+/ Gr:egor: Ekit'ft and StepltenE.

Hans,n

Sztompka, Piotr. r99u. Dilemmas of the Great Transformation. Sisyphus2,:9-27.

Sztics,Jeno. 1983. Three Historical Regions of Europe. In John Keane, ed., Ciuil
z
Societyarul the Sate: NeLUEuropean Perspectiues, pp. L9r-f2. London: Verso.
Thelen,I(arirleen.r999. Historrcai Institutionalismin Comparative Politics.Annwal
Reuiew of Political Sciencez: 369-qo4.
Accounting for Postcommunist Regime Diversity
ooo. Time and Temporalirv in the Analysis of Institutional Evolution and Change.
:2. .What
Studiesin American Poliicel Deuelopment14(r): ror-8. Countsas a Good Cause?
Tilly. Charles. r99o. Coerciort, Capital, and European States, A.D. 99o-r99o.
Cambridge,Mass.: Basil Blackwell.
HerbertKitschelt
Tismaneanu, Vladimir. 1998. Fantasiesof S"iluation:Democracy,Nationalism, and Myth
in Post-CommunistEurope. Princeton: Princeton University Press.
Tucker, Robert C. t97r. The Souiet Political Mind: Stalinism and Post-Stalin Change.
New York: !f. rW.Norton.
!'enables, Anthony, and Nuno Lrmao. 1999. Geographicai Disadvantage: A
Heckscher-Ohlin-von Thunen Model of International Specification. I7orld
Bank, Policy ResearchWorking Papers, no. 2256. A single hegemonic power ruled the East European and Central Asian re-
Verdery, Katherine. ry96. What Was Socialism, and What Comes Next? Princeton: gions until the late r98os and stamped its ideologicai doctrine and basic
Princeton Universiry Press. institutions on every single polity, with the partial exception of Yugoslavia
Walder, Andrew. :1986.Comn'tunistNeo-Traditionalism: Work and Autbority tn Chinese and Albania. Its demise,however, has resulted in a highly diverse set of more
Industry. Berkeley: University of California Press.
than rwenry-five sovereign polities with features that range from those o{
Wallerstein, Immanuel. r974. The Modem \X/orld-System:Capitalist Agriculture dnd
full-fledged competitive democracieswith well-protecred civic and polirical
tbe Origins of the Eurcpean Vlorld-Economy in the Sixteenth Century. New York:
Academic Press.
rights all the way to authoritarian, personalist. if not desporic, rule. Ivlea-
t99ra. Untbinking Social Science: The Limits of Nineteenth-Centtrry Paradigms. sured in terms of the civic and politicai rights indexes deveioped by Freedom
Cambridge, Ivlass.:Poliry Pressin association with Basil Blackwell. House, there is no region or set of countries on earth with a currently larger
ry9rb. Geopolitics and Geocubure: Essays on the Changing World-System. diversity of political regimes.' The fact that before ry89 all of today's East
Cambridge: Cambridge University Press;Paris: Editions de la Maison des Sci- Central European and Central Asian poiities had communist single-party
encesde I'Homme. rule and socialist economic planning systemscannot possibly account for
'!?eber,
N,'Iax.t968. Economlt and Societ,"'t.3vols, Berkele,rz:
University of California the tremendous diversity of political regimes that emerged in this region in
Press. the early 199os. This postcommunist diversiry came about in the short win-
\Whitehead, Laurence. 1999. Geography and Democratic Destiny.
Journal of Democ- dow of about three years ft99o-93). Since that time, new regime structures
racy ro\rl:74-79.
have been more or less "iocked in" in almost all polities. Countries that
Woodward, Susan. 1993. Balkan Tragedry:Chaos and Disso/tfiion after the Cold War.
by ,gg+ were more democratic have stayed that way. Countries that were
Whshington, D.C.: Brookings Institution.
authoritarian have not reversedcourse and become democraric. There may
even be a tendency toward a polarization of regime typessuch that polities with
initially intermediate levels of civic and/or political rights eventually became
entirely democratic or fully authoritarian.' In a similar vein, postcommunist

' The standard deviation of countries' civic and political rights

scores for the set of post-
communist polities serves as the mathematical indicator of regional regime diversiry. The
comparatively high standard deviation of civic and political rights is so high because,in con-
trast to other regions, the postcommunist area polities display no central tendency. In Latin
America and East Asia, bv contrast, that central tendency has gravitated roward democracy
or mixed regimes since the r98os. In the Nliddle East and A{rica, regimes are overwhelmingly
authoritarian.
I Several countries in the intermediate range of the Freedom House indexes (l.j-j.o) in
the mrd-r99os appear to be moving toward more-democratic civic and political rights

t ,to
 Herbert Kitschelt
countriesthar were leadersin economic market reform rn r99z-g) are still
in that position by the en,Jof the millennium. if we accept European Bank
of Reconstructionand Development (EBRD) inderes of economic reform
eftort as or:r crirical measur€.
The impressivediversity of political regimes and economic reform efforts
evidencedb'r the postcommunistpolities posesa provocative puzzle for the
socialsciences.Why is there no uniform and persistent"communist legacy"
detectableamong the numerouscommunist successorregimes?\X/hatled to
the appearanceof great variance in the modes of dismantling communrst
rule as well as the resulting new polirical power relations and institutional
codificationsalmost overnight in the early r99os? The great physical,eco-
nomic, social,and cultural diversitltof postcommunistcountriesallows tts to
eliminatesomeexplanationswith great ease.Conversely,someessentialsim-
iiaritiessharedby all polities also rule out important explanationsfor lasting
regime diversity. For exampie, the best confirmed generalization about de-
mocratization is that rich countriesare more likely to becomeand then stay
democratic rhan poor countries (Bollen, ry79; Barkhart and Lewis-Beck,
r994; Geddes>1999).But differentialsin economicwealth and development
cannot account for the observedpattern in postcommunist Europe and Cen-
tral Asia. On that account, Belarusshould display more civic and political
democratic rights and economic reform effort than Lithuania, Armenia more
than Poland, and Russia more than Hungary. Broadly speaking, all formerly
communist polities are "middle income" economies whose political regime
patterns are underdetermined by socioeconomic ievels of wealth and devel-
opment. The only generalizationfrom theoriesof economic modernization
that applies to all of them is that members of this group are likely to display
diverse and volarile regime properties (Huntington, r99ri Przeworski and
Limongi, 1997).
Although similar levelsof economic developmentand sharedexperiences
with a "Leninist legacy" cannot serve as explanations of postcommunist
regime diversity, scholars have proposed a bountv of competing hypothe-
ses to account for that outcome. Students of postcommunist regimes have
nominated religion, geographical location, precommunist regimes and state
formation, post-Stalinist reforms in communist regimes,modes of transition
to postcommunism, and winners or losers in the "founding elections" of
the new regimes as explanations of current regime diversiry to name only
some of the most important arguments.Most of thesevariablestend to be
plausiblecandidatesfor explanationbecausethey display striking patternsof
(Croatia, Macedonia, and Moldova). Orhers initially in a semiauthoritarian halfway house
have lately gravitated toward authoritarianism (Belarus and Kvrgyzstan, as well as possibly
Albania or Armenia). Some countries have staved in this intermediate category (Ceorgia,
Russia, ukraine). But no countrv has joined the intermediare category srnce r995
rhrough a degradation of pree-ristingdemocraric accomplishments or an lmprovement ove[
authoritarianism.
Postcommunist Regim e D iuersity
covariance with the dependent variables, political regime properties, or eco-
nomic reform effort. Nevertheless,scholarsdisagreeon which of rhem qual-
ify as reasonablecausalexplanations.Becausethe rival explanatory variables
are highlv collinear borh in cross-narionalas well as inrerremporalperspec-
tive after the initial outburst of diversification in :'989-9-. , it is impossible
to settle on a single superior causal accolint based on inductive statistical
computation alone, regardiessof which multivariate statisricalmodel one
might acceptas mosr adequatefor the explanatory task. In order to discrim-
inate among explanations,one has to draw on more-subtlemethodological,
metatheoretical,and ontological considerationsabout the conceptualization
of causalirywhen faced with the theoreticalproblem of expleining posrcom-
munist regime diversitlr.
The nature of this chapter is rherefore primarilv epistemologicai. I do
not develop any single substantiveexpianation of postcommunist regime
diversity in detail, but I draw upon rhem in order ro illustrate different
conceptionsof causality.Empirical evidencedoes constrain whar we can
reasonablyconsider or dismiss as a good cause of postcommunist regime
diversiry. But this empirical evidence, by itself, leavessufficient ambiguities
to necessitatea reflection on turther criteria and qualifications in order to
discriminateamong more or lesssatisfactorycausalexplanations.
What We Can(not) Ask Causal Analvsis to Achieve in the Social Sciences
Let me begin by rebutting a common misunderstanding of what causai anal-
ysis should accomplishin the social sciences.We will make headway in rhe
causalanaiysisof social processesonly if we abandon misplacedaspirations
and erpectations. To pur rt in one proposition: sciencesof complexiry in
general, and the social sciencesin particulaq cannot explain singular events
and, conversely,therefore cannot advance point predicrions of what is likely
ro happen in a particular insrance.
In the mid-r99os Japan was the last technologically advanced country
to close down its earthquake prediction program. All other countries with
significant earthquake risks, including the united States.had done so earlier.
Policy makers and scientistsconcluded that earthquake prediction is unable
to reduce the margin of error sufficiently to sound operational warnings, for
example, to evacuatecertain areas.Does the fact that earthquake prediction
programs have globallv failed lead us to conclude that geology in generai
and geotectonicsmore specifically have failed? Scientistswouid respond to
that questionwith a resoundingno. Our understandingof continentalplate
tectonics has substantially improved over time, and scientists can rdentify
the areas that are vulnerable to major earthquakes quite precisely. I7hat
cannot be achieved, however, is to predict the specific occurrence of earth-
quakes.Scientificknowledge about earthquakesis probabiiistic becauseit is
practically impossible to assemblea fuil account of all forces that impinge
.1 Herbert Kitscheh
ol1a time-spaceevent.Such singuiar eventsare the result of multiple causal
chains that interact. Even if all causal mechanismslvere known, an extraor-
dinary amount of data would have to be gatheredfrom countlesssites,some
of them deep underground, to be fed into an encompassingtheoretical model
of plate tectonics.As a consequence,it is not feasibleto predict a particular
earthquake event with sufficient certainty to allow policy makers to take
expensivedecisions based on such information - for example, the evac-
uation of a city. The weather is a case in point. In most regions of the
world, meteorologistshave so far failed to predict seasonalweather fea-
tures with sufficient precision to enable growers to adjust the crops they
plant.
The caseof earthquakeprediction can be generalizedto ali sciencesthat
deal with moderateor high complexiry.It is all but impossibleto predict sin-
gular events with a sufficient measure of precision and certainry to improve
the ability of policy makers and citizensto act in a more rational, strate-
gic, future-oriented fashion in addressingparticular situations. For ontolog-
ical reasons,the prediction of eventsis even more difficult in the social sci-
encesthan in the study of natural phenomena with high complexity. lvlerton
OSSZ) drew our attention here to the problem of reflexiuity, precipitating
self-fulfilling or self-destroying prophecies. Ivlany of the people who make
public event predictions are not iust disinterestedsocial scientists(even if
we grant that there are such people), but actors in whose interest it is to
bring about or avert the predicted events. They will attempt to change the
boundary conditions assumed in the event prediction. Thus, sociai predic-
tions often generate their own social dynamic of hvpothesis falsification or
verification. In light of recent experienceswith stock markets around the
world, any reader can spell out the logic of stich individual and coliective
conduct for stock market bubbles.
In addition to complexity and reflexivitn also uncertainty from the per-
spective of the actors sets limits on the causal account of individual sociai
events.With the benefit of hindsight and with time and resourcesto collect
data, scholars may have a much better grasp of the objective constraincs
and the diverse actors' perceptions and calculations at the time of political
decision making than any of the actors had themselvesex ante. Particularly
in times of regime crisis, political actors often may not haye the time and the
accessco gather the information that would allow them to choose their best
course of action. Facedwith ambiguities about the past and the present, they
have to interpretthe prospective yield of alternative coursesof action. These
interpretationsmay havea systematicand a random component.The system-
atic component is amenabie to causal analysis within a cognitive "political
culture" framework. I7hat are the theoreticalconceptsand idioms that allow
actors to interpret their choice situation? \7hat were the causal propositions
within their zone of attention that enabledthem to construct payoff matri-
cesresulting from alternative strategiesof action chosen by themselves,their
f mrd-r99os
PostcornntunistRegntzeDiuersity ,_3
(potential) allies,and their adversaries?In social scienceaccountsof regime
change,the cultural semanticsof breakdown becomesone causalchain that,
on average,may show a staristically significant probabilistic relationship ro
the occurrence of regime breakdowns.
The other component of uncertainry and interpretation is random. How
individual and collectiveacrors define a particular historical situation may
depend a great deai on contingent social nerworks (who gers to know
what) and idiosyncratic psychologicalprocesses(the personaliry of actors
in high-impact positions). Whether "persuasive," "charismatic," and thus
"entrepreneurial" and "innovative" leadersare situatedat critical nodes of
sociai networks .in rimes of acceleratingregime decay and thus may organize
collective action that brings about regime collapse can never be predicted
by a systematiccausaily oriented social science.Similarly, sociai sciencecan
never determine the conditions under which entrepreneurial leaders have
a "correct" or "foresighted" inrerpretation of the situation in which they
choosea course of action. such leadersand their enrourage,just like other
citizens,are subject to cognitive errors. Revolutionary situations may thus
slip by simpiy becauseleadershipfailed at crirical moments, either becauseit
was insufficiently bold and persuasive and/or becauseit cognitively miscon-
strued the situation. Conversely,political leadershipmay define situations as
revolutionary and attempt insurrections when the objective circumstances
are not conducive. Scholarshere face a random component of collecdve ac-
tion with the intent of regime change that will always make it impossible to
predict singular events with certainry.
The challengeto theorizepolitical regimechangethus exemplifiesthe gen-
eral limits of explanation and prediction in the social sciences.Complexity,
reflexiui4t, and actor uncertaintyabout the parameters of the situation make
it impossible for social scientiststo predict singular events.The postdiction
of past events("explanation") always leavesresidualunexplainedvariance.
.We
may know that, on average, a higher level of affluence, the onset of a
sharp economic recession,and discord within rhe ruling authoritarian elites,
particularly if they are military juntas or single-party regimes, make a tran-
sition of authoritarian rule ro durable democracy more likely. This general
knowledge, however, will help us little to predict rhe occurrence and precise
timing of democratizationin individual cases,suchas Indonesiaor Malavsia,
Singaporeor South Korea. Nevertheless,as scholarswe work on making our
generalknowledgeabout cause-effect relationsmore preciseand calibratethe
average contributing effect of each variable to rhe phenomenon of regime
breakdown. This will, however, still never permit us to predict individual
eventswith certainty.
In the retrospecriveaccount of regime change,we may distinguish befween
conditions that made regimes more conducive to change in the long run, as
well as conditions rhat acted upon regimesover short periods of time and
precipitated their ultimate collapse. Nevertheless,this does not allow us to
'o'l
"pp.r.
range of
rournsrowlrd the
,o-;;^;'^_"1ate
.;;:
Freedonr H^,,.^
;fi:::*1"^.,::
, ,
,"r"."* (,;.r_j.or
 Herhert Kitschelt Postcomrnttnist Regime D iuersih, 55

draw a simple divide between "necessary" and "sufficient" causesof regime From a purely social scientific point of view, interested in the causal anal-
change. Each cause may have functional equivalents. No single long-term yus of human social action, institutions, and processes,the prediction of
or short-term condition may be necessaryor sufficient in its own right, singular historical events is thus neither interesting nor feasible. The focus
but oniy the concatenation and configuration of forces that yield particular of attention is on the general mechanisms and causai linkages that make
historical constellations(cf. Gerring, t999i z6). These considerationsonly the occurrence of certain events more likely. For this reason, the knowledge
r e p € a t a t h e s i sa d v a n c e db y N l a - x W e b e r ( r 9 8 8 : e s p . 2 8 6 - 8 9 ) a l m o s t o n e of financial economistsdoes not necessarilyrender them to be better stock
hundred years ago. The social sciencesdeal with general correlations and market speculators,businesseconomistswili not necessarilyhelp a particu-
causationsthat permit only probabilistic explanationsand predictions.It is lar company to beat its profit targets, and political scientistscan never claim
up to historiansand historical-comparativesociologyto show how multiple to tell politicians with certainry how to improve the margin of their elec-
causalchains interact in unique conte-\tsand produce often irreproducible toral support. The hiatus between analytical, cause-seekingsocial science
results. and public policy remains large. Sudents of political regime changehave to
If eventsresult from multiple causal chains, the construction of "analytic appreciatethat as much as practitioners,whether they are averagecirizens
narratives" (Bates et aI., ry98) may be instructive, but of limited use for or political professionals,including professionalrevolutionary cadres.
causal analysis.Analytic narratives "are motivated by a desire to account for A further exampie may clarify the limits and possibilities of causal analysis
particular eventsand outcomes" (r).They reconstructthe logic of reasoning in the social sciences.According to most accounts, both the Czech Repub-
and interaction that is presumedto motivate actors in a particular case.As lic and Poland are examplesof postcommunistpolities that have advanced
long as we examine individual cases,however,it is difficult to draw causal market-liberal economic reform most, when compared ro the entire cohort
inferencesabour any logic of action attributed to social actors in a theoreti- of polities that resulted from the breakdown of communism (cf. Heliman,
cally conceptualized structure of constraints, even though we can ascertain 1996; 1998; EBRD, 1998; Fish, ry98b; Kitschelt and Maleskn zooo). Yet
which theoretical accounts fit the facts of the case (tS-tz). In that sense, the Czech Republic experiencedan economic recessionand then a weak re-
the study of particular casesonly provides "data points," each of which is covery in the final years of the rwentieth century, whereas Poland's economy
open to multiple causal inferences.Taken together, in a comparative analy- sustainedvery high rates of growth in that period. Theories that predict why
sis of multiple cases,they ma,vnarrow down the range of plausiblecauses both the Czech Republic and Poland are better off in terms of reform effort
of an outcome and identify distinct causal chains that contribute to that and reform outcome than most postcommunist economiescannot account
outcome. for the particular details of differential performance since r996.
Becausemultiple causal chains interact in many contexts, it is misleading Other, more short-term causal chains may influence these particuiarities,
to biame the social sciencesfor not predicting the "event" of communism's including the role of leadership and of subjective interpretations of eco-
sudden demise rn r989-9t in most countries governed by that system of nomic uncertainty by influential politicians, something that social science
governance.,'Beginning w-irh rhe debate between Ma-x Weber and his an- is less equipped to explain than psychology.For example, we may argue
archist students in r9r9, and documented in Weber's ruminations about that for electoral reasonsthe governing politicians in the Czech Republic
socialist economics tn Economy and Society,soon followed by von Mises's engagedin a hasry privatization program in r99z that overlookedproblems
and the Austrian school's work on planned economies,a great deal has been of moral hazard in the governance structures of banks, investment funds,
known for a long time about the principal weaknessesof economic insti- and corporations outside the financial sector. \7e may also surmise that id-
tutions with a hierarchical allocation of scarceresources.Of course,these iosyncrasiesof the political leadership,such as Vaclav Klaus's exposureto
basic economic liabilities of socialism, taken by themselves,were neither Chicago-style economics and his admiration of Margaret Thatcher, trans-
necessarynor sufficientconditions for the breakdown of communism. Any lated into a fanaticism of market-liberal reforms that ignored the importance
satisfactoryhistorical account of the communist breakdown must incorpo- of market-complementing political-regulatory institutions that avert market
rate the many facets of decay and cumulative comparative inferiority on failures generated by unrestrained contracting and lower transaction costs
multiple economic,political, social,and cultural dimensionsof performance in the privatization of economic assets.From the vantage point of social
that characterizedtheseregimesas they enteredthe r98os. sciencetheory, the leaders' psychological dispositions or the details of the
electoral calendar are entifely coincidental external "shocks" intertwined
with a particular historical situation, which will always guarantee a measure
; F o l l o w i n g a c h o r u s o f s c h o l a r ss i n c e r 9 8 9 , K a l y v a s ( r 9 9 9 ) s u b s c r i b e st o t h i s c h a r g e a n d
of randomnessin explanationsand predictionsof postcommunisteconomic
provides an impressive review of the voices that have tried to explain communism's fall after .$flhat
reform effort and success. is consolinq for a social science account
the fact.

.--,D !uvyqru 'rere_democraric

civic and p"liii""t.dgh;
 tr'/
16 Herbert Kitscheb
of postcommunisreconomic reform. however,is that the particularitiesof
Czech politics did not let the counrrv's per{ormance drop to levels encoun-
tered among countries to which scholarsattest much lesspromising social
and political capabilities for economic reform.
Having no,vvapproximatell'clarified what causalanalysiscannot do in the
study of posrcommunist political regime change and economic reform, let
us inspectmore ciosely what social sciencemay be able to contribute, or at
'!7hen
least what social scientistsaspire for it to accomplish. social scientists
talk about causes,they usuallyhave in mind whar Aristotle referredto as rhe
"efficient" cause(s)of some ourcome under study - that is, the forces that
brought it about, produced it, or crearedit (Gerring, zoor: chap. 7). It is no
simple question, however,to determine what exactly is meant by creation and
what qualifies a good causal analysis.Before we turn to concrete e-xamplesof
rivai causalaccountsof postcommunistpolitical and economic trajectories,
we need to generate some epistemological criteria to evaluate the status of
causaiclaims.
What Counts as a Good Cause?
Assessingthe quality of inf'erencesin general and of causal inferencesin par-
ticular involves multiple dimensions and calls for judgments that go beyond
the arbitration among claims purely grounded in formal logic.a Ultimately,
what "counts" as an acceptablecausein a community of scholarsinvolves
trade-oftt among a variefy of criteria. Debates among scholars of postcom-
munism are thus often as much about the qualiry of causal attribution as
the correct specification of empirical rests or the narure and validiry of the
data brought to bear on the theoretical question. Let me first sketch an un-
satisfactory conceptualization of causal analysis, then move on to crireria
of causal analvsis in the social sciences,and finally address the problem of
trade-offs between such criteria.
The "Coueing Law" Schema as Explication of Cawsal Analysis
The debate about Hempel-Oppenheim'sstandard "covering law" explana-
tion iilustrates the ontological rather than the logical nature of the episte-
mological debate about causal analysis.For these authors, a complere ex-
planation involves a time-spaceinvariant proposition about the relationship
between two variables and a staremenr of empirical conditions rhat assert
-We
the presenceof one of the rwo terms. can then draw the inference that
the facts corresponding to the other rerm must be present too. Explana-
tion, then, is the inference of an observable fact from knowiedge of a gen-
eral relationship among antecedentand consequentvariablestogerherwith
+ In this section, I heavily draw on Gerring's (zoor: r:8-5r) extremely useful discussion of
formal criteria and empirical confrmation of causal inferences.
Postco mmunist Regime D iuersit.t i7
empiricalknowledge about the antecedentconditions.The problem is, how-
ever, that this standard "covering law" explanation cannot separate law-
ful (causal)and accidental("correlational") generalizations(Salmon,r989:
r 5 ) . 5T h e s a m ep r o b l e m i s p a t e n ti n S t i n c h c o m b e '(sr 9 8 7 : z 9 ) c o n c e p t i o no f
a causal law as a "statement that certain values of rwo or more variables are
connected in a certain wa5" even though the author later reminds us that
correlation is not the sameas causation(lr).A simiiarly deficientstatistical-
epistemologicaldefinition of causation appearsin King, Keohane, and Verba
(r9g+r 8r-8u ): "The causal effect is the differenceberweenthe systematic
component of observations made when the explanatory variable takes one
value and the systematiccomponent of comparableobservationswhen the
explanatory variable takes on another value." The ontological intuition be-
hind causation is that some x helps to create some y in the senseof " bringing
it into existence."Substantivescientifictheories,not reconstructiveepiste-
mologn must elaborate conceptual primitives and entities that constrtute an
acceptable ontology of causation in a particular domain. To demonstrate
that questions of causation involve ontological and pragmatic considera-
tions unique to each science is the purpose of Richard Miller's Fact and
Method:
[A]n explanationis an adequatedescriptionof the underlyingcausesbringingabout
a phenomenon. Adequacy, here,is determinedby rulesthatarespecilicto particalarfields
at particulartimes.The specificiryof the rules is not just a featureof adequacyfor
given specialpurposes,but characterizes all adequacythat is relevantto scientific
explanation.The rulesarejudgedby their etfrcacyfor the respeaiuelieldsat the respec-
tiuetimes- which makesadequacyfar more contingent,pragmaticand field-specific
than positivistsallowed,but no lessrationallydeterminable. (ry87: 6, emphasis
added)
Onnlogical Criteria of Causal Analysis
In order to specify the meaning of what Miller calls "adequacy" of expla-
nation for a particular scholarly field, such as the study of postcommunrst
regimechange,let us draw on Gerring's(rg99, Lo-29, Loori t1o-46) three
criteria pertaining to causal inferenceand four criteria involved in the confir-
mation of causal relationships.They demonstrate how difficult it is for social
I This critique is different from the following advanced by Somers (1998:777), whrch is inap-
propriate: "Because it [the Hempel-Oppenheimer scheme] subsumes explanation under the
rubric of predictive universal laws, the covering law model cannot disclose the underlying
causal mechanisms underwriting chains of eventsl it cannot allow for the contingency of
outcomes or explain how temporal sequences,conjunctures, and spatial patferns matter to
theory construction." There is nothing in the H-O schemethat would not allow us to explain
events by the conjuncture of multiple general laws and their corresponding initial conditions.
Goldstone (r998: 834) is correct that also Somers's insistenceon historical specrficity does
not ger around general lawlike propositions.
 l*/}h, flru'.2 l,
,/
58
Pry'' Herbert Ki
scientiststo agreeon an account of a phenomenon's causes.The ilheqla are
temporal priority of the cause vis-d-vis the consequence,the independlto
of the causefrom the effect,and the (ab)normality condition. The first two
need no explication. The last criterion follows from comparative logic: all
sorts of processesare necessaryto bring about a certain outcome, but only
few of them covary with the specific outcome consistently. They are "ab-
normal" againstthe background processesthat go on all the time without
bringing about the phenomenonto be explained.The four empiricalcriteria
to identify causality are covariation between causeand effect,comparison
with alternative explanations across settings,processtracing to eitablish
the temporal order of things, and plausibility argumentsabout cause-effect
relations.
The third criterion of causalinference([ab-]normality)and the first two of
the criteria of empiricalverificationare manifestlyinsufficientto allow iden-
tification of causality for familiar reasons:covariation does not establish
causality. The first and second criterion of causal inference and the second
criterion of causal empirical analysisgo to the heart of our subjectmatter:
temporal sequence,independenceof causefrom effect, and processtracing.
Gerring'sfourth empirical criterion, plausibility,is too vagueto be of much
'what
use for our discussion. is striking is that the key analytical and em-
pirical criteria of causal analysis- temporal sequence,independence,and
processtracing - are inherently ambiguous and leave a great deal of room
for interpretation by social scientistsand the scholarlycommunitieswithin
which they operate.
'With
regard to sequence,how much temporal priority do we require a
variable to exhibit in order to count as a cause?Gerring notes that prox-
imate causestemporally close to the outcome "are often referred to as
'occasions'
for specificoutcomes,rather than causes" (Gerring, 1999: zz).
In a similar vein, how independentmust an "independentvariable" be from
its purported effect to count as a cause?The problem that looms large over
explanatory hypothesesis that they border on tautologies.Gerring (zoor:
t38-4o, r42-43) quite righdy claims that there are hardly any tautologies
in the strict sense,but that the informational value of a proposition with
causalintent is a matter of degree.How informative is it to say thar party
identification,measuredas a citizen'selectoralregistrationas supporter of
a particular party, is the best predictor of voting for that particular party
in an actual election,conrrolling for citizens' political attitudes and social
position?This proposition is clearly not a tautology in the logical sensebut
may construct a sufficientlyshort causal nexus to count as an "occasion"
more than as a "cause" in what many social sciencescholarswould aspire
to in the causaIaccount of a phenomenon.They would try to ,.endogenize,'
party identification by uncovering the elementsof a citizen's social, cultural,
and political experiencethat contribute to party identification.
Postcommunist Regime D iuersitY
' No abstract logical proof or quantifiable decision rule but substantlve
59
debatesabout alternativeaccountsin a scholarly community draw the line
betweeninsightful explanation and uninformative tautology.Let us refer to
the problem of temporal priority as tl'rat of causaldeptb.
, Thor. who are seekingconsiderablecausaldepth may run into a conflict
with another empirical criterion of verification in causal analysis,that of
processtracing.A causal account needsto demonstrate a temporal sequence
of eventsand processes. The "deeper" the causesare, the more distant they
are from the ultimate explanandumand the more tenuousmay be the causal
sequencethat leadsto the ultimate outcome.The criteria of causaldepth and
processtracing thus involve trade-offs. Before we examine the trade-offs, let
us illuminate the activity of processtracing in the social sciencesby drawing
on a closely related concept, that of a causalmechanisnt.
To acceptsomethingas a causeof a social phenomenon,it must involve
mechanism(s)that brought about the effect.Mechanismsare processes that
convert certain inputs into outputs (Hedstrom and Srvedbetg, 1998: 7).ln
addition to temporal precedence over outputs,theremust be someintelligible
linkage between antecedentand consequence.In the spirit of Max'Weber
(:rgZB),we may say that a causalmechanismis intelligiblewhen it involves
:therational
deliberationof human beingsin the production of someoutcome.
This proposal has severalimplications.
First, the concept of causal mechanism ip the social sciencesimplies
,methodologicalindiuidualisrnin the weak sensethat causal mechanismsrely
on human action, even though eachaction may be constrainedby collective
iand aggregatephenomenaexternal to the individual actors (resourcedistri-
butions, the temporal structure of events,the physical distribution of actors,
and the collectivelyunderstoodrules and anticipatedconsequences of action
that are often codifiedin formal institutions).Employing causalmechanistns
in explanation)we move from a highty aggregatelevel of socialentities(sets
Ibf human beings,structures,and institutions)to the individual-levelconduct
Of particular actors in order to account for higher-leveloutcomes.This weak
inethodologicalindividualism is not inimical to the considerationof struc-
tural and collectivephenomena.It only requiresthat we treat individuals'
actions as critical ingredientsin any account of structural transformation,
suchas that of political regimes.
il ,Second,a mechanismworking through human action is intelligible only
. r r:t -,-r---^-^^ -^L^l
f it is basedon the actors'more or lessexplicit, deliberatepreferencesched-
. As Boudon ltggl) argues,the finality of a social scienceexplanation
is grounded in the understandabilityof human action. This requires that
actorscan invoke some higher obiectivesthat inform their conduct (Weber's
itwertrationalifaf).Pushed to its ultimate consequence,intelligible action is
i;,lationalwhen actors choose among different objectives, explicitly ranked
ln, theit preferenceschedules,in light of scarcemeans at their disposal in
 ,r'/
,/,/60 Herbert Kitscbeh 6r
Postcommunist Regime D iuersitY

such ways as to maximize their overall value satisfaction (Weber'sZweckra-

gap or lag between input and output, causeand effect" (25). This statement
tionalitiit). Purely affective or "traditional," habitual conduct is unintelligible
signals a potential conflict with at least one formulation of the criterion of
and doesnot count as a social mechanismin explanatory accounts.The same
temporal causal depth elaborated earlier. Advocates of the need for causal
appliesto physicalcausesof behavior.6
mechanismscould embrace the call for temporal causal depth of an expla-
Third, mechanisms require that we specify the social knowledge actors
nation, as long as linkages acrosstime can be partitioned into infinitesimally
bring to bear on intelligible action. Given that actors often encounter risk
small steps. Each step is linked to the next by a causal mechanism' Causal
or uncertainty, they must make empirically unproved assumptionsabout the
depth results from assemblingvery long chains of proximate causesexpress-
consequences of alternativestrategiesin the pursuit of their objectives,usu-
ing causal mechanisms.No single stepr however' must be too large.
ally without being able to assign objective probabilities ro rhem. Identifying
What is a small or a large step, of course' is a question of ontology and
these interpretations of the situation is often as important in understand-
not logic or methodology. If we take the imperative literally, however, that
ing human action as determiningthe actors'preferencesand the "objective"
causalmechanismsshould be very close to the explanandum in spatiotem-
constraints of the situation, observed by the social scientist with the benefit
poral terms, the proximity criterion has, in my view, little merit in many
of hindsight.
explanatory contexts of the social sciences,most certainly in the case of
This ontological conception of mechanism (weak methodological indi-
postcommunist political and economic regime change.
vidualism with intelligibility or rationality and social knowledge) differs
Precommunist political regimes acrossEastern Europe displayed striking
from a purely epistemologicalconception of mechanism recently invoked by
diversity.TIn some regards, postcommunist polities reproduce this pattern.
Elster(r998). Elsterwishesto reservethe conceptionof mechanismfor causal
Countries that were democratic or semiauthoritarian in the interwar period,
linkagesthat operate sometimesbut not always. They explain, but do not
such as the Baltic countries, particularly if they also had older semiauthori-
predict, according to the logic "If A, then sometimes B" (Elster, ry98: 49).
tarian roots in the Habsburg Empire with its mechanismsof political repre-
There may always be intervening causesthat suppressthe causallink between
sentation, such as Czechoslovakia,Hungary, Poland, Slovenia' and Croatia,
A and B or functionally equivalentcausesthat create B without A. More-
tend to becomefull-fledged democracieswith civic and political rights imme-
over, the relationship between A and B may be of an inherently probabilistic
diately after 1989, with the Slovak and Croatian laggards fully converging
rather than a deterministickind. There are no laws in the social sciences
toward the rest of the group by the late r99os. Countries with authoritarian
in the strict senseof deterministicprocesseswithout interveningvariables,
patrimonial regimestend to yield greater postcommunist regime diversity
functional equivalents, or stochastic fuzziness.
and instability. Some of them gravitate toward formal electoral democracy
I proposeto separatethe problem of nondeterministicrelations between
with tenuous levels of civic and political rights (Russia,Ukraine); some start
causesand consequences, broadly conceived (stochasticlaws, intervening
out in this category but appear to make headway toward full democracy
causes,etc.), from the ontological conception of causal mechanismin the
(Bulgaria, Macedonia, Romania, and, to a lesserextent within the former
social sciences.The latter calls for social explanationsthat shift the level of 'soviet
. Union, Armenia, Georgia, and Moldova); others relapseinto author-
causalprocessanalysisfrom that of aggregatesand structuresto individual , itarianism after a fleeting moment of liberalization (Belarus, Serbia until
action under constraints. It is entirely open whether thesecausalmechanisms I
2ooo). The colonial periphery of the former Soviet Union' finall5 almost
are stochasticor deterministicor whether interveningcauses(mechanisms?)
i iuniformly drifts toward authoritarian solutions. Here' there are no precom-
suppresstheir effects.
,',munistprecedentsof civic mobilization.

