See discussions, stats, and author profiles for this publication at: [Link]

net/publication/325253765

The Comparative Method

Chapter · February 1993

CITATIONS READS

0 1,405

1 author:

David Collier
University of California, Berkeley
75 PUBLICATIONS 11,264 CITATIONS

SEE PROFILE

All content following this page was uploaded by David Collier on 19 May 2018.

The user has requested enhancement of the downloaded file.

 1n Ada W , Finifter, ed. ( 1993) Political Science: The State of the Discipline n.
Washington, D. C.: American Political Scienee Associafiou.
5
The Comparative Method

David Collier

Comparison is a fundamental tool of analysis. It Choosing lo study few cases routinely poses the
sharpens our power of description, and plays [Link] problem of having more rival explanations lo assess than
role in concept-formation by bringing into focus cases to observe, or the quand~ of •JJW1y [Link],
suggestive similarities and contrasts among cases. small N" (Lijphart 1971, 686). Elementary statistics
Comparison is routinely used in testing hypotheses, and it teaclles us that as the number of e<plaoatory factors
can contribute to the inductive discovery of new approaches the number of cases, the capacil)I to
hypotheses and to theory-building. [Link] the explanations through statistical
The forms of comparison employed in the [Link] rapidly diminishe.s. This problem bas stimu-
discipline of political science vary widely and include lated much discussion of how most productively to
those contained in statistical analysis, e,perimcntal analyze a small N.
research, and historical studies. At the same lime, tlie The late 1960s and early 1970s ssw a boom in
label "comparative method" has a standard [Link] _ writing on comparative method (e.g . , Merrill and [Link].o
within the discipline and in the social sciences moii 1966; Klllleberg 1966; Verba 1967; Smelser 1968;
broadly: it refers to the methodological issues that arise Lasswell 1968; Przeworski and Teuoe 1970; Sattori
in the systemalic analysis of a small number of cases, or 1970; Menitt 1970; Etzioni and Dubow 1970; Lijpbart
a "small N. • 1 This chapter examines alternative 1971; Vallier 1971; Zelditch 1971; Armer and Grimshaw
perspectives on the comparative method that have 1973). This literature established a set of norms and
emerged over roughly the past two decAdes. Although practices for smnll•N research, proposed alternative
the primary focus is on discussioos located in the fields of strategies for conducting such analyses, and creared a
comparative politics and international srudjes, the base line of understanding that has played an important
application of the comparative method is by no means role in the ongoing practice of smnU·N studies. This
restricted to those fields. chapter assesses the issues of [Link] method that
The [Link] to analyze only a few cases is have been debated in the intervening years and considers
strongly influenced by the types of political phenomena their implicatioos for ongoing resea(Ch. The point of
under study and how they are conceptualiwl. Topics for departure is Arend Lijphart's (1971) article ·comparative
which it is -productive to examine relatively few cases Polities and Comparative Method. · Among the studies
include revolutions, particular type.s of national political published in that period, Lijpbart's piece stands out for its
regimes (e.g., post-communist regimes), or particular imaginative synthesis of basic issues of comparison and
forms of urban political systems. This focus on a small of the relation between comparative method and other
number of cases is adopted because there e xist relatively branches of methodology.' It therefore provides a
few [Link] of the phenomenon under consideration that helpful framework for examining, and building upon,
el\hibil the attributes of [Link] to the analyst. new developments in the field.
Alternatively, some analysts believe that pol itical A central theme that emerges in the discussion
pbeoomena in general are best understood through the below is that refinements io methods of small-N analysis
careful [Link] of a small number of cases. In the have substaJltially broadened the range of techniqu~
fidd of comparative and international studies, tho practice available lo comparative researchers. The most fruitful
of focusing on few cases bas achieved greater legitimacy approach is eclectic, one in which scholars are willing
in recent years in conjuoctioo with the rise of the school and able to draw upon these diverse techniques.
of ·comparatlve historical analysis,· in which small
numbers of countries are studied over long periods. This
close scrutiny of ea4' country limilS the number of
national cases a schol:t.r can consider. 2

Electronic copy available at: [Link]

JOG The Comparative Method

Synopsis of Lijphart statistical method for evaluating hypotheses, duo 10 the

lack of experimental control and the problem of many
Lijpbatl defines the comparative method as the variabl<:5, sm.U N. Yct it does offer a stronger basis for
aoalysis of a small number of =s, entailiog al least two evaluanng hypoth~ than do case studies. Despite the
obs¢rvarions., yet too few to permit the application of constramt of addressing more variables than cases the
conventional statistical analysis. A central goal of his comparative method allows systematic compariooi: tb3l, ii
article is to assess the comparative method in relation to appropriately utilized, can contribute to adjudicating
among rival explanations.
three other meth<llls--experimental, statistical, and case-
Ahboug'b tbe data requirements of the
study- and to evaluate these different approaches by two
comparative method may be much greater than for case
criteria: 1) how well they achieve the goal of testing
theocy through adjudicating among rival e·xplanatioos, and studtes, Lijphart argues that !hey are less demanding th2n
for experimental or sta![Link] research. He therefore
2) bow difficult it is to acquire the data needed to employ
views. the comparative me~od as most appropriate in
each method (see Figure 1).
research based on modest resources, and be suggests tb>t
The experimental method has the merit of
studies using the comparative mQthod might ofteo serve
providing strong criteria for eliminating rival explanations
as a first step [Link] statistical analysis.
through experimental control, b ut unfortUnately it is
impossible to generate appropriate experimental data for lf at all possible one should generally use lhe
most topics relevant to political analysis. The statistical statistical (or perhaps even the"cxpcrlmcntal)
method has !he merit of assessing rival explanat ions method instead of the wukcr ¢<lmp:i.C3tive.
through the weaker but still valuable procedure of method. Bot often, given the inevitable
s-tatistical control, but ii is often not feasible to collect a [Link] of time, energy, [Link] financial
•-uffieieotly large set of reliable oota to do this form of rc,wu~. the intensive comparati..,e analysis
analysis. or a. few cases may be more promising than
The case-study method bas tbe merit of a more. superficial $talistical ana;Jysis of
providing a framework in wbic.h a scholar wilh modest many cases. 1n such a situation, the most
CruiLful approach would be to regard the
time and resources can generate what may potentially be
coml)Orative analysi, u the first <tage of
useful data oo a particular case. UofortUo>tely, r'C$carch, in which hypotheses [Link] carefully
opportuniiies for [Link] testing hypotheses are far f()[Link], and the stati$tical analysis as lhe
more limited than with the otber methods. Yet Lijphart second stage, in which these hypothese:.-. are
(pp. 691-93) insists that case studies do make a tested in a., large a sample as -possible.
contribution to testing hypotheses and building theocy, (1971, 685)
Md he offer.; a suggestive typology of case studies [Link]
on the nature· of this contribution. He distinguishes Lijphart also proposes solutions to both sides of
amoog arheoretical case studies; in1erpre1a1ive case the problem of many variables, small N ( 1971, 686 fl).
studies (that self=osciously use a theory to illuminate a With regard io the small number of c..ses, even if
particular case); hypo1hesis-genera1ing case studies; researchers stop short <if a statistical study, tbey can
1heory-confirming case studies; 1heo,y-infirming case nonetheless try to increase the number of cases used in
studies (that, although they cannot by tbemselves assessing bypotbeses. With regard to the large numl>er of
disconfinn a theory, can raise doubts about it); and variables, be suggests two appto~ches, First, analysts
deviam case analyses (that seek to elaborate and refine can focus on 'comparable cases,• that is, on cases that
Oieocy tbrougb a close examination of a case that departs a) are matched on many variables th.1t are Mt central to
from the predictions of an established theory). Lijpbart the study, thus in effect ·contr<illing' for these variables;
emphasizes that 'certain types of case studies can even be and b) differ in terms of tbe key variables that are the
considered implicit parts of the comparative metbod' (p. focus of analysis, thereby allowing a more adequate
691), and to the extent that the assessment of hypotheses assessment of their influence. Hence, the selection of
does occur in some case [Link], it is often because the cases ac<s as a partial substitute for statistical or
case studies are placed in an implicit or e,plicir experimental control. Second, aruilysts can reduce the
comparative framework. Yet even within this number of variables either by combining variables in a
framework, be emphasizes that fuldings from a single single s<:al.e or through theoretical parsimony, that is,
ca.'!e should not be given much weight in the evaluation of through developing a theocy that focuses on a smaller
hypotheses and theory (p. 691). number of [Link] factors.
Th<> comparative method, as defined by Lijpbart, 'Thus, Lijpbart provides a compact formulation
bas an intermediate status in terms of both his criteria. It of the relationship between the comparative method and
provides a weaker basis than the experimental or

Electronic copy available at: [Link]

 Collier 107

Flgur• 1. s;1uating the Comparative Method u of 1971: Lljphart·, Schesn<

CHe Study Method Comparative Method Experimental Method

Defined u: Syste-- Merit: Eliminates rival

Meril; Permits lntefl-
sive examination or [Link] analysis of small explanations lhrough
cases even with Hm· number of cases experirr\en1al control
tied resources ('' smaJI-N" analysis) lnher•nt Probl•m:
Inherent Prob!em: Merit: " Given fnevft· Experimental control iS
Contribu1es Jess to able scarcity of time, impossible for many o.r
building 1heory ll'lan enetgy, and financial most topics of
studl es wiln m()(e resourc&s, the [Link] relevance to fietd of
eas.&s slve anaJysis o-t a few comparative polities
cases may be more
Typea ot c,se Stu·
promising [Link] the
dlu·:
1. Atheoret1cal
supetficial s1iitlstlcaJ Statlsttcol Method
analysis of many
2. lnlerpret[ve
caS<ls" (Ujphart. p. Mtrft: Assesses rival
3 . Hypolh~~iS·
685)
1 generating explanations throuoh
4 . Theory-confirming Inherent Probl•m: stadstlcal control
1 5. Theory-mfltming Weak capacity to son
lnhertnt Problem:
I (i.e., case [Link] out ri.,aJ explanations,
Oitticutt to collect ade-
specifically, the prob-
i :~:~~·::;l~ally) lem of "[Link] vari-
quate information in a
sufficient number ot
!0. Deviant case studies ) ables. few cases"
cases, duo to limited
Polent11I Solut1on1:
time and resources
1. ln<:cease number or
cases
2. Focus on compar-
able cases
3. R&duce number of
variables
a. Combine vari-
abl. .
b. Employ more pat·
simonious th80l)'
108 The Comparative Method

olhcr melhodologies, and he offers solutions to lhe these approaches (pp. 196-197). [Link] ~ycle arises in
characteristic djlemmas of lhe comparative method. response to the problems that emerge as scholars push
each approach up to - or beyond - the limits of its
usefulness. For example, a •parallel demo,ostratioo"
Further Perspectives on Small-N scholar might introduce a new theory and show bow it
applies to m.,ny cases. "[Link]-testing" scholars,
Analysis wanting to specify the conditioos uoder which the theory
does not hold, eould m.'3cc further comparisons with the
Tho two decades following Lijphart's study have goal of discovering these eooditions. Hypothesis-testing
seen the emergence of new perspectives on sma!J-N studies that too brashly compare cases that are p(ofoundly
analysis, as well as a renewed focus on methodological different might, in tum, stimulate ·CQntraSt of coot~xts·
alternatives already av,ulable before be wrote his article. scholars to examine more urefuUy the meaning of the
Though many of these innovations appear io. work difforences among the cases. ft is thus useful to lo<>k
explicitly concerned with the comparative method, beyond an exclusive focus on the role of comparisoo in
conventionally understood, others appear io. writing on broad causal analysis, to an understaoding that
the experimental, statistical, and case-study methods. encompasses the different elements in this n:scorch cycle.
The result bas been an intellectual cross-fertilization of lrus is .n ot to say that assessing hypolhcse,s doe,i
great benefit to the comparative method. Figure 2 not remain a P.a,[Link] goal of comparison, and many
provides an ovecview of these innovations. scholars insist that it is the paramount goul. Yet this
broader perspective offers a valuable account of how
Innovations in the Comparative Method comparative work proceeds within a larger =r<.'h
community, pointing usefully to the interaction among
Innovations in <be comparative melhod can be different goals of comparison.
discussed in terms of the issues introduced above, encom-
passing lhe goals of comparison, lho justification for JuSlijicatwn for Small N
focusing on few cases, and lhe problem of =Y
variables, s~ll N. A =ond trend is toward a mora elaborote
justification of a focus on relatively few cases. Lijpbart's
Goals of Comparison rationale seems in retrospect rather modest, in that it
emphasizes only the problem of inadequate resources and
A central and legitimate goal of comparative treats the small-N eomparisoo as a way station on the
analysis is assessing rival explanations. However, as route to more sophistica!ed. statistical analysis.
Theda Skocpol and Margaret Somers (1980) argue, A very different defense of wocking either with•
comparative studi"8 should be understood not merely in
tecms of this single goal, but in tecms of three distioct,
small N or with = studies had previously been
available io arguments favoring a ·configuralive·
yet ultimately connected, goals.' The first is that approach (Heckscher l957, 46.- Sl, 85-107), and this
considered above: the systematic examination of perspective was elaborated a few years before the
[Link] among [Link] for the purpose of causal [Link]·s analysis in Sianey V«ba's
analysis.' The second is the examinatioo of a number of (1967) review essay advocating the "disciplined
cases with the goal of showing. that a particular model or configurative approach.• In evaluating Robert A. Dahl's
set of eoncepts usefully illuminates these cases. No real Political Opposirions in \Veste:m Democracies ( 1966},
test of the theory occurs, but ralher the goal is the Verba points both to the [Link] of the hypotheses
parallel denwnsiration of theory. This use of eomparison entertained io the book and to the difficulty of assessing
plays an important role in the process lhrough which them adequately. except through a close eonunand of the
theories are developed. The third type of comparison is cases, leading him to. advocate this disciplined
the examinati.o n of two or more cases io order to eonfigurative mode of research. Verba's formulation is
highlight bow different they are. thus ·e stablishing a appealiog [Link] he is eoneemed with systematic .
framework for interpreting how parallel processes of hypothesis testing and theory building. At the same Ul!ll>,
change are played out in different ways wilhin each he links this priority with a more explicit approeiation of
context. [Link] conrrast of conte:xts is central to the more the difficulty of testing hypotheses adequately and the
0
ioterpretive• side of the social sciences and reflects yet value of properly executed case studies in providing
another way that comparison is frequeotly used. subtle [Link] of hypotheses.
Tn addition to providiog a more multifaceted It might be claimed that the difficulty of
accow,t of the goals of comparison, Skocpol and Somers adequately testing bypolheses ultimately derives from the
suggest the intriguing idea of a research ·cycle· among
 Collier 109

Agu,.. 2 . [Link] R~kvanl to~ Comparative [Link]

Comparative Mtlhod Experiment1J MtthOd

Broadened UndemtandinQ of Type• ot Olttu1lon of Older Ide.. and lntrodue·

Comparttlve Studiea lion of Hew ldcH on [Link]..
Experimental OoaJgn
1. Emphasl$ on 1merpretrve ul\Clerstanding
1. MetnOOOlogy o1 quasl•.xpetiments and
2. Idea 01a " resaa,c-..n eyc-!e" among the
interrupted time-cseries deS>gn beCOme,
rype.s- (Skoqx>J and Somers)
mote wkioty known
Funtu~r Ju11Jfk:.11Jona for Focu, on•
2. Analysis of the Connecticut Ct3.'ci<[Link]
Small-N on $peeding as o•empj_ar ot integuptod
Case Study Method 1. lo pursue "d!.$<:•pfineo configurativa t\n;H,8:fies amltysis (Campbell and
approach" (Verba, remfotcetl by i... RO$$)
Hew Perspectlvu on CHe- Studies Almond end Genco)
3. Citi:;~io.~-cf idc~.: ;10CU: c;~~!,
1. Nawvo:ei'i.;6 "i ;r,~ "-~~ :ii.1,[Link] m~\;!oJ Z, To avoid proo!em ol " conceptuol experiments- encourogod by ooc;;tlfica~ion
(Campbell) Sl(OIC-hlr,if' ~S!!tOri) of evaluation [Link]
2. Refinements In L1johan 's typology ol 3 . To faciti1a1c "lhickdescnp11on" B!ld 4. Proposed statistical solution to problem
case studtes (fcks.1e1n, George}
La o~hef forms of interpre11vo understand- of selection bias. in quasi--experiments
Ing (Geeru and many otheus)
3 lmprovttmem of causal an3tys1s in case (Achen)
studies througn ..process ,racing" 4. To echiev& ti;na_1y;1c depth ot cas.-
(George and [Link]) o,,en1ed approach (Ragtn) Statfatlul Me·thod
.c. Cntique of value ol c a.s ~ .stvdies in O~bates on Solutions to Problem ot
New w,mlng& and Nbw ·So1utlona
assessing rat1<1na1 choice t~ry (Achen Many Varlabfu, Stnall•N
1. Criticism of [Link] ~a1istJ.c•1 practice
and Snidat) 1. Yatu-e of inc,[Link] numberot ~
in [Link] soaal scienc.e.s (Freedman)
2 Comparaole cases vttr~s contrasting ,,_ 2. New sla1istical tech.l'liquss re~vant to
case$ (U)pnart versu$ Pn:eworski and
small-N [Link]:Jy$is
f eune)
3 . Aooucing r'lumoer of vari&blH in con•
3. Etfon ,o ro!ine statis11cal analysi$ of a
small-N in tl'le debate on CO!'P¢ratism
junction with using slronger theory
and economic groWlh In Westcm
Europe ([Link]-riicks-
Panerson)
110 The Comparative Method

