefore that the collective ranking was
the noncylical B > C > A. If so, then on the vote between B and C,
the Republicans would be tempted to vote strategically for B in order to
get their rst-ranked A, but would refrain, for if they did the northern
Democrats would counter with a strategic vote for C which would avoid
their last-ranked A and preserve their second-ranked C.
One way to decide between Farquharsons interpretation and mine is to
examine the record. The minority Republican report from the subcom-
mittee of jurisdiction opposed Democrat Gores measure and proposed
a substitute that would accomplish Republican President Eisenhowers
road program. The minority report specically objected to the Davis
Bacon provision, with about 1,200 words of argument against it (Con-
gressional Record, May 20, 1955: 6724). Republicans Prescott Bush of
Connecticut, Martin of Pennsylvania, and Cotton of New Hampshire
signed the report, and sought to move their broad substitute, which was
defeated. Thus, evidence from the record suggests that the Republicans
ranked A > C > B, and therefore the conclusion that there was no cycle.
Gaubatz (1995) claims that public opinion is indeterminate in recent
US policy debates on military intervention overseas, and that this inde-
terminacy is due as much to intransitivities in collective preferences as to
the technical difculties of polling or the complexity of the issues. Citing
Riker, he relates that it is unlikely that aggregations of opinion will re-
ect a notion of democratic preference in any philosophically acceptable
way (540). According to Gaubatz, Hinckley proposed that public opin-
ion on US foreign policy sorts into three dimensions, which results in
six types of respondents: unilateralists who favor force (18 percent); uni-
lateralists who oppose force (7 percent); multilateralists who favor force
(26 percent); multilateralists who oppose force (19 percent); isolationists
who favor force (13 percent); isolationists who oppose force (17 percent).
Hinckley identies four possible courses of action withdrawal or doing
nothing (W), multilateral sanctions (S), unilateral military intervention
(U), multilateral military intervention (M) then proposes an ordering
of the four actions for each of the six types. Pairwise majority aggregation
of Gaubatzs proposed orderings results in multiple cycles.
Gaubatzs proposedorderings are not basedonany evidence about peo-
ples actual preferences over foreign-policy choices, but rather entirely
on his intuitions of plausibility. It is easy, however, to propose equally
plausible orderings that do not result in any cycles. For example, one
tiny change in his orderings eliminates all cycles, as he concedes (548).
He claims that forceful isolationists would favor multilateral intervention
over multilateral sanctions; but if they didnt, then no cycles. Thus, his
exercise must be seen as the illustration of a logical possibility, and not,
as he claims, a demonstration that these questions are not simply aca-
demic or restricted to highly articial and constructed examples(540).
Incidentally, there are errors of exposition worth noting. His Table 4
(546) is a pairwise-comparison matrix, but inverted such that the winner
is stated column against row rather than row against column, and thus
may confuse readers accustomed to the row-against-column convention.
Further, Table 4 says M> W, 51 > 49, but from his proposed rankings
it should read W > M, 56 > 44. His Figure 1, illustrating intransitive
policy paths, is mistaken. The gure repeats the error in Table 4, and is
confusing in that an arrow pointing into an alternative indicates its supe-
riority (his statement that removing sanctions from consideration would
not eliminate the cycles is based on the incorrect gure; actually, re-
moving S results in noncyclical W> U > M). A corrected gure would
show three cycles: S > W > M > S, M > S > U > M, and W > U >
M> S > W. Note that M> S is involved in each cycle; reversing that to
S > M eliminates all cycles and yields the Condorcet collective ranking
S > W > U > M. The Borda collective ranking, both before and after
reversing M > S, is S > W > M > U; Gaubatz objects that the absence
of cycles depends on a shift in the relative preferences of 13 percent of
the population over their two least favored options (549), but the Borda
results allay such concerns. He also objects that if S were the Condorcet
winner, still it would be the rst choice of only 19 percent of the pop-
ulation (548), but this is to enshrine the plurality rule as our standard
of choice among many alternatives, which we have seen already is not
well-advised.
The puzzle Gaubatz seeks to explain is ndings such as these: 70 per-
cent of respondents agreed that the US should take all necessary ac-
tion, including the use of military force, to make sure Iraq withdraws its
forces from Kuwait; 45 percent that the US should engage in combat
if Iraq. . . refuses to leave Kuwait; and 32 percent that the US should
increase the level of its troops to force Iraq to leave Kuwait (542).
The conventional explanation is that responses differ according to sub-
tle differences in question wording. Note that the variation on the three
questions is not along Hinckleys three dimensions; they do not variably
appeal to the six different types. Thus, Gaubatzs alternative explanation
seems to lack relevance, for the example he offers anyway. He would have
been more persuasive had he offered an example of questions varying
across Hinckleys dimensions. Apublic opinion survey typically measures
single attitudes, often measures the strength of those attitudes, and sums
responses on single attitudes; and sometimes responses undergo various
statistical treatments suchas multiple regressionor factor extraction. Pair-
wise majority voting ignores strength of individual attitudes, compares
alternatives pair by pair, and chains the results. These are quite different
exercises. Gaubatz urges public-opinionsurveyors to move frommeasure-
ment of lone attitudes with its attendant problem of implicit alternatives,
variable across respondents but unmeasured, to explicit alternatives and
measurement of attitudes relative to one another.
If that were done, then an opinion surveyor could aggregate individual
rankings into a collective ranking. Such an aggregation, if carried out by
pairwise majority rule, could conceivably result in cycles. But why should
the researcher limit herself only to pairwise alternatives and aggregation
by pairwise majority voting? Why not use the Borda count, if the data
are of only ordinal quality? There is no hazard of strategic voting. Or,
if data can legitimately be construed as cardinal, and each respondent
counted as one, then a cardinal summation of attitudes could be carried
out. An agnostic researcher might report aggregation results by several
methods (I believe the reasonable methods would tend to converge). It
is only dogmatic insistence on the IIA(A) condition that limits one to
pairwise majority voting, introduces the possibility of cycling, and leads
to declarations that democracy is meaningless. Gaubatz goes further and
recommends that institutions be crafted that would allow elites to shape
the agenda in ways that overcome or even exploit public intransitivities
(553). Arguably, democratic leaders should have some latitude in devis-
ing foreign policy, and might justiably carry out policies that receive
retrospective rather than prospective approval. But Gaubatzs direct call
for antidemocratic manipulation goes beyond such considerations. Such
counsel is especially disturbing in that it ultimately relies on the normative
claim that the preference ranking of any two options should be indepen-
dent of the inclusion or exclusion of any third option (538). But that
contraction consistency condition, IIA(R-M), leads to possibility, not to
Arrows impossibility. It is IIA(A), that the preferences over two alterna-
tives shouldnt change if preferences over any third alternative change,
that drives Arrows impossibility result.
Rose-Ackerman (1995) scrupulously reiterates that hers is a hypothet-
ical exercise. Thus, the article should not be faulted. She examines the
1993 International Olympic Committee vote on choice of location of
games in 2000 (Sydney won). The IOC proceeds by alternative vote: on
each round the location with the least votes is dropped. She has some
rankings revealed by votes, and assumes others. Crucially, for the 32 of
the 88 voters who voted for Beijing, she assumes that their further ranking
was Manchester > Berlin > Sydney > Istanbul. If so, then by pairwise
majority voting there are three cycles, including one among all locations
except Istanbul. Beijing, not Sydney, would win by the Borda count,
and by several other voting rules. I suggest, however, that the 32 voters
who ranked Beijing rst, may have further ranked Sydney > Berlin >
Manchester > Istanbul. Then there would be no cycles, Sydney would
win by pairwise majority vote, and by Borda count, as it did actually by
alternative vote.
As I have explained my work to political scientists in recent years, I have
encountered many of what we might call coffee-break cycles. I am told,
for example, that departmental hiring decisions are collectively irrational,
but interlocutors neglect alternative explanations, such as the use of near-
unanimity rule in many places, and neglect to consider that full rankings
typically are not known. I am told that researcher X found a cycle, but
whenever I track down X I nd out that she supposed preferences were
such and such in order to illustrate the concept of cycling to students,
or was looking for a cycle and didnt nd one, or that the story is utterly
garbled. In Australia, a half-dozen people told me that researcher Y had
discovered that the frustrating defeat of the republic referendum was
due to cycling disclosed by survey research data. Wrong. Many people
knew that Y looked at survey data on the question, few knew that her
ndings were negative (and thus unpublished), and one insisted that I
must have misunderstood what she told me about her results. I have
stopped pursuing coffee-break cycles.
Conclusion
Given the wide inuence of Rikers teachings, the vastness of the political
universe, and the certainty of academic publication for the most trivial
cycling claim, it is remarkable that there are few such claims in the liter-
ature. That the few are almost all wrong is astonishing. One explanation
for the errors is that most cycle claims are from Riker and that he suf-
fered from conrmation bias: he was looking so hard for cycles that he
noticed indications that would support the nding of a cycle but did not
notice indications that would not support the nding. Rikers nding
of a cycle in the Wilmot Proviso is based on gross factual error and
Blydenburghs nding of a cycle in tax legislation is based on gross con-
ceptual errors. Rikers nding of a cycle in the 1860 election is based on
an unwarranted inference of voters rankings of the four candidates, and
in the Constitutional Convention is based on misspecication of the alter-
natives. Otherwise, the major source of error is confusion about sincere
and strategic voting. Rikers study of agricultural appropriations in the
US House of Representatives, and Neufeld, Hausman, and Rapoports
study of Muscle Shoals each nds a cycle in expressed preferences, each
concedes that there was a bloc of strategic voters, for each we can easily
conclude that sincere preferences were in equilibrium, and in each the
Condorcet winner was selected. Some of Rikers versions of the Powell
amendment, his account of the Depew amendment, and Bjurulf and
Niemis Swedish telephone and telegraph cycle each inadvertently as-
sumes irrational voters who fail to vote strategically when they should.
I suspect that most future cycle claims will arise from cases that involve
confusion between sincere and strategic voting.
16 New dimensions
Introduction
We have been through a lengthy odyssey, more than a dozen case studies,
each an island of terrors and delights. Nowas we approach Ithaca it is time
to recall why we started our journey. The Arrow theorem disclosed the
logical possibility of a majority cycle, of perpetual political instability. But
we observe stability rather than instability in democratic politics. Riker
(1958) initially responds that cycles are common but rarely detected.
Simulations show and empirical studies corroborate, however, that cy-
cling is an empirical improbability. Riker then concedes that uncontrived
cycles are quite rare, but that on major issues actors will contrive cycles
by introduction of new alternatives. Actors also engage in strategic vot-
ing, agenda control, and the introduction of new dimensions in order to
contrive multidimensional disequilibrium, according to Riker.
I object that he must show that manipulation is frequent, harmful, and
irremediable. Rikers position is that it is either theoretically impossible or
empirically difcult to detect such manipulation. He is able, however, to
produce spectacular anecdotes that showharmful manipulation on major
issues, including a demonstration that the biggest event in American his-
tory was the consequence of a contrived cycle. We have worked through
each of his examples, only to nd that each is mistaken, and thus Rikers
case fails on its own terms. In summary, theoretical considerations about
the distribution of preference orders suggest that cycles are most unlikely;
empirical studies show that cycles are of no practical importance; nally,
almost every developed and published example of a political cycle has
now been refuted. Thus, after fty years of scholarship, from the rst
publication of Arrows theorem, almost no one has satisfactorily demon-
strated the existence of a normatively troubling cycle in the real world.
Over time, Rikers argument increasingly came to rely upon the notion
of creating disequilibrium, and thereby turning losing alternatives into
winning alternatives, by the introduction of new issues and dimensions.
We have already examined his major empirical illustration of the notion
378
New dimensions 379
in the history of the politics of the antebellum period. Now we try to
understand exactly how this mechanism is supposed to work. Almost
everyone in politics is a loser in some respect, so why dont we observe
millions of attempts to introduce newissues and dimensions? There must
be constraints on such introductions. The constraint is that the speakers
introduction of a new issue or dimension must be freely rejected or ac-
cepted by the listeners. Deliberation is not subject to disequilibrium,
and can indicate central outcomes such as the intersection of medians
in multidimensional issue space; disequilibrium is a consequence only of
the unfair assumptions of the McKelvey voting model.
Further, I remind that the Arrow theorem, spatial voting theory, and
the McKelvey theorem are only models, useful but not true. When a
model does not agree with observations, at some point it is the model
that must go. I continue that democratic discussion is a complement of,
not a substitute for, voting. I reiterate that if there were a problem with
cycles, the problem would be in the preference prole not in the aggre-
gation function. If cycling were a problem, deliberation would usually
increase the similarity among preference rankings on the individual level,
and by appeal to external principles at the aggregate level could rank the
alternatives in a cycle. It is at least as easy to subtract as to add dimen-
sions in public discussion; I provide an illustration of how, by rening
predeliberative preferences, public discussion might eliminate a cycle.
Discussion also permits the identication of the center of opinion and
the organization of strategic voting so as to defeat unfair agenda control.
Introduction of issues and dimensions
Riker attempts to rescue the pervasive-disequilibrium hypothesis from
the improbability objection with the notion that on major issues actors
introduce new alternatives or dimensions in order to generate disequilib-
rium where there was perhaps an equilibrium before. He illustrates the
new-dimensions mechanism with his story of the Depew amendment to
the proposal for the direct election of senators. In one dimension there
was an equilibrium for passage of the constitutional amendment. Depew
supposedly introduced a new dimension of federal regulation of Con-
gressional elections in order to split and defeat the supporters of direct
election. The institution in place required a two-thirds vote for departure
fromthe status quo of legislative selection of senators, the alternative that
Depew favored, so that by contriving a cycle Depew thwarted the major-
ity and determined the outcome arbitrarily in his favor. Riker presented
erroneous estimates of cyclical preferences over three alternatives under
consideration. I showed that there were no cyclical preferences and that
the proposed constitutional amendment for direct election of senators
rather failed for lack of sufcient votes.
Let us linger to consider just how this mechanism of introducing new
dimensions is supposed to work. Alternative B, along a rst dimension,
would win on a straight vote. Amanipulator, Deborah, prefers alternative
Aalong a rst dimension, which is not the median voters preference, and
she introduces a new second dimension in order to contrive a cycle. If
Deborahs favored alternative A is the status quo and if the voting rule
in place is the backward-moving agenda such that the status quo is voted
on last, then A is selected and Deborah need do nothing further. For
Deborah to succeed she needs to know that there are cyclical prefer-
ences in the two dimensions. Also she must somehow have the power to
unite the two dimensions together into one decision and the majority she
thwarts must somehow lack the power to divide the question in order to
consider matters dimension by dimension. Her success is barely plausi-
ble in the real world. If A is not the status quo under a backward-moving
agenda then Deborah must additionally have control over the sequence
of the agenda in order to have A selected and additionally the majority in
support of B must vote myopically against its own interests. This is even
less plausible in the real world. We can imagine a minority actor getting
away with chicanery on a single occasion, but could an actor get away with
this again and again? Why would the thwarted majority let her get away
with it?
On any unidimensional issue there may be a minority that would prefer
a different decision. Say that there are 100 voters as in the US Senate and
a unidimensional issue with a minority of 40 voters. Each of those 40 then
has an incentive to introduce second, third, and nth dimensions in order
to generate disequilibriumand snatch victory fromthe jaws of defeat. If all
it takes to turn a loss into a win is the introduction of a new dimension,
then we should routinely observe the introduction of 40, 80, 120 new
dimensions by potential losers on any issue. That is not what we observe,
however. Therefore, there must be constraints on the introduction of new
dimensions. What might those constraints be? The answer is not clear in
Riker.
In his general treatment of contrived cycles, just as the natural en-
vironment selects among genes, it is the environment of institutions or
constitutional structures (1982, 211) that selects among the newalterna-
tives or dimensions that politicians randomly introduce. Earlier, however,
Riker (1982, 192) held that such institutions themselves are in disequi-
librium and thus subject to overthrow by introduction of new alterna-
tives or dimensions; it is unclear what force selects among institutions.
New dimensions 381
Moving from the general treatment to the illustrative example of the
commercialists in antebellum America succeeding in splitting the agrar-
ians after forty years of effort, this particular outcome was not due en-
tirely to the wit and persistence of the losers (emphasis added, 213).
The outcome was conditioned not just by constitutional structure, but
also by underlying values and the state of the technology and the econ-
omy, he says. These external circumstances as well as the constitu-
tional structure select from among many attempted manipulations only
a few as winners. Why did the commercialists nally succeed with the
issue of slavery in the territories? The issue matured when Whigs had
sufciently frightened Northern Democrats that they too became, defen-
sively, enthusiastic Free-Soilers, as in the Wilmot Proviso episode, the
Free Soil party, and ultimately the Republican party (229). Lets think
about Rikers explanation. It couldnt have been the southern Whigs who
had so frightened the northern Democrats, because the southern Whigs
were not Free-Soilers. It must have been the northern Whigs then. How
did the northern Whigs frighten the northern Democrats? By inducing
irrational beliefs in them? Or were the northern Democrats rationally
frightened? What would they be afraid of? They would be afraid of los-
ing elections. And why would they lose an election? Because a candidate
strong on containing slavery might beat a candidate weak on containing
slavery. And why would one candidate beat another? Because the winning
candidate would receive the votes of those who thought that containing
slavery is important. And why didnt free soil matter before this point?
Because before the voters thought that other issues were more important.
And did northern Democratic incumbents change positions to accom-
modate the changed preferences of their constituents? Not much, we
know from Poole and Rosenthal (1997, 90) that the Congressional issue
space in this period changed mostly due to the replacement of legislators,
indicating that the preferences of the electorate were the main force of
change. And what caused the changing preferences of the voters? Their
replacement by birth, death, and immigration, and the development of
their preferences in response to events and changing circumstances.
Prospective losers attempt to add a dimension to the issue space,
according to Riker (1990a, 51). Some seek to raise dimensions (the
Magnuson example, discussed below), some seek to reinterpret old di-
mensions (the Whigs reinterpreted slavery to be a national issue rather
than a local issue, he says), and some seek to dismiss dimensions (by
claiming, thats not an issue). The mechanism of manipulation is the
displaying of the relevance of a dimension, recalling it from latent stor-
age to the center of psychic attention (1990a, 51). Is it then that the
dimensions are already there in the citizens mind and the politician like
a pianist manipulates the keyboard and the pedals? Can anyone at anytime
add or subtract any dimension? What is the constraint? The constraint
is what Riker is concerned to deny free public reason. Rikers spatial
metaphor is motivated by a doctrine of democratic irrationalism, and is
a cumbersome and misleading translation of what we ordinarily under-
stand as the public consideration of reasons for and against a proposed
action. Although there is no widely accepted account of how persuasion
and attitude change might work, we are able to sketch the preconditions
of public deliberation, among them:
the inclusion of everyone affected by a decision, substantial political equality
including equal opportunities to participate in deliberation, equality in methods
of decision making and in determining the agenda, the free and open exchange of
information and reasons sufcient to acquire an understanding of both the issue
in question and the opinions of others. (Bohman 1996, 16)
As an aside, it is instructive to consider how J.Q. Adams, according to
Riker the master manipulator in the Yankee cause, attacked the slavery
issue. The fanatical abolitionists relied on moralistic aggression, attack-
ing the very character of the slaveholder. Abolitionists were disliked in
the North for their self-righteous extremism, and Adams made plain that
he was no abolitionist. Adams rather, on a regular basis over nine years,
attacked in the US House of Representatives the gag rule that prohib-
ited not only discussion but receipt of petitions on the topic of slavery.
The fully intended exclusion of an issue from consideration is a clear vi-
olation of the deliberative ideal. Northerners were eventually moved not
so much by sympathy for slaves but by the fear that their own concerns
could be excluded from parliamentary consideration. Adams also found
a way to emphasize the exclusion of slaves from the democratic ideal of
free and equal citizenship. He asked the Speaker of the House for a ruling
on whether he could submit a petition, he said perhaps of doubtful au-
thenticity, signed by 22 slaves. The southerners went berserk, and ranted
and raved for a whole week over this evil suggestion. It was proposed to
expel Adams; to take the petition outside the House and burn it; Adamss
request was denounced as an outrage that has no parallel in parliamen-
tary history (Miller 1997, 230236). Adams let the furor ripen for days.
When he did nally rise to defend himself, he explained that he doubted
the petitions authenticity because the slave petitioners had begged for
continuation of the institution of slavery (apparently concocted by an
overzealous slaveholder). That occasioned a further four days of invec-
tive. The southerners defended their cause in the name of liberty and
equality (for example, the principle of slavery is a leveling principle; it
is friendly to equality. Break down slavery and you would with the same
New dimensions 383
blowbreak down the great democratic principle of equality among men,
Virginia Representative Wise, quoted in W. L. Miller 1997, 439). Finally,
Adams was put on trial in the House in 1842 on a motion to censure,
which was withdrawn after his defense evoked an outpouring of pub-
lic support. Again, it was more the slaveholders threat to the principles
of liberty and equality than the issue of slavery that rallied support to
Adams. An important factor in the ending of slavery was the gradual
developmental maturation of the deliberative ideal.
To continue, participants in public deliberation must of course be free
to accept or reject an argument on their own will. An agenda-controller
with appropriate institutional power could theoretically force the joinder
of two or more dimensions so as to create pairwise, ordinal, majority-rule
voting disequilibrium and carry out Rikerian manipulation, assuming no
germaneness rule, no strategic voting, and the other unfair requisites of
the chaos model. Discourse, however, is somewhat different from voting
in this respect. True, an unconstrained monopoly rhetor as in a total-
itarian regime would have unfair inuence on auditors. A rhetor can
argue that a second or third orthogonal dimension is relevant to the topic
of concern, but in a free and equal society the rhetor cannot force the
auditor to accept her argument. The auditor, normatively and practically,
must freely assent to the relevance asserted by the rhetor. That is one gi-
ant constraint on the introduction of new issues and dimensions. When
an auditor does freely assent to the relevance of an orthogonal dimension,
the auditor is not thereby set adrift in a sea of internal disequilibrium. If
making up ones mind does resemble the spatial model of voting, in that
one locates one ideal point on one dimension and another ideal point on
a second orthogonal dimension, and so on a contentious proposition
that I assume only to explore Rikers argument even then one does not
make up ones own mind on the basis of pairwise, ordinal, majority-rule
voting. Ones position would be simply the multidimensional ideal point
at the intersection of the ideal points from each dimension. Thus, no
matter how many dimensions might be involved, the manipulator cannot
generate disequilibrium within the individual.
To repeat, the introduction of new alternatives and dimensions is con-
strained by the consent of the audience. Then, the manipulator can
only succeed in the frictionless, unrealistic world of the McKelvey voting
model, only when granted the openly unfair advantage of being an uncon-
strained monopoly agenda-setter among myopic voters. Riker (1990a,
53) later conceded that, in the real world, agenda-setters ordinarily
do not have control over all amendments [and] there are probably no
direct real-world analogues of this theoretical possibility of manipula-
tion. In that same article he tries to shore up his hypothesis with other
mechanisms: manipulation is possible if manipulators vote strategically
andothers do not (appealing to his mistakenanalysis of the Powell amend-
ment), or if the manipulator is informed about the distribution of prefer-
ences but other voters are not (appealing to his mistaken construction of
the Plott and Levine experiment), or if voters are denied the motion to
divide the question (no example). Notice, however, that for each of these
mechanisms, unfair output depends on unfair input. To conclude, cycles
are naturally rare, and politicians are constrained from arbitrarily intro-
ducing newalternatives or dimensions in order to generate disequilibrium
where there was none before.
Model or reality?
The Arrow possibility theorem assumes among other things that indi-
viduals have complete preferences over all possible states of the world,
that each individuals rankings are transitive, that preferences must be
aggregated in an ordinal pairwise fashion, and that there is no natural
similarity among individuals preference rankings. This may be a useful
approximation of reality recall that Arrowintends his assumptions to be
descriptive but all the same it is a model and not a reality. The model
implies that political decisions would be constantly unstable yet this is
inconsistent with our observations of political life. The model implies
that there is no public good, yet this is inconsistent with our moral intu-
itions. At this point we can either reject the model and its assumptions
as insufciently unrealistic, or we can attempt to account for the dis-
crepant observations and intuitions. There are no knockdown arguments
that determine either choice; it is a matter of judgment, but judgment is
not arbitrary. Those who follow the path of Arrow and Riker attempt to
explain observed stability as a consequence of structure-induced equilib-
ria such as agenda control, but in my view the sand slips through their
ngers, the attempts are not persuasive (agenda control is countered by
strategic voting, proposed structures are not realistic descriptions, and
so forth). I say that it is the model that is wrong, that randomly ordered
preferences are not descriptively realistic and that pairwise comparison
is not normatively compelling. The irrationalists do not because they
cannot say that our observations of stability are an illusion. They do
say, however, claiming the mantle of scientic authority, that our intu-
ition that there is a public good is an illusion. They do not realize that
our moral intuitions are data of a sort that themselves we want to t
into some coherent scheme, that are constrained by considerations of
coherence. Thus they do not notice, that without an idea of the public
good we become unable to say that one political institution is better than
New dimensions 385
another our moral world no longer makes any sense. To give up the idea
of the public good leaves only nihilistic criteria such as superior force or
blind tradition for the evaluation of political institutions. We can accept
the nihilistic consequences, or we can challenge the realism of the model.
To give up democracy, for example, in order to avoid violation of Arrows
condition of the independence of irrelevant alternatives seems to me to
be an absurd bargain.
Much the same goes for the multidimensional spatial model of voting.
The spatial model additionally assumes that the political preferences of
individuals and publics can be represented such that an issue inhabits a
line and that alternative preferences on the issue inhabit points on that
line. Another issue inhabits a second line orthogonal to the rst and al-
ternative preferences on the second issue inhabit points on the second
line; a third issue inhabits a line orthogonal to the rst two, and so on.
Within that multidimensional space an individual occupies an ideal point
and points away from that ideal point are less preferred. The governing
metaphor is inspired by money accounting; an issue is a category of ex-
penditure, and a preference on that issue is the choice of one number out
of some range of expenditure. Say that it is possible to spend from 0 to
100 units on the issue, then one person most wants to spend 22 units,
another person most wants to spend 87 units, and so on; there is another
category of expenditure with a range from0 to 100 units and the rst per-
son most wants 43 units of this issue and the second wants 88 (the model
can be more general than this, I amtalking about its inspiring metaphor).
This way of thinking is probably a useful approximation of reality, a simple
model that provides insights that otherwise would not be available. But
again, it is a model, not a reality. For one thing, the frictionless McKelvey
model tells us that anything might happen, but since chaos is not what
we observe then we must suspect that anything might be wrong with the
model, which motivates the search for amendments to the model such
as permitting the agents to vote strategically which does indeed return
outcomes to the center, and so on. For another thing, although a use-
ful approximation, it is only proposed and not at all established that the
political preferences of individuals and publics are veridically portrayed
as ideal points and indifference contours within a multidimensional issue
space. Psychologically and socially, is this really how people are oriented
to action in the political world? I do not have a better model to offer, but
I do sense that there is much about the development and expression of
political preferences that is not captured by the model, and thus that it
would be a mistake to believe that the model is true rather than just use-
ful. If the model violates observations or intuitions, we must remember
that it may be the model that is at fault.
The concept of issue dimensions is somewhat obscure. It is plain as day
when graphically portrayed by familiar Cartesian geometry. But when
you try to grasp the concept it slips away. How do issues and alterna-
tives along issues emerge? Does a public share a common denition of
issues and alternatives so that they inhabit the same issue space? If they
do, how did that happen? How do we know when something is an issue
and when it is not? Why do alternatives inhabit one issue rather than
another? Does every issue deserve its own dimension? How do you dis-
tinguish one issue from another? What distinguishes an issue from an
alternative? If the distribution of the populations preferences on dimen-
sion Apredicts their preferences along dimension B, then are Aand B the
same dimension or are they different dimensions? Riker (1993, 4) came
to concede that issue spaces tend to be one dimensional over time.
This conforms to my intuitions about politics, but why things should be
this way is not obvious to me. Is every point in multidimensional issue
space feasible? I think not; issues are constrained by communication, by
commitment, and by budget, according to the interesting work of Hinich
and Munger (1997, 193194) on ideology; for example, one cant prefer
that 100 percent of the budget be spent on tanks, another 100 percent
on schools, and a third 100 percent on solar energy. Is it these ideolog-
ical constraints that collapse the dimensions? Is it the party system? Are
preferences rather already nearly unidimensional in voters minds prior to
institutional constraints? The mathematics of the spatial model are nicely
developed, but its interpretation is less developed. There is much yet to
discover. The state of knowledge does not license a doctrine so radical as
democratic irrationalism.
Deliberation and disequilibrium
Many people in political science accept some version of the Arrow and
Riker story about the instability of democratic voting. A few might have
hoped that democratic discussion would solve the formal problems of
democratic voting: deliberation would be a normative substitute for ag-
gregation. But, in the absence of unanimity, the formal problems of
democratic voting remain when democratic discussion concludes. I do
not think that anyone has pointed out that Rikers theory concerning the
rhetorical introduction of new issues and dimensions is in fact an (erro-
neous) theory of deliberative disequilibrium. The irrationalist hypothesis
of pervasive political disequilibrium must be attacked at its source, I be-
lieve, not evaded by vague appeals to democratic discussion. It is better,
I think, to conceive of deliberation and aggregation as complements. The
preconditions of free and equal discussion are much the same as the
New dimensions 387
preconditions of free and equal voting. On the one hand, when discus-
sion ends without agreement, remaining issues are decided by a method
of free and equal voting. On the other hand, problems of unfreedom and
inequality in democratic voting are remedied by democratic discussion,
I shall argue.
Uponreection, the Condorcet paradox is not so surprising: why would
an aggregation function work to reduce widely distributed disagreements?
Voting does not reduce disagreements, it can only register them. Discus-
sion is what reduces (or, unfortunately, increases) disagreements between
people. Why is the Condorcet paradox intuitively disturbing? Because for
a social choice to cycle among A, B, and Cseems simply arbitrary mean-
ingless, as Riker says. But the meaninglessness is prior to the social choice,
it lies in the particular cyclical prole of individual preferences; where
A > B > C, and B > C > A, and C > A > B are together possible. There
is a trivial class of cyclical proles: when the population is nearly indif-
ferent among alternatives. The few cycles found in the empirical studies
were mostly of this trivial variety (alternatives adjacent by Borda count).
Suppose a population involved in many issues, and one of those issues is
the collective purchase of ice cream. If there happened to be a cyclical
prole on this issue Vanilla > Chocolate > Strawberry; Strawberry >
Vanilla > Chocolate; Chocolate > Strawberry Vanilla we would not be
too surprised. The choice is a matter of mere taste, and the choice among
avors is a matter of near indifference compared to the choice between
having ice cream or not. Considered collective indifference is as norma-
tively insignicant as considered individual indifference among alterna-
tives. Then there is a repugnant class of cyclical proles. Pick any three
dissimilar states of the world I suggested previously personal prosperity,
the torture of kittens, and suicidal nuclear war. One voter prefers Pros-
perity > Torture > Suicide, a second Suicide > Prosperity > Torture, a
third Torture > Suicide > Prosperity. There is a cycle among the three
alternatives and a Rikerian manipulator could arrange for his favored
alternative, such as suicidal nuclear war, to win. I have just shown that
majority-rule voting can destroy all life on earth. What is wrong with
this picture is not majority-rule voting but the absurdly unstructured
preference prole. Choice over those three alternatives is not a matter
of mere taste; there are a few freaks in the population, but the huge
majority prefers personal prosperity and that is the choice among the
alternatives it is very easy to understand and explain the preference for
personal prosperity.
Imagine a place where the distribution of preferences over all possible
states of the world is random. One couldnt make sense of the people
in such a place. One could not interact with such people for lack of an
approximation of what other people want. Repugnant cyclical proles
might be possible in rare instances, but they could not be prevalent,
else other minds would be unintelligible, the unintelligibility not follow-
ing from aggregation mechanisms but rather from the obscurity of other
minds desires. The cyclical order of preferences is arbitrary, individuals
preferences are unstructured, not ordered by any external principles open
to public discussion. If the ordering of individual preferences is unstruc-
tured and purely arbitrary, then the ordering of social preferences will be
unstructured and purely arbitrary as well.
Accepting the arbitrary ordering of individual preferences is a conse-
quence of the doctrine of consumer sovereignty or subjective welfarism
from welfare economics. Taking as given the self-regarding preferences
of price-taking individuals and rms, a competitive equilibrium is Pareto
optimal: laissez-faire leads to the common good; and, almost any Pareto-
optimal equilibrium can be attained with taxes and transfers on individ-
uals and rms; but, there is no Arrow social-welfare function (Feldman
1991). Elster argues that social choice theory fails to capture the distinc-
tion between the isolated and private expression of preferences on the
market from the open and public activity of politics:
The notion of consumer sovereignty is acceptable because, and to the extent that,
the consumer chooses between courses of action that differ only in the way they
affect him. In political choice situations, however, the citizen is asked to express
his preferences over states that also differ in the way in which they affect other
people. This means that there is no similar justication for the corresponding
notion of the citizens sovereignty, since other people may legitimately object to
social choice governed by preferences that are defective in some of the ways I
have mentioned. (Elster 1986b, 111).
Elster (1986b) and Sunstein (1990, 1993) categorize various sources of
defective individual preference formation.
1
Rather than emphasize de-
fective individual preference formation to be reformed in principled pub-
lic justication, I will emphasize the continuity between principles of
individual ordering of preferences and principles of public ordering of
preferences.
A competent adult is almost always the best judge of choices that af-
fect only her, on informational grounds alone, if not on larger normative
grounds. From the variety of conicting desires and beliefs, the individ-
ual strives to construct a consistent order of desires, and a consistent
order of beliefs, each as consistent and complete as needed for the con-
duct of life. In experimental situations, individuals often make intran-
sitive choices in single experiments, but in repeated trials their choices
become steadily more consistent (Elster 1979, 154, citing Davidson).
New dimensions 389
When individual inconsistencies are discovered, consistency is regained
by individual deliberation with reference to substantive principles. If we
ask someone why he prefers A >B >C >A, we are likely to get a correc-
tion. We understand immediately why someone likes winning a lottery
ticket better than cleaning the barn better than death by ring squad.
If we ask someone in a harder case why he likes A > B > C, we will
hear understandable (possibly mistaken) reasons; often the provision of
more and more context provides sense to the ranking. Even in the case
of mere tastes, the framing case for welfarism, there is an understandable
and nonarbitrary reason: it tastes better to me. Notice that reasons
of taste are completely convincing in justifying only a portion of pri-
vate choice and are completely unconvincing in justifying most of public
choice. Stepping from private life to public life, to choices that affect oth-
ers, information on desires and beliefs goes from certain to less certain,
the construction of possible alternatives and their consequences is less
certain than in private life, and individual preferences over public states
of affairs are initially incomplete and intransitive. In the process of public
deliberation, from the local assembly to the larger public sphere, individ-
uals gain evidence and principles by which to order their own conicting
desires and beliefs over public affairs. The public existence of evidence
and of substantive principles of consistency induces some similarity in
individual preference orderings over public affairs, because beliefs are in-
ternally related by correspondence to independent objects, and because
principles for ordering desires, although diverse, are on net necessarily
more similar than random.
The skeptic tells us that nothing has changed. Before deliberation there
is a set of individual preference orderings to which certain possibility the-
orems apply. After deliberation there is a new set of individual prefer-
ence orderings to which the same theorems apply. But deliberation may
contribute the minimal consistency in individual ordering probably suf-
cient for consistent public ordering. Moreover, if public inconsistency
remains, just as does the individual when discovering a private incon-
sistency, so does the public body turn to deliberation over substantive
ordering principles to resolve the inconsistency, if it is an important one.
If serious inconsistency remains, the polity is in trouble, but the source
of the trouble is in individuals preferences not in the voting mechanism.
The skeptics trick was to exclude principled consistency by axiom at the
individual level, only to demand it by intuition at the aggregate level.
The construction of alternatives and dimensions is subject to the prin-
ciple of interpretive charity and its constraints of logical consistency
and correspondence of beliefs, and of transitive consistency and simi-
larity of desires. Following Hinich and Munger (1994, 1997), multiple
dimensions collapse as one dimension depends on another; the entire
multidimensional space is not feasible because of the constraints of com-
munication, ideological coherence, and budget. Here is an example of
how deliberation and the budget constraint might collapse dimensions.
At one time in my political life I was much enlightened by a naively
designed public-opinion poll on county services. A huge majority wanted
less money spent on tax collection, a smaller majority wanted more money
spent on parks, and an even smaller majority thought the amount spent
on the countys main business, public safety, was about right. Each indi-
vidual might be perfectly consistent in expressing his or her prepolitical
preferences, but taken together the preferences are incoherent: less money
spent on tax collection would mean lower taxes, and thus lower services.
Suppose there are three alternatives:
r
A = less tax collection
r
B = more parks
r
C = same public safety.
Since there are more than two alternatives, and majorities support every
alternative, there is the possibility of a majority-rule voting cycle. Suppose
that a majority faction of 60 percent prefers more parks to the same public
safety (B > C), but a minority faction of 40 percent prefers C > A > B
and knows that the majority is evenly divided between A > B > C and
B > C > A, so that the minority faction plans to use agenda control
in order to gain a victory for A = lower taxes. The agenda-controlling
minority faction would propose B against C, which B would win (if the
minority faction proposes A against B rst, or A against C rst, it would
end up with its last-ranked alternative B). The minority faction then
would propose B against A. The minority faction of 40 percent prefers A
over B, and half of the majority faction, another 30 percent prefers Aover
B, making a majority of 70 percent for A over B. The cycle and hence the
manipulated outcome would collapse after public deliberation, however,
because of the incoherence of predeliberative preferences. If the minority
faction pressed ahead it would win its second-ranked A, lower taxes, but
this would make impossible its rst-ranked preference, C, the same public
safety. The minoritys inconsistency would not be accepted as a matter of
preference sovereignty, but would be justly ridiculed in debate. If lower
taxes result in less services, then there are only two possible preference
rankings: lower taxes > same or more county services, and same or more
county services >lower taxes. With preferences developed by deliberation
there is no longer a cycle. The critic may object that all Ive done is
manufacture an example where three alternatives easily reduce to two.
But my point is that it is just as easy rhetorically to subtract alternatives
and dimensions as it is to add them, and that the subtraction here is not
New dimensions 391
arbitrary, but due to a coherence constraint discovered in public debate.
To go further, because of the communication, coherence, and budget
constraints on issue space, it is probably easier to subtract dimensions
than to add them.
There is empirical evidence for the proposition that deliberation can
add to stability. Lets call a collection of individuals preference orders that
are unrelated to one another, such as under the impartial-culture assump-
tion, unstructured preferences. Call another collection where there is more
resemblance among preference orders more structured. The more closely
a collection of individuals preference orders approximate a collection of
single-peaked preference orders, the more likely is it that the paradoxes
and instabilities of social choice would be avoided. McLean, List, Fishkin,
and Luskin (2000; see further, List 2002) show that deliberation can add
structure to unstructured preference orders: certain perspectives on the
subject become more salient, and random and ill-formed opinions drop
away. The deliberative opinion poll selects citizens at random, assembles
them, provides neutral information, and encourages public deliberation
over the issues at stake. Attitudes are measured before and after deliber-
ation. Deliberation might improve structure in three ways: rst, it might
increase the proportion of preference orderings that can be arranged
single-peaked, second, it might reduce the proportion of preference or-
derings that do not conform to the largest single-peaked arrangement,
and third, it might reduce indifferent or incomplete preferences among
the population. Seven deliberative opinion polls commissioned by Texas
utility regulators provide data that show deliberation delivering all three
effects. There was not much structure to incoming preference orders in
the Texas case. McLean et al. also examined a deliberative opinion poll
concerning the Australian Republic Referendum. The voters of Australia
were to decide among, (1) change to a republic with president directly
elected, (2) change to a republic with a president appointed by
2
3
vote
of the legislature; and (3) the status quo, subjects of Queen Elizabeth.
There was much structure to incoming preference orders, so structure
did not improve by deliberation. Instead, the outcome changed from
a stable predeliberative social preference for alternative (1) to a stable
postdeliberative social preference for alternative (2). Curiously, alterna-
tive (2) was sent to the people by the legislature, and it failed the popular
vote, suggesting perhaps that popular deliberation over the measure was
incomplete or defective.
Here is another way that deliberation might tame majority-rule voting
disequilibrium, if one considers it a realistic problem. Ordinal pairwise
voting contains no information about the intensity of preferences, and
that is how the possibility of cycles arises. Public deliberation provides
information about the intensity of preferences, and information about
intensity allows for knowledge of the likely center of opinion. It will be
objected that anyone can assert any intensity claim they please in debate,
but those who have worked on committees will understand that the mem-
ber whose positions are increasingly fervent and urgent is increasingly
ignored. One need not resort to a controversial cardinalist framework for
this effect; the relative ranking of ordinal preferences as indicated by the
Borda count is enough. As for multidimensional disequilibrium in the
McKelvey model, again deliberation can disclose the location of the in-
tersection of medians, among other things. That knowledge of the center
of opinion would motivate majorities to resist, with strategic voting or
otherwise, a manipulator aiming for an extremist outcome. Something
like this can be seen when legislative deliberators warn that the pending
amendment is a killer amendment and advise their allies to vote strate-
gically in order to obtain the majority outcome. A deliberative assembly
can also open for public discussion the unfairness of the institution of
an unconstrained monopoly agenda-setter, or any other shenanigans that
violate the principles of freedom and equality. Riker says there are many
transient majorities. If he is right about that, then the members of those
transient majorities all have one thing in common: the demand for fair
treatment.
Lincoln at Freeport
After Liberalism against Populism in 1982 came Rikers The Art of Polit-
ical Manipulation in 1986, intended for a more popular audience. The
text and its examples are used in undergraduate political science courses.
The 1986 volume illustrates Rikers doctrines by way of twelve simple
case studies. Riker coined the term heresthetic, the art of manipulation, in
contrast to rhetoric, the art of persuasion.
2
Rhetoric succeeds by chang-
ing others preferences towards those of the rhetoricians, heresthetic wins
instead by setting up the situation so that others must support the prefer-
ences of the heresthetician. Heresthetic is managing, manipulating, and
maneuvering to get the decisions one wants.
We examine two chapters in depth. The rst is about the Freeport de-
bate between Lincoln and Douglas. Riker believes that Lincolns rhetoric
forced a dilemma upon Douglas in which Douglas was forced to choose
between losing supporters in the North or supporters in the South. Riker
relies on outmoded historical interpretations of the Freeport debate,
however. Modern scholarship does not support these highly dramatized
interpretations. Moving from the empirical to the theoretical, Douglas
did indeed face a dilemma, but it was one that was forced upon him
New dimensions 393
by the changing preferences of the northern and southern populations,
not by Lincolns discourse. The second chapter is about Senator Warren
Magnuson and a controversy over the shipment of nerve gas to the United
States during the VietnamWar. Riker believes that with a rhetorical our-
ish Magnuson added a dimension to the debate that forced certain sena-
tors to vote for an action they opposed and thereby engineered a roll-call
victory for an action supported only by a minority of the Senate. Riker
gets several details of the story wrong and he does not fully understand the
parliamentary situation. The record shows that Magnusons bill would
have passed without fuss, perhaps without even a roll-call vote; the de-
bate was over a more radical version of Magnusons proposal offered by
Gravel. Senators stated that they voted for the more radical version of the
proposal in order to gain the attention of the White House, which had not
been responsive on nerve-gas controversies across the United States, and
that they would withdraw Gravels radical proposal in conference com-
mittee if the White House became more responsive on the issue. Later in
fact the Gravel proposal was effectively withdrawn in conference and the
Magnuson bill effectively retained. Rikers story asserts that Magnusons
rhetoric forced 13 senators to vote against their own beliefs. But 40 sena-
tors felt free to ignore Magnusons rhetoric and voted against the radical
Gravel proposal.
The rst chapter, meant to illustrate the theme of dividing the opposi-
tion with a newalternative, tells the story of a debate between Lincoln and
Douglas at Freeport on August 27, 1858, during their race for US senator
from Illinois. Lincoln asked Douglas: Can the people of a United States
Territory, in any lawful way, against the wish of any citizen of the United
States, exclude slavery from its limits prior to the formation of a state
constitution? (Riker 1986, 12). Much of the story is a recapitulation
of our earlier discussions of the antebellum period. The story continues
that Douglas was running for Senate in 1858 needing to win Illinois but
was planning to run for President in 1860 needing to win the nation. If
Douglas answered yes, that the people of a territory can exclude slavery,
then he would please northern Democrats in Illinois in 1858 but displease
southern voters in the nation in 1860. If Douglas answered no, then he
would displease northern Democrats in Illinois in 1858, repudiate his
long-held popular-sovereignty position, but please southern voters in the
nation in 1860. Douglas answered yes, and won the 1858 election in Illi-
nois, but in 1860 split the Democratic convention and lost the election
in the country. Lincolns great heresthetical maneuver was at a stroke
to consign Douglas to political doom. Riker leans toward but does not
commit to the position that Lincoln foresaw that he would run against
Douglas for President in 1860. Lincolns question was the capstone of
the Republican strategy of splitting the Democratic majority, however,
and such was certainly Lincolns full intention, Riker says. His source for
the nal assertion? A long folksy quotation from a historical novel about
Lincoln published in 1901.
The Freeport Question was once a standard story in American history,
but its melodramatic version has since been dismissed by historians as
folklore. Potter (1976), the author of the standard history of the decade
prior to the Civil War, says that the Freeport question was one of the
great nonevents of American history (338) and introduces the topic
with sarcasm: When folklore appropriates a scene . . . it begins at once,
unfortunately, to improve upon history by adding certain characteris-
tic ctitious touches . . . it dramatizes an ordinary contest into an epic
struggle . . . virtue invariably overcomes wickedness by some simple but
supernaturally effective device a silver bullet, a magic phrase, a sling
for David against Goliath (334). The folkloric version neglects context.
There are several problems. Lincoln was not a national gure and could
not have known that he would be the Republican candidate in 1860, and
could not have been certain that Douglas would be a Democratic candi-
date. Next, Douglas had already addressed the question in a speech in
Springeld, Illinois, on June 12, 1857, 14 months prior to Freeport. He
attempted to avoid the dilemma by afrming that the people of a terri-
tory could decline to enact local legislation necessary to the support of
the institution of slavery (they may or may not have the right to exclude
slavery, but they do not have the duty to enact legislation supportive of
slavery).
Meanwhile, Douglas was occupied with the much greater contro-
versy over the Lecompton constitution, the biggest issue in Congress
that year. In Kansas, free-staters outnumbered slave-staters by two to
one. A proslavery gerrymander, however, put the states constitutional
convention in the hands of the slave-staters, and free-staters boycotted
the election in order to deny it legitimacy. The slave-staters at the
Lecompton convention practiced Rikerian agenda control in designing
a referendum on the state constitution that forced voters to choose be-
tween approving either existing slavery or expanded slavery in Kansas.
Illustrating that ordinal pairwise voting is not the motor force of his-
tory, the free-stater majority in Kansas boycotted the unfair referendum,
and the constitution with expanded slavery was approved by the minor-
ity slave-state voters. Elections to the territorial legislature were more
representative, and its free-stater majority referred three alternatives to
the voters: the Lecompton constitution with existing slavery, Lecompton
with expanded slavery, or rejection of Lecompton entirely. The slave-
staters boycotted this election, and the majority free-staters voted almost
New dimensions 395
unanimously to reject the Lecompton constitution. The conict migrated
to Washington, DC. Democratic President Buchanan made acceptance
of the Lecompton constitution a question of party loyalty, but Douglas
and many northern Democrats would not could not go along. Admis-
sion of Kansas with the slave Lecompton constitution passed the Senate
but failed the House. Buchanan opted for Lecompton because of pres-
sures of the southern majority of the Democratic Party on him. Douglas
came out against Lecompton because its fraudulent birth made a mock-
ery of the popular sovereignty principle that he both seemed sincerely
to believe in and needed politically in order to maintain the Democrats
as a bisectional party. If Douglas had not opposed Lecompton he would
have been dead in Illinois in 1858 and dead in the North in 1860. For
purposes of Presidential election, the Democratic Party at that point had
the South but needed the North.
Before Freeport, the question arose three times in campaign events
in July 1858, and Douglas answered, Slavery cannot exist a day in the
midst of an unfriendly people with unfriendly laws (337). Moreover,
Lincoln in a private letter written on July 31 acknowledged that Douglas
had dodged the question and expressed the opinion that anyway because
of the Lecompton controversy, Douglas cares nothing for the south; he
knows he is already dead there (337). Lincolns purpose in asking the
Freeport question, says Potter, was to return attention to the already es-
tablished fact that Douglas could only reconcile Dred Scott and popular
sovereignty with the lame proposition that constitutionally guaranteed
rights could legitimately go unenforced; I would add that Douglass an-
swer reinforces the suspicion that he was too expedient a character. But
Lincoln did not vigorously pursue the question in any of the ve joint
debates that followed. Lincoln, in fact, did not want to overemphasize
the question of the best policy for containing slavery, according to Potter,
because there was little difference between the practical outcomes of the
candidates policy proposals. Lincoln attempted rather to shift the debate
from the policy dimension to the moral dimension. Lincoln said that the
contest was between the Republicans who considered slavery wrong, and
Douglas and the Democrats who did not consider slavery a matter of
right and wrong. Further, Lincoln did go after Douglas aggressively on
another question more likely to rouse the audience, according to Potter
(1976, 349351). Lincoln warned that the next slavery decision from the
Supreme Court would go beyond Dred Scott to declare that no state has
the right to exclude slavery from its limits. Douglas used his strongest
language in the debates denouncing this charge.
Douglass dilemma was larger than the Freeport question, and it was
not posed by Lincoln. His dilemma was posed by preferences of the
people and by how those preferences developed in response to events.
The Whig Party had already collapsed. The Democratic Party was held
together by the doctrine of popular sovereignty. On March 6, 1857, the
Dred Scott decision ruled that the US Congress lacked constitutional au-
thority to prohibit slavery in the territories and Chief Justice Taney,
although not the Supreme Court, declared that territorial legislatures
lacked authority to prohibit slavery which threatened to nullify the doc-
trine of popular sovereignty. In February and March of 1858 the southern
wing of the Democratic Party further threatened the credibility of popular
sovereignty by attempting to force admission of Kansas with a minority
slave constitution. It was the widening gap between the two sections of the
party that was Douglass problem, not Lincolns discourse. Nor was it the
Republicans who split the Democrats. Proximately, it was the Democrats
who split themselves at their 1860 Charleston Convention, and ultimately
it was because of preference development especially the determination
of the re-eaters in the Lower South to secede. On December 29, 1858,
one southern Democratic politician wrote to another urging that the slav-
ery issue be kept alive. He wrote that such a policy might result in a
Southern party which would either succeed and thus govern the country
or fail and thus form a compact Southern party ready for action [i.e.,
secession] (Nevins 1950, 179). The gap between North and South was
most widened by the Dred Scott decision, which emboldened the South
to take its intransigent stand. There is a hidden functionalism in Rikers
argument. Did the Republicans benet from the split in the Democratic
Party? Yes; indeed, how could they have thrived without it? Did the Re-
publicans try to split the Democratic Party? Yes, just as Douglas tried
to split the Republicans by adhering to popular sovereignty, and just as
Calhoun tried to split the Whigs. Did the Republicans cause the split in the
Democratic Party, however? That would have to be shown. Did Lincoln
cause the preferences of the secessionist gentleman I quoted? Evidence
that Republicans discursively encouraged a split is not sufcient as, ac-
cording to Rikers story, the Yankees had tried that for at least forty years
already without any success. Is it that they had nally found the magic
phrase, or is it that preferences and constraints had changed over the
forty-year period?
Magnuson and nerve gas
This story, say Riker and Weingast (1988, 391),
involves a motion that likely would have failed if unadorned, but that passed
triumphantly when a single legislator embellished it rhetorically with an interpre-
tation that other legislators knew to be meretricious, and that a majority probably
New dimensions 397
opposed. Thus a single person brought about what appears to be a majority-
opposed outcome.
Rikers earlier cycling stories seemed to recede as Riker developed his
notion of heresthetic, which is exemplied by the story of Magnuson and
the nerve gas. Riker frequently retells the story, in Riker (1986), in Riker
and Weingast (1988), and in Riker (1990a). Well focus on the lengthier
1986 version.
In 1970, during the Vietnam War and when Republican Nixon was
President, the US Department of Defense decided that it had to ship
a large quantity of nerve gas from Okinawa in the Pacic Ocean to the
United States. It would be landed in Seattle, Washington and then be
shipped by train to a military depot in the neighboring state of Oregon.
This became the federal issue of supreme concern to the Pacic North-
west of the United States in that period, as I can attest having lived there
at the time. On May 29, 1970, Senator Warren Magnuson, Democrat
from Washington State, offered an amendment to the Foreign Military
Sales bill stating that No funds authorized or appropriated pursuant to
this Act or any other law may be used to transfer chemical munitions
from Okinawa to the United States (Riker 1986, 107). The amend-
ment was cosponsored by Senator Henry Jackson, another Democrat
from Washington, a war hawk and friend of the Pentagon, by the two Re-
publican senators from Oregon, Mark Hateld, a dove on the war, and
Bob Packwood, and by antiwar Senator Mike Gravel, Democrat from
Alaska. Riker says that Magnuson as a Democrat was trying to discredit
the Republican President.
The amendment had been discussed on the oor but had not yet
been voted on when the Defense Department announced that it would
not ship the nerve gas to Oregon, but rather to the Kodiak Naval Sta-
tion in Alaska. Senator Gravel of Alaska on June 29, 1970 moved an-
other two-sentence amendment to the still-pending Foreign Military
Sales bill. The rst sentence of Gravels amendment was identical to
Magnusons amendment. The second sentence of Gravels amendment
read: Such funds as are necessary for the detoxication or destruction
of the above described chemical munitions are hereby authorized and
shall be used for the detoxication or destruction of chemical munitions
outside the United States (Riker 1986, 108). The intent of the motion
was to force the containment and encourage the destruction of the nerve
gas on Okinawa. Magnusons great heresthetical maneuver was in sup-
port of Gravels amendment, according to Riker. Riker relies in part on
Redman (1973, 207), whose reportage he recognizes was quite inaccurate
(Riker 1986, 113; for example, Redman wrongly believes that Senator
Harry Byrd of Virginia is Senator Robert C. Byrd of West Virginia).
Redman (1973, 206207), Magnusons aide at the time, wrote:
I prepared another memorandum cataloguing the arguments he had marshaled
against the shipment and took it to him at his desk on the Senate Floor. He
surveyed the memo cursorily, then handed it back with an annoyed No, no,
no! Bewildered, I retreated to the staff couch and waited to hear what argument
he intended to use instead of the familiar ones of possible sabotage, dangerous
sections of track along the proposed route, and populations that would have to be
evacuated as a precaution against leakage. When the time came, he took a wholly
novel and ingenious approach. The issue, he told his colleagues, was not one of the
people versus the Pentagon, as the news media seemed to assume. Instead it was
another case of the President versus the Senate. The Senator from West Virginia
(Robert C. Byrd) had recently offered a resolution, which the Senate had passed,
stating that the Senate expected the President to keep it informed throughout the
treaty negotiations with the Japanese government on the subject of Okinawa. The
Presidents sudden decision to move the nerve gas off Okinawa must reect some
aspect of those treaty negotiations, Magnuson insisted and the Senate had not
yet been informed of, much less consented to, any such agreement. To allow the
nerve-gas shipment under these circumstances, he asserted, would be to abandon
the Byrd Resolution and to abdicate the Senates rightful role in treaty-making
generally. The President, Magnuson said, might get the idea that he could ignore
the Senate and its constitutional prerogatives whenever he wished. Jolted by this
reasoning, the Senator fromWest Virginia and his Southern colleagues friends of
the Pentagon almost to a man, but vigilant guardians of the Senates constitutional
responsibilities voted down the line with Magnuson. The amendment, which
had been doomed a few minutes earlier, passed overwhelmingly.
A sling for David against Goliath?
The ChurchCooper amendment to the same bill, with the purpose
of forcing President Nixon to withdraw from his Cambodian incursion,
was voted on the next day. For the Gravel nerve-gas amendment, Riker
thinks that Magnuson already had the votes of nine regional senators
from Hawaii, Alaska, Washington, Oregon, and Idaho, but not of Demo-
cratic hawk Jackson of Washington state. Jackson supported the original
Magnuson amendment, but, once the danger to Washington state had
passed, Riker believes, Jackson voted against the Gravel amendment.
He also thinks that Magnuson had the votes of the anti-Nixon forces,
those who voted both for the Gravel amendment and for the Cooper
Church amendment; there were 40 of these, including 7 of the 9 re-
gional senators. That means 40 + (9 7) = 42 natural votes for the
Gravel amendment, out of a Senate of 100 members voting by majority
rule. There were 24 senators who voted against both the Gravel amend-
ment and the ChurchCooper amendment. That leaves 33 senators, 26
of whom were present for the vote on the Gravel amendment. The 26
New dimensions 399
divide into two groups, according to Riker. The rst group of 16 voted
against Gravel and for ChurchCooper, they were not impressed by
Magnusons heresthetic. The second group of 10 voted for Gravel but
against ChurchCooper, they were indeed moved by Magnusons her-
esthetic, even though they probably recognized it for what it was, namely
a tactic to manipulate them (1986, 112). The only warrant offered for
these imputations is the two vote counts. In the end, Magnuson added
10 to his natural tally of 42 to obtain a victory with 52 yeas and 40 nays,
the story goes. This is heresthetic at its best (1986, 112). Magnuson
did not persuade, I think, but maneuvered so that those who would have
lost in one dimension won in two (1986, 113).
Alas, Riker detected some errors in Redmans bewildered account but
he did not detect them all. The story manifest in the Congressional Record
and The New York Times is not the one that either Redman or Riker
tells. Redmans account is regrettably nave and starry-eyed according to
Riker (1986, 108), and these traits do probably contribute to Redmans
errors. Rikers account, however, is cynical and murky-eyed and these
traits probably contribute to his. Lets start all over again.
The US was at war in Vietnam. In 1968 thousands of sheep died in
Utah due to a military nerve-gas accident (Congressional Record, May 21,
1970: 16482). This was an unnerving incident, the eerie pictures of dead
sheep were known and discussed quite widely in the West, more than
many other political issues, I recall. The Pentagon announced contro-
versial plans to ship deteriorating nerve-gas materials, 27,000 tons, from
Colorado, Alabama, and Kentucky to a New Jersey port for burial at sea.
Japan had been pressing for the return of the Ryukyu Islands, including
Okinawa, controlled by the US but home to a million Japanese subjects
(The New York Times Index, 1969, 1,279). There were mass demonstra-
tions, electoral agitation, and minor riots in Japan and in Okinawa. One
obstacle to returning control to Japan was the need to remove US nu-
clear weapons from Okinawa. In the midst of this tension came a news
report that 25 American military personnel in Okinawa were hospital-
ized after accidental discharge of toxic gas. The presence of the nerve gas
became a big political issue in Japan, and shortly the Defense Depart-
ment announced that it would remove chemical weapons fromOkinawa
13,000 tons or ve shiploads full. In November 1969 President Nixon
announced unilateral abandonment of biological weapons and proposed
negotiations on the matter of chemical weapons (The New York Times In-
dex, 1970, 339340). In March of 1970 the Democratic National Com-
mittee called on Nixon not to ship the gas to the United States. The Re-
publican governors of Washington and Oregon led suit in April in US
District Court seeking to enjoin the shipment. Magnuson, Hateld, and
Packwood joined the cause. Gravel joined the cause; he and Magnuson
introduced legislation to prevent shipment to the United States including
Alaska. On May 24, three days after Magnuson introduced his amend-
ment on the oor of the Senate, President Nixon phoned Senator Jackson
to tell him the gas would not go through Washington to Oregon, but the
White House was considering Kodiak Island in Alaska as an alternative
storage site. The Army announced that no nerve gas would be moved
while the amendment was pending. In June, the Pentagon announced
that it was studying Johnston Island, a US possession uninhabited by
civilians 700 miles southwest of Hawaii. Meanwhile, deteriorating nerve-
gas weapons from Army depots in Alabama and Kentucky were shipped
by rail to South Carolina for urgent burial at sea off Florida. On April 30
Nixon announced that he would send US combat troops into Cambodia,
which lit a restorm of protest (Congressional Quarterly Almanac, 1970,
930931). There were demonstrations and riots across the United States.
On May 4 four students were shot dead by the National Guard at
Kent State University. On May 9, 60,000 to 100,000 students demon-
strated in Washington, DC. The Congress erupted as well, meetings
were held between relevant committees and the President, and numer-
ous proposals arose to prohibit expanding the war to Cambodia. Nixon
backpedaled.
Senator Harry Byrd (D-VA), Senator Thurmond (R-SC), and Senator
Hollings (D-SC) did fuss on the oor about President Nixons negoti-
ations with Japan over the reversion of Okinawa to Japan (Congressional
Record, April 7, 1970: 1047010478). On November 5, 1969, Harry Byrd
had offered a resolution that the administration should not conclude ne-
gotiations concerning change in the status of Okinawa without seeking
the advice and consent of the Senate (under the Senates constitutional
right to approve treaties), and that resolution had passed the Senate,
63 to 14 (reported in Congressional Record, April 7, 1970: 10474). The
speaking senators concerns were threefold: to uphold the prerogatives
of the Senate, to protect the military interests of the United States in
Okinawa, and, explicitly (10470), to restrain Japanese textile imports.
For these three senators represented states with a textile industry dam-
aged by cheaper Japanese imports. Jumping forward, in May and June
of 1971, some southern senators threatened to defeat any treaty brought
to the Senate concerning the return of Okinawa to Japan unless Japan
agreed to curb textile imports (The New York Times Index, 1971, 1,276).
By September of 1971, the Japanese government, which much desired
the reversion of Okinawa for domestic political reasons, agreed to reduce
voluntarily textile exports to the US in exchange for Okinawa. The
Senate then approved the Okinawa treaty in November by a vote of 84
New dimensions 401
to 6 (Harry Byrd and Magnuson were among the few nay votes, I dont
know why).
Magnuson introduced the Magnuson amendment to the Foreign Mili-
tary Sales Act on May 21, 1970 (Congressional Record: 1648116482). He
made several arguments about the hazards of the proposed shipment. He
also mentioned the treaty issue, stating that he was aware of no commit-
ment to remove the nerve gas fromOkinawa in the Nixon administrations
negotiations with Japan. He boasted of his staunch support for military
appropriations, a true claim. He said that he thought chemical munitions
should be destroyed but that the decision about maintaining or destroy-
ing the Okinawa arsenal was properly the Pentagons. He mentioned that
a chemical and biological warfare treaty was on its way to the Senate.
Magnuson was perplexed: If anyone understands this movement, ex-
plain it . . . I have not found anyone who supports this movement except
Pentagon ofcials (16481). Magnuson reported that Evans, the Repub-
lican governor of Washington, and McCall, the Republican governor of
Oregon, were beseeching the administration on the issue, and quoted
McCalls call to the Republican Vice President: For heavens sake Ted,
give the feelings of Oregonians a little consideration and ease up on the
bullheadedness that is forcing so many supporters of the administration to
the wall in Oregon (16482). Church, Democrat of Idaho, and a mem-
ber of the committee with jurisdiction over the bill, stated his support
for the Magnuson amendment. Gravel of Alaska added that he believed
that the gas should remain on Okinawa or be detoxied or destroyed,
that it should not go to Alaska, and stated that Magnusons amendment
prohibited its shipment to Alaska.
The introduction of the Magnuson amendment did not stop the
Pentagon from studying Alaska as a potential storage site, however.
Five weeks later, on June 29, 1970 (Congressional Record: 2201622023)
Gravel returned with a newamendment. The effect of the Gravel amend-
ment was to prohibit removal of the nerve gas to any US state or posses-
sion and to permit funding for destruction of the materials. Gravel was
fulminating about chemical weapons. In Rikerian terms, he was adding a
dimension to the debate Nixons delay in presenting the 1925 Geneva
chemical and biological warfare treaty to the Senate but in a man-
ner that palpably would lose him rather than gain him votes. Church
suggested that the language of Gravels amendment would not prohibit
removal to US possessions Guam or Johnston Island, but Gravel insisted
that it would. Jackson objected. Jackson, Church, and Fulbright from the
committee of jurisdiction tried to narrow the amendment to the states of
the United States, but Gravel stubbornly refused. Magnuson intervened
to explain that he had a more broadly sponsored amendment already
pending, now joined by a senator from New Jersey, that would prohibit
shipment of the Okinawa gas to the United States. Jackson and Tower
objected that the Gravel amendment would weaken the US position in
negotiations with the Soviet Union on biological and chemical weapons.
Tower further objected that the Gravel amendment tied the hands of the
executive in negotiating with the Japanese (an evident Senatorial norm
is to pressure but usually not to dictate to the executive on foreign pol-
icy). Gravel replied, Obviously, if it does impair the secret negotiations
with the Japanese, I am sure that the conferees can be so informed and
could strike it from the conference report (22021). In other words, pass
the Gravel amendment as a statement, but recede from it in the House
Senate conference on the bill. Gravel stated that his larger purpose was
to promote biological and chemical disarmament.
Now Magnuson made his argument. He explained that his more nar-
rowly drawn and more broadly supported amendment was still pending.
He thought that the Senate should take some kind of position on the
affair. His basic complaint was that:
on all of these occasions when we were discussing Okinawa, nothing was done
to inform us of administration plans. We passed a resolution saying, Please,
Mr. President, before you start doing all these things concerning Okinawa, come
to the Senate and discuss some of them. The President never did this. That is
what we have been arguing about for 5 long weeks. That would have saved us a
lot of trouble. It would have saved a lot of trouble for the Senators from Alaska
and Washington and Oregon. (22022)
I infer that Magnuson and the regional senators had been bushwhacked
by the nerve-gas controversy and were angry about being kept in the
dark by the Pentagon. Further, the Pentagon had tricked Magnuson;
they said they would not ship the nerve gas to the continental United
States, which he accepted, but then they said that Alaska was not part
of the continental United States. It was at that point, he explained, that
he and the remaining Washington and Oregon senators had joined with
Gravel to introduce the Magnuson amendment (22021; The New York
Times Index, 1970, 340). He also mentioned the political problems that
other senators in other regions had with nerve-gas storage and ship-
ments that year (by implication Utah, Colorado, Alabama, Kentucky,
New Jersey, South Carolina, Florida): if we are going to have chemi-
cal and biological weapons, then I think that the Senate should know a
little bit about what is done with them. . . it has been a constant prob-
lem (22021). He also argued that Nixon had ignored the Harry Byrd
resolution asking that the Senate be consulted on Okinawa negotiations
(Riker is focused on this issue, the prerogative of the Senate, but is not
New dimensions 403
aware that the phrase is code for restrictions on imports of Japanese
textiles).
On May 21, Magnuson said he was aware of no agreement with the
Japanese on the nerve-gas subject, on June 29, he said, in the meantime
this agreement on nerve gas was made (22021); on June 29, at one point
Gravel referred to an agreement with Japan to remove the gas (22018)
and at another point disclaimed awareness of such an agreement (22020).
I speculate that the senators had received condential information about
the Okinawa negotiations and that they had learned that a provisional
agreement had been reached with Japan to remove the nerve gas. I spec-
ulate further that the senators nowplanned to take a strong stand in order
to prevent further local political nightmares at the clumsy hands of the
Army. Magnuson said that he had heard that the Pentagon would ship the
nerve gas to Johnston Island, but had heard nothing formal about such a
proposal in other words, he had no commitment from the Army about
its plans. It may very well be that US negotiators needed a domestic re-
action to counterbalance the diplomatic demands arising from domestic
reaction in Japan.
Church, on behalf of the committee of jurisdiction, asked Gravel again
to narrow his amendment so that the gas could be shipped to US posses-
sions such as Guamor Johnston Island. The implication was that if Gravel
did so the committee would consent to the amendment and it would pass
without controversy, probably even without a roll-call vote. Gravel again
refused. Gravel said that he wanted to make a statement about biological
and chemical weapons, but was willing to have the prohibition of shipping
to Johnston Island withdrawn in the conference committee if the adminis-
tration came and said it would be a problem. A clear understanding was
offered by Gravel. Church responded that he could not accept Gravels
broad amendment on behalf of the committee, but did declare that he
would vote for it himself. Church said, The Nixon administration has
played rather loose on this matter. This discussion and vote should bring
them up abruptly. It will force them to come to the Senate and the con-
ference and lay their cards on the table (22022). Church accepted the
understanding offered by Gravel. Magnuson added that in the event the
Gravel amendment failed the upcoming vote, Magnuson would again of-
fer the Magnuson amendment, which by implication would be accepted
by the committee and passed without controversy. The effective choice
was between the Magnuson amendment that would pass without con-
troversy, and the more confrontational Gravel amendment that would
require a roll-call vote.
The Gravel amendment passed by a roll-call vote of 51 to 40. We
must redo Rikers analysis of the vote. It is not obvious that the vote
on the Gravel amendment on nerve gas is related as Riker would like
to the vote on the ChurchCooper amendment to prohibit expansion of
the war into Cambodia; 66 senators voted or expressed positions either
for both or against both measures, but 30 voted in a mixed fashion on
them, a rather weak correlation. The vote on the Gravel amendment does
not t Poole and Rosenthals (1997) usually predictive spatial model of
ideology (the models predictions err on 22 out of 95 votes, the issue
scores 0.488 in terms of their measure of proportional reduction of error,
Voteview, 91st Senate, Roll-Call #414). The ChurchCooper amendment
does adequately t the ideological model (9 errors out of 99 votes, 0.775
proportional reduction of error, Voteview, 91st Senate, Roll-Call #425).
To continue, 38 senators voted for both Gravel and ChurchCooper,
39 if we count a pair from Nelson (not 40 as in Riker). There were
9 Senators in Hawaii, Alaska, Washington, Oregon, and Idaho voting
for the Gravel amendment (Jackson was the exception). That means,
continuing to correct the arithmetic, that there were 39 + (9 7) = 41
(not 42) natural votes for the Gravel amendment. There were 23 senators
who voted against both Gravel and ChurchCooper, 27 if we count a pair
from Long and announced nays from Miller, Mundt, Russell (not 24).
That leaves 32 (not 33) senators, 30 (not 26) of whom were present for
the vote on the Gravel amendment. The 30 divide into two groups. The
rst group of 17 (not 16) voted against Gravel and for ChurchCooper,
the ones Riker said were not impressed by Magnusons heresthetic. The
second group of 13 (not 10) voted for Gravel but against ChurchCooper,
the ones Riker said were manipulated by Magnuson. Riker claims to see
patterns distinguishing the 17 from the 13, but all I see is noise. Except
that there is one slight pattern not mentioned by Riker among the 13 who
voted for Gravel but against ChurchCooper: of the ve Democrats, three
are from states with nerve-gas transportation controversy (Alabama and
Kentucky). The eight Republicans are ideologically and geographically
scattered; I speculate that they had personal ties with the two Republican
senators from Oregon who voted for the Gravel amendment.
There is one more interesting pattern. Riker confounded two of
Magnusons appeals. Magnuson did complain to the Senate about
nerve-gas controversies that year involving Hawaii, Alaska, Washington,
Oregon, Idaho, Alabama, Kentucky, New Jersey, Florida, and other
states, and of failure by the White House to respond straightforwardly
to the problems. He also appealed to those associated with Harry F. Byrd
who sought to get the Okinawa agreement into the Senate so they could
extract textile concessions from Japan. Riker notices that Harry Byrd and
Thurmond ignored Magnusons appeal, but Riker is not aware that their
issue was textile imports. The appeal of Magnuson a free-trader whose
New dimensions 405
state exports military equipment, airplanes, grain, logs, and lumber to
the textile protectionists fell utterly at: not a single senator fromVirginia,
North Carolina, South Carolina, or Georgia voted as Magnuson asked
for the Gravel amendment. The great heresthetical maneuver was a big op
in this respect. Redman is also deluded, as the southerners did not vote
down the line with Magnuson after hearing his honeyed tongue they
were more solidly against him than any other region. All the premises
of Rikers argument are mistaken. He believes the Gravel amendment
passed only because of Magnusons speech, but we know from the record
that the Magnuson amendment would have passed without controversy.
Magnusons appeal to the textile interests failed completely. His other
appeal was not so much to the formal prerogatives of the Senate, but
was about mistreatment of supportive senators by the Pentagon, its trick-
ery on Alaska, and unwanted nerve-gas controversies across the United
States. Democrat Magnusons purpose was not to discredit Republican
Nixon Magnuson, the two Republican governors of Washington and
Oregon, and the two Republican senators from Oregon were trying to
save their skins with their home publics. Jackson did not bail out as
soon as the state of Washington was spared Jackson was a cospon-
sor of Magnusons amendment which explicitly included Alaska. Jackson
was for the Magnuson amendment but against the Gravel amendment,
no doubt because Jackson the hawk loathed Gravels aggressively dovish
views. Absolutely no evidence is offered to support Rikers imputation
that the senators who voted as Magnuson asked did not believe in the
sincerity of his appeal. In sum, one persons rhetoric did not bring about
a majority-opposed outcome.
There is a saying about how to get the attention of a donkey. You hit
it over the head with a two by four. That is what Magnuson did. Given
their generous support for the military establishment, the two Democratic
senators from Washington should have had upon request a quick com-
mitment from the Pentagon not to ship the gas to the Pacic Northwest
(of course the military were busy ghting a major war in Southeast Asia
at the time). Then the Pentagon tricked Magnuson over Alaska with the
phrase continental United States, and he got out the small board of the
Magnuson amendment. That did not get results. So then he let wild man
Gravel loose, and whacked the Pentagon over the head with a big board.
That did get results. The Pentagon eventually committed to shipping
the gas to Johnston Island. Apparently, the administration did approach
the conference committee, because the conference report on the For-
eign Military Sales Act added a third sentence to the Gravel amendment:
For purposes of this section, the term United States means the several
states and the District of Columbia (Congressional Record, December 31,
1970: 44305). Johnston Island was a possession, not part of any state.
The understanding between Gravel and Church was carried out.
Magnuson got what he wanted. Gravel got what he wanted (although,
I think, not as a result of his effort): in August 1970 Nixon sent the
Geneva protocol on bacteriological and chemical weapons to the Senate.
The textile industry got what it wanted: restriction on Japanese textile
imports.
Here we encounter a newcontradiction. Riker conceded that cycles are
unlikely onroutine issues, but here is a routine issue where a single speaker
purportedly turned a minority into a majority by dimensional manipula-
tion, and such dimensional complication looks to be easy for any issue no
matter howsmall. Indeed, there were many dimensions involved in this
small issue, among them: safety of storage and transport of biological and
chemical weapons in the northwest; safety of storage and transport of bi-
ological and chemical weapons elsewhere in the country; the expectation
of senators to be spared needless political embarrassments; the biological
and chemical warfare treaty; SenatePresident relations; USSoviet rela-
tions; USJapanese relations; US strategic military capacity; the Vietnam
War; protection of the US textile industry; and party, sectional, and per-
sonal ties among senators. Was it a case of anything can happen? No,
again because discourse is unlike agenda control. A rhetor cannot force
a dimension on an auditor. Magnuson complained that the Pentagon
tricked him on Alaska, that a number of senators across the country had
gone to political hell on nerve gas, and that the military was still not
forthcoming on the issue he asked if that is how SenateExecutive rela-
tions should be in the future. In the absence of evidence to the contrary,
the inference must be that the senators who voted with Magnuson ac-
cepted his argument. Riker and Weingast (1988, 392393) believe that
this manipulation succeeded even though the senators inuenced by it
doubtless recognized that they were being manipulated, and probably
resented it. Ignoring the several important errors of interpretation, the
original story depends on this curious and unwarranted inference. We
are to believe that 13 seasoned senators including tough Senator Barry
Goldwater (R-AZ) were forced to vote against their own position be-
cause of Magnusons clever rhetoric, when another 40 senators felt no
hesitation whatsoever in ignoring Magnusons plea entirely. Why did the
13 succumb but the 40 resist? We are offeredno explanation. I suggest that
the 13 senators may have been concerned about either their colleagues
on whom they depended to build coalitions or their own fortunes given a
pattern of increasing executive neglect of senatorial relations. As for the
many other dimensions, discussion is not voting. Discussing the issues
does not create a majority-rule disequilibrium. That can only happen
New dimensions 407
when it comes to voting and then only under the unrealistic and unfair
conditions of the McKelvey model.
The art of political manipulation
There are twelve stories in Rikers (1986) The Art of Political Manipulation.
The point of diminishing returns has already been passed in discussing
such stories; it is no longer fruitful for my purposes to subject each new
one to a searching examination (students may wish to take on the re-
mainder). I have not patiently examined Rikers chapter on heresthetic in
C.P. Snows (1951) novel The Masters, about the election of a master in
a Cambridge college; the chapter on trading votes at the constitutional
convention; and the chapter on Speakers of the House Reed and Cannon.
The rst part of the chapter about the ying club tells the story of how
Plott and Levine hoodwinked Levines ying club into buying the planes
Levine wanted by unfairly manipulating the choice of voting rules. I nd
the story distasteful in several ways, and have not examined it in detail.
The second part of the chapter reports Plott and Levines experiment,
the one I described as suggesting that agenda control is possible only
if the agenda-setter has unconstrained power over the sequence of con-
sideration and unconstrained power over the distribution of information
of preferences; in other words, if one party is given an unfair advantage
then the procedure has an unfair outcome. The chapter about camouag-
ing the gerrymander tells the story of a city manager covertly managing a
gerrymander of city-council districts in order to keep her job; it is said to
be real, but is undocumented. The chapter on Pliny the Younger and par-
liamentary law we have already encountered; it is an example of agenda
control countered by strategic voting such that in the end there was no
harmful manipulation. The chapter on how to win a roll-call vote by
not voting reports an isolated instance of parliamentary chicanery. Five
of the stories, all alleging cycling and instability in one way or another,
we have already examined and found wanting: Lincoln at Freeport,
Chauncey Depew and the 17th Amendment, Gouverneur Morris in
the Philadelphia Convention, Warren Magnuson and nerve gas, and
exploiting the Powell amendment. The story from Riker (1958) about
a cycle in Congressional agricultural appropriations is not included in
the volume, and the celebrated Civil War stories in Riker (1982) about
the cycle in the Wilmot Proviso and the cycle in the election of Lincoln
are curiously absent.
In response to my volume, there may be a renaissance of cycling claims
in the literature. But such claims would miss the point, in two ways.
First, anecdotes are not sufcient. The only persuasive evidence would
be a showing of frequency from a properly drawn sample of some dened
universe of voting, together with a substantive analysis of whether each
instance of cycling is harmful. Second, even if cycling were shown by
that standard, in order to salvage the doctrine of democratic irrational-
ism it would have to be shown further that undemocratic outcomes are
irremediable.
17 Plebiscitarianism against democracy
Introduction
Riker summarizes his case against populism (democracy) and then of-
fers his case for liberalism. Populismis arbitrary and meaningless, there
is no identiable will of the people nor public good, he says. His liber-
alism, in contrast, requires only that it be possible for citizens to reject
ofcials of whom they disapprove. Rikers alternative of liberalism fails,
I maintain, because it reduces to merely the random removal of ofcials,
and because its justication unavoidably appeals to a will of the peo-
ple or a public good that Riker is concerned to reject. If Rikers larger
argument were correct, then neither democracy nor Rikers minimal lib-
eralism would survive. The chapter concludes by tracing the provenance
of the doctrine of democratic irrationalism through James Burnham to
the elite theorists of the early twentieth century, particularly Pareto.
The summary case against populism
Rikers case against populism depends crucially at every point on what I
have called his basic argument pattern concerning the obscurity of prefer-
ences. Most commentators fail to appreciate the centrality of this premise
to his total argument. His closing brief also crucially relies for evidence on
his erroneous case study of the 1860 American presidential election. First
in his summary case is the claim that democracy is arbitrary. If there are
more than two alternatives on an issue of political concern, then any one
of a number of reasonable voting rules could be applied. Different rules
lead to different outcomes from the same prole of voters preferences,
however, and none of the reasonable voting rules is normatively better
than any of the others, on his account. He conjectures that there are few
voters proles such that different voting rules would report the same
winner (we reviewed evidence showing that the conjecture fails). Fur-
thermore, most of the voting rules in wide use do not collect information
on ranking of all alternatives, so we are not aware of how different rules
409
might yield same or different outcomes. This means that, even if some
method produces a reasonably justiable amalgamation [e.g., a Condorcet
winner], we do not know it (emphasis in original, Riker 1982, 235). For
example, his reconstruction of the election of 1860 shows, he believes,
that different voting rules would have reported different outcomes, but
we can not be certain about his or anyones reconstruction, he says, hence
we can never be certain about how voting rules perform. Outcomes are
the function of both the preference prole and of a voting rule inherited
from some prior constitutional arrangement, in other words, the method
of counting partially determines the outcome of counting. How can we
judge that an outcome is fair if we cannot know the preference prole
and thus the extent to which the outcome is a function of the voting rule
rather than of the voters preferences? Outcomes might be accurate amal-
gamations and they might not be, but we seldom possess the information
to judge which is which. The arbitrariness claim depends on his basic
argument about the obscurity of preferences.
Second in his summary case is the claim that democracy is meaning-
less. Suppose that a society has decided to use a particular method of
voting and dene as fair the outcomes of that method. But even with
one method of voting the same prole of preferences can yield different
outcomes; this we know from the theorems of Arrow on cycling, Gibbard
and Satterthwaite on strategic voting, and McKelvey and Schoeld on
chaos in multiple dimensions. Every reasonably fair method of vot-
ing can be manipulated in several ways. Since we cannot know whether
manipulation has occurred, the truth and meaning of all outcomes is
thereby rendered dubious (Riker 1982, 236). One way to manipulate
is by strategic voting, which is probably commonplace, thus all voting
is rendered uninterpretable and meaningless (237); manipulated out-
comes are meaningless and we cant tell whether or not outcomes are
manipulated. Another way to manipulate is by agenda control, and we
can be fairly certain that it is commonplace; but we can never be sure
when and howit occurs or succeeds, thus again all outcomes are rendered
uninterpretable and meaningless. One variety of agenda control is the in-
troduction of newdimensions in order to introduce disequilibriumwhere
there was none before. This is demonstrated by his analysis of the election
of 1860, he says. It is hard to say . . . that the most momentous election
in American history was a fair or true amalgamation of individual val-
ues, mainly because the decision was thoroughly and. . . deliberately
confused by the inclusion of several issues (Riker 1982, 237). Be-
cause of the obscurity of preferences we can never know for certain
whether any one outcome was manipulated or not, thus all outcomes are
meaningless.
Plebiscitarianism against democracy 411
Rikers basic argument pattern is also the basis for his rejection of what
he calls populism and what the rest of us call democracy. Riker (1982,
238) says that populism can be summarized in two propositions: rst,
what the people, as a corporate entity, want ought to be social policy,
and second, the people are free when their wishes are law. We do not
and cannot know what the people want, according to Riker (1982, 291):
Populism is supposed to reveal a substantive will, a proposition with content.
Yet if voting can fail to reveal such propositions accurately and if we do not and
cannot know in any particular instance whether failure has occurred, then none
of the propositions supposedly revealed can be believed.
The rst proposition falls, says Riker, because we do not know the peo-
ples wishes, and the second falls because if we do not know their wishes
the people cannot be made free by enacting their wishes. Populism fails,
he says, not because it is morally wrong, but because it is empty. Before
the Reformation, he says, the popes decrees as to the will of God were
authoritative, but the proliferation of such claims after the Reformation
rendered the batch uninterpretable and meaningless no one could be
sure what God wanted even if all postulated such an entity. Similarly, he
continues, the discoveries of modern political science will lead our next
generation to reject the will of the people as the basis of government.
There may be a people, but we cannot know them.
It is important to realize that Rikers case against democracy is not
based on the claimthat there are inevitable imperfections in social choice,
that the variety of real decision procedures only imperfectly approximate
one or another of a family of justiable ideals. No, Rikers case is based
on the claim that the preferences of other people are unknowable. If,
for one reason or another, one does believe that it is possible to infer the
preferences of others, then one may reject his case against democracy
without much ado.
The defense of liberalism
According to Riker (1982, 242), the essence of his liberal interpretation of
voting is that voting permits the rejection of candidates or ofcials who
have offended so many voters that they cannot win an election. Voting
does not provide a statement of the popular will; all it does is provide that
an ofcial is retained or rejected. Then follows a remarkable argument.
We begin by assuming the existence of what we already know does not
exist namely, a fair and accurate amalgamation of voters values (242).
This will not taint the analysis, he says, it is an initial standard, not an
instrument of interpretation (I shall explicate this claimshortly). He then
comments on four possibilities: (1) that a good ofcial is retained; (2) that
a good ofcial is rejected; (3) that a bad ofcial is rejected; and (4) that
a bad ofcial is retained. As to the rst, if a good ofcial (one who has
not offended enough voters for them to reject him in a fair and true
amalgamation of their values, [242]; Riker does not explicitly use good
or bad) is retained, then whatever voting method is in use is working, and
there is no problem. Second, if a good candidate is rejected, for example,
if she is the victim of a plurality voting rule, strategic voting, agenda
control, dimensional manipulation, and so on, then there is no problem
with the liberal interpretation of voting, according to Riker. How is that?
The liberal interpretation requires the rejection of the offending, not the
retention of the unoffending. If a decision procedure permits rejection
of an alternative, there is no reason to require that the mechanism of
rejection work perfectly, he says.
Furthermore, the liberal interpretation does not require that ofcials
be agents of the voters will since there is no such will, according to Riker.
This means that an ofcial in the Rikerian regime would abandon any
effort at reading the voters will, not because the ofcial is sophisticated
enough to know that there is nothing there to read, but merely because
he or she knows by experience that voters rejection may be random
(Riker 1982, 243). (An implication would be that politicians obsession
with opinion polling is an extravagant illusion.) The liberal interpretation
intends to prevent abuse of ofce or authority on the part of ofcials
by threat of removal at election; the randomness of the threat does not
matter, says Riker, in fact, it motivates ofcials to try even harder to avoid
offending the voters. The older Rikers claim about random punishment
is just plain wrong, however. The younger Riker (1953, 110) states the
reason:
The process of government can be controlled by citizens only when elections are
a transmission belt of ideas and decisions fromthe voters to the rulers. If elections
have no relevance to public policy, then the policy makers need not respect the
electoral sanction.
True, if there is a random chance of getting a ticket for driving faster
than the speed limit, then my speeding is constrained by the expected
value of the penalty. But if there is a random chance of getting a ticket
whether or not I am driving faster than the speed limit, then I will simply
do as I please. If I drive 30 miles per hour there is a 10 percent chance of
getting a $500 ticket, and if I drive 100 miles per hour there is a 10 per-
cent chance of getting a $500 ticket. Under those circumstances why
should I drive 30 miles per hour if I would rather drive at 100 miles per
hour?
The third possibility is that a bad ofcial is rejected. If that happens,
the voting method is working, and there is no problem. The fourth possi-
bility is that a bad ofcial is retained. Failure to reject a bad ofcial does
not violate the liberal interpretation of voting, Riker says. In discussion
of the second possibility (rejection of a good ofcial), Riker said that all
his liberal interpretation requires is the rejection of bad ofcials, not the
retention of good ones. Now, in discussion of the fourth possibility, we
learn that failure to reject bad ofcials is not a problem. All that his liberal
interpretation of voting requires is that it be possible to reject a putatively
offending ofcial, not that the rejection actually occur (emphasis in orig-
inal, Riker 1982, 243). We know from the social choice results that good
ofcials can almost always be defeated because in multiple dimensions
there are almost always win-sets demarcating policies preferred by differ-
ent majorities over any given status quo; so there are always majorities
who can defeat bad ofcials too, he says. And if success is even some-
times possible, then the liberal interpretation can be sustained (243).
In a footnote Riker (1982, 290) addresses the criticism that he accepts
liberalism if it works occasionally (that is, randomly), but rejects pop-
ulism if it fails at all. He admits that he imposes different standards. He
maintains that populism requires knowledge of the peoples will. As we
have just seen, his rejection of populism essentially depends on his ba-
sic argument pattern: any one instance of voting might be manipulated,
we cannot know which, hence preferences are obscure in all instances
of voting, therefore the peoples will is unknowable. He maintains that
the random rejection of ofcials is sufcient for liberalism. To ll in his
argument, if liberal voting randomly rejects any candidate it succeeds,
but if populism is anything less than perfectly correct in rejecting bad
candidates it fails. Riker does not discuss tolerances, although he should:
if there is a 5 percent chance that a candidate is randomly removed by
voting, is that liberal voting device better than a hypothetical populist
voting device that rejects lesser alternatives 85 percent of the time?
I shall begin with an analogy. A liberal engineer is asked by a vend-
ing company to invent a device that will distinguish the American 25-cent
piece known as a quarter from washers (the generic term for such fakes is
slugs) that one can obtain at ten for a penny at a hardware store. The com-
pany wants the device to accept quarters and to reject slugs. The engineer
collects his fee and returns with a prototype. The device randomly accepts
or rejects inputs. When it randomly accepts quarters, the engineer claims
that it is working correctly. When it randomly rejects quarters, the engi-
neer claims that this does not matter, what should matter is that the device
reject slugs. When the device randomly rejects slugs, the engineer claims
that it is working correctly. When the device randomly accepts slugs, the
engineer audaciously claims that when a user inserts a slug there is the
possibility that it might be rejected. Those who would pass slugs would
be deterred by the knowledge that slugs are sometimes rejected; these
counterfeiters would be even more deterred by the knowledge that rejec-
tion of slugs is random, the engineer claims. The engineer concludes his
presentation with the observation that although he assumed a distinction
between quarters and slugs for purposes of discussion, in fact both are
just meaningless pieces of metal, the rst possessing higher value only as
a matter of social convention, and so it does not matter anyway which
inputs are accepted and rejected, which is the true beauty of his random
mechanism. Such is Rikers argument.
He asked us to assume at the outset the existence of something that he
says does not exist, a collective judgment or will concerning the good or
bad performance of an ofcial. So drop the initial working assumption
and replace it with the supposedly correct assumption that there really is
no such judgment or will as to good or bad performance of an ofcial.
Now his claim for the adequacy of random rejection gains clarity: if there
is no good or bad then it doesnt matter who is retained and who is
rejected and thus a random device is as good as any device. The random
replacement of ofcials is all we can hope for from Rikers liberalism.
It seems worthwhile to point out just how little is contained in the liberal inter-
pretation of voting . . . Since social decisions are not, in liberal theory, required to
mean anything, liberals can cheerfully acknowledge that elections do not neces-
sarily or even usually reveal general will. All elections do or have to do is to permit
people to get rid of rulers. (emphasis added, Riker 1982, 243244)
Now suppose that the vending company hires a populist engineer. The
populist offers several devices that implement various techniques to mea-
sure various characteristics of the input its size, its weight, its shape,
the notching on its edge, and so on and variously aggregate the mea-
surements into a nal accept-or-reject decision (just as there are different
reasonable methods of decision for human collectivities). One device is
extremely costly, but will avoid both false positives and false negatives
99.95 percent of the time. Another device is quite affordable, and avoids
both false positives andfalse negatives 95 percent of the time. The vending
company rejects the liberals device, rejects the populists costly device,
and accepts the populists cheap device. The company knows expected
failure rates fromdirect tests before implementation. In practical applica-
tion, the company does not know in advance whether any one particular
input will result in a false positive or a false negative, but this does not
lead them to accept the liberal engineers fantastic advice that quarters
are unknowable.
Do not be confused by Rikers claim that his liberal interpretation per-
tains to the rejection of candidates who have earned the disapproval of
voters. For the purposes of the social choice theorems that Riker inter-
prets as defeating populism, the rejection of candidates by disapprovers
is formally identical to the acceptance of candidates by approvers. If a the-
orem holds for A > B > C then it holds for C < B < A. Further, say that
there are three candidates in the cycle so central to the irrationalist vision:
A > B > C > A. For each of the candidates we can say both that she is
approved of by a majority of the voters and that she is disapproved of by
the voters. Riker does not deny this; he emphasizes in his exercise that any
candidate, good or bad, can be arbitrarily defeated. One aw of Rikers
demonstrative exercise worth noting is the assumption that candidates
are good or bad on their own without reference to an alternative. We can
call this the NixonAgnew problem. You may think that Nixon is bad,
but you wont want to get rid of him when the alternative is Agnew, who
is even worse, a consideration that former President Nixon perhaps had
in mind when choosing his mediocre sidekick. Correcting for the absence
of comparisons gives us eight different possibilities: (1) A better incum-
bent could beat worse challengers, which is ne. (2) A good incumbent
could beat equally good challengers, which is ne too. (3) A bad incum-
bent could beat equally bad challengers, which is arguably bad. (4) A
worse incumbent could beat better challengers, which is bad. (5) A bet-
ter incumbent could lose to a worse challenger, which is bad. (6) A good
incumbent could lose to an equally good challenger, which is okay. (7) A
bad incumbent could lose to an equally bad challenger, which is arguably
bad. (8) A bad incumbent could lose to a good challenger, which is good.
The important case that Riker neglects to consider is when voters are
forced to accept a bad candidate because the alternative candidates are
worse. The majority of voters indeed disapprove of the bad candidate but
their disapproval does not remove him from ofce.
But all of this is irrelevant, because for Riker there is no good or bad, no
better or worse, to the matter. He says that ofcials who abuse ofce and
authority will be rejected, but he seems not to notice that there is no test
for such abuse apart from the collective judgment or will of voters, which
he has already dismissed as a fantasy. If ofcials cannot further the public
good because it does not exist then neither can they further a nonexis-
tent public bad. If there is neither public good nor public bad, another
problem arises. What justies the institution of the random rejection of
rulers, or indeed any political institution? There is no public good for the
inhabitants of the Peoples Republic of China; hence in terms of aggregate
subjective welfare there is no way to distinguish between the party dicta-
torship and some democratic alternative. There is no way to distinguish
the governments of Serbia and Slovenia. The Italian government of 1930
is no different than the Italian government of 1990. And so on. Riker
does concede the existence of one kind of common interest: a common
or public interest is held in common, so voting is unnecessary to reveal it:
Any randomly chosen member of the society can articulate public interest
was well as any other (Riker 1982, 291). Presumably he has the Pareto
criterion in mind. If everyone in Italy wanted to depose Mussolini with
the exception of Mussolini himself, we could not on Rikers denition say
that Italy had a common interest in changing its leadership. If only one
member of a society rejects the principle of elections for public ofce,
then elections would not be in the common interest as Riker denes it.
Riker goes so far as to say that there may be an objectively right public
interest for a society even when people do not agree. But in the absence
of unanimity each interpretation of that objective public good is merely
arbitrary, he says; leftist reformer or demagogue Ralph Nader might cor-
rectly state the public interest and rightist reformer or demagogue George
Wallace might correctly state it, but voting short of unanimity provides
no evidence as to which of the two best approximates the objective public
good. I would add that even a unanimous vote might miss the public
interest if not implausibly there were a universally held false belief among
the voters; the proper question is whether there is a probabilistic con-
nection between voting and the public interest. Further, note well the
absence of public reason in Rikers scheme. He seems not to realize that
by his eschewal of public reason he has painted himself into a corner. The
judgment inherent in his liberal interpretation of voting that the random
rejection of ofcials would be in the public interest, perhaps even objec-
tively, is invalidated if only one member of the society disagrees with it,
and surely one of the ofcials to be rejected would disagree, otherwise
there would be no need to force his removal.
Riker continues that his liberal interpretation of voting satises three
desiderata of democracy: participation, liberty, and equality. His argu-
ment is contradictory, however. He says that rejection is possibly random,
which is consistent with his argument against populism, but he thereby
implies that the rejection is possibly not random. If the rejection is not
random then that implies a public will, but if there is a public will then
we are back to the populismthat Riker denounces. His liberalismsatises
a kind of participation, he says, because each voter enjoys the possibility
of rejecting an ofcial. At best ofcials are responsive to a (possibly ran-
dom) threat of expulsionfromofce. But this may leadthemto avoidgross
offense to groups of citizens who can eject them from ofce (emphasis
in original, Riker 1982, 245). What is this may? If rejection is random
then ofcials are not constrained. If rejection is not random then there
is a public will and we are back to populism. He also says that his liberal
participation engenders self-direction and self-respect within individuals,
but how can that be if there is a random, that is, no, relationship between
individual action and collective result? In his rst chapter Riker defended
his liberalism on the ground that it made ofcials an approximate agent
of the electorate and thus achieved democratic self-control. Now in his
last chapter he implicitly repudiates that earlier characterization (242
243), yet he continues to insist that a random relationship between the
electorate and ofcials would also achieve democratic self-control.
The liberal veto also promotes a kind of liberty because it is a curb on
tyranny, he says. The chance to express disapproval of an ofcial and the
chance that the disapproval might result in rejection of the ofcial satises
the ideal of negative liberty, of freedom as absence of restraint. There are
two problems. First, in Rikers scheme, absent unanimity of judgment or
will there could be no such thing as tyranny; because if there is no public
will then there is no public will concerning what is tyranny and what is
not. Second, if the removal of tyrannical ofcials were imperfect, then
ones liberty would suffer each time the voting rule failed; and worse, if,
as it must be in order to avoid the supposed populist error, the removal of
public ofcials were randomthen again there is no connection between an
individuals action and the collective result. In the Soviet Union one had
the opportunity to cast a vote against some ofcials, and occasionally an
unpopular ofcial would be replaced by another nominee from the party,
yet there was no relationship between citizens preferences and policy
outcomes. The Soviet voting system satises the older Rikers (1982)
liberalist criterion of formally possible rejection, yet would he want to call
Soviet voting democratic? Contemplate this objection from the younger
Riker (1953, 9192):
Truly responsible government is only possible when elections are so conducted
that a choice of men is a decision on policy, that a decision on policy is soon trans-
formed into action, and that action taken is popularly supervised. . . Consider,
for example, plebiscites in the Soviet Union. . . The popular will is not really
consulted; and the people, however avidly they vote, indeed do not rule. Elections
are a fa cade . . . because the structure of government does not permit elections to
inuence policy making.
Indeed, actual tyrants are inordinately fond of plebiscites, which if they
were held on some regular basis would be democratic by the Riker cri-
terion. Finally, the liberal veto promotes a kind of equality, according to
Riker. Voters have an equal right to participate, but the claim for equality
carries only if one accepts that Rikers liberal participation the formal
possibility to reject an ofcial is adequate.
Populism and democracy
What does Riker mean by populism? In his theory, it is some version of the
notion rst theoretically articulated by Rousseau, that the legitimacy of
the state is founded in the will of the people (1982, 11). What the rest of
the world calls democracy Riker calls populism; that way he can remain a
democrat even though he rejects the idea that government should respond
to what its citizens judge best. This is convenient for his case, as populism
is a label whose connotations range fromthe weakly to the strongly pejora-
tive. Rikers liberalism, however, unwittingly resembles what is pejoratively
called plebiscitarianism possibly tyrannical rule justied by merely for-
mal opportunities for electoral rejection. Thus, we can accept Rikers
contrast of liberalism (positive) against populism (negative), or we can
substitute the contrast of plebiscitarianism (negative) against democracy
(positive).
Democracy, or what Riker calls populism, includes but is by no means
limited to the simplistic implementation of the peoples will that declares
the public good to be whatever a bare and direct majority chooses from
moment to moment in ordinal pairwise majority voting. Even the demo-
nized Rousseau avoided this error: for Rousseau, the will of all revealed
by voting may imperfectly identify what would really be good for all, the
general will. A democrat may quite consistently and defensibly recom-
mend broadly accepted institutions that neutrally rene and enlarge the
will of the people. Raw political preferences may be poorly informed and
may be contradictory even within individuals. Public deliberation within
multiple overlapping arenas, outside and inside elections and the repre-
sentative assembly, serves to informand to order rawpolitical preferences.
A democrat may hold that the people (and he among them) can be seized
by myopic and self-defeating passions, and thus that it is advisable that
the legislative power be conned to the promulgation of general laws and
refrain from applying law to particular individuals, that it is advisable for
legislation to be approved by two or more differently composed bodies
(lower house, upper house, executive, constitutional review), or for cer-
tain types of legislation to require supermajorities or perhaps better ap-
proval by two succeeding assemblies, and so on. A democrat may believe
that representative democracy rather than direct democracy is necessary
for a group the size and the nature of the territorial state. Representatives
may specialize in political information, and may professionalize and de-
personalize political conicts. A democrat may insist that there are essen-
tial preconditions to democracy, such as rights to life, liberty and personal
