Methodology Forum

Comparative Political Studies

43(2) 230­–259
The Strategy of Paired © 2010 SAGE Publications
DOI: 10.1177/0010414009350044
Comparison: Toward a http://cps.sagepub.com

Theory of Practice

Sidney Tarrow1

Abstract
Paired comparison is a strategy of political analysis that has been widely used
but seldom theorized. This is because it is often assimilated to single-case
studies or regarded as a degenerate form of multicase analysis. This article
argues that paired comparison is a distinct strategy of comparative analysis
with advantages that both single-case and multicase comparisons lack. After
reviewing how paired comparison has been dealt with in comparative politics,
the article details a number of its advantages and pitfalls, illustrates them
through the work of four major pairing comparativists, and proposes what is
distinct about the strategy. It closes with a number of suggestions for using
paired comparison more effectively.

Keywords
paired comparison, methodology, comparative theory

This article examines a strategy of comparative analysis that has been widely
used but little theorized—paired comparison. It has been used since Alexis de
Tocqueville used it implicitly across his two most famous books, Democracy
in America and The Old Regime and the French Revolution. And it has
recently been used effectively by authors as different as Valerie Bunce, Peter
Hall, Peter Katzenstein, Seymour Martin Lipset, Robert Putnam, and Richard
Samuels. Tocqueville adopted the method as part of a voyage of discovery.

1
Department of Government, Cornell University, Ithaca, NY, USA

Corresponding Author:
Sidney Tarrow, Department of Government, Cornell University, Ithaca, NY 14851, USA
Email: [email protected]
Tarrow 231

But like many who have used the strategy more systematically—including
the present author1—he had no theory of practice to guide him. Others have
thought about the method but have usually seen it as a variant of single-case
study or as a degenerate case of large-N analysis. Developing a theory of
practice for this widespread strategy of comparative politics is the goal of
this article.
What do I mean by “a theory of practice”? I mean something very similar
to what Charles Tilly meant when he wrote that no one

can pursue empirical social research effectively without deploying and

testing two interdependent bodies of theory simultaneously: a theory
embodying explanations of the phenomenon under investigation, and
another theory embodying explanations for the evidence concerning
that phenomenon. (2008, p. 47)

The “theory of practice” to which the subtitle of this article refers is the
second one: how evidence collected from the practice of paired comparison is
assembled, is evaluated, and can be judged. I argue that paired comparison is a
method of political analysis distinct from both single-case studies and multicase
analysis. Moreover, although paired comparison is most often associated
with qualitative approaches, it is compatible with a variety of specific
methods. First, I review the ways in which the method has been used and
then how it has been treated in the growing how-to literature in comparative
politics. Second, I argue that it is compatible with both most-similar and
most-different cases and with both correlational and causal-process logics
of analysis, illustrating my argument with summaries from the work of
four distinguished comparativists. Next I say what I think is distinct about
paired comparison and offer some reasons why scholars in comparative
politics have been attracted by it. Fourth, I add some cautionary notes about its
major pitfalls but try to respond to each of them. Finally, I propose some
provisional answers to the question “How can the strategy of paired
comparison be used more broadly and more effectively?”

A Widespread and Successful Practice

The practice of paired comparison has been used in a variety of sites and
settings. It has been used to compare similar and different countries; to study
large processes; to compare voters, parties, and public opinion; to study politi-
cal contention; and to examine different levels of the polity.2 With such
widespread use of a method in comparative politics, one might suppose that
232 Comparative Political Studies 43(2)

it would be easy to find practitioners who have theorized carefully about its
uses and limitations. But apart from the obvious idea that two cases tell us
more than one, scholars have not yet addressed four questions that are
important in constructing any theory of practice:

• What is distinct about paired comparison?

• What are its strengths and weaknesses?
• What are the different ways in which paired comparison has been
used?
• How can it be used more systematically and effectively?

This article proposes some answers to these questions.

Thinking About Paired Comparison

Although the literature on comparative methodology has contributed much
to our understanding of both single-case studies and large-N analysis and
about qualitative comparative analysis (Ragin, 1987, chapters 6-8), it has
had little to say about paired comparison per se. Many authors either see
paired comparison as “case study plus one” or as a degenerate version of
large-N comparison.
Neither of the two major theoretical statements in the quantitative tradi-
tion gives paired comparison more than passing notice. Adam Przeworski
and Henry Teune’s Logic of Comparative Social Inquiry (1970) assumed the
regression equation as the modal form of comparative analysis. Although
they occasionally cited work that used paired comparisons (Converse &
Dupeux, 1962; Dogan, 1967) and even compared pairs of cases here and
there (1970, pp. 80, 128). In dismissing “proper names” as viable subjects
for study, Przeworski and Teune implicitly guided their readers away from
paired comparison.
Gary King, Robert Keohane, and Sidney Verba’s Designing Social
Inquiry (1994) is more nuanced. Generous to a fault in offering the advan-
tages of causal inference to their qualitatively oriented cousins, King,
Keohane, and Verba found much to praise in single-case analysis but rele-
gated paired comparison to the status of a residual category, one that “can
yield valid causal inferences” but usually remains “essentially descriptive”
(King et al., 1994, p. 45).
When we turn to the qualitative tradition, we find that even its staunchest
advocates, Henry Brady and David Collier, offer little guidance for using
paired comparison in their diverse edited volume Rethinking Social Inquiry
Tarrow 233

(2004). Collier, together with his partner, Ruth Berins Collier, is a past master
of the method (Collier & Collier, 1991). But among the contributions by Brady,
Collier, and their collaborators, I found only three direct references to “paired
comparison” (pp. 94, 247, and 265) and only one to “matching cases” (p. 108).
Practitioners of comparative politics have recently begun to think seri-
ously about the practice of paired comparison. In their landmark book,
Case Studies and Theory Development, Alexander George and Andrew
Bennett made a strong case for the method of “controlled case comparison”
(2005, pp. 59, 80, chap. 8). George and Bennett do not ignore the difficul-
ties in using paired comparison. But they argue that some of the problems
can be addressed by careful attention to process tracing, an argument to
which I will turn below. Their main advice is to shift from lateral paired
comparison to before-and-after analyses of the same case (pp. 81-82). But
this is only another way of carrying out the method of process tracing that
they favor and offers little help to those who wish to compare different
political systems.
The most exhaustive examination of case methods, John Gerring’s Case
Study Research (2007), gives the technique of “matching cases” two pages
but then assimilates paired comparison to the single-case study, noting that
“case study research may incorporate several cases, that is, multiple case
studies” (p. 20). Gerring does offer an excellent examination of paired com-
parison in his discussion of case selection (chap. 5) and of internal validity
(chap. 6). But his definitional assimilation of “a few cases” to single-case
analysis does not help us to understand if there is anything particular about
paired comparison.3
We have a long way to go in understanding the strengths and pitfalls of
paired comparison. We can begin by distinguishing two dimensions that appear
regularly in the literature on paired comparison: between most-different and
most-similar systems and between correlational and causal process analysis.

Comparing Most-Different
and Most-Similar Systems
Scholars who have reflected on the strategy of paired comparison almost
invariably turn to the methods proposed by John Stuart Mill—the methods of
similarity and difference. But what Mill was describing were varieties of the
experimental method and not of the comparative analysis of social or politi-
cal phenomena. And he saw that there would be many situations in which
neither the method of agreement nor the method of difference would suffice
but in which a combination of the two would be necessary—what he called
“the joint method of agreement and difference” (1973, pp. 394-396).
234 Comparative Political Studies 43(2)

Some scholars have seen Mill’s two logics as similar to Przeworski’s and
Teune’s distinction between most-similar and most-different systems analysis
(Gerring, 2007, p. 139). I think these are different dichotomies and that the
latter is more useful to social or political analysis than the former. According
to Przeworksi and Teune and Gerring, there are two main ways of carrying out
comparative analysis:

• Most-similar systems design, in which common systemic character-

istics are conceived of as “controlled for,” whereas intersystemic
differences are viewed as explanatory variables (Przeworksi &
Teune, 1970, p. 33)
• Most-different systems analysis, in which “variation on the inde-
pendent variable is prized, while variation on the outcome is
eschewed” (Gerring, 2007, p.139).4

The first method implies minimizing the differences between the systems
being compared, whereas the second implies maximizing these differences.5
Przeworski and Teune and others have criticized most-similar systems
designs because “the experimental variables cannot be singled out” (1970,
p. 34). This is certainly true, but as Gerring and others have pointed out,
comparing similar systems has some advantages:

• Even if the specific variables that unite systems cannot easily be

sorted out from one another, the method can direct attention to the
ways in which they differ (Gerring, 2007, pp. 133-135).
• In exploratory research or in the exploratory phase of more rigorous
designs, a second case can confirm a tentative finding from a single
case (Gerring, 2007, p. 131).
• “Matching” techniques can be used to select cases for intensive
comparison within a large-N sample (Gerring, 2007, pp. 135-137).
• Besides, some cases are inherently interesting because they take
contrasting routes to similar outcomes—like the differences in the
paths of democratization of Britain and France (Tilly, 2004).

The most-different approach implies a different strategy. As Gerring

puts it, in this approach “the researcher tries to identify cases where just
one independent variable…as well as the dependent variable…covary, while
all other plausible factors…show different values” (2007, p. 139). An example
is the “most-different systems” design used by Valerie Bunce in her Do New
Leaders Make a Difference? (1981). In the face of the enormous differences
between these two types of system, leadership succession appeared to produce
Tarrow 235

a fundamental similarity—early policy innovation. Using a combination of

budgetary data and policy case studies, the similarities proved to be robust
across policy areas and using different kinds of data. New leaders made a
difference because they implemented new or revised policies in a spurt of
policy innovation that slowed down as their terms proceeded.
Bunce saw the process of succession in two parts: the campaign for office
(e.g., how power is won and maintained) and the honeymoon of new leaders
once they are in office. For both periods Bunce outlines a number of processes
leading new leaders to innovate: In the West the opening up of issues during
the campaign puts pressure on successful candidates to take action once
elected (1981, pp. 225-226); in the East successions, “which provide provi-
sional mandates…are solved only when one contender manages to convince
the other…elites around him that he in fact can govern by responding to their
needs and those of the system” (p. 22).
Most-different case analysis is particularly prone to the “many variables,
too few cases” criticism that has been leveled at small-N studies in general.
But it has advantages that elude similar case analysis:

• Drawing attention to similar or identical processes in a wide variety

of cases can expand or limit the scope conditions of established
research findings (McAdam, Tarrow, & Tilly, 2001, chap. 10).
• Examining outliers within a large-N population of cases can help to
identify the variable or variables responsible for general outcomes
of interest in core cases when these are reversed in the case of the
outlier.
• Widely dispersed cases may be of political interest even when they
break some of the canons of best-case analysis.

Neither most-different nor most-similar system analysis is inherently

superior. Which we choose depends on the problem under investigation
and not on some imaginary “best practice.” Tocqueville (without knowing
the term) chose a distinctive most-different systems analysis in his two
books on France and the United States, whereas Putnam, although in debt
to Tocqueville, chose a most-similar systems analysis in comparing
northern and southern Italy. Each illustrates both the virtues and the pitfalls
of the strategies they chose.

Tocqueville in America
Because Democracy in America and The Old Regime and the French Revolu-
tions appeared decades apart and are read by different groups of readers, their
236 Comparative Political Studies 43(2)

fundamentally comparative structure is often overlooked. But both books

address the same fundamental problem: the role that state centralization
plays in relation to intermediate associations and democracy. Between
them, they reveal that Tocqueville was engaged in a “voyage of discovery,”
one that left him susceptible to seeing America through French eyes and,
incidentally, left him unable to see the potential for rupture in America’s
contentious politics.
Why did the revolution break out in France, Tocqueville asked, rather than
in one of Europe’s more backward countries (1955, p. x)? His answer was that
state centralization denuded the aristocracy and other corporate groups of
their positive functions, reducing them to parasitic burdens on society and
leaving that society without its “natural” sources of civic initiative and bal-
ance. Because a society stripped of intermediate bodies lacked a buffer
between state and society, Frenchmen became “self-seekers practicing a
narrow individualism and caring nothing for the public good” (1955, p. xiii).
The result was jealous egalitarianism, sporadic and uncontrolled mobilization
and, ultimately, The Revolution: “a grim, terrific force of nature, a newfangled
monster, red of tooth and claw” (1955, p. 3). The lesson Tocqueville drew was
that the stronger the state, the weaker would be civic participation and the
greater the incentive to violence when collective action did break out. No one
would want to live in such a state, and after a decade of terror and chaos, a
despotism more absolute than the Old Regime resulted. Thus Tocqueville
arrived at one of his more brilliant insights: that rather than the antithesis of
the Old Regime, Napoleon’s empire was its natural successor.
Traumatized by the French experience of democracy/despotism, when
Tocqueville came to America he found the lack of a strong central government
invigorating. In his view of America, decentralization played an important
role, both in leaving space for intermediate associations to flourish and
fracturing potentially oppressive majorities that might—as in the First French
Republic—have led to despotism. Thus, the history of his own country’s cen-
tralization lay implicitly behind his understanding of the wellsprings of
pluralistic democracy in America.
But there was an aspect of American politics that Tocqueville didn’t
understand very well: the nature of our contentious politics. Because he did
not see the kind of ideological politics he had gotten used to in France
(recall that he came to America to escape the results of the 1830 revolu-
tion), Tocqueville failed to see the turbulence that lay just below the surface
of American politics. Slavery he abhorred, both on ethical grounds and
because it produce a lazy and corrupt white plantocracy; but because he
admired decentralization, he failed to see the potential for states’ rights to
produce a civil war.
Tarrow 237

Tocqueville’s chain of causation from centralization to civic atomization,

and from there to extremist democracy and thence to a new despotism, may
or may not have been correct for France. The important point is that it led
him to transfer to the United States—in inverted form—the connection he
saw in France between centralization and intermediate bodies and to ignore
the fundamental differences in the two countries (1955, pp. 61-68). Turning
the coin of centralization/decentralization onto its American side, he pos-
ited that a state without a centralized bureaucracy would not smother local
initiative and would avoid the dangers of the cycle of extremism and tyr-
anny he had witnessed in France.
Beginning from structuralist preconditions, Tocqueville elaborated
what we today would call a most-different systems design. Combined with
his intuitive talent as an observer, his contrasts between centralization/
decentralization and old regime/new state gave zest to his accounts of both
countries’ histories. But his voyage of discovery to America was too
deeply impregnated with a Frenchman’s fears and preconceptions, and he
failed to see that the differences between the French old regime’s corpo-
rate structures and the new American state’s associations were too wide
for effective comparison. The method of comparing most-similar systems
adopted by Robert Putnam effectively attacked just this problem.

Putnam in Italy
Using a most-similar systems design, Robert Putnam and his collaborators
studied the determinants of policy performance in the two major regions of
Italy. Structured by the same political institutions, the two regions responded
in diametrically opposed ways to the same regional reform. Putnam and his
collaborators used a series of paired comparisons of these similar systems to
show how unevenly the institutional innovation of the regional reform affected
the practice of administrative politics. They argued that the differences were
the result of the two regions’ very different cultural “soils.” The central
inference in the book was that the same reform in the same country produced
regional policy performance differences because of demonstrable differences
in the civil involvement and civil competence in each region.
When he traveled back to the late-medieval origins of north-central Italy’s
city–state governments and to the simultaneous development in the south of
an autocratic Norman regime, Putnam (1993) discovered the regional differ-
ences he had found “had astonishingly deep historical roots” (p. xiv). In both
regions, he found analogies to the divergent civic capacities that he had iden-
tified in his data (pp. 121-123, 131). When he reached the 19th century,
238 Comparative Political Studies 43(2)

statistical data became available that showed that “the same Italian regions
[in the north] that sustained cooperatives and choral societies also provided
the most support for mutual aid societies and mass parties” and that “citizens
in those same regions were most eager to make use of their newly granted
electoral rights” (p. 149). In the south, in contrast, “apathy and ancient vertical
bonds of clientelism restrained civic involvement and inhibited voluntary,
horizontally organized manifestations of social solidarity” (p. 149). Most
important, he found that civic participation in the late 19th and early 20th
centuries correlated strongly with his contemporary indices of civic capacity
(r = .93) and with institutional performance (r = .86).
But there was a problem. When Putnam and his collaborators turned to
causal inference about their findings, they never considered the impacts of state
and social structure on civic capacity in the two regions. Every regime that gov-
erned southern Italy from the Norman monarchy to the unified government that
took over in 1861 governed the region with a logic of colonial exploitation. Nor
did southern Italy’s semicolonial status suddenly disappear with unification.
The new Italian regions were installed in different civic soils in northern and
southern Italy, but part of that difference was a public culture shaped by over a
hundred years of political and administrative dependency on the central state
(Pizzorno, 1971, pp. 87-98). There were unmeasured intervening variables
between the civil and uncivil cultures that Putnam focused on that, for Putnam’s
critics, were a more plausible explanation for the differences he encountered.

Correlational Logic Versus Process Tracing

The second major dimension in how paired comparison has been used is
between the correlational strategy used in large-N analysis and the causal pro-
cess models more familiar from single-case studies. Although the classical
way of structuring a paired comparison begins with differences or similarities
in outcomes and posits independent variables that are thought to explain these
outcomes, causal process analysis relates variables to one another over time.
Paired comparison of this type is what we can call “duel-process tracing.”

The Logic of Correlation

Despite the small number of cases involved, the logic of relating antecedents
to outcomes is the same as the strategy of correlation. The researcher puts
side-by-side sets of antecedent conditions—usually at the systemic level—
with outcomes of interest, to infer causal relations between them. So far, this
Tarrow 239

sounds very much like large-N research, except for two factors, one negative
and one positive:

• Negatively, the comparison of two cases lacks the degrees of free-

dom of large-N comparisons.
• Positively, it allows for and indeed demands a degree of intimacy
and detail that inspires confidence that the connections drawn
between antecedent conditions and outcome are real.

The Logic of Process Tracing

Although most applications—and many criticisms—of paired comparison
assume the logic of correlation, process tracing has earned a revival among
students of comparative politics (Brady & Collier, 2004; George & Bennett,
2004). In successive work, Bennett has argued for a merger between pro-
cess tracing and a Bayesian approach to evidence. In the latter article, he
argues that “the goal is to document whether the sequence of events or
processes within the case fits those predicted by alternative explanations of
the case” (2008, p. 705). The founding of a new section of the American
Political Science Association (APSA) on “qualitative and multimethod
research” has helped to diffuse these approaches.
The “step-by-step” approach to process tracing has been criticized as a form
of story telling (Tilly, 2002). Stories, he argues, are an instinctive human way
of ordering experience, but they do not necessarily aid in producing causal
analysis. In a recent article, Caporaso takes a more conciliatory view, arguing
that narratives can be broken down into mechanisms (2009). This is similar to
the view taken by McAdam et al. (2001), who argue for breaking down large
political processes (e.g., democratization, revolution, nationalist episodes) into
their constituent mechanisms, defining mechanisms as delimited class(es) of
events that alter relations among specified sets of elements in identical or
closely similar ways over a variety of situations.
A number of important studies in comparative politics are based on the
strategy of what I call duel-process tracing. For example, Bunce’s study of
leadership turnover in the Soviet Union and the West (1981) showed that the
leadership selection process in both countries produced similar “honeymoon
periods” in which policy innovation was heightened and public spending rose.
By observing these mechanisms and processes, Bunce specified the linkages
between independent and dependent variables of interest. For Bunce, as for
others who use paired comparison to identify what is key in political processes,
“mechanisms seek to connect background factors (often structural) with a defi-
nite but more remote outcome” (Caporaso, 2009, p. 71).
240 Comparative Political Studies 43(2)

Critics loyal to the logic of correlation sometimes point out that process
tracing leaves a vacuum at the end of the analytical trail: For example, they
want to know what are the outcomes of the mechanisms and processes that
are so lovingly detailed by process analysts? There are two main answers to
this charge:
First, laying out the mechanisms and processes that connect independent
and dependent variables can show how independent and dependent vari-
ables are connected, as Caporaso argues (2009). A good example is Jeremy
Weinstein’s comparison of two types of violent civil war, in one of which
civilians are more brutally and more widely targeted than in the other.
Weinstein might have been satisfied by discovering this outcome and trac-
ing it to structural preconditions. But he was more interested in the
mechanisms that produced the different degrees of violence, one of which
he called an “investment” strategy and the other a “consumption” strategy.
In the cases he studied, although the first mechanism produced less violence
directed against civilians, the second produced more indiscriminate violence
(Weinstein, 2006).
The second, and more far-reaching, answer is that scientific causation does
not always take the form of the explanation of specific outcomes. In an equally
distinguished scientific tradition, the process itself is the outcome; accounting
for it involves a logic of process tracing (Bunge, 1997; Tilly, 2002). An exam-
ple is Collier and Levitsky’s important article “Democracy With Adjectives”
(1997). Properly seen, their goal was not simply to classify different kinds of
democracies but to specify the different mechanisms that can drive the process
of democratization and their interaction. If we want to know why a particular
outcome emerged, we need to understand how it occurred.

Skocpol on Revolutions and Almost-Revolutions

Skocpol’s States and Social Revolutions provides a useful example of both
the strengths and the pitfalls of the correlational design in paired comparison.
Skocpol posited two main systemic independent variables to explain the rev-
olutions she studied in France, Russia, and China: external pressure and
fiscal crisis. This was a largely structuralist model of explanation. It led her
to give short shrift to factors internal to the revolution, as in her oft-quoted
aphorism dismissing purposive action as a cause of revolution (1979, pp. 5,
14-18).6 She then briefly discussed two shadow cases, Japan and Prussia (pp.
99-10), in which social revolution did not occur, pointing to the absence in
these cases of the antecedent conditions that, she argued, produced revolu-
tions in her three main cases.
Tarrow 241

But in dismissing the importance of purposive action, Skocpol radically

underspecified the political processes in both the three cases she studied in
detail and even more so in her shadow cases. Between revolutionary situations
and revolutionary outcomes lie a host of mechanisms and processes, some of
them purposive, others environmental, and still others relational (Tilly, 2000).
That the revolutionary process itself is not a worthwhile object of analysis is
surely wrong, as many of Skocpol’s successors have argued (Foran, 1994;
Goldstone, 1998; Tilly, 1993).

Samuels in Tokyo and Rome7

Richard Samuels took a voyage of discovery every bit as wide geographically
as Tocqueville’s but narrower than Skocpol’s in terms of the systems exam-
ined. In Machiavelli’s Children (Samuels, 2003), he examined the variety of
ways in which leaders mobilize fragments of the past in order to bring about
change. Samuels carried out a loosely parallel narrative of two states—Japan
and Italy—whose modern forms developed at the same time and in a similar
international context: both were attempting to catch up to the early risers.
Despite the differences between the two countries (Samuels at one point
recalls Neil Simon’s The Odd Couple), he insists that “few nations have as
many important common features”:

Neither Italy nor Japan even existed as modern states when Great Britain
and the United States embarked on their industrial revolutions. