5 In Ada W. F"mifler, ed. (1993) Political Scjence: The State of the Discipline!!. Washington, D.C.: American Political Science Associat.ion.

The Comparative Method

David Collier

Comparison is a fundamental tool of ~ysis. II sharpens our poweT of de$cription, and plays a central role in concept-formation by bringing in!O focus suggestive similarities aJ'.ld cont""'1s among cases. Qimparisoo is routinely used in testing hypotheses, and it can contribute !O the inductive discovery of new hypotheses and !O thoory-buiJding.

The forms of comparison employed in the discipline of political science vary widely and include those contained in statistical analysis, experimental research, and historical studies. At the same time, the label "comparative me!hod" hos a standard me.wing within tho discipline and in the social sciences mori broadly: it refers !O the -tbodological issues that arise in the systematic analysis of a small number of eases, or a "small N. · • This chapter t.'tamines alternative perspectives on the comp•rative me!hod that have emerged over roughly the po.~t two decades. Although the primary focus is on discussions located in the fields of comparative politics and international srudies, the application of the comparative methOO is by no means restric~ to those fields.

The decision to analyze ool y a few eases is strongly influenced by the types of political phenomena under study and how they are conCCiJtua1i1.ed. Topics for which il is productive to examine relatively few cases include revolutions, particular types of Mtionol political regimes (e.g .• post«>mmuoist regimes). or particular forms of urb.ul political sysiew.s. This focus oa a small number of eases is adopted because there e~ist relatively few in.stances of the phenomenon under consideration that exhibit the attributes of interest to the analyst. Altemativdy, some analysts believe that political phenomena in general are best understood through the careful examination of a small number of ca5C$. lJ1 the fidd of comparative and international studies, tho practice of focusing on few cases ba.s achieved greater legitimacy in n>:ent ytllrs in conjunction with the rise of !lie school of · comparative historical analysis, · in which small numbers of countries are studied over long periods. This close scNtiny of each country liuUts the number of national cases a scholar COii consider.1

Choosing to study few cases routinely pos.:s the problem of having more rival explanations to assess than cases to observe, or the q\l2Ddary of "many variables, small N" (Lljphart L971, 686). Elementary statistics teadies us that as the number of expl1U1atory factors approaches the number of eases, tho capacity to adjudicate among the explanations through statistical compariSPn rapidly diminishes. This problem has stimu· la~ much discussioo of how most productively to analyze a small N.

The late 1960s and early 1970$ saw a boom in writing on comparative method (e.g .. Merritt :!Jld Rokho 1966; K.alleberg 1966; Verba 1967; Smelser 1968; Lasswell 1968; Pr:r.eworski and Teune 1970; Sarwri 1970; Merritt 1970; Eb:iooi al!d Dubow 1970; Lijpbar1 1971; Vallier 1971; Zdditcld971; Armer and Grimshaw 1973). This literarurc established a set of norms and practices for small-N research, proposed alternative strategies for conducting •"Ucb analyses, and creared a base line of understanding that bas played an important role in the ongoing practice of small·N studies. This chapter assesses the issues of companuive method that have beeo debated in the iniervetling years and considers thelr implications for ongoing research. The point of deparrure is Arend Lijphart's (1971) article "Comparative Politics and <Amparative Method. • Among the studies published in that period, Lijphart's piece stands out for its imaginative synthesis of basic issues of comparison and of the relation ~- comparative melhod and othet branches of methodology.' U therefore provides a helpful framework for examining, and building upon, new developments in the field.

A central theme that emerges in the discussion below is that refiaemcots ia lllelbods of small· N analysis have substantially broadened the range of leclullques available to comparative reselU'Cbers. The most frui1ful approach is eclectic, one in which scholm are willing and able to draw upon these diverse techniques.

106 The Comparative Method

Synopsis of Lijphart

Lijpbart defines che comparative method u the analysis of a small number of cases, entailing at least two obseivations, yet too few to permit the application of conventional statistical analysis. A central goal of his article is to assess the comparative method in relation to lhnoe othec metbods~xperimeotal, statistical, and C&Sb­

study-and to evaluale these different approaches by twO

criteria: 1) how well they achieve the goal of teSting theory through adjudicating among riV11! explanations, and 2) how difficult it is to acquire the data needed to employ each method (see Figure I).

The experimental method has the merit of providing strong criteria for eliminating rival explanations through experimental cootrol, but Wlfortunalely it is impossible to generate appropriate <>Xperimental data for most topics relevant to political analysis. The statistical method bas the merit of assessing rival explanations through the weaker but still valuable procedure of statistical control, but it is oil.co not feasible to collect a sufficieotly large set of reliable data to do this form of analysis.

The case-study method bas the merit of providing a framework in wbich a scholar with modest timo and resources can generate wh•t may polelltially be useful data on a particular case. Uofortuoately, opportunities for systematically testing hypotheses are fat more limited than with tho other mclbods. Ye< Lijphart (pp. 691-93) insists that case studies do m:iko a contribution to testing hypotheses and building theory, and he offers a suggestive typology of case studies based oa the nature of this contribution. He distinguishes amoog a1heore1ical ca..t;e studies; ;n1trpre1a1h•e case studies (that se!f-<:ansciously use a theory to illuminate a par1icular case); hypothesis-generating case studies; th<ory-confinning case studies; theory-infirming case studies (that, although they cannot by themselves disconfirm a theory, can raise doubts about it); and ,U..iQ/11 case analyses (that seek lo cl•borate and refine theory through a close cxamin•tioo of a case that depatts from the predictions of an established theory). Lijpbart empbasitcs that "certain types of case studies can even be considered implicit parts of the comparative method• (p. 691), and to the exlent that the as=meot of hypotheses does occur in some case studies, ii is ofteo because the case studies are placed in an implicit or explicit comparative framework. Yet even within this framework, he emphasizes that findings from a single case should oot be given much weight in the evaluation of hypotheses and theory (p. 691).

The comparalive method, as defined by Lijphart, has an intermediate status in tenns of both his criteria. It provides a weaker basis than the experimental or

statistical method for evahuting hypotheses, due l-0 the lac~ of experimental cont.rol and the problem of nuny vanobles. sm.11 N. Yet 11 does offer a stronger b:i.<is for eV11!uating hypotheses than do case studies. Despite the constrain~ of addressing more variables than c..ses, the companuve melhod allows systematic comparison tlui, 1r approprialdy utilized, can contribute to adjudicating among rival explanations.

Although the data requirements of the comparative method may be much greater than for caSe

studies, Lijpbart argues thllt Ibey ore I= demanding tlun for experimental or statistical rc:=n:h. He therefore views the comparative niclbod as most appropriate in re.search based on niodesl resources, a.ad be suggests that studies using the comparative ~thod might ofieo serve as a first step ioward statistical analysis.

U at all ponible one should gcnctally use the stotisticol (or perhaps cvca the cxpcrimcftuO method inst.cad of lhc weaker compsn.tiVC­mcthod. But oft.en. &ivcn lbc inevitable scarcity of time, cncri)', and fina.nc_ial ?CSOUrct!3, lhe intensive comparative an:tlysi.! of & few cuct may be more promising than a more. superficial Atatlatical analysis of many cases. In such a sjtuation, the most fruiLful approach would be to regard lhc comp1n.tive analy1iJ u lhe fll"l ..uge or reocarch, in whU:h hypo<hc<cs ~carefully formultu:d, and the mtistical analysis u the seoood s<agc, in which lhcse hypotba<:s arc le$tCd in as larie a sample u po$Sible. (l9il, 6&S)

Lijphart also proposes solutions to both sides or the problem of many variables, small N (1971, 686 ff). Wilh regard io the Stllllll ounibcr of eases, even if researchers stop short of a statistical study, lhey can nonetheless try to increase !he number of cases used in a.ssessillg hypotheses. With regard to the large number of vuiables, be suggests two approaches. First, 3ru1lysts cao focus 0<1 "comparable cases, · that is, on cases that a) are matched on many variabl<>S that are nor ccotral lD the study, thus io eff<>et •controlling" for these variables; aod b) differ in terms of the key variables that au the focus of analysis, thereby allowing a more adequate assessment of their influence. Hence, the selection of cases acts as a partial substitute for statis tical or expcrimeotal cootrol. Second, analysts caa reduce the number of variabl<>S either by combio.ing variables in a single scale or throueh theoretical p;mimony, that is, through developing a theory that focuses on a smaller oumbet of explanatory factors.

Thus, Lijphart provides a compact formu!atioo of the relatiooship between the comparative method aod

Rgur• 1. Situating tht Comparttivt P.icthod as of 1971: l.ijpht.r1°• Scheme

~ Study Method Comparative Method Exj>erlmen!AI Method

l ~eri~ Permi~ 1nten--Slve exam,naoon ot C-a.$eS even witn lun· lled resources

Inherent Problem: Contnbutes les.s 10 bu1ld1ng theory 1nan stuc!les wl1n mOfe eas&s Typu of Cue Stu· dlu: 1. A1heoretical

2. lnterpretivt

13. Hypo1h~sis·

generahng J • · Theory-confirming t 5. Theoty-tnhrm1ng I (i.e •• ease s1ud1e1 I tha1 weaken a 1 1heo<y marginally) I 6. Deviant ease StU018S

Defined u: Syste­matic 1n11ys.s ol small

number of cases

f"•m•• ·N .. enalysi$) Merit· "G1v1n inevn· able scarcHy ot time. enetgy, and financial resources, the inten-­si'Ve analysis o-t a few caso1 may be more promising lh'\n the superficial statlstJcaJ analysis of many caHs"' (Ujpha11. p. 685)

I tnherent Proble m: Weak capacity to .on O\lt nvaJ ex:pla.nat.ons. spec1ficafly. lhe prob­lem of .. many vao-abt11. few cases ..

Po11ntlal SoJutlona: 1. lncrea.se number ot

cases 2. Focus on comper·

able cues 3. R&duce number of

vanables a. Combine vari­

ables b. Employ more pat•

11monious thecxy

Me rtt: Elimlnat es nval explanations lhlOugh experiinental control Inherent Problem·

Expenmen1a! control ls impossible for many or most topics of

relevance to ttetd of

comparativ~ pc>lllics

Statl1Uc1l Method

Merit; Assesses rival explanations thtOIJOh

statistical control

lnher9nl Proble m· OilflCUlt 10 collect OO•

quate mtorrnanon 1n a sutfic:ient number of cases, due to limited time and resoutce1

Collie< 107

108 The Comparative Method

olhcr melhodologies, and he offers solutions to lhe characteristic djlemmas of the comparative method.

