Documente Academic
Documente Profesional
Documente Cultură
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
12.2 State-level aggregation of rst-place winners,
Middle America 274
12.3 State-level aggregation of rst-place winners,
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
4 Democracy Defended
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.
A long, dark shadow over democratic politics 5
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.
6 Democracy Defended
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
A long, dark shadow over democratic politics 7
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.
8 Democracy Defended
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.
A long, dark shadow over democratic politics 9
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,
10 Democracy Defended
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)
A long, dark shadow over democratic politics 11
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
12 Democracy Defended
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
A long, dark shadow over democratic politics 13
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)
14 Democracy Defended
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
A long, dark shadow over democratic politics 15
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
16 Democracy Defended
(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
A long, dark shadow over democratic politics 17
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
18 Democracy Defended
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.
A long, dark shadow over democratic politics 19
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).
20 Democracy Defended
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
A long, dark shadow over democratic politics 21
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
22 Democracy Defended
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
24 Democracy Defended
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
26 Democracy Defended
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.
The doctrine of democratic irrationalism 27
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
28 Democracy Defended
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.
The doctrine of democratic irrationalism 29
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
30 Democracy Defended
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
The doctrine of democratic irrationalism 31
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
32 Democracy Defended
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
The doctrine of democratic irrationalism 33
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)
34 Democracy Defended
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
The doctrine of democratic irrationalism 35
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).
36 Democracy Defended
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
The doctrine of democratic irrationalism 37
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.
38 Democracy Defended
(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
The doctrine of democratic irrationalism 39
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
40 Democracy Defended
(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
The doctrine of democratic irrationalism 41
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
over C
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
42 Democracy Defended
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.
The doctrine of democratic irrationalism 43
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
46 Democracy Defended
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
Is democratic voting inaccurate? 47
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
48 Democracy Defended
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
Is democratic voting inaccurate? 49
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
50 Democracy Defended
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
Is democratic voting inaccurate? 51
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.
52 Democracy Defended
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.
Is democratic voting inaccurate? 53
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
54 Democracy Defended
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,
Is democratic voting inaccurate? 55
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
56 Democracy Defended
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
.
58 Democracy Defended
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
Is democratic voting inaccurate? 59
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
60 Democracy Defended
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
Is democratic voting inaccurate? 61
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
62 Democracy Defended
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
Is democratic voting inaccurate? 63
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
64 Democracy Defended
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
Is democratic voting inaccurate? 65
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
66 Democracy Defended
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
Is democratic voting inaccurate? 67
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
68 Democracy Defended
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
Is democratic voting inaccurate? 69
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
70 Democracy Defended
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
Is democratic voting inaccurate? 71
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
74 Democracy Defended
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
The Arrow general possibility theorem 75
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
76 Democracy Defended
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
The Arrow general possibility theorem 77
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
78 Democracy Defended
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
The Arrow general possibility theorem 79
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
80 Democracy Defended
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
The Arrow general possibility theorem 81
(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
1
, . . . , R
n
be two sets of individual orderings
and let C(S ) and C
i
y,
then C(S ) and C
in 1957:
these 13 are unequivocally P > Q > O
, and Qbeat 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
The Powell amendment 207
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.
208 Democracy Defended
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
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
> 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
244 Democracy Defended
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
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,
The Wilmot Proviso 245
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
246 Democracy Defended
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
The Wilmot Proviso 247
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
248 Democracy Defended
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
The Wilmot Proviso 249
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
250 Democracy Defended
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
The Wilmot Proviso 251
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
).
But on March 3, 1847, the third occasion, the House defeated the Wilmot
amendment (OA
and D
and F
.
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
> B
, 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
on July 24 and A
>B
. 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
C, C > A
,
and A
> 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
More of Rikers cycles debunked 317
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
318 Democracy Defended
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 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
and B
, but B
> C
by eight votes to two although B
> B
. That yields B
> B
C > A
, and
B > A
A > B
> B
C > B > A
A > B
is equivalent to (B
or D), and
thus that #361 shows that C (B
C>A
>B
implies 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
(C > A
, Roll
Call # 356; A
> B
to be equivalent to
(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
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 D> C)
also imply that (C >(B
and C >D
More of Rikers cycles debunked 325
imply that (C > (B
and C is preferred to D, we do
not know anything about whether C is preferred to B
and D together (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
326 Democracy Defended
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
More of Rikers cycles debunked 327
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.
328 Democracy Defended
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,
More of Rikers cycles debunked 329
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
330 Democracy Defended
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
More of Rikers cycles debunked 331
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
332 Democracy Defended
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
More of Rikers cycles debunked 333
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
334 Democracy Defended
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
336 Democracy Defended
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
338 Democracy Defended
Table 15.1. Blydenburghs analysis, Revenue Act 1932
Group Number in Vote
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
Other cycles debunked 339
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
340 Democracy Defended
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
Other cycles debunked 341
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
342 Democracy Defended
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
Other cycles debunked 343
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.
344 Democracy Defended
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
Other cycles debunked 345
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.
346 Democracy Defended
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
Other cycles debunked 347
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
348 Democracy Defended
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
Other cycles debunked 349
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
350 Democracy Defended
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,
Other cycles debunked 351
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
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
Other cycles debunked 359
Table 15.5. (cont.)
Votes
Senators 1,2,3 4,5 6 Comments
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
360 Democracy Defended
Table 15.5. (cont.)
Votes
Senators 1,2,3 4,5 6 Comments
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,
= 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,
370 Democracy Defended
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).
Other cycles debunked 371
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
372 Democracy Defended
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
Other cycles debunked 373
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
374 Democracy Defended
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
Other cycles debunked 375
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
376 Democracy Defended
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
Other cycles debunked 377
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
380 Democracy Defended
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
382 Democracy Defended
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
384 Democracy Defended
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.
386 Democracy Defended
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
388 Democracy Defended
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
390 Democracy Defended
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
392 Democracy Defended
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
394 Democracy Defended
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
396 Democracy Defended
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
398 Democracy Defended
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
400 Democracy Defended
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
402 Democracy Defended
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
404 Democracy Defended
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,
406 Democracy Defended
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
408 Democracy Defended
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
410 Democracy Defended
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
412 Democracy Defended
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?
Plebiscitarianism against democracy 413
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
414 Democracy Defended
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.
Plebiscitarianism against democracy 415
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
416 Democracy Defended
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
Plebiscitarianism against democracy 417
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.
418 Democracy Defended
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
Plebiscitarianism against democracy 419
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
420 Democracy Defended
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
Plebiscitarianism against democracy 421
(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
422 Democracy Defended
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)
Plebiscitarianism against democracy 423
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
424 Democracy Defended
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
Plebiscitarianism against democracy 425
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.
426 Democracy Defended
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
Plebiscitarianism against democracy 427
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
428 Democracy Defended
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.
Plebiscitarianism against democracy 429
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
430 Democracy Defended
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
Plebiscitarianism against democracy 431
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
434 Democracy Defended
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
Democracy resplendent 435
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
436 Democracy Defended
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.
Democracy resplendent 437
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
438 Democracy Defended
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.
Democracy resplendent 439
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
440 Democracy Defended
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.
Democracy resplendent 441
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
442 Democracy Defended
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
Democracy resplendent 443
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.
Cambridge University Press.
Groseclose, Tim, and James M. Snyder, Jr. 1996. Buying Supermajorities.
American Political Science Review 90: 303315.
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.
American Political Science Review 70: 12021214.
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.
Cambridge University Press.
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.
American Political Science Review 87: 102115.
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.
American Political Science Review 78: 116.
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.
Cambridge University Press.
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
and academic attacks on democracy
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
and multidimensional issue spaces 192,
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 the APA presidential elections 49
and Arrows independence
condition 123, 124, 125126,
127, 128, 138, 141
consequences of 147, 150
and independence 151153
justication of 144146
and manipulation 154155, 156
and cardinality 78
and Condorcet efciency 48
and the Copeland method 50, 51
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
and the Lincoln election 259, 273,
277, 279, 281
and multidimensional issue spaces 184,
191
and the Muscle Shoals vote 356, 357
and numbers of parties 66
and plurality grading 6061
as positional voting method 46
and the Powell amendment 200
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
and strategic voting 65
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 the APA presidential elections 49
and Arrows condition of unrestricted
domain 99100, 108
and Arrows independence condition
124, 128, 138139
consequences of 147
and independence 151153
justication of 144145, 146
and manipulation 154155, 156
and Arrows theorem 8, 7879, 80, 82,
84, 85, 89
Condorcet efciency 4748
and social homogeneity 49
Condorcet-consistent rules 56
and the Copeland method 50
and cyclebusting 114
and the Danish Parliament election
371
and the Depew amendment 219
and dictatorship 434, 435
evaluation 55, 58, 59
and the Finnish electoral college 364,
365
Jury Theorem 63
and logrolling 163
and multidimensional issue spaces 176,
184, 186, 187
pairwise ordering 49
rankings 52, 53
and the Revenue Act (1932) 337
and self-interest 113
switch to other methods 442
see also pairwise comparisons
Condorcet winners 4445, 46, 49, 50,
51, 7071
and agenda control 169
and Arrows condition of unrestricted
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
independence condition 125
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
and academic attacks on democracy
441
and agenda control 166, 167168
and Arrows condition of unrestricted
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
and individual preferences 391
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
and multidimensional issue spaces 178,
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 self-interest 113
and simulations of homogeneity 9699
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 dictatorship 440, 442
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
and Rikers basic argument pattern 16,
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
and Arrows independence condition
141
and democracy 440, 442
and instability 434
and majoritarian processes 10
direct democracy 418
disequilibrium 21
and agenda control 169
and democracy 440, 441
and equilibrium in economics 435
436
and the introduction of new
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
and the Lincoln election 260, 280,
410
and new dimensions 378379
and Ordeshook 435
and Pareto 428
and the Revenue Act (1932) 337
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 election 16, 288289
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
and the Missouri Compromise 264
Dryzek, John 13, 15
Dutch parliamentary elections 88
Duvergerian equilibrium 66
Duvergers Law 289292
and the election 16, 289290
and the election 16, 290292
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
and multidimensional issue spaces 175,
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
and the APA presidential elections 49
evaluation 59, 66
preference voting 53, 54
harmful manipulation 159, 160, 378
and agenda control 168
and the Depew amendment 239240
frequency of 193196
and logrolling 163
and the Powell amendment 197, 198,
201
in Scandinavian parliaments 353
and strategic voting 162
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
and multidimensional issue spaces 182,
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
independence condition 143
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
and academic attacks on democracy
441
and the Depew amendment 217
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
and Arrows independence condition
154156, 157
and Arrows possibility theorem 70,
7294
economic 436, 438440
and the introduction of new
dimensions 381
and the Magnuson amendment 406
in politics 436
and Rikers basic argument pattern 37,
38, 258, 280
and Rikers case against populism 410
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
and the Missouri Compromise 261
and the Wilmot Proviso 241, 242, 248,
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
and the Wilmot Proviso 245, 255
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
and Rikers basic argument pattern 37,
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
and the Lincoln election 278, 281
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
and Rikers case against populism 410
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 agenda control 167168
and agricultural appropriations (1958)
333
and Arrows independence condition
137139, 141
and the Depew amendment 236
and dictatorship 435
and the Lincoln election 259, 277,
278, 279, 281
and logrolling 163
and multidimensional issue spaces 192
and the Muscle Shoals vote 356357
and the Revenue Act (1932) 342
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 Arrows possibility theorem 8486
and Condorcet efciency 48, 49, 60
and elimination 65
evaluation 59, 5259, 60, 70, 72
and the Finnish electoral college 363,
367
and the impartial-culture assumption
47
and the Lincoln election 278, 279, 281
and social homogeneity 49
and strategic voting 160
plurality voting 59, 44, 49, 51
accuracy 64
and the APA presidential elections 49
and Arrows independence condition
142, 157
and Arrows possibility theorem 84
and the Borda count 6
and the Copeland method 50, 51
and the Lincoln election 259, 277,
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
and social homogeneity 49
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
and the Lincoln election 276
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
and Arrows independence condition
147148, 150
proportional representation 66
and the Finnish electoral college 363,
386
Przeworski, Adam 13, 441
public choice theory 13, 27, 29, 30
and markets 433, 434, 437
and self-interest 111112
and the Wilmot Proviso 246
public good
and democracy 36, 419, 424425, 430
and the general will 34
and Rikers liberalism 35
public interest
and Rikers defense of liberalism 416
and Shepsle and Boncheks cycles 362
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
and the Depew amendment 220221
and the Powell amendment 205
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
independence condition 146148
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
and the Depew amendment 20, 91,
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 Shepsle and Boncheks cycles 361
and skepticism 42
on the slavery issue 192, 282, 293303,
304, 307308, 361
and social choice theory 1314
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
and the Wilmot Proviso 43, 240, 302,
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
and the Revenue Act (1932) 343344
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
and social choice theory 432
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
and Arrows independence condition
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
and the introduction of new
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
and individual preferences 388
irrationalist interpretations of 16, 52,
443
and majority voting 14, 15, 44, 5556,
83
origins of 7278
Rikers interpretations of 425
and Shepsle and Boncheks cycles 362
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)
and Arrows condition of unrestricted
domain 8081, 82, 92
and Arrows independence condition
123, 129, 143, 149, 146149, 150,
157
and dictatorship 434435, 442
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
and Arrows condition of unrestricted
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
and academic attacks on democracy
441, 442
and agenda control 166, 168
and agricultural appropriations
(1958) 330331
and Arrows possibility theorem 91
and the Condorcet method 9
and cycle claims 337, 377
and the Depew amendment 233, 236,
239
and elimination 66
and the Finnish Electoral College 367
frequency of 193
and the Iowa corn issue 369
and the Lincoln election 282, 289
and majoritarian processes 10
and multidimensional
disequilibrium 378
and multidimensional issue spaces 196
and the Muscle Shoals issue 336, 355,
361
and new dimensions 384
and the Powell amendment 20, 197,
198, 204, 347350, 384
and public choice theory 13
Riker on 37, 194
and Rikers case against populism 410
and Rikers defense of liberalism 412
in Scandinavian parliaments 345353
and social choice 14
and voting methods 9, 65
and the Wilmot Proviso 257, 345350
Stratmann, Thomas 102, 165
Strom, Gerald S. 170, 180, 181, 188, 329
strong positionalist independence (SPI),
and Arrows independence
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
and Rikers defense of liberalism 417
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
and academic attacks on democracy
442
and agenda control 167, 169
and Arrows condition of unrestricted
domain 95, 108, 99108
and Arrows independence condition
138146, 147
and Arrows possibility theorem 7982
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
and multidimensional issue spaces 173,
177, 182184
preference rankings 20, 21, 44
prole of individual preferences
387391
and Rikers case against populism
409410
and self-interest 108113
and simulations of homogeneity 9699
unknowable nature of 23, 37, 38, 39,
70, 72, 178, 279
and Rikers case against populism
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
and Arrows independence condition
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 the Lincoln election 259
and rational choice theory 27
and Rikers case against populism
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),
and Arrows independence
condition 149150
Weale, Albert 13, 15
Weingast, Barry R. 14, 15, 24, 103104,
294, 302, 307, 440, 441, 442
model of universalism 102
and the Revenue Act (1932) 337
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
and the introduction of new
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
and agenda control 168
and Arrows independence condition
123, 150, 154155
and cyclebusting 117, 118, 121
and the Muscle Shoals vote 356
and the Powell amendment 200
Zimbabwe 108
Zywicki, Todd J. 238