property, regular elections, equal voting rights, freedom of association,
and freedom of speech, such that violations of these preconditions by
minorities or majorities should be constitutionally prohibited and that
their violation would justify rebellion by aggrieved parties. A democrat
may propose to free and equal citizens an independent standard of the
public good based on generalized interests discovered by hypothetical
choice behind some veil of ignorance that yields principles similar to
Rawlss equal basic liberties, fair equality of opportunity, and the differ-
ence principle; or an independent standard such as Benthamite summa-
tion of utilities, which yields similar practical results; or, more modestly an
aggregate subjective standard such as in a multidimensional issue space
the point that is the intersection of medians; and propose consent to
constitutional arrangements including methods of collective deliberation
and decision that would approximate to an independent standard. All of
these proposals are debatable, but they are all democratic since each is
based, one way or another, on the will of free and equal people. When
Riker takes on populism he takes on all of these conceptions, not just
the simplistic conception of democracy that has few serious advocates.
Populism in the standard sense of the term refers to a collection of
distinct political movements related by family resemblance:
These people and movements, then, are populist, and have much in common: the
Levellers; the Diggers; the Chartists (Moral and Physical Force); the Narodniki;
the US populists; the [Russian] Socialist-Revolutionaries; [Mahatma] Gandhi;
Sinn Fein; the Iron Guard; Social Credit in Alberta; Cardenas; Haya de la
Torre; the CCF [Cooperative Commonwealth Federation] in Saskatchewan;
Poujade; Belaunde, Nyerere. (Wiles 1969, 178)
An adequate but imperfect denition of populisms central premise is:
virtue resides in the simple people, who are the overwhelming majority,
and in their collective traditions (Wiles 1969, 166)). The rst move-
ment with the name, and still the exemplar, was the American populist
movement of the later nineteenth century. The American populists were
a mass movement of small vulnerable rural producers in the South and
the West, with a hostility to all large-scale agencies contrary to the in-
terests of the farmer, with a desire for perfected competitive capitalism
rather than an economy dominated by monopolies and trusts, with al-
lies among frontier miners on the basis of free-silver monetary policies,
and with aspirations for alliance with workers in the cities (Worsley 1969,
220). The populist ideal was that the general interest should prevail over
the special interests. Most populist demands were eventually enacted by
the two major parties. The Narodniki were a movement of intellectu-
als in later nineteenth-century Russia who envisioned a society built on
the traditional mir (collective peasant village), and who were revolution-
ary anticapitalists; their label is translated into English as populism. The
Narodnya Volya (the Peoples Will) was a terrorist organization which
assassinated Tsar Alexander II in 1881 (Worsley 1969, 220). Lenin dis-
dained the peasantist policies of the Narodniki and their successors the
Socialist Revolutionaries. Another sort of populismwas identied in post-
colonial Asia and Africa, whereby the nation is opposed to the outside
world including the ex-colonial powers, and the nation is represented
by an authoritarian partystate regime (Worsley 1969, 229). Yet another
use of the term populism is for the charismatic presidential authoritari-
ans such as Vargas in Brazil and Peron in Argentina who promised the
urban masses a redistribution of wealth from the compradors and their
American imperialist masters to the people (Hennessey 1969, 30). The
avowed traditionalist Shils (as related by Worsley 1969, 242244) of-
fers a general denition of populism, based, however, on the American
variant, that exhibits a waspishness on the topic in American conserva-
tive discourse. For Shils, the two principles of populism are rst, the
supremacy of the people over every other standard, over the standards
of traditional institutions and over the will of other strata. Populism iden-
ties the will of the people with justice and morality (244). Second, a
direct relationship between leaders and the people, unmediated by insti-
tutions, is desired. This seems to be a slamagainst Roosevelt and his New
Deal. Nevertheless, there are many varieties of what I termed democracy
that are not included within Shilss contentious denition of populism.
Populism is a theoretical category for Riker, but what does he think are
its empirical referents? Populism in Rikers pejorative sense perhaps orig-
inates with the experiences of the French Revolution, when Robespierre
and his Committee of Public Safety in the name of peoples will instituted
the Reign of Terror. Populism in this sense is any tyranny that claims to
rule in the name of the people. Oddly enough such regimes often orig-
inate in or are periodically sustained by manipulated plebiscite, which
by Rikers theory would provide legitimacy so long as there is a regu-
lar but merely formal possibility for rejection of the regime. Since, for
some reason, most modern states claim to embody the will of the peo-
ple, the scope of Rikers term populism is vast. Communists (Riker 1982,
245), such as the ghastly Khmer Rouge in Cambodia of the 1970s, were
populists in his sense, since they claimed to rule on behalf of the peo-
ples true interests. He does not mention the case, but the Iron Guard
in Romania of the 1930s viciously oppressed and murdered Jews and
other minorities on behalf of the ethnic majority. Vargas of Brazil, Indira
Gandhi of India, Peron of Argentina each an elected leader who re-
sorted to demagogic authoritarianism are quite denitely populists,
according to Riker (1982, 245). Riker explicitly includes Latin American
constitutional dictatorships in the category of actually existing populism
(1982, 246). Furthermore, anyone who sees liberty as not only negative,
as the absence of restraint, but also as positive, as the capacity to direct
ones own life, the distinction offered by Isaiah Berlin, is a populist who
will end up with rulers who can oppress both the minority and the very
majority whose will they are supposed to work (245).
Who is the most populist philosopher? Not Rousseau, not Pol Pot, but
Benthamand his fellowradicals are the worst intellectual offenders (Riker
1982, 257). Elsewhere, Rawls, and contemporary social democrats, are
denounced as populist utilitarians who have consistently subordinated
humane values to some arbitrary and imposed virtues they prefer (Riker
1980b, 4243). Which is the most populist society? The reader will be
surprised to learn that Great Britain is the politically worst of all nominally
democratic societies. In a footnote, Riker (1982, 256257) explains that
what he calls liberalism and populism is called liberalism and radicalism
by Beer (1960, 33) in his comparative analysis of the British and American
polities. Quoting Beer (1960), radicals such as Bentham, would make
government the instrument of the will of the people, a unied and au-
thoritative force in which he found the only sovereign for the polity and
for which the majority spoke. Riker does not indicate the remainder of
Beers thought, that the radicals contrasted the general interest to the
various special interests (as did the American populists). The supercial
individualism of the radicals evolved into the collectivist notions of the
Labour Party, Riker continues, precisely Berlins development of coercion
out of positive liberty. (It may not matter for the argument, but we should
note that Berlin did not consider the policies of the parliamentary Labour
Party to be a threat to British liberty.) Beers analysis in no way resembles
Rikers, however. In fact, Beer (1960, 31) considers the British system to
be quite practically successful in combining popular participation with
effective and coherent governance. Remarkably, his major reservation is
that the British parties so strongly inuence public opinion as to render
the electorate too homogeneous (5253), that the system may not ade-
quately respect the rawer formof the popular will. In Great Britain, Riker
believes, the populist elimination of constitutional limitations threatens a
constitutional dictatorship (Riker 1982, 248), it seems unlikely that
the liberal sanction [of elections] can survive populist institutions, he
predicts (249).
Which is the most liberal, in Rikers terms, of all societies? Regular
elections are sufcient for the existence of Rikers (1982, 250) liberal
democracy, but further institutions are required for its maintenance, he
says. Those are the institutions found in the US Constitution, such as fed-
eralism, the legislative, executive, and judicial branches of government,
and the separation of powers. One would be more comfortable if it were
not an American informing us that America has the best political institu-
tions in the world. The younger Riker (1953, 161, 164) was of a different
mind:
the separation of powers was designed to impede majority action. It is, as it
was intended to be, the primary obstacle to those effective, legitimate ma-
jorities which. . . are an indispensable necessity for realizing the democratic
ideal . . . although the separation of powers has been justied as a protection of
minorities, it has in reality the opposite effect. Genuine protection for minority
groups is the process of compromise inherent in democratic politics . . . Ambition
must be made to counteract ambition, said Madison; but after 160 years of
experience we say, ambition must be subordinated to majority will.
That the Constitution with its prohibition on the taking of property,
part of its implicit endorsement of slavery; its Senate, presidency, and
Supreme Court designed so as to oppress an African-American minor-
ity and to defy a European-American majority which opposed the evil
institution of slavery; its stable but brittle adherence to any inherited mi-
noritarian status quo no matter how unfree, unequal, unparticipatory,
and undemocratic it was; its federalism that invited the secession of the
southern states; that this Constitution had anything to do with the onset
of the bloody Civil War is not a proposition within Rikers ken. The Civil
War came about because of a meaningless cycle according to Rikers
investigations, not from a tragic conict between irreconcilable princi-
ples and interests exacerbated by defective antimajoritarian institutions.
1
Nowa newconict threatens. The present situation in the United States
is . . . that, although the constitutional limitations remain, populists per-
sistently seem to undermine them. . . our homegrown populists may well
succeed. Populism puts democracy at risk (Riker 1982, 252). The rem-
edy, he says, is to defend the Constitution in the short run and to widely
disseminate the discoveries of social choice theory in the long run.
Rikers a priori political theory approves of the presidential democracy
originating in the US Constitution and disapproves of the parliamentary
democracy originating in Great Britain and the continent of Europe.
Presidential democracy protects liberty but parliamentary democracy
threatens it. This is curious, as empirical political scientists nd that
presidentialism seems to involve greater risk for stable democratic
politics than contemporary parliamentarism (Linz 1994, 70):
with the outstanding exception of the United States, most of the stable democ-
racies of Europe and the Commonwealth have been parliamentary regimes
and a few semipresidential and semiparliamentary, while most of the countries
with presidential constitutions have been unstable democracies or authoritarian
regimes. (Linz 1994, 4)
Particularly, it is hypothesized that demagogic-authoritarian tendencies
in some Latin American countries are attributable precisely to their
presidential constitutions adopted as a result of admiration for the
great American democratic republic . . . all presidential democracies were
inspired by the US model (Linz 1994, 45). Following Stepan and
Skach (1994, 128129), under pure parliamentarism the institutional
incentives for politicians are to seek single-party or coalitional majori-
ties, minimize legislative impasses, inhibit the executive from outing
the constitution, thereby discouraging support for military coups in
political society. Under pure presidentialism, there is less incentive to
form parliamentary majorities, this maximizes legislative impasses, exec-
utives are more tempted to out the constitution, and political society
is more encouraged to call for military intervention to surmount im-
passe. Under presidentialism, the president and the parliament are sepa-
rately elected, they can be of different parties yet each claim legitimacy,
and they serve xed terms no matter how unpopular and illegitimate
each may become. Under parliamentarism, a prime minister, or a ma-
jority coalition, are each terminated whenever they lose the condence
of the public. There is evidence that parliamentary regimes are more
long-lived (stably democratic) than are presidential regimes. The author-
itarian menace that Riker strives to blame on the will of the people may
rather originate in the liberal presidentialist constitutionalism that he
recommends.
Rikers populism includes Communism and the popular fascisms, Her
Majestys Government and Sinn Fein, Mahatma Gandhi and Indira
Gandhi, Kim Il Sung and John Rawls, Theodore Roosevelt and Franklin
Roosevelt, John Lennon and Vladimir Lenin, Switzerland and Tanzania,
the younger Riker (1953), in short, every political theory, movement,
and regime that claims legitimacy in the peoples will, rather than in
Gods will, hereditary succession, tradition, or frank elitism. Rikers lib-
eralism includes the older Riker (1982), some but by no means all of
Rikers colleagues and students, the United States Constitution as the
Supreme Court interpreted it in the Gilded Age (Riker and Weingast
1988), William Rusher, an editor at the American conservative activist
journal the National Review (Riker 1982, 15), we may surmise others of
the libertarian constitutional-traditionalist persuasion; but not Madison,
a major architect of the US Constitution.
The distinction between Madisonian and populistic democracy
originates in Dahls (1956) A Preface to Economic Democracy, according
to Riker (1982, 255), who comments that his own usages are quite dif-
ferent from Dahls. The Madisonian US Constitution is the exemplar of
Rikers (1982, 252253) liberalism. Dahl (1956, 28) rejects Madisonian
democracy, because, among other reasons, its zeal to avoid majority
tyranny licenses minority tyranny:
If the freedomof some majority is already curtailed in such a way that only positive
governmental action will eliminate that deprivation, and if a minority with a veto
dislikes the measures proposed to increase majority freedom, then by exercising
its veto a minority can maintain deprivations of the freedom of a majority and
hence can tyrannize over it.
Dahl has in mind social-democratic measures of the twentieth century,
but something similar could be said about the abolition of the right to
hold property in slaves in the nineteenth century. For Dahl, populistic
democracy is a theory that would maximize political equality and pop-
ular sovereignty, but which fatally lacks empirical content. Dahl himself
opts for a third alternative, polyarchal democracy, an operationalization
of populistic democracy that is both measurable and more or less approx-
imately attained in practice.
Rikers liberal Madisonianism is not Madisons, because, unlike
Riker, Madison was not a nihilist with respect to the public good. Riker
admits that his liberalism would not satisfy Madison: random rejection
would generate false negatives, and Madison would have been troubled
by this case (242); random rejection would generate false positives, and
Madison would have believed this case impossible (243). Elsewhere,
Riker (9) claims that Madison was unconcerned by the quality of demo-
cratic decision, whether good or bad, but it is easy to show that Rikers
claim is false. Madison is concerned that transient majorities might dam-
age both the rights of individuals and the permanent and aggregate interests
of the community (The Federalist No. 10 by Madison, in Hamilton, Jay,
and Madison n.d. (1937)/1787, 54). Madison repeatedly appeals to the
public good in Federalist No. 10. Under inuence of their passions, a
majority may enact measures contrary to the permanent interests of the
community including those of a majority faction. Under inuence of their
partial interests, a majority may enact measures that damage the common
good of the community including that of a majority faction (e.g., by fo-
menting factional strife that makes all parties worse off ). Indeed, even
Madison would be an outright populist by Rikers criteria: for Madison,
the advantage of representative democracy is that the assembly would
rene and enlarge the public views, and the patriotism and love of jus-
tice of its members would best discern the true interest of the country,
and not sacrice that interest to temporary or partial considerations. Fur-
thermore, according to Madison, the citizen of the larger republic would
inhabit multiple majorities and minorities, and thus would be reluctant to
permit majorities to invade the rights of minorities. Madison would limit
majority rule so as to promote both the public good and private rights.
Riker is plain we are not reading it into him that there is no public
good. It is an implication of his views that there are no private rights in the
absence of actual unanimity about their content. Thus, there is no yard-
stick of public good or private right by which to compare constitutional
alternatives; thus, there is nothing for his liberalism to promote.
Riker, Burnham, and Pareto
The themes of Rikers Liberalism against Populism are often taught, in my
experience anyway, as the scientic gospel on democracy. This unfortu-
nate circumstance is somewhat due to the fact that many in the political
science profession accept his interpretations of social choice theory, even
as they repress his irrationalist conclusions. I have shown that one need
not accept his irrationalist interpretation of social choice theory. Now I
want to show that Rikers irrationalist doctrine is an old vinegar poured
into the bright new bottles of social choice theory. The major elements of
that doctrine come straight from the musings of the early elite theorists,
Mosca, Sorel, Michels, and especially Pareto, as transmitted by the self-
described Machiavellian James Burnham, a founder with William F.
Buckley of the National Review. Riker seems to borrow many ideas, and
even some phrases, from Pareto, but does not acknowledge or mention
Pareto as an inuence. The one major element of the irrationalist doc-
trine that is not immediately apparent in Pareto the content of Rikers
contrast of liberalism to populism looks like it comes straight from
Burnham, again without acknowledgment.
The overlapping doctrines of Pareto and Riker are of interest because
of their potentially malign inuence on practical politics. The (classical)
liberal Paretos doctrine of the irrationality of democratic politics directly
helped inspire Mussolini, and Pareto ended his life as a liberal fascist. The
liberal Rikers doctrine of the irrationality of democratic politics has been
cited against the democratic struggle in China: political science shows
that democracy is arbitrary and meaningless, so it is better to maintain
paternalistic party rule, the argument goes. This twist of fate is not a com-
plete surprise, since ex-Communist Burnhams Machiavellian innovation
was to adapt Leninist methods to the defense of American interests. Nei-
ther Pareto nor Riker intended the illiberal application of his doctrine,
but ideas do have consequences.
2
Liberal autocracy made an unmistak-
able appearance in Pinochets Chile, is a current of inuence in China
today, and may be what troubles some people about the antidemocratic
excesses of international economic institutions.
Since it is not the details of Paretos theory but the provenance of Rikers
that is my theme, I shall be brief about Paretos ideas. Pareto distinguishes
logical conduct fromnonlogical conduct (I amfollowing Burnham1943,
124133 on Pareto, perhaps Rikers source). Conduct is logical when:
(1) it is motivated by a deliberate end; (2) when the end is possible; and
(3) the means for reaching the end are appropriate. Logical conduct is
common in arts, crafts, sciences, and economic activity. Conduct is not
logical if any one of the three conditions of logicality is absent: if the end
is not deliberate, if it is not possible, or if the means are not appropriate.
Nonlogical conduct is common in the social and especially in the politi-
cal arena. For Riker (1982, 210), there is a strong objective connection
between needs and the consumers private choice, but only the weakest
objective connection between needs and the voters public choice. Riker
(1982, 200206) further contrasts the economic context and the political
context. In the economic context, there is a Pareto optimal compet-
itive equilibrium; everybody is better off relative to where they started;
and even if individuals are dissatised with their endowments, the mar-
ket does not leave its participants worse off. Political or moral scarcity,
however, when contradictory values (such as the dispute over slavery)
are believed by some participants to be universal, differs from economic
scarcity; in politics the choice is among mutually exclusive alternatives
so that there are almost always losers and the losers are usually worse off
relative to where they started. It may seem that Rikers contrasts are only
a weak parallel to Paretos, but permit the case to develop. Riker (1982,
205) continues that the most common kind of political scarcity has to do
with manipulating markets and money:
An assertion of the general virtue of rural life on the family farm justies farm
subsidies. An assertion of the general moral value of the health of communities . . .
justies tariffs, subsidies, and noneconomic government contracts. An assertion
about the general moral value of helping the unfortunate justies a huge vari-
ety of welfare subsidies such as social security. An assertion about the general
moral value of a fair wage justies excluding some laborers from the market
in order to lessen competition for others. Assertions about the general justice
of rewarding inventors, investors, or consumers justify monopolies (almost all
of which are granted or maintained by governments and regulation). Assertions
about the general moral value of labor peace and fair bargaining power justify
the cartelization of labor in unions. Assertions about the moral repugnance of the
spoils system justify grants of permanent tenure to civil servants. Etc.
Compare Pareto (Les Syst` emes socialistes, in Finer 1966, 139140):
The direct production of economic goods is often a very laborious process,
whereas appropriating those goods produced by others is sometimes a very easy
matter . . . some manufacturers produce merchandise of a certain type; through
protective duties on the materials they use, they pay tribute which goes to
other groups of manufacturers, to farmers, merchants, etc. Other tribute is
exacted from them by the circulation of paper money or by government mea-
sures of monetary policy; they pay tribute money to politicians, laying out cash
to maintain certain prejudices which they judge favourable to their interests.
In compensation they receive tribute from consumers in the shape of protec-
tive duties on foreign products which might compete with theirs, and from
the workers through the issuing of paper money or through measures taken by
the government to prevent the workers from freely negotiating the sale of their
labour . . . the most paltry reasons nd acceptance when they serve powerful in-
terests or minister to xed inclinations . . . most men make convictions of their
interests.
Economics is rational, politics is irrational; politics is how the parasites
expropriate the producers; conviction is constructed from interest on
these points Riker and Pareto agree.
The two major features of Paretos system of sociology are residue the-
ory and elite theory. Each is found updated in Rikers Liberalism against
Populism. What is Paretos residue theory? For Pareto, derivations are the
varying rationalizations and verbal associations connected to constant
residues. Residue is a sociological concept not a psychological one, but
residue can be thought of roughly as typical human sentiments. Deriva-
tions differ from country to country and from era to era, but the underly-
ing residues remain the same (Burnham 1943, 134145). Riker (1980a,
433, emphasis added) seems to be Paretian when he seeks to explain the
appeal of the supposed illusion that democracy is a device to combine
individual values into decisions of government:
the contemporary force associating individual values and social outcomes is
wholly secular, though probably derived. . . from Christian modes of thought.
In the ideology of democracy, which may well be a kind of secularized Christian
theology, that form of government is often, though I believe quite inaccurately,
dened as the rule of the people . . . this picture of democracy is internally incon-
sistent and cannot be sustained. . . Nevertheless inconsistencies and inaccuracies
do not deter most ideologues.
To translate further the notions of residue and derivation interests al-
ways determine principles, principles never determine interests; moral
discourse, or appeals to the common good, ultimately reduce to self-
serving cant if not hypocrisy, for both Pareto and Riker. Rikers (1982,
224) cynical analysis of the American Civil War insinuates that both the
supporters and the opponents of slavery were primarily motivated by po-
litical opportunism, for example. Pareto (Les Syst` emes socialistes, in Finer
1966, 137) conrms generally that what historians depict as a battle for
liberty is merely the clash of competing elites; historians believe . . . that
the elite which in reality is seeking to get hold of power to use it and
misuse it in just the same way as the elite it is opposing is moved only by
pure love of its fellow men. Riker (1982, 221) acknowledges that moral
concerns can coincide with political interests, Pareto (1963, 1,295, sec-
tion 1859) that many adherents of the democratic religion are sincere,
although for each such is merely the sincerity of the deluded.
3
What is Paretos elite theory? All the apparent variety of political his-
tory reduces to the circulation of elites. History is not an unfolding ad-
vance of the democrats against the oligarchs, it is the perpetual struggle of
one elite against another. Using Paretos (Les Syst` emes socialistes, in Finer
1966, 134) vocabulary, suppose that A is the elite in power, aspiring elite
B competes with A for the loyalty of the masses C; once A is defeated and
B is in power, then aspiring elite D arises, and so on. Rikers political the-
ory posits the identical image of an unstable equilibrium of institutions.
For Riker, disequilibrium is the characteristic feature of politics, even
for apparently stable institutions such as the American Constitution
the American Civil War came about because the political losers in the
minority commercial faction of antebellum American politics sought for
sixty years an issue that would split the majority agricultural faction, and
found an adventitious one in slavery to generate the cycle they needed
for ascendance, consolidated by force. True, Riker (1980a, 445) con-
cedes, institutions might temporarily stabilize the chaos of aggregated
tastes, but institutions are only congealed tastes subject themselves to
instability. Compare Pareto (Treatise on General Sociology, in Finer 1966,
254): Every individual . . . endeavours to obtain a maximum of indi-
vidual utility . . . an innite number of positions of equilibrium with the
requirements of individual maxima of utility becomes possible. Public
authority intervenes to impose some and exclude others. Because of po-
litical disequilibrium, political outcomes are not the will of the people,
they are rather the will of the smarter, bolder, more powerful, more cre-
ative, or luckier people, says Riker (1982, 200). Paretos (Les Syst` emes
socialistes, in Finer 1966, 134) aspiring elite is distinguished by energy,
character, intelligence. The virtue of liberal constitutionalism, accord-
ing to Riker (1982, 253) in the concluding paragraph of his volume, is
that it guarantees some circulation of leadership so that great power is
usually eeting and no vested interest lasts forever. Riker (1980a, 443)
says that political institutions are established by force, not by the summa-
tion of wills. Pareto (Les Syst` emes socialistes, in Finer 1966, 136) says that
for right or law to have reality in a society, force is necessary. Interests
determine principles, institutions originate in force, political history is
the circulation of elites the echo of Pareto in Riker, not only in ideas,
but in phrasing, is remarkable.
Each proclaims his views in a relentlessly scientistic tone. Burnham
(1943, 124) on Pareto:
He is not offering any programme for social improvement nor expressing any ideal
of what society and government ought to be. He is trying merely to describe what
society is like, and to discover some of the general laws in terms of which society
operates. What could or should be done with this knowledge, once obtained is a
question he does not try to answer.
Such a judgment belies Paretos lifelong political agitation. Riker (1980a,
432) urges a science of politics, and the essence of science is, of course,
the accumulation of more or less valid generalizations (1990b, 166).
Riker (1980a, 446) states that his political theory appears to be mathe-
matically irrefutable. Each is passionate and oddly bitter in his denunci-
ation of the democratic ideology (Riker) or the democratic religion
(Pareto).
4
Another remarkable parallel is Rikers duplication of one of Paretos
lesser-knowndoctrines. Pareto distinguishedthe objective utility of a com-
munity, its survivability in terms of political and military power, from the
subjective utility for a community, the individual welfares of the commu-
nitys members. Objective utility and subjective utility seldom coincide,
says Pareto; for example, conducting war decreases the subjective utility
of the community but increases its objective utility. Riker (1982, 291):
the notion of a public interest, so cherished by populist [democratic] propa-
gandists, is not, technically speaking, rendered meaningless simply because the
populist interpretation of voting is meaningless. A public interest is an interest
attached to the collective body of society; and as long as society exists, it has,
presumably, some interests, which are its common or public interests . . . By def-
inition, however, a common or public interest is held in common, so voting is
unnecessary to reveal it . . . A public interest may even exist when people do not
agree. There really may be an objectively right but not indisputably evident policy
for the society . . . the public interest cannot be revealed by nonunanimous voting.
Unanimity is Pareto optimality (Riker 1982, 117); only unanimous gov-
ernment decisions but any voluntary market exchange satisfy the Pareto
criterion.
If voting is always irrational and public discussion is always opportunist
then that leaves only unanimity or force as the means for deciding public
policy; as Pareto (A Few Points Concerning a Future Constitutional
Reorganization, in Bucolo 1980, 273) put it, strength and consent . . .
are the foundations of government, note the absence of public reason
in Pareto just as in Riker. Riker seeks to justify liberalism as a defense
against the thoroughgoing irrationality of politics, but fails to realize that
liberalismis not the unique conclusion fromthe irrationalist premise. The
alternative conclusion is that if democracy is irrational and fraudulent,
then those with energy, character, and intelligence should impose their
interpretation of the objective public interest by force, and Pareto even-
tually shifted from the liberal conclusion to the alternative conclusion.
Pareto, more than anyone, imported the concept of equilibrium into
social theory, as an alternative to organicism. Riker accepts equilibriumin
the economic sphere, and his advance on Pareto is a claimto demonstrate
permanent disequilibrium in the political sphere, yet thereby vindicating
the masters judgment as to the irrationality of politics. Rikers core dis-
tinction between liberalism and populism is not immediate in Pareto,
but does appear in Burnhams transmission of Pareto and the remaining
Machiavellian elite theorists. Burnham (1943, 174175, 180, 182):
Democracy is usually dened in some such terms as self-government or
government by the people. Historical experience forces us to conclude that
democracy, in this sense, is impossible. The Machiavellians have shown that the
practical impossibility of democracy depends upon a variety of factors . . . The
theory of democracy as self-government must, therefore, be understood as a
myth, formula, or derivation. It does not correspond to any actual or possi-
ble social reality . . . The truth is, however, that there are other meanings com-
monly associated with the word democracy, which have nothing to do with
self-government . . . democracy means a political system in which there ex-
ists liberty . . . The crucial difference that freedom makes to a society is found
in the fact that the existence of a public opposition (or oppositions) is the only
effective check on the power of the governing elite.
Compare Riker (1982, xviii, 242, 245):
The populist interpretation of voting (i.e., that what the people, as a corporate
entity, want ought to be public policy) cannot stand because it is inconsistent
with social choice theory. If the outcomes of voting are, or may be, inaccurate
or meaningless amalgamations, what the people want cannot be known. Hence
the populist goal is unattainable . . . The essence of the liberal interpretation of
voting is the notion that voting permits the rejection of candidates or ofcials
who have offended so many voters that they cannot win an election. . . the liberal
veto generates freedom because of the very fact that it is a curb on tyranny.
Burnham (1943, 176) holds that the aw of populism is its degeneration
fromparliamentary democracy to Bonapartism: if it is the will of the peo-
ple that justies the extension of suffrage and parliamentary supremacy,
then that same will of the people can justify abandonment of the demo-
cratic institutions. For Riker (1982, 249), with a populist interpretation
of voting it is easy for rulers to believe their programs are the true will of
the people and hence more precious than the constitution and free elec-
tions. Each is caught in a conundrumby his rejection of the rationality of
public deliberation in attaining the public good. Burnhams contribution
to the war effort against Fascism was a celebration of the elite theorists
whose doctrines contributed to the intellectual formation of the totalitar-
ian ideologies; Burnham does not once mention that Mosca and Pareto,
his defenders of freedom, welcomed Mussolinis Fascism. It is down-
right strange to oppose Bonapartism by appeal to its most prominent
apologists, and Burnham was too clever to have overlooked this paradox.
There seems to be a hint of esoteric glee in Burnhams construction.
5
Neither does Riker acknowledge Pareto or Burnham as forebears.
Pareto, Mosca, and Riker each have the best of liberal intentions. Pareto
did value liberty and probably would not have remained loyal to the com-
plete trajectory fascism followed after his death in 1923. But his love for
liberty was blinded by a hatred for democracy. The Pareto (AFewPoints
Concerning a Future Constitutional Reorganization, in Bucolo 1980,
275) that Burnham recommends to us recommended to Mussolini that
he preserve freedom of speech but merely the appearance of democracy:
The only aim must be that of freeing oneself from the democratic ideologies of
the sovereignty of the majority. Let this sovereignty retain its shadow it atters
powerful emotions but let the substance pass to an elite for the objective good.
If he had lived, perhaps Pareto would have seconded Moscas (quoted
in Albertoni 1987, 10, emphasis added) sentiments in a speech to the
Italian Senate in 1925: I who have always been sharply critical of par-
liamentary government must now almost regret its fall. Or, who knows,
if he had lived even longer, perhaps he would have applauded the an-
tidemocratic views of the Chicago boys in Chile, for whom the virtue
of authoritarianism was that it permitted the scientically correct pol-
icy. Said one of them in defense of the military dictatorship: A positive
science with ideology ceases to be a positive science; ideology which is
only positive science does not have an element of ideology (Barahona,
quoted in Barber 1995, 1946). Riker calls his enterprise positive political
theory and those of his theoretical opponents ideology.
I rmly believe that Rikers intentions are wholly liberal. What I amsay-
ing is that one may innocently endorse a doctrine which unforeseeably
necessitates consequences one would not endorse. Some fellow travelers,
as Burnham and other anti-Communists pointed out repeatedly, irre-
sponsibly endorsed doctrines of peace and freedom which had the unin-
tended consequence of furthering violence and oppression. I suggest that
the doctrine of democratic irrationalism may have illiberal consequences
in the world that its liberal adherents do not intend.
18 Democracy resplendent
Introduction
In this nal chapter, I recommend other scholars to those who want
a more formal approach to these issues. Next, I show that all the in-
stability and manipulation results for the polity have parallels for the
economy, but that there is a double standard which endorses the re-
sults for the polity but rejects them for the economy. Finally, I return
to the hall of quotations, with answers to the new academic attack on
democracy.
Those looking for a more formal approach can turn for complemen-
tary insights to Sen and his constructive social choice theory, beginning
with his Nobel Lecture (1999). Sen reports that the rst response to
Arrows theorem was, in politics pessimism about democratic decision
making, and in economics despair about evaluating social welfare. The
background to Arrows theorem was Robbinss incredible claims that ev-
ery mind is inscrutable to every other mind and that no common denom-
inator of feelings is possible. Sens diagnosis, made in many rich formal
contributions over several decades, is that the impossibility is due to un-
justied informational restrictions: It is not surprising that the rejection
of interpersonal comparisons must cause difculties for reasoned social
decision, since the claims of different persons, who make up the society,
have to be assessed against each other (365). He also points out that
Arrows original impossibility result should be no surprise, as in aiming
to identify a unique rule one may undershoot and yield multiple possi-
bilities, or one may, as did Arrow, overshoot and yield none. Saari, in
advanced (2000a; 2000b) and in introductory (2001a; 2001b) texts, also
provides comprehensive and innovative perspectives on the problems of
aggregation. The current and forthcoming work of Christian List (e.g.,
Dryzek and List 2003, List 2001, 2002, 2003, List and Goodin 2001)
also merits attention.
432
Democracy resplendent 433
Instability: neither everywhere nor nowhere
Rowley (1993, xiii), in my hall of quotations, spoke most condently
about howArrows theoremprovides incontrovertible support for market
process and for the minimization of democratic government. W. Dean
Burnham (1999, 2250), with less enthusiasm, concludes that, In pol-
itics, unlike in economics . . . virtually no naturally occurring equilibria
exist. Kuttner (1996, 333345), in his book on the virtues and limits
of markets, says that such a view is widespread. In public choice theory,
according to his muscular rhetoric:
the demonized state makes an almost perfectly Manichaean mirror image of the
idealized market. The sacred economy is at constant risk of being violated by a
profane polity. The core claim is that systematic error and opportunism are as
endemic and logically inevitable in the political enterprise as self-purication is in
the marketplace. That premise then gives Public Choice theorist an all-purpose
trump to any demonstration of market failure: Yes, the market does perhaps fail
fromtime to time, but political interference will only make it worse. Public Choice
theorists, in their zeal to impeach economic intervention, go further and impeach
democracy itself. (333)
He explains that these conclusions are deduced fromaxiom, logical infer-
ence, and extrapolation of the market model. The presumptions lead to
a series of syllogisms that supposedly prove that politics leads to chaotic,
rapacious, or perverse outcomes. The inevitable conclusion is that the po-
litical realm should be made as narrow as it can be. He says that Arrows
impossibility theorem is the Rosetta stone of public choice and is cited as
if to demonstrate once and for all the futility of political efforts at social
betterment. The celebration of the market has become an insidious form
of contempt for political democracy, he says, and public choice poses as
an expert witness for the claim that political intervention in the economy
should be minimized.