Further Perspectives on Small-N Analysis

Tho two decades following Lijphatt's study have seen the emergence of new perspectives on sma!J-N analysis, as well as a renewed focus on methodological alternatives already av:tilable before be wrote his article. Though many of these innovations appear io. work explicitly concerned with the comparative method, conventionally understood, others appear io. writing on the experimental, statistical, and case-study methods. The result bas been an intellectual cross-fertilization of great benefit to the comparative method. Figure 2 provides an ovecview of these innovations.

Innovations in the Comparative Method

Innovations in <be comparative method can be discussed in terms of the issues introduced above1 encom­passing the goals of comparison, tho justification for focusing on few cases, and the problem of DlMY variables, s~ll N.

Goals of Comparison

A central and legitimate goal of comparative analysis is assessing rival explanations. However, as Theda Skocpol and Margaret Somers (1980) argue, comparative studies should be understood not merely in teems of this single goal, but in teems of three distinct, yet ultimately connected, goals.' The first is that considered above: the systematic examination of eova.riation among c.'\ses for the purpose of causal analysis.' The second is the examination of a number of eases with the goal of showing. that a particular model or set of concepts usefully illuminates these eases. No real test of the theory occurs, but rather the goal is the parallel denwnsiration of theory. This use of eomparison plays an important role in the process through which theories are developed. The third type of comparison is the examinati.on of two or more cases in order to highlight how different they are. thus ·establishing a framework for interpreting how parallel processes of change are played out in different ways within each context. Tb.is conrrast of contt!Xls is central to the more • interpretive• side of the social sciences and reflects yet another way that comparison is frequently used.

Tn addition to p roviding a more multifaceted account of the goals of comparison, Skocpol and Somers suggest the intriguing idea of a research ·cycle· among

these approaches (pp. 196-197). Tb.is ~ycle arises in response to the problem5 th•t emerge as scholars push each approach up to - or beyond - the limits of its usefulness. For example, a •parallel demoJ)Stratioo" scholar might introduce a new theory and show bow it applies to m.iny cases. "Hyp.othesis-testing" scholars, wanting to specify the conditions under which the lheory does not hold, could m.'3cc further comparisons wilh the goal of discovering these conditions. Hypothesis-testing studies that too brashly compare cases that are p(ofoundly different might, in tum, stimulate ·contra.St of coot~xts" scholars to examine more carefully the meaning of the differences among the cases. ft is thus useful to lo<>k beyond an exclusive focus on the role of comparison in broad causal analysis, to an understanding that encompasses the different elements in this cescorch cycle.

'This is .not to say that assessing bypothese,s doe:i not remain a P.a,rnmollOt goal of comparison, and many scholars insist that it is the paramount goul. Yet tbis broader perspective offers a valuable account of how comparative work proceeds within a larger r=rch community, pointing usefully to the interaction among different goals of comparison.

JuSlijicatWn for Small N

A S<leond trend is toward a mora elaborote justification of a focus on relatively few cases. Lijphart's rationale seems in retrospect ratber modest, in that it emphasizes only the problem of inadequate resources and treats the small-N comparison as a way station on the route to more sophisticated. statistical analysis.

A very different defense of wocking eilher with a small N or with ~ studies had previously been available in arguments favoring a "configurative· approach (Hecksche.r 1957, 46.-S l , 85-107), and this per.;pective was elaborated a few years before the publication.ofLijpbart's analysis in Sianey V«ba's (1967) review essay advocating the "disoipline'd configurative approach.• In evaluating Robert A. Dahl's Political Opposirions in \Veste:m Democracies ( 1966}, Verba points both to the sophistication of the hypotheses entettamcd in the book and to the difficulty of assessing them adequately. except through a close command of the cases, leading him to. advocate this disciplined configurative mode of research. Verba's formulation is appealing becausa he is coneemed with systematic . hypothesis testing and theory building. At the same u..,, he links this priority with a more explicit approeiation of the difficulty of testing hypotheses adequately and the value of properly executed case studies in providing

subtle a.<.sessmcnts of hypotheses. It might be claimed that the difficulty of

adequately testing hypotheses ultimately derives from the

Collier 109

Agure 2 . lnnO\>ltioru Rekv1nt to d~ Comp1rativ~ Mdbod COmparatlv• ... thOd Experimental MethOd

Broadened UndentandinQ of Type• or Diffusion of Older Ide•• and tntroduo-CDmp1r.1lve $1udi•• lio.n of New ldo•• on Ou.Mi•

1, Emphasis Of'l lnterpreuve u~rsland1ng EXperime.nut 0.a.lgn

2. Idea ol • "tesea1cn cye!e" among the 1. MetnOOOlogy ot quasJ•.x,porlmen1s •nd

rypes iSkocpo,l Ind Som en.) interrupted tui'l•·series d•slgn t>tCOm••

Funher Jusllfte11Jon.1 for Focua on• m0te 'Mdoty known

Small·N 2. Analysis of the Connectievt cr3r,.1«k>wn

Case Sluely M11th0d , . ro pur$U8 "di$C1pt1neo COnfi.guretiVll on tpced1ng as 011emplar of lntertuptod

approach" (Vorba, temtorced t>y [,... tl trfO-$.Elrid onaiym (Campbtll and

Hew Pet1pectJvtt on Cate Studies Almond and Genco) Ross)

1. Ne;; uo:e.1.>0 ~i ;r,11 "'''• •it.1Vr 1flu(;,.;;y 2 , To avo1<1 prooiem ol "conceptual 3. Citt:.i:io~ cf i~c:e "1>out. q~~I·

(Campboln sttetchir.g" !S!-'tOri) experiments enoouraQOd »y «>0•1\callon

2. Relincmenls In L1jgha11'1 typology o t 3 To lacm1a1c "lhjck descnphon" a/Id of ev&luation research

c.au studl(!5 (Eckstein, George) 1 .... othor k>rms of 1ruerpre11vo unclerstano-4. Proposed s..;atislica.I solution to problem

3 1mprovemen1 ol causal an.•tys11 in ca.se Ing (G&er\l ancJ many othets) ol selection bias in ciuuJ.ox,,.nm•nts

studtes uuough "~rocess tracing'" 4. To tchieve anntyuc dop•h o1 cu.-(AchM)

(George and M<K-") 011ente<f approach (Aagu\) Statiatlcal M•tbOd

4. Cn11que 01 va!ue ot ca,se stucl·•1 1n Otb•les on Soluhon.s to Probtem of

a.s.s.e.ss.ing rat.on.al cnoict lf\.ary (Achen I Many V•tl-1ates, SmaU··M New Waming-11nd New Solut~ON

aJ1d Sno<I.,) , VaJue oC increasing number°' cases 1 Criticism oC stana.1rd llatiM&I prectit;t

I 2 Compa.ra:>te cases ver$VS conlta.sting "'-

•n the sooal sc1ence.s (Freedman)

eases (L,pnan vet$US Pneworskt and 2... New statistical t.ChntQU'eS reMvant to

Teu.ne) $11\all-N analyslS

3 Reoucang numnier ol vatlabo.ft in ~

I 3~ Etfort IO refine stabStlC&I anatysis ol a

I tunction Wlll'I U:SlnQ suonget lheOf"y

~fl..N 111 lhe debate on corpor1hsm .and economw; grow1h In W11tem Eur- (Langt-GilltOU.J-P1nerson)

110 The Comparative Method

problem of limited re&<>Ul'CC$ discussed by Lijpbart. If enough taleoted researchers worked long aod hard, they could carry out a Polit/cal Oppositions study for many dou:ns of countries. Yet the problem here is somewhat different from that emphasized in Lijpbart's initial formu­lation. 11 is AOt so much that reso= ate limited, but !hat coostructing adequate comparisoas has proved more difficult Iha.a had ofleo been thought in the 1960s and early 1970s, in lhe ioitw days of eothusiasm for comparative statistical r=rcb. Among these difficulties, that of the Vlllid application of concepts across divetse contexts bas been e.o;pecially vexing.

Within the literature on comparative method , a key step in elucidating these problems of validity, and thereby strengthening the justification for a small N, is Giovanni Sartori's (1970) classic discussion of "Concept Misfonnalioo in Comparative Politics, • the basic themes of which are elabonated in hls later book Social Sci~nce Ccnttpts (1984). Sartori suggests that the application of a coacept to a broader range of cases can lead to conceprual "stretching, • as some of the meanings associated with the concept fail to fit the new cases. The concq>ts th.at cao most easily be applied to a broad range of cases are often so general that they do not bring into focus the similarities and contrasts among cases that are OSS<Otial building block.s in worthwhile comparative analysis. Consequetitly, 1 study focused on coaoepts !hat ate carefully adapted to this "finer slicing" of a given set of cases should be extended to other cases only with areat caution. From this perspective, it may be argu<\CI that the most interesting studies will ofleo be those th.at focus on a smaller number of cases.

With regard to the probleru.s of increasing the number of CIS08 uodor study.' A<Lim Przeworski and Heory Teune's 7lie Logic of Comparative Social /llqulry (1970) is a major source of insight. Although they arguo that achleviog a hlgb level of generality should be a basic 11oal of socilll science, their frameworic is centrally cooc:emed with the difficuhies that can arise in ceocraliziog beyond m initial set of cues. With regard to problems of validity, they advocate the use, wbeo aecessary, of ·system-specific" indicators that serve co openationaliz.o the sanu: concept in distinct ways io different contexts (pp. 124-130). For the scholar seeking to move toward a larger set of cases, the potential need for system·specific indicators necessitates the close examination of every new case.

Prz.eworski and Teuoe also address tho problem that as the analyst incorporate$ more cases into a srudy, distinct causal panems msy appear in the new cases. To deal with this problem, Pruworslci md Teune advocate "replacing proper names" of social systems by idcntifyini those systems in terms of the explaoatory factors that account for why causal relations take a particular form

within each system (pp. 26-30).' This approach m.'ll:eo the invaluable coatributioo. of providiag a theoretical rather than an idiosyncratic and case·specific, basis for analyzing differences in causal patterns. However, extending an analysis to additional cases on tho basis of this prooedure again requires a pai.nstalciog assessrncn1 of each new context. Thus, Pruworslci and Teune provide 1 valuable tool for adequately analyzing a larger number of cases, but their approach again shows th.at this must bo done with caution.