problem of limited mouices discussed by Lijpbart. If within each system (pp. 26-30).' 1rus approach make,
enough talented researchers worked long and hard, they the invaluable contribution of providing a theoretical
could carry out a Polirical Oppositions sl\Jdy for many rather than an idiosyncratic and case~[Link], OOSis for
dozens of countries. Yet the problem here is somewhat analyzing differen= in causal patterns. However,
different from that emphasized in Lijpbart's initial formu- extending an analysis to additional cases on the basis of
lation. II is J10t so much that mouices are limited, but this procedure again requires a painstaking assessment of
that constructing adequate comparisons has proved more each new context. Thus, Paewoi'ski and Teune provide
difficult than bad olt¢n been thought in the 1960s and a valuable tool for adequately analyzing a larger numb,:,
early 1970s, in tho initial days of enthusiasm for of cases, but their approach again shows that this must be
comparative statistical research. Among these done with caution.
difficulties, that of U,e valid application of concepts Since 1970, the renewal of a Weberian concern
a,:ross diverse contexts has beeo especially vexing. with interpretLve unden.'landing, i.e., with dec,phenng the
Within the litecature on comparative method, a meaning of bcliavior and institutioo.s to the actocs
key step in elucidating these problems of validity, and involved, has also strengthened the justification for
thereby strengthening the justification for a small N, is advancing cautiously with one or very few =.s.
Giovanni Sartori' s (1970) classic discussion of "Concept Clifford Geertz's {1973) label "thick description" is
Misformation in Comparative Politics," the basic themes commonly evoked to (efer to this concern,• and this
of which are el.a borated in his later book Social Scien'Ce focus has appeared in many guises relevant to political
Concepts (1984). Sartori suggests that the application of research, including Gabriel Almond and Stephen J.
a cone<:pt to a broade, J11Dge of cases can lead to Genco's analysis of "Clouds and Clocks' (1977) and
conceptual "stretching, · as some of the meanings Skocpol and Somers' "contrast of cootextS" approach,
associated with the concept fail to fit the new cases. The which encompasses studies that use comparison to richly
concepts that can most easily be applied to a broad range conle><.tualiz.e research findings. Charles C. Ragin's TM
of cases are often so general that they do not bring intti OJmparaJive Mtthod (1987) explores ·another facet of this
focus the similarities and contrasts among cases that are concern in bis analysis of the 'holistic" orien!:Jtion of
essential building blo,cks in wonbwhile comparative what be calls "case-oriented· [Link] and the complex
analysis. Consequently, a study focused on concepts that problems of "coojunctural causation" - that is, causal
are carefully adapted to this "finer slicing" of a given set pallcms that vary a,:eording to the context - to which
of cases should be extended to other ca..<es only with great configurative scholars are typically far more sensitive.
caution. From this perspective, it may be argued that the Finally, the intellectual success in recent years of
most interesting studies will often be those that focus on a the school of comparative historical analysis bas played
smaller number of cases. an important role in lcgitinl.,ting a focus on a small N.
With regard to the problems of increasing the 1rus appcoach was pioneered in works such as Reinhard
number of c,ises under study,• Adam Przeworslci and Bendix (1964), Barrington Moore {1966), and Lipset and
Henry Teune's 77,e Logic of Comparative Social lnqufry 'Rokkan (1967), and moro recent works include Rokkan
(1970) is a major source of insight. Although they argue (1970), Tilly (1975), Paige {1975), Bendix (1978),
that achieving a high level of generality should be a basic Trimberger (1978), Skocpol (1979), Bergquist (1986),
goal of social science, their framework is centrally Luebbert {1991), Goldston.• (1991), Collier and Collier
concerned wit!,_ the difficulties that can arise in (1991}, and Rueschemeyer, Stephens, and Stephens
generalizing beyond an initial set of cases. With regard (1992). Methodological statements focused on this
to problems of validity, they ad,·ocate the use, wbeo tradition include Skocpol and Somers (1980), Skocpol
nee<OSSary, of "system-specific" indicators that serve to (1984), Tilly (1984), and Ragin {1987).
operatiooalizo the same concept in distinct ways in The. p'ltlicular form of analysis in these studies
different contexts (pp. 124-130). For the scholar seeking varies considerably, as suggested by Skocpol and Somers'
to move toward a larger sol of cases, the poteritial ne<:d typology noted above. In varying combinaiions, these
for system-specific indicators oecessit;ites the close studies employ both rigorous qualitative comparisons that
examination of every new case. extend across a number of nations, and also hjstorical
Przeworski and Teune also address the problem analysis that often evaluates each national case over a
that as the analyS1 incorporates more cases into• study, number of time periods.9 This t,adition of research thus
distinct causal patterns may appear in the new cases. To combines well-thought-out comparison with an
deal with this problem, Przeworski and Teune advocate appreciation of historical conte>t, thereb:,, contributing to
• replacing proper names" of social systems by identifying ao effort. to ·rustorici:ze• the social sci.:.nces.
those systems in terms of the e,planatory factors that Although the uses of CQmp•riS!>D in this literature
account for why causal relations take a particular foan are diverse, as Skocpol and Somers emphasized, it may
 Collier 111

be argued that a major .consequence of the growing whlch employ causal tests that are weak, unconvincing,
in,por1ance of comparative historical studies is to further or inappropriate (Ragin 1987, chap. 4).
legitimate the approach that was L ijpllart 's original Yet the fact that broad quantitative comparison
concern: the assessment of rival explanations, based on has not become a predominant approach should not lead
systematic, qualitative comparison of a small number of scholars to overlook what has been accomplished. Rober1
ca.=. In light of a spectrum of studies from Barrington Jackman (1985) insists-that comparative statistical
Moore's ( 1966) pioneering analysis of the emergence of research has had more sue= than is recognized, and
alternative forms of modem regimes, to Skocpol's ( 1979) Lijphart1 s own recent work moves in this direction
study of revo lution, to Luebber1's (1991) [Link] of the ( 1990). The fai lure to seii.e good opportunities to do
c:merg-eoce of UberaJis.m, fascism, and social democracy quantitative research could certainly be viewed as being
in inteN.·ar Europe, it is evident that this literature has as much of a mistake as premarure quantification, and the
given new legitimacy to the use of broad historical fruitful d ebate ·o n corporatism and economic growtl, io
comparison for systematic causal analysis. Efforts lO Western Europe discu~ below is one of many
codify procedures for assessing hypotheses in this type of ex.a mples of bow statistical methods can effectively
analysis, such as that in Ragin's Comparcm\•e Method address interesting analytic issues. Further, the
( 1987), fur1her reinforce the plausibility of insisting on availability of new statistical 1""'1niques (also d iscussed
the viability o f small-N analysis as a middle ground below) has made it far more productive to do qu.-uitita(ive
between case studies and s tatistical studies. analyses with as few as ten 10 fifteen [Link]. Conse·
quenlly, the option of increasing the "N" at least to that
Solutions 10 the Problem of Many Variables, Small N level is still worth pursuing, and it should probably be
pursued more often.
The evolving debates on comparative method 2. Focus on Comparable Cases The
have suggesttd fur1her refinements in Lijphart's original re<:ommendation that analysts focus oo carefully matched
three solutions for the problem of m:l!ly variables, small cases has been both reinforced and cballeoged . In a
N, i .e.: 1) increasing the number of cases, 2} focus ing discussion published in the mid-1970s, Lijphart (1975)
on matched cases, and 3) reducing the number of explores further the trade-<>ff be noted in 1971 between
variables. the goal or increasing the number of cases and the goal of
l. Increase the Number of Cases Al the tinle matching cases as a substitute for statistical control .
Lijphart wrote, it was believed in some. circles lbat Obviously, if a researcher is 10 select cases that are really
comparative social science would in¢re.."I.Sio_gly be oriented similar, however that similarity is defined. the number of
toward large.. N comparative studies, based on extensive appropriau, cases is likely to become limited. _[n the face
quaatitative da.t.a SetS and rigorous statistical analysis. of this trade-off, Lijpbart opts in favor of the more
Today 1here can be no question that, for better or worse, careful matching of fewer cases, and he goes so far -as to
quantitative cross•national research in the subfield of restrict the application of the term · co111parative method.,
comparative polities, and quantitative international politics to analyse-s that focus on a small number of carefully
in the subfield of international relations, ruive not come lo matched cases. This emphasis paralle ls a ·much earlier
occupy as dominant a position as many bad expected. perspective on the comparative method referred to as the
Within these two subfields, they hold the stalUS o f one method of "controlled comparison" (Eggan 1954).
approach among many. Arthur Slinchcombe's (1978) advocacy of the
Vari9us factors have placed limits oo the success methodology of "deep analogy,· i .e., the comparative
of large-N [Link] ba.,ed on quantitative data. sets, analysis of very fow, extremely closely matched, cases
amoog which is certainly the [Link] coacem with pushes this app roach even further.
closely conu,xtuali ze<I analysis and interpretive studies. A contrasting strategy is advocated by
Broad quantil3tivc comparison lll3Y have been set back a., Przeworski and Teune (1970, 32-39) and Przeworski
many scholars discovered how extraordinarily ( 1987, 38-41). They suggesr that even with careful
time,..consuming it is to construct appropriate data sets. matching of cases in what they label a • most s imilar·
often out of propor1ion to the professional rewards that sys-terns design , there remains a problem of
seem to be forthcoming. This is por1iculady a problem •overdcteanioation, • in that this design fails to climiaatc
when the focus of analysis extends beyond the advanced many rival explanations, leaving the researcher with no
industrial countries to regions for which it is often criteria for choosing among them. They prefer instead •
extremely difficult to develop reliable data. In addition, "most different" systems-design, based on a set of cases
the quantitativt-<:[Link] approach bas probably been which are highly diverse and among whlcb the analyst
hur1 by the publication o f too many stud[es in which ,races similar p rOC<:SSes of change.•• Przeworski
concepLs are operationalized with dubious validity and suggests that the strength of this design is in par1
 112 The Comparacive Mechod

responsible for [Link] advances in the lileralUre on of mitigating this problem is to introduce greater varia-
democ('atization. such as the work of O'Donnell, bility [Link] intemal comparison.
Sclunitter, and Wbicehead (1986). Przeworslci maincains The ongoing debate on mos! similar versus most
that this literature addresses such a broad range. of cases d ifferent systems designs bas implications for the statu.,
that BJ1alysts are forced to dislill o ut of Iha! diversity a set [Link] sCUdies. Dankwa,1 Rustow ( 1968) a,gued so,oe
of common elements that prove l-0 have greal explanatory time ago in favor of moving beyond an area studies
power. 11 approa¢h, and many scholars agree that cases [Link]~ i,.,
This discussion can be placed in perspective by selected in response to the analytic requirements of
recognizing that cases that are closely matched from one particular research projecl<, rather than on the basis of 1
poinl of view may contras! sharply from BJ1otber. My geographic proximity that at best is often a poor
own iecenc work (ColJier and Collier 1991) combines the substilU!e for the. analytic m.,tching of cases. Recenc
two strategies by starting with. a set of eight [Link].n · cross-area· studies on su=sful export-led growth 3lld
American COUJ1lries that are roughly matched on a on democracy suggest lbat this alternative pen.-[Link] is
number of broad dimensions. Among the eight coUJ1tries, gaining ground. "
the 8J1Jllysis focuses on pairs of coUJ1tries that are However, the area studies approach is a booming
nonetheless markedly diffcreol. The overall malcbing business today for a variety of reasons, including the
assures that the conlexts of analysis are analytically impressive funding of area studies by U.S. foundacions in
equivalent, at least lo a significant degree, and !he paifed rcc.e,,t decades, as well as in.<titucional momentum. ln
comparison places parallel processes of change in sharp fact, from the point of v iew of the theoretically orien(ed
relief because they are operocing in settings that are very small-N comparalivist, this isnol a bod outcome. The
different in many respects. country case studies produced by area speciolists :ire
In conjunction with the debace over che merits of crucial building blocks in most comparative work, and
mosc similar and most different sysleros designs, ii is without them cross-area studies would be on far w"1ker
important to recognize that in many studies, the groUJ1d. It is essential io recognize Iha! these ease studies
conclusions reached in the overall comparison of cases beoefic greally from the inlellectual leverage gained when
are also assessed - impLicilly and sometimes explicitly - individual scholars develop, over many years, a
through [Link]-case analysis. 1n the section on case cumulative aod well-contextualized underst.,nding of a
studies below, the discussion of 'pauern matching· and particular region. Particularly in light of [Link]
"process tracing' suggests some of the forms this cakes. concerns that broad comparative studies Should be
lt is no coincidence thal wilhin the school of comparative attentive to Che context of- [Link], the contribution of
historical analysis, findings are often reported io books, area specialists is essential.
rather than articles. Part of the reason is that the 3. Reduce the Nwnoer of Variahles: The third
presen~ition of detailed informacion on each case serves solution to the small-N problem is to reduce the number
to further validate the conclusions drawn from of explanal-Ory factors, either through combining
compirisoas across cases. variables, sometimes referred to as ·data reduction, • or
These within-case comparisons are critical to the through employing a theorecical perspective that focuses
viability of small-N analysis. As St3Dley Lieberson on a smaller set of explanatory factors. One of Che
0991, 312-315) has correctly insisted, taken by promising sources of parsimonious explanatory theory is
themselves, comparisons across a ~mall number of cases, the 'rational choice' approach that has gained increasing
using either a most similar or a most different syslems attention among poJitjcal scientists. Rational choice
design, provide a weak basis for causal inference. modeling offers a productive means of simplifying
However, if ooe considers the role of these internal arguments that contain a multitude of int~restiog
comparisons, the 'N' is substantially increased, thereby variables, bul that may fail to specify the most critical
strengthening causal analysis. 12 ones. Within the field of comparative analysis, Geddes's
This use of within-<ase comparison can also help (1991) study of aoministrative reforu, in Latin America,
protect the analyst from a problem that arises in tbe most which models the impa,:t of different eleetoral and pa(ty
different sys-rems design, in which councries are matched systems on the incentives of lcgislato"' to adopt reform,
on the dependent variable and differ in terms of a series provides an excellent example of a productive
of background variables. Barbara Geddes (1990) has simplification of a [Link] topic. As such models gain
shown chat if cases are selected on the basis of scores on increasing acceptance in the comparative field, analysts
the dependent variable, which is how most different will B£quire a useful tool (or addressing the small-N
systems designs are often carried ouc, the lack of variance problem."
on the [Link] be explained incroduces a ' selection More work on concept formation is alSO needed,
bias' that can greatly weaken causal inference. One way notwithstanding the sustained cpotribution of Sartori
 Collier U3

(1970, 1984, 1991, 1993, and Sartori , [Link]: and Teuoe design. In this design the analyst looks al a long serie.s
1975); the work of authors such as McKinney (1966), of observations over tlme, so that the values of the
Kalleberg ( 1966), and Defelice (1980); and also Burller's observed variable are examined not on!y at two points
(1976) invaluable synth<-~is of the Weberian approach to immediately before. and aher the policy change or other
concept formation. CompMativists do not devote eoough innovation (which "interrupts" the series), but also well
attention to thinking [Link] bow well or poorly concepts before and well after. To illustrate th<> risk of restricting
are serving them and therefore may have insufficient the Wllllysis w these two observations, the authors present
ground for knowing whether they arc malcing appropriate several hypothetical configurations of data in which
choices io Lhe effort to achieve theoretical parsimony. restricting lhe analysis to two observations leads to a
The field of cognitiv:e science bas recently fLDding of sharp discontinuity' whereas the fuU time
provided insights into categorization that may be useful in series reveals continuity. Causal inferences about the
refining the concepts employed in comparative studies. impact of discrete evcat.s can be risl")' without an
The application of these insighis is illustrated hy George [Link] series of observations. Comparativists
Lakoff's (1987) challenge to frameworks. such as that of employing stm1ll-N analysis must heed this warning, since
Sartori, that employ what l..akoff calls "[Link] they routinely analy2o the impact of discrete ,>vents,
categorizatfon, • in which the meaning of conCcpt~ is ranging from wars. revolutions, and military coups to
uoders-tood in terms of defining cba:[Link] that are speeiJie public policies.
[Link] as giviog the concepts well-defined bouodari~. Donald Campbell and [Link] Ross's (1968) subse-
This understanding is crucial to Sartori 's framework, in quent analysis of the impact on traffic fatalities of the
that the problem of conceptual stretching which be Connecticut crackdown on. speeding in the t950s p(ovides
analy= hinges on these boundaries. Cognitive scientists a [Link] •exemplar· of the imaginative application of a
argue that in ordinary language. the meaning of concepts quasi-experimental design to public policy analysis.
derives [Link] from defining characteristics, but from -an [ndeed, Przeworsk.i (1987, 3.1) has argued that
implicit ·cognitive wodel • that underlies the concept and methodology is intluenced far mor:e by exemplar.a than by
from "[Link]· ca...;es that serve to aochor the concept's formal allcmpls to "legislat~· correct methods, and the
meaning and provide a point of reference for identifying Conuecticut crackdown article bas cert.,inly played that
better and worse cases. This perspective provides a role. u
different view of the question of boundarie.5, and h~oce of The ~ appears to be a simple one. When the
coaceptuaJ stretching. More wor'k is needed to discover Sule of C'.onnecticut initiated strict enforcement of the
tbe deg ree 10 which these patterns in ordinary language vehicular speed limit in the I 950s and tr,ffic deaths
are also present .i.u social science usage, and if so, the dropped sharply, the cause and effect relationship seemed
iroplicntions for the use of concepts in comp~r'dlive obvious. Yet in evaluating [Link] c.1usul Hnk~C.1ml)bcll [Link]
analysis (sec Collier and Mabon 19.9 3). Ross do an impressive analysis of potential lhreals to its
"internal validity · (was that really the cause in
Innovations Suggested by Work on Other Connecticut?) and i~s •e,;1ert1al validity· (can tl1e finding
be geoeralired?). No sen.si1ive analyst can read this
.Methods artiele without acquiring a more sobt!r view of the
problems of evaluating policy impacts.
Experimental Method Ideas about quasi-experimental and interrupted
time series design have also been disseminated through
Although the experimental me:bod itself may be
the large body of writing on evaluation ,esearch. This
of linle relevance 10 lhe wpics addressed in most
includes studies that apply these ideas to the analysis of
comparative rese.1rch, ideas dcri_;cd from the
political developmeot (Hoole 1978), as well as excellent
experimental method can improve small-N studies.
treatment5: of expcrimeotal design and evaluation research
Donald Campbell and Julian Stanley's classic
in introductory textbooks on social science methodology,
Experimentol ond Quasi-Experimental Designs for
. such as Babbie (1992).
Research (1963i shows bow the logic of experimental Although much writing oo quasi-experiments
design can be applied to ·quasi-experiments,• that is, to appears to offer helpful guidance and practical advice to
"obsc.["V'ational • studies that include some event or i_
[Link]- small-N analysts, Christopher H. Achen's 11,e Sro1is1ical
tion that bas a form analogous to an experimental Anolysit of Qua.,i-Experimen1s (1986) may leave them
intervcntioo, but that occurs in a "natural" setting. An
feeling that the methodological challenges posed by this
example would be the initiation of 11 new public policy type of design are overwhelming. [n studies of tho
whose impact one wishes to assess. impru:t of public policy, the core problem is the laek of
Campbdl and Stanley underline 1he great value ·randomiziltion· in the application of the policy, which
in quasi~[Link] of the "intemipted time series"
 114 The Comparative Method

may result in selection bias. For example, the beoefits of Mooney and Duval 1992). These techniques use
a policy are commonly received by some groups and not
computer simulation to [Link], from an icitial set of re:ll
by others, on the basis of cert.1in attributes possessed by
data, a large number of hypothetical replications of u,
the g roups, and ii is possible that these prior auributes 0
study, which can then be used in statistical tests !hat .,.
will [Link] reinforce the outcom.:s that the policy
not as vulnerable to violations of distributional as,,,unp-
seeks to promote. In the absence of true experimental
tions as are more conventional tests. These technique;;
data, this po.~ the challenge of disentangling the impact may be especially useful when !here is great
of the policy from the impact of these prior attributJ;S. heterogeneity among units, as may re;idily =ur in
This causal riddle can be addressed by constructing a cross~narionaJ comparisons.
model of how citizens are selected lo be ~ipicnts of the The development of "robust" and "n:sist'Ult"
policy. This model then becomes a building block in the statistical measures (Hampel e t al. 1987; Hartwig 1979;
analysis of the policy's impact, in that these prior
Mosteller and Tukey 1977) is promising in much the
considerations can be • factored out• in assessing the
same way, These measures are relatively unaffect¢d by
policy. Achen shows that solving the riddle requires a extreme or deviant values and can therefore help
complex form of "two-stage• statistical analysis. overcome tho problem in small•N analysis that findings
The implications of Achen's book may be may be seriously distorted by a single observation tb3t is
discouraging fo r analysts working with a small number of grutly in err:or.
cases. An adequate solution 1o the lack; of randomization Another Sel of techniqu.:s concerned with rhis
requires a form of statistical analysis which can be same p roblem is "regressioo diagnostics· (Bollf-0 and
opplied lo an elaborate quantitative data set, but this Jackman 1985; Jackman 1987). These ase tests used in
U:chnique would be bard to apply in a small-N study. A [Link] with conventional cegression analysis-to
more hopeful view might be that the literature on assess whether unusual values on particular obS:eJVatioas.
experiment~ and quasi-experiments at least provides caUed influential casas, have distorted the fu)dings. Tho
useful warnings about the perils of anal)'Ziog discrete advantage of regression diagnostics in comparison witb
events as if they were true experimenL1J interventions. In robust and resistant statistics is that one can employ them
the absence of appropriate data sets, the researcher must with the more familiar coefficients associated with
exercise caution in making causal claims. ~g=ioo analysis.
The use of regression diagnostics is nicely
Innovations In Statistics illustrated in the recent debate oo the relationship between
corporotism and economic growth in 15 Western
Recent work on statistical analysis has provided European countries (Lange and Garrett 1985, 1987;
both new warnings about the risks of statistical studies Jackm:m 1987, 1989; Hicks 1988; Hicks and Patterson
and new opportunities for doing meaningful statistical 1989; Garrett and I..aoge 1989). The starting point of
work. with relatively modest case bases. The statistician this debate is l'eter I..aoge and Geoffrey Garrett's 1985
David Freedman has launched a major assault on the use article, which presents an interesting and complex idea in
of multivariate quantilative analysis in the _social sciences a simple form. They argue that the organizational
(1987, 1991), which ho claims fails because the strength of unions in the labor market and the political
underlying research design is generally inadequate and strength of the left in the ele<:toral and govemmenl'11
because th.e data employed fail to meet the assumptions of arenas both have an impact on economic growth, but that
the statistical techniques. His criticism may bring this impact is shaped by a complex interplay between
considerable satisfaction. to those who have been skeptical these two factors, which they represent 1hrougb an
about statistics all aloag and who take comfort in the "interaction" term in their regression analysis of the 15
g,taler "control" of the material they feel derives from cases.
analyzing relatively few cases through more quali1->tive In a reanalysis of their article, Robert W.
techniques. Ct is realistic to expect that we nuy go Jackman ( I 987) employs regression diagnostics to
through a period of greater questioning of the use of exa mine certain influential cases !hat be believes distort
statistics in the social sciences. However, as with the their findings. In the ensuing discussions among these
rejection of quantitative cross-national research d iscussed five authors, an expanded model with further control
above, it would be unfortunate if a reaction against variables is proposed, this expanded model is both
quantitative studies went too far. challenged and defended, and I..aoge and Garrett
Toe e mergence of new statistical techniques that subsequently defend their original model and call for new
are helpful in the analysis of relatively few cases makes data and further tests.
s-uch a blanket rejection unwarranted. One example is the Tus scholarly debate brings together an
development of "n:sampling strotegies" such as lhe important subst:uitive problem, a rugb level of area
"bootstrap" and "jackknife" (Diaconis and Efron 1983, expertise and knowledge of specific cases, the inventive
 Collier llS

use of a relatively straightforward statistical model, a 'o[ almost no scientific value· (1963, 7). He shows
constructive-critjque based on regression diagnostics, [Link] instead that case. studies are the basis of most comparative
a sustained process of cumulative knowledge geoeratioo research, that they offer many more opportunities than is
based oo the scrutiD.y of a shared data set. Just as the ofteo recognized for falsifying the researcl>er's main
Campbell and Ross article on the Connecticut speeding hypotheses, and that mucb can be learned from making
crackdown is ao exemplar of a quasi-experimental design, explfoit the comparisons tl!JI~ [Link] often implicitly .built into
this debaie should stand as an exemplar of a case studies. For example, any given hypothesis about a
methodologicaUy sophisticaied effort by several scholars case has implications for many facets of tho case.
to solve an important problem within the f~ork of Campbell uses -the label "pattern matehing • to refer lo the
small-N quantitative analysis. 1bis debate also shows process· of discovering whe1hcr the,;e implications an;
that although an "N" of 15 might often be approached realized. The analyst can thereby increase th.e "N" by
through qtL1litative small-N comparison, it can likewise multiplying the opportunities to test hypotheses wi([Link]
be subjected to statistical analysis, with interesting what may initially have been viewed as a "single' [Link].
resuits. Thi$ procedure of pattern matehing is Jielpful in
Another area il1 which potential problems of addressing the long-standing concern that case studies are
statistical analysis are amenable to solution concerns the useful fo r [Link] hypotheses, but that the same case
issue or "average eff«:ts~ in regression studies. The cannot then be used to test the hypothesis because it
results of the simpler forms of regression analysis are offers no possibility of discolllirmation. This is
based on an average of the strength of causal relations sometimes referred to as the problem of ex post facto
across the cases being studied. For the coefficients hypotheses." The procedure of pattern matching opens
produced by regression analysis to be meaningful, it is the possibility that an hypothei;is initially generated by a
necessary that these causal relations be homogeneous particular case could subsequently fail lO be supported by
across the cases. Yet Ragin (1987, cl!Jlp. 4), among the same case. Thus, the problem of ex post facto
others, bas forcefully argued that this assumption hypotheses can be partially overco01e. "
cx,mmonly does not hold, givco the complex form,; of Harry Eckstein (197S, 113· 123) is likewise
·multiple coojunctural causation'" often [Link] in ronceroed with testing, as opposed to generating,
comparat,ve studies. In different rontexts of analysis, the hypotheses in case-study analysis, Md be argues
interaction among causal factors may vary. forufully that many analysts have greatly underestimated
However, solutions to [Link] problem [Link]-available. the value of case studies for hypothesis testing. lo
John E. Jackson (1992) shows bow it can be addressed particular, the carefully constructed analysis of a ·critical
with advanced statistical techniques, and the interaction case· - for example, one about which the analyst bas
term in the l..aoge-Oarrell regression analysis, discussed particularly strong expectations that it will fit the
above, deals wilb precisely this problem: that the effect hypothesized causal pattern - can provide an invaluable
of nne explanatory factor \/aries depending on the value opportunity lo falsify the relevant hypothesis.
of another explanatory factor. FinaUy, Przeworski and Alexander George and Timothy McKeowo
Teune's procedure of ·replacing proper names,· also (1985), building on George (1979), present a helpful
di$Cussed above, takes this problem of causal romplexity synthesis of two key building blocks in the process
and turns it into ·a n [Link] to deal more theoretically through which hypotheses are tested in case studies. The
with the diversity of causal patterns. first corresponds lo the conventional approach lo placing
a case il1 romparative perspective, wnich they call the
lnno,·ations in tht Case-Study Method •congruence procedure.• Tue scholar examines the
values of an hypothesized indepeodent aJJd dependent
When Lijphart wrote bis 1971 article, he variable for a given case and determines, in light of
apparen~y fell some hesitation about including a explicit or implicit comparison with other cases, whether
discussion of case studies in an assessment of the these values are consistent with the predictions of the
romparative method." Yet the two topics are closely hypothesis under consideration (pp. 29-30). The second
linked. aod nis helpful typology of the uses of case is "process-tracfog, • through which the researcher
studies il1 hypothesis testing and theory building set the engages in a close processual analysis of the uofoldiog of
stage for refinements in case study aoalysis later eveuts over time within the case (pp. 34-41). The goal is
introduced by other scholars. lo assesses whether the dynamics of change withi.o eacb
One of the most suggestive discussioos of the case plausibly reflect the same causal pattern suggested
case-study method is that of Campbell (1975). He by the compantive appraisal of the case in relation lo
dramatically recants the bold assertion be mad<: in his other cases. Process tracing may be seen as a specific
earlier book with Stanley that •one-shot" case studies are [Link] of Campbell's pattern matching, and as-with
116 The Comparative Method

pattern matching the analyst makes a series of within-case continuing intellectual and institutional strength of ara
observatio11s against which the hypothesis can be further studies, and deep skepticism in some circles abouc the
assessed. validity of broad comparison.
Overall, these articles, along with works such as Second, it is evident that quantitative tecbniques
Robert 1{_ Yin's Case Study Research (1984), offer a employing a relatively small number of cases can
systematization of case-study procedures that provide a successfully address important substantive questions.
valuable point of refere nce for scholars· concerned with This approach merits altention in light of the new
small-N analysis. At tho same lime, the debate continues sts(istical tests suitable for small-N analysis. TI,e
on the proper role of case studies in .assessing and opportunity fo, cumulative scholarly learning provided by
building theory. An inreresting part of this debate, statistical studies is [Link] illustrated by the
published as a special issue of the journal World Politics Lange-Oarrett-Jackmnn-Hicks-Patter.;on debate. This
(1989) focuses on the contribution of case studies lo debate is also relevant to the issue of linking rival
evaluating one [Link] of rational choice analysis, researeh traditions, because it shows thac insights derived
i.e., rational deterrence theory in internatjonal re1atjoos. from case studies aod from more qualitative comp,:uatiYe
The o ~ g article by Acben and Snidal (19.89) argues work can, af\e( all, serve as stepping-stones on the path
that the case studies employed by many inremational toward statistical analysis.
relations specialists do not adequately address the central The third altem,tive bas been reinforced as well:
ideas of this body of theory, thereby raising an issue the systematic comparison of a small number of case3,
perhaps not often enough considered in discussions of the with the goal of causal analysis, which is the approach
comparative method: How can the methodological that Lljpbart originaUy advocated. In !his perspective,
concern with executing good comparisons be linked to the broad qualitative comparison is seen as both possible and
key analytic issues posed by the particular theories that productive. The growing influence of the school of
are to be evaluated? Acheo and Snidal also note the comparative historical analysis has substantially eohancod
problem of selection bias in case studies of de terrence the credibility of this approach, and ii p lays an importanc
theory, that is, the problem that case studies usually focus role as an analytic middle ground between the case-study
on deterrence failure, whereas much or most of the time tradition and small-N statistical analysis.
deterrence works. The issue of the journal includes a All three of thOS<l approaches will persist, and a
series or articles by S<:bolars cloS<l to the case-study key question is how well they can be linked. The
tradition who debate the issues raiS<ld by Achen and tradition of research on Western Europe provides an
Soidal. These articles constitute a valuable effort to think encouraging model, in that the findings of quaoti1,ttive
through how case studies have functioned in relation lo comparative [Link] play an important role in general
the assessment of a particular body of theory, a line of debates in that field." ln research on Latin America,
inquiry that should be taken up more often. by contrast, quantitative comparative work receives
In this debate oo deterrence theery, an considerably less attention from mainstream scholars.
intellectual tension emerges that bas been a recurring Yet the kind of cross-fcrtilb:ation fouod in the West
theme in this chapter. between analyses that seek to European field can make an important contribution to
achieve a generic understanding, based on relatively few strengthening research. With good communication,
variables and encompassing many cases, as opposed to country specialists and expertS in qualitacivesmall-N
analyses that seek to draw out the complexities of comparisoa can push the comparative [Link] toward
particular cases. more carefully contextualized analysis. Likewise, the
comparative quantifiers can push the couotry specialists
and expertS in qualitative comparison toward more
Conclusion systematic measurement and hypothesis testing. A central
goal must be to sustaio such communication.
The implications fo, graduate !raining are clear.
Among the diverse approaches discussed in this
Tf Ph.D. candidates are to be prepared to address these
chapter, three major analytic alternatives stand out.
issues of comparison, they should have enough training io
First, new perspectives on the case-study method have
statistical methods lO evaluate quantitative studies that
strengthened the viability of that approach. Discussions
employ old, and new, methods of statistical analysis and
of opportunities for within-case comparisons have in ·fact
to use such methods when appropriate. Those more
begun to blur the distinction between case studies and the
oriented toward statistical analysis should have enough
comparative method, although the case-study approach
backgrouod in qualitative small-N comparison and case
does remain a distinct tradition. Interest in case studies
study analysis to be able to build oo the analytic
has been reinforced by several factors, including the
contribution of those approaches. Both groups should
renewed concern with interpretive social science, the
have substantial exposure to basic writings on the
 Colliu 117

philosophy of science and logic of inquiry that can co1TC,pond, rc:3PCctively, LO John Stuart Mill'• {1974) method of
provide the framewor\c fo r more informed choices about dill'ereoc.e- and method of ai'J'.U-cncnt. Wheteu Pr-""won:tj and Teuoc•a
these methodological alternatives. label, of ·,[Link]" aud ..different· refer to [Link] the ca,u are
[Link],;bcd, at oppO,lic:410 coaltUlin,. oa • .acriu of bdekgrowtd Variable,.
In this way, th~ foundation can be laid for an Mill'• [Link] o( "'diffcrenc~· aod ·1;grccmet1.1· refer co whether the cue.a
eclectic practice of smaU-N analysis !hat takes advantage are cootr."'1ing, a, oppoled toma\ched, OQ the depaukns variable,
of opportunities that present themselves on both side.s of 11. Pe1'110na.l comnw..aieatiOft from Adl.m [Link].
what could otherwise be a major intellectual divide. ll. [Link] A,ehco, pcrsooa.l ¢OrntnUnkaiion, bu lot1t
insisted 00 lhi, point.
13. For c,umplc, Gcrcffi [Link] Wymaa (1990), Haggard
(1990). [Link] (1991), &.od [Link]., 11.1\d Sttpbera
Notes (1992}.
14. for• di.,;uuion of lt(atcg:ic choice modcla {• elotcly
11:Li• ii • revised aod upandcd Vtrtion of 1n article [Link] r [Link] type of model) [Link] have been applied to che [Link], o( political
publi&hed in [Link]::'wart A. [Link] [Link] Ke:MClh Ptul [Link], edt., r-cfonn. dcmocniti:talion. aad democn:[Link] cotu0lid,1tion i.a:1. l..[Link].
Compa.ra1iw Palitico/ Dyf'1Mlics: Gl<>b41 Ra~o,rch Ptrspc-ctiw.s (New Amcrfoa, aad that Likcwi..c ofl'er fruitful aja-q,li(icatK>Ni o( complex
Y<>rt: Harper Colliru, 1991). Pcrmiuion lo re-print gre.n~ed by Harper pbcno~na, au CoWc-r and Notdc--n ( 1992),
Colliru. Rulh Bcrins CoWcr, [Link] Paul Ericbon, Leonardo JS. The rcpri.f'!li-ng of lhia [Link]:k- io a teadcr oo ,oeial
Morfoui-. Elizabeth &&bee, [Link] Caro) A,. Mc4lio made parti-culuiy [Link],:c mcthodoloa:y ([Link].c l970) made ii widely available 10 [Link]
useful suggecflions on culicr-dnfb. I also [Link],[Link] co-mmcnu Crom [Link], and it.a inOuc~ bu bee-a [Link].
Chrffloph,:r Achu, Su:phcn ColGcr, June, Furoh, David Freedman, 16, .Pcaocal [Link] from Arend Lijpbart.
Dcb<>rul Norden, Robert PowdJ, ~foTill ShuW, and Ulura Stoicet. 17, Thia problem i• roudncly ditcuuc.d in [Link]-y
Ada F'.[Link] and two anonym0tu [Link].·er, likewise nude helpful mclh-odologyu:ru, e.g •• &bbk ( 1992., 24-2.S, 427).
commenu. Thi• f'[Link]~b hu been supported b)' 1 G\lgg~cim 18. Although paatm ('!Utch.i.n, within [Link] [Link] ¢He·
Fellowship, lhe SociaJ Science Re~src:h Coumil, and chc lnaitute of introduce• [Link] P.:OSs:ibility of fahi(ying th~ hypothc.m~ il doc., 00(
GovcmmcntaJ Srudics a l Berkeley. Finally, I would lib to note:• Very ovcn:ome all of the pi;cblcma or ex pGJ.t facto bypothue•. Th.u•,
ptomii ing ttl4nut<!rip1 (Kfog, Verba, and [Link] Im) lhal pattern [Link] w ilJ prob•bly DOl O\.'ercomc, a [Link]
urtfortuNtcly [Link] lO my attention too l:itc Lo be di.J.::U-»Cd Ul thi• [Link]:loCnlstiveDC':' wbjch may arise due to acl~t:tioa bi.u or to [Link]
ch1ptcr. [Link] telection or an atypical c-uc.
19. See, for example,~ debale oo in1creai medi&tioo and
I, ·t1· ia tii&c:d IO r1:for lo th&numbor of [Link] tM.l)'Ud ia e-orporatism i"n We11tem Europe, including [Link] .(1976)1 Hibba
•ny riv-en ,tudy, (1978). Sc:hmi:lter (198 1), abd Cameron ( 1984). TI1e 'dcbat.e ll&!Vd by
2, Reference• lO ttpn:$CnLllivc. worb of [Link] Lange and [Link] (198S) it• contin"llation of this line or •naJyli1.
histori:.al [Link].U. •~ prc:[Link] below.
3, In his compuison of these n,e1hodis, Lijphart
,;;koowlcdgc.s h.i.a debt lO S~lsi:r·, (l96S) cxceUcm [Link]•i• that
employed a [Link] [Link]:, See also Smclitr (1976). Bibliography
4 . Thi, perspective ·hu beco claborstcd by St([Link] {1984,
chap. 11), and a pa"Ud formulation i1 found in Charita Tilly (1984.
chap. 4). Achci:i, Christophe, H. 1986. 1hc Suu1srle-aJ [Link] q/~i~
S. Skocpot [Link] S.(ut)Cr-. ( 1930, 131-81) rc.f,:r lO thit u E:q,~rl~u. Berkeley and U)II At\gelu: [Link] of
~macro-causal" [Link]. Yet 8([Link]·N 11t1.1di.e1 [Link] generate [Link] lu:I California Prc.11.
hypotbcs.::s ean have botb o tnA<:ro [Link] a. micro [Link], and it doc..s not Ac hen, Cbri$l()S)be-r H., [Link].1· Duix:,.q SnidaJ, 1989. ·[Link].1
1«m pro4uciivc. to exclude from thi, category tho~ v.·hh a mi:ro f«:u,. [Link] Thc.M)' and [Link] C.6C Swdlea. • World
Hent:e, lhu [Link] [Link] ia u~d. P<Jlilics 41:143-69.
6. [Link] Pruwottlci and Teune are eentnally co-oeemcd Alm:oad, Gabriel A. 1 and S1tphcn J. Ococo. 1m. "'Cl.oud1. Clocks.
wilh issues lhal [Link] wttca [Link] [Link]:t arc: add,c.4 to an [Link]•. the ~ the Srudy of PtJlitic1," World Politics '29;489-'Sll.
problem, lhcy dl~un are also more likely lO OC¢Ur if Ot\e it [Link] Almer, Mkhael·, -and Allen Orim9haw, ed•. )973. Cc-rnporodw: Sodol
wi:lh • larger N IO begin wilb. Rc:1~a~h. New Yori;: lobn WiJey.
7. For ex-ample. [Link] of referring to ·[Link], • one Babbie., .Eld. 1992. Th~ [Link]:e ofSodol &ua~h. 6th edi1ion.
would re(c.r to • country in wbicb, due lO the impact of mauive oil Bc1motl,l, CA: Wad,w-orib.
1Y>[Link], • particular caual relationship [Link] a distinct (onn. Bend.~ .[Link]. 1964. Nanon-Building an# C'idunshlp: S1udla ()/
8. ,[Link]-ription· is aomctimca [Link]:nly understood Our Qi(tng1'11g Scx:ial Ordu, New Yori:: JObn \\,..lie)',
1.0 refer ,imply to •[Link] de&Cription, • which i• oot what G«-r tt Bendix.. [Link]. 1978. Kings or Pt'-opk: Power and tlu [Link] u,
inltnc'5. [Link]. [Link] lcy and Los Ao,elca: UNvcni1y c,fCaliforni.t
9 . Give.a that dioe atudiea of\eo focu, oo long periods of Pr-..,.
time [Link] «ch ca»e, i1 rtUgbt ~ argued thtt the number of cuc.s could Bergquist, Olatles. 1986. l..[Link] Lalin Am~rka: C<>n:.paraiiw
be. g,-eatly inc:teoca through cacnpariJOn over time, thereby fl\111:ing Eua,$ en O;ik. A.f'[Link]<J., Vie-~.:.sula, and C<,IQmbla.
them somethin,s olhcr lha.n $.l"l)&ll•N" srudiu, However, aince the i,-ocl of Sl.a..n(ord: Slaruord Univen:i1y Prus.
[Link] studiea in [Link]• tradition is I.O explain [Link] configuration& o( Bollco, ~ l h A ., and Robert W. Jaekroan. J98S. "'Rcgt¢uio.o
riational outcomes u they ate [Link] over long ptriodt , th_c1-t Diagnostics: An E~sitory Trulmcnt or OutUcrt and
outi;omc, ot\ca. ~[Link] be di14ggrcg1tcd into a .cric.s of longitudinal [Link] Cuc,, .· Sociologic41 Mtthods ar.d &uarrh
ob~r,[Link].s. Hence, [Link] number o( [Link] [Link] ruliSlic&lly be 13,SI0-42.
ineruscd lhrougb the uac of comparitoo onr time. Burge,, Thoma,. 1916. Mar Web<r'1·1Aeoryo/C<,~ep1Form¢,:m.;,
10. The mott $iOUl.u and moll. diffcrcn1 ,y,tcm, duiga,. Hisu,ry. L..awr. 411d [Link] Ty~$. [Link].: Duke Universily',
118 The Comparative Method

Cameron, D•vi.d R.. 1984. ·social Dctr«raey, [Link].&: POJM ()j (ndustrlafi:llliOfl in IAnn Alr-Ldl a -4
[Link]$10, .Labour Quiescence, and the Reprcs,[Link] of ~ [Link]. Princ:etoa: Prino;.,cton Uoivenity Pru,.
&onomk U'llete~ i..o [Link] Ca~itali.1t Society.• 1A Ottkr Ooldstono, [Link] A._ l99l. lk-,-0411:0n wkl Rt:~llion in~ &riy
ond C-o"JUct in Contemporary Copirallnn, ed. John H, M~m Wqrld. &rkeley and Lot Angele-.: Uruvcl'$it o(
Ookhborpe. New Yott Oxford University~- Cali(omia Pteu.. y
Camp,beU, Oor:aJd T. 1975. ''Degrees ofFtudom• aod lhe Case Hau,itd, Secphao. 1990. Padt~s from lM Peripkry: 1>-< [Link]~
Swdy. • Comparolfre- Polirl'cdl Stud iu 8; l 7&-9). of Gf'O'WUI ill ti:< Newly lntbutrlaG:ing <:.owu:rir.f, ltha.c, ·
Campbell, OooaldT. • aod H. l..au!'t,nce Ro». 1968. tho [Link].:ut CotncU Unh·<:ni1y Pru,. •
Cn,[Link] o n Spc.:dio.g: Tune Seriea Data in Qua,i.. Hampel, Fnnk R~, e1 al. l987. Robw1 [Link],: 1"4: Ap;,~>i
Experimental [Link].s.• Lttw and Society Review '3:'l3-S'3. &ucd"" Jn,[Link].e,Funcdo,u. New York.: John Wiley.
C1mpl>ell, Donald T,. [Link] Juliao C. Stanley. 1963. £.qxritMnl41 ();JtJ Hartwig, Frcdcri.;:k. w ith Brian E·. Oc4ring. 1979. [Link](J,y ~
Qu4..tl-&Mrimauol Duignsfor Rt:st:<Jrdt. Chica,go: Rud Analydl. S•go·Univeriity Paper, Serie, No. 07-0)6,
McNally. .Beverly Hilla, CA: Sage Pub(i(:.s1iocu,
Co11ier, David, and Jame• E . [Link]. l993. ""Conceptual ' Strct.:;hing' HecUChec. [Link] . 19S7. 1kt Study o/Om1P4ratli-e Go~·.i:mnun, w
or
Revisited: Altt0\[Link] View, [Link] in Comparative Poliric.s. London: Geotte. Allen and Uawi.o.
[Link]$. • Am(CriC''1!'1 Politicol Scknu Rrvkw. Wbb,, Oougl,11 A., It, 197&. ·0n the [Link] &:onorriy of lo~·Ru,11
Collier. David. [Link] 0-:botah L. ~ordcn. 1992, "[Link] Choice T rends lD Strik.G Ac-tivhy. · 8ritirli Journal of PoU1l~al
Moddt' Q( Poli.t:ic:al Change-in .[Link] Amc:rka. • Q,,,,,parodv< Same, 1'153-115.
P()/Jlia 14:219-243, Hieb, .[Link]. 1988. ' Social [Link]:ra~i<: Corponti.-a.m ..Dd
Collier, Rotb [Link], and Oa..,Jd Collier. 1991. Shaping /Ju [Link],:al Ecooomio Growth.· Jownal of PoliJJcs 50:677-704,
Anmo: [Link] Junctures, Ihe UJbor Movemau. 4fl4 R~gf,ru fficb:, AJe:uQdcr, aod William Dav-id Ps tu.l"llOn, 1989. 'On 1he
DyntJmlr:s in /.,,aJm Amuico, Princeioo: Priocctoo University .Robua1,t1,[Link] of the Left COfPOnitiJl Mode.I or [Link]
Pr-cu. Orowlh." Jounw1 of Poli4cs 51:662-<115.
Dthl, Robc-tt A., ed. 1966. Political Opposid(NLS'in Wr:skm Hoolc, [Link].1 W. 1978. [Link] [Link]\!h and Del·dopmatl
[Link]. New Haven: Y•le Unlv¢nity Prcu, ActMtfu. Bevcriy Hill,, CA: [Link] Pubtioatiooa.
lxfdicc, E. Qenc. 1980. ·Compari.$00 Mi..[Link]: Commoo Jackmt.n, Robcn W . 1985. ·cro•N•tiona.l [Link].l Rucan;h •od
NonS<n8C in Comp,ntive Po1ilks. • Compo:NJdw PoU:ks the Srudy o( Comparative Politic 1 ... [Link] loum,aJ of
13: 119-26. PoUdcal [Link] 29: 161-82.
Oiaconi.s, PcNi, and Bnidley Efro1, . 198'3. ·compuu:r· lrltcMive h c [Link], R,obcr1 W. 1987. "The Politics-of [Link]. Growth in
[Link] Stati$1..ic~. • Scicrllific [Link]:ric;an '248:1 16•.30. [Link] Dcn:xx:-,..c:ic.i, 1974--30: Ld\is:i: Strength or tfo'1Jl
[Link].n~ Hany. 197S. · ease Study and Theory in Politi.:al Scicoce. • Su Oill" Jocmu,J of ~olltl'cs.49:1.41·56.
ln ffandl,oqk of folf6col Sci,n,;c, vol, 7. ed. Fn:d Oo:en>1<,io J,ctm.w,'Robe~ W. 19&9. "The P<Jlitio1 of Boononuo Growlh. O..:e
utd Ncl'10o W. Polllby. Steading, MA: Add'i-&00-Wc11lcy. Agai.a.• Journal ()/ Pt>lldct ·Sl.:646-66 1.
Eggan. Fred. 1954. ·social Aothropology and tbe-McUlod o( l [Link], lob..o E. l992. "Elltimation or ~{ode ta with V1ri1bk
Controlled Compari,on. • [Link],crica:n [Link] 56:743- Coefficicou.• fo Pollti'c-al A,w}[Link]$, vol. 3. ¢d . Jan,e1-A.
63. [Link]. Ann .Arb<it": Uni..,ctsily of Miobigao ~u.
[Link], Amitai, and Frederic L. Dubow, eda·. 1970. [Link],atfr< [Link]~ Ar,huc L. 1966. ibc Logic o(COmp&ri$0n; A
Perspurlw:s: 1hcori,:4 and A(,:rhnd.r. 8o:ll(>n: Link, Mclhodologieel Noto on the Comparilive Study of [Link]
Btow11. S).':stems. • World PoJitic-s 19 :69~82.
FrccdtJUn, OAvid A, 1987. ·~s {)then Sec. U1: A CueS~dy in Path, King, Guy, Sidney Verba, snd~obetl O. ~.tnc, 19~- Sd<Nifi<
[Link]. • Joumol of Educo:i()(I.Q/ Stolisrics 12:101·28. lnfermc< fr, Q,.,aliuitive /l.,:.se'1rch , [Link] l'l"ltlnultript,
freedman, [Link]. A. 1991. ~St.s!htical Mockl, and Shc>c (...[Link].:r.• 1n Department o f Go"Yeminl:·nl, Hatvaril Univottity,
SociologicoJ Medwdology 1991, ed. P<:tec Maf'ldcn, $10 [Link]', Gcorgc. 1987. WOll'WI, Fi~~ and Dangcrc1,u 1hb:t1s: WM1
Fn~iscO: JQ1scy~[Link]. [Link] &-.~al about~ .Mind. [Link]: UnivcAit)' oC
0-artctt, Gcoffny, [Link] Peter Lange. l9&9. · ao..,emment Pu\[Link] Chicago l"rcu.
and .Ec-onomi.c Performance; Wlu:.n aod How Doe, 'Who Unge, Pck:c. and Geoffrey Garren. 19&5. 'The Polilics or Oro~:
Govccns' Malter?· Jo&1rr-ol of Po/ides 51:676,93. S~tcgic [Link] and ~ooomic 'fer(QMl'.Jll)Ce i.n Adv&no:cd
Geddes~ [Link]>1n.. 1990. ·»ow the Cues You Choose Affect th-e (nduitrial Ocl'l'IC)(;noie1, 1914-1.930. • /oumal <if fol/du
Anxwe~ You Gel: S<l~lioo Biu io Comparative Politiu. • 47,792,827.
ln Politkal AnalJN, vul. 2, ed. Jama A. Stinuon. Ann L..aJta:e, Ptu:r. [Link] GeoO'.rey G~u~ l9S7. -n,c. Polili'e.-ofGCOWl.b
[Link]: University of Miehigaa Prcu. Re<:oa.:sidc:reJ. • Jc,ur,w,J 'of Po4tks •9:l.S7-7•.
Gcdde,, &rbon, 1991. • A Game Theotttic Model of Rc(orm in Lalin LauweU,, Harold O. 1968. -rhe Future. o( the [Link] Method. ·
Amet'Kao Ocmo:ncie,. • Anuric-an Pofirlcol Sd<,r1e< Review C«[Link] Politics 1:3-18.
8.S:371-392. U ~el'90n, [Link].• 1991 . "SrMU N'a [Link] Big [Link]; /t,.n
Gec-rtz-. Cliffo rd. 1973". -rl:iick Dcscriptlon: Toward •n lnletprcLive [Link] of lhc [Link] in Companitive Studie, Bucd
Theory or OJllurc," In The lrurpu«1a'oll of Cult&1us. ed. oo a [Link] Number of C..sc11. • s«ial Forots 70:307~20.
CUfTord Gecttt. New Yort: Bui¢- Boob. Lijpb1rt. Arend. 1971. "CompmUve Polllica and Con.v-rativc
Gcori:c, Alexander L. 1979. · [Link] [Link] lbcocy Oevel<1pmtnt : Method.• American Polillcol &:it.n,cc. ~view 65:682-93.
The Method of Stn1.:tu:rcd, Fo,cu..sed [Link]&oo," In [Link]'\, Mend. l975. -rh~ Compan,,l,[Link]:.e Strategy in
l>lplc!macy: New Approo<'-ha in Hi!.klry, 'TMory. tJl'ld Policy, [Link],uc:[Link] ." C<N,tppraiivc Pt>lrdcal Swdiu
ed. Paul Gordon [Link]. New York: The Free Prea,, 1,153-77.
George. Akxaodec t..., and TimoQly J. Mc~owtt. 193.S. "Cue Ujphiri. Ar•md. 1990. "The- Poli1icil Cooteqoences or EJc:[Link]
Studic• and Theories [Link] ().:cis ion M ak ing.· Lawa. 1945--1985. • [Link] Politi, (1} $clt!nCC Re\;ew
Advancu 1·n (nf(Jn,ttJJllm Pme~.ssing in OrganJ![Link].r, vQ). 2. 84,4'1-96.
San~ [Link], CA: JAJ: Prcs.11, Lipsct. Scymo:ui:-Martin, and Stein [Link].o. 1967. · [Link]•,
(icn::ffi, Gary, 1.00 Dooald L. Wynv.n, cd1. 1990. Mc:.rw/~.turing Sttucturu, Pattr Sy~me:1 [Link] V()tc.r AligflJ'[Link].a,; An ldtto,.
 Collier 119
duC'ti<to, • 1n Pany [Link].s and Voter AligMfmU, eds. ~hmittu, Philippe C. 1981 . "[Link]:.,;t lntcrmc<[Link].a and R.c:,imc
......
Se)'mOur Martitl Upset and [Link] Rollio. New York: Ft"Cc Govet"l'.IAbllity in [Link] [Link] Europe [Link] N"onh
Ameriu. • In Organlvng ln4ruu in WUk-m Luvpc, ed.
Lueblx-n., Gregory M. 1991 . [Link], F&dsm, or Social
Dcmoeraq; &,[Link] 04s$.es and h PoJili'"al OrigiN of
Rt:gima in ll\krwo-, [Link],e. New York: O~ford Vaivenif)'
......
Su:wloi: D. &rgor. Cambridge: C..mlnidr, Uaivcnity-

Skocpol, Theda. 1979. St4la" a:,t,tJ. [Link] &,,o"4dons: A C«nporazlw

"'-- A.1-uuy.N ofFrcnc~, Ruula, and C1-Jna. Cambridge:
[Link], John C. l966. Cons~dw- Ty/>('l"gy and &xial TMory. C'4mbridgc University Prcu.
New York: [Link] Publi..sbiog Company. s ~ . Thcd.t. 1934. Vlti0n -1 Mellwd;,, H'u u,r!cal s«Io/Qgy.
M«Tia, Rkbard. 197-0. S•yst<matic Appl"()a~s lQ OJm.,?araJivc · (AmbridJe [Link] New Yotk: c..[Link] [Link] Prcu.
Polllfcs. Chic;,go: Rand McNally, SkocpoJ, Theda;, and Margan: Somer1. 1980. "tbc Uae. or
Mctriu, Richud, [Link] Sttin Rokkao, ed,. 1966. Comparing [Link]: Come• rativc HiMty [Link] MaCroso,dal [Link]. • ~ l i w
1ht- Uu of [Link]<J4ve DaS,tJ in Cr<ns-NaJioru,l Rest:[Link]. SIUdlu In Society dNJ. Hinory 22: 174-97.
New [Link]: Yale Uni-;ersl1y Prew. [Link], Neil. I 968. "-The Methodology of Comparative AAalylU ot
Mill, [Link]. (1843119?4. A Sys1an ofLogic. Tot0nto: Ecoo6tnic Activity. · lo &sap In SodologlcoJ
Univen;ily ofTo~ Pt.:$$. lnlerprctoti()(S, ed. Noil SmellC'r. En,lcwood CUfta:, NJ:
Moooey, C'hrist<>phcr 2 . , aod Robcn O. OuvtL 1992. · Boouu-ep Prentice-.H-tll.
lnfctencc.: A Prt-titn,i(\acy Moate Cado Evaluation.• Paper Smets.:r, Neil. 1976. Comporctiw Met>.ods in IM Socid1 Sd~c.r.
[Link] ol the umuaJ rnc:etulg$ o( lhe Amerkan Po)jlical Eaal eW'()()d Cliff,, NJ: Prcmicc Hall.
Stieo.:e AuociALion, CMcago~ Stinebeotnbc, [Link] L. 1978. ~orelical Mcuwd.s in Sodttl H'11ic,ry,
Moore, Barrington. 1966. Social Origins of DittaL(Jrship arid New Yori:: Ac•dcmic l>rca...
Dt:mtXrocy; LAnJ and Pt:[Link] in 1M Moldng of W Mo.:km 1iUy1 [Link].a. 1975 . TM Formarlon <>/ Ntaf()ntl./. $tas.c1 llt Westt:m
World. Boston: &.~on Prc-u. Euro~. Princeton: Prmcd.l.>o Uoivemly Pn:A.
Moste.l!er, Frederick, arid John W, TuUy. l977. DtU.a Anal)'$tl and Tilly. Chari~. 1984. Big~,~ lArge Prr>cu:us, IJJ<g~
Regrts~<111, R«ding, MA: Addi$on--We,l?y. ComporiM,,t.s. New Yort: Ruilt-11 Saac.
O'Donnell, [Link]. J'bjJippe C . Schmincr, and Uwrcoc.e Whitch<d, Trimb¢rger. Elko Kay. 1978. &volwfonjrom A.b,n -e: Mili'IQry
eds. I 9S6. Tru:n.t'idons .fiw,t AuJhcn·[Link] ~ . [Link]: 8"'1!~1'uU liN! Dcwlopmen: In Japan, Turiq, EgypJ, '1NI
John R<>pldru Univcnity Pre». Peru. NeWBt\.UUwiet, NJ: [Link] Boob.
Plige. Jeffrey. 1975. Agrarian RcYOlu:iOfl: Soeial Mt>Vancr.1$ and Tuf\e, Edward R. 1970. The ~tali~-e ~yd$ ofS«ial Probkms.
[Link]'r1 A&dculture in W UNkrdei.·t-UJpcd World. New Rudirig, MA~ Addit0n--Wcdcy.
York:: Free Prus. Y1lli<r, lvlO, "1, 1971, wmpat01iw: M<liwdl in s«iol<Jty: Eltay1
Pruworski, Ada1n. 1987. ·t,.[Link] Crou-NaliotuJ R.t-Jeatth, on Ttmd.t dnd Appll~(ll}(ln.f. Bccblcy and..Loa Angele s:
1970-1983: An O....ctvW:.w . • Mcinolf Dierket, lli.o.s N. Uftivoriity afCal'irart1i.a Pren.
Weiler, aod Ariuc Bc:nhoin Ant.o.1, eds. O,,nparadve Policy Vcrt,a, Sidoc)'. 1967. "Some Dilemmas in Comparative Reaeucll ...
&seo.rd1: U amlngfrqm bpcn'e1,u. Brook"f,dd, VT: World Politics 20:11 J-27.
Cower. Wllcnsl'.y, Harold L. £970. 7M 'New CorporMism, ' unl1dllf,4tion,
Pri.-.wunki, /\dam, l99 l . D<m«raryand tlu: M~,.- P()litkal a11d and tM [Link] Stal<. Profcs.skm1J P•pcn fo [Link]
U<;nomic Reform¥ In £[Link]:rn £u:r(lpe a,wJ [Link] Americ-0. Pol1tiul Soo:iology Seriu, Beverly Hilla: Sage- Publicatioo,.
[Link]: Cambridge Onivc.ni1y Prc.s.a. Vii,, Robert K 1984..Que Stud)' &search: Design and Methods .
[Link], Ad,un, and Henry Teune. 1970. Tht'. Wglc Qf >..ppli_
cd Scx:i~I Rctc&rcb [Link].i, vol. 5 . .Bovcriy
Comµrati\•£ Social Inquiry. New Yon:'~ John Wiky. Hill,; [Link] Public:ationa,
Ragin, Charles C. 19S7. Tht Camptu"(,l!i-.·c Mt-Utod: Moving beyond [Link], Jr., Morris. 1971. ~1n1eUi2-ihlo Con'[Link]. • In
(lu'211tadyt: ar4 Quantir.o:iw Straugies. ~rktfc.y [Link] Lot Qmip,m:Ui~ M'e:Jwds in Sociology: &i~i Oft Trends and
Angel«; Universily of [Link] Prcw. [Link].s, ed. lv.a.o [Link]. &ri:clcy and Lo. Angele,;:
R:[Link], Stein, 1970. Cithau, EJu-riom, Pdrrit:s: Appr:[Link] u, 1M UniverJity o( CaJjfomia Pru.,
C-omparo.r.l11c Study of fr<x.C$US ofDewlopment. New York:
David McKay.
Rue&eahemeyer, Oieln-tb, Evelyne Hu.~r S~hen•, and John 0 .
Steph-cns. 1992. Caplui141 Dei,[Link] and Democracy.
Ch1cal!o: University of Chfe-ago Prus.
Ru5W\I>' , Dtinkwan. 1968. "Mode ~ Qn and Compu11ivc Politi,:,-:
Pto11r1:cu in Research and Theory,• [Link].e Politfos
l:[Link] .
Sar<ori . [Link]. l 970. "Conctpl MiJfomatiOn in Cornpantivc
Potilic,. • ,tmcri4,a11 [Link] Science Review 64: JOJJ--S·J.
Sanori. Giovanni, ed • .l934. Socio! Sclawt' Concepu: A Sy-&umadc
[Link]$, Beverly Hilla. CA: Sage.
Senori, Gi<lvl:llN. 1991. "Com;:,erif\B ead [Link]. • Joul1'JIJJ of
Thel)uliCol l'qflllr:1 J :143,57.
Sanori, Giovanni. 1993 . -rOl&[Link], Model Mania, and Leaming
from Em>r.· l t>[Link] of1hcort-ti4-ol P<>liriu,
Sa.rt0ri, GiovaMi., Fttd W. Rigss, aod Henry Tcune. l97S. Towu of
&ib<.l: On th< Dejiiildo,t and AM}ysfs of C<m«p4 in Ike
Sodal Sci~nc-es. lmematiotul Swdies Association,
O,;.[Link] Paper No. 6, Unive~i1y of Piu.&bofgb.

View publication stats