The typical student of philosophy, politics, and economics these days
would be taught, rightly on my view, the advantages of the market econ-
omy as compared to the command economy. In the worst case, the stu-
dent would only be taught propaganda about the miracle of the market,
but in the typical case she would be taught in a nuanced way about market
successes and also about market failures: monopoly; undersupply of pub-
lic goods; incomplete markets; externalities; information failures; busi-
ness cycles, unemployment, ination, and deation. In the best case the
student might debate some normative problems of the unmodied mar-
ket economy, such as unjustied inequalities in social ofces and product,
unfair disparities in bargaining power, invasion of liberties by private cor-
porations, displacement of family and friendship, expropriation of desires
by advertising, and so on. One of the main points of the public choice
movement is that welfare economics uncritically assumed that democratic
government would unproblematically rectify market failures. They argue
that government failure is as possible as market failure. This is a valuable
point, but possibility does not mean necessity. The destructive response
is to prohibit the democratic government fromaddressing market failures
(actually, to prohibit it from correcting all market failures except for the
protection of private property). The constructive response is to assess the
actual likelihood of government failures and from theoretical and empir-
ical considerations to devise institutions, beyond undiscerning inaction,
that would minimize failures of both market and government.
Although a successful democracy protects its citizens from military
invasion, civil war, autocracy, market failures, poverty, criminality, and
other injustice, these days one seldom hears anything about the mira-
cle of good governance, despite the conspicuous variations in govern-
ment performance across the world. The students in Harvards core
course in politics are told that politicians are venal, immoral, disgust-
ing scoundrels (Shepsle and Bonchek 1997, 5). If the market economy
works better than the command economy, does democracy work bet-
ter than autocracy? Not necessarily. Shepsle and Bonchek (67) continue
that the choice is between incoherence and dictatorship: There is, in
social life, a tradeoff between social rationality and the concentration of
power. These are disturbing teachings. Would autocracy really be more
socially rational than democracy? Just as it is an error to compare an
actual market with an idealized government, so is it an error to com-
pare an actual democracy with an idealized dictatorship. If egomaniacal
redistributional instability is possible in a democracy, even contrary to
its commitments to fairness, then instability is even more possible, and
certainly more dangerous, in a dictatorship which lacks democratic com-
mitments to fairness. Although under a dictatorship there is no longer any
Condorcet-rule voting creating the potential for cycles, certainly there are
many possible military coalitions who would want to depose the dictator
and seize the social ofces and product each for itself. If there are insta-
bilities, they afict not just democracy, but also dictatorship, and also the
market.
The student who is taught the doctrine of democratic irrationalism is
not taught the parallel ndings about the economy. The prejudice is per-
vasive. The third fundamental theorem of welfare economics, Arrows
impossibility theorem, is stated pessimistically rather than optimistically:
dictatorship is the only possible social-welfare function, and so forth,
as we have seen in this volume. A more optimistic version of the third
theorem would state that given individual orderings there are many
possible social welfare functions that would yield a social ordering, and
a few good ones. The optimistic version would go on immediately to
state qualications. It would say that if voters were wholly selsh and
lacked any preferences for fair outcomes, then redistributional insta-
bility would follow with Condorcet pairwise voting. It would go on to
say if Condorcet voting were to be excluded for any reason, then any
other social-welfare function for three or more alternatives would re-
quire more information than is available from pairwise comparisons. It
would note in a jocular aside that if pairwise Condorcet voting were ex-
cluded, yet pairwise comparison insisted upon, then the only remain-
ing voting rule would be the dictatorship of one. It would not write the
Condorcet paradox of voting on the blackboard and declare democracy
to be meaningless.
The student is rarely taught that the Arrow impossibility theorem ap-
plies to the economy as well as to polity, although Riker and his students
are aware of the point. Most strikingly, Riker and Ordeshook (1973)
state that the market allocates resources in nontransitive ways (85), and
continue that they are not worried by the Condorcet paradox of voting,
because:
People are not invariably disturbed by the inconsistencies and incoherencies of
market outcomes such as the oft-discovered fact that society spends more on
liquor than education although surely a majority would wish otherwise. Markets
have been churning out such inconsistencies for centuries without leading us to
reject them as useful tools. (114)
Perhaps this passage is due to the inuence of Rikers coauthor, as the
message of Rikers later Liberalismagainst Populism(1982) is not so placid.
Ordeshook (1982) contested Rikers later hypotheses of the pervasive dis-
equilibrium of politics. Ordeshook responded that the presumed stabil-
ity of markets resides principally in our abstract description of them and
not necessarily in reality (26) and that social-choice results do not prove
something unsavory or even disturbing about democratic processes in
particular and political processes in general (31). In the same volume,
Fiorina and Shepsle (1982) suggest that the differing emphases on equi-
librium in economics and on disequilibrium in political science are due
to professional incentives. In economics, only equilibrium-preserving
extensions of models are of interest (i.e., publishable) (60). Positive po-
litical theory, however, follows in the footsteps of Arrows impossibility
theoremin emphasizing disequilibrium. Presumably, models showing po-
litical equilibrium and democratic possibilities are not as publishable as
those proclaiming paradoxes and impossibilities. It could have been oth-
erwise, they suggest: Political theorists might have decided early on that
unidimensionality was a basic assumption of all political models, akin
to the regularity conditions imposed on consumption and production
sets by economists (Fiorina and Shepsle 1982, 6061), and economists
could have followed up on an early observation that instability seems
to be a universal phenomenon in competitive economies rather than an
exceptional one. Further, they insist, as I have, that Equilibria and dise-
quilibria are properties of models. It remains to be demonstrated whether
they are descriptive of empirical phenomena (62). In the same volume,
Schoeld (1982), in response to Rikers observations about the possibil-
ities of manipulation in the polity, reviews possibilities of manipulation
in the price system, and explores further parallels between political and
economic instability. I applaud all this good sense, but I am afraid that
it is not often declared in the rhetoric and pedagogy of the doctrine of
democratic irrationalism.
In contrast to the pessimistic phrasing of the third fundamental the-
orem, the rst fundamental theorem of welfare economics is phrased
optimistically, as a marvelous possibility: the economy is in competitive
equilibrium and the equilibrium is Pareto optimal, or, laissez faire leads
to the common good (Feldman 1991). Riker (1980, 434) celebrates
the scientic and intellectual success of the price theorists; he says their
discovery of the competitive equilibriummakes economics the most pres-
tigious of the social sciences. The rst theoremis a formalization of Adam
Smiths invisible hand. In the equilibrium each consumer maximizes her
utility given her budget constraint, each rmmaximizes prots given mar-
ket prices, and the market for each good clears (there are no shortages or
surpluses). Yes, the theoremis taught in a nuanced fashion; qualications
are immediately stated. The theorem holds only if each agent is selsh,
only if each agent is a price-taker rather than a price-maker, and only if
each agent knows all prices for all goods. It is sometimes acknowledged
that Pareto optimality is a troublesome welfare criterion: those who start
off rich end up rich, those who start off poor end up poor, and non-
exchange transfers from rich to poor are prohibited because it would
make the rich worse off by the Pareto criterion.
Why not phrase the theorem in a pessimistic fashion? We could ob-
serve that not all agents are price-takers, there are monopolists. We could
observe that not all agents are selsh, that some care about what hap-
pens to others, or observe that there are many other externalities. We
could observe that it is the rule and not the exception for agents to have
asymmetric information about goods and prices. We could observe any
of these facts about the actual economy, and then go on to state that high
economic theory proves that, given a number of innocuous conditions,
there is no competitive equilibrium in the economy.
Even small information costs can have large consequences and many of the stan-
dard results including the welfare theorems do not hold even when there are
small imperfections of information. . . whenever information is imperfect or mar-
kets (including risk markets) are incomplete that is, essentially almost always
competitive markets are not constrained Pareto efcient. (Stiglitz 2000)
1
The market is arbitrary and meaningless. The evidence is all around
us, witness depressions, speculative bubbles, involuntary unemployment,
useless consumerism, have we got some stories! Dot-com entrepreneurs
who waste other peoples money are paid a thousand times what teachers
or nurses are paid for taking care of human beings! Americans buy un-
safe gas-guzzling SUVs that waste a nonrenewable resource, contribute
to global warming, make them international pariahs, and involve them
abroad with detestable autocracies! That would be an unhelpful interpre-
tation of the rst theorem, but I make it to illustrate that there is a double
standard: the market is presumed to be stable and good, and democratic
governance is presumed to be unstable and bad. Similarly, elsewhere
in public choice, markets are presumed to be efcient, no matter how
far-fetched or unempirical the explanation required to excuse monopo-
listic behavior or other economic asco, and democratic governance is
presumed to be inefcient, even if empirically vindicated (Hovenkamp
1990b; and see Wittman 1995).
What else? The competitive equilibrium assumes that there is no force
or fraud on the market. Homo economicus will haggle to death over price
but will never take what he wants by force; he operates ruthlessly within
a strictly dened bubble of sainthood, according to Skaperdas (2002).
If the constraints of peace and honesty are removed, then the pursuit
of material self-interest degenerates into a political economy of lord and
serf. One agent specializes in forceful expropriation, and gets more com-
pensation, and the other agent specializes in production and gets less
compensation, says Skaperdas. I add that force and fraud are potential
threats on markets, but are generally low in incidence; similarly, unfair-
ness is a threat in democratic voting, but is generally lowin incidence; and
presumably there are various moral and material incentives in each realm
that constrain such threats. Nevertheless, its considered legitimate for the
market model to assume the absence of force and fraud, but illegitimate
for the democratic model to assume a minimal concern for fairness.
The competitive equilibrium of the economy stated in the rst theo-
rem is static, not dynamic. What students are usually not taught about
the economy is that it is possible that there is no dynamic process that
leads to the competitive equilibrium, or, worse, that it is possible that
the competitive equilibrium is unstable in that the smallest perturbation
leads the economy away fromit and into a cycle of prices or into complete
chaos. Our friend, the cycle, as a matter of logical possibility, aficts the
market as it aficts voting. Scarf (1960) began from the standard setup
individuals with utility functions for the same commodities, who trade
beginning from initial endowments of those commodities, and respond
to the announcement of an initial vector of prices, and if that vector is
the competitive equilibriumthen all markets clear. The rst theoremsays
nothing about what happens if the initial vector of prices is not the com-
petitive equilibrium. Scarf assumed three agents and three goods, and
assumed utility functions obtained by cyclic permutation of the goods
and the initial endowments. What happens is a process such that prices
fall away from the unstable equilibrium and revolve in an endless cycle.
We could apply ideas from the doctrine of democratic irrationalism and
argue further that because such a cycle could happen in any one instance,
it could happen in all instances, and thus that preferences on the market
are unknowable, and hence that market outcomes are meaningless, etc.
It is logically possible that the dynamic economy is in complete chaos,
paralleling the McKelveySchoeld model of voting chaos. The math-
ematician Donald Saari (1995b) says that standard price-adjustment
models admit highly chaotic behavior:
I have no idea whether Adam Smiths invisible hand holds for the real world,
but, then, no one else does either. This is because, even though this story is used
to inuence national policy, no mathematical theory exists to justify it. Quite
to the contrary; what we do know indicates that even the simple models from
introductory economics can exhibit dynamical behavior far more complex than
anything found in classical physics and biology. (222)
Saari explains what is called the Sonnenschein, Mantel, and Debreu the-
orem in economics. If there are at least as many agents as commodities
in the economy, then there exist endowments and individual preferences
such that anything can happen! (224, emphasis added). Demand could
go up and price would go down. Saari goes on to deny that this insta-
bility nding is due to implausible preferences or to an overly simplistic
model. He does nd, in what he believes to be a more comprehensive
model, that while an unregulated free market might not work as widely
advertised, if correct regulations are imposed, the market now might be-
have as desired (227). To further the mischief, I would add that perhaps
multidimensional instability is more tractable in democracy than in the
market, because in a democratic setting any one person can move to di-
vide the question and thus reduce to single dimensions the issues under
consideration.
The irrationalists main point is not the inconsistency of the polity, but
its manipulability and consequent meaninglessness, it may be objected.
Unlike democracy, the market is not susceptible to such manipulation.
Wrong. First, just consider the process of buying a used car. Second, as an
abstract possibility, the problem is general in the economy, but is claimed
not to be of wide practical importance. The GibbardSatterthwaite the-
orem about manipulation by strategic misrepresentation of preferences
does not directly apply to market mechanisms, because economists do not
demand the universal domain condition for the economy. Satterthwaite
(2001) argues nevertheless that, no attractive social choice functions ex-
ist for markets that are both strategy-proof and efcient. Market theory
disposes of the problem by assuming that all agents are price-takers and
thus nonstrategic.
He illustrates market manipulability with an example. The cost of a
product is 0.55 for a seller, its value 0.8 for a buyer, and since value
exceeds cost, efciency demands trade. The Walrasian auctioneer an-
nounces a price of 0.5, the seller declines, and the buyer is still will-
ing. The auctioneer announces 0.6, but the seller decides to hold out
for 0.7, and the buyer is still willing. A price-taking seller would have
stopped at 0.6. The auctioneer announces 0.7, and trade is consum-
mated at that price. But what if the sellers value had been 0.68? Then the
buyer would not have been willing at 0.7, and there would have been no
trade. There is both an incentive for misrepresentation and a potential for
inefciency.
It is interesting how such a problem is addressed in the discipline of
economics. Satterthwaite observes that people often act as if markets
were strategy-proof, indicating the empirical implausibility of the ma-
nipulability result. He presents a model of a double auction as a Bayesian
game, which shows that as the number of agents in the market increases,
equilibrium strategic behavior decreases rapidly toward zero and full ef-
ciency is rapidly approached. Numerical examples suggest that a market
of size 8 or better would be approximately, not exactly, strategy-proof.
Thus, he argues, price theory is justied to ignore the possibility of manip-
ulation, and exact strategy-proofness is too strong a requirement. Scarf s
exposition of the market cycle suggests as one interpretation that such
cycles actually happen, but alternatively suggests that the model is not
realistic, or that similarity of utility functions would avoid the result.
In positive economic theory, theoretical bads are trumped by empirical
goods, and empirical bads are trumped by theoretical goods. In positive
political theory the obverse obtains.
I like how Saari (2001b) narratively generalizes the aggregation prob-
lem. Such problems are everywhere: the market, voting, sports or scholar-
ship or any kind of nonmarket and nonpolitical ranking, engineering
decisions, individual choice among multiattribute alternatives, some
statistical manipulations. With respect to engineering, for example, a
complex project is analytically decomposed into a number of simpler
tasks, and then the results of the simpler tasks reassembled. In the pro-
cess of analysis and synthesis, however, crucial information can get lost,
resulting in perverse syntheses. That is just what happens with Arrows
independence condition: it requires social decisions to ignore the full in-
formation available from individual orderings, and from discussion, and
insists on only the diminished pairwise information from them. The in-
dependence condition is like a reader who only counts letters, rather than
considering their relationship to one another: 3-t, 3-a, 2-s, 1-h, 1-m, 1-i,
1-k, 1-e, 1-.
2
Yes, a well-informed person should be aware that poorly
designed aggregation can yield perverse results, but she should also be
aware that well-designed aggregation can yield useful results.
Talking back in the hall of quotations
Think back to the hall of quotations presented in Chapter 1. Among the
voices, Wolff (1970) initially seems to be the most consistent. He says
that the Condorcet paradox of voting infects all democratic social choice,
and suggests that there is no alternative but to embrace the doctrine of
anarchism. Rowley (1993), to avoid the sting of Arrows impossibility
theorem, recommends a minimal state rather than no state, and a full
market. Shepsle and Weingast (1984) argue that the cycling legislature
cannot reliably correct market failures. Tribe (1988) recommends a fuller
state, but suggests that because of the Arrowtheorem, courts knowbetter
than legislatures. Tushnet (1988) corrects Tribe, noting that the Arrow
theorem would apply to the courts as well. If judicial guardianship fails,
that would seem to leave nothing but the market, but we have seen in this
chapter that the market is tainted as well. Thus, even Wolff falls short.
Perhaps the most consistent position would be to abolish both state and
market? But that would violate Condition P (if everyone prefers Metallica
to AC/DC, then society prefers Metallica to AC/DC). All social life is
impossible, according to consistently extended irrationalism.
Katznelson and Milner (2002) say that the fall of the Weimar Republic
and other democracies, and the rise of Fascismand Bolshevismin the rst
half of the twentieth century, were due to the democratic instability shown
by Arrow. However, if democracy is inherently unstable, but dictatorship
is not, then howis it that Fascismand Bolshevismare gone fromthe scene,
but many democracies remain? W. Dean Burnham (1999) ascribes the
rise of the Nazis and, in contrast to Katznelson and Milner, the fall of
the Bolsheviks, to political disequilibrium. The Soviet regime lasted some
75 years, however, hardly unstable compared to typical regime durations.
Observe that the disequilibrium hypothesis explains everything and
nothing.
Runciman (1963) says that the intuitive criteria of the Arrow theorem
showthat strict democracy is impossible, Tuck (1993) that Arrowshowed
scientically that the programof the citizens making social decisions does
not make sense, Cain (2001) that all voting systems have some norma-
tive blemish, and Samuelson (1977) that an ideal voting scheme cannot
possibly be found. There is no ideal voting system, in the same sense,
however, that there is no ideal dinner, no ideal residence, and no ideal
holiday, simply because there are always tradeoffs among desiderata. Even
if one voting rule is close to ideal, anyone can propose a new desideratum
that the scheme is bound to lack. Given a set of desiderata, we are able to
say that some voting rules are better or at least as good as some others,
and to say why. It is especially disappointing that economists would be
surprised to nd tradeoffs in the choice of one voting rule over another
in various circumstances, since emphasis on tradeoffs and constraints is
a hallmark of that discipline.
Riker and Weingast (1988) say that cycles are ubiquitous, Plott (1976)
that cycles are the case not the exception, Sunstein (1988) that cycles
make accurate aggregation highly unlikely, and Katznelson and Milner
(2002) that instability is an immanent feature of liberal democracy. This
volume studied simulations, actual preferences, and anecdotal allegations
and found a nearly complete absence of cycles. Further, it offered the-
oretical explanations for the rarity of cycles, suggested that most cycles
which might occur would be of trivial consequence, and that there exist
defensible voting methods that avoid cycles. Hardin (1993) says that no
government of a complex society is likely to be coherently democratic, Bell
(1974) that public decisions have no rationality, Przeworski (1991) that
voting results do not identify any unique social preference, and Shepsle
and Bonchek (1997) that its nearly impossible to arrange for the making
of fair and coherent group choices. Yet there are no systematic or casual
observations of the radical instability predicted by irrationalist theory.
Riker and Weingast (1988), Mashaw (1989), Sunstein (1988), and
Cain (2001) warn of the ubiquity of strategic voting and agenda control,
but we have seen that these are only of consequence when institutions
unfairly grant some actors more formal power than others, directly con-
trary to the democratic ideal, and that such defects are remediable. Next,
who knows how many boxes of chalk and barrels of ink have been ex-
pended on explications of McKelveySchoeld multidimensional chaos
(from [1976] on, political science as a discipline faced no more press-
ing challenge than to interpret and incorporate these profound instabil-
ity results, G. Miller 1997, 1,185)? Yet the simple parliamentary rule
allowing any one member to request division of the question disposes of
the problem, as does strategic voting in many circumstances, constraints
on leadership, or a switch from Condorcet voting to Borda or other
methods.
Nevertheless, Riker and Weingast (1988) say that there is a fundamen-
tal arbitrariness to social choice under majority rule, Mashaw (1989)
maintains that literally anything can happen when votes are taken and
that apparently democratic decisions are the artifact of decision pro-
cesses controlled by manipulators, Cain (2001) that we cannot validly
infer anything about the preferences of the society based on laws pro-
duced by a legislature, and Shaviro (2000) that legislative enactments are
random and purposeless. This is a startling hypothesis and one is entitled
to ask: Are there any demonstrations that actual democratic decisions
from a proper sample are uncorrelated with the preferences of the voters
making those decisions? There are not. Feldman (1980) doubts all as-
sertions about a general will, a social good, or even a social benet; Plott
(1976) says that the public good cannot, in principle, exist; Ordeshook
(1986) that there is no public interest or community goals; and Riker
and Weingast (1988) that the will of the people has no meaning. If so,
then no one, including scholars of the irrationalist bent, would be jus-
tied in recommending to us social institutions or social policies of any
kind. Katznelson and Milner (2002) insist on a tradeoff between stability
and democracy, Arrow (1963/1951) claims that in the noncomparabilist
framework dictatorship is the only satisfactory social-welfare function,
and Shepsle and Bonchek (1997) suggest that only permitting dictator-
ship avoids social irrationality. Is Arrows independence condition, that
all voting rules should proceed only by pairwise comparison of alterna-
tives, more normatively compelling than the choice of democracy over
dictatorship?
Conclusion
The old academic attack on democracy in the late nineteenth century and
early twentieth century, especially by the elite theorists Mosca, Michels,
and Pareto, contributed to the retreat of democracy and the rise of
Fascismand Bolshevism. If democracy is impossible and fraudulent, then
superior individuals should impose the objective good, is the conclusion
that others drew from the elitists hard-headed theories. There is a lim-
ited range of reexivity in social life, such that, for instance, a power-
ful belief that democracy is impossible and fraudulent creates the situ-
ation it denes. Elitism was eventually refuted, and democracy revived,
by the fruits that each bore. The new academic attack on democracy
resuscitates discredited elite theory with the formally authoritative tools
of social choice theory.
The new academic attack on democracy fails, theoretically and em-
pirically. The irrationalist interpretation of the Arrow theorem and as-
sociated social choice results is one of the bigger intellectual errors of
the second half of the twentieth century. Its long, dark shadow over
democratic politics is now lifting. Democracy is on the march in the
world today. The Chinese students constructed a goddess of democracy
in the days before their blood stained the agstones.
3
The statue shines
bright in images of the demonstration. For all those who battle against
tyranny and for democracy, know that in theory, too, democracy shines
resplendent.
Endnotes
2. 1nc boc1ni Nc or bc:ocna1i c i nna1i oNaLi s:
1. Rochester school refers to an intellectual tendency, not a place. It should
not be assumed that former or current students or faculty from the Rochester
Department share all or any of Rikers views.
2. Quoted from http://www.rochester.edu:8000/college/PSC/intro/history.php
3. Quoted from http://www.rochester.edu:8000/college/PSC/graduate/intro.php
4. For a rational-choice account of political leadership, particularly the chairing
of a political science department, see Shepsle and Bonchek (1997, ch. 14).
Compare to Jane Mansbridges (1994, 156) remarks on public-spiritedness
and chairing of an academic department.
5. I belong to the Public Choice Society, which welcomes scholars of all varieties
to its ranks.
6. See Brennan and Lomasky (1993) for a profound examination of the paradox
of participation.
7. See Cox 1999 for an alternative defense of rational choice, and generally for
examples of healthy rational-choice research.
8. The exemplar of constructive social choice theory is Amartya Sen. See his
Nobel lecture (1999).
3. i s bc:ocna1i c vo1i Nc i Nacccna1c:
1. See Grofman and Feld (1988) and Grofman and Owens (1986) symposium
on information pooling.
2. See Levin and Nalebuff 1995 on further pragmatic criteria for selecting a
voting rule.
3. See Grofman and Reynolds (2001) for a recent inventory of main ndings
on electoral systems.
4. There is an emerging public discourse in America on alternatives to plu-
rality elections. See Hill (2002), and Center for Voting and Democracy
(www.fairvote.org) for introductions.
5. My account is adapted from Farrell (2001, 150).
6. Bradley (1995); although there is small evidence of nonmonotonic results in
a recent election, Gallagher (1999).
7. Disputedby Dummett (1997), who suggests, without argument, a conservative
2 percent incidence rather than Allards 0.28 percent.
444
Endnotes 445
4. 1nc annow ccNcnaL iossi ni Li 1v 1nconc:
1. Philosophy too has moved many decades beyond the logical positivism that
originally justied the ordinalist revolution. See Grifn (1986, esp. 106126)
for an example.
5. i s bc:ocnacv :caNi NcLcss: annows coNbi 1i oN or
cNncs1ni cb bo:oi N
1. Michael Munger gave me this story.
2. I am adapting from Collies (1988) useful summary. See Lutz and Williams
(1976) for a decisive survey of empirical evidence against the minimum-
winning coalition, and Hardin (1976) and Grofman (1984) for further ob-
jections.
3. Shepsle and Weingast (1981), Niou and Ordeshook (1985).
4. For legislative discourse see, for example, Fenno (1966) on appropriations,
Wildavsky (1974) on the budget, and Conlan, Wrightson and Beam(1990) on
taxation. Wildavsky (1974, 17) points out that legislative budgeters distinguish
the base, from which annual considerations begin, from the fair share, what the
budgeted item is due.
5. That some give some to an anonymous recipient indicates some altruism; and
contributions triple when the recipient is truthfully identied as a reputable
charity (the Red Cross, see Eckel and Grossman 1996).
6. According to http://www.theindependent.co.zw/news/2002/August/Friday16/
muckr.html, he made the statement to the Sunday Times (presumably of
Africa).
7. If people rank fair distribution second, but use plurality rule, which only counts
rst preferences, then the voting rule would force unfair outcomes from mod-
erately fair voters. See Reilly (2001).
8. Lewin (1991) also nds that the public-choice hypotheses of politicians as
vote-maximizers and bureaucrats as budget-maximizers are unsupported by
the evidence. There is an important difference between exploring such as-
sumptions on an as-if model-building basis, and mistakenly believing them
to be true.
6. i s bc:ocnacv :caNi NcLcss: annows coNbi 1i oN or 1nc
i NbcicNbcNcc or i nncLcvaN1 aL1cnNa1i vcs
1. Keith Dowding, Christian List, and Bruno Verbeek saved me frommany major
and minor errors in an earlier draft of this chapter. They are not to blame for
those that remain.
2. Arrows Condition I is the conjunction of two conditions that can be written
separately: one requiring ordinal measurability and no interpersonal compar-
isons, and another purely the independence of irrelevant alternatives.
3. Arrows book was dated 1951, and he published a 1952 article summarizing
its ndings. The 1951 book was reprinted in 1963, containing an important
addendum which updates and responds to critics.
446 Endnotes
4. Sen (1970), Plott (1971), Hansson (1973), especially Ray (1973), McLean
(1995), but for another view see Bordes and Tideman (1991).
5. My example is adapted from Goodman and Markowitz (1952).
6. Sen (e.g., 1982, 330) blames the exclusion of non-utility information and the
exclusion of any utility information involving interpersonal comparisons for
the impossibility result. Both Saari and Sen nger the exclusion of available
information as the culprit.
7. Christian List (2003) argues that even within the vericationist framework,
the empirical meaninglessness or underdetermination of interpersonal utility
comparisons does not imply the impossibility of such comparisons.
8. A positive afne transformation is f(x) = a + bx, where a and b are real
numbers and b > 0.
9. Social choice rules such as do whatever the Bible, or the leader, says to do
need not weight voters.
10. The ranking computation of the Borda count is a purely formal operation
on ordinal comparisons and should not be interpreted as a cardinal utility
(Kelly 1988, 71).
11. The Condorcet case, however, does not violate either of the independence
conditions.
7. s1na1cci c vo1i Nc aNb accNba coN1noL
1. Stratman 1997 is my main source for this section, although the controversial
interpretations are mine.
2. Stratmann (1997, 330) claims that logrolling implies cycling, but among the
references he cites is Bernholz (1975, 961) who is concerned to correct this
very error: it is easy to show that logrolling does not necessarily imply the
paradox of voting, nor does the paradox of voting imply the existence of
logrolling situations.
3. Strom 1990, 183 attributes to Riker (1980a) the conclusion that legisla-
tors have little incentive to grant agenda control to either individuals or
groups. A case might be made that this is an implication of Riker (1980a),
but this is certainly not Rikers published position. Riker (1980a) was re-
cycled into Chapter 7 of Riker (1982) on agenda control, under examina-
tion here, which claims that agenda control is ubiquitous. Rikers (1993, 1)
introduction to his edited volume on agenda formation, continues the
theme that making agendas seems just about as signicant as actually passing
legislation.
8. :cL1i bi :cNsi oNaL cnaos
1. For example, Ordeshook (1986, 81): Some interpret [McKelveys theorem]
mistakenly to mean that anything can happen. . . . In his original essays,
however, McKelvey is careful to limit the implications of his analysis.
2. This and the next two paragraphs borrow from Strom (1990, 115125).
3. Humes Enquiries was rst published in 1777.
Endnotes 447
9. assc:i Nc i nna1i oNaL ac1ons: 1nc iowcLL a:cNb:cN1
1. Then and now, Alabama schools, contrary to the national pattern of local
revenue support, rely on state and federal funds, more even than neighbor-
ing southern states (Thomas and Stewart 1988, 85; personal communication,
2002, Brad Moody, Auburn University). Thanks to its New Deal Democrats,
Alabama was the most liberal state in the south in the early 1950s; and civil
rights controversies affected state elections later in Alabama than in neighbor-
ing states (Barnard 1974, 4; personal communication, Moody).
2. For the record, those who voted nay on the passage of the School Construc-
tion Bill in 1956 (Congressional Record Roll-Call Vote No. 92, Congressional
Quarterly Vote No.48) who also voted nay in 1957 on the motion to strike
the enacting clause of the School Construction Assistance Act (Congressional
Record Roll-Call Vote No. 154, Congressional Quarterly Vote No.56: Andrews
(D-AL), Elliot (D-AL), Grant (D-AL), Huddleston (D-AL), Jones (D-AL),
Rains (D-AL), Roberts (D-AL), Selden (D-AL), Hays (D-AR), Trimble
(D-AR) Natcher (D-KY), Siler (R-KY) Ford (R-MI), Judd (R-MN), Jones
(D-MO), F.P. Bolton (R-OH), Albert (D-OK), Steed (D-OK), Fenton
(R-PA), Kearns (R-PA), McConnell (R-PA, not recorded vote for 1957),
Reece (R-TN), Byrnes (R-WI), and Laird (R-WI).
3. For the record, those who voted aye on the passage of the School Construction
bill in 1956 (Congressional Record Roll-Call Vote No. 92, Congressional Quarterly
Vote No. 48) who also voted aye in 1957 on the motion to strike the enact-
ing clause of the School Construction Assistance Act (Congressional Record
Roll-Call Vote No. 154, Congressional Quarterly Vote No. 56, Hosmer (R-CA),
Scudder (R-CA), Sadlak (R-CT), LeCompte (R-IA), Fallon (D-MD), ONeill
(D-MA), Meader (R-MI), Cannon (D-MO), Hull (D-MO), Becker (R-NY),
Bosch (R-NY), Derounian (R-NY), Keating (R-NY), Ostertag (R-NY),
Radwan (R-NY), Taylor (R-NY), Feighan (D-OH), Dogue (R-PA).
10. assc:i Nc i nna1i oNaL ac1ons: 1nc bcicw a:cNb:cN1
1. Voteview is an indispensable computer program which contains all US Con-
gressional Roll Call votes, and allows for location of Representatives and
Senators in a two-dimensional space. See Poole, Rosenthal, and Shor (1999).
11. cN:aNi icLa1i Nc 1nc :aNi icLa1i oN: 1nc wi L:o1 inovi so
1. My account relies mostly on Morrison (1967); somewhat on DeVoto (1957)
and Sellers (1966); and on relevant portions of the congressional record cited
in the text.
2. I rely on Morrison (1967, 2137).
12. cN:aNi icLa1i Nc 1nc :aNi icLa1i oN: 1nc cLcc1i oN or Li NcoN
1. Throughout this chapter I rely primarily on McPherson (1993), with much
detail from Potter (1976) and some detail from Nevins (1950). These are
standard texts on the history of the period prior to the Civil War.
448 Endnotes
2. On immigration I rely on Fogel 1992.
3. The platforms are reprinted in Morison (1971, 11231127).
4. Here we turn to Dumonds close study (1931, 9296).
13. aN1cncLLc: ioLi 1i cs coNcLcbcb
1. The subeld of international relations could suggest additional causal hypothe-
ses relating to the outbreak of war in general and civil war in particular.
2. Although I differ in minor details of the analysis, I ampersuaded by Jenkins and
Morris (2002) that the new evidence they develop strongly supports the view
that the Southern Democratic leaders behind Breckenridge wanted a Lincoln
victory in order to better justify secession.
14. :onc or ni kcns cvcLcs bcncNkcb
1. The records of the Federal Convention are found in Farrand (1966). Refer-
ences to the volumes of Farrand (1966) are indicated by the speakers name and
a Roman numeral followed by a page number, for example Madison, II 500.
References to Roll Calls for example, #12 are also to Farrand (1966).
2. At this time there was slavery in both the North and the South, but it was
much more important in the South.
15. o1ncn cvcLcs bcncNkcb
1. In Lagerspetz (1997) there was doubt about the position of the SFP. Since
then, Lagerspetz has done further research, and communicates to me that the
ofcial position of the 25 SFP members was SV > ST > KA > TA.
16. Ncw bi :cNsi oNs
1. See also Cowen (1993), Kavka (1991), compare Gillroy and Wade (1992).
2. I have grown to dislike the word heresthetic, and do not encourage its adoption;
I use it only in order to expound Rikers doctrine.
17. iLcni sci 1ani aNi s: acai Ns1 bc:ocnacv
1. After writing these passages I discovered that the younger Riker (1953, 158
160) argued forcefully and at length that the Civil War was due to the antima-
joritarian features of the constitution.
2. A thinkers political allegiances are irrelevant to judging the quality of his or
her argument. However, if a thinker makes normative political recommenda-
tions, and those recommendations fail to work as intended, then the fact of
failure does reect on the content of the argument. A thinker might make a
brilliant argument that, if party members were well intentioned, a one-party
state would have benecial consequences; but the consistent failure of such
schemes relevantly undermines the claim.
3. Paretos Mind and Society (1963) was originally published in 1935.
Endnotes 449
4. A sympathetic commentator notes Paretos private prejudice against parlia-
mentary democracy. He was always reiterating that he held no such prejudice,
that his work was scientic, not subjective. This is absurdly false. In some cases
he let his prejudices obtrude by slipping in implicit value-judgments, in oth-
ers by using loaded terms, by sarcasms, abuse and imputations of baseness
(Finer 1966, 65). Compare Rikers vituperation of ideologues, quoted above.
5. In the McCarthy years, the Leninist but anti-Stalinist Independent Socialist
League (ISL) sought removal from the Attorney Generals list of subversive
organizations. Anti-Communist Burnhamwas subpoenaed to be the main, but
reluctant, witness against his former comrades in the ISL. The governments
case collapsed when Burnham testied under oath that he would lie under
oath if he thought it his patriotic duty (Wald 1987, 277). This is both a double
deception and a triple deception.
18. bc:ocnacv ncsiLcNbcN1
1. See Bowles and Gintis (2000) for another sophisticated retrospective on the
competitive-equilibrium model.
2. Or, Thats a mistake.
3. Given recent events, I must note that democracy is a matter for the people
involved to develop, not a matter for outside powers to impose by deception
of force.
References
Albertoni, Ettore A. 1987. Mosca and the Theory of Elitism. Oxford: Basil
Blackwell.
Allard, Crispin. 1995. Lack of Monotonicity Revisited. Representation 33:
4850.
Alt, James. 1983. The Evolution of Tax Structures. Public Choice 41: 181223.
Amadae, S.M., and Bruce Bueno de Mesquita. 1999. The Rochester School:
The Origins of Positive Political Theory. Annual Review of Political Science
2: 269295.
Anderson, Thornton. 1993. Creating the Constitution: The Convention of 1787 and
the First Congress. University Park: Pennsylvania State University Press.
Arrow, Kenneth J. 1952. The Principle of Rationality in Collective Decisions.
In Arrow (1984, 4558).
Arrow, Kenneth J. 1960. Introductory Notes. Portfolio and Art News Annual 4:
7678.
Arrow, Kenneth J. 1963/1951. Social Choice and Individual Values, second edn.
New Haven, CT: Yale University Press.
Arrow, Kenneth J. 1967. Values and Collective Decision-Making. In Peter
Laslett and W.G. Runciman, eds., Philosophy, Politics and Society, third series.
Oxford: Basil Blackwell. Reprinted in Arrow (1984).
Arrow, Kenneth J. 1973. Formal Theories of Social Welfare. In Arrow (1984,
115132).
Arrow, Kenneth J. 1984. Social Choice and Justice: Collected Papers of Kenneth J.
Arrow. Oxford: Basil Blackwell.
Arrow, Kenneth J. 1987. Oral History I: An Interview. In Arrow and the Ascent
of Modern Economic Theory, ed. George R. Feiwel. New York: New York
University Press.
Arrow, Kenneth J. 1997. The Functions of Social Choice Theory. In Kenneth J.
Arrow, Amartya Sen, and Kotaro Suzumura, eds., Social Choice Reexamined,
vol. I. New York: St. Martins Press.
Ashworth, John. 1995. Slavery, Capitalism, and Politics in the Antebellum Republic,
vol. I, Commerce and Compromise: 18201850. Cambridge University Press.
Austen-Smith, David. 1987. Sophisticated Sincerity: Voting over Endogenous
Agendas. American Political Science Review 81: 13231329.
Austen-Smith, David, and Jeffrey S. Banks. 1998. Social Choice Theory, Game
Theory, and Positive Political Theory. Annual Review of Political Science 1:
259287.
450
References 451
Baddeley, Alan. 1994. The Magical Number Seven: Still Magic After All These
Years? Psychological Review 101: 353356.
Barber, William J. 1995. Chile con Chicago: A Review Essay. Journal of Eco-
nomic Literature 33: 19411949.
Barker, Ernest. 1951. Principles of Social and Political Theory. London: Oxford
University Press.
Barnard, WilliamD. 1974. Dixiecrats and Democrats: Alabama Politics, 19421950.
Tuscaloosa: University of Alabama Press.
Baron, David P., and John A. Ferejohn. 1989. Bargaining in Legislatures.
American Political Science Review 83: 11811206.
Barry, Brian, and Russell Hardin. 1982. Rational Man and Irrational Society.
Beverley Hills: Sage Publications.
Bartholdi, John J., III, and James B. Orlin. 1991. Single Transferable Vote Resists
Strategic Voting. Social Choice and Welfare 8: 341354.
Bartholdi, John J., III, Craig A. Tovey, and Michael A. Trick. 1989. Voting
Schemes for which It Can Be Difcult to Tell Who Won the Election.
Social Choice and Welfare 6: 157165.
Bartholdi, John J., III, Craig A. Tovey, and Michael A. Trick. 1992a. The Com-
putational Difculty of Manipulating an Election. Social Choice and Welfare
6: 227241.
Bartholdi, John J., III, Craig A. Tovey, and Michael A. Trick. 1992b. How
Hard Is It to Control an Election? Mathematical and Computer Modelling 16:
2740.
Beer, Samuel H. 1960. NewStructures of Democracy: Britain and America. In
Samuel H. Beer, Leon D. Epstein, Louis Hartz, Charles E. Lindblom, and
J. Roland Pennock, eds., Democracy in the Mid-Twentieth Century. St. Louis:
Washington University Press.
Bell, Daniel. 1974. The Coming of Post-Industrial Society. London: Heinemann.
Berg, Sven. 1985. Paradox of Voting under an Urn Model: The Effect of
Homogeneity. Public Choice 47: 377387.
Bernholz, Peter. 1975. Logrolling and the Paradox of Voting: Are They Logically
Equivalent? American Political Science Review 69: 961962.
Bjurulf, Bo, and Richard Niemi. 1978. Strategic Voting in Scandinavian Parlia-
ments. Scandinavian Political Studies 1(n.s.): 522.
Black, Duncan. 1958. The Theory of Committees and Elections. Cambridge Uni-
versity Press.
Block, Cheryl D. 1998. Truth and Probability Ironies in the Evolution
of Social Choice Theory. Washington University Law Quarterly 76: 975
1036.
Blydenburgh, John C. 1971. The Closed Rule and the Paradox of Voting.
Journal of Politics 33: 5771.
Bohman, James. 1996. Public Deliberation. Cambridge, MA: MIT Press.
Bordes, Georges, and Nicolaus Tideman. 1991. Independence of Irrelevant
Alternatives in the Theory of Voting. Theory and Decision 30: 163186.
Bowles, Samuel, and Herbert Gintis. 2000. Walrasian Economics in Retro-
spect. Quarterly Journal of Economics 115: 14111439.
Bradley, Patrick. 1995, STV and Monotonicity: A Hands-On Assessment.
Representation 33: 4647.
452 References
Brennan, Geoffrey, and Alan Hamlin. 2000. Democratic Devices and Desires.
Cambridge University Press.
Brennan, Geoffrey, and Loren Lomasky. 1993. Democracy and Decision: The Pure
Theory of Electoral Preference. Cambridge University Press.
Bucolo, Placido, ed. 1980. The Other Pareto. London: Scolar Press.
Budge, Ian. 1993. Issues, Dimensions, and Agenda Change in Postwar Democ-
racies: LongtermTrends in Party Election Programs and Newspaper Reports
in Twenty-Three Democracies. In WilliamH. Riker, ed., Agenda Formation.
Ann Arbor: University of Michigan Press.
Burnham, James. 1943. The Machiavellians: Defenders of Freedom. London:
Putnam.
Burnham, W. Dean. 1955. Presidential Ballots: 18361892. Baltimore: Johns
Hopkins University Press.
Burnham, W. Dean. 1999. Constitutional Moments and Punctuated Equilibria:
A Political Scientist Confronts Bruce Ackermans We the People. Yale Law
Journal 108: 22372277.
Byrd, Robert C. 1988. The Senate, vol. I. Washington, D.C.: U.S. Government
Printing Ofce.
Cain, Michael J.G. 2001. In William F. Shughart III and Laura Razzolini,
eds., The Elgar Companion to Public Choice. Cheltenham, UK: Edward
Elgar.
Calvert, Randall L., and Richard F. Fenno, Jr. 1994. Strategy and Sophisticated
Voting in the Senate. The Journal of Politics 56: 349376.
Caplan, Russell L. 1988. Constitutional Brinksmanship: Amending the Constitution
by National Convention. New York: Oxford University Press.
Caplin, Andrew, and Barry Nalebuff. 1988. On 64% Majority Rule. Economet-
rica 56 (July): 787814.
Caplin, Andrew, and Barry Nalebuff. 1991. Aggregation and Social Choice: A
Mean Voter Theorem. Econometrica 59: 123.
Chamberlin, John R. 1985. Investigation into the Relative Manipulability of
Four Voting Systems. Behavioral Science 30: 195203.
Chamberlin, John R., Jerry L. Cohen, and Clyde H. Coombs. 1984. Social
Choice Observed: Five Presidential Elections of the American Psychological
Association. Journal of Politics 46: 479502.
Citrin, Jack and Donald Philip Green. 1990. The Self-Interest Motive in
American Public Opinion. In Samuel Long, ed., Research in Micropolitics,
vol. III. Greenwich: JAI Press.
Coady, C.A.J. 1992. Testimony: A Philosophical Study. Oxford: Clarendon Press.
Cohen, Joshua. 1986. An Epistemic Conception of Democracy. Ethics 97 (Oc-
tober): 2638.
Cohn, Jonathan. 1999. Irrational Exuberance: What DidPolitical Science Forget
about Politics? New Republic, October 25.
Coleman, James. 1990. Foundations of Social Theory. Cambridge, MA: Harvard
University Press.
Collie, Mellisa P. 1988. The Legislature and Distributive Policy Making in
Formal Perspective. Legislative Studies Quarterly 13: 427458.
Colman, Andrew M., and Ian Pountney. 1978. Bordas Paradox: Theoretical
Likelihood and Electoral Occurrences. Behavioral Science 23: 1521.
References 453
Congressional Quarterly Almanac, vol. XIII. 1957. Washington, DC: Congres-
sional Quarterly News Features.
Congressional Quarterly Almanac, vol. XXVI. 1970. Washington, DC: Congres-
sional Quarterly News Features.
Conlan, Timothy J., Margaret T. Wrightson, and David R. Beam. 1990. Taxing
Choices: The Politics of Tax Reform. Washington: Congressional Quarterly.
Cooter, Robert and Peter Rappoport. 1984. Were the Ordinalists Wrong about
Welfare Economics? Journal of Economic Literature 22: 507530.
Cowen, Tyler. 1993. The Scope and Limits of Preference Sovereignty. Eco-
nomics and Philosophy 9: 253269.
Cox, Gary W. 1987. Electoral Equilibria under Alternative Voting Institutions.
American Journal of Political Science. 31: 82108.
Cox, Gary W. 1990. Centripetal and Centrifugal Incentives in Electoral Sys-
tems. American Journal of Political Science. 34: 903935.
Cox, Gary W. 1997. Making Votes Count. Cambridge University Press.
Cox, Gary W. 1999. The Empirical Content of Rational Choice Theory: AReply
to Green and Shapiro. Journal of Theoretical Politics 11: 147169.
Cox, Gary, and Timothy N. Tutt. 1984. Universalism and Allocative Decision
Making in the Los Angeles County Board of Supervisors. Journal of Politics
46: 546555.
Dahl, Robert A. 1956. A Preface to Democratic Theory. University of Chicago
Press.
Dahl, Robert A. 1989. Democracy and its Critics. NewHaven, CT: Yale University
Press.
Davidson, Donald. 1980. Essays on Actions and Events. Oxford: Clarendon Press.
Davidson, Donald. 1986. Judging Interpersonal Interests. In Jon Elster and
Aanund Hylland, eds., Foundations of Social Choice Theory. Cambridge Uni-
versity Press.
Davidson, Donald. 1991a. Epistemology Externalized. Dialectica 45: 191202.
Davidson, Donald. 1991b. Three Varieties of Knowledge. In A.J. Ayer Memo-
rial Essays, ed. A. Phillips Grifths. Cambridge University Press.
Davies, Todd, and Raja Shah. 2003. Intuitive Preference Aggregation: Tests of
Independence and Consistency. Presented at Annual Meeting of the Public
Choice Society, March 21, Nashville.
Denzau, Arthur,William Riker, and Kenneth Shepsle. 1985. Farquharson and
Fenno: Sophisticated Voting and Home Style. American Political Science
Review 79: 11171134.
DeVoto, Bernard. 1957. The Year of Decision, 1846. London: Eyre and
Spottiswoode.
Divine, Robert A., et al. 1999. America: Past and Present. New York: Longman.
Dobra, John L. 1983. An Approach to Empirical Studies of Voting Paradoxes:
An Update and Extension. Public Choice 41: 241250.
Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper
and Row.
Drew, Elizabeth. 2000. The Corruption of American Politics: What Went Wrong and
Why. New York: Overlook Press.
Dryzek, John. 2000. Deliberative Democracy and Beyond: Liberals, Critics, Contes-
tations. Oxford University Press.
454 References
Dryzek, John, and Christian List. 2003. Social Choice Theory and Deliber-
ative Democracy: A Reconciliation. British Journal of Political Science 33:
128.
Dummett, Michael. 1997. Principles of Electoral Reform. Oxford University Press.
Dummett, Michael. 1998. The Borda Count and Agenda Manipulation. Social
Choice and Welfare 15: 287296.
Dumond, Dwight Lowell. 1931. The Secession Movement: 18601861. New York:
Macmillan Company.
Eakin, Emily. 2000. Political Scientists Are in a Revolution Instead of Watch-
ing. The New York Times, November 4.
Eckel, Catherine C., and Philip J. Grossman. 1996. Altruism in Anonymous
Dictator Games. Games and Economic Behavior 16: 181189.
Elster, Jon. 1979. Ulysses and the Sirens. Cambridge University Press.
Elster, Jon. 1986a. Introduction. In Jon Elster, ed., Rational Choice. New York
University Press.
Elster, Jon. 1986b. The Market and the Forum: Three Varieties of Political
Theory. In Elster and Aanund Hylland, eds., Foundations of Social Choice
Theory. Cambridge University Press.
Elster, Jon, ed. 1998. Deliberative Democracy. Cambridge University Press.
Enelow, James M. 1997. Cycling and Majority Rule. In Dennis C. Mueller,
ed., Perspectives on Public Choice: A Handbook. Cambridge University
Press.
Farquharson, Robin. 1969. Theory of Voting. Oxford: Basil Blackwell.
Farrand, Max, ed. 1966. The Records of the Federal Convention of 1787, 4 vols.
New Haven: Yale University Press.
Farrand, Max. 1913. The Framing of the Constitution of the United States. New
Haven: Yale University Press.
Farrell, David M. 2001. Electoral Systems: A Comparative Introduction. Bas-
ingstoke: Palgrave.
Fay, Brian. 1996. Contemporary Philosophy of Social Science: A Multicultural
Approach. Oxford: Blackwell.
Fehr, Ernst, and Urs Fischbacher. 2002. Why Social Preferences Matter. The
Impact of Non-Selsh Motives on Competition, Cooperation and Incen-
tives. Economic Journal 112: C1C33.
Fehrenbacher, Don E. 1995. Sectional Crisis and Southern Constitutionalism. Baton
Rouge: Lousiana State University Press.
Feld, Scott L., and Bernard Grofman. 1986. Partial Single-Peakedness: An
Extension and Clarication. Public Choice 51: 7180.
Feld, Scott L., and Bernard Grofman. 1992. Whos Afraid of the Big Bad Cycle?
Evidence from 36 Elections. Journal of Theoretical Politics 4: 231237.
Feldman, Allen M. 1980. Welfare Economics and Social Choice Theory. Boston:
Martinus Nijhoff.
Feldman, AllenM. 1991. Welfare Economics. InJohnEatwell, Murray Milgate,
and Peter Newman, eds., The New Palgrave: The World of Economics. New
York: W.W. Norton.
Felsenthal, Dan S., and Moshe Machover. 1995. Who Ought to Be Elected and
Who Is Elected? An Empirical Investigation of 92 Elections under Three
Procedures. Electoral Studies. 14: 143169.
References 455
Felsenthal, Dan S., Zeev Maoz, and Amnon Rapoport. 1993. An Empirical
Evaluation of Six Voting Procedures: Do They Really Make Any Difference?
British Journal of Political Science 23: 127.
Fenno, Richard F., Jr. 1966. The Power of the Purse: Appropriations Politics in
Congress. Boston: Little Brown.
Ferejohn, John, Morris P. Fiorina, and Richard D. McKelvey. 1987. Sophisti-
cated Voting and Agenda Independence in the Distributive Setting. Amer-
ican Journal of Political Science 31: 169193.
Fine, Benjamin. 1957. President Pushes School Help Bill. The New York Times,
July 4, sec. 1, p. 15.
Finer, Samuel E. 1966. Vilfredo Pareto: Sociological Writings. London: Pall Mall
Press.
Fiorina, Morris P., and Charles Plott. 1978. Committee Decisions under Ma-
jority Rule: An Experimental Study. American Political Science Review 72:
575598.
Fiorina, Morris P., and Kenneth A. Shepsle. 1982. Equilibrium, Disequilibrium,
and the General Possibility of a Science of Politics. In Peter C. Ordeshook
and Kenneth A. Shepsle, eds., Political Equilibrium. Boston: Kluwer-Nijhoff.
Fischer, David Hackett. 1989. Albions Seed: Four British Folkways in America.
Oxford University Press.
Fogel, Robert William. 1989. Without Consent or Contract. New York: W.W.
Norton.
Fogel, Robert William. 1992. Problems in Modeling Complex Dynamic Inter-
actions: The Political Realignment of the 1850s. Economics and Politics 4:
215254.
Freehling, William W. 1990. The Road to Disunion, vol. I, Secessionists at Bay:
17761854. New York: Oxford University Press.
Gallagher, Michael. 1999. The Results Analysed. In Michael Gallagher and
Michael Laver, eds., How Ireland Voted, 1997. Boulder: Westview Press.
Gaubatz, Kurt Taylor. 1995. Intervention and Intransitivity: Public Opinion,
Social Choice, and the Use of Military Force Abroad. World Politics 47:
534554.
Gehrlein, WilliamV. 1983. Condorcets Paradox. Theory and Decision. 15: 161
197.
Gehrlein, William V. 1995. Condorcet Efciency and Social Homogeneity. In
WilliamA. Barnett, Melvin J. Hinich, and Norman J. Schoeld, eds., Political
Economy: Institutions, Competition, and Representation. Cambridge University
Press.
Gehrlein, William V. 1997. Condorcets Paradox and the Condorcet Efciency
of Voting Rules. Mathematica Japonica 45: 173199.
Gibbard, Allan. 1973. Manipulation of Voting Schemes: A General Result.
Econometrica 41: 587601.
Gillroy, John Martin, and Maurice Wade, eds. 1992. The Moral Dimensions of
Public Policy Choice. University of Pittsburgh Press.
Ginsberg, Allen. 1956. Howl and Other Poems. San Francisco: City Lights Books.
Glazer, Amihai, and Henry McMillan. 1982. Amend the Old or Address the
New: Broad-Based Legislation When Proposing Policies is Costly. Public
Choice 74: 4358.
456 References
Goodin, Robert. 1992. Motivating Political Morality. Oxford: Basil Blackwell.
Goodin, Robert. 1995/1986. Laundering Preferences. In Utilitarianism as a
Public Philosophy. Cambridge University Press.
Goodin, Robert, and Hans-Dieter Klingemann. 1996. A New Handbook of Polit-
ical Science. Oxford University Press.
Goodin, Robert, and K.W.S. Roberts. 1975. The Ethical Voter. American
Political Science Review 69: 926928.
Goodman, Leo, and Harry Markowitz. 1952. Social Welfare Functions Based
on Individual Rankings. American Journal of Sociology 58: 257262.
Green, Donald P., and Ian Shapiro. 1994. Pathologies of Rational Choice Theory.
New Haven, CT: Yale University Press.
Grifn, James. 1986. Well-Being: Its Meaning, Measurement, and Moral Importance.
Oxford: Clarendon Press.
Grofman, Bernard. 1984. The General Irrelevance of the Zero SumAssumption
in the Legislative Context. In M. J. Holler, ed., Coalitions and Collective
Action. W urzburg: Physica-Verlag.
Grofman, Bernard, and Scott L. Feld. 1988. Rousseaus General Will: A Con-
dorcetian Perspective. American Political Science Review 82: 567576.
Grofman, Bernard, and Guillermo Owen. 1986. Information Pooling and Group
Decision-Making. London: JAI Press.
Grofman, Bernard, and Andrew Reynolds. 2001. Electoral Systems and the Art
of Constitutional Engineering: An Inventory of the Main Findings. In Ram
Mudambi, Pietro Navarra, and Giuseppe Sobbrio, eds., Rules and Reason.
Groseclose, Tim, and James M. Snyder, Jr. 1996. Buying Supermajorities.
Gross, Donald. 1979. Conference Committees, Sophisticated Voting, and
Cyclical Majorities. Legislative Studies Quarterly 4(1): 7994.
Habermas, J urgen. 1984, 1987. Theory of Communicative Action, vols. I and II.
Boston: Beacon Press.
Hamilton, Alexander, John Jay, and James Madison. n.d. (1937)/1787. The
Federalist: A Comment on the Constitution of the United States. New York: The
Modern Library.
Hamilton, Charles V. 1991. Adam Clayton Powell, Jr. New York: Atheneum.
Hansson, Bengt. 1973. The Independence Condition in the Theory of Social
Choice. Theory and Decision 4: 2549.
Hardin, Russell. 1976. Hollow Victory: The Minimum Winning Coalition.
Hardin, Russell. 1993. Public Choice Versus Democracy. In David Copp, Jean
Hampton, and John E. Roemer, eds., The Idea of Democracy. Cambridge
University Press.
Hauptmann, Emily. 1996. Putting Choice Before Democracy: A Critique of Rational
Choice Theory. Albany: SUNY Press.
Haynes, George H. 1938. The Senate of the United States: Its History and Practice.
Boston, MA: Houghton Mifin.
Hennessy, Alistair. 1969. Latin America. In Ghita Ionescu and Ernest Gellner,
eds., Populism: Its Meanings and Characteristics. London: Weidenfeld and
Nicolson.
References 457
Henrich, Joseph, Robert Boyd, Samuel Bowles, Colin Camerer, Ernst Fehr,
Herbert Ginitis, and Richard McElreath. 2001. In Search of Homo
Economicus: Behavioral Experiments in 15 Small-Scale Societies. Amer-
ican Economic Review 91: 7378.
Hesseltine, WilliamB. 1962. The Tragic Conict: The Civil War and Reconstruction.
New York: George Braziller.
Hill, Steven. 2002. Fixing Elections: The Failure of Americas Winner Take All Politics.
New York: Routledge.
Hinich, Melvin J., and Michael C. Munger. 1994. Ideology and the Theory of
Political Choice. Ann Arbor: University of Michigan Press.
Hinich, Melvin J., and Michael C. Munger. 1997. Analytical Politics. Cambridge
University Press.
Hoebeke, Christopher Hyde. 1995. The Road to Mass Democracy: Original Intent
and the Seventeenth Amendment. NewBrunswick, NJ: Transaction Publishers.
Holmes, Stephen. 1990. The Secret History of Self-Interest. In Jane
Mansbridge, ed., Beyond Self-Interest. University of Chicago Press.
Holt, James. 1967. Congressional Insurgents and the Party System: 19091916.
Cambridge, MA: Harvard University Press.
Hovenkamp, Herbert. 1990a. Arrows Theorem: Ordinalism and Republican
Government. Iowa Law Review 75: 949973.
Hovenkamp, Herbert. 1990b. Legislation, Well-Being and Public Choice. Uni-
versity of Chicago Law Review 57: 63116.
Hume, David. 1975. Enquiries Concerning Human Understanding and Concern-
ing the Principles of Morals, 3rd edn., ed. P.H. Nidditch. Oxford: Clarendon
Press.
Jacobsen, Kurt. 2001. Unreal, Man. The Guardian (UK), April 3.
Jay, William. 1849. A Review of the Causes and Consequences of the Mexican War.
Boston, MA: Benjamin D. Mussey.
Jenkins, Jeffery A., and Irwin Morris. 2002. Running to Lose? John C.
Breckinridge and the Presidential Election of 1860. Manuscript.
Jevons, William Stanley. 1871. Theory of Political Economy. London: Macmillan.
Jillson, Calvin C. 1979. The Executive in Republican Government: The Case
of the American Founding. Presidential Studies Quarterly 9: 386402.
Johannsen, Robert W. 1973. Stephen A. Douglas. Oxford University Press.
Jones, Bradford, Benjamin Radcliff, Charles Taber, and Richard Tampone.
1995. Condorcet Winners and the Paradox of Voting: Probability Calcu-
lations for Weak Preference Orders. American Political Science Review 89:
137146.
Joslyn, R.A. 1976. The Impact of DecisionRules inMulti-candidate Campaigns:
The Case of the 1972 Democratic Presidential Nomination. Public Choice
25: 117.
Kasza, Gregory. 2001. Perestroika: For an Ecumenical Science of Politics.
PSOnline, September, www.apsanet.org.
Katznelson, Ira and Helen Milner, eds., 2002. Political Science: The State of the
Discipline. New York: W.W. Norton.
Kavka, Gregory S. 1991. Is Individual Choice Less Problematic than Collective
Choice? Economics and Philosophy 7: 143165.
Kelly, Jerry S. 1988. Social Choice Theory: An Introduction. Berlin: Springer-Verlag.
458 References
Kinder, Donald R. and D. Roderick Kiewiet. 1979. Economic Discontent and
Political Behavior: The Role of Personal Grievances and Collective Eco-
nomic Judgments in Congressional Voting. American Journal of Political
Science 23: 495527.
Kinder, Donald R. and D. Roderick Kiewiet. 1981. Sociotropic Politics: The
American Case. British Journal of Political Science 11: 129161.
King, Ronald F., and Susan Ellis. 1996. Partisan Advantage and Constitutional
Change: The Case of the Seventeenth Amendment. Studies in American
Political Development 10: 69102.
Koford, Kenneth J. 1982. Centralized Vote Trading. Public Choice 39: 245268.
Kramer, Gerald H. 1973. On a Class of Equilibrium Conditions for Majority
Rule. Econometrica 41: 285297.
Kramer, Gerald H. 1977. A Dynamic Model of Political Equilibrium. Journal
of Economic Theory 16: 310334.
Krehbiel, Keith, and Douglas Rivers. 1990. Sophisticated Voting in Congress:
A Reconsideration. Journal of Politics 52: 548578.
Kuga, Kiyoshi, and Hiroaki Nagatani. 1974. Voter Antagonismand the Paradox
of Voting. Econometrica 42 (November): 10451067.
Kurrild-Klitgaard, Peter. 2001a. An Empirical Example of the Condorcet Para-
dox of Voting in a Large Electorate. Public Choice 107: 135145.
Kurrild-Klitgaard, Peter. 2001b. Intransitivity and Robustness in Social Choice:
Evidence from Danish Election Surveys. Public Choice Society, Confer-
ence, San Antonio Texas, 911 March.
Kuttner, Robert. 1996. Everything for Sale: The Virtues and Limits of Markets.
University of Chicago Press.
Kyvig, David E. 1996. Explicit and Authentic Acts: Amending the U.S. Constitution,
17761995. Lawrence: University Press of Kansas.
Lagerspetz, Eerik. 1993. Social Choice in the Real World. Scandinavian Political
Studies 16(1): 123.
Lagerspetz, Eerik. 1997. Social Choice in the Real World II: Cyclical Preferences
and Strategic Voting in the Finnish Presidential Elections. Scandinavian
Political Studies 20(1): 5367.
Laing, J.D, and S. Olmstead. 1978. An Experimental and Game-Theoretic
Study of Committees. In Peter C. Ordeshook, ed., Game Theory and Po-
litical Science. New York University Press.
Laing, J.D., and Benjamin Slotznick. 1987. Viable Alternatives to the Status
Quo. Journal of Conict Resolution 31: 6385.
Laing, J.D., and Benjamin Slotznick. 1991. When Anyone Can Veto: A Lab-
oratory Study of Committees Governed by Unanimous Rule. Behavioral
Science 36: 179195.
Le Breton, Michel, and Michel Truchon. 1997. A Borda Measure for Social
Choice Functions. Mathematical Social Sciences 34: 249272.
Levin, Jonathan, and Barry Nalebuff. 1995. An Introduction to Vote-Counting
Schemes. Journal of Economic Perspectives 9: 326.
Lewin, Leif. 1991. Self-Interest and Public Interest in Western Politics. Oxford
University Press.
Lewis, Anthony. 1956. School Aid and Tangled Politics. The New York Times,
July 8, sec. IV, p. 1.
References 459
Lichbach, Mark I. 1996. The Cooperators Dilemma. Ann Arbor: University of
Michigan Press.
Lindsay, Alexander Dunlop. 1935. Essentials of Democracy. London: Oxford
University Press.
Linz, Juan. 1994. Presidential or Parliamentary Democracy: Does It Make a
Difference? In Juan Linz, and Arturo Valenzuela, eds., The Failure of Presi-
dential Democracy, vol. I, Comparative Perspectives. Baltimore: Johns Hopkins
University Press.
Lippincott, Benjamin Evans. 1938. Victorian Critics of Democracy. Minneapolis:
University of Minnesota Press.
List, Christian. 2001. Mission Impossible? The Problem of Democratic Aggregation
in the Face of Arrows Theorem. D.Phil. thesis in politics, Nufeld College,
Oxford. http://www.nuff.ox.ac.uk/users/list/mission impossible.htm
List, Christian. 2002. Intradimensional Single-Peakedness and the Multidimen-
sional Arrow Problem. Theory and Practice 52: 287301.
List, Christian. 2003. Are Interpersonal Comparisons of Utility Indeterminate?
Erkentniss 58: 229260.
List, Christian and Robert E. Goodin. 2001. Epistemic Democracy: Generaliz-
ing the Condorcet Jury Theorem. Journal of Political Philosophy 9: 277306.
Little, I. M. D. 1952. Social Choice and Individual Values. Journal of Political
Economy 60: 422432.
Lutz, Donald S., and James R. Williams. 1976. Minimum Winning Coalitions in
Legislatures: A Review of the Evidence. Beverley Hills: Sage Publications.
Ma, Barry K., and Jeffrey H. Weiss. 1993. On Some Generalizations of a Mean
Voter Theorem. Economics Letters 42: 185188.
Mackay, Alfred F. 1980. Arrows Theorem: The Paradox of Social Choice. New
Haven, CT: Yale University Press.
Mackie, Gerry. 1996. Ending Footbinding and Inbulation: A Convention
Account. American Sociological Review 61: 991017.
Mackie, Gerry. 1998. All Men are Liars. Is Democracy Meaningless? In Jon
Elster, ed., Deliberative Democracy Cambridge University Press.
McDonald, Forrest. 1985. Novus Ordo Seclorum: The Intellectual Origins of the
U.S. Constitution. Lawrence: University Press of Kansas.
McKelvey, Richard D. 1976. Intransitivities in Multi-dimensional Voting
Models and Some Implications for Agenda Control. Journal of Economic
Theory 12: 472482.
McKelvey, Richard D. 1979. General Conditions for Global Intransivities in
Formal Voting Models. Econometrica 47: 10841111.
McKelvey, Richard D. 1986. Covering, Dominance, and Institution-Free Prop-
erties of Social Choice. American Journal of Political Science 24: 6896.
McKelvey, Richard D., and Peter C. Ordeshook. 1990. ADecade of Experimen-
tal Research in Spatial Models of Elections and Committees. In James M.
Enelow, and Melvin J. Hinich, eds., Advances in the Spatial Theory of Voting.
McKelvey, Richard D., and Norman J. Schoeld. 1986. Structural Instability
of the Core. Journal of Mathematical Economics 15: 179198.
McLean, Ian. 1995. Independence of Irrelevant Alternatives before Arrow.
Mathematical Social Sciences 30: 107126.
460 References
McLean, Ian, and Fiona Hewitt. 1994. Condorcet: Foundations of Political Theory
and Social Choice. Aldershot: Edward Elgar.
McLean, Ian, and Arnold B. Urken, eds. 1995. Classics of Social Choice. Ann
Arbor: University of Michigan Press.
McLean, Ian, Christian List, James Fishkin, and Robert Luskin. 2000. Can
Deliberation Induce Greater Preference Structuration? Evidence from De-
liberative Opinion Polls. Proceedings of the American Political Science
Association.
McPherson, James M. 1993. Ordeal by Fire, vol. I, The Coming of War. New York:
McGraw-Hill.
Mansbridge, Jane. 1990. Beyond Self-Interest. University of Chicago Press.
Mansbridge, Jane. 1994. Public Spirit in Political Systems. In Henry J. Aaron,
Thomas E. Mann, and Timothy Taylor, eds., Values and Public Policy. Wash-
ington: The Brookings Institution.
Mashaw, Jerry L. 1989. The Economics of Politics and the Understanding of
Public Law. Chicago-Kent Law Review 65: 123160.
May, Kenneth O. 1952. A Set of Independent, Necessary, and Sufcient Con-
ditions for Simple Majority Decision. Econometrica 20 (October): 680684.
Merrill, Samuel, III. 1988. Making Multicandidate Elections More Democratic.
Princeton University Press.
Michaelsen, William B. 1987. Creating the American Presidency: 17751789.
Lanham, MD: University Press of America.
Miller, Arthur H., Charles Tien, and Andrew A. Peebler. 1996. The American
Political Science Review Hall of Fame: Assessments and Implications for an
Evolving Discipline. PS: Political Science and Politics 29 (1): 78.
Miller, D.W. 2001. Storming the Palace of Political Science. The Chronicle of
Higher Education, September 21.
Miller, Dale T., and Rebecca K. Ratner. 1996. The Power of the Myth of Self-
Interest. In Leo Montada and Melvin J. Lerner, eds., Current Societal Con-
cerns About Justice. New York: Plenum Press.
Miller, Dale T., and Rebecca K. Ratner. 1998. The Disparity between the Actual
and the Assumed Power of Self-Interest. Journal of Personality and Social
Psychology 74: 5362.
Miller, Gary J. 1997. The Impact of Economics on Contemporary Political
Science. Journal of Economic Literature 35: 11731204.
Miller, Gary J., and Joe A. Oppenheimer. 1982. Universalism in Experimental
Committees. American Political Science Review 76: 561574.
Miller, George A. 1994/1956. The Magical Number Seven, Plus or Minus Two:
Some Limits on Our Capacity for Processing Information. Psychological
Review 101: 343352.
Miller, Nicholas, Bernard Grofman, and Scott L. Feld. 1989. The Geometry of
Majority Rule. Journal of Theoretical Politics 4: 379406.
Miller, William Lee. 1997. Arguing about Slavery. New York: Alfred A. Knopf.
Moore, Glover. 1967. The Missouri Controversy: 18191821. Gloucester, MA:
Peter Smith.
Morison, Elting. 1971. Election of 1860. In Arthur M. Schlesinger, Jr., Fred L.
Israel, and William P. Hansen, eds., History of American Presidential Elections:
17891968, vol. II. New York: Chelsea House.
References 461
Morris, John D. 1957a. School Aid Bill Killed in House by 208203 Vote. The
New York Times, July 26, sec. I, p. 1.
Morris, John D. 1957b. Democrats Blame President for School Aid Bill Defeat.
The New York Times, July 27, sec. I, p. 1.
Morrison, Chaplain W. 1967. Democratic Politics and Sectionalism: The Wilmot
Proviso Controversy. Chapel Hill: University of North Carolina Press.
Moulin, Herv e. 1988. Axioms of Cooperative Decision Making. Cambridge
University Press.
Mueller, Dennis C. 1989. Public Choice II. Cambridge University Press.
Mueller, Dennis C. 1996. Constitutional Democracy. Oxford University Press.
Murdoch, Iris. 1992. Metaphysics as a Guide to Morals. London: Penguin.
Mutz, Diane, and Jeffery J. Mondak. 1997. Dimensions of Sociotropic Behavior:
Group-Based Judgments of Fairness and Well-Being. American Journal of
Political Science 41: 284308.
Myerson, Roger B. 1995. Analysis of Democratic Institutions: Structure, Con-
duct and Performance. Journal of Economic Perspectives 9: 7789.
Myerson, Roger B., and Robert J. Weber. 1993. A Theory of Voting Equilibria.
Neufeld, John L., William J. Hausman, and Ronald B. Rapoport. 1994. A Para-
dox of Voting: Cyclical Majorities and the Case of Muscle Shoals. Political
Research Quarterly 47: 423438.
Nevins, Allan. 1950. The Emergence of Lincoln, vol. II, Prologue to Civil War: 1859
1861. New York: Charles Scribners Sons.
Niemi, Richard G. 1969. Majority Decision-Making with Partial Unidimen-
sionality. American Political Science Review 63 (June): 488497.
Niemi, Richard G. 1983. Why So Much Stability?: Another Opinion. Public
Choice 41: 261270.
Niemi, Richard G., and J.R. Wright. 1987. Voting Cycles and the Structure of
Individual Preferences. Social Choice and Welfare 4: 173183.
Niou, Emerson, and Peter C. Ordeshook. 1985. Universalism in Congress.
American Journal of Political Science 29: 246258.
Nitzan, Shmuel. 1985. The Vulnerability of Point-Voting Schemes to Preference
Variation and Strategic Manipulation. Public Choice 47: 349370.
Nurmi, Hannu. 1987. Comparing Voting Systems. Dordrecht: D. Reidel.
Nurmi, Hannu. 1992. An Assessment of Voting System Simulations. Public
Choice 73: 459487.
Ordeshook, Peter C. 1982. Political Disequilibrium and Scientic Inquiry:
A Comment on William H. Rikers Implications from the Disequi-
librium of Majority Rule for the Study of Institutions. In Peter C.
Ordeshook and Kenneth A. Shepsle, eds., Political Equilibrium. Boston, MA:
Kluwer-Nijhoff.
Ordeshook, Peter C. 1986. Game Theory and Political Theory. Cambridge
University Press.
Ostrom, Elinor. 1990. Governing the Commons. Cambridge University Press.
Pareto, Vilfredo. 1963. The Mind and Society: A Treatise on General Sociology. New
York: Dover.
Pellikaan, Huib, and Robert J. van der Veen. 2002 Environmental Dilemmas and
Policy Design. Cambridge University Press.
462 References
Pettit, Philip. 1996. Institutional Design and Rational Choice. In Robert
Goodin, ed., The Theory of Institutional Design. Cambridge: Cambridge
University Press.
Pildes, Richard H., and Elizabeth S. Anderson. 1990. Slinging Arrows at
Democracy: Social Choice Theory, Value Pluralism, and Democratic Poli-
tics. Columbia Law Review 90 (8): 21212214.
Plamenatz, John. 1973. Democracy and Illusion. London: Longman.
Plott, Charles. 1971. Recent Results in the Theory of Voting. In M. Intriligator,
ed., Frontiers of Quantitative Economics. Amsterdam: North-Holland.
Plott, Charles. 1976. Axiomatic Social Choice Theory: An Overview and Inter-
pretation. American Journal of Political Science 20: 511596.
Plott, Charles, and Michael Levine. 1978. A Model of Agenda Inuence on
Committee Decisions. American Economic Review 68: 146160.
Poole, Keith T., and Howard Rosenthal. 1997. Congress: A Political-Economic
History of Roll-Call Voting. New York: Oxford University Press.
Poole, Keith T., and Howard Rosenthal. 1999. D-NOMINATE After 10 Years:
A Comparative Update to Congress: A Political Economic History of Roll Call
Voting. Available at http://voteview.gsia.cmu.edu.
Poole, Keith T., Howard Rosenthal, and Boris T. Shor. 1999. Voteview for Win-
dows 2.0.0: Roll Call Displays of the U.S. Congress, 17891988. Available at
http://voteview.gsia.cmu.edu.
Popkin, Samuel. 1991. The Reasoning Voter. University of Chicago Press.
Potter, David M. 1976. The Impending Crisis: 18481861. New York: Harper.
Przeworski, Adam. 1991. Democracy and the Market: Political and Economic Re-
forms in Eastern Europe and Latin America. Cambridge University Press.
Radcliff, Benjamin. 1993. The Structure of Voter Preferences. Journal of Politics
55: 714719.
Radcliff, Benjamin. 1994. Collective Preferences in Presidential Elections.
Electoral Studies 13: 5057.
Radner, Roy, and Jacob Marschak. 1954. Note on Some Proposed Decision
Criteria. In R. M. Thrall, C.H. Coombs, and R. L. Davies, eds., Decision
Processes. New York: John Wiley.
Rae, D. W. 1969. Decision Rules and Individual Values in Constitutional
Choice. American Political Science Review 63 (March): 4056.
Rawls, John. 1971. A Theory of Justice. Cambridge, MA: Harvard University
Press.
Rawls, John. 1993. Political Liberalism. New York: Columbia University Press.
Ray, Paramesh. 1973. Independence of Irrelevant Alternatives. Econometrica
41: 987993.
Redman, Eric. 1973. The Dance of Legislation. New York: Simon and Schuster.
Regenwetter, Michel, and Bernard Grofman. 1998. Approval Voting, Borda
Winners, and Condorcet Winners: Evidence from Seven Elections. Man-
agement Science 44: 520533.
Regenwetter, Michel, James Adams, and Bernard Grofman. 2002. On the
(Sample) Condorcet Efciency of Majority Rule: An Alternative View of
Majority Cycles and Social Homogeneity. Theory and Decision 53: 153186.
Reilly, Benjamin. 2001. Democracy in Divided Societies: Electoral Engineering for
Conict Management. Cambridge University Press.
References 463
Reston, James. 1956. The Race between Politics and Education. The New York
Times, July 8, sec. IV, p. 9.
Riker, William H. 1953. Democracy in the United States. New York: Macmillan.
Riker, William H. 1958. The Paradox of Voting and Congressional Rules
for Voting on Amendments. American Political Science Review 52: 349
366.
Riker, William H. 1961. Voting and the Summation of Preferences: An Inter-
pretive Bibliographic Overview of Selected Developments during the Last
Decade. American Political Science Review 55: 900911.
Riker, William H. 1963. The Theory of Political Coalitions. New Haven, CT: Yale
University Press.
Riker, William H. 1965. Arrows Theorem and Some Examples of the Paradox
of Voting. In John M. Claunch, ed., Mathematical Applications in Political
Science. Dallas: The Arnold Foundation.
Riker, William H. 1980a. Implications from the Disequilibrium of Majority
Rule for the Study of Institutions. American Political Science Review 74:
432446.
Riker, William H. 1980b. A Reply to Ordeshook and Rae. In Peter C.
Ordeshook and Kenneth A. Shepsle, eds., Political Disequilibrium. Boston:
Kluwer.
Riker, William H. 1982. Liberalism Against Populism: A Confrontation between the
Theory of Democracy and the Theory of Social Choice. Prospect Heights: Wave-
land Press.
Riker, William H. 1984. The Heresthetics of Constitution-Making: The Pres-
idency in 1787, with Comments on Determinism and Rational Choice.
Riker, William H. 1986. The Art of Political Manipulation. New Haven, CT: Yale
University Press.
Riker, William H. 1990a. Heresthetic and Rhetoric in the Spatial Model. In
James M. Enelow, and Melvin J. Hinich, eds., Advances in the Spatial Theory
of Voting. Cambridge University Press.
Riker, William H. 1990b. Political Science and Rational Choice. In James E.
Alt, and Kenneth A. Shepsle, eds., Perspectives on Positive Political Economy.
Riker, William H. 1993. Introduction. In William H. Riker, ed., Agenda For-
mation. Ann Arbor: University of Michigan Press.
Riker, William H. and Peter C. Ordeshook. 1973. An Introduction to Positive Po-
litical Theory. Englewood Cliffs, NJ: Prentice-Hall.
Riker, William H., and Barry R. Weingast. 1988. Constitutional Regulation
of Legislative Choice: The Political Consequences of Judicial Deference to
Legislatures. Virginia Law Review 74: 373402.
Risse, Mathias. 2001. Arrows Theorem, Indeterminacy, and Multiplicity Re-
considered. Ethics 111: 706734.
Robbins, Lionel. 1937/1932. An Essay on the Nature and Signicance of Economic
Science, 2nd edn. London: Macmillan and Co.
Rose-Ackerman, Susan. 1995. HowChina CouldHave Won: The Nonneutrality
of the Olympic Voting Rules. Journal of Policy Analysis and Management
14(1): 140142.
464 References
Rousseau, Jean-Jacques. 1968/1762. The Social Contract, trans. Maurice
Cranston. London: Penguin.
Rowley, Charles K. 1993. Introduction. In Rowley, ed., Social Choice Theory,
vol. I. Hants, UK: Cheltenham.
Runciman, W.G. 1963. Social Science and Political Theory. Cambridge University
Press.
Saari, Donald G. 1990. Susceptibility to Manipulation. Public Choice 64: 24
41.
Saari, Donald G. 1995a. Basic Geometry of Voting. Berlin: Springer.
Saari, Donald G. 1995b. Mathematical Complexity of Simple Economics.
Notices of the AMS 42: 222231.
Saari, Donald G. 1997. The Generic Existence of a Core for q-Rules. Economic
Theory 9: 219260.
Saari, Donald G. 2000a. Mathematical Structure of Voting Paradoxes: I. Pair-
wise Votes. Economic Theory 15: 153.
Saari, Donald G. 2000b. Mathematical Structure of Voting Paradoxes: II Posi-
tional Voting. Economic Theory 15: 55102.
Saari, Donald G. 2001a. Chaotic Elections. American Mathematical Society
Saari, Donald G. 2001b. Decisions and Elections. Cambridge University Press.
Saari, Donald G. 2003. Capturing the Will of the People. Ethics: 113: 333
349.
Samuelson, Paul. 1977. The Collected Scientic Papers of Paul A. Samuelson,
vol. IV, ed., H. Nagatani and K. Crowley. Cambridge, MA: MIT Press.
Satterthwaite, Mark. 1975. Strategy Proofness andArrows Conditions. Journal
of Economic Theory 10: 187217.
Satterthwaite, Mark A. 2001. Strategy-Proofness and Markets. Social Choice
and Welfare 18: 3758.
Scarf, Herbert. 1960. Some Examples of Global Instability of the Competitive
Equilibrium. International Economic Review 1: 157172.
Schoeld, Norman. 1978. Instability of Simple Dynamic Games. Review of
Economic Studies 45: 575594.
Schoeld, Norman. 1982. Instability and Development in the Political Econ-
omy. In Peter C. Ordeshook and Kenneth A. Shepsle, eds., Political Equi-
librium. Boston, MA: Kluwer-Nijhoff.
Schoeld, Norman. 1995. Democratic Stability. In Jack Knight and Itai Sened,
eds., Explaining Social Institutions. Ann Arbor: University of Michigan
Press.
Schwartz, Joseph M. 1995. The Permanence of the Political: A Democratic Critique
of the Radical Impulse to Transcend Politics. Princeton University Press.
Schwartz, Thomas. 1982. Rationality and the Myth of the Maximum. Nous
6: 97117.
Sears, David O., and Carolyn L. Funk. 1991. The Role of Self-Interest in Social
and Political Attitudes. In M. P. Zanna, ed., Advances in Experimental
Social Psychology, vol. XXIV, 291. New York: Academic Press.
Sellers, Charles. 1966. James K. Polk: Continentalist, 18431846. Princeton
University Press.
Sen, Amartya. 1970. Collective Choice and Social Welfare. San Francisco:
Holden-Day.
References 465
Sen, Amartya. 1982. Choice, Welfare and Measurement. Cambridge, MA: Harvard
University Press.
Sen, Amartya. 1993. Internal Consistency of Choice. Econometrica 61 (May):
495521.
Sen, Amartya. 1995. Rationality and Social Choice. American Economic Review
85: 124.
Sen, Amartya. 1999. The Possibility of Social Choice. American Economic
Review 89: 349378.
Shapin, Steven. 1994. A Social History of Truth. University of Chicago Press.
Shaviro, Daniel. 2000. When Rules Change. University of Chicago Press.
Shepsle, Kenneth A., and Mark S. Bonchek. 1997. Analyzing Politics: Rationality,
Behaviors, and Institutions. New York: W.W. Norton.
Shepsle, Kenneth A., and Barry R. Weingast. 1981. Political Preferences for
the Pork Barrel: A Generalization. American Journal of Political Science 25:
96111.
Shepsle, Kenneth A., and Barry R. Weingast. 1984. Political Solutions to
Market Problems. American Political Science Review 78: 417434.
Shiffrin, Richard M. and Robert M. Nosofsky. 1994. Seven Plus or Minus
Two: A Commentary on Capacity Limitations. Psychological Review 101:
357361.
Skaperdas, Stergios. 2002. Restraining the Genuine Homo Economicus:
Why the Economy Cannot Be Divorced from Governance. University of
California, Irvine, manuscript.
Snow, C.P. 1951. The Masters. London: Macmillan.
Stampp, Kenneth M., ed. 1974. The Causes of the Civil War. Englewood Cliffs,
NJ: Prentice-Hall.
Stepan, Alfred, and Cindy Skach. 1994. Presidentialism and Parliamentarism
in Comparative Perspective. In Juan Linz and Arturo Valenzuela, eds., The
Failure of Presidential Democracy, vol. I, Comparative Perspectives. Baltimore:
Johns Hopkins University Press.