Since 1970, the renewal ofa Weberian concern with inteipretivo understanding, i.e., with deciph<ring the meaning of behavior and institution.< to the actors invol ved, bas also strengthened the justification for advancing cautiously with one or very few cases. Oifford Geertz's (1973) label "thiclc description" is commonly evoked to refer to this concern, 1 and this focus bas appeared in many guises relevant to political research, including Gabriel Almond and Stephen J. Geoco's analysis of "Cloud., and Clocks· (19n) and Skocpol and Someno' · contrast of contexts" approach, which encompasses studies that use comparison to richly contellluali.to research findings. Charles C. Ragin's The ComparaJi>oe Method (1987) explores another facet or thit concern in hls analysis of the •holistic" orienution of what he calls "case-<>rieoted• research and the complex problems of "conjuoctural causation" - thal is, caUSlll patterns th.at vary accordins to lhe context - co which contigurative scbohus ate typically far more sensitive.

Finally, the intellectual success in recent years of the school of comparative hlstorical analysis bas pbycd an important role in lej1itim.11ing a focus on a small N. This approach was p ioneen:d in works such as Reinlurd Bendix (1964), Barrington Moore (1966). and Lipsot anJ Roklcan (1967), and moro recent wod(s include Rok.k•n (1970), Tilly (1975), Paige (1975), Bendix ( 1978), Trimberger (1978), Skocpol (1979), Bergquist (1986), Luebbert (1991), Goldstone (1991), Collier and Collier (1991), and Ru=bemeycr, Stepheas, and Stephens (1992). Methodological statements focused on this tradition include Skocpol md Somet'S (1980), Skocpol (1984) , Tilly (1984), and Ragin (1987).

The particular form of analysis in these studies varies coasiderably, as suggeswd by Slc.ocpol and Somcrt' typology noted above. In varying combinations, th0$0 studies employ both rigorous qualitative comparisons that exlald across a number of nations, and also historical analysis that ofleo evaluates each national case over a number o f time periods.' This tradition of research thus combines weU-tbought-<><1t compo.risoo with an appnociation o( hlstorical context, thereby contributing to an effo~ to "hlstoriciu" the soc:ill scieooes.

Although the wes of comparison in this l iterature arc diverse, as Slcocpol and Somers emphasized. it may

be argued that a major .consequence of the growing i.ropor1ance of comparative historical studies is to further legitimate the approach that was Lijpllart's original concern: the assessment of rival explanations, based on systematic, qualitative comparison of a small number of ca.=. In light of a spectrum of studies from Barrington Moore's (1966) pioneering analysis of the emergence of alternative forms of modem regimes, to Skocpol's (1979) study of revolution, to Luehber1's (1991) analysis of the c:merg-eoce of UberaJis.m, fascism, and social democracy in inteN.•ar Europe, it is evident that this literature has given new legitimacy to the use of broad historic.al con1parison for systematic causal analysis. Efforts lO codify procedures for assessing hypotheses in this type of analysis, such as that in R·agin's Comparcui~·e Method (1987), fur1her reinforce the plausibility of insisting on the viability of small-N analysis as a middle ground between case studies and statistical studies.

Solutions 10 the Problem of Many Variables, Small N

The evolving debates on comparative method have suggesttd fur1her refinements in Lijphan's original three solutions for the problem of m:l!ly variables, small N, i .e.: 1) increasing the number of eases, 2} focusing on matched cases, and 3) reducing the number of variables.

I. Increase the Nuntber of Cases Al. the tirue Lijphart wrote, it was believed in some circles lbat comparative social science would in¢re..<t.Siogly be oriented toward large .. N co1nparative studies, based on extensive quaacitative da.t.a SetS and rigorous statistical analysis. Today •here can be no queslion that, for better or worse, quantitative cross-national research in the subfield of comparative politics, and quaniitative international politics in the subfield of international relations, ru.ve not come to occupy as dominant a position as many bad expected. Wil.b.in these two subfields, they hold the sta!US of one approach among many.

Vari9us factors have placed limits oo the success of large-N re.=rcb ba.,.,,J on quantitative data. sets. amoog which is certainly the re.newed concern with closely cont<>xtuali ze<I analysis and interpretive studies. Broad quantil3tivc comparison lll3Y have been set back a.• many scholars discovered how extraordinarily time-consuming it is to construct appropriate data sets. often out of proportion to the professional rewards that seem to be fonhcoming. This is por1iculady a problem when the focus of analysis extends beyond the advanced industrial countries to regions for which it is often extremely difficult to develop reliable data. In addition, the quantitativt-<:omparative approach bas probably been hur1 by the publicatioo o f too many studies in which concepL• are opcratioaalized with dubious validity and

Collier 111

which employ causal tests that are weak, unconvincing, or inappropriate (Ragin 1987, chap. 4).

Yet the fact that broad quantitative comparison has not become a predominant approach should not lead scholars to overlook what has been accomplished. Rohen Jackman (1985) insists· that comparative statistical research has had more suw:s:s than is recognized, and Lijphart1s own recent work moves in this direction (1990). The failure to seize good opportunities to do quanlitative research could certainly be viewed as being as much of a mistake as premarure quantification, and the fruitful debate ·oa corporatism and economic growth in Western Europe discu~ below is one of many ex.amples of bow statistical methods can effectively address interesting analytic issues. Further, the availability of new statistical a..cluliques (also discussed below) has made it far more productive to do qu.wiita(ive analyses with as few as ten to fifteen cases. Conse­quently, the option of increasing the "N" at least to that level is still worth pursuing, and ii should probably be pursued more often.

2. Focus on Comparable Cases The re<:ommendation that analysts focus oo carefully matched cases has been both reinforced and cballeoged. In a discussion published in the mid-1970s, Lijphart (1975) explores further the trade-<>ff be noted in 1971 between the ·goal or increasing the number of cases and the goal or matching cases as a substitute for statistical control . Obviously, if a researcher is to select cases that are really similar, however that similarity is defined, lhe number of appropriat<> cases is likely to become limited. Jn the face of this trade-<>ff, Lijpbart opts in favor of the more careful matching of fewer cases, and he goes so far -as to restrict the application of the term · co111parative method"' to analyses that focus on a small number of carefully matched eases. This emphasis parallels a ·much earlier perspective on the comparative method referred to as the method of "controlled comparison" (Eggan 1954). Arthur Stincheombe's (1978) advocacy of the methodology of "deep analogy,· i.e., the comparative analysis of very fow, extremely closely matched, cases pushes this approach even further.

A contrasting strategy is advocated by Przeworski and Teuae (1970, 32·39) and Przeworski (1987, 38-41). They suggesr that even with careful matching of cases in what they label a ·most similar" s.y•iemS design, there remains a problem of •overdcteanioation, • in that this design fails to climiaatc many rival explanations, leaving the researcher with no criteria for choosing atn0ng them. They prefer instead a "most different" systems· design, based on• set of cases which are highly diverse and among which the analyst traces similar prOC<:SSeS of change. '0 Przeworski suggests that the strength of this design is in par1

112 The Compantive Method

responsible for important advances in the literaturn on democr:atiution, such as the work of O'Donnell, Sclunitter, and Whitehead (1986). Pneworsld maintai.n.s that this literature addresses suob a broad range of cases that ~lySIS are foruxi to distill out of that diversity a set of common ele111CDIS that prove to have great explanatory po~·er.11

This discussion can be placed in perspective by recognizing that ca.'leS that are closely w..icbed from one point of view may contnist sharply from rutotber. My own iecent work (Collier and Collier 1991) combines the two strategies by starting with a set of eight Latin American countries that ue roughly matched on a number of broad dimensions. Among tho eight COUt>tries, the analysis fOC<ISC$ on pairs of countries that ate

nonetheless markedly different. The overall matching &SSUl'e$ that tho contexts of analysis are anolytically equivalent, at least to a significant degree, and the paired oomparison places parallel processes of change in sharp relief because they are operating in scttin~s that are very differem in many l1'SJ>OCIS.

In ooojuoctioo with the debate over the merits of most similar and roost different syste~ designs, it is imporUJlt to recognize that in many studies, the conclusions reached in the overall comparison of cases are also asse"8ed - implicitly and sometimes explicitly -through within°"'ase an•lysis. Jn the section on case studies below, the discussion of 'pauem matching• and 'process tracing" suggeSIS some of the ro~ this takes. It is no ooincidence that within the school of oomparative historical analysis, findings are often reported in books, rather than articles. Part of the reason is that the presen~~tioo of detailed information oo eoch case serves to further v.lidnte the conclusions drawn from comparisons across cases.

These within-<:ase comparisons are critical to the viability or small·N analysis. As Stanley Uebersoa (1991, 312·315) bas correctly insisted, taken by themselves, comparisons across a small number of cases, using either a moSt similar or a most different systems design, provide a weak basis for cauS'11 inference. However, if one eoosiders the role of the..~ internal comparisons, the • N' is substantially increased, thereby strengthening ca~! analysis. 11

This use of wiihi.n""'2SC oomparisoo can also help protect the analysr from a problem that arises in the most different systems design, in wbicb oountries are matched on the dependent variable and differ in ienns of a series of background variables. Barbara Oeddes (1990) bas shown that if cases arc selected on the bll$iS of scores on the dependent variable, which is bow most different systems designs are ofteri carried out, the lack of variance on the outcome to be •~plained introduces a •selection bias• tbar can g~tly weaken causal inference. One way

of mitigating this problem is to introduce greater varia­bility through internal ooruparison.

The ongoing debate on most similar venms moot

different systems designs has implications for the status of area studies. Danlcwart Rustow (196&) ar~ed 50.,.,

time ago in favor or moving beyond an area studies approach, and many scbolara agree that cases •bould ho selected in response lo the analytic requirements of particular research projecl~. nther than on tho b11Sis of a geographic proximity that at best is often a poor •"Ubstitute for the analytic matching of cases. Recent 'cross-2tea' s1udies on successful export-Jed growth 2"'1 oo democracy sugge.<t thar this altttnative p<rsp<ctive i$ gaining growid."

