John Gerring - Cuali

8/3/2019 John Gerring - Cuali

1/15

What Is a Case Study and What Is It Good for?

Author(s): John GerringSource: The American Political Science Review, Vol. 98, No. 2 (May, 2004), pp. 341-354Published by: American Political Science AssociationStable URL: http://www.jstor.org/stable/4145316 .

Accessed: 26/10/2011 17:11

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

American Political Science Association is collaborating with JSTOR to digitize, preserve and extend access to

The American Political Science Review.

http://www.jstor.org
http://www.jstor.org/action/showPublisher?publisherCode=apsahttp://www.jstor.org/stable/4145316?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/4145316?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=apsa


2/15

AmericanPoliticalScience Review Vol. 98, No. 2 May2004

W h a t I s C a s e S t u d y a n d W h a t I s G o o d f o r ?JOHN GERRING Boston Universityhis paper aims to clarify the meaning, and explain the utility, of the case study method, a methodoften practiced but little understood. A "casestudy,"I argue, is best defined as an intensivestudyof a single unit with an aim to generalize across a larger set of units. Case studies rely on thesame sort of covariational evidence utilized in non-case study research. Thus, the case study method iscorrectly understood as a particular way of defining cases, not a way of analyzing cases or a way ofmodeling causal relations. I show thatthisunderstandingof thesubjectilluminates some of thepersistentambiguitiesof case study work,ambiguitiesthatare, tosome extent,intrinsicto theenterprise.The travailsof the case study within the discipline of political science are also rooted in an insufficientappreciation ofthemethodological tradeoffs thatthismethodcallsforth. Thispaperpresentsthe amiliar contrast betweencase study and non-case study work as a series of characteristicstrengthsand weaknesses-affinities-ratherthanas antagonisticapproaches to the empirical world. In the end, theperceived hostility betweencase study and non-case study researchis largely unjustifiedand, perhaps, deserves to be regardedas amisconception. Indeed, thestrongestconclusion to arisefrom thismethodological examination concernsthe complementarityof single-unit and cross-unit researchdesigns.

he case studyoccupiesa vexed positionin thedisciplineof politicalscience.On the one hand,methodologistsgenerallyview the case studymethod with extreme circumspection(Achen andSnidal1989;King,Keohane,andVerba1994;Lieberson[1991]1992, 1994;Njolstad1990).A workthatfocusesits attentionon a single exampleof a broaderphe-nomenon s aptto be describedas a "mere" asestudy.At the same time, the disciplinecontinues to pro-duce a vast numberof case studies,many of whichhaveentered hepantheonofclassicworks Allen1965;Allison1971;Dahl1960;Johnson1983;Kaufman1960;Lazarsfeld,Berelson,andGaudet1948;Lijphart 968;Pressmanand Wildavsky1973). Judgingby recentscholarlyoutput,the case studymethodretainscon-siderable appeal, even among scholars in researchcommunities not traditionallyassociated with thisstyleof research-e.g., amongpoliticaleconomistsandquantitativelynclinedpoliticalscientists(Acemoglu,Johnson,andRobinson2003;Bates et al. 1998;Rodrik2003). By the standardof praxis,therefore, t wouldappear hat the methodof the casestudy s solidlyen-sconcedand,perhaps, ven thriving.Thus,a paradox:Althoughmuch of whatwe knowabout the empiricalworld is drawn rom case studiesandcasestudiescontinueto constitutea largepropor-

tion of work generated by the discipline, the case studymethod is held in low regardor is simply ignored. Evenamong its defenders there is confusion over the virtuesand vices of this ambiguous research design. Practi-tioners continue to ply their trade but have difficultyarticulatingwhat it is that they are doing, methodolog-ically speaking. The case study survives in a curiousmethodological limbo.How can we understand the profound disjuncturethat exists between the case study's acknowledged con-tributions to political science and its maligned statuswithin the discipline? If case studies are methodologi-cally flawed, why do they persist?The paper is divided into two parts.The firstpart fo-cuses on matters of definition. I argue that for method-ological purposes a case study is best defined as anin-depth study of a single unit (a relatively boundedphenomenon) where the scholar's aim is to elucidatefeatures of a larger class of similar phenomena. It isdemonstrated that case studies rely on the same sortof covariational evidence utilized in non-case study re-search. Thus, the case study method is correctly un-derstood as a particular way of defining cases, not away of analyzing cases or a way of modeling causalrelations. I show, finally,that this understanding of thesubject illuminates some of the persistent ambiguitiesof case studywork, ambiguities that are,to some extent,intrinsic to the enterprise.In the second part of the paper I proceed to examinethe contrast between case study and non-case studywork. The central argument here is that the differencesbetween these two genres are best understood as char-acteristic strengths and weaknesses-affinities-ratherthan antagonistic approaches to the empirical world.Tradeoffs,rather thandichotomies, characterize the on-going case study/non-case study debate.

WHAT S A CASESTUDY?What is a case study, and how is it differentiated fromother styles of research? Regretfully, the term "case

John Gerring is Associate Professor, Department of Political Sci-ence, Boston University, 232 Bay State Road, Boston, MA 02215([email protected]).Helpful comments were received from Robert Adcock, AndrewBennett, Henry Brady, Bear Braumoeller, David Collier, MichaelCoppedge, Colin Elman, Peter Hall, Alan Jacobs,Evan Lieberman,Jim Mahoney, Jason Seawright, David Waldner, and three anony-mous reviewers for the journal. William Barndt and JoshuaYesnowitzprovidedcrucialassistance at the finalstages of manuscriptpreparation. George and Bennett 2004 reached me in manuscriptform as I was developing the ideas for this paperand hada substantialimpact on my own understandingof the case study-although somedifferences of opinion remain. Funding for this research was gener-ously provided by the Frederick S.Pardee Center for the Study of theLonger-Range Future. Work was completed while the author wasin residence at the School of Social Science, Institute for AdvancedStudy.

341


3/15

What Is a Case Study? May2004

study" is a definitional morass. To refer to a work as acase study might mean (a) that its method is qualita-tive, small-N (Yin 1994); (b) that the research is ethno-graphic, clinical, participant-observation, or otherwise"inthe field" (Yin 1994); (c) that the research is charac-terized by process-tracing (George and Bennett 2004);(d) that the researchinvestigates the properties of asin-gle case (Campbell and Stanley 1963,7; Eckstein [1975]1992);or (e) that the research investigates a single phe-nomenon, instance, or example (the most common us-age). Evidently, researchers have many things in mindwhen they talk about case studyresearch.1As a result ofthis profusion of meanings, proponents and opponentsof the case study marshal a wide range of argumentsbut do not seem any closer to agreement than whenthis debate was first broached several decades ago.How, then, should the case study be understood? Thefirst three options enumerated above (a-c) seem inap-propriate as general definitions of the topic since eachimplies a substantial shift in meaning relative to estab-lished usage. One cannot substitute case study for qual-itative,ethnographic,orprocess-tracingwithout feelingthat something has been lost in translation. These defi-nitions are best understood as describing certain kinds(subtypes) of case studies, rather than the general phe-nomenon itself. The fourth option (d) equates the casestudy with the study of a single case, the N = 1 researchdesign. This is simply wrong, as argued at length below;case studies always employ more than one case. Thefifth option (e), centering on phenomenon, instance, orexample as the key term, is correct as far as it goesbut also ambiguous. Imagine asking someone, "Whatis your instance?" or "What is your phenomenon?" Acase studypresupposes a relatively bounded phenome-non, an implication that none of these terms captures.As a substitute for these flawed definitions, I proposeto define the case study as an intensive study of a sin-gle unitfor thepurpose of understandinga larger classof (similar) units. A unit connotes a spatially boundedphenomenon-e.g., a nation-state, revolution, politicalparty,election, or person-observed at a single point intime or over some delimited period of time. (Althoughthe temporal boundaries of a unit are not always ex-plicit, they are at least implicit.)2To clarify this definition we must establish the rela-tionship of the case study,so defined, to other terms inthis crowded semantic field. Following is a set of nesteddefinitions, which should be read carefully. A "pop-ulation" is comprised of a "sample" (studied cases),as well as unstudied cases. A sample is comprised ofseveral "units,"and each unit is observed at discretepoints in time, comprising "cases." A case is comprisedof several relevant dimensions ("variables"), each ofwhich is built upon an "observation" or observations.

For those familiar with the rectangularform of a datasetit may be helpful to conceptualize observations as cells,variables as columns, cases as rows, and units as eithergroups of cases or individual cases (depending uponthe proposition and the analysis).The most important point is that all these terms aredefinable only by reference to a particular propositionand a corresponding research design. A country mayfunction as a case, a unit, a population, or a case study.It all depends upon what one is arguing. In a typi-cal cross-country time-series regression analysis (e.g.,Przeworski et al. 2000), units are countries, cases arecountry-years, and observations are collected for eachcase on a range of variables. However, shifts in theunit of analysis of a proposition change the referen-tial meaning of all terms in the semantic field. If onemoves down one level of analysis the new populationlies within the old population, the new sample withinthe old sample, and so forth, such that an observationin the original proposition now becomes a case. Pop-ulation, unit, case, and observation are nested withineach other. Since most social science research occursat several levels of analysis these terms are generally influx. Nonetheless, they have distinct meanings withinthe context of a single proposition, which defines theprincipal unit of analysis.I do not issue this somewhat novel definition of casestudy (an intensive study of asingle unit for the purposeof understanding a largerclass of units) with any hopesof displacingcommon usage. Indeed, there is no harmincontinuing to refer to a case study in the various wayslisted above (options a-e). What is important is thatwe have recourse to a narrower and clearer definitionwhen methodological confusions arise so that we have away to arbitratesuch confusions. The definition chosenhere is useful in this regard. Moreover, it captures theessential features of other extant definitions; it is reso-nant (Gerring 2001, chap. 3). Finally,as the succeedingportions of this paper show, it clarifies the distinctivefeatures of a broad class of work in the discipline ofpolitical science and in neighboring fields of the socialsciences. It is theoretically useful.

TheCase StudyMethodConsideredas an EmpiricalEndeavorThe distinctiveness of the case study is most clearly un-derstood when placed within a broader set of method-ological options. To understand what a case study is,one must comprehend what it is not.All empirical evidence of causal relationships is co-variational n nature. A purportedcause and effect mustbe found to covary. They must appear and disappear,wax and wane, or perform some other transformationin tandem or at some regular,more or less predictable,intervals. Even where this covariation is imagined, asin a counterfactual thought experiment, the evidencewe imagine is of a covariational sort. Conversely, theabsence of such covariation is taken as disconfirm-ing evidence. If the appearance and disappearance(waxing/waning et al.) of X and Y are not associated

1 In addition to sources cited above, see Brady and Collier 2004,Campbell (1975) 1988, Davidson and Costello 1969, Feagin, Orum,and Sjoberg 1991, George 1979, McKeown 1999, Ragin 1987, 1997,Ragin and Becker 1992, and the symposium, "The Case StudyMethod in Sociology," in Current Sociology, Volume 40, Number1 (Spring 1992).2 Similarunderstandings of the term "unit"can be found elsewhere(e.g., King, Keohane, and Verba 1994, 76-77).

342


4/15

AmericanPoliticalScience Review Vol. 98, No. 2

TABLE . ResearchDesigns:ACovariational ypology TemporalariationNo YesNone 1 unit) [Logicallympossible] (a)CasestudyWithin-unit (b)Casestudy I (c)Casestudy IISpatial Variation Across-unit (d)Cross-sectional (e)Time-seriesross-sectionalAcross- ndwithin-unit f)Hierarchical (g)Hierarchicalime-series; omparative-historical

in any waythat can be rationally xplained,andhencepredicted or postdicted), henthe empirical videncesuggests hata causalrelationshipdoes not exist.3Thisprovidesa useful way of typologizingvariousresearchdesigns.Covariationmaybe observed(a) ina single unit diachronically,b) within a single unitsynchronically,c) within a singleunit diachronically,(d) acrossunits synchronically, e) across units syn-chronicallyand diachronically,f) across and withinunitssynchronically,r (g) acrossandwithinunitssyn-chronically nddiachronically,sdepicted n Table 1.It will be seen that the case studyoccupiesone ofthreepossiblecells.TypeI case studiesexamine vari-ation in a single unit over time, thus preserving heprimary nitofanalysis.Othercase studiesbreakdownthis primaryunit into subunits,which are then sub-jectedto covariational nalysis-either synchronically(typeII)orsynchronicallynddiachronicallytypeIII).These are the three logicallyconceivableapproachesto the intensivestudyof asingleunitwherethat unit isviewed as an instanceof some broaderphenomenon.Consequently, henone refers o thecasestudymethodone is in factreferring o threepossiblemethods,eachwith a differentmenu of covariational vidence.The bottom halfof Table1 laysout variousacross-unit researchdesigns (wheresome important lementof the empiricalanalysis nvolvescomparisonsacrossunits).Here I have listedthe methodsmostcommonlyidentifiedwith hese researchdesigns.Across-unit nal-ysis without any explicit temporalcomponent(d) isusuallyclassifiedas "cross-sectional"even thoughatemporal omponentsusually imulatedwith ndepen-dentvariables hat are assumed o precedethe depen-dent variableunderinvestigation).When a temporalcomponent s includedwe oftenreferto the analysisas"time-seriescross-sectional"TSCS)or pooled time-series (e). When one examinesvariationacross- andwithinunits in the same researchdesignone is said tobe employinga "hierarchical" odel(f). Finally,whenall formsof covariation re enlisted n asingleresearchdesign the resultingmethod is described as "hierar-chicaltime-series"(if quantitative)or "comparative-historical"ifqualitative) g).Of allcross-unit esearch

designsthe casestudy s probablyclosestto the latter,where evels of analysismoveupanddownmoreorlesssimultaneously ndwhere a small numberof unitsaresubjected o intensivestudy. ndeed,the comparative-historicalstudy may be looked upon as a series ofcase studies combinedwithexplicitcross-unitanalysis(MahoneyandRueschemeyer2003).Havingplaced hesestandard ross-unit esearchde-signswithina covariationalypologyonemustalsotakenote that each of these methods might also be em-ployedas a casestudy.A casestudymay employcross-sectional, TSCS,hierarchical, ierarchicalime-series,andperhapseven comparative-historical odels.It alldependsuponthepropositionn question.Specifically,it is the purposesto whichthese analysesareput,andhence the definitionof a unit,that determineswhetheror nottheyareappropriatelyeferred o ascase studies.Thiswillbecome cleareras we proceed.The N QuestionI have arguedthat what distinguishes he case studymethod romall othermethods s itsrelianceon covari-ation demonstrated y a singleunitandits attempt,atthesametime,to illuminateeaturesof abroader etofunits.It follows fromthisthat the niumber f cases(N)employedby a case studymaybe eithersmallor largeand,consequently,maybe evaluated n a qualitative rquantitativeashion.4To see whythismustbe so let usconsiderhow a casestudyof a singleevent-say, the FrenchRevolution-works.Intuitively, ucha study providesan N of one(France). fonewere tobroaden heanalysis o includea second revolution e.g.,theAmericanRevolution), twould be common o describe he studyas comprisingtwo cases.Yet, as I have arguedpreliminarily,his isa grossdistortionof whatis really goingon. It wouldbe more correct o describe ucha studyascomprisingtwo units,ratherthantwo cases,for a case studyof asingleevent generallyexaminesthat event over time.France s observedbefore,during,andafterthe eventto see whatchangedandwhat remained hesameafterthis cataclysmicevent. These patternsof covariationoffer theempirical luesoneneedstoreachconclusionsabout causation.Theyalsocreatemultiplecasesout ofthat ndividual nit.N= 2, at thevery east(e.g.,beforeand after a revolution), n a case studyof typeI.

3 Note that covariation (or correlation) refers to the mutual relation-ship between X and Y; variation, to the behavior of a single variable.These words are often used interchangeably. Hume's word for thiswas constant conjunction, and others have been employed as well. Ishould clarify thatalthoughthe empirical component of a causal argu-ment is covariational in nature, successful causal arguments dependupon more than just covariation. Among other things, a convincingcausal account must identify a causal mechanism (see below).

4 This section explains and elaborates on a theme first articulated byCampbell (1975) 1988, itself a revision of Campbell's earlier perspec-tive (Campbell and Stanley 1963).

343


5/15

WhatIs a Case Study? May2004

If, instead, there is no temporal variation-if, for ex-ample, the French Revolution is examined at a singlepoint in time-then the object of investigation will becovariational patterns within that unit, a case study oftype II. Within-unit cases consist of allcases that lie at alower level of analysisrelative to the inference under in-vestigation. If the primaryunit of analysis is the nation-state, then within-unit cases might be constructed fromprovinces, localities, groups, or individuals. The pos-sibilities for within-unit analysis are, in principle, in-finite. Indeed, within-unit N often swamps across-unitN, particularlywhere individualscomprise the relevantwithin-unitcase. A single national surveywill produce alargersample than any conceivable cross-country anal-ysis.Thus,in many circumstancescase studies of type IIcomprise a larger N than cross-sectional analyses orTSCS analyses.Evidently, if a case studycombines both temporal andwithin-unitvariation, as in case studies of type III, thenits potential N increases accordingly. This is probablythe most common genre of case study analysis.These covariational facts hold true regardless ofwhether the method is experimental or nonexperimen-tal. It is also true of counterfactual reasoning, whichtypically consists of four cases-the actual (as it hap-pened) before and after cases and the before and aftercases as reconstructed through counterfactual reason-ing (i.e., with an imagined intervention). In short, thecase studydoes not preclude high-N;it simplyprecludesacross-unit N (by definition).What, then, of the classic N= 1 research design,which haunts the imaginations of social scientists ev-erywhere? This hypothetical research design occupiesthe empty cell in Table 1. Its cell is empty because itrepresents a research design that is not logically fea-sible. A single unit observed at a single point in timewithout the addition of within-unit cases offers no evi-dence whatsoever of a causal proposition. In trying tointuit a causal relationship from this snapshot-a sin-gle case without within-unit covariation-we would beengaging in a truly random operation, since an infinitenumber of lines might be drawn through that one datapoint.

Ambiguities-Necessaryand UnnecessaryThe effort in this section has been to clarify what itmeans to conduct a case study.I have argued that a casestudy is most usefully defined as the intensive studyof a single unit wherever the aim is to shed light ona question pertaining to a broader class of units. Al-though this definitional exercise does not settle all theambiguities besetting the case study research design, itdoes provide a way of understanding ambiguities thatremain. Six issues deserve emphasis.The first ambiguity concerns the problem of distin-guishing different types of covariational evidence. Wehave pointed out that case studies may observe a sin-gle unit through time (type I), synchronic within-unitvariance (type II), or synchronic and diachronic within-unit variance (type III). Notice that types II and III, but

not type I, involve a change in level of analysis, sincecases are drawn from phenomena within the primaryunit (as defined by the proposition of interest). Thus,some case studies-but not all-involve a change in theprimaryunit of analysis.To complicate matters further,case studies often combine observations of the primaryunit over time (type I) with synchronic and diachronicobservations of within-unit covariation (types II andIII). Many case studies are thus hybrids of all threeresearch designs. A final complication is introduced bythe fact that it is often difficultto figureout which sortofcovariational evidence isbeing mobilized at aparticularjuncture.The difficultyowes something to the complex-ities of within-unit analysis.Although the primaryunitof analysis is usually clear, within-unit cases are oftenmultiple and ambiguous.A second source of ambiguity concerns the blurryline between a unit that is intensively studied-the casestudy-and other adjacent units that may be broughtinto the analysisin a less structuredmanner. Recall thatbecause a case study refers to a set of units broader thanthe one immediately under study, a writer must havesome knowledge of these additional units (a) to choosea unit for special treatment and (b) identify plausiblecausal hypotheses. Case studies are not immaculatelyconceived; additional units always loom in the back-ground.Tospeak of a case studyat all it is helpful to introducea distinction between formal and informal units. Theformal unit is the unit chosen for intensive analysis-the person, group, organization, county, region, coun-try,or other bounded phenomenon of which the writerhas in-depth knowledge. Informal units consist of allother units that are brought into the analysis in a pe-ripheral way,typically in an introductory or concludingchapter. Often, these informal units are studied onlythrough secondary literature;they are always more su-perficially surveyed than the formal unit under study.Sometimes, the status of informal units is left implicit.This may be warranted in circumstances where the rel-evant comparison or contrast between the formal unitand other units is obvious or generally accepted. In anycase, the distinction between a formal and an informalunit is always a matter of degrees. The more equality oftreatment granted to peripheral units, the more a studyleans toward a cross-unit style of analysis. The greaterthe predominance of a single unit, the more it meritsthe appellation case study.A third ambiguity occurs whenever a single workcombines single-unit and across-unit analysis in a for-mal manner. This would be true of comparative-historical work as well as any work in which an inten-sively studied unit is "nested" within a broader researchdesign (Coppedge 2002; Lieberman 2003). Indeed, theonly thing that distinguishes the single-unit study froma sample (which is of course also designed to elucidatethe features of some larger phenomenon) is that thelatter is generally understood as composed of morethan one unit. Case studies, like samples, seek to rep-resent, in all ways relevant to the proposition at hand,a population of cases. A series of case studies mighttherefore be referred to as a sample; it is a matter of

344


6/15

AmericanPoliticalScience Review Vol. 98, No. 2

emphasis and of degree. The more case studies onehas, the less intensively each one is studied, and themore confident one is in their representativeness (ofsome broader population), the more likely one is todescribe them as a sample rather than a series of casestudies.A fourth ambiguity afflictingcase studies is that suchworks generally partake of two empirical worlds. Theyare both studies toutcourt andcase studies of somethingmore general. As a study,the population is restricted tothe unit under investigation. As a case study, the pop-ulation includes adjacent units-perhaps quite a largenumber of them. This tension is evident in GrahamAllison's (1971) renowned work, whose subtitle, Ex-plaining the Cuban Missile Crisis, invokes a narrowtopic, whereas the title, Essence of Decision, suggestsa much larger topic (government decision-making).To complicate matters further, different propositionswithin the same work commonly applyto different pop-ulations.Some maybe restrictedto the unit understudy,whereas others have a wider ambit. This is clearly thecase in Allison's study and is noted explicitly in theintroduction.To complicate matters further, the status of a workmay change as it is digested and appropriatedby a com-munity of scholars. "Meta-analyses" are systematic at-tempts to integrate the results of individual studies intoa single quantitative analysis, pooling individual casesdrawn from each study into a single dataset (with vari-ous weightings and restrictions). The ubiquitous "liter-ature review" often aims at the same objective, albeitin a less synoptic way. Both statistical meta-analysesand narrative literature reviews assimilate a series ofstudies, treating each of them as case studies in somelarger project-whether or not this was the intentionof the original authors.A final ambiguity concerns the sort of argument thata case study is intended to prove or demonstrate. Onespecies of case study examines a loosely defined generaltopic-war, revolution, gender relations-in a particu-lar setting but offers no specific proposition that mightbe applied across a largerset of units.E. P.Thompson'sTheMaking of theEnglish WorkingClass (1963) is usu-ally construed as a case study of class formation. Thissuggests a very general purview, perhaps applicable toall countries in the modern era. Yet Thompson doesnot proffer a theory of class formation, aside from therather fuzzy notion of a working class participating inits own development. Thus, his work is probably cor-rectly understood as a study of how a more generalphenomenon occurred in one country setting. Virtu-ally any intensive study of a relatively bounded topicqualifies as a case study in this minimal sense, so long asit can be linked with some larger topic via a key word(e.g., class formation). Indeed, the narrowest terrainssometimes claim the broadest extensions. Studies of awar are studies of war, studies of a farming commu-nity are studies of farming communities everywhere,studies of individuals are studies of humanity, and soforth.

A very different style of argumentation informsBenjamin Reilly's (2001) study of the role of electoral

systems in ethnically divided societies. Reilly argues,onthe basis of several case studies,thatsingle-transferablevote (STV) electoral systems have a moderating effecton group conflict relative to first-past-the-post (FPP)electoral systems. Here is a good example of a casestudy that is more than simply suggestive (for otherexamples see Eaton 2003, Elman 1997, Lijphart 1968,and Stratmann and Baur 2002). For present purposes,what is significantis that both styles of argumentation-the suggestive and the falsifiable-are legitimately re-ferred to as case studies. Evidently, they have very dif-ferent methodological implications. But these implica-tions should not be confused with the case studyformat,which can be implemented in interpretivist as well aspositivist modes.Having flagged these six ambiguities of the casestudy,the question is begged: Are they necessary? Arethey intrinsic to the research design, or might they beavoided?In many instances, ambiguities can be removed sim-ply by more careful attention to the task of specifi-cation (Gerring 2001, 90-99). Writers should be clearabout which propositions are intended to describe theunit under study and which are intended to apply to abroader set of units. Regrettably, many studies focusedon some element of politics in the United States frametheir analysisas a study of politics-by implication, pol-itics in general (everywhere and always). One is left towonder whether the study pertains only to Americanpolitics, to all contemporary polities, or, in varying de-grees, to both. Indeed, the slippage between study andcase study accounts for much of the confusion that weencounter when reading single-unit analyses.To the ex-tent that propositions-and their attendant cases, units,and populations-are stated clearly and explicitly, theauthor avoids confusion and the work attains a higherdegree of falsifiability.This may involve some sacrificein narrative flow, but it is rightly regarded as the entryprice of social science.However, it hardly seems plausible that the six am-biguities noted above arise solely from the sloppy orunduly belletristic habits of case study practitioners.Indeed, a certain degree of ambiguity is inherentin theenterprise of the case study.This pertains, most of all,to the study/case study distinction.It would be difficult to write a study of a single unitthat does not also function as a case study, and viceversa, for reasons already explored. Indeed, it may bedifficult to neatly separate the study and case studycomponents of a work (e.g., into different chapters ordifferently labeled propositions). The reason for thisstructuralambiguity is that the utility of the single-unitstudy rests partly on its double functions. One wishesto know both what is particular to that unit and whatis general about it. It should be kept in mind that casestudies often tackle subjects about which little is pre-viously known or about which existing knowledge isfundamentally flawed. The case study typically presentsoriginal research of some sort. Indeed, it is the opportu-nity to study a single unit in great depth that constitutesone of the primary virtues of the case study method(see below). If a writer were to restrict herself only to

345


7/15


TABLE . Single-UnitVersusCross-UnitResearchDesigns:Tradeoffs ndAffinitiesAffinity

Case Study Cross-UnitStudy1. Typeof inference (a) Descriptive +(b) Causal +2. Scope of proposition (a) Depth +(b) Breadth +(c) Boundedness +3. Unithomogeneity (a) Case comparabilityinternal) +(b) Representativeness external) +4. Causalinsight (a) Causalmechanisms +(b) Causaleffect +5. Causalrelationship (a) Invariant +(b) Probabilistic +6. Strategyof research (a) Exploratorytheorygeneration) +(b) Confirmatorytheory esting) +7. Usefulvariance (a) Foronlya single unit +(b) Formanyunits +8. Ontology Indeterminate

elementsof theunitthatweregeneralizablei.e.,if sherigorouslymaintains he "casestudy"mode of anal-ysis), a readermightjustifiably omplain.Such rigorwould clarify the population of the primaryinfer-ence,butit wouldalsoconstitutea considerablewasteof scholarlyresources.Imaginea study of economicgrowth hatfocusesonMauritius s a casestudyyetre-fusesto engagecausalquestionsunlesstheyareclearlyapplicable o othercountries sincethisis a casestudyof a moregeneralphenomenon,growth).No mentionof factorsspecificto the Mauritian ase is allowable;all propernouns are convertedinto commonnouns(Przeworski ndTeune1970).Such a studyseems un-dulynarrow;ts conclusionsmaymislead.Indeed,it is often difficult o tellwhichof the manyfeaturesof a given unit are typicalof a largerset ofunits(and hence fodder for generalizablenferences)andwhichareparticularo theunitunder tudy.Theap-propriate esponse o suchambiguitysfor thewriter oreportallfactsandhypotheses hatmightberelevant-in short, o overreport.Muchof the detailprovidedbythe typicalcasestudymaybe regardedas "fieldnotes"of possibleutility orfutureresearchers-perhapswitha ratherdifferentset of inferences n mind.Again, itseemsjustifiableorcasestudiesto work on twolevelssimultaneously,he unit itselfandsome broaderclassof (perhapsdifficult o specify)units.As ageneralobservationwemightsaythatmethods,strictlydefined,tend to lose their shapeas one lookscloserat their nnards.A studymerges ntoacasestudy,a single-unit tudymergesinto a studyof a sample,alongitudinaltudymerges nto alatitudinaltudy,nfor-malcasesmerge ntoformalcases,and so forth.Meth-ods thatseemquitedissimilar n designbleedinto oneanotherwhenput into practice.There are few "pure"methods.And this is probablya good thing.Chastitysnotnecessarilyanattribute o be cherished nresearchdesign.