Stigler, George, and Gary Becker. 1977. De Gustibus non Disputandum.
American Economic Review 67: 7690.
Stiglitz, Joseph. 2000. The Contributions of the Economics of Information
to Twentieth Century Economics. Quarterly Journal of Economics 115:
14411478.
Stratmann, Thomas. 1997. Logrolling. In Dennis C. Mueller, ed., Perspectives
on Public Choice: A Handbook. Cambridge University Press.
Strom, Gerald S. 1990. The Logic of Lawmaking: A Spatial Theory Approach.
Baltimore: Johns Hopkins University Press.
Sunstein, Cass R. 1988. Constitutions and Democracies: An Epilogue. In Jon
Elster and Rune Slagstad, eds., Constitutions and Democracy. Cambridge
University Press.
Sunstein, Cass R. 1990. Preferences and Politics. Philosophy and Public Affairs
20(3): 999.
Sunstein, Cass R. 1993. The Partial Constitution. Cambridge, MA: Harvard
University Press.
Tabarrok, Alexander, and Lee Spector. 1999. Would the Borda Count Have
Avoided the Civil War? Journal of Theoretical Politics 11: 261288.
466 References
Tangian, A.S. 2000. Unlikelihood of Condorcets Paradox in a Large Society.
Social Choice and Welfare 17: 337365.
Tanguiane, A.S. 1991. Aggregation and Representation of Preferences: Introduction
to Mathematical Theory of Democracy. Berlin: Springer-Verlag.
Taylor, Michael J. 1969. Proof of a Theorem on Majority Rule. Behavioral
Science 14 (May): 228231.
Thagard, Paul. 1992. Conceptual Revolutions. Princeton University Press.
Thomas, James D., and William H. Stewart. 1988. Alabama Government and
Politics. Lincoln: University of Nebraska Press.
Tocqueville, Alexis de. 1969/1851. Democracy in America, trans. George
Lawrence. New York: Doubleday.
Tovey, Craig A. 1995. Dynamical Convergence in the Euclidean Spatial
Model. In William A. Barnett, Herve Moulin, Maurice Salles, Norman
J. Schoeld, eds., Social Choice, Welfare, and Ethics. Cambridge University
Press.
Tribe, Laurence. 1988. American Constitutional Law, 2nd edn. Mineola, NY:
The Foundation Press.
Tsetlin, Ilia, Michel Regenwetter, and Bernard Grofman. 2001. The Impartial
Culture Maximizes the Probability of Majority Cycles. Manuscript.
Tuck, Richard. 1993. The Contribution of History. In Robert E. Goodin and
Philip Pettit, eds., A Companion to Contemporary Political Philosophy. Oxford:
Basil Blackwell.
Tullock, Gordon. 1992. Is There a Paradox of Voting? Journal of Theoretical
Politics 4: 225230.
Tushnet, Mark. 1988. Red, White, and Blue. Cambridge, MA: Harvard University
Press.
Van Deemen, Adrian. 1999. The Probability of the Paradox of Voting for Weak
Preference Orderings. Social Choice and Welfare 16: 171182.
Van Deemen, Adrian, and Noel P. Vergunst. 1998. Empirical Evidence of
Paradoxes of Voting in Dutch Elections. Public Choice 97: 475490.
Wald, Alan M. 1987. The New York Intellectuals. Chapel Hill: University of
North Carolina Press.
Waldron, Jeremy. 1999. The Dignity of Legislation. Cambridge University Press.
Walt, Stephen. 1999. Rigor or Rigor Mortis? Rational Choice and Security
Studies. International Security 23: 548.
Weale, Albert. 1984. Review Article: Social Choice Versus Populism? An
Interpretation of Rikers Political Theory. British Journal of Political Science.
14: 369385.
Weale, Albert. 1995. William Riker and the Theory of Democracy. Democra-
tization 2: 377395.
Weber, Robert J. 1995. Approval Voting. Journal of Economic Perspectives 9:
3949.
Weingast, Barry R. 1979. A Rational Choice Perspective on Congressional
Norms. American Journal of Political Science 23: 245262.
Weingast, Barry R. 1998. Political Stability and Civil War: Institutions, Com-
mitment, and American Democracy. In Robert H. Bates, Avner Greif,
Margaret Levi, Jean-Laurent Rosenthal, and Barry R. Weingast, eds.,
Analytic Narratives. Princeton University Press.
References 467
Wiebe, Robert H. 1995. Self-Rule: A Cultural History of American Democracy.
University of Chicago Press.
Wildavsky, Aaron. 1974. The Politics of the Budgetary Process, 2nd edn. Boston:
Little Brown.
Wiles, Peter. 1969. A Syndrome, not a Doctrine. In Ghita Ionescu and Ernest
Gellner, eds., Populism: Its Meanings and Characteristics. London: Weidenfeld
and Nicolson.
Williamson, Oliver E., and Thomas J. Sargent. 1967. Social Choice: A
Probabilistic Approach. The Economic Journal 77 (December): 797813.
Wittman, Donald. 1995. The Myth of Democratic Failure. University of Chicago
Press.
Wolff, Robert Paul. 1970. In Defense of Anarchism. New York: Harper and
Row.
Worsley, Peter. 1969. The Concept of Populism. In Ghita Ionescu and Ernest
Gellner, eds., Populism: Its Meanings and Characteristics. London: Weidenfeld
and Nicolson.
Young, H. Peyton. 1974. An Axiomatization of Bordas Rule. Journal of
Economic Theory 9: 4352.
Young, H. Peyton. 1988. Condorcets Theory of Voting. American Political
Science Review 82: 12311244.
Young, H. Peyton. 1997. Group Choice and Individual Judgments. In Dennis
C. Mueller, ed., Perspectives on Public Choice: A Handbook. Cambridge
University Press.
Zywicki, Todd J. 1994. Senators and Special Interests: APublic Choice Analysis
of the Seventeenth Amendment. Oregon Law Review 73: 10071055.
Index
Adams, John Quincy 14, 56, 299300,
301, 382383
Africa 1, 420
AfricanAmericans, voting rights and the
Depew amendment 218, 219,
223, 224
agenda control 5, 17, 35, 72, 158, 159,
160, 166171, 384
and academic attacks on democracy
441
and the Depew amendment 219
frequency of 193
and majoritarian processes 10
and multidimensional disequilibrium
378
and multidimensional issue spaces 178,
196
and Riker
Art of Political Manipulation 407
basic argument pattern 37
case against populism 410
defense of liberalism 412
and voting methods 9
see also new dimensions
agenda-setting 15, 344, 383
agricultural appropriations (1958) 21,
310311, 329334, 377
and the Andersen amendment 330
and the Javits amendment 329, 330,
331
and the OToole amendment 329,
331
and the Whitten amendment 329, 330,
331332
Alexander II, Tsar 420
Amadae, S.M. 24, 28
American Civil War 20, 191, 302, 309,
407
and the Depew amendment 223, 225
and Pareto 427
and Rikers disequilibrium hypothesis
428
and Rikers liberalism 422
and slavery 293, 294, 295
and the Wilmot Proviso 240, 243
American Political Science Review 23
American Psychological Association
(APA), presidential elections 49,
86
anarchism 440
antidemocratic doctrines 3
antiplurality see negative plurality
approval voting 44, 45, 48, 66
accuracy 6465
and the APA presidential elections 49
and Arrows possibility theorem 89
and the Copeland method 50, 51
evaluation 70, 72
and the Lincoln election 260, 277,
278, 279, 281
Arab world 1
Argentina, populism in 420
Aristotle 10, 29
Arnold, Matthew 2
Arrow, Kenneth 4, 10, 15
condition of unrestricted domain (U)
17, 9394, 95122
and cyclebusting voting rules
113122
and redistributional instability
99108
and self-interest 108113
and simulations of homogeneity
9699
contraction-consistency independence
condition (IIA (RM)) 124, 140,
146, 151, 375
and democratic instability 440
on dictatorship 442
impossibility theorem 3, 10, 15, 16, 17,
70, 72, 329, 375
440, 441, 443
collective choice rule 80
468
Index 469
Condition D (nondictatorship) 81,
93
Condition P (Pareto principle) 81,
9293
and the Condorcet paradox 8,
7879, 80, 82, 84, 85, 89
and cycling 83, 8692, 410
and the economy 434435
independence of irrelevant
alternatives 81
justifying the theorems conditions
9294
and the market 22, 30
and new dimensions 378, 379, 384
and the paradox of voting 5, 35
and positive political theory 435
and the Powell amendment 43
and public choice theory 433
and Sens constructive social choice
theory 432
and social choice theory 70, 72, 76,
78, 85
and social states 79
and voter preferences 39
independence condition (IIA) 17, 56,
123157, 440, 442
consequences of 146150
and constitutional engineering 157
as defending the wrong principle
124131
awed irrelevance justication in
136142
and independence 151156
and justication of voting rules
142146
and the market model of democracy 28
196
public choice theory 1013
Rikers attempt to demonstrate an
Arrovian cycle 310311
and Rikers theory of dimensional
manipulation 292303
and the Rochester school 24
Social Choice and Individual Values 76,
83, 127, 143
and strategic voting 161
Asia, populism in postcolonial 420
Athens, elections in 29
Austen-Smith, David 24, 162, 215
Australian House of Representatives 65
axiomatic approach 16
balanced and unbalanced cycles 117122
ballots, infeasible or irrelevant alternatives
on 136
Banks, Jeffrey S. 24
Barker, Ernest 2
Baron, David P. 102
Barry, Brian 123, 143
Bayesian Nash equilibrium, and the Powell
amendment 197, 207
Beer, Samuel H. 421
Bell, Daniel 441
Bell, John 20, 258, 259, 267, 268, 269,
270, 271279, 285
Benson, Lee 300
Bentham, Jeremy 419, 421
Bentham voting method 52, 53, 54
Berlin, Isaiah 421
bipolar-culture assumption, and
Condorcet efciency 48
Bjurulf and Niemi, on Scandinavian
parliaments 21, 335336,
344353, 377
Blacks rule 48
Blydenburgh, John C. 21, 362
and the Revenue Act (1932) 21, 335,
337344, 376
Bohman, James 382
bolshevism 10, 440, 442
see also communism
Bonapartism 430, 431
Bonchek, Mark S. 15, 29, 6869, 336,
434, 441, 442
Borda count 5, 6, 7, 8, 9, 44, 45, 51, 55,
442
accuracy and fairness 64, 68, 7071
and agenda control 167168, 169172
and agricultural appropriations (1958)
333, 334
and Arrows independence
condition 123, 124, 125126,
127, 128, 138, 141
consequences of 147, 150
justication of 144146
and manipulation 154155, 156
and cardinality 78
and Condorcet efciency 48
and cyclebusting 114, 117, 118, 120,
121
and the Danish parliament election
371, 372
evaluation 5556, 59, 61, 62, 63, 64,
67, 68
and the Finnish Electoral College 364,
365, 366
and impartial-culture assumptions 48,
54
470 Index
Borda count (cont.)
and individual preferences 392
and the Iowa corn vote 369
277, 279, 281
191
and the Muscle Shoals vote 356, 357
and numbers of parties 66
and plurality grading 6061
as positional voting method 46
and public opinion on US military
intervention overseas 374, 375
rankings 52, 53
Rikers recommendation 7071
and simulations of homogeneity 97
and social homogeneity 49
winners 50
Boudinot, Elias 298
Brazil, populism in 420
Breckinridge, John C. 259, 260, 267,
268, 269, 270, 271, 272, 273,
274276, 285
Brennan, G. 102, 136, 166
Britain
elections 59, 87
Labour Party working party on electoral
reform 69
as a populist society 421
Buchanan, James 28, 264, 266, 288289,
290, 291, 395
Buckley, William F. 425
Budge, Ian 91
Bueno de Mesquita, Bruce 24, 26, 28,
344
Burnham, James 409, 425, 429,
430431, 433, 440
Butler, Pierce 323
Cain, Michael J.G. 15, 441, 442
Californians for Electoral Reform 53
Calvert, Randall L. 215
Cambodia, Khmer Rouge 420
Cambreleng, Churchill 249
Carlyle, Thomas 2
Chamberlain, John R. 49, 86
chaos theorems see McKelvey and
Schoeld chaos theorems
Chile 425, 431
presidential election (1970) 59
China
and democratic irrationalism 425
student movement for democracy 2, 4,
443
Citrin, Jack 109
coffee-break cycles 337, 372376
Cohen, Jerry L. 49, 86
Cohen, Leonard 443
Cohn, Jonathan 25, 28
Cold War 1
collective decision making, paradoxes of
15
Collective Rationality, and Arrows
independence condition 127
communism 1, 420, 423
Condorcet criterion 55, 56, 6162,
68
see also pairwise comparisons
Condorcet voting method 5, 79, 28, 35,
44, 55, 387, 435, 440
and American presidential elections 88
as an ideal rule 68
and Arrows condition of unrestricted
domain 99100, 108
and Arrows independence condition
124, 128, 138139
consequences of 147
justication of 144145, 146
and manipulation 154155, 156
and Arrows theorem 8, 7879, 80, 82,
84, 85, 89
Condorcet efciency 4748
Condorcet-consistent rules 56
and the Copeland method 50
and cyclebusting 114
and the Danish Parliament election
371
and dictatorship 434, 435
evaluation 55, 58, 59
and the Finnish electoral college 364,
365
Jury Theorem 63
and logrolling 163
184, 186, 187
pairwise ordering 49
rankings 52, 53
and the Revenue Act (1932) 337
switch to other methods 442
see also pairwise comparisons
Condorcet winners 4445, 46, 49, 50,
51, 7071
and agenda control 169
domain 99100
Index 471
and public opinion on US military
intervention overseas 374
and simulations of homogeneity
9699
and US Senate deliberations on Muscle
Shoals 336
consumer economics, and Arrows
contraction consistency 126, 127
Coombs, Clyde H. 49, 86
Coombs voting method 48, 49, 65, 66
Cooter, Robert 7576
Copeland voting method 48, 50, 51, 64,
154155
cumulative voting method 5259
cycling 11, 15, 35
441
domain 17, 99
and Arrows impossibility theorem 17,
83, 8692
balanced and unbalanced cycles
117122
coffee-break cycles 337, 372376
and the Condorcet voting method
5, 8
cyclebusting voting rules 113122
and the Danish prime minister post
336
and democratic irrationalism 408
and the Depew amendment 197, 217,
218, 219220, 233, 235237
as an empirical improbability 378
Iowa senators and anticorporate farming
legislation 336, 369370
and the Lincoln election 54, 258, 259,
279, 301, 376
and logrolling 164, 165
and market failures 440
mistaken claims of 2021, 335
Danish Parliament election (1994)
336, 370372
Finnish electoral college 336,
362369
Muscle Shoals 21, 336, 353361
Revenue Act (1932) 335
and Riker 376377
Scandinavian parliaments 335336
Shepsle and Boncheks cycles 336,
361362
196
structured preference orders 182
and the Powell amendment 20, 197,
198, 201, 353
in Rikers Art of Political Manipulation
407
and Rikers basic argument pattern 37,
178
Rikers cycles 181, 310334, 335,
353
and the agricultural appropriations
vote (1958) 21, 310311, 329
334
and the slavery issue 293
and the US Federal Convention
(1787) 21, 310, 316, 311316,
321328, 329, 376
and stability 10
unimportance of 21
and the US Civil War 20
what-if cycles 336, 372376
and the Wilmot Proviso 241, 243, 258,
301, 376
Czech Parliament 91
Dahl, Robert A. 2
A Preface to Democratic Theory 8384
A Preface to Economic Democracy
423424
Davidson, Donald 40
Davis, Jefferson 268
Davis, Senator John 250251, 253
deadlock, and Arrows possibility
theorem 84
democracy
academic attack on
and elite theory 427, 428, 442
literature on 916
Madisonian and populistic 423424
parliamentary 422423
and plebiscitarianism 409, 418
polyarchal 424
preconditions to 418
presidential 422423
and the public good 36, 419, 424425
revival of 443
Riker on populism and 418425
democratic irrationalism 34, 16, 22,
2343
academic opinion on 916, 426
and cycling 408
and democracy defended 2731
doctrine of 361, 409, 425431, 436,
438
and the economy 434435
472 Index
democratic irrationalism (cont.)
and the introduction of new
dimensions 382, 384385, 386
and knowledge of other minds 3943
and multidimensional issue spaces 192
and rational choice theory 23
Riker on democracy as arbitrary
409410
Riker, William on democracy as
meaningless 410
17, 3739, 258
and Rikers Liberalism against Populism
4, 23, 3136
and strategic voting in Scandinavian
parliaments 352353
Democratic Party
and the election 16, 290
and the Dred Scott decision 396
and the KansasNebraska Act 307
and the Missouri Compromise 306
see also Lincoln election
Denmark
Danish national elections 52, 8889
Danish prime minister post 336
general election for parliament (1994)
370372
strategic manipulation in 352
Denzau, Arthur 204, 212
Depew amendment 20, 91, 193, 195,
217240, 297, 407
alternative interpretation 231240
and the Bristow amendment 223, 226,
228230, 231, 234235
and cycling 197, 217, 218, 219220,
233, 235237
and the Force Bill 224
historical background 221231
and the Insurgent Republicans 231,
238
and irrational voters 220221
and new dimensions 379
and the Oregon voting system 221222
and the Rayner amendment 226228
and the Sutherland amendment 218,
220, 228231, 233, 234235, 239
developing world, and democracy 1
Dewey, John 2
dictatorship
and Arrows impossibility theorem
141
and democracy 440, 442
and instability 434
direct democracy 418
disequilibrium 21
and democracy 440, 441
and equilibrium in economics 435
436
dimensions 380, 383
and McKelvey and Schoelds chaos
theorems 17, 196
and multidimensional issue
spaces 180181
and positive political theory 435
and Rikers case against populism 410
Rikers hypothesis 916, 17, 20, 42,
192, 196, 240
410
and Ordeshook 435
and Pareto 428
and slavery 293
and the Wilmot Proviso 241, 242,
246
and Shepsle and Boncheks cycles 362
Douglas, Stephen A. 20, 258, 259260,
261, 268269, 271279, 363
and the Baltimore Convention 267,
287
choice of Lincoln over 281292
and the Congressional Party 266
debate with Lincoln at Freeport 22,
392, 393396
and the Democratic Party 267, 269,
292
and the doctrine of popular
sovereignty 193, 265, 269
and the Lecompton constitution
394395
Downs, Anthony 189
Dred Scott decision (Supreme Court)
396
and the Lincoln election 264265,
269, 290, 303
and Lincolns debate with Douglas at
Freeport 395
Dryzek, John 13, 15
Dutch parliamentary elections 88
Duvergerian equilibrium 66
Duvergers Law 289292
Index 473
Eastern Europe, fall of communist
regimes 1
economic libertarianism, democratic
irrationalist justication of 30
economics
and Arrows impossibility theorem
434435
competitive equilibrium in the
economy 436438
doctrine of noncomparable utility 17
economic instability 436
and market manipulation 439
material-welfare school 26
Post-Autistic Economics Movement
7476
and social choice theory 72
see also markets
Eisenhower, Dwight D. 198, 210, 211,
212, 213
electoral colleges
and the Federal Convention (1787)
318319
Finnish 336, 362369
and the Lincoln election 282, 284, 292
electoral heart 188
elimination rules 56, 6566
elite theory 442
in Pareto 427, 428
Ellis, Susan 238
Elster, Jon 27, 107, 323, 388
empirical failures, and multidimensional
issue spaces 178182
Enelow, James M. 92
equality, Riker on 32, 36, 416, 417
equilibrium
Bayesian Nash 197, 207
competitive equilibrium in the
economy 436438
in economics 435436
Kramers dynamical model of political
equilibrium 185
178, 181182, 191
adding friction to 188
and disequilibrium 180181
structure-induced 178
Pareto-optimal 388, 426427, 436, 437
and Paretos social theory 430
in public choice 13, 14
see also disequilibrium
Essex result 327
ethical voters 111
Europe, democracy in the interwar
period 2
Everett, Edward 267
factvalue distinctions 74
Farquharson, Robin, Theory of Voting, on a
US Senate vote 372373
fascism 1, 10, 423, 431, 440, 442
Federal Convention (1787) 310,
311329, 376
and the Brearley Committee 323
Committee on Remaining Matters 316,
318319, 320, 321, 325
and electoral college 318319
and the Essex result 327
and executive selection
by electors 314, 315316, 317318,
320, 321, 324325
by joint ballot 315, 316, 317,
324325, 327
by national legislature 311, 314,
316, 317, 320
and the Great Compromise 311, 318,
319, 321
and Houstons motion 312313
and the New Jersey plan 312
and political opportunism 325326
and presidential term of ofce
314
and rational action 313
and the separation of powers
movement 316
small and large states 313314,
319320
and the Virginia Plan 311, 312
Fehr, Ernst 104
Feld, Scott L. 50, 85, 87, 182, 183
Feldman, Allen M. 10, 442
Felsenthal, Dan S. 50, 51, 87
Fenno, Richard F. 215
Ferejohn, John A. 102
Fillmore, Millard 264, 288289, 290,
291
Finland
civil war in 363, 366
electoral college 336, 362369
history of democracy in 368
strategic manipulation in 352
Fiorina, Morris P. 102, 180, 435
Fischbacher, Urs 104
Fischer, David Hackett 326
Fogel, Robert William 263, 305
Franklin, Benjamin 295, 328, 329
Fremont, John C. 263, 264, 290
French Revolution 420
Funk, C.L. 109
game theory 24, 27, 67
and universalism 102104
474 Index
Gaubatz, Kurt Taylor, on public opinion
on US military intervention
overseas 373375
Gehrlein, William V. 49, 98, 115
general will, in Rousseau 23, 3334
Germany
fall of the Weimar Republic 10, 440
Nazism 13
Gibbard, A. 155
GibbardSatterthwaite manipulation
ndings 22, 161, 410, 439
Giddings, Joshua 300, 301
Glazer, A. 102
global scale of democratization 1
Goldwater, Senator Barry 406
Goodin, Robert 24, 107108
government failure and market failure
434
Green, Donald 27, 28, 29, 109, 178179,
181, 194195
Grofman, Bernard 50, 52, 85, 87, 89,
182, 183
Gross, Donald 369
Habermas, J urgen 2
Hamlin, A. 136
Handbook of Political Science (Goodin and
Klingemann) 24
Hansson, B. 148, 149
Hardin, Russell 11, 15, 123, 143, 441
Hare voting system 48, 56, 65, 66, 140
evaluation 59, 66
preference voting 53, 54
harmful manipulation 159, 160, 378
frequency of 193196
and logrolling 163
201
in Scandinavian parliaments 353
Hauptmann, Emily 2829
Haynes, George H. 221223
hegemony, and rational choice theory 26
Heseltine, William B. 303
Hinich, Melvin J. 188189, 190
Hitler, Adolf 13
Holmes, Stephen 110, 113
Hume, David 110, 194, 195
ideological coherence, constraints of 196
impartial-culture assumption 48, 49, 54,
55, 8889
and the Condorcet efciency 4748
183
and simulations of homogeneity 96, 97
inaccuracy
and the Condorcet voting method 9
democratic voting as inaccurate 44
and Rikers basic argument pattern 37
Rikers inaccuracy hypothesis 52
India, populism in 420
individual motivation, and preference
rankings 192
Indonesia 1
International Olympic Committee vote
375376
Iran 1
irrationalism see democratic irrationalism
IsraeliPalestinian conict 1
Jackson, Andrew 296, 299
Jackson, Senator Henry 397, 400, 405
Japan, and US nerve gas materials 399,
400401, 403
Jefferson, Thomas 295
Jillson, Calvin C. 320
judicial review 12, 15
justice, Platonic and Marxist conceptions
of 31
KansasNebraska Act (1854) 288, 289,
294, 295, 307
Kasza, Gregory 26
Katznelson, Ira 440, 441, 442
Kiewiet, D. Roderick 108109
Kinder, Donald R. 108109
King, Ronald F. 238
Klingemann, Hans-Dieter 24
Koford, K.J. 165
Kramer, Gerald H. 185
Krehbiel, Keith 197198, 206, 207208,
211, 214215, 361
Kuga, K. 97
Kuttner on the virtues and limits of
markets 433
Kurrild-Klitgaard, Peter 52, 8889, 336,
370372
Lagerspetz, Eerik 21
Latin America 1, 420, 423
Lecky, William 2
Lecompton constitution 394395
legislative power, and democracy 418
Lenin, V.I. 2, 420
Levin, Jonathan 50
Levine, Michael 169, 407
Lewin, Leif 109
liberal political theory 2
Index 475
liberalism
Rikers case for 25, 409, 411417
and the NixonAgnew problem 415
and random rejection of candidates
413416
and the rejection of populism 413,
416
and the retention or rejection of
ofcials 411413
liberalist democracy 3, 22
and the doctrine of democratic
irrationalism 31
libertarianism 4
liberty, Riker on 32, 33, 36, 416, 417
Lincoln, Abraham 301, 392396
debate with Douglas at Freeport 22,
392, 393396, 407
Lincoln election 20, 54, 92, 258280,
407
and antebellum politics 281
and the Baltimore Convention
267268, 287
choice of Lincoln over
Douglas 281292
and the Civil War 287
and the Congressional Party 266
and the Constitutional Unionists
259260, 267, 269270, 285
and cycles 54, 258, 259, 279, 301,
376
and the Democrats 259, 265, 266, 267,
269, 271, 285, 286, 289, 292
and Duvergers Law 290
northern 261, 262, 263, 269, 286
southern 262, 269, 286
and the doctrine of popular
sovereignty 265, 269, 270
and Duvergers Law 289292
and immigration to the north 262
and the Lower South and
secession 286287
and the nativist American Party 263
and the Republicans 263, 264265,
269, 285, 286, 290
Riker on the 20, 54, 92, 240, 259,
258260, 280, 281282, 376, 409,
410
and the case against populism 202,
409, 410
and the slavery issue 294
and the Supreme Court Dred Scott
decision 264265, 269, 290, 303
voter preferences 271280, 281
voting system and candidacies 285286
and the Whigs 261262, 263, 264
Lindsay, Alexander Dunlop 2
Lippmann, Walter 2
Little, I.M.D. 76
logical positivism, and Arrows
logrolling (vote trading) 17, 159, 160,
163166, 171
and agricultural appropriations
(1958) 333
as welfare-enhancing or
welfare-reducing 163164, 171
Lomasky, L. 102, 166
McDonald, Forrest 314, 319
McGovern, Senator 213
Machover, Moshe 51, 88
McKelvey, Richard D. 22, 102, 179
McKelvey and Schoeld chaos
theorems 17, 22, 30, 170, 184,
185, 196, 410
441
and the economy 438
and new dimensions 379, 385
Rikers interpretation of 17, 173176,
186190, 196
McKelvey voting model 383, 392, 407
Mackie theorem 171
McMillan, H. 102
Machiavellian elite theories 430
Madison, James 22, 35, 314, 320, 322,
328
Madisonian democracy 423424
Magnuson amendment and nerve gas 22,
361, 381, 393, 396407
and the ChurchCooper amendment
398399, 401, 403, 404
and the Gravel amendment 393, 397,
398399, 400, 401, 403404, 405
and the Harry Byrd resolution 400,
402, 404
Maine, Sir Henry 2
majoritarian democracy, Riker on 23,
4951, 159, 302
majoritarian processes, and dictatorship
10
majority cycling, and Arrows possibility
theorem 71, 72, 8586
majority voting 44, 107
cycles and individual preferences 390
and the Lincoln election 283
and plurality rule 5
and social choice 14, 15, 44, 5556,
83
and stability 10
voting problems with 5
476 Index
Mandeville, Bernard 110
manipulation 17, 35, 161
154156, 157
and Arrows possibility theorem 70,
7294
economic 436, 438440
dimensions 381
and the Magnuson amendment 406
in politics 436
38, 258, 280
and Rikers defense of liberalism 412,
413
Rikers manipulation argument 160,
158160, 162, 171, 194196,
292303, 344353, 384
in Scandinavian parliaments 21, 336
see also agenda control; harmful
manipulation; multidimensional
issue spaces; strategic voting
Maoz, Zeev 50, 87
markets
arbitrary and meaningless nature
of 437
and democracy 28
force and fraud in 437
and instability 433434
manipulation of 438440
market failure 30, 434, 440
and the Pareto-optimal competitive
equilibrium 426427
and politics 14, 15, 22
Martin, Luther 322
Marx, Karl 2, 31
Mashaw, Jerry L. 10, 441, 442
Massachusetts state constitution, and the
Essex result 327
mean voter theorem 183
meaningless, democracy as see democratic
irrationalism
Merrill, Samuel 48, 49, 55
Mexico
249, 250, 251
Michaelsen, William B. 317, 321,
327333
Michels, Robert 2, 3, 442
Miller, D.T. 112113
Miller, G.J. 103
Milner, Helen 440, 441, 442
Missouri Compromise
and the Dred Scott case 264
and Kansas 306
and the Lincoln election 258, 261
and northern politics 302
and Riker on slavery 294, 295, 296,
297, 298, 307
and slavery votes in Congress 305
Monroe, James 305
Moore, Glover 297
Morris, G., and the Federal Convention
(1787) 316, 320, 322323,
325326, 328, 407
Mosca, Gaetano 2, 3, 431, 442
movement and action, and voters
preferences 38
Mueller, Dennis C. 156
multidimensional chaos 173
multidimensional disequilibrium 378,
392
multidimensional issue spaces 2122, 35,
72, 159, 170, 173176, 310
adding back friction 185191, 196
and democracy 419
experimental and empirical failures
178182
and harmful manipulation 193196
nominatively attractive point in
176178
178
and structured preference orders
182185
and voter preferences 178
see also new dimensions
Munger, Michael C. 188189, 190
Murdoch, Iris 74
Muscle Shoals (Senate deliberation on)
21, 336, 353361, 377
and the Coolidge administration 355
Jones proposal 354, 356
Norris proposal 354
and the Republicans 356, 357
and the southern Democrats 356
Underwood proposal 354355, 356,
357
Mussolini, Benito 425, 431
Mustasa, Didymus 108
Nagatani, H. 97
Nalebuff, Barry 50
Nansons voting system 48
Nash voting method 52
negative majorities, and the Revenue Act
(1932) 337338
negative plurality 66
Index 477
Neufeld, John L. 21
Neufeld, Hausman and Rapoport on
Muscle Shoals 21, 336, 353361,
377
new dimensions 378408
deliberation and disequilibrium
386392
introduction of issues and dimensions
379384
models of 384
see also multidimensional issue spaces
New Republic 25
Niemi, Richard G. 21, 86, 182, 335336
Nixon, Richard M. 212
and the Magnuson amendment and
nerve gas 397, 398, 399400, 401,
402, 405, 406
and the NixonAgnew problem 415
normative democratic theory 2
Northern Ireland 69
Norway, strategic manipulation in 352
Nurmi, Hannu 47, 49, 69
objective utility, Pareto on 429
Olsons logic of collective action 111
opinion polls 412
Oppenheimer, J.A. 103
opportunity costs, of rational-choice
scholarship 28
opposition to democracy 2
Ordeshook, Peter C. 11, 24, 179, 435,
442
Packwood, Bob 397
pairwise comparisons 54
and agricultural appropriations (1958)
333
137139, 141
and dictatorship 435
278, 279, 281
and logrolling 163
and the Muscle Shoals vote 356357
Papua New Guinea 67, 68
paradox of voting 5
see also Condorcet voting method
Pareto criterion 76, 416
Pareto, Vilfredo 2, 3, 22, 409, 425431,
442
and elite theory 427, 428
on logical and nonlogical conduct 426
on objective and subjective utility 429
and residue theory 427428
Pareto-optimal equilibrium 388,
426427, 436, 437
parliamentary democracy 422423, 430
Parsons, Talcott 344
participation, Riker on 23, 32, 36,
416417
path dependence, and the Condorcet
voting method 9
Pellikaan, Huib 111
perestroika-glasnost movement in
American political science 26
Philippines 1
Pierce, Franklin 288, 289
Plato 2, 31
plebicites 417, 420
plebiscitarianism 409, 418
Pliny the Younger 43, 168, 407
Plott, Charles 11, 169, 180, 407, 441,
442
plurality grading 6061
plurality runoff 44, 48, 56
and Condorcet efciency 48, 49, 60
and elimination 65
evaluation 59, 5259, 60, 70, 72
367
and the impartial-culture assumption
47
and the Lincoln election 278, 279, 281
plurality voting 59, 44, 49, 51
accuracy 64
142, 157
and the Borda count 6
281, 282284, 285, 292
with more than two parties 6667
in Papua New Guinea 67
problems with 59
pure plurality 68
rankings 52, 53
and Rikers defense of liberalism 412
political tactics, and social welfare 15
political/moral scarcity, and the
Pareto-optimal competitive
equilibrium 426
478 Index
politicians and opportunism, and the
Federal Convention (1787)
325326
Polk, James K. 246247, 248250,
254256
Pollard, Edward 302
polyarchal democracy 424
Poole, Keith T. 90, 91, 102, 189190,
191, 193, 304306
populism
American populist movement
419420, 421
in Britain 421
Riker on populism and democracy
418425
Rikers case against 23, 3137,
409411
populist democracy 3, 13, 22, 23, 31
Portugal, democratization of 1
positional voting 46, 56
positive political theory 3, 22, 24, 159,
181182, 431
and disequilibrium 435
postmodernist hegemony, and rational
choice theory 26
Potter, David M. 306, 308, 394, 395
Powell, Adam Clayton 198, 204, 205,
209, 213
Powell amendment 20, 43, 43, 91, 193,
197216, 361, 407
and the Ayres amendment 214215
and the Brown v. Board of Education
Supreme Court decision 198, 205,
209, 210
and the Civil Rights Act (1956) 206
and cycling 20, 197, 198, 201, 353
Desegregationist Republicans 202,
204
and harmful manipulation 197, 198,
201
and the NAACP 204209, 210211
northern Democrats 207, 207208,
209, 211, 212
Powellites 202, 204205, 206, 207,
209
recorded votes 199200
School-Aiders 202, 206, 207, 210
Segregationist Republicans 202204,
206
southerners 202
and strategic voting 20, 197, 198, 204,
347350, 384
and the Wainwright amendment
213214, 215
power, social rationality and the
concentration of 1415
preference rankings 20, 21, 44
preference voting 53
preference-development hypothesis
303309
presidential democracy 422423
probabilistic voting 196
147148, 150
proportional representation 66
386
Przeworski, Adam 13, 441
public choice theory 13, 27, 29, 30
and markets 433, 434, 437
and the Wilmot Proviso 246
public good
and democracy 36, 419, 424425, 430
and the general will 34
and Rikers liberalism 35
public interest
public opinion, on US military
intervention overseas 373375
pure plurality 68
Radcliff, Benjamin 88
radical interpretation, Davidsons
principles of charity in 40
Rapoport, Amnon 50, 87
see also Neufeld, Hausman and Rapoport
Rappoport, Peter 7576
Raskin, Marcus 36
rational actions, and self-interest
109110, 113
rational choice theory 23, 2829, 106,
194
critics of 2526, 27
Rochester school of 13, 22
Ratner, R.K. 112113
Rawls, John 2, 419, 421
reciprocal fairness in voting
behaviour 104106
Redman, Eric 397398, 399
Regenwetter, Michel 51, 89
Reilly, Benjamin 67
repeated alternative-vote procedure 50
representative democracy 418
Republican Party
and cheap land 306
Index 479
and the KansasNebraska Act 307
and the Lincoln election 263,
264265, 269, 285, 286, 290
residue theory in Pareto 427428
Revenue Act (1932) 21, 335, 337344
Riker, William 4, 1213, 15, 17, 22, 158,
441, 442
on agenda control 166, 168171
on agenda setting 383
and the agricultural appropriations vote
(1958) 310311, 329334, 377
on the American Civil War 20, 407
and Arrow
condition of unrestricted domain
101
impossibility theorem 3, 83, 84,
8990
The Art of Political Manipulation 22,
329, 361, 392, 407408
cycling stories in 407
and the axiomatic approach 16
basic argument pattern 16, 17, 3739,
40, 46, 71, 72, 178, 409, 411
and contrived cycles 378379
and democratic irrationalism 3, 20, 29,
30, 31, 409
193, 195, 217240
and disequilibrium 916, 17, 20, 42,
196, 240
and empirical cycles 355
inaccuracy hypothesis 52
liberal interpretation of voting 35
and liberalism 22, 25, 409, 411417
Liberalism against Populism 4, 17, 23,
3136, 91, 193, 196, 279280, 329,
344, 392, 425, 435
Paretos theories in 427
and libertarianism 4
and the Lincoln election 20, 54, 92,
240, 258280, 281282, 376
and the case against populism 202,
409, 410
on logrolling 163
and McKelvey and Schoelds chaos
theorems 17, 173176, 186190,
196
and the Magnuson amendment 361
and majoritarian democracy 23,
4951, 302
manipulation argument 160, 158160,
162, 171, 194196, 344353, 384
theory of dimensional
manipulation 292303
on the market 435
and mistaken cycles 376377
and multidimensional disequilibrium
378
on multidimensional issue spaces 175,
186, 191193, 196
on new issues and dimensions
379384, 386
and Pareto-optimal
equilibrium 426427, 436
on Pliny the Younger 43, 168, 407
and populism 23, 3136, 409411,
418425
and positive political theory 3, 22, 24,
159, 181182, 431, 435
and the Powell amendment 20, 43, 43,
91, 193, 197216
and the Revenue Act (1932) 337, 343
and the Rochester school of rational
choice theory 13, 15, 29
and skepticism 42
on the slavery issue 192, 282, 293303,
304, 307308, 361
on strategic legislators 345350
on transient majorities 392
and the US Federal Convention
(1787) 21, 310, 316, 311316,
321328, 329
on the US Supreme Court and property
rights 302, 303
and voting rules 44, 46, 54, 55
evaluating 56, 64, 68, 7071
376
Rivers, Douglas 197198, 206, 207, 208,
211, 214215
Robbins, Lionel 74, 74, 75
Robinson, Dave 53
Rochester school of rational choice
theory 13, 2326, 28, 29, 30, 37,
175
Romanian Iron Guard 420
Roosevelt, Franklin 36, 197, 336, 420
Roosevelt, James 212
Rose-Ackerman, Susan, on the
International Olympic Committee
vote 375376
Rosenthal, Howard 90, 91, 102,
189190, 191, 193, 304306
Rousseau, Jean-Jacques 32, 3334, 107
Social Contract 33
and the will of the people 418
480 Index
Rowley, Charles K. 12, 15, 433, 440
Runciman, W.G. 11, 441
Rusher, William 36, 423
Ruskin, John 2
Russia
Narodniki movement in 419
see also Soviet Union
Saari, Donald G. 46, 54, 61, 117, 184,
432, 438, 439
Samuelson, Paul 10, 11, 144
Satterthwaite, M. 155, 439
Scandinavian parliaments 21, 335336,
344353
and the Swedish telephone and telegraph
company expansion 347350, 377
and the Swedish voluntary riemans
association 350351
Scarf, Herbert 438
Schoeld, Norman J. see McKelvey and
Schoeld chaos theorems
Schumpeter, A. 28
Schwartz rule (method of transitive
closure) 58, 64, 116117
Scott, Wineld 262
Sears, D.O. 109
secret ballot, and the Finnish Electoral
College 363
self-interest, and voting behaviour
108113
Sen, Amartya 75, 77, 8081, 82, 99107,
112, 144
Senators, election of US see Depew
amendment
Seventeenth Amendment (to the US
constitution) see Depew amendment
Seward, William 267, 308, 309
Shapiro, Ian 27, 28, 29, 178179, 181,
194195
Shaviro, Daniel 15, 442
Shepsle, Kenneth A. 14, 15, 23, 24, 25,
29, 89, 204, 212, 336, 434, 440,
441, 442
Shepsle and Boncheks cycles 336,
361362
Shils on the principles of populism 420
single transferable vote 4951, 66, 69
157
and the Finnish electoral college 366
single-peakedness, and multidimensional
issue spaces 182
Skach, Cindy 423
skepticism 42, 43
slavery issue 21, 282
abolition of 304
dimensions 381, 382383
and Lincolns debate with Douglas at
Freeport 393396
and the preference-development
hypothesis 303309
Riker on the 192, 282, 293303, 304,
307308, 361
rise of antislavery agitation 299
and the US constitution 223
see also Missouri Compromise; Wilmot
Proviso
Smith, Adam 110, 436, 438
Snow, C.P., The Masters 407
social choice theory 1314, 15, 16, 23,
24, 432
and Arrows possibility theorem 70, 72,
76, 78, 85
constructivist 29
dening 27
and democratic means 31
irrationalist interpretations of 16, 52,
443
and majority voting 14, 15, 44, 5556,
83
origins of 7278
Rikers interpretations of 425
and voting rules 61, 7071, 7778, 107
social dilemmas
and individual cooperation 106
and logrolling 165
social homogeneity, and voting rules 49
social rationality, and the concentration of
power 1415
social welfare function (SWF)
domain 8081, 82, 92
123, 129, 143, 149, 146149, 150,
157
social-utility-efciency of voting rules 48,
55
Sonnenschein, Mantel and Drebreu
theorem in economics 438
sophisticated voting see strategic voting
South Africa 1
Soviet Union
fall of communism 1, 13, 440
voting system 417
Spain, democratization of 1
Index 481
spatial voting 10, 116, 179, 181, 191,
196, 379
multidimensional 385386
Spector, Lee 276, 279
stability and democracy 10
domain 95
and deliberation 391
instability in markets and government
433440
instability in Scandinavian parliaments
21
tradeoff between 442
Stepan, Alfred 423
Stephen, Sir Leslie 2
Stephens, Alexander H. 303
strategic voting 5, 17, 35, 71, 72, 158,
159, 160162, 171, 344345
441, 442
and agricultural appropriations
(1958) 330331
and the Condorcet method 9
and cycle claims 337, 377
239
and elimination 66
and the Finnish Electoral College 367
frequency of 193
and the Iowa corn issue 369
and multidimensional
disequilibrium 378
and the Muscle Shoals issue 336, 355,
361
198, 204, 347350, 384
and public choice theory 13
Riker on 37, 194
in Scandinavian parliaments 345353
and social choice 14
and voting methods 9, 65
Stratmann, Thomas 102, 165
Strom, Gerald S. 170, 180, 181, 188, 329
strong positionalist independence (SPI),
condition 150
structure-induced equilibrium 178
subjective utility in Pareto 429
Sunstein, Cass R. 13, 441
Sweden, strategic manipulation in 352
Tabarrok, Alexander 276, 279
Tallmadge, James 297298
Tangian, A.S. 55, 97
Tanguiane, A.S. 52
Taylor, Zachary 261
Teapot Dome scandal 354
Tennessee Valley Authority 197, 336
Tervo, Penna 367
Thatcher, Margaret 6
Tocqueville, Alexis de 112
Tovey, Craig A. 188
transitive closure (Schwartzs method)
58, 64, 116117
Tribe, Laurence 12, 15, 440
Truman, Harry S. 204
Tuck, Richard 11, 441
Tushnet, Mark 12, 15, 440
tyranny
battle against 443
and Madisonian democracy 424
see also dictatorship
United Nations 91
United States
agenda control in the US Congress 170
agricultural appropriations votes in the
House of Representatives 21,
310311, 329334
American populist movement
419420, 421
American progressivism 36
Cambridge City Council
elections 5354
campaign nance problem 30
candidate elections 59
Civil Rights Act (1964) 195
and democracy 1, 2
Federal Convention (1787) 310,
311329
House of Representatives 65
Iowa senators and anticorporate farming
legislation 336, 369370
and the Lecompton constitution
394395
Muscle Shoals in the US Senate
(1925) 21, 336, 353361, 377
presidential elections 84, 8687, 88,
106, 311
see also Lincoln election
482 Index
United States (cont.)
Revenue Act (1932) 21, 335, 337344,
376
and Rikers hypothesis of preferences in
disequilibrium 192
Rikers liberalism and the US
Constitution 421, 423, 428
roll-call votes in US Congress 304305
Seventeenth Amendment to the US
Constitution see Depew
Amendment
slavery 21, 192
strategic voting in primary elections
162
Tax Reform Act (1986) 362
see also American Civil War; Powell
amendment; Wilmot Proviso
universalism 103, 102104
utilitarian voting methods, and Arrows
independence condition 146147,
148, 150
utilitarianism 7273
utility, interpersonal comparisons of
142146
Van Buren, Martin 246, 248, 291
Van Deemen, Adrian 88
van der Veen, Robert J. 111
Vergunst, Noel P. 88
Victorian England 2
Vietnam War 397
Virginia school 20
vote trading see logrolling
voter information, and multidimensional
issue spaces 188
voter preferences
442
domain 95, 108, 99108
138146, 147
choices and underlying preferences 38,
39, 4042
and cyclebusting 114
and democracy 418
indirectly inferring 37
individual preferences 47
defective 388
structured and unstructured 391
in the Lincoln election (1860)
271280, 281
and manipulation 160
177, 182184
preference rankings 20, 21, 44
prole of individual preferences
387391
and Rikers case against populism
409410
unknowable nature of 23, 37, 38, 39,
70, 72, 178, 279
411
and Rikers defense of
liberalism 412, 413
voters paradox, inconsistency of the 11
voting, Riker on liberalism and
populism 22, 32, 3436
voting rules 59, 31
accuracy in 64
124, 137142
axiomatic approach to 16, 6869, 71,
72
and convergence 54
cyclebusting 113122
democratic voting as inaccurate 44
elimination rules 56, 6566
evaluating 55
fair voting methods 44
and rational choice theory 27
409410
and social choice theory 61, 7071,
7778, 107
in the Soviet Union 417
tradeoffs in choice of 441
Waldron, Jeremy 30
Walt, Stephen 28, 344
Washington, George 312
weakened independence conditions (PI),
condition 149150
Weale, Albert 13, 15
Weingast, Barry R. 14, 15, 24, 103104,
294, 302, 307, 440, 441, 442
model of universalism 102
Wentworth, Senator, on the Wilmot
Proviso 256257
what-if cycles 336, 372376
will of the people 418, 419, 420, 421,
430, 442
Index 483
Wilmot Proviso 20, 43, 91, 193, 240,
241257, 258, 294, 295, 296, 302,
407
alternative interpretation 251
257
and cycles 241, 243, 258, 301,
376
Ingersoll substitute 245, 246
dimensions 381
McHenry amendment 245
and the Missouri Compromise 245,
255
and northern Democrats 241, 247,
253, 254, 255
and the Oregon Territory 247, 248,
250, 255
and southern Democrats 248
and strategic voting 257, 345350
Wick amendment 245
Wilson, James Q. 25
Wolff, Robert Paul 12, 15, 440
World War I 1
Wright, J.R. 86
YoungKemeny voting rule 5556,
5859, 6364, 70, 7294
123, 150, 154155
and cyclebusting 117, 118, 121
and the Muscle Shoals vote 356
Zimbabwe 108
Zywicki, Todd J. 238

Democracy Defended

Încărcat de

Informații document

Descriere originală:

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Democracy Defended

Încărcat de

Drepturi de autor:

Formate disponibile

This page intentionally left blank

and the legislator votes

, contradicting yesterdays vote for C over B. At this point

) are the same. Condition I

(S ) be the corresponding social choice functions. If, for all

(S ) are the same. (Arrow 1963/1951, 27)

and Q. The vote was 208 > 203 (actual votes

. Federal aid to schools was

. Therefore, Rikers calculation of

by 215 >209. There were distributional

in 1957, but they only strengthen

in 1957 was the Republican-crafted distributional formula, so

lacked a desegregation amendment, and possessed the

is further conrmed by the

in 1957 (even when including 13

. A yea vote means to kill the bill

). Wainwright kept his promise to support

Nay vote on 62nd Senate Roll-Call # 25 (S > N)

> N, 62nd Senate Roll-Call # 25), which

= winner (it takes >

= winner (it takes >

= winner (it takes >

without any Wilmot Proviso on March 3, 1847,

(Recorded vote March 3,

OA > SQ Inference OA > SQ Revealed vote

> SQ, Voteview, 29th House, Roll-Call #637).

, Voteview, 29th House, Roll-Call

> WP, Voteview, 29th House, Roll-Call #635).

> WP), 97 yeas and 102 nays. This would only be

to enter, and another might inadvertently reward

. Although most of the time most reasonable

. On July 24, the convention voted for A >B

is expensive election of electors as correctly understood on July 24, then

. Second, New Jersey

, then the convention

just as the convention preference merely for elec-

. For our purposes A

> A, but neither of these inferences are essen-

as amended by joint ballot as C; then C > A

. Next, the small states

. The Convention vote was tied and thus

C. Finally on August 24, the delegates decided to postpone further

on July 26. Election by electors in the states

C. Earlier on August 24 national legislative election (with joint

], this cycle would not have been revealed indeed it would

, for a noncyclic or transitive ordering of B

. On September 5, Rutledge (SC) moved

> C. To continue the former analogy, if the delegates were in-

A. There is no direct vote between B

C by four votes to four, thus I

; from that data I think it is plausible to assume

. Alternative B is rst, tied for sec-

, and if we did then alternative A would be tied for

, selection by electors in the abstract

, and selection of executive by joint ballot of the national legislature

. Elster claims that B

or D). My suggested ordering of

. Roll Call #359 showedthat C>D. Next,

and C > D are equivalent to C > (B

versus C), he and the delegates were

, Roll Call #215 and others), and C > D (Roll Call

or D), we presume that he was not wasting his time by introducing

contain the most satisfactory compromise over (1) method