However. tho area studies approach is a boomiog business today for a variety of reasons, inc luding Che impressive funding of area srudies by U.S. foundlltions io iecent decades, as well as instirutional momentum. In fact, from the point of view of the theoretic•lly oricnt.ed small·N comparativist, this is oot a bad outcome. The COUt>try case studies produced by "'"" spec~lius are crueial building bloclc.s in most oomparative wor1c, 211d without them cross-area studies would be on far weoker ground. It is essential to recognize that these case studies benefit grcally froo1 tho intoll<OC:Cual leverage gained when individual scholars develop, over many years. • cumulative and well""'°nte>1ualized understanding of a particular region. Parueularly io light of current concems that broad comparative studies should be atteorive to the cootext of Malysis, the oontribution of area specialiSIS is essential.

3. Reduce t/14 Number of Variabw: The third solution to the small-N problem is to reduce tho nurober of explanatory factors, either through combining variables., sometimes referred to as "data reduction,· or through employing a theoretical perspective that focuses oo a smallec set of e.\pl2D.111ory factors. One of the promising soun:es of parsimonious explanatory rhoory is the ·rational choice· approach that bas gained increasing attention amcng political scientists. Rational cboico modeling offers a productive means of simplifying arguments that contain• mullirudo of interesting variables, but that may fail to specify the most critical ones. Within the field of comparative analysis, Geddes's (1991) study of administrative refonn io Latin America. which models the impact of different electoral and party systems on the incentives of lcgislatoJS to adopt reform, provides an excellent example or. productive simplification of a comp)Oll topic. As such models gain iocr=ing acceptance in the «1mparative field, analysts will acquire a useful tool for addressing the smoll·N problem."

More wor1c on concq>t formation is also needed, notwithstanding the sustained contnl>utioo of Sartori

(1970, 1984, 1991. 1993, and Sartori, Riggs, and Teune 197S); tbc work of autbors such as McKinney (1966), IUtllebctg (1966), and DeFelire (1980); and also Burger's (1976) invaluable synthesis of the Weberian approo<:h co con«pl form:uion. Comp3111tivists do not devo1e cnooi:h atlention lo thinking through how well or poorly concepts arc serving them and therefore may have insufficient ground for knowiJ1g whether they are ruJllcjng :ippropri•t" choices in the effort to achieve theoretical parsimony.

The field of cognitiv:e science has recently providlld insights into categorization that may be useful in n:fining tho concepts employed in comparative studies. TI1e applicalion of Chose insights is illus1111ted by Coorge Ukoff's (1987) challenge to frameworks. such as lhat of Sar1ori, lhat employ what Lakoff calls "classical categoriza11on, • in which the meaning of conC<:plS is und.ot1>t00d in lerms of defining cbaracceriscics thac arc seen as giving lhe roncepts weU~efined boundoric.s. TIUs underswiding is crucial to Sanori's framework, in lhac th" problem of conceptual sttecching which he analy= hingo:s on these boundaries. Cognitive sciencists ari:ue lh•t in ordinary langwge, the meaning of concepls derivc.4i n.ot from defining characteristics, but fro1n an implicit ·co~'11itive model" that underljes cbe concept and rrom "cxi.:mpla.r" ca.~ th:tl serve to anchor che concept's mcnning aJ1d provide a point of reference for identifying b<ller and worso c=. Tills perspeclive provid.:.s a differen1 view of lhe qoe.5tion of boundaries, and hcnoc of conceptual £!retching. More woric is needed lo discovor the degree 10 which lbese patterns in ordinary language arc also prcsc.nt in social science usage. and if so, the implicationJ for Che use of concepts in companuive analysis (sec Collier and Mahon 1993).

Innovations Suggested by Work on Other Methods

Experimental Method

Allh()ugb rbe experimenlal mecbod itself may be of liule relevance lo the klpics addressed in cnosc comparacive reS<l"reh, ideas deri,;cd from lhe experimental method can improve small-N srudies. Donald Campbell and Julian Stanley's classic Experimtntal and Quasi-Experimental Designs for Rt.rtarch (1963) shows bow lbe logic of experimental desi~"' can be •pplied to • quasi-<oxperiments, ' thal is, IO "observational" studies lhal include some ev"'1t or i.nnova­tion that bas a form analogous to an cxp:rimeot.:tl intervention, but tb3t OCCUtS in a ·natural· selling. An example would be the ioiciacion of a new public policy wb0$e impact one wishes 10 assess.

Campbell and St.10.lcy underline the gr"'11 value in qw1si-expcri1ne.nts of tbe • incerruptOO tin1~ series•

Collier U3

design. In this design lhe analyst looks al a loog series of observations over time. so !hoc the values of the observed variable are examined not only at lwo points immediately before and after the policy change or other innovation (which "interrupts" the series), but also well before and weU after. To illustraw Ibo risk of restricting the analysis IO these two observations, the authors present several bypotherical configuralions of dacu in whicb restricting the analysis 10 lwo Observalions leads 10 a fwding of sharp discontinuity, whereas tl1e ruu tioie series reveals continuity. CaU$1ll inferences about the impact of discrete events can be risl-y without an extended series of observations. Comp3ratlvists employing small-N analysis "'"-" heed this warning, since they routinely analyze the impact of discrete llVents, ranging from waN, revolutions, and military coups to SpCCiJic public policies.

Donald Campbell and Laun:ncc Ross's (!968) subse­quenc analysis of the impact on traffic fatalities of the Connecticuc crackdown on speedinjl in lhe 1950s provides a srunning "exemplar' of the> 1Ul3;iru11ivc application of a quasi-experimental design lo public policy analysis. Indeed, Przeworski (1987, 31) has argued that melhodology is influenced for more by exemplar.i Chan by fornial allcmpts to · 1egi:l\late"' correct rnethods, antl tbe Connecticut crackdown ar1icle ha.< certainly played that role. 11

The ca.'.i¢ appears IQ be a simple one. When lhe SlJlte of Connec1icu1 initiaced scricc cruo=ment of the vehicular speed limit in the 19SOs and traffic deaths dropped sharply, the cause and effect n:lacionship sumcd obvious_ Yet in ,;valuacing chis cau5"1 link, Campbell and Ross do an impl'<'SSive analy$iS of pocential threats co its "inlernal validity' (was that really tbc cause in Co~cicut?) and its "exccmal validity" (can lhe finding be generalized?). No seosuive analyst can read this article without acquiring a more sober view of the problems of evaluating policy impacts.

Ideas about quasi-cxperiment'11 and incerrupled time series design have also b.li;n disseminaled through lhe largo body of wriling on evaluation """"1tcb. This iocludcs sludics chat apply these ideas lo the analysis of p01itical developmenc (Hoole 1978), a.• well as excellent treatment~ of exporimeotal design and cvalualion res<Oatch in incroduccory textbooks on socfol science mechodology, such as Babbie (1992).

Allhough much wriling on quasi~xperimeots appears lo offer helpful guidance and practical advice lo

small-N analysts, Christopher H. Acheo's The Statistical Anal)~is of Quasi·Exf'<'riments (1986) may lea\'C chem feeling that the methodological cballtng<11 poS<ld by Ibis type of design are overwhelmin&. In studies of lhe impact of public policy, lhe core problem is the lock of "r•ndomization" in the application of cbu policy, which

114 Tbe Comparative Method

m.ay rtSUlt in sel..,tion bias. For example, I.be benefits of a policy are commonly received by SOUIC groups and aot by other$, oo I.be basis of cer13in attributes poSSCS3CCI by the groups, and it is possible I.bat lbesc prior attributes will lbem.<elves reinforce I.be outcom.:. I.bat I.be policy seeks to promote. In the abscoce of true experimental data, this poses the cballeoge of disentaogling the impact of the policy from I.be impact of these prior attribuiea. This causal riddlo cao be addrcs.sod by constructing a model of how citizens are scl"'ted to be recipients of the policy. This model theo becomes a building block in the analysis of the policy's impact, in that these prior considerations CllO be • fuctored out• in assessing the policy. Acheo shows that solving the riddle requires a complex form of "cwo-stage" statistical analysis.

The implications of Acbeo's book may be discouraging for analysts worlcing wilb a small number of cases. All adequate solution to the lack of randomiutioo iequires a form of statistical analysis which cao be applied to an elaborate quantitative data set, but this t<:ehnique would be bard to apply in a small·N study. A morco hopeful view might bo that the literature on expcrimeots and qu:isi-expcriments at least provides useful warnings about the perils of analyzing discrete events a.• if they were true experimeota l intcrveotions. In the absence of approprU!a data sets, the nisearcber must exercise cautioo i1l nuking causal claims4

lnnovatlons In Statistics

Recent work on statistical analysis bas provided both new warnings about the risks of statistical studies and new opportunities for doing meaningful statistical work with relatively modest case bases. The statistician David Freedman bas launcb...1 a major assault on the use of multivariate quantillltivc analysis in the social sciences (1987, 1991), which be claims fails becsuse the uoderlyiag research design is aeoerally inadequate and because the dalll employed fail to meet the assumptions of the s12tistical techniques. His criticism may bring coo.sidcrable satisfaction to I.base who have been skeptical aboul statistics all along and who talce comfort in the gtealCr "cootrol" of the material they feel derive$ from analyzing relatively few cases through more qualitative techniques. It is realistic to expect that we may go through a period of greater questioning of the use of statistics in the social sciences. However, as with tho rejection of qu:wtitative cross-national research discussed above, it would be uofortunatc if a reaction against quaotibtiv<>studies weot too far.

The emergence of new statistical t~qucs that are helpful in the analysis of relatively few cases makes such a blalllcet rejectioo unwarr.inted. One example is the development of "n:sampliag strategics" such as the "bootstrap" and "jaclclcnife" (Diaconis and Efron 1983,

Mooney and Duval 1992). These techniques u.se computer simulatioo to create, from an initial set or red data, a large number of hypolheti°"I replicatioos of the study, which can theo be used in statistical tesL• that are not as vulnerable to violations of distributional IU>'Ulllj>­

tions as are more cooveotiooal tests. These tcchniqces may be especially useful when there is great heterogeneity among units, as may readily occur io cross-national comparisons.

The development or "robust" and "resistant' statistical measures (Hampel et al. 1987; Haitwig 1979; Mosteller and Tukey 1977) is promising in much the same way. These measurcos are relatively unaffec.ttd by extreme or deviant values and can lherefore help overcome the problem ia small·N analysis that finding> 011y be seriously distorted by a single observation that is greatly in error.

Another set of techniques concerned with this woe problem is "rcgressioo diagnostics" (Bolleo and Jaclcmao 1985: Jackman 1987). These are tests used in conjWJ.ction with conventional regression analysis to assess whether unusual values on particular obscrva:tioos, called influential cases, have distorted the findiags. ~ advantage or regression diagnostics in comparison with robust and resistant statistics is that one can employ them with the more. familiar coefficient.$ IWX'iarecl with regremoo analysis.

Tbo use of regression diagnostics is nicely illustrated in the receot debate on the relationship between corporatism and economic growth ia IS Western European countries (tange and Garrett 1985. 19117; Ja.ckman 1987, 1989; Hicks 1988; Hicks and Patlcrsoo 1989; Garrett and Lango 1989). The starting point of this debate is Peter Ullge and Geoffrey Garreu's 1985 article, which presents an interesting md complex idea in a simple form. They argue wt the organil'ltional strength of unions in the labor market and the political stttngth of the left in the electoral and govemmcntal arenas both have an impact oo ""°oomic growth, but that this impact is shaped by a complex interplay betweeo these two factors, whiob they repreS<:ot through an "iotcntetioo' term ia their regression analysis of the IS

In a reanalysis of I.heir article. Robert W. Jockruan (1987) employs regression diagnostic~ ID

examine certain iaflueotiaJ cases that be believes distort their findiags. In the eosuiag discussions among these five authors, an expanded model with further control variables is proposed, this cxpaoded model is both challenged and defended. and lan~c and Garrett subsequenUy defend their original model and call for new d•la and further tests.

This scholarly debate brings together an important sub5tanlive problem. a hi"1> level of •rea expertise and lcoowledge of specific cases, the inventive

use of a relatively slraighlforward statistical model, a constructive-critjque based on regression diagnostics, and a sustained process of cumulative knowledge geaeratioa based oa the scrutilly of a shared data set. Just as the Campbell and Ross article on the Conaecticut speeding crackdown is an exemplar of a quasi-experimental design, this debaie should stand as an exemplar of a methodologicaUy sopb.isticaied effort by several scholars to solve an important problem within the framework of small-N quantitative analysis. 1bis debaie also shows that although an "N" of 15 might often be approached through qtL1litative small·N comparison, it can likewise be subjected to sutistical analysis, with interesting resuits.

Another area in which potential problems of statistical analysis are amenable to solution concerns the issue or "average eff«:ts~ in regression studies. The results of the simpler forms of regression analysis are based on an average of the strength of causal relations across the cases being studied. For the coefficients produced by regression analysis to be meaningful, it is necessary !hat these causal relations be homogene<>us across the cases. Yet Ragin (1987, cl!Jlp. 4), among others, bas forcefully argued that this assumption commonly does not hold, given the complex forlll!i of ·multiple conjunctural causation'" often eocowue-red in comparat(ve studies. In different contexts of analysis, the interaction among causal factors may vary.

However, solutions to lhis problem a.re-available. John E. Jackson (1992) shows bow it can be addressed with advanced statistical techniques, and the interaction term in the l..auge-Oarrell regression analysis, discussed above, deals witb precisely this problem: that the effect of nne explanatory factor '1aries depending on the value of another expl3natory factor. FinaUy, Przeworski and Teune's procedure of ·replacing proper names,· also discussed above, takes this problem of causal complexity and turns it into ·an opport\ln.ity to deal more theoretically with lbe diversity of causal patterns.

lnno>"ations in tht Case-Study Method

When Lijphart wrote bis 1971 article, he apparen~y fell some hesitation about including a discussion of case studies in an assessmeo.t of the comparative method.16 Yet the two topics are closely linked. and rus helpful typology of the uses of case studies ill hypothesis testing and theory building set the stage for refinements in ease study analysis later introduced by other scholars.

One of the most suggestive discussioos of the case-study method is that of Campbell (1975). He draruatieally rtcants the bold assertion be mad<: in b.is earlier book with Stanley tbal •one-shot" case studies are

Collier llS

"o[ almost no scientific value" (1963, 7). He shows instead that case studies are the basis of most comparative research, that they offer many more opportunities than is ofteo recognized for falsifying the researcl>er's main hypotheses, and that mucb can be learned from making explfoit the comparisons tbJI~ are often implicitly .built into case studies. For example, any given hypothesis about a case bas implications for many facets of tho case. Campbell uses -Ole label "pattern matching· to refer lo the process· of discovering whether thC!» implications an; realized. The analyst can thereby increase th.e "N" by multiplying the opportunities lo test hypotheses wiUi.in what may initially have been viewed as a "single" case.

This procedure of pattern matching is liolpful in. addressing the long-standing concern that ease studies are useful for gene.rating hypotheses, but that the same case cannot then be used lo test the hypothesis because it offers no possibility of discolllinnation. This is sometimes referred to as the problem of ex post facto hypotheses." The proceduno of pattern matcrung opens the possibility that an hypotbei;is initially geocrated by a particular case could subsequently fail to be supported by the same ease. Thus, the problem of ex post facto bypolhesei; can be partially overco01e. "

Harry Eckstein (1975, 113- 123) is likewise concerned with testing, as opposed to generating, bypolheses in case-study il.nalysis, Md be argues fori:efully !hat many analysts have greatly underestimated the value of case studies for hypothesis testing. lo particular, the cartfully constructed analysis of a ·entieal ease• - for example, one about which the analyst bas particularly strong expectations that it will flt the hypothesized causal pattern - can provide an invaluable opportunity to falsify the relevant hypothesis.

Alexander George and Timothy McKeown (1985), building on George (1979), present a helpful synthesis of two key building blocks in the process through which hypotheses are tested in case studies. The first corresponds to the conventional approach tn placing a case ill comparative perspective, wruch they call the •congruence procedure.• Tue scholar examines the values of an hypothesized independent aJJd dependent variable for a given case and determines, in light of explicit or implicit comparison with other cases, whether these values are consistent with the predictions of the hypothesis under consideration (pp. 29·30). The second is "proccss-ttaciog, • through wb.icb the researeber engages in. a close processual analysis of the unfolding of eveuts over time within the case (pp. 34-41). The goal is to assesses whether the dynamics of cbaoge within each case plausibly reflect the same causal pattern suggested by the compantive appraisal of the case in relation lo olher cases. Process tracing ruay be seen as a specific illstancc of Campbell's pattern matching, and as· with

116 The Comparative Method

pattern matching the analyst makes a series of within-case observation.~ against wJiicb the hypothesis can be fur1ber assessed.

Overall, these articles, along with works such as Robert 1C Yin's Case Study Research (1984), offer a systematization of case-study procedures that provide a valuable point of reference for scholars· concerned with small-N analysis. At tho same lime, the debate continues on the proper role of case studies in .assessing and building theory. An inreresting part of this debate, published as a special issue of the journal World Politics (1989) focuses on the contribution of case studies lo evaluating one applic-.atioa of rational choice analysis, i.e., rational deterrence theory in internatjonal re1atjoos. The o~g article by Acben and Snidal (19.89) argues chat the case srudies employed by many inremational relations specialists do not adequately address the central ideas of this body of theory, thereby raising an issue perhaps not often enough considered in discussions of the comparative method: How can the methodological concern with executing good comparisons be linked to the key analytic issues posed by the particular theories that are to be evaluated? Acheo and Snidal also note the problem of selection bias in case studies of deterrence theory, that is, the problem that case studies usually focus on deterrence failure, whereas much or most of the time deterrence works. The issue of the journal includes a series or articles by S<:balars close to the case-study tradition who debate the issues raised by Achee and Soidal. These articles constitute a valuable effort to th.ink through how case studies have functioned in relation la the assessment of a particular body of theory, a line of inquiry that should be taken up more often.

In this debate oo deterrence theory, an intellectual tension emerges that bas been a recurring theme in this chapter. between analyses that seek to achieve a generic understanding, based on relatively fe-w variables and encompassing many cases, as opposed to analyses that seek to draw out the complexities of particular cases.

Conclusion

Among the diverse approaches discussed in this chapter, three major analytic alternatives stand out. First, new perspectives on the case-srudy method have strengthened the viability of that approach. Discussions of opportunities for within-ase comparisons have in ·fact begun to blur the distinction between case studies and the comparative method, although the case-study approach does remain a distinct tradition. Interest in case studies has been reinforced by several factors, including the renewed concern with interpretive social science. the

continuing intellectual and institutional strength of are. studies, and deep skepticism in some circles abouc the validity of broad comparison.

Second, it is evident that quantitative tecbniques employing a relatively small number of cases can successfully address important substantive questioo.s. This approach merits attention in light of the new

statistical tests suitable for small-N analysis. Tiie opportunity foI cumulative scholarly learning provided by statistical studies is n.icely illustrated by the Lange-Oarrett-Jackm:m-Hicks-Patter.;on debate. This debate is also relevant to the issue of linking rival researeb traditions, because it shows that insights derived from case studies aod from more qualitative comp:uative work can, af\e( all, serve as stepping-stones on the path toward statistical analysis.

The third altemotive bas been reinforced as well: the systematic comparison of a small number of Case3,

with the goal of causal analysis, which is the approach that Lljpbart originaUy advocated. In !his perspective, broad qualitative comparison is seen as both possible and productive. The growing influence of the school of comparative historical analysis has substantially eohancod the c redibility of this approach, and i~ plays an important role as an analytic middle ground between the case-study tradition and small-N statistical analysis.

All three of these approaches will persist, and a key question is how well they can be linked. The tradition of research on Western Europe provides an encouraging model, in that the findings of quantitative comparative schola.rs play an important role in general debates in that field." ln research on Latin America, by contrast, quantitative comparative work receives considerably less attention from mainstream scholars. Yet the kind of cross-fcrtilb:ation fallDd in the West European field can make an important contribution to strengthening research. With good communication, country specialists and experts in qualitativesrnall-N comparisoa can push the comparative qu.a.otifiers toward more carefully contextualized analysis. Likewise, the comparative quantifiers can push the COllDtry specialists and expertS in qualitative comparison toward more systematic measurement and hypothesis testing. A central goal must be to sustain such communication.

The implications fo< graduate !raining are clear. Tf Ph.D. candidates are to be prepared to address these issues of comparison, they should have enough training in statistical methods to evaluate quantitalive studies that employ old, and new, methods of statistical analysis and to use such methods when appropriate. Those more oriented toward statistical analysis should have enough backgrollDd in qualitative small-N comparison and case srudy analysis to be able to build on the analytic contribution of those approaches. Both groups should have substantial exposure to basic writings on the

philosophy of science and logic of inquiry that can provide the framework for more informed choices about these methodological alternatives.

In this way, th~ foundation can be laid for an eclectic praetice of smaU-N analysis lbat takes advantage of opportunities that present themselves on both side.• of what could otherwise be a major intellectual divide.

Notes

Th.i• ii • revised aod expanded VtrtioR of 1n artidc c.uiic r publi&hed in Da.nl::'wart A. Jt.1,1.1row a.Dd Ke:MClh Ptul Erick.Jon, edt., Compa.raiiw Palitico/ Dyrtamlcs: Gln"41 Rcs~"rch Ptrspcctiw.s (New Y<>rt: Harper Colliru. 1991). Pcrminion to tc-print gra.n~ed by Harper Colliru. Rulh Bcrins CoWcr, Kc.nncth Paul Erkbon, Leonardo Morli1ui, Elizabeth &&bee, a.Dd Caro) A.. Mc4lio made parti-cullrly useful RJggcflions on catticr-dnfb. I al.so actoowfc.dgc c(Hnmcnu Crom Chrinophr:r Achcn, Su:phcn ColGcr, Junes Fcaro.n, David Freedman, IXb<>rui Norden, Robert PowdJ, ~{cnill Sha.ob, and U ura Stoic et. Ada F'.mi.Rer and two anonym0tu re.vicv.·ert likewise nude helpful commenu. Thia f'Ctu~b ha• been supported b)' 1 G\lgg~cim Fellowship, lhc SociaJ Science Re~src:h Coumil, and che lnaitut.e of GovcrnmcntaJ Srudies al Berkeley, Finally, I would llke to note:• Very pronlisiog a'l4nut<!rip1 ()Ong, Verba, and Xeoh-.oo Im) lhal urtf()rtii.Ntcly c.amc: lO my auentioo too l:itc Lo be di.J;;U-»ed in th.is eh1pter.

I, .,..,. ia \l&c:d io R:for lo th& numbor of C.IJC.J tM.1)7.cd ia •ny giv-en stl>dy.

2. Refe-rencea lO ttpn:$CnLllive. worb of comp.ualive histori: .al atalys.U. •~ prc:JCnt.ed below.

3. In his compecison of these n~1hod1, Lijphart e;;koowledgu hi.a debt lO S~lKr"• (l96S) cxce.Uc.nt a.na.ly•i• that employed a patt.!ld fruticwork:. See also Smclitr (1976).

4. This perspective ·has be.co e.laborstcd by Sto:-pol {1984, chap. l l ), and a pareUd f<>rtnulation i1 found in Chatlt1 Tilly (1984. chap. 4).

5. Skocpot 1.nd $.oit)C.-. (1930, 131-81) re.for to thit l.'i ~macro-causal' analysts. Yet •null·N 11Wdi.e1 th&t generate 1.od 1.u:1

hypotbes.::s ean have botb o tnACro a.nd a. micro foe.us, a.ad it dou OOl teem pro<,Jueiive-lO exclude from thi• category tho~ v.·hh a mi:ro f«:u•. Hent:e, this altcm.ative- la.bcl ia u~d.

6. Allhough Pruwottlci and Teune are eentr.1lly eooeemcd wilh issues lhal a.rise wtica additioMI eaJ1C:t arc: added to an 11o&ly1i1, the problcm1 \hey dl~un are also mote likely lO OC¢Ur if Otle it dcali.ng wi:lh •larger N IO begin with.

7. For ebmple. inst-ead of tcferring to ·vc.nezuda, • ooe would refer to • country in wbicb, due lO the- impact of mauive oil 1Y>Ven.uc1, a patiicular c1u...U relationship usun'les a distinct fonn.

8. 11lic.rdeac-ription· i.1 aome.timea miR.akcnly understood l.O refer $imply to •de.tailed de&Cription, ·which i• oot what Gec.rtz

inltnc'5. 9 . Give.a tha1 ihe-fC atudica of\co focu1 oo long periods ()f

time w.ithin e.cb ca~. i1 nUgbt ~argued thtt the number of cues could be. gttatly inc:tee,ka lhroogh cacnparison over lime, thereby mo king them somcthin,s olhcr than J.ma-ll·N" srudic1, However, 1ince the*°*' of ma.oy srudiea in lhia tradition is l.O explain ovenJI configuration& or riational outcome• a1 they are nu.nirctl.ed over long ptriodt , th_ca.t outcomes otka. ~i.nnot be diN.ggrcgatcd into a .criu o( longitudinal ob~r.·alion.s. Hence, I.he number o( c.aJCt c.annol realiSlic&lly be inerca~d through lhe uae. o f c:ompari.oo over time.

10. The mott $ioUl..r and moll. diffcrcn1 •YSkml dulgas

Colliu 117

co1TC1pond, .te3{1Celively, LO John Stuart ~f'ill'• (1974) method of dill'ereoc:.e and method of a~cncnt. Whetca1 Pr-""won:kj and Teuoc•a label• of ·1im.iJar" aud .. different• refer LO whether the C•tU are m.ati;bcd, •• opJ>Ofic:d lo coaltUlin,. oa a .ac.riu of bdekgrou:nd Vari.able•. Mill'• la.bel1 or "'diffcrenc~· a~ "'1;grccmet1.1• refer co whether the caac.1 &l'e cootn~ing, a• oppoled toma\ched, OQ the depaukns variable;

11. Pcl'llQOA.I c:omn'IU.DieatiOft from Adl.m Prz.cwor:d::i. ll. Christ.opbct A,ehco, pcr:soo.al communkaiion, bu Ion&

Wisted oo lhi• point.

13. For C°':t1mplc. Getcffi aJ'ld Wymaa (1990), Haggard (1990). Pruwon.ki (1991), &nd ~:uc.1ebcJneyer.Stcpheos, 11.M Sttpbcra (1992}.

14. for a di.,;uuion of ltfatcg:ic choice modela {a eto.C"ly rel.ti.Cd type of model) that have been applied to che an.aty•i• or political rcfonn. democniti:tation. a.ad democ:n:itic comolid.stion i.&:1 l..aJ.io. America, aad thai 1.ikcwl.e ofl'er fruitful ajarpli(ication. o( complex pbcoomeGa, .u Collier a.nd Notdc-n (1992).

JS. The rcpri.f'!li-ng of lhi1 a.nick- io a teadcr oo JOCial Kico:c methodoloa:y (f\lftc l970) made h widely avaihble 10 politic.al llC~li.tu. and iu inOuc~ ha.a bee-a 1Ubata.ntial.

16 • .Peaoc.tl coDUtw.nication from Arend Lijpbart. 17. Thia problem i1 roudncly ditcuuc.d in iru.roduetory

mclh-odologyu:ru, e.g •• Babbie ( 1992., 24-2.S, 427). J 8. Although patttm ('fUtch.i."i within lhe same caae­

introd1;1cc1 I.he P:OSs:ibili ty of fahi(ying th~ hypothem~ il doc.s 00(

oven:ome all of the pl;Cbfcma or ex poJ.t facto bypothc&e.a. Th.ua, pattern rr.stcbing wilJ probably DOl overcome> • p.roblccnof unreprcs.cnUtivcOU;I wbjcb may a.rise due to 1elet:tiOt1 bi.a• or to I.be. ch.s.DCc telection of an atypical c.uc.

19. Sec, for example,~ dcb•le oo inlcresi mcdiatioo and c.orporati1m in Wc.11tem Europe, including Wilensky .(1976), Hibb• (1978). Sc:hmi:Lter (198 1), Uld Cameron (1984). TI1e 'dcbat.c lll.Vd by Lange and GArrett (198S) it. contin'llation of th.ii line or •naJyai1.

Bibliography

Achci:i, Cbristophct H. 1986. 1hc Suu1srleaJ A.MlytU q/ ~i~ Ex~rl~u. Berkeley and Lm At\gelu: Uni.venity of California Pre .....

Ac hen, Cbti$lOl)be-r H., a.nd-Duix:1_q SnidaJ, 1989. ·Rational Dctcrtt.nee Thcoty and Compan.livc. C.6C Swdlea. • World P"1ilics ~1: 143-69.

Alm:oad, G•briel A.1 and Stephen J, Ococo. 1m. "'Cl.oud•. Clocks. ~ the Srudy of PtJlitic:1, • World Pt>litics '29;489-'Sll.

Almer, Mkhael .. -•nd Alleo Grim9haw, ed•. 1973. Ccrnporod't'e Sodol Rc:1~Q~/s . New Yori;: John Wik.y.

Babbie, Eld. 199'2. The PrtJC.cit:e of Sodol &s,ea~h. 6th cdi1ion. Belmont, CA: W•d,worib.

Bend~.Rei.ahttd. 1964. Nonon·Bullding an# Cldunshlp: S1ud/a()/ Our ~gi'ng Scx:ial Onkr. New Yurt: JObn \\,..lie)'.

Bendix. lleinh.ard. 1973. Kings or Pe,opk: Power and tJu Matll/.au IO Ru.le. Bc.r~clcy and Loa A;oselca: UNvcni1y of Californi.t

"""· Bergquist, Charles. 1986. l.4.borin Lalin AmerlC:tJ: Comparaii'w &says en Otik~ A.f'8pi.dn<J, Vie-~VU'la, and CAIQmbla. SIAO!ord: SwUord Uruven:i1y P'Rs.s.

Bolleo, ~lh A •• and Robert W. Jackman. J98S. ·Rcgt¢11io-0 Ditgnostica: An Expository Tretlmcnt of Ou\Jjcrt and lnOueotial Ca5e:,r. • Sociologic41 Methods ar.d &i-earrh 13,SlG-42.

Burget, Thomas. 1916. Mar Web<r'1·1Aeoryoj(A)~ep1Form¢t:Jn<.

Hiszory. Lowr. 41ld ld.tQJ Ty~$. Outb.am: Duke Universily'.

118 The Comparative Method

Cameron, D•vi.d R.. 1984. ·social Dcn'IOCraey, Cocpor.li$:lll, Labour Qu:ksconce, and the Reprc&enC.tion of &oQQmic lol.cte~ i..o Adva.nccd Ca~itali.1t Society.• IA Ottkr ond C>"JUct in Contemporary Copirallnn, ed. John H. Ookhborpe. New Yott Oxford University~-

CampbeU, Oor:aJd T . 1975. ''Degrees ofFtudom• aod lhe Case SWdy. • Comparoli\'e-Poli.o'cdl Srudiu 8; l 7&-93.

Campbell, OooaldT . • aod H. l..au{'t,nce Ro». 1968. tho Connc-.eti.:ut Cn.('kdowrt on Spc.:dio.g: Time Seriea Da.ta in Qut.•i· Experimental Analy.si.s.' Lttw and StJdcty RcvieW '3:'33-S'l.

C1mpl>ell, Donald T .. 1.od Juliao C. Stanley. 1963. £.qxritM11'41 o;Jtd

QU4sl-&Mrim.mtol Designs/or Rt:st:<Jrdt. Chice.go: Rand McNally.

Co11ier. David, and Jame• E . Mabon. l993. 'Conceptual 'Strct;:;hing' Revisited: AlttO\llJ.ive VieWJ o( Cat.cgorie1 in Comparative Ana.ly!i.i$. • A.m('.riC''1!'1 Politico/ Scf~ru:~ RM<w.

Collier. David. 6.nd 0-:botah L. ~ordcn. 1992, 'Stn.1.cgic Choice f\.foddt'Qf Poli.t:ic:al Change-in J.A1jn Amc:ri.:a. • c.tJm.parodv< P()/Jlia 14:219·243.

Collier, Ruth Beri.ru, and Oa..,id Collier. 199l. Shaping /Ju PoliJical A~: Crltkol Junctures. lhe L4bor Movemau. afl4 R~gf!M DyntJmlr:s in lalm ~rica. Prin«ioo: Princctoo University Pren ..

Dthl, Robc-tt A., ed. 1966. Political OpposidO'ILS'in Wr:skrn Democ.rodcs. New Haven: Y•le Unh··¢nity Pru&,

CkFdicc, E. Qenc. 1980. ·compari'°o ~ii.J<:oni;;civcd: Commoo Nons<l\$C in Com;pentive Po1iti-e.s." Compo:NJdvc Poli:!<.·s

13: 119-26. Diaconi.s, PcNi, and Bnidley Efro1, . 198'3. ·co1npuu:r·lrltct1.iive

~ictJ1od1.in Stati$lic~. • Scicrllific A.mr:ric;an '248:116·.30. F.cbtcin~ Hany. 197S. ·case Study and Theory in Politi.:al Scicoce. •

ln l(andhoqk of folf6col Sci=<. vol. 7. ed. Fn:d Oo:e0>1<·io . utd NcllOo W. Polllby. Steading, ~iA: Add'i-&00-Wc11lcy.

Eggtn, Fred. 1954. "Social Aotb.ropology and tbel-fcUiod o(

Controlled Compari.t00. • .Am,crica:n A.nlhropologisr 56:743-

63. Et7.ioni, Amitai, and Frederic L. Dubow, eda-. 1970. C.Ompa,ad\•c

Pr:rs~crlw:s: 1hr:oric4 and A(erhnd.r. Bo~n: Little, Btow11.

FrccdtJUn, DAvid A. 1987. ·~s {)then Sec. U1: A C•SeSf:!Jd)' in Path. Ana.lysis. • Joumol of Educo:i()(l.Q/ St.alisJics 12:101-28.

freedman, O•viJ A. 1991. ~st.slh1.ical ~{ockl• i nd Sh<>c (...calhcr.• 1n SociologicoJ ~f<ihodoWgy 1991, ed. P<:tec ~11f"ldcn, S1D Fn~iscO: JQ11<:y~Bass.

0-artcu, Gcofrny, 1.nd Peter Lange. l9&9. · ao..,cmment Plr\4.aruhip and .Ec-onomi.c Performance; Wlu:.n aod How J)oe, 'Who Govccns' ~ialter?• Jo&Jrr.ol of Politics 51:676·93.

Gcddcs1

Ba.rbtra. 1990. •ttow the C.11es You Choose Affect th-e An:ilwc~ You Get: St:l~tioo Bias io Compantivc Politics. • ln Politi~al AnalJ.sii, vul. 2, ed. Jama A. Stinuon. Ann Att>ot: University of Michigan Ptcu.

Gcddet , &rbon, I 99t. "A Game Theotttic ~iodd o( Rc(orm iR t.lin Amerk•D Oemo:nci'e1." Anuric-an Pofirlcol Sd<.M< Review 85:371-392.

Gec-rtt-. Clifford. 1973'. -rl:iick Ocscriptlon: Toward •n lnletprcLive Theory or OJIUJrc,. In The lnurpl'l'Ulll'oll of Cult&Jrl'S, ed. Clifford Gecttt. New Y<Xt: Basic- Books.

Gcori:c, Alexander L. 1979. "Cue.Swdies 8-nd lbcocy Devel<ipmtnt: The ~{ethod of Sttuctu:rcd, Fo.cu.sed Coqui&oo." In l>lp~acy: New Approo<"-ha i.n Hi!o·klry, 17w-ory. f1l'l4 Policy, ed. Paul Gordon 1.Aun:n. New York: The free Pre11.

George. Alc.xaodec t.., and Timothy J, Mc~owo.. 19SS. ·ease Studic• and Theories of Organization.I ().:cision Making.· Advances 1·n fn/(Jnp.tuilJn i>mel'.s.ring in OrganJUJ.ti()n.t, v()). 2.

San~ &rb•ta, CA: JAJ: Prcs.11. (;cn::ffi, Gary, aOO Donald L. Wynv.n, td1. 1990. Mc:rtf.lf~Juring

Miro.ck.&: PotM ()j /ndustrlali:IJliOfl in IAnn Alr<Ln'l a -4 ~ Ario. Princ:etoo: Prino::ecoo University Pre ....

Coldstono, l•ck A._ 199l. R<V041lion and Rt:bdlion in~ &riy M~m WQrld. &rkeley and Lot Angeka: Uruvcl'$it o( Cali(omia Pteu. 1

Hawtd. Secphao. 1990. Padov~s from lM Pcripkry: 1>-< Poll.ti~ of Gf"O'Wrh ill tJ:" Newly lntbutrlaG:ing <:cwvrie1, ltha.ci· CotncU Univ<:n.i1y Prut. •

Hampel, Fnnk ll~. ct al. l 987. Robwt StaJi.stiu: 7"4: Appr(Nk)i &ucd on lnjLl.enc:e,Funcdon.s. New York~ John 'N'iky.

Hartwig, Frcdcri;;k. with Brian e-. Oc4ring. 1979. E.xplorGJ(Jry ~ Analydl. S•go·UniveC3ity Pe.per, Serie• No. 07-0)6, Beverly Hilla, CA: Sage Publj(:s1iocu,

Hecbicher. Gunn.st. 19$7. Tkt Study o/Om11'4rarl''' Go~·cmnwuw Politics. London: Geotte. Alica aod Uawi.o.

Wbbt, Ooug_l,1 A., Jr, 197&. ·0n the Polifu: . .t &:onon'iy of Lo~·Ru,c Trends ln StrikG AC-ti'o'hy. • Br!JU/I Journal of PoU1l~al Sdenc< 1dS3-175.

Hic:b, Ale.under. 1988. "Social Oc.moc:ra~i<: Corpontl-a.m e.Dd Ecooomic Growth.· Jownal of PoliJJcs 50:677-704,

H1cb:1 A!ex.aQdcr. and William Da\'ld Pstul'"JIOn, 1989. 'On ~e .Robual.tlit-ti <Jf the. Left CotpOntiJl 1'-iodel or F.conomK Orowlh: Jounw! of Poli4cs Sl:662..S1S.

Hoolc, FttOCi.t W. 1978. E~'on Rts<ol\!h and De..·,Jqpmatl Aca~;tfu. Bcvcriy Hille, CA: S.gc Pubticatiooa.

Jackmt.n, Robcn W. 19&5. ·cro•N•tiona.I Stat.illieal Rcx.an;h aod the Srudy o( Comparative. Politic• .... Am.4!rican loumal of PoUdcal Sclcnu 29:161-82.

)•c k.nun, R.obcr1 W. 1987. 1'he Politics-of F.conomic. Growth in lndUAriAl Dcn:xx:·,..c:ic.11, 1974--30: l.A:i\is:i: Strength or N'oo.b ~a QilJ• }Ocund.J of ~Dlfo'cs.49:241·56.

Joctm.w,'Robe~ W. 19&9. "The P<Jli1io1 of Boonomio Growlh. o.>:e Agai.a. • Journal ()/ Pt1l1dct $ l :646-661 •

l ac-boo, John E. l992. ·Elltimation or l\.{odcta with Vari1bk CQef(icicou.• In Pollo'c'al Analy.sl$, vol. 3. ¢<1 , Jan,e1-A. Stirn.on. Ann Arb<it': Uni..,crsity of Micbigao ~u.

K.alkberg~ AMuc L. 1966. Ille Logie o( Compari$0n: A ~tctliodologieel Noto on the Comp.tr.live Study of J>Qlitic.al S).'stents. • World Polidc-.& 19:69~82.

King, Gary, Sidney Verba, snd~obetl O. ~.tnc. 19~. Sd<ndfi< lnfermcl' fr• Qualiuuive Rc.se'1rch. Uopublisb.ed l'l"ltlDu~pc. Department of Go"Yeminl:·nl, Hatvaril Univctticy.

Lakoft", George. 1987. WQll'WI, Fi~~ and Dang"rc"'' 'Thb:tfS.' Wiiar <:aJesorla &-.~al about~ .. 'lind. Chic.ago: UnfvcAit)' oC

Chicago l"rcu. Unge, Pck:c. and Geoffrey Garren. 19&5. 'The Polilics or Gro~:

S~tcfic lftlerac.tion and Fhooomic 'fcrfQMl'.Jlf)ee in AdvAEKcd (J)duitrial Ocl'l'IOCr.cie1, 1914-1930. • /oumal ef foUd~s 4"792-82 7.

l..aJta:e, Ptu:r. Uld Geoffrcy G~u. l9S7. 1be Polilie-.-ofGCOWl.b Re<:oa.:sidereJ. • JounJ.DJ 'of Po4tlc-s 49:l.S7-74.

LauwcU,, lfarotd O . l968. -rhe Future- o( the Comp.traiive Method. • Cortfpamlive Politics l :3-18.

U~cr900, Sl.tn.lcy .• 1991 . ·srruiU N'a and Big CotKlu.Jioiu; /\n Examinatioa of Ute Rc.alaniog in Companitive Studiet S.acd oo a Sn.all Number of Ca.sc:11. • S«ial Fore.es 70:307~20.

Lijpb•rt. Arend. 1971. ·compmUve Polllica and Con.v-,..tivc Method." Am<rlcan Polillcol &:it.n.cc ~view 65:682-93.

Ll.jpb.tr\, Mend. l975. lbe Co.mp•n.l>le·C•.k.I Stretegy in Compar.tivcJt.etcateb. • Cofrtpprasivc Polidcal Swdiu

1:153-77. Ujph.et\, Mnd. 1990. '"The- Poli1ic:il C<Joteqocnces o( EJc1>t.oRJ

Law1, 1945 .. 1985. • Am.trlcan Politic(1./ $dt!ncc Re\;ew

84:4'1·96. Lipset. Scymo:ui:- .. fattin, and Stein R.okh.o. 1967. ·cteavaie•,

StnKturc•, Pattr Sy~me:i 11.Dtl Vour Alig~ An ldlt<>-

d..UO... • !ft Pony~ tnd y....,. A&_...,. eclt.

Seymour Martin Lip.et and Steio R.otb.o. New Yoft: Fru ........ Lu<bb<n. Gt<ro<y M. 1991. Ube,..,,.. F4Kism. or Sodal ~: Social Oouu and <N PoJidcal Orl,V.. of R11lm<s 111 lflkrwor Euro~. Ne.w York: 0.tford Univcni1y ........

McKJnncy, John C. l966. ConstnlCdW "J'ypologyand S«ltsl 1"l'Oty. New York:: Meredith PubliJbio,g Company.

~it.CTiU, R.ich1rd. 1970. Syst<m.ofic Approadi,es 4J O:Jmporodw Polldt:s, Cbic1go: Rand M~N•ll)'.

~1cniu, Rlch1nf, 1od St.tin Rollao. ed•. J966. ComparlA1 Mu!ON: TM Use of Qu.wuis.¢vt Dal(: in CrouvflaJion-;;sl hsto~lt.

New H•Vcl'.l! Yale Univenlty Prca. M~l, lollnSw.an. (184311974. A Sys1<m o/Lofic. Ton>alo'

Unt.,.e:nity ofTOl"OrtO Prua .. Mooooy, Chri@rhcr Z... tod Rohen D. Dunl. 1992. ·&cu1np

ln(ct"Cntc:: A Pn:fimiMq Moak C4do EvaNatioa. • Paper

prc.Mnkd et lhc afttllJ&f mcctin&• of Che Amcriuia Po6ciul Sctctw:e ~ Ohago.

Moon, Bani.""°n. 1966. $ocioJ OrilfrJ of D1ft0Uln1Jp Md OetMCrocy· Lord and Pu:.scriJ in IA< MtNtittt o/tN N<Jtd<m World 8oMG: ~on f'rcu.

1\1olkller, Frcdetiek", and Joho Vt', 'I\Jhy. 1977. DllJa AMlyN and

Rttr~~rioo. lie.ding, MA: Addi..on-We.Jcy. O'Donnell, Cuillcnno, PbiJippc C . Schmiu.cr, ind Uwrcocc Wl1itchc•d,

e:dt. 19&6. TraAfitlonsftomAuslwn'l4riM RMU. 0..llhno"': John lh>pl::ina Un.ivcnity Pre».

Vci,ec. JclTroy. 197$. Agrari'11! .ikvoluzi0fl: Soo"a.I Mow.mm.ts and E.rf")rt Agd cultun ttJ W UNkrdci.·cl()~d World. New York: Ftc.c. Pr-tu.

Pnc• onk:i.. Ada1n. 1987 . • ,._{et.bod$-Of Crou-NctlonaJ R~.a.can:h, 19'1().1983· All Ovctvic""··· ~{cinoU'.Dicrtu. HUI.IN. W~ilcr , and Ari!!IC Berthoio Amlil. ed.$, OJm.parrulv~ Pt>Oey ~1,arrlt: Uarn!ngfivna ErpuWtc~. Btoott'aclJ. VT: t;o..cr

Pfu•onl., Ad.am 1991. CJcnw>aoey end du Mrukt.. Po61kol OltJ & ..-I< Rcfa<wol In £4stLrn f;Mropc a.J '-""" "-"<•. C.mbride<: Cambridge Uaivenity Pre•.

Ptuv.on.ki. Ad..n'~ and HenryTcunc-. 1970. ~ LDfk of Comparati'oL Social lr..qulry. New Yort..: J<>On WI.Icy.

baln. Qartc.s C. 1987. »it: <Mr-.paroti'-c Mctho4: Mo'V.I MyMd Quohtaow aN/ Qumuito:iw Stro.ugi~A . Ekrkcley and LbJ ~elet• Univcl"lity of CAlifoml.a Prcw.

Rotbn, $1.ein 1970. Ciduns. Ek<rions, P41'fl't1: ApproM.lt~1 ID Ill~ Comp<Jrad'tlr Study~/ fr«e;u6 of Dn«lopmc111. New York: David ~i .:Kay.

Ructchcmcyct, OJctri-::h, Evelyne Hulkr Su-phcn1, and Jchn 0 . S1c9hcn1. l992. Capi«i!lst Dei>elopmcnl t»td D<.r1t«roq. Ch1ca~o: Uftivetsity of Qica,o PttJt.

Ru#OW, O.nkwatt. 1968. ' Modemi.tat.ion and Compantlvr. Politic-1: Pro.p«o in :Ret«rc:b and Theory. · Compa~w. Polida l :J7-51.

S•Mn. GtOVal'lni l970. "C<Jocept ~fuformatioc in Comp•nl.ivc: Poll•K•.· AmericCl'S Po!iricoJ Scitnc' ~ 64:1033~S).

Sanon. G"tOYanni. ed. 1914. Sod4! SekNc l:otte'f"S: A Sy~t AMlyril . lkv<riy lmh. CA: Sag<.

S&nori, Giovanni. 1991 . ·eomp.ruc •ad Mitcomp,arina. • J"""'41 <>j 7Non~ />¢iz;q 3;l.43-S7.

Sanori. Giovanni 1993. I~ Modd ~tW., and Lumin,c from En-or.~ J(JU.rNJ} q/Tr:.eore='col Polldu.

Sanon, Gtovan.N, Fred W. Rit1JS, a.Qd Hcttty Tcu.ne. 197S. Towtr of

Babtl: On IM DtfitUdorr and A/lolyN of ~'/HI In IN Sotfol ~itNts. lntctN.liocul Swdic1 A.No:iltioa. Occ1.sional Pi per No. 6, Univer.by or Piu.t>ur}tb.

CoJJ;u 119

Sdunitur. Philippe C. 1911. •lalLtt.M latumcdi.Uoeud R.ciU:nc::: Gcwc.mability in Coa&c~ry WULcna Europe a4d Nodb Am<riu. • 1o °'Jonl:Jni '""""' 1n w.,...~ ir-op., ..s. Su:uncc D. Be.rice. C.mbridec: Cambridp Uaivcnity ........

Skocpol, Thedt. 1979. S.- and Social-...,,. A Comporad,.

AMlyN •I FNn«, l/Mulo, and °"""· C.mhndg.: C.mbridge Uoivertity PR ...

SkO<pOI, Thcd.t. 1984. Yul°" <WI MeW>d In f/ul()rl<ol S«lology. · ~mbridtc: aod New YCNt: C.n\bridao Vniveraity J>au..

Sko,:pal, Thed.i, and ""f• rgarct Somen. 1980. -iM UK.a o(

Compan.tlvc: ffUc:ofy ln MIC:roaoei..I lnqui.ry.• ~w_ S:udlt:1 In S.OCldJ tJ1td Hi.l""Y 2'2:174-97.

Smelter, Neil. 1961. J'hc: Mc:tbodolos'Y of ~atatiw. Ana.ly.U. of E.!oooa:iic ActiYict. • la Eu.,_ "" Sodolotkal lr..LuplltaJiOt'I , ed~ Neil Smelacr. En,kwoodCUffa., NJ: Pm:b::e-lWl.

Sm<I><•, N<il. 1976. ~"< Wc<Mds In Ill< SodaJ Sdco<a. Eltale""'"'1C1ifft. NI: .......... Hall.

Scifte~ Ml>•• I.. 1978. ""-1/NI N- In S<>O.i /r..-y. New Yori:: ~- Prus.

Tilly, Cl.alb. 197S. Thi: Fol'rfWUlott of/llal1<JMJ Sl4k1 M Wuiem &.rope. Princet00: Plmcc&On UfUvcndy PR:•.

Tilly, Clarlcs. 1984. m, ~1. lArze houuu, 111'g~ ~. New Yort: Ruue.11 S1ac.

Trimbc>rgcr, Ell«:1 Kay. 197&. Rew>J;.ufonjr'Otf1 Abo~: Mili14ry Bun~ro.u tiNI DevtWpm.mJ In JapM, Tu'*<}'. Egyp1. and Pt:ru. New Bf\Uuwick, NJ: T rautclion Book.II.

Tufte, Edw•nl R. 1970. l1>e {Jwtlnlluul•~ AMlym •/Social Problnns. Rc.ading, MA: Addil()n-We1Jey,

V1Ui<r, l•111, <4. 1971. Wmpor.#r< MclMdl In S«loloty: ~«'!' on Trmd.t and AppJicodfJN. Bcct'clcy and LOI AngcJe1; Uttivenit)' orCafiforoia Ptt ...

Vert>., Sidaey. 1967. ·Some DUcmmas lo eoo,pantivc Reaea.n:b. · World Politia 20:111·27.

W~osi:y. Httcld L 1976. TM 'll<w o,,po..-.' ~. and du Wdfan Stale'. Pro(c..uional Papen iA Coakmporvy ~I Soc. leg)' Scnu. 8e:vcri)' Hilll. Sa10 Pl.Jbliutioc».

Yio, Robcti K. 1984 • . Co.u Study ks,~lt l>t-"1" ""'1 Alt:rhod.t. Afftlie.d Soc:U.l Ruca.ttb Method• Serie•. vol. S. &veriy Htll1; Sage Publiu\iona.

Ztld)ub, Jr., Monil. 1971. •Inr.t.Hiaiblc Comp1.Nons. · ln CcmP<JrlZli~ M~:A.ods i11 S«k>/40: En"Ys ot1 Trerttl.s and Appli<aliOt'lS, ed. lv•o VaUict. lk:rkcky aDd Lot Angclu:

Uah·crsity o( California ~ ...