Trade-offs behueen Causal Depth and Catrsal Mechanisms
According to Hedstrom and Swedberg ft998: z4-z1), the virtue of causal
mechanismsis their relianceon action, their precision,their abstractionfrom
concrete contexts, and their reductionist strategy of opening black boxes.
This, they claim, is in line with the striving of all science"for narrowing the
6 Of course,physically induced conduct,
such as the lethargy of a starving population, has social
consequences.Moreover, the physical processes(famine, etc.) may themselves be brought
about by social mechanismsbased on deliberate action, such as the destruction of food crops
in warfare.
ii , At first sight, this account of postcommunist diversity lacks actor-related
ilcausalmechanisms.There are two not necessarilyconflicting ways to rem-
fedy that problem, but they remain inherently controversial. One of these
luays operatesthrough cognitive processeswithin and acrossgenerationsof
lactors. Those who were alive in the interwar period' survived communism,
jand lived to welcomepostcommunistpolitics draw on skills and experiences
ftheyneve.quitelost duringcommunism.Sucha cognitivecapitalstockhelps
fthem formulate expectations about the dynamics of the new polities and
i7 In this and the following paragraphs, I am drawing on Kitschelt et al' (1999), chaps
and z.
/62
Herbert Kitschelt
contributes to political rejuvenation - for example, in the formation of po-
litical parties with distinctive programmatic appeals. Moreover, even where
postcommunist actors are too young to have experiencedprecommunist rule
or even much of communism, their eldersmay have handed them down cru-
cial skills, interpretations,and experiences.In addition to politics, this may
apply also to the businesssphere.R6na-Tas and Bcircjcz(zooo: zzr), for ex-
ample,found that family history mattersfor the new postcommunistpolitical
entrepreneurs,and here the experience of grandparents who lived most of
their lives before communism is of substantially greater influence than that
of their parents.Moreover, postcommunistentrepreneurialismis negatively
correlated with precommunist family landownership in some countries, such
as Bulgariaand Poland,where it involved essentiallysubsistence farming di-
vorced from market involvement.
Those who accept cognitive causal mechanismsover the span of more
than fifty years and two generationswould argue that the human mind is a
robust depositof ideasand information. In periods of societalcrisis,people
are capable of activating their long-term memory and scan its content in
order to interpret their strategic options under conditions of uncertainty.
Moreover, technical and institutional memory enhancers(scripture, literacn
media of communication, education,professionalsin charge of preserving
memories)and Kubik's "cultural entrepreneurs"(Chapter ro in this volume)
extendthe capacityof human actorsto retrieveand processinformation over
lengthy periods of time.8
The other way to make plausible the efficacy of intermediate distance
explanations focuses on political practices and institutional arrangements
rather than beliefs and cultural orientations. In this vein, the effrcacy of
causal mechanismsin the choice of political regimes and economic activities
operating across periods of more than fifty years can also be made plausi-
ble by decomposing their operation into smaller steps, each linking more
proximate phenomena.Thus, precommunistpolitics and economicshad an
impact on the way communist stateswere formed under Soviet tutelage with
differentialbargainingpower of the new communistrulersvis-i-vis what was
left of civil society after the Nazi war. These nationally and even subnation-
ally different bargaining constellations reassertedthemselvesin the period
of de-Stalinizationfrom :'953 to 1956, when Stalinist rulers had to con-
front domestic societal forces that were acting on precommunist experiences
then only five to fifteen years old. The increasing post-Stalinist diversifica-
tion of communist rule inside and beyond the Warsaw Pact, including that
between republics of the Soviet Union itself, constitutes a mechanism that
then explainswhy communist countriesand regionsrespondeddifferentlyto
the challengesof technologg political elite turnover, and changing societal
d Foracritiqueof anoverlynarrowconstructionof
t h e c r i t e r i o n o f s p a t i o t e m p o r a lp r o x i m i t y ,
s e ea l s o G o l d s t o n e ( r 9 9 8 : 8 3 8 ) .
61
Postcommunist R egim e D iuersitY
demands (induced by education, different family structures, etc.) from the
r96os through the r9Sos. The diversificationof communist rule, in turn,
causesdistinct patterns of oppositional mobilization and incumbent elite re-
sponse in the ultimate crisis of communism. These configurations, finally,
influence the initial and ultimate political and economic outcomes of the
communist breakdown. Grzymala-Busse'schapter on communist successor
parties (Chapter 5 in this volume) highlights the linkage among the last two
causalstepsin the chain.
Even this cascadeof probabilistic causal linkages between precommunist
politics and society, the establishment of communist rule, its post-Stalinist
transformation, and its ultimate collapse,however, may not satisfy those
who insist on very close spatiotemporal proximity in causal mechanisms
(Kiser and Hechter, ry9r; ry98). Drawing on the epistemological principle
of Ockham's razor, they tend to discard "deeper" explanations as ineffi-
cient and causally irrelevant for an outcome. This epistemological move in
the evaluation of alternative causal accounts constitutes the main bone of
contention between different camps in the study of postcommunist regime
transition.
As a rule of thumb, causal mechanisms that are temporally proximate
("close") to the effect they claim to explain account for more statistical
variance in the effect than deeper and less proximate causes most of the
time, even though not always. Intervening external shocks reduce the path
dependenceof ultimate outcomes and dilute the measurableeffects of tem-
I porally more distant causes.But is a shallower causal explanation realiy
:i "better" just becausethe statistically explained variance is greater than that
of a deeper alternative? Advocates of causal proximity engagein a statisti-
cal modeling strategy to which their adversariesobject' The former rely on
single-equation statistical models to account for an ultimate outcome and
enter rival candidates for causal explanation on the right-hand side of the
equation, regardlessof their temporal position within a chain of causation.
ilemporally deeper causeswith less direct impact on the ultimate outcome
then tend to wash out in the hunt to find the statistically most efficient
e*planation. Single-equation models constitute the technical implementa-
tion of Ockham's razor. By emphasizing multivariate statistical efficiency
,bver ontological criteria of temporality and human action in causal mecha-
,hisms,Ockham's razor promotes explanatory shallowness,with the ultimate
danger of explanations approaching tautological reasoning.
r,, Alternatively, one might conceive of the statistical explanation of an ulti-
lnate outcome (regime change) as a multiequation model that links tempo-
,tdlly prior to subsequentcausesof an outcome in a stepwiseestimation pro-
ure. This analysis reveals the temporal interconnections between causes
hat are directly pitted against each other in the single-equationmultivariate
tatistical tournament inspired by Ockham's razor. A multiequation model
causal chains may reveal that some proximate causeis, to a considerable
 ,/,/
"4 Herbert Kitscheb
extent, endogenousto a deeper cause. Moreover, it may reveal interaction
effectsbetweendeepercauses("structural conditions") affecting shallower
causes("triggers"). Finally, the same variable may appear in different cases
at different stagesin the temporal causal chain that must be reconstructed
to arrive at a satisfactory causal account. These are just examples for a gen-
eral epistemologicalrule adversariesof ockham's razor advance:respectthe
temporal complexity of causalrelations.
The searchfor causalmechanismsand causaldepth are murually supporr-
ive but also mutually constraining criteria in the construction of satisfactory
explanations. The requirement that causal mechanisms run through inten-
tional action limits the spatiotemporaldepth of explanations,while causal
depth criteria militate against minimizing lags between cause and effect,
accompanied by the dangers of tautological reasoning. How far we wish
to push back the envelope of causal analysis in terms of spatiotemporal
depth and history is a pragmatic and ontological question, not a mafter of
epistemology,logic, or statistics.Reasonableminds may disagreeon what
counts as a good explanationof a particular phenomenon.I employ the next
two sections of this chapter to persuade readers that many explanations
currently offered to account for postcommunist regime variance are either
too shallow (temporally proximate, bordering on tautology by blurring the
line betweenexplanansand explanandum)or too deep (without a chain of
causalmechanisms)to be fully satisfactoryin an account of postcommunist
pathways.
ExcessivelyDeep Explanations of Postcommunist Regime Diversity
The current paradigmatic caseof an excessivelydeep compararive-historical
explanation in political scienceis Robert Putnam's GSgl) account of demo-
cratic processesand performance in northern and southern ltaly, claiming
that twelfth- and thirteenth-century polity formation in the two parts of Italy
shaped both the institutional practices as well as rhe economic outcomes in
the different Italian regions in the secondhalf of the twentieth century. lfhat
Putnam lacks is a convincing account of the transmission from thirteenth-
century Italian conditions to those of the twentieth century. No mechanisms
translatethe "long-distance"causalityacrosseight hundredyearsinto a more
proximate chain of closercausalforcesacting upon eachother. Margaret Levi
(r996:46) thereforeidentifiesin Putnam'swork a "metaphoricaluseof path
dependencewithout the rigorous analytics a compelling application of the
conceptrequires." lfhen one tries to supply such causalmechanismsacross
historicaltime, one soon discoversthat civicnessand governancewildly fluc-
tuated in Italy and that, if anything, the eighteenth-century governmental
institutions should serve as the referencepoint for fwentieth-century devel-
opment, not the thirteenth- and fourteenth-century city-states in northern
Italy and the Kingdom of Sicily (Sabetti,1996: esp./7-7).
PostcommunistRegime Diuersiry 65
In a similar vein, in the analysisof postcommunistpolitical change,ob-
servers often invoke different religious doctrines and zoftesof administratiue-
political control under tutelage of either Prussia, Habsburg Austria, Russia,
or the Ottoman Empire as determinants of late twentieth-century politics but P
I
generallydo not specify the mechanismsthat lead from these antecedentsto
the political consequence.This explanatory gap occurs even in the better
historicalcomparativeanalyses,such as Sch6pflin'sftSSl) broad compara-
tive sweep, even though his study provides, in principle, the facts and tools
to work out the causal mechanisms that link the institutional transforma-
tions of diverse polities across different historical "stages" and "rounds of
struggle"in EasternEurope.eSchopflinG9%t r9-zzl discusses the different
professionalization the of post-World War
levels of in state bureaucracies
I East and Central European polities, yet his later description of commu-
nist regime differentiation in the r96os and r97os does not systematically
draw on such earlier regime variations and their influence on actors' strate-
gic capabilitiesand constraints.For example, in contrast to Hungary and
Poland with a bureaucracy penetrated by the gentry, Schcipflinattributes to
interwar Czechoslovakia "a relatively well-functioning administration and
considerableautonomy of the law" (zo). Yet he does not explore a linkage of
thoseearlier administrativepracticeswith the later prominenceof "illiberal
technocrats"(zr4) in communist Czechoslovakiaat a time when the state
party apparatus of Hungarian and Polish communism was already crum-
bling and a generation of reformist apparatchiks took control of the ruling
parties.
In terms of correlational statistics,religion provides a surprisingly strong
association with regime outcomes.'o But it is unclear what mechanisms
causal analysesare really picking up with religion. Is it religious doctrines
about the relationship between economic activity, politics, and religious or-
ganization that matter for the founding of new political regimes?Or is re-
ligion simply a tracer of institutional historical correlates - for example,
the timing of state formation and the construction of rational bureaucratic
administrations that facilitated the rise of market capitalism and the devel-
' op*ent of civic societies?In this latter case, there is nothing intrinsically
crucial about religious beliefs that affect postcommunist regimes.Instead of
starting from religion, we would need to construct a causalchain starting
from state formation to account for postcommunist regime diversity.
9 Incomplete "deep" accounts of postcommunist regime variation also tend to characterize
A g h ( r 9 9 8 : c h a p .r ) , C r a w f o r d ( t 9 9 6 : c h a p .r ) , a n d t h e t y p o l o g i c a l a p p r o a c h i n O f f e ( r 9 9 4 :
z4r-49).
' to
The precise statistical strength of that association depends on intricacies of operationaliza-
tion. It is slightly weaker, if dummy variables signal the cultural dominance of a religion
(cf. Fish, rgq8b). It is somewhat stronger if we create a single ordinal-scale variable with
the highest value for compatibility with democracy and market liberalism going to !0estern
C h r i s t i a n i t y ,f o l l o w e d b y E a s t e r nO r t h o d o x y a n d t h e n I s l a m ( c f . K i t s c h e l t ,r 9 9 9 ) .
//
bb Herbert Kitscheh Postcotnmunist Regime D iuersitY o/

Religion could also be a variable that simply measuresproximity to the componentsmost influential in establishingneighbor effectsare o.fa simple
Vest (cf. Fish, 1998a: z4r). Long-distancecausality accounting for post- modernizationtheoreticallogic. Moreover, as I already suggested,some el-
communist regimessometimestakesthe form of geostrategicconsiderations ementsof the index (FDI, trade) are in fact endogenousto the explanatory
within the international system. According to this view, proximity 1o tn. variable.Second,Kopstein and Reilly do not submit their neighborhoodef-
'West
and a resentmentof Russian domination under the shadow of the fectsindex to what f h"u. called the multivariate tournament of variables.
Brezhnevdoctrine translate into adoption of a market-liberal representative How collinearis their index with other variables,and how can they establish
democracy.It is not easn however,to spell out the causal linkage between causalprimacy for the neighborhood effect?
geographicallocation and postcommunistregimechange.There are at least Finally,none of the casestudiesexplainshow neighborhoodeffectstrump
four analyticalmodesto bring geographyto bear on the problem of causally other variables.All structural theorieswould predict that Slovakiaand also
explaining postcommunistregime diversity.First, the causalchain is histor- Croatia would have returned to the cohort of postcommunist economic and
ical in nature and would run from stateformation through interwar politics politicalreform vanguardpolities.In a similar vein' structuraltheorieswould
and post-Stalinismto the regime transitionsof the r989-9r period. In that predict backslidingin Kyrgyzstanand an intermediate,volatile traiectory in
sense,geographicalIocation doesnot add anything that could not be gleaned Moldova, Georgia, and Macedonia, as well as in Armenia, all countries
from variablesmore informative in the socialscientificsenseof permitting the where new titular majorities find it imperative to protect the autonomy of
constructionof causalmechanismsthat account for the ultimate outcome. their new polities by cateringto Westerninstitutional principlesin order to
The secondview of geographytreats it as a proximate causeof regime receive,.ripo.t and resourcesfrom new allies. Kyrgyzstan,a country dis-
forms. The implied causal mechanismhere is focusedon the economic in- tant from the West but initially pushing for economic and political reform
centives'Westerncountrieshave offered to their postcommunistneighbors, after the collapseof the SovietUnion, may have experienceda relapseinto
for now more than a decade, to adopt political and economic rules of the authoritarianism not so much because of its distance from the West and
game that are similar to their own. If this account is correct, trade flows be- the political practicesof its neighbors. but simply becausefrom the very
tween the OECD West and postcommunist countries, as well as patterns of beginningits reform politicians were unable to build political parties and
foreign direct investment(FDI) should signal a closelink to postcommunist professionalbureaucraticadministrations'Western in a domesticenvironmentinhos-
political regime form. As a matter of fact, the statistical association between pit"Ut. to the institutional correlatesof a capitalistand democratic
thesevariablesis quite modest, however,and certainly much weaker than but maybe not geographical theory can explain the back-
ioliry. Structural
that betweenregimeform and "deep" variables,suchas religion,precommu- ilidi"e into semiauthoritarianism among a final group of countries,those in
nist regime, or mode of communist rule.rr Somecountries,such as Poland, lelatiie physical proximity to the West but emerging from institutional lega-
engagedin vigorous economic reform, but did not benefit from FDI flows biesand-practicei of collective action that should make the establishmentof
until some time later.FDI herecertainly did not work as a proximate stimu- liberal dimocraciesrelativelydifficult (Albania' Belarus,Serbia)'
lus fo economicreform. Moreover, it appearsthat the causaldirection goes !l ,The strongest,most convincing version of geographic influence trumping
in opposite ways, namely that FDI and trade are endogenousto previously btructural faitors would have to be establishedthrough a fourth mechanism
enacted economic reform. rnicrologic, the future-oriented expectations of postcommttnist politicians.
' . l.r,r rr-:^-/EIT\
A third way to think about geography is in the fashion proposed by al elites in countries close to the European Union (EU)
"nJeco.romic
Kopstein and Reiliy (Chapter 4 in this volume), namely as physicalcontigu- te benefits through trade and FDI in the near or medium-term future)
ity and communicationwith neighborswhoseeconomicand political reform .theVembrac" ..ono-t reform and democratic civic and political rights'
strategiesmutually influence each other. Kopstein and Reilly have made an Prleworski (r99r: r9o) suggestedearly on, "geographyis indeedthe sin-
'rqurol
interesting opening move on this worthy avenueof research,but in order to to hope that Eastern European counrries will follow the path to
-prosperity."
convinceme they would have to addressat least the following issues.First, racy and The best way to show the independent influ-
their index of neighborly communication is a murky amalgam of numerous of g"og.aphy through expectations would be to detect extraordinary
, ..r 1 -,-, ----,-^l -,^):-t^-^
elementsthat I would like to seeanalyzedseparatelyin order to determine i. ..for* efforts among countries with weak structural predictors
(r) the unidimensionality of the concept that is postulated to inform the ist"blith democracy and economic reform, yet plausible expectations to
index and (z) the relative influenceof index componentson the dependent Lthe EuropeanUnion in the foreseeablefuture (Bulgariaand Romania).
variable. My suspicionis that the index is not unidimensionaland that the far. the .r,id.n.. on these casesis mixed. Both countries clearly stand
ithin the cohort of formerly patrimonial communist countriesin terms
" I have reported evidence in Kitschelt ftggg, jZ). runn8 liberal democracies with firmly protected civiil and political
ituting
,/ Herbert Kitschelt
68
rights.Yet their economicreform efforts,in terms of short-termstabilization,
medium-term market liberalization,and long-term constructionof market-
supporting political institutions certainly lends little support to the propo-
sition that expectations of EU accessionenabled them to move ahead of
the pack of other formerly patrimonial communist polities and to closethe
g"p to the lead countriesof the postc<lmmunistregion in Central Europe (cf.
Kiischelt,zoor ). Geographyas a "deep" explanationmay thereforestill lack
an empirically corroborated mechanismto count as a plausiblereason for
patterns of regime diversiry and economic reform strategy among postcom-
munist countries.
ExcessivelyShallow Explanations of Postcommunist Regime Diversity
Shallow explanations provide mechanisms and high statistical explanatory
yields but little insight into the causal genealogyof a phenomenon.This
danger is evidencedby Philip Roeder's(rgg+; 1997; zoot) analysisof di-
versity in the regime types of Soviet successorstates.Let me focus on his
most refined recent statement of that perspectiveand particularly on the
analysis of regime outcomes from 1 994 to 2ooo. Roeder recounts in ad-
mirable detail how bargaining configurations between different interests
propelled different fission products of the Soviet Union toward democra-
cies-with full contestation,exclusionarydemocracieswith disenfranchised
Russian minorities, oligarchies, or autocracies.The more incumbent power
eliteswere fragmented before the initial republic-levelelectionsof r99o, then
before rhe initial postindependenceelections,and finally in subsequentyears
to the turn of the millennium the more the dynamics of bargaining promoted
a democraticoutcome.The two key variablesare the dominanceof agrarian
or urban-industrial agenciesbefore r99o and the presenceof indigenized
or Russianparty and managerialelites.The prevalenceof urban-industrial
agenciesand nonindigenized Russian elites threatened the viability of hege-
monic autocratic solutions, presumed to be the first preference of incum-
bent Union Republic party elites belonging to the titular majority. lfhere
urban-industrialagenciesprevailed,the result was either full democracy(in
Armenia, Lithuania, Russia, and the Ukraine) or exclusionarydemocracy
(in Estonia and Latvia). Four caseswith mixed control structuresoscillate
between autocracy (predominant in Azerbaijan, Kazakhstan,and Belarus)
and oligarchy (Georgia). Starting from rural-agricultural dominance and
nonindigenizedparty elites,Kyrgyzstan ends up with a halfway house be-
tween democracy and authoritarianism, whereas Moldova inches toward
democracy. Three cases of rural-agricultural agency dominance yield au-
thoritarianism (Tajikistan,Turkmenistan' and Uzbekistan).
Roeder'smeticulousprocesstracing (Roeder.zoor : 5z) remains unsatis-
factory in two ways that mutually reinforce each other. First, the model says
nothing about the origins of the power configurations on the eve of the Soviet
Postcommunist Regim e D iuersi \t 69
union's disintegration.second, the model fits the data only becauseof the
ratherodd characterizationof someof the regimeoutcomesin the late r99os,
and even then it is sometimesdifficult to seethe linkage berween stipulated
mechanismsand regime consequencesover time.
Let me work backward from the unsatisfactory characterization of regime
outcomesand show that "deeper" explanationsactuallyaccountfor properly
describedpostcommunist regimesin more sarisfactoryways, even th-ugh
such explanations treat some of the short-term event history, coalitional
struggles,and victories of individuals and their followers as not worthy of
detailedtheoretical explanation." In contrast to Roeder,I consider the long-
term practice of state building and the mobilization of secondaryassocia-
tions (interestgroups,parties)beforeincorporation into Russiaor the Soviet
union as well as the mode of incorporation in the Soviet Union as criti-
cal variables that shape the capacity of counterelites to threaten autocratic
self-transformations of the elites. Roeder's approach focuses on intraelite
bargaining in order to explain short-term regime outcomes, ignoring the po-
tential for mass action as a critical element in elite calculations.I do not
deny the importance of elite fragmentation, but I emphasizethe collective
action potentials of citizens and incipient counterelites that result from the
history of state building, precommunist patterns of political mobilization,
and resulting modes of organizing civic compliance under communism..l
Roeder'sclassof democraciesis incoherent.Seriousinfringementson civil
and political rights in Armenia, Russia, and the ukraine (even including
electoralmanipulation) make it impossibleto place them in the samecaregory
as full democracies.conversely, not only Lithuania but also Estonia and
Latvia qualify as full democracies after their revision of citizenship laws.
The Freedom House rankings reflect this, but not Roeder's r.rnusualand
contrivedmeasureof accountability.'+y6", we needto explain is a division
12 Actualln it is hard to see how some of Roeder's own process tracing - for example, of the
'
oscillationsof regime patterns in Azerbaijan, Belarus,oi Talikirtar, - follow a tight analytical
'
l9qi. Qoor 4o-42). Had the outcomes been otherwise, Roeder could have told a slightly
'
different story about victory or defeat of factions at war.
r3 Roeder attempts
to disqualify my approach as building on the distinction between "rational-
i l e g a l a n d p a t r i m o n i a l c u l t u r e s " ( z o o r : 5 r ) . H e m a k e s t h e r h e t o r i c a l m o v e r o a s s o c l a r ea n y
. '' analysisof temporally deeper causeswith the pursuit of "bad" cultural explanation against
. "good" (and shallower) rational-instrumental and institutional explanation. The final seoion
; of my chapter discussesthe inadequacy of this dichotomy. Needless
to say, I have always
', placed the emphasis of my approach on political practices and institurions,
but I would not
.f want to exclude the importance of cultural elementsin a very specificsense:even good game
;' theorists would not want to discount elementsof culture (corporate and otherwise), such as
' ' memories of past (inter)actions and beliefs about the expectations those episodesgenerated
I among current players in order to model how current actors calculate the payoffs resulting
. from alternative strategic choices.
I a F o r g o o d r e a s o n ,t h e F r e e d o mH o u s e a n n u a l r a n k i n g s a t t r i b u t e t o R u s s i aa n d t h e U k r a i n e
barely "semi-free" status, scoring them below the median value on their democracy scales
 Herbert Kitschelt PostcommunistRegimeD iuersiry 7r
/,"
of four groups: (r) full democracies(Baltic countries);(z) semidemocracies and Bessarabia/Moldova).Here the new mastersengagedin inconclusive
with a pattern of improvement throughout the r99os (Armenia, Georgia, efforts to assimilate indigenous ruling classes,such as the Armenian dias-
Moldova); (3) initial semidemocracieswith a pattern of slower or faster pora pariah entrepreneuri and the Georgian nobility and Orthodox Church.
deteriorationin the r99os (Belarus,Russia,Ukraine);and (4) full autocracies. in ..i."t.d episodesof struggle,indigenousethnic movementspressedfor
Group r countries were precommunist interwar semidemocracieswith grearerautonomy, even under the Soviet administration. The presenceof
considerableassociationalmobilization basedon class.nation. and economic a patrimonial state apparatusand the lack of twentieth-centuryexperiences
sector in an environment of beginning industrialization and bureaucratic with democraticor semidemocraticassociationalself-organization,however,
state building with a formal-legal rule of law. Soviet rule built on this orga- constrainthe transition of postcommunistregimesin thesecountriestoward
nizational infrastructureand respondedto the collectiveaction threat with full democracy.The anti-Russiandispgsitionsof the new titular maiority
a mixture of repressiveand accommodativepractices to co-opt and depoliti- elitesmake them receptive to Western incentives to adopt democratic prac-
cize a potentially formidable civic mobilization. rVhen hegemony crumbled, tices in exchange for economic resourcesand political assurancesthat prop
this mobilizational potential came to the fore and made indigenous and non- up their independenceagainst an overbearing Russian neighbor. As noted by
indigenous elites seek a regime transition through negotiation. Eventually, Kopstein R.illy, geographyand power distribution in the international
"nd
even the nonindigenouspopulation 6nds it advantageousro assimilateinto ryri.* may, atthe margin, tip the balanceof regimedynamicsin thesecoutl-
the new independent,'Western-oriented democracies(Laitin, r998). tries toward democratization.
By contrast, in group 2, the core of long-standing patrimonial Russian The six countriesof group 4, fr.nally,have all evolvedinto unambiguously
state building without broad precommunist interest group and party mo- authoritarian regimes (Azerbaijan, Kazakhstan, Kyrgyzstan, Tajikistan,
bilization and only very limited ethnic differentiationand autonomy move- Turkmenistarr. Uzbekistan).Settingasidemany differencesamong them'
"nd
mentsin different historicalepisodes(in Belarusand Ukraine in r9o5, r9r7- they have in common a legacyof (r)virtual statelesssocietiesbefore (z) very
zr, after 1989), the realm of communist elite action in the late r98os was late i.rco.poration into the Russian empire in a (l ) more colonial rather than
much lessconstrained by bottom-up challengesand involved factional efforts assimilativemode (Kapp eler,r993: 174).Earmarking theseregions for strate-
to engineerpreemptive reform while converting public assetsinto the private gic resourceexploitation, Russianrulers sought to subordinateindigenous
property of managerialand party elites.It resultedin an unstableoscillation ilit.r, org"nized along primary relations of kinship and tribe, unde-ra light
between semidemocratic and oligarchical regime forms in an environment colonialadministrationthat did not fully assertevenpatrimonial techniques.
of volatile parties and interest associations.The high level of education of i.ater. u.rd.. Soviet rule, those indigenous clientelist and kinship-based
at least the urban population, however, makes a suppressionof autonomous practices of governance were not entirely displaced, but transformed into
secondaryassociationsinauspiciousin the longer run. Even elite fragments fegional-spatialclientelistpatrimonial networks co-opting the existing net-
bent on authoritarianism,such as those that assertedtherlselvesin Belarus puorksof eld.6 and tribal headmen (JonesLuong, zoor). The assimilation
in the late r99os, are likely to encounter increasingly stiff resistanceto the bf indieenous ethnic elites into the communist apparatus of domination per-
for collective action.
realizationof their preferences.t5 $.t.t"t.d the incapacig of clients and constituencies
Group 3 are countries incorporated late into the Russian empire in the iboth the earlier breakdown of tsarist rule ft9r7-zr) and the recent fall
late eighteenthand first half of the nineteenthcentury (Armenia, Georgia, bf the SouietUnion thus did not spawn powerful ethnopolitical autonomy
ts in any of these incipient countries, with the partial exception of
Baku region of Azerbaijan in rg:l7. when the soviet Union imploded,
on civil rights and political liberties ranging from r (highest) to 7 (lowest). By zoor, Russia ither the indigenous elites simply endowed their existing system of rule
receivesamere5*5andtheUkraine4f4,whereasailBalticcountriesareratedasr+2.
' 5 I n l i n e w i t h P r z e w o r s k ie t a l . ( z o o o ) , I a s s e r tt h a t e c o n o m i c d e v e l o p m e n ti n t h e l o n g e r r u n
ith a new ethnonational ideology and perpetuatedwell-worn practices
Kazakhstan,Uzbekistan),or brief interludesof uncertainty,internal compe-
matters for political mobilization and durable democracy,though constrained by the legacies
ofpast regime formatjon and associatedexperiences.Roeder (zoor: zo) initially appears to ition, and civil war enablednew (or returning)rulers to reestablishtheseold
deny the explanatory power of economic development to predict regime outcomes but later tteins of autocracywithin the short spanof a handful of years(Azerbaijan,
brings this variable in through the back door by attributing to regions with a prevalence yrgyzstan,
-I Tajikistan).
of urban-industrial elites a greater threat potential vis-i-vis the autocratic designs of party analysisin orderto show
havegonebeyonda merecritiqueof Roeder's
elites.His early assessmentis faulty becausehe does not correct per capita GDP in the Soviet
t sha[Lw expianations may not only be shortsightedbut, i' their zeal
repubhcs by purchasing-power parity. I)oing so places them exactly in that second-highest
quintile of economic development to which Huntington (r99r) and Przeworski et al. (zooo) iexplain processesin all their details, also distort the characterization
loutcomei. Instead, by homing in on state formation (including ethnic
attribute the highest probabiliry of making a durable move toward democracy.
./r, Herbert Kitschelt Postcommunist Regime D i uersity 73

incorporation), economic development,and political experiencewith au- igreater or lesserexecutive and legislative powers of a presidentialoffice in
tonomous secondaryassociationsand collectiveaction before the advent of .new postcommunist democracies as the ultimate independentvariable ac-
',countingfor economic reforrn trajectories.'z
and under communist rule, a more satisfactoryexplanation emergesthat
involves longer but, frorn the perspectiveof an actor framework relying on
beliefsand experiences,not excessivelylong causalchains.The specifictech- Deep and Shallow Explanations: Rivalry or Complementarity?
niquesof communist rule to createcivic complianceserveas the transmission
belts that link precommunist to postcomnlunist experiences.My explanarory The recent prominence of causally shallow explanationsis due not exclu,
strategyis quite similar to that proposedby Laitin (tSSS) ro accountfor dif- ,sivelyto fashionable multivariate reasoning, driven by the criterion to max-
ferent patterns of languageassimilation by Russian minorities faced with imize explained variance in statisticai terms. The historical experience of
non-Russian-speaking titular majorities since r99r. the "Third Wave" of democratization (Huntington, 199r) powerfully influ-
Shallow explanationsof regime and policy changeare quite common in encesthe penchant for such approaches.\What appearedas the sudden sweep
the comparative literature on posrcommunist polities. After having covered of democracyacrossLatin America, SoutheastAsia, the communist hemi-
Roeder in some detail, becausein my view his work representsthe most ,sphere,and even parts of Africa in the r98os and early r99os made deep
rigorous and interesting effort to date to spell out the logic of postcommu- structuralist and comparative-historical theories appear to be of little use
nist transformation inspired by the imperative of short cause-effectchains (O'Donnell and Schmitter,r986; Przeworski, ry9r: 3). The problem here is
and "mechanisms" operating in small time-spaceintervals,I can now only that the explanatory focus is on the "event" of authoritarian regime break-
mention a few more examplesof "shallow" explanationswithout detailed down, but, as I argued before, all sciencesof complexitS and not iust the
discussion. Higley, Pakulski, and Wesofowksi's GSSS) work, building on social sciences,are bad at making point predictions to account for individual
Higley and Gunther (r992), arguesthat stabledemocraciesresult from elites events.This event-orientedframing of the object of explanation is furthered
that agreeon basic regime parameters,while simultaneously displaying plu- by studies that follow the now fashionable pooled cross-sectionaltime-
ralist differentiation into parties, interest groups, and movements with their ,seriesanalysesin which the dependentvariable is really short-term histori-
'cal fluctuation, rather than lasting regime parametersthat have entrenched
unique objectives. \Thereas the authors conceive of these attributes of in-
traelite relations as an explanation for the rise and persistenceof democracy, Ithemselves.'8
I see it as a simple redescription of democratic practices. In a similar vein, ;il In order to avoid misunderstanding,let me neverthelessreiterate two
Fish's (zoor) intentional explanation of authoritarian backsliding in post- ,,1points. On the one hand, I do not doubt that strong correlationsbetweenwhat
communist regimes borders on tautology. If chief executives promote the ilRoeder, Fish, and others offer as explanations for postcommunist regime
degradation of democracy, and if that process is particularly pronounced i.typesand regime performance in terms of economic reform actually do ex-
where superpresidential constitutions have been adopted, in order to pro- i ist, but I question their status in explanatory accounts of regime diversity,
vide a causal account we had better ask why superpresidentialism could euenif they turn out to prevail over rival causal candidates by purely sta-
entrench itself and subiugate the fate of political regimes under the whims tistical criteria of significancein single-equation tournamenrs. On the other
of individual rulers. The self-servingintentions of rulers faced with this op- hand, I do not deny that short-term accountssometimes,but not often, do
portunity structure certainly do not supply a satisfactory causal account of provide the ultimate explanation and that further backward-orientedpro-
the regime outcomes. In a similar vein, the fact that the replacement of old cesstracing is futile. Beforewe concludethat explanatory chains cannot be
communist successorparties by noncommunist parties and alliancesin ini- temporally deepened,however, we must have carefully specifieda model of
tial postcommunist elections is the statistically most efficient predictor of explanatory layers that takes into account the temporal ordering of forces
economic reform effort in subsequentyears (Fish, ry98b) is not much of an that may impinge on rhe final outcome.'Withthesequalificationsin mind, let
insight. And that this variable "beats" deeperexplanatory variablesof eco-
nomic reform in a single-equationstatisticalmodel is not surprising,given
the proximity berweenfavored causeand effect.'6I would voice similar con- This critique applies to Hellman's (r996) attribution of causalefficacy to executive-legislative
relations in the comparative analysis of economic reform. For a different perspective,see
cernsabout explanatory shallownesswith regard to propositions that treat
K i t s c h e l ta n d M a l e s k y ( z o o o ) a n d K i t s c h e l t ( z o o r ) .
This focus is technically entrenched by employing rhe lagged dependent variable (regime
' 6 S i m i l a r p r o b l e m s a p p l y t o F i s h \ ( r g g 8 a ) e x p l a n a t i o n o f d e m o c r a t i z a t i o n .I n b o t h c a s e s , change) as a control on the right-hand side of the equation and/or by including a full set of
single-equation models, often with high to very high collinearity among the independent country dummies that suck out a great deal ofcross-sectional variance. Structtrral variables
v a r i a b l e s ,d o n o t h e l p t o s h e d l i g h t o n t e m p o r a l p a t r e r n so f c a u s a l i t y . are superior to country dummies but require more theoretical work.
,/
74 Herbert Kitschelt
me outline four different ways to frame the relationship between short-term
and long-term factors in explanatory accounts.
First, one can argue that long-term factors actually trump short-term factors
as causalexplarations. Although ideally it is neat to specify micromecha-
nisms that establishan unbroken chain of causal linkages between deep,
distant causesand ultimate outcomes, sometimesan external shock may
yield a crisis in a system such that actors engagein a randomized trajectory
of trial-and-error in search of new solutions. This trial-and-error process
may be unpredictable experimentation with fluctuating short-term political
coalitions that do not lend themselvesto any systematicexplanations.In
the longer run, however, those experiments will prevail that are consistent
with long-term structural parametersof resourcedistributions and actors'
capabilities,beliefs,and aspirations.In that spirit, Ekiert (Chapter 3 in this
volume) assertsthat broadly perceivedlegaciesof the past offer the most
consistentexplanationsof successfulpostcommunisttrajectories.The short-
term problem solvingof actorsmay generatea greatdeal of noisethat cannot
be explained or patterns of action that are not worth explaining if the ex-
planatory objectiveis more durable long-run steadystatesof postcommunist
p o l i t i c a lr u l e a n d e c o n o m i cg o v e r n a n c e .
Second, one can argue that short-term factors serueas proximate links in
the chain of causation.In this view, deeper structural and shallower, agency-
related explanations are mutually complementary in some kind of funnel
of explanation.Shallow explanationsrely on proximate causalmechanisms
that become useful only if complementedby causally deeper analysesof
regime diversity. Conversely,the searchfor depth must not ignore the provi-
sion of causal mechanismsthat make plausible how structural, institutional,
and cultural parameters translate into strategic,calculated action, which, in
turn, creates new macrolevel outcomes. Intermediary "links" in the chain
betweendeeperand shallowercausesof current outcomesmust be specified.
In this sense,Grzymala-Busse's (zooz and Chapter 5 in this volume) discus-
sion of the transformative capacitiesof communist parties in the r97os and
r98os providesan intermediatecausalmechanismto link precommunistand
postcommunist political rule.
To adapt and modify a statement from the German philosopher Kant,
shallow explanations without depth are empty, deep explanations without
mechanismsare blind. The complementarityof "deep" and "shallow" ex-
planationsechoesthe call for "layered" structural,institutional,and interac-
tional levelsof analysisproposed by Ekiert and Hanson in the introduction
to this volume. Let me neverthelessthrow some cold water on this happy
"peace formula" in the battle between structuralist and process-orientedan-
alysts of regime change. Where deeper structural causes(such as diversity of
communist regimes.predicatedon varianceamong precommunist interwar
regimes)are highly collinear with proximate causes(such as the outcome
Postcommunist Regim e D iuers ity 75
of founding elections,or the power of presidenciesin postcommunist con-
stitutions) of the ultimate outcomes (regime form, economic reform effort),
then at least two different conceptualizations of the relationship between
deeperand shallower causesare empirically equally plausible. Either the
deeper causesx "work through" the shallower cause y to bring about the
final outcome z (x -+ y -->z); or the causesx bring about both what appears
as the shallowercausey as well as the outcome z (x -->y; x -+ z). Becauseof
collinearity, we cannot statistically distinguish between these alternatives.
Our only way to conduct at leasta plausibility checkof which alternative
is more reasonableis to conduct case-orientedprocesstracing of outliers
that conform either to the structural or to the process-orientedexplanation
'We
or to neither of them. are interested in two temporal perspectiveswhen
dealing with outliers. Retrospectively,what causesoutliers not to conform
to the expected patterns on the deeper structural or the shallower process-
oriented variables? Prospectively,do we detect changesthat make outliers
gravitate toward the expected patterns on one or both of them? In other
words, is outlier status a transitional nonequilibrium state or permanently
"locked into" a new institutional and political-economiccompact?
Third, one can argue that sbort-term factors serue As the ultimate causes
, of outcomes.Consider outlier outcomes that conform to shallower, process-
oriented explanations but not to deeper structural explanations. Particularly
if this configuration persistsover time and there is no "equilibrium" process
: that makesoutcomes gravitate toward the result expectedbasedon structural
i lrredictions,then short-term causal arguments are key to bringing about the
i ultimate result of what appear to be outliers in the structural perspective.In
I the face of otherwise high collinearity between deeper and shallower causes
I of someultimate outcome,such outlier constellationsgenerateat leastsome
plausibility for the proposition that deeper causesaffect outcomes only by
"working through" shallowercauses.
, iNow consider other constellationsand processes.An outcome initially
conforms to predictionsbasedon shallow causation,but over time changes
suchthat it is in line with oredictionsbasedon structural causationas well.
r initially an outcome is at variance with both structural and process-
based forces, but over time gravitates to what underlying structure would
ict. In thoseinstances,suchoutliers suggestthat the collinearitybetween
i.t<'andy vis-i-vis z really involves a long-run equilibrium with a structural
determination of both y and z through x.
Fourth, one can argue that someutsesarepure outliers;neithershort-runnor
factors haue explanatorypower. Some outliers may always be out-
iers,regardlessof whatever systematicshallow or deep causal analysiswe
lnay explore. There man however, be structural reasonsfor randomness.As
i and Limongi GgSll point out, in middle-incomecountriespo-
regimescan be quite volatile,tipped by minor disturbances.Structural
76 Herbert Kitscbelt
background conditions tend to lock them in lessfirmly than those that char-
acterizevery poor or very affluent polities.
To move beyond this rather abstractdiscussion,let us examinea few out-
liers and their dynamics in studies by Fish (ryg8b) and papers I recently
(co)authored(Kitschelt and MaleskS zooo; Kitschelt, zoor). Fish (r998a:
z4z) arguesthat economic reform has a "possible causal effect" on democ-
racy, while in another piece he advancesthe proposition that the victory of
democraticforcesin initial electionsfurtherseconomicreform (Fish, rqg8b).
In that piecethere are no substantialoutliers, so it is impossibleto conduct
an analysisof divergentcases.'eTaking the two piecestogether shows how
ambiguouscausalattributions are when the temporal priority of the forces
to which causality is attributed is so tenuous. This critique, however, under-
lines only what I have already said about shallow explanationsand is not
my main point here. I rather focus on outliers in the analysis of economic
reform as predictor of democratization.
rVhen Fish employs economic reform as predictor of the democratic
quality of postcommunist regimes, major outliers are Albania, Croatia,
Kyrgyzstan, and Slovakia, all of which should have exhibited more polit-
ical democracy in 1996, if economic reform was the critical driving force.
Converseln Slovenia displayed too little reform, given its level of democracy
(Fish,r998a: tablesand figureson pp. 2r7 ) zz5, and zz7). Becausestructural
theoriesusually also predict a high correlation of political democratization
and market liberalization,they cannot account for theseanomalies.We can,
however, examine the adjustment processoutlier countries have undertaken
over time.
In Kyrgyzstan and Albania, since ry96 the outlier status has becomemore
pronounced. Thesedevelopmentsare consistentwith a structural theory that
identifies citizens' skills and resourcesas unconducive to democratic stabi-
lization. Here, democraticcivil and political rights have eroded,while eco-
nomic reform has made only small advances.'oIn a third country, Slovenia,
increasedeconomic reform caught up with the level of democratization al-
ready reached by the mid-r99os. In all three processes,either the causal
relationship is the inverse of what Fish claims (dernocracyinfluenceseco-
nomic reform), or there are underlying structural factors that, over'While
time,
generate a compelling equilibrium between politics and economics.
these casestend to be inconsistent with Fish's theoretical logic, two others
'e The only countries marginally approaching outlier status are Albania (too Iitde initial democ-
ratization for its level of economic reform) and Armenia (a great deal of democrac6 but linle
reform). In both cases,there may be measurementproblems that create the anomalies. Com-
pared with other measuresof economic reform effort in Albania (cf. Kitschelt and Malesky,
zooo: table A-r), Fish's measure is on the high side. In Armenia, there may have been less
initial decoupling from communist elites in the first election than Fish's score implies.
'o For change rates of economic tefornl berween r994 and a999, seeKitschelt and Malesky
(zooo: table r).
Reginte Diuersity 77
:m his argument through process analysis. The outliers Croatia and
kia have recently become more democratic, in line with hieh revelsof
ebonomicreform achievedearlier." Nevertheless,only a minority of outliers
show dynamic processesconsistentwith Fish'scausal-temporalargument.
; Now let us see how dynamic processesof economic ieform ieflect o.
proximate and structural causes(Kitschelt and Malesky, zooo). Regardless
i of whether one selectsgeography (proximity to the vest), religion, mode
of
co.mmunlstrule, and rnterwar regimes,or evencloselyrelatedproximate cor-
felates (pervasivenessof corruption, displacementof .o**u,iirt, in the first
election), measures of structural divergence among postcommunist coun-
tries are quite strongly related to institutional arrangiments (powers of the
presidency)in these countries in ry94 and very closely related to the
same
measurein t999." Both, in turn, are excellentpredictorsof postcommunist
economic reform effort, regardlessof the point in time or. .hoor.r, at least
for the entire universe of postcommunisr countries (Kitschelt and Maleskv.
zooo: table z). can we infer from these correlations among structural anJ
institutional predictors of reform that structural conditions "-ork throueh"
; proximate executive-legislativearrangementsto achievethe economi. poli.y
il outputs? In other words, is high presidential power the proximate cause
of
weak economic reform effort?
If strucfural and proximate explanations of economic reform were com-
plementary, we should observe that outliers with structural features con-
ducive to economic reform (closenessto the west, western christianit*
bureaucratic-authoritarian or national-accommodationist .om-u.rirt r.-
gime, low levelsof corruprion), but comparativelystrong presidentialpow-
1ibrs, should display weah economic reform effort. conversely, we should
i dbservepolitieswith inauspicioussrrucruralbackgroundcondiiions and par-
liamentary government, but strong economic reform effort.
, r only one of the six postcommunistoutlier countriesthat do not contrib,re
to the strong correlation between structural conducivenessand institutional
arranBements,however, has economic reform efforts that correspond to a
pattern suggesting that institutions are the proximate cause of economic
reform efforts. Three countries with good structural conditions had com-
paratively strong presidential powers from r99o on, but they nevertheless
engaged in vigorous economic reform efforts from the very start. Here,
clearly, structure trumps institutions (croatia, Lithuania, poiand). Not by
chance,politicians in at least two of these reform-orientedcountries have
l
aI of course,
Kopstein and Reilly's argumenr (chapter 4 in this volunre) that leaders and
' citizensadjust both political structures and economic reform
strategiesto the expectation 6f
i (rJTest) European integration provides a possible alternative mechanism for the equjlibration
betweenpolitical and economic reform. Furthermore, rhe change in Croatia's anj Slovakia,s
levelsof democratization is broadly consistent with a structuralist argument as well.
" " B a s e d o n t h e o p e r a t i o n aml e a s u r e s otfh e s e c o n c e p t s i n K i t s c h e l r a n d M a l e s k y ( z o o o ) , t h e s e
c o r r e l a t i o n sa r e + . 6 6 i n t 9 9 4 a n d + . 7 9 i n r y 9 9 @ f . p . z l a n d t a b l e s r B a n d A _ r ) .
 ,/
.'' z8 Herbert Kitschelt
attempted to weaken presidential powers basedon the legislativeexperience
of the first decade.
Two further countries had weak presidenciesup to 1994 and underly-
ing structuresnot conducive to economic reform (Belarus,Ukraine). Con-
trary to the institutional hypothesis, but in conformity with the structural
hypothesis, these configurations yielded weak economic reform efforts. In
both countries,the introduction of strongerpresidenciessince r994 has not
changedthat outcome. Only in the sixth case,Russia,do changingrates of
economic reform between r994 and r9g9 appear to reflect institutional ef-
ficacy.Until r993, with a formally weak presidencythat was enhancedby
presidential emergency powers granted after the falled ry9r coup, Russia
engagedin rather intensive economic reform efforts, given its comparatively
unfavorable structural circumstances.Once Russia empowered the presi-
dency through the r993 constitution, however, its economic reform effort
languishedthroughout the subsequentsevenyears.Empirical support by only
one of six casesis not reassuringfor the institutional thesis that weak presi-
denciesfurther economicreform. Reviewingeighteencaseswith collinearity
betweenstructural conditions and institutional arrangementsand six out-
liers with separableeffectsof thesevariableson economic reform provides
preciously little evidencethat economic reforms have to "work through"
institutionai arrangementsas the proximate cause.
The examples taken from Fish's and my own work are meant to cast
doubt on epistemologicalprescriptions that demand very tight spatiotempo-
ral proximity of causesand consequences. Of course' my defenseof "struc-
ruralism" in theseinstancesdoesnot suggestthat political actors and strategic
. action play no role. However, what affects deliberate, calculated political
^1Y actionworks often through longer chains of causaldetermination than short-
fi 1y'" rerm mechanisms.Nevertheless,my analysisdoes not suggesta historical de-
V terminism that puts everythinginto structural conditions.There are several
/O limitations to structuralarguments,evenwhen accompaniedby a micrologic
'\\
of action that makes strategicchoicesintelligible.First of all, there is the el-
\
.-.nt of uncertainty in the crisis of a political regime. Depending on the
personalitieswho are at the right place at the right time, whole polities
rnay take a "leap into the dark" for which no systematictheory, whether
building on structural or proximate causes'can account. For example, at
least one country in the postcommunist universe is recalcitrant to whatever
structural and institutional theories predict about economic reform efforts -
Kyrgyzstan.Despiteinauspiciousconditionsand a strong presidentialistcon-
stitution, Kyrgyzstan has engagedin quite vigorous reform efforts.
Second,actors in new political regimesundergo rapid learningprocesses,
triggered by the successor failure of initially chosen strategies.Policy feed-
backsbecomethe causeof new initiatives(Pierson,r991).Learningmay yield
resultsnot predicted basedon structural background conditions or proxi-
mate factors.An important mechanismin postcommunistdemocracieshere
communist Regime D iuersity 79
trospective economic voting in the face of manifest failure of economic
olicy strategies.Thus, the failed social protectionist economic policies of
B u l g a r i a n ,M o l d o v a n , a n d R o m a n i a n c o m m u n i s t s u c c e s s opr a r t i e si n
b mid-r99os eventuallyled to the electoralvictory of market-liberalizing
rties promoting more-vigorousreform. In other instances,where voters
iilterpretedunsuccessful
effortsto bring abouteconomicreform asthe cause
6f theirsocioeconomic electedon socialpro-
misery,victoriouschallengers,
pctionist tickets, then changed their tune and actually promoted reform
'(i.g.,
in the Ukraine). Even without electoral politics, under authoritari-
dnism rulers may learn from economic policy failure and try new strate-
t$ies(cf. Ames, ry82).In this vein, authoritarian leadersin Kazakhstanand
Kyrgyzstan have engagedin rather vigorous reform efforts in the r99os that
cannot be coveredby other explanations.
irii: Finalln external "shocks" that no systematicdomestictheory of economic
feform effort incorporates affect the trajectory of postcommunist countries.
i:Thus,the generaltrend toward rnarket liberalism in the global systemhas
subjectedall postcommunistcountriesto pressuresto accommodate,though
4t different speeds.Moreover, economic reform efforts receivea boost from
the end of civil and international wars. The sharp leaps in economic reform
effort in Armenia, Georgia, Moldova, and Tajikistan since t994 certainly
$uggesta linkage between pacification of a country and economic reform.
,[: The relevanceof innovative learning and policy feedback,conjunctural
luncertainty,and exogenousshocks shows the limits of systematic,causally
drientedsocialsciencemore generally.It underlinesthe impossibilityof crisp
point predictionsfor individual cases,regardlessof whether structural con-
'ditions or precipitating factors are the analytical focus. Sfith regard to the
cohort of postcommunist countries, this indeterminacy is structurally en-
hancedby the very fact that they are middle-incomecountriesthat, as a co-
[ort, display very high levelsof regime volatility (Przeworskiand Linrongi,
tssz).
,:t,
of Causal Explanation and Social ScienceParadigms
[n this final section,let me reject a stereotypicalrendering of principles of the-
constructionand explanation often encounteredin epistemologicaland
lftetatheoreticaldebatesin comparative politics. According to this stereo-
ftbpe, th" particular model of social actor and choice,the explanatory depth
i$f the theorS and the relianceon causalmechanismsare necessarilylinked to
babhother.Those who rely on an instrumentalrational choiceconceptionof
human action, centeringon actors' pursuit of "interests" in fungible private
$oods (wealth and power), also emphasizeproximate causal mechanisms
dnd rely on "shallow" explanationsaccordingto the stricturesof Ockham's
lazor. ConverselSsupporters of "deep" explanationsallegedlyemphasize
culture and discourseand thereforerelv on a lessinstrumentalconceDtionof
 Herbert Kitscheh Regime Diuersily 8r
8o

for empirical social outcomes. The ontological criteria I support

human action concerned with collective identities and idealistic preferences
(solidarity,salvation,beauty).They are said to shun the elaborationof causal do not necessarilyimply an affinity to what is now called "historical"
mechanisms. ).institutionalism (cf. Thelen and Steinmo, r99z). The latter's distinc-
claim is that institutions shape actors' preferences(March and Olsen,
A closer examination of the substantivecornparativepolitics literature,
but see Thelen, ry99) and that historical processeshave contingent
however,revealsno a priori associationbetweenconceptionsof human ac-
tcomes (Immergut, 1998).With regard to preferenceformation, I have not
tion, causaldepth, and the relianceon causalmechanisms.The electiveaffini-
a singleavowedlyhistoricalinstitutionalistaccountthat would explain
ties constructed between rational-interest-driven, shallow, but mechanism- ' preferences
rather than merely their strategies,constrainedby insti-
endowed explanations,on the one hand, and cultural, normative)and deep
tions and cognitive frameworks. Concerning the historical contingency of
explanationswithout mechanisms,on the other,are misleading.Someof the
ial phenomena,every comparativistrecognizesthe stochasticnature and
shallowestaccountsof socialand political action are cultural, particularly in
ry of social processes.But focusing on the random component of
the currently popular streamof studiesabout discourseformation and fram-
$trchprocessesgives up the quest for causal explanations that imply some
ing. They offer mechanisms without causal depth.'; I see a similar danger
in much of what is now advertisedas a "historical institutionalist" expla- fdlianceon causalmechanismspertaining to a mulritude of cases.Only gen-
iral causal propositions, applying to an indefinite number of cases,are em-
nation of political processes.Such undertakingsoften do not move beyond
irically testable. A historical institutionalism that focuseson idiosyncratic
the thick descriptionof historical processes.'4Conversely,as Thelen (tgSS\
vidual eventsand unrepeatableprocessesis empirically irrefutable.
correctly points out, many explanatory accountsthat rely on rational actor
Finally, the recognition that structuralist and actor-oriented, voluntarist
calculationsare averseneither to spatiotemporallyextendedchainswithout
roaches, those with long and short causal chains, are often mutually
or with "long distance" causal mechanismsnor to cultural analysis'Long
tary should not lead to the searchfor some grand "synthesis."
causalchains,for example,play a role in DouglassNorth's (r98r) rational
XVhat is advertised as such is usually not much more than a taxonomic
choice analysisof why England dominated the early stagesof the Industrial
ition of the different frameworks (e.g.,Snyderand Mahoney, 1999). For
Revolution, while other European powers, such as Spain or France,fell be-
tive theory building and empirical analysis,it is in fact more fruitful
hind. Rogowski's (r989) influential study of trade-relatedpolitical cleavage
'not to emphasize synergisms but to take each mode of explanation and
patternslacksmicrofoundations(cf. Alt et al., 1996).More recentlSa whole
itheory building in its purity and push it as far as possible.Hence I agreewith
host of rational choice theoristshas realizedthe importance of actors' cog-
nitive frameworks and cultural orientationsin accountingfor their strategic I.ichbach's\1998l.4or) conclusion,derivedfrom an analyticalreconstruction
6f rival collective action theories, that "we need creative confronrations,
choicesin light of instrumental self-interest(cf. Denzau and North, r994;
;lvhich should include well-defined combinations rather than grand syntheses,
Greif, ry94; Bates, Figueiredo, and rffeingast, 19981.
rationalist and structuralistapproachesto contentiouspolitics."
In this sense,the benchmarkthat good explanationsshould involve causal
mechanisms,but also causal depth, rules out neither rational choice nor
cultural (cognitive,normative) mechanisms.These ontological requisitesof -Conclusion
causal analysisdo not prejudicethe nature of the substantivetheoriesthat
So what "counts" as a good causein explanationsof postcommunistpoliti-
lial regime diversity? I have suggestedthat to answer ihis question is to em-
tl An example is discursive franre analysis in studies of social movements. See Snow et al.
.bark on an ontological and transscientificenterprisemore so than a narrowly
(r986) and the subsequentsizable literature derived from this paradigm.
'4 Metatheoretical works on the new historical institutionalism reveal and often even recognize rhethodologicaland empirical one. Different scientificcommunities may rea-
t h i s d a n g e r .S e eI m m e r g u t ( r 9 9 8 ) , S o m e r s( r q g 8 ) , a n d T h e l e n ( r 9 9 9 ) ' I n a s m u c ha s h i s t o r i c a l iSonablydisagreeon the appropriate answer to the challengeof causality.A
institutionalism focuseson conGgurative,conjunctural interactions of causalchains and their itreatisethat tries to seDarateeood from bad causalanalysisinvolvesas much
unique temporal sequencing,it postulates an unpredictability of collective outcomes, It thus
iersuasion about what should be important in social scientificanalysisas
concentrates on what cannot be causally explained either in terms of actors' preexisting straightforward logical inference from patterns of empirical evidence.
cognitive and cultural frames or strategiesresulting from instrumental interestsconstrained
-carce resources,institutions, and rival players. The danger is that this institutionalism I This chapterhas arguedfor a deepversionof causalanalysis,yet one that
- i v e su p o n e x p la n a t i o n .F o r g o o d r e as o n ,I m m e r g u t ( r 9 9 8 : z 7 ) w o r r i e s t h a t h i s t o r i c aI doesnot lose sight of socialmechanisrns. The latter show how human beings
rist accounts lack falsifiability and therefore cannot promote alternative theories with deliberativefaculties and capacitiesto choose objectivescan act on
r by principles of structuralist or rational choice institutionalism. Immergttt constraintsand opportunities.It is human action that brings about collective
r s e q x e n c e ." i n e s c h e w i n gs 1 ' s t e m a t i z a t i o n
t h, e h i s t o r i c a li n s t i t u t i o n a l i s t s
outcomes,even though broader socioeconomic,institutional, and cognitive
impact of their work."