WHAT S A CASE STUDYGOODFOR?It hasbeen demonstrated hat the differencebetweena casestudyand a study(toutcourt) s rarelyclear-cut.Indeed,the case studyis probablybest understoodasan ideal-typerather han a methodwithhard-and-fastrules. Yet the fact that the case studyis fuzzyaroundthe edgesdoes not meanthat it is lacking n distinctivecharacteristics.When consideredas an ideal type thecasestudyresearchdesign, ike all researchdesigns, x-hibits characteristictrengthsandweaknessesrelativeto its across-unit ousin.These pertainto the type ofinferenceunder consideration descriptiveor causal),the scope of the proposition(its depth,breadth,andboundedness), he degreeof unit homogeneity oundamongcasesand betweenthe sampleand the popula-tion, the sort of causal insightdesired(causaleffector causal mechanism),the strategyof research(ex-ploratoryor confirmatory), nd the kindof empiricalevidence available.Tradeoffs longthese sevendimen-sions are summarizedn Table2.Ontologicalpresuppo-sitions are also importantbut of indeterminatemport(as indicated n Table2).It should be underlinedthat these tradeoffsrep-resent methodological affinities,not invariantlaws.Exceptionscanbe foundto eachof thegeneral enden-cies identifiedhere. Even so, the strengthsandweak-nesses often noted in case studyresearch,reproducedin many subfieldsand disciplinesover the course ofmanydecades,are not the productof a purelystochas-tic process. General patterns suggest general inter-pretations.I should also emphasizethat each of these trade-offs carriesa ceterisparibuscaveat.Case studiesaremoreuseful orformingdescriptivenferences, llotherthings beingequal.Since ceterisis not alwaysparibusthereader houldnotjump oanyconclusions bout heresearchdesignappropriateo a givensettingwithout

346


8/15

AmericanPolitical Science Review Vol. 98, No. 2

considering the single-unit/cross-unit options availablewithin that research context.Finally, readers should note that although many ofmy examples are drawn from the subfield of compara-tive politics, with nation-states as the principal unit ofconcern, these examples could be replicated with otherunits and in other research settings. The problem of thecase study is not limited to a single subfield.

Typeof Inference:DescriptiveVersusCausalDescriptive inference remains an important, if under-valued, trope within the social sciences (Gerring 2001,chap. 6;King, Keohane, and Verba 1994, chap. 2). Thus,it is not at all pejorative to observe that there is amethodological affinity between descriptive inferenceand case study work. When one is examining correla-tive relationships or proximate causal relationships thecase study format seems less problematic and is oftenhighly informative. Indeed, many of the most famouscase studies in anthropology, political science, and so-ciology are primarily descriptive in orientation (e.g.,Fenno 1978, Hartz 1955, Lynd and Lynd [1929] 1956,Malinowski [1922] 1984, and Whyte [1943] 1955). Howcan we understand this affinity?What?andHow? questions are easier to answer with-out recourse to cross-unitanalysisthan Why? questions.The simplest genre of descriptive case study asserts thatthe unit under study (A) is like, or unlike, other similarunits (B and C). A more complicated descriptive casestudy might assert a classificatory relationship amongA, B, and C, such that A falls into a certain typologicalrelationship with B and C. The latter, of course, is morecomplicated and is more likely to require some explicitcross-unit examination. However, a descriptive infer-ence does not make any assertions about causal rela-tionships (beyond the most proximal) occurring withinA, B, and C. In this sense, descriptive inference is sim-pler, methodologically speaking.To be sure, descriptive case study propositions areimplicitly comparative and these comparisons musthave a cross-unit reference point. To say "green" isto imply "not blue." However, it is usually fairly easyto make such comparisons without conducting a studyof the presumed variation. One knows what blue iswithout going in search of blue cases. This illustratessomething important about the structure of descrip-tive propositions in social science. They are held to-gether by language-by ordinary or technical termsand their definitions. When describing a phenomenonone is usually comparing it to an ideal-type definition.American political culture is "liberal" or "republican"insofar as it conforms to standard definitions of thesetwo concepts (Smith 1997). Todescribe is to categorize,and to categorize is to rely on language to divide upthe world into identifiable entities. Language, in thissense, provides the "laws" that allow for consistent in-terpretations of the phenomenal world (Sartori 1984).Chairs are different from tables in predictable ways;labeling entities as one or the other thus allows the useof nomothetic descriptive generalizations without ac-

tually studying the phenomenon in question each timeone sallies forth with a new pronouncement about theworld.5It should be clear that the affinitybetween case studyresearch and descriptive inference does not denigratethe possibility of causal analysis through case study re-search, of which one might cite many illustrious exam-ples. Indeed, the discussion that follows is primarilyconcerned with propositions of a causal nature. Mypoint is simply that it is easier to conduct descrip-tive work than to investigate causal propositions whileworking in a case study mode.

Scope of Proposition:BreadthandBoundedness VersusDepthThe variable utility of the case study is also partly aproduct of the scope of the causal argument that awriterwishes to prove or demonstrate. Arguments thatstrive for great breadth and boundedness are in greaterneed of cross-unit cases; causal arguments restricted toa small set of units more plausibly subsist on the basisof a single-unit study. The extensive/intensive tradeoffis commonsensical. Insofar as one wishes to make anargument about the universe of nation-states, one ison surer ground if one has studied more than one ofthose nation-states. A case study of France probablyoffers better evidence for an argument about Europethan for an argument about the world. Propositionalbreadth and evidentiary breadth generally go hand inhand. The evidence should be commensurate with thescope of the proposition.This statement, like all others, has a ceteris paribuscaveat. There is a variety of ways in which single-unit studies can credibly claim to provide evidencefor causal propositions of broad reach-e.g., by choos-ing cases that are especially representative of the phe-nomenon under study or by choosing "crucial"cases(Eckstein [1975] 1992). Even so, a proposition witha narrow scope will be more conducive to case studyanalysis than a proposition with a broad purview. Thebreadth of an inference thus constitutes one factor,among many,in determining the utilityof the case studymode of analysis.Similarly, the boundedness of an inference is oftenrelated to the degree to which it exploits cross-unitvariance. Precisely because their focus is so tight, casestudies often produce inferences with poorly definedboundaries. It is clear that an inference extends be-yond the unit under study, but it is often unclear howfar the inference extends. Cross-unit research may alsosuffer from poorly bounded inferences; however, it isless likely to do so since the research design allows oneto test the limits of an inference in an explicit fashion.

5 Granted, cross-unit examination may be useful, particularly whenthe terms in question are broad and/or ambiguous. If, for exam-ple, one is examining American political culture as an example of abroader class of political cultures deemed "liberal," t will be difficultto reach firm conclusions without a larger sample of nation-states.Even so, it will be easier to describe the subject without cross-unitreference-points than to explain it.

347


9/15

WhatIs a Case Study? May2004The sharpness of the boundaries of the population--what is and is not "covered" by an inference-is easierto establish when units in the population also constitutecases under study.By the same token, one of the primaryvirtues of thecase study method is the depth of analysis that it offers.One may think of depth as referringto the detail, rich-ness, completeness, wholeness, or degree of variancethat is accounted for by an explanation. The case studyresearcher'scomplaint about the thinness of cross-unitanalysis is well taken; such studies often have little tosay about individual cases. Otherwise stated, cross-unitstudies are likely to explain only a small portion of thevariancewith respect to a given outcome or to approachthat outcome at a very general level. A cross-unit studymight be satisfied to explain the occurrence of a revo-lution in Country A, while a case study of Country Amight also strive to explain specific features of thatevent-why it occurred when it did and in the way thatit did.I shall return to the advantages of takingan in-depth,Gestalt-oriented look, at a single unit. For present pur-poses, the conclusion is simple.Research designs invari-ably face a choice between knowing more about lessand knowing less about more. The case study methodmay be defended, as well as criticized, along these lines(Ragin 2000, 22).

UnitHomogeneity:Case ComparabilityVersusRepresentativenessSingle-unit studies provide cases that are likely to becomparableto one another. After all, they are all drawnfrom the same unit (by definition). Cases drawn fromdifferent units, in contrast, often force the researcherto make heroic assumptions about the comparabilityof concepts and causal relationships across the chosencases. After all, they are different units.Yet the strength of the case study also suggests a cor-responding weakness. Single-unit research designs of-ten fall short in their representativeness-the degree towhich causal relationships evidenced by that single unitmay be assumed to be true for a larger set of (unstud-ied) units. Are the men chosen by Robert Lane (1962)typical of the American male, white, immigrant,work-ing class? Is Middletown representative of America(Lynd and Lynd [1929] 1956)?The tradeoff between comparability and representa-tiveness is a general feature of cross-unit sample size.Naturally, there are many ways to overcome the cor-responding problems of comparability and representa-tiveness in both case study and non-case study research.Even so, it should be pointed out that the addition ofunits to a research design can hardly increase the casecomparability of a sample. Similarly,it is unlikely thatthe addition of units will decrease the representative-ness of a sample (though it is certainlypossible). Thus,itseems appropriateto regardthe tradeoff between com-parability and representativeness, like other tradeoffs,as intrinsic to the study/case study choice of researchdesign.

CausalInsight:CausalEffectVersusCausalMechanismsA fourth tradeoff concerns the sort of insight into cau-sation that one is able to achieve by examining empir-ical evidence of a particular X:Y relationship. Tradi-tionally, quantitative researchers have been concernedprimarilywith the estimation of a causal effect-the ef-fect on Y of a given change in X, taking all backgroundcircumstances into account (King, Keohane, and Verba1994, 81-82)-and, equally important, an estimate ofthe probability of that effect, captured statistically inthe error term. Assuming that the causal relationshipis probabilistic in nature (see discussion below), theresearcher must examine several instances of this phe-nomenon to gauge the average causal effect of X on Yand the random element of that variation. The calcu-lation of a causal effect presumes the investigation ofcross-unit variation precisely because, for a probabilis-tic cause, one cannot assume that the behavior of oneunit will be indicative of the behavior of other units.Units may behave differently. Thus, the example of asingle unit, even if subjected to iterated testing, is nota good way to estimate causal effects and is certainlyinadequate to the task of estimating probabilities.But causal arguments depend not only on measur-ing causal effects. They also presuppose the identifi-cation of a causal mechanism (Gerring, forthcoming;Hedstrom and Swedberg 1998). X must be connectedwith Y in a plausible fashion; otherwise, it is unclearwhether a pattern of covariation is truly causal innature. The identification of causal mechanisms hap-pens when one puts together general knowledge of theworld with empirical knowledge of how X and Y inter-relate. It is in the latter task that case studies enjoy acomparative advantage.Case studies, if well constructed, allow one to peerinto the box of causalityto the intermediate causes lyingbetween some cause and its purported effect. Ideally,they allow one to "see" X and Y interact-Hume'sbilliardball crossing the table and hitting a second ball.Clayton Roberts (1996, 66) describes process-tracingas "the minute tracing of the explanatory narrative tothe point where the events to be explained are mi-croscopic and the covering laws correspondingly morecertain" (see also George and Bennett 2004). Often,the connections between a putative cause and its ef-fect are rendered visible once one has examined themotivations of the actors involved. Intentionality is anintegral part of causal analysis, as interpretivists havebeen claiming for some time (Taylor 1970). Similarly,the investigation of a single unit may allow one to testthe causal implications of a theory, thus providing cor-roborating evidence for a causal argument. This is of-ten referred to as pattern-matching. Here, the theoryof primary interest "generates predictions or expecta-tions on dozens of other aspects of the culture,and [thewriter] does not retain the theory unless most of theseare also confirmed. In some sense, he has tested the the-ory with degrees of freedom coming from the multipleimplications of any one theory" (Campbell [1975]1988,380).

348


10/15

AmericanPoliticalScience Review Vol. 98, No. 2One can readily see why the investigation ofcausal mechanisms (including both process-tracingandpattern-matching) is commonly associated with thecase study research design. The in-depth analysis of asingle unit is useful in elucidating causal mechanismsbecause its characteristicstyle of evidence-gathering-over-time and within-unit variation-is likely to pro-vide clues into what connects a purported X to a par-ticular Y.Cross-unitvariation,in contrast, isoften mutewith respect to causal mechanisms. The Xs and Ys maybe at a considerable remove from one another; onedoes not know, or must simply intuit, what connectsthe dots.The caveat here is that cross-unit evidence is not al-

ways or necessarily mute with respect to causal mech-anisms, and single-unit analysis is not always revela-tory on this score. A cross-unit analysis is opaque if,or insofar as, (a) there is great causal distance be-tween the X and the Y variables, (b) the interven-ing variables thought to lie between X and Y cannotbe tested empirically in a cross-unit research design,and (c) the X:Y relationship cannot be intuited fromcommon sense or deductive reasoning. There is a gen-eral perception--common at least among protagonistsof the case study-that cross-unit studies entail largeblack boxes with no peepholes and that, therefore, theymust be supplemented by the in-depth analysis of keyunits. This is not always the case. One can think ofplenty of studies in which the relationship is quite clear,and where a case study would be superfluous.And onecan point out any number of studies in which interven-ing variables are in fact investigated with a cross-unitresearch design.6Again, it is important to stress that weare examining typical, not definitional, characteristicsof the case study.Ceteris paribus,case studies are morelikely to shed light on causal mechanisms and less likelyto shed light on true causal effects.CausalRelationship: nvariantVersusProbabilisticCausal arguments may be either invariant ("determin-istic") or probabilistic. Invariant causal relationshipsare asserted to be always true, given some set of back-ground circumstances.They take the form of necessary,sufficient,ornecessary and sufficientarguments.Proba-bilisticarguments,in contrast, are true in a probabilistic

fashion; a cause increases the likelihood of an outcomeand/or the magnitude of a (scalar) outcome.7Ceteris paribus, case study research designs havean easier time addressing invariant causes. Considerthat a necessary or sufficient causal proposition can bedisproved with a single case study (Dion 1998). Provingan invariant causal argument generally requires morecross-unit cases. However, it is not nearly as compli-cated as proving a probabilisticargumentfor the simplereason that one assumes invariantrelationships;conse-quently,the single unit understudycarries more weight.Where the causal relationship is assumed to be proba-bilistic, on the other hand, case study evidence is easierto dismiss; it is, after all, just one example of a generalphenomenon assumed to have a stochastic component.Recall that an error term may be incorporated into casestudy work since the N of a case study is indeterminate;however, this error term remains a property of single-unit analysis. In contrast, the error term in a large-Ncross-unit analysis represents the stochastic nature ofan (assumed) probabilistic relationship.Strategyof Research:ExploratoryVersusConfirmatorySocial science research involves a quest for new theo-ries as well as a testing of existing theories, a series of"conjectures and refutations" (Popper 1969). Regret-tably, social science methodology has focused almostexclusively on the latter. The former is quickly dis-missed as a matter of guesswork, inspiration,or luck-aleap of faith, in any case, and hence a poor subject formethodological reflection. Yet it will readily be grantedthat many works of social science, including most gen-erally acknowledged classics, are seminal works. Theirclassic status derives from a new idea, a new perspec-tive, that is subsequently subjected to more rigorousanalysis. Indeed, it is difficult to devise a program offalsification the first time a new theory is proposed.Path-breaking research is, by definition, exploratory.Subsequent research on that topic is confirmationistinsofar as its primarytask is to verify or falsify a preex-isting hypothesis or a set of hypotheses. Thus,the worldof social science may be usefully divided accordingto the predominant strategy of research undertaken,exploratory or confirmatory/disconfirmatory (Gerring2001, chap. 10). These constitute two moments of em-pirical research, a generative moment and a skepticalmoment, each of which is essential to the progress of adiscipline.Case studies enjoy a natural advantage in researchof an exploratory nature. These same advantages,

6 For example, a vast literature addresses the putative causal con-nections between trade openness and the welfare state. The usualempirical finding is that more open economies are associated withhigher social welfare spending. The question then becomes why sucha robust correlation exists. What are the plausible interconnectionsbetween trade openness and social welfare spending? One possiblecausal path, suggested by David Cameron (1978), is that increasedtrade openness leads to greater domestic economic vulnerability toexternal shocks (due, for instance, to changing terms of trade). Insubsequent work, writers have attempted to test this hypothesis byexamining the correlation between terms of trade and social welfarespending. If Cameron's causal mechanism is operative, one shouldfind a robust correlation between these two variables in a cross-national regression. As it happens, results are equivocal (Alesina,Glaeser, and Sacerdote 2001). The point is,writers can and do exploitcross-unit variation to test assumptions about causal mechanisms.

7 I avoid the term "determinism" since it has multiple meanings,onlyone of which-invariance-is relevant here. I assume that to makean invariantcausal argument does not commit one to a view that allcauses are perfectly determined; some causes may be invariant andothers probabilistic. Useful discussions of invariance/determinismcan be found in Adcock 2002, Dion 1998, 141, Goertz and Starr2003, and Waldner 2002. Goertz (2003, 76-94) includes a sample of150 necessary condition hypotheses deployed in various fields of thesocial sciences (Goertz 2003, 76-94).

349


11/15

WhatIs a Case Study? May2004

however, often serve as impediments in work of a con-firmatorynature. Let us explore why this might be so.8Traditionally,scientificmethodology has been identi-fied with the segregation of conjecture and refutation;one should not be allowed to contaminate the other.Yet in the real world of social science, inspiration arisesfrom perspiration. "Lightbulb" moments build on aclose engagement with the particular facts of a partic-ular case (unit). Ragin (1997) notes that case study re-search is all about "casing"-defining the topic, includ-ing the hypothesis(es) of primary nterest, the outcome,and the set of cases that offer relevant information vis-A-visthe hypothesis. It is a highly circular process. Astudy of the French Revolution may be conceptualizedas a study of revolution, of social revolution, of revolt,of political violence, and so forth. Each of these top-ics entails a different population and a different set ofcausal factors. A good deal of authorial intervention isnecessary in the course of defining a case study topic,for there is a great deal of evidentiary leeway. Yet thevery "subjectivity"of case study research allows for thegeneration of a great number of hypotheses, insightsthat might not be apparent to the cross-unit researcherwho works with a thinner set of empirical data acrossa large number of units and with a more determinate(fixed) definition of cases, variables, and outcomes. Itis the very fuzziness of case studies that grant thema strong advantage in research at exploratory stages,for the single-unit study allows one to test a multitudeof hypotheses in a rough-and-ready way.Nor is this anentirely conjecturalprocess.The covariational relation-ships discovered among different elements of a singleunit have a prima facie causal connection: They are allat the scene of the crime. This is revelatory when oneis at an early stage of analysis, for there is no identi-fiable suspect and the crime itself may be difficult todiscern. The fact that A, B, and C are present at theexpected times and places (relative to some outcomeof interest) is sufficient to establish them as indepen-dent variables.Proximal evidence is all that is required.Hence, the common identification of case studies as"plausibility probes," "pilot studies," "heuristic stud-ies," and "theory-building"exercises (Eckstein [1975]1992; Ragin 1992, 1997; Rueschemeyer and Stephens1997).A multiple-unit study, in contrast, generally allowsfor the testing of only a few hypotheses but does sowith a somewhat greater degree of confidence, as isappropriate to work of a confirmatorynature. There isless room for authorial intervention because evidencegathered from a cross-unit research design can only beinterpreted in a limited number of ways. Another wayof stating the point is to say that whereas case studieslean toward Type 1 errors (falsely rejecting the null hy-pothesis), cross-unit studies lean toward Type 2 errors(failing to reject the false null hypothesis). Perhaps thisexplains why case studies are more likely to be theory-generating,whereas cross-unit studies toil in the prosaicbut highly structured field of normal science.

I do not mean to suggest that case studies neverserve aconfirmatoryrole. As discussed, evidence drawnfrom a single unit may disconfirm a necessary or suffi-cient hypothesis. Case studies are also often useful inconjunction with a cross-unit study for the purpose ofelucidating causal mechanisms, as discussed. However,general theories rarely offer the kind of detailed anddeterminate predictions on within-unit variation thatwould allow one to reject a hypothesis throughpattern-matching (without additional cross-unit evidence). Thepoint is, theory confirmation/disconfirmation is not thecase study's strong suit. The selection of "crucial"casescannot overcome the fact that cross-unit N is minimal.We are unlikely to reject a hypothesis, or to consider itdefinitively proved, on the basis of the study of a singleunit, particularly f the hypothesis has behind it a corpusof scholarly work. Eckstein himself acknowledges thathis argument for case studies as a form of theory con-firmation is largely conjectural. At the time of writing,several decades ago, he could not point to any existingstudy where a crucial case study had performed theheroic role assigned to it (Eckstein [1975] 1992, 172).I suspect that this is still more or less true. Indeed, itis true even of experimental case studies in the naturalsciences (Campbell and Stanley 1963, 3). A single casestudy is still a single-shot, a single piece of evidencelying at the same level of analysis as the propositionitself.The tradeoff between exploratory and confirmatoryresearch helps us to reconcile the enthusiasm of casestudy researchers and the skepticism of case study crit-ics. They are both right, for the looseness of case studyresearch is a boon to new conceptualizations just as itis a bane to falsification. The problem is that work ofan exploratory nature, although it may receive praisefrom the discipline, is unappreciated, and greatly un-dertheorized, by methodologists.

Useful Variance: ingle-VersusMultiple-UnitThe analysis of any causal relationship hinges on thecounterfactual assumption-that without X (or withmore or less of X), Y would be different. In investi-gating this assumption the preferred research designsare, in order of preference: (a) laboratory or field ex-periments, (b) "natural"experiments (where a singleunit undergoes unmanipulated change through timethat approximates a true experiment), (c) thought ex-periments (counterfactuals), or (d) statistical controls(a quasi-experimental method of neutralizing irrele-vant variables so as to isolate the true causal effectsof one or a few factors of theoretical interest). Theimplication of this hierarchy of research designs is thata laboratory/field experiment, natural experiment, orthought experiment involving a single unitmay be moreuseful than multiple units that attempt to mimic thevirtues of the experimental method with purely "statis-tical" evidence.

For example, in investigating the relationship be-tween campaign efforts and voter turnout one mightbe more convinced by a field experiment conducted in8 For discussion of this tradeoff in the context of economic growththeory see Temple 1999, 120.

350


12/15


a single community than by multiple cross-communitystudies or individual poll data that rely on a host ofmore or less unsatisfactory quasi-experimental con-trols (Gerber and Green 2000). Similarly,in investigat-ing the role of electoral systems in conditioning pub-lic policy outcomes one might be more convinced bya single natural experiment-a change in a country'selectoral system-than by a cross-country study em-ploying statistical controls to examine these complexcausal relationships (Horowitz 1985, 602). Even whereone is forced to rely solely on counterfactual thought-experiments to evaluate causal claims, one might stillprefer the case study mode if useful variance is notavailable across units. The time-honored question ofwhether early democratization leads to a quiescentworking class and lower levels of social welfare devel-opment (Lipset 1963) is difficult to investigate cross-nationally for the simple reason that only one coun-try granted suffrage to the (male) working class priorto industrialization. In this circumstance, a historicalstudy focused on the United States (i.e., a case study)may provide the most compelling evidence of a generalproposition.To clarify, the issue is not whether naturalexperiments-or real experiments for that matter-aredesirable. They are always desirable. The issue is howmany experiments are available and how (truly) exper-imental are they? If many experiments are possible,as they typically are in natural science settings, thena cross-unit research design is probably justified. Butthis is uncommon in social science settings. Similarly,returning to our previous examples, if nation-statesexhibit substantial variance on the independent vari-ables of interest-early democratization and electoralchange-then one would probably choose a cross-unitresearch design (with a time-series dimension). Thiswould not preclude a supplementary case study, butsuch a case study would undoubtedly carryless weight.In the event, however, one is thrown back on moreprimitive expedients. Much of the utility of the casestudy is "contextual" in this special sense. The sortof natural or experimental variation that would makecross-unit analysis fruitful is lacking; such units simplydo not exist and cannot be generated.This helps to explain why case studies often focuson rare ("historical") events. Let me define an eventprovisionally as an instance of substantial and rela-tively quick change in an independent or dependentvariable of theoretical interest. Now imagine a uni-verse of empirical data in which such events occur inonly 13 instances. This approximates the situation ofscholars whose work focuses on the phenomenon ofsocial revolution (e.g., Skocpol 1979, 287). In princi-ple, the field of empirical evidence is virtually bound-less, including all nation-states over the past threecenturies. (Let us suppose that this constitutes the do-main of the inference.) Yet useful variation is severelylimited. The world does not provide many units thathave experienced revolutionary upheaval. Arguably,one learns more about this phenomenon through anintensive case study of France, or several case stud-ies focused on the 13 or so social-revolutionary coun-

tries (the comparative-historical method), than by across-country TSCS research design. The same prob-lem affects the study of many social science questionswhere only one unit, or a small number of units, un-dergoes the event that is to be explained: early in-dustrialization (England and the Netherlands), fascism(Germany, Italy), the use of nuclear weapons (UnitedStates), world war (WWI, WWII), democratization inAfrica (Botswana, Mauritius, South Africa), single-nontransferable vote (SNTV) electoral systems (prere-form Japan,Jordan,Taiwan,Vanuatu), and settler soci-eties (Australia, Canada, New Zealand, South Africa,United States).

OntologicalConsiderationsThus far, I have looked upon the choice between casestudy and non-case study methods as a matter of logic(the logic of causal inquiry) and empirics (the inves-tigation of the empirical world). However, this choicerests also on ontological presuppositions. An ontologyis a vision of the world as it really is, a more or less co-herent set of assumptions about how the world works,a research Weltanschauung analogous to a Kuhnianparadigm (Hall 2003; Kuhn [1962] 1970). As a resultof its all-embracing character and ambiguous claims,an ontology cannot be proved or disproved, at leastnot in the usual (verificationistor falsificationist) sense.Although it seems odd to bring ontological issues intoa discussion of social science methodology, it may begranted that social science research is not a purely em-pirical endeavor. What one finds is contingent uponwhat one looks for, and what one looks for is to someextent contingent upon what one expects to find.The significance of ontological presuppositions be-comes apparent whenever questions of breadth, casecomparability,and representativeness are vetted. Con-sider the possibility that a sample of units will becomemore and more dissimilar (less comparable) as thatsample is enlarged, but with no obvious break-points.Here it evidently is the choice of the researcher how todefine the population of a given inference and, hence,what the relevant units of analysis will be. Where dolike cases end and unlike cases begin?If adjacent units are thought to be entirely noncom-parable, the case study method is impossible. The per-fectly ideographic universe displays such uniquenessamong units that absolutely nothing can be learnedabout one unit by studying another. The notion ofa "case study" is nonsensical. At the other extreme,where all units of a given type are perfectly comparable,the case study is equally nonsensical. Why focus on asingle unit when other units will do just as well? This isthe nomothet's way of looking at things.Case study researchers are situated between thesetwo extremes. They are dubious about the viabil-ity of comparisons drawn across many apparently di-verse units. "Variable-oriented"research involves "ho-mogenizing assumptions" (Ragin 2000, 35). Yet theyare equally suspicious of the claim, implicit in muchhistorical and anthropological work, that each unit is

351


13/15


FIGURE. TheOntologyof Case StudyResearchDesign

Case StudyIdealctL-

Ideographic Nomothetic

AssumedComparability f PotentialUnits

unique. The case study thus occupies a tenuous onto-logical groundmidwaybetween ideographic and nomo-thetic extremes, as depicted in Figure 1 (see Collier andCollier [1991] 2002, 13-14).To be sure, there is no profit in dwelling on ontologi-cal differences between case study and non-case studyresearchers. Ontological debates are, by definition, ir-resolvable. Once one has defended one's position asa matter of ontology, further discussion is superfluousexcept as it might bear upon matters of logic and co-herence. If social science is understood as an evidence-based form of inquiry then matters of ontology aresimply not relevant or are only tangentially relevant.Nonetheless, insofar as our ontological presuppositionsinfluence our construction of cases, we had best be cog-nizant of this fact. Indeed, the middle-range positionof case study research on this crucial question mayhelp to account for its ambiguous position in the so-cial sciences. It is neither fish nor fowl, ontologicallyspeaking.

CONCLUSIONThis paper has attempted to shed light on the apparentdisjuncture between an often-maligned methodologyand a heavily practiced method, both of which go bythe name of case study. The torment of the case studybegins with its definitional penumbra. Frequently, thiskey term is conflated with a set of disparate method-ological traits that are not definitionally entailed. Thefirst task of the essay, therefore, was to craft a narrowerand more useful concept for purposes of methodologi-cal discussion. The case study, I argued, is best definedas an intensive study of a single unit with an aim togeneralize across a larger set of units.In other respects, the predicament of the case study isnot merely definitional but ratherinheres in the methoditself. To study a single unit with intent to shed light

upon other units brings in its train six methodologi-cal ambiguities that are, to some extent, ineradicable.First, case studies may build upon a variety of covari-ational evidence; there is no single type of case studyevidence, but rather three (see Table 1). Second, casestudies assume a distinction between formally and in-formally studied units that is never entirely clear sincethe latter must be brought into the analysis in somefashion but would compromise the notion of a casestudy if fully integrated. Third, individual case studiesare often grouped together in a single study, thus con-founding the distinction between single- and cross-unitanalysis.Fourth, case studies usually perform a doublefunction; they are studies (of the unit itself) as wellas case studies (of a broader class of units). Finally,the inference(s) pursued by a case study may be eitherillustrative or falsifiable. These methodological issuesbedevil the case study research design. And they shallcontinue to do so, for they areinherent in the enterprise.The travails of the case study within the disciplineof political science are rooted, finally, in an insuffi-cient appreciation of the methodological tradeoffs thatthis method calls forth. I have argued that at leastseven characteristic strengths and weaknesses must beconsidered. Ceteris paribus, case studies are generallymore useful (1) when inferences are descriptive ratherthan causal, (2) when propositional depth is prizedover breadth and boundedness, (3) when (internal)case comparability is given precedence over (external)case representativeness, (4) when insight into causalmechanisms is more important than insight into causaleffects, (5) when the causalproposition at issue is invari-ant rather than probabilistic, (6) when the strategy ofresearch is exploratory, rather than confirmatory,and(7) when useful variance is available for only a singleunit or a small number of units. (Ontological consider-ations also come into play when one chooses betweena single-unit and a cross-unit research design, thoughthe methodological implications of these assumptionsare equivocal.)Of these seven considerations, the last is perhaps themost important. There is little point in pursuing cross-unit analysis if the units in question do not exhibit vari-ation on the dimensions of theoretical interest and/orthe researchercannot manage to hold other, potentiallyconfounding, factors constant. Of course, a preliminarycanvassing of these units is necessary to perceive thesefactsabout the sample. But beyond this,cross-unit anal-ysis may be of little consequence. By the same token,individual units may offer useful variance or they maynot. In any case, the most important single questionresearchers should ask themselves as they contemplatevarious research designs is which one of these optionsmost closely approximates the experimental ideal.Throughout this discussion I have avoided the discus-sion of methodological considerations that arepracticalin nature or rooted in specific research contexts. Myconcern has been with general methodological issues.Before concluding this essay, however, it is importantto acknowledge that practical and contextual consid-erations are often paramount in the choice between acase study and a non-case study research format.

352


14/15


The collection of original data is typically more dif-ficult in cross-unit analysis than in case study analysis,involving greater expense, greater difficulties in iden-tifying and coding cases, learning foreign languages,traveling, and so forth.9 Whatever can be done for aset of units can usually be done more easily for a sin-gle unit. Similarly,case studies commonly afford multi-ple observations of a single case, thus providing firmerevidence of the factual accuracy of a given proposi-tion than would be possible in the analogous cross-unitstudy.A second practical consideration concerns the stateof research on a given topic. Here one is concerned withthe "triangulation" of evidence rather than the easeof evidence-gathering. Social scientists are accustomedto the idea that research occurs within the context ofan ongoing tradition. All work is dependent, for theidentification of topic, argument, and evidence, on thisresearch tradition. What we need to know, and henceought to study, is to some extent contingent upon whatis already known. It follows from this that the utility ofcase study research relative to non-case study researchis to some extent the product of the state of researchwithin a given field. A field dominated by case studiesmay have little need for another case study. A fieldwhere cross-unit studies are hegemonic may be des-perately in need of in-depth studies focused on singleunits.

Indeed, much of the debate over the utilityof the casestudy method has little to do with the method itself andmore to do with the state of current research in thatfield. If both case study and cross-unit methods havemuch to recommend them (the implicit argument ofthis paper), then both ought to be pursued-perhapsnot in equal measure but at least with equal diligenceand respect. There is no virtue, and potentially greatharm, in pursuing one approach to the exclusion of theother or in ghettoizing the practitioners of the minorityapproach. The triangulation essential to social scien-tific advance demands the employment of a variety of(viable) methods, including the case study.This paper is manifestly not a brieffor the case study.Rather, it is a brief for the better understanding of thecase study. We may or may not need more case stud-ies in political science. It is hoped, however, that theforegoing discussion will encourage better case studiesand a greater appreciation of their utility within thediscipline.It should also be apparent that the perceived hos-tility between case study and non-case study researchis largely unjustified and, perhaps, deserves to be re-garded as a misconception. Case studies may be small-or large-N, qualitative or quantitative, experimental orobservational, synchronic or diachronic. The case studyresearch design comports with any social-scientific the-oretical framework including behavioralism, rationalchoice, institutionalism, and interpretivism. What dif-ferentiates the case study from the cross-unit study isits way of defining cases, not its analysis of those cases

or its method of modeling causal relations. The casestudy researchdesign constructs cases from asingle unitwhile remaining attentive to inferences that span simi-lar units outside the formal scope of investigation. Non-case study research designs construct cases across unitsexemplifying the principal causal inference. This dif-ference in "casing"has important consequences, as wehave shown. However, it does not render the case studyepistemologically distinct from the cross-unit analysis.Indeed, the two modes of analysis are interdependent,and this is as it should be. One is at pains to imaginecross-unit research that does not draw upon case studywork or case study work that disregards adjacent units.The strongest conclusion to arise from thismethodolog-ical reflection concerns the complementarity of single-unit and cross-unit research designs.REFERENCESAcemoglu, Daron, Simon Johnson, and James A. Robinson. 2003."An African Success Story: Botswana." In In Search of Prosper-ity: Analytic Narratives on Economic Growth, ed. Dani Rodrik.Princeton, NJ: Princeton University Press, 80-122.Achen, Christopher H., and Duncan Snidal. 1989. "Rational Deter-rence Theory and Comparative Case Studies." World Politics 41(January): 143-69.Adcock, Robert. 2002. "Determinism and Comparative-HistoricalAnalysis: Clarifying Concepts and Retrieving Past Insights."Pre-sented at the Annual Meeting of the American Political ScienceAssociation, Boston.Alesina, Alberto, Edward Glaeser, and Bruce Sacerdote. 2001. "WhyDoesn't the US Have aEuropean-Style Welfare State?"BrookingsPapers on Economic Activity2: 187-277.Allen, William Sheridan. 1965. The Nazi Seizure of Power: The Ex-

perience of a Single German Town,1930-1935. New York:Watts.Allison, Graham T. 1971. Essence of Decision: Explaining the CubanMissile Crisis. Boston: Little, Brown.Bates, Robert H., Avner Greif, Margaret Levi, Jean-LaurentRosenthal, and Barry Weingast. 1998.Analytic Narratives.Prince-ton, NJ: Princeton University Press.Brady, Henry E., and David Collier, eds. 2004. Rethinking SocialInquiry:Diverse Tools,Shared Standards.Lanham, MD: Rowman& Littlefield.Cameron, David. 1978. "The Expansion of the Public Economy:A Comparative Analysis." American Political Science Review 72(December): 1243-61.Campbell, Donald T. [1975] 1988. "'Degrees of Freedom' and theCase Study."In Methodology and Epistemology for Social Science,ed. E. Samuel Overman. Chicago: University of Chicago Press.Campbell, Donald T., and Julian Stanley. 1963. Experimental andQuasi-Experimental Designs for Research. Boston: Houghton

Mifflin.Collier, David, and James Mahoney. 1996. "Insightsand Pitfalls: Se-lection Bias in Qualitative Research." WorldPolitics 49 (October):56-91.Collier, Ruth Berins, and David Collier. [1991]2002. Shaping thePo-litical Arena: CriticalJunctures,the Labor Movement, and RegimeDynamics in Latin America. Notre Dame, IN: University of NotreDame Press.Coppedge, Michael J. 2002. "Nested Inference: How to Combine theBenefits of Large-Sample Comparisons and Case Studies." Pre-sented at the Annual Meeting of the American Political ScienceAssociation, Boston.Dahl, Robert A. 1961. Who Governs?: Democracy and Power in anAmerican City.New Haven, CT: Yale University Press.Davidson, P. O., and C. G. Costello, eds. 1969. N = 1: ExperimentalStudies of Single Cases. New York: Van Nostrand Reinhold.Dion, Douglas. 1998. "Evidence and Inference in the Comparative

Case Study."ComparativePolitics 30 (January):127-45.Eaton, Kent. 2002. Politicians and Economic Reform in New Democ-racies.University Park:Pennsylvania State University Press.9 Granted, a good deal of cross-unit work involves the reanalysis ofexisting datasets;in this situation the barriersto entry are not so high.

353


15/15


Eckstein, Harry. [1975] 1992. "Case Studies and Theory in PoliticalScience." In RegardingPolitics: Essays on Political Theory,Stabil-ity,and Change. Berkeley: University of California Press.Elman, Miriam, ed. 1997. Paths to Peace: Is Democracy theAnswer?Cambridge: Cambridge University Press.Feagin, Joe R., Anthony M. Orum, and Gideon Sjoberg. 1991. ACasefor the CaseStudy. Chapel Hill: University of North CarolinaPress.Fenno, Richard E, Jr. 1978. Home Style: House Members in theirDistricts.Boston: Little, Brown.George, Alexander. 1979. "Case Studies and Theory Development:The Method of Structured,Focused Comparison."In Diplomacy:New Approaches in History, Theory,and Policy, ed. Paul GordonLauren. New York: Free Press.George, Alexander L., and Andrew Bennett. 2004. Case StudiesandTheory Development, Cambridge, MA: MIT Press.Gerber, Alan S., and Donald P. Green. 2000. "The Effects of Can-vassing, Direct Mail, and Telephone Contact on Voter Turnout: AField Experiment." American Political Science Review 94: 653-63.Gerring, John. 2001. Social Science Methodology: A CriterialFrame-work. Cambridge: Cambridge University Press.Gerring, John. Forthcoming. "Causation: A Unified Framework for

the Social Sciences." Journal of Theoretical Politics.Goertz, Gary.2003. "TheSubstantive Importance of Necessary Con-dition Hypotheses." In Necessary Conditions: Theory,Methodol-ogy and Applications, ed. Gary Goertz and Harvey Starr. NewYork:Rowman and Littlefield.Goertz, Gary, and Harvey Starr, eds. 2003. Necessary Conditions:Theory, Methodology and Applications. New York: Rowman andLittlefield.Hall, Peter A. 2003. "AligningOntology and Methodology in Com-parative Politics." In ComparativeHistorical Analysis in the SocialSciences, ed. James Mahoney and Dietrich Rueschemeyer. Cam-bridge:Cambridge University Press.Hartz, Louis. 1955. The Liberal Traditionin America. New York:Harcourt, Brace & World.Hedstrom, Peter, and Richard Swedberg, eds. 1998. Social Mech-anisms: An Analytical Approach to Social Theory. Cambridge:Cambridge University Press.Horowitz, Donald L. 1985. Ethnic Groups in Conflict.Berkeley: Uni-versity of California Press.Johnson, Chalmers.1983.MITIand theJapaneseMiracle:TheGrowthoflndustrial Policy, 1925-1975. Stanford,CA: StanfordUniversityPress.Kaufman, Herbert. 1960. The ForestRanger:A Study in Administra-tive Behavior. Baltimore: Johns Hopkins University Press.King, Gary, Robert O. Keohane, and Sidney Verba. 1994. DesigningSocial Inquiry:Scientific Inference in Qualitative Research. Prince-ton, NJ: Princeton University Press.Kuhn,Thomas S. [1962]1970. TheStructureof ScientificRevolutions.Chicago: University of Chicago Press.Lane, Robert. 1962. Political Ideology: Whythe American CommonMan Believes What He Does. New York: Free Press.Lazarsfeld, Paul E, Bernard Berelson, and Hazel Gaudet. 1948. ThePeople's Choice:How the VoterMakes Up His Mind in a Presiden-tial Campaign.New York, Columbia Univ. Press.Lieberman, Evan S. 2003. "Nested Analysis as a Mixed-Method So-lution to Cross-National Research." Ms. Princeton University.Lieberson, Stanley. [1991] 1992. "Small N's and Big Conclusions:An Examination of the Reasoning in Comparative Studies Basedon a Small Number of Cases." In What Is a Case? Exploring theFoundations of Social Inquiry, ed. Charles S. Ragin and HowardS. Becker. Cambridge: Cambridge University Press.Lieberson, Stanley. 1994. "More on the Uneasy Case for Using Mill-Type Methods in Small-N Comparative Studies." Social Forces 72(June): 1225-37.Lijphart,Arend. 1968. The Politics ofAccommodation: Pluralism andDemocracy in the Netherlands. Berkeley: University of CaliforniaPress.Lijphart, Arend. 1975. "The Comparable Cases Strategy in Com-parative Research." Comparative Political Studies 8 (July): 158-77.Lipset, Seymour Martin. 1963. The First New Nation: The UnitedStatesin Historical and Comparative Perspective. New York:BasicBooks.

Lynd,Robert Staughton, and Helen Merrell Lynd. [1929] 1956.Mid-dletown:A Studyin American Culture.New York:Harcourt, Brace.Mahoney, James, and Dietrich Rueschemeyer, eds. 2003. Compara-tive Historical Analysis in the Social Sciences. Cambridge: Cam-bridge University Press.Malinowski,Bronislaw. [1922] 1984.Argonauts of the WesternPacific.Prospect Heights, IL: Waveland.McKeown, Timothy. 1999. "Case Studies and the Statistical WorldView." InternationalOrganization 53 (Winter):161-90.Njolstad, Olav. 1990. "Learning From History? Case Studies andLimits to Theory-Building."In Olav Njolstad, (ed.), Arms Races:Technological and Political Dynamics (Thousand Oaks, CA: Sage)220-46.Popper, Karl. 1969. Conjecturesand Refutations.London: Routledgeand Kegan Paul.Pressman, Jeffrey L., and Aron Wildavsky. 1973. Implementation.Berkeley: University of California Press.Przeworski, Adam, and Henry Teune. 1970. The Logic of Compara-tive Social Inquiry. New York: John Wiley.Przeworski, Adam, Michael Alvarez, Jose Antonio Cheibub, andFernando Limongi. 2000. Democracy and Development: PoliticalInstitutions and MaterialWell-Being n theWorld,1950-1990. Cam-

bridge: Cambridge University Press.Ragin, Charles C. 1987. The Comparative Method: Moving BeyondQualitative and QuantitativeStrategies. Berkeley: University ofCalifornia.Ragin, Charles C. 1992. "Cases of 'What Is a Case?" In What Is aCase? Exploring the Foundations of Social Inquiry, ed. Charles C.Ragin and Howard S. Becker. Cambridge: Cambridge UniversityPress.Ragin, Charles C. 1997. "Turningthe Tables: How Case-OrientedResearch Challenges Variable-Oriented Research." ComparativeSocial Research 16:27-42.Ragin, Charles C. 2000. Fuzzy-Set Social Science. Chicago: Universityof Chicago Press.Ragin, Charles C., and Howard S. Becker, eds. 1992. What Is a Case?Exploring the Foundations of Social Inquiry. Cambridge: Cam-bridge University Press.Reilly, Benjamin. 2001. Democracy in Divided Societies. Cambridge:

Cambridge University Press.Roberts, Clayton. 1996. TheLogic of Historical Explanation.Univer-sity Park:Pennsylvania State University Press.Rodrik, Dani, ed. 2003. In Search of Prosperity: Analytic Narrativeson Economic Growth.Princeton, NJ: Princeton University Press.Rueschemeyer, Dietrich, and John D. Stephens. 1997. "ComparingHistorical Sequences: A Powerful Tool for Causal Analysis."Com-parative Social Research 16:55-72.Sartori, Giovanni. 1984."Guidelines for Concept Analysis."InSocialScience Concepts:A SystematicAnalysis. Beverly Hills, CA: Sage,15-48.Skocpol, Theda. 1979. States and Social Revolutions: A Compara-tiveAnalysis of France,Russia, and China. Cambridge: CambridgeUniversity Press.Smith, Rogers M. 1997.Civic Ideals:ConflictingVisionsof Citizenshipin U.S.History. New Haven, CT:Yale University Press.Stratmann, Thomas, and Martin Baur. 2002. "PluralityRule, Propor-

tional Representation, and the German Bundestag: How Incen-tives to Pork-Barrel Differ Across Electoral Systems."AmericanJournal of Political Science 46 (July):506-14.Taylor,Charles. 1970. "The Explanation of Purposive Behavior." InExplanation in the Behavioral Sciences, ed. Robert Borger andFrank Cioffi. Cambridge: Cambridge University Press.Temple, Jonathan. 1999. "The New Growth Evidence." Journal ofEconomic Literature (March) 37: 1112-56.Thompson, E. P 1963. The Making of the English Working Class.New York:Vintage Books.Waldner, David. 2002. "Anti Anti-Determinism: Or What HappensWhen Schrodinger's Cat and Lorenz's Butterfly Meet Laplace'sDemon in the Study of Political and Economic Development."Presented to the Annual Meeting of the American Political ScienceAssociation, Boston.Whyte, William Foote. [1943] 1955. StreetCorner Society: TheSocialStructureofan Italian Slum. Chicago: University of Chicago Press.Yin, Robert K. 1994. Case Study Research: Design and Methods.Newbury Park, CA: Sage.

354

John Gerring - Cuali

Documents

Transcript of John Gerring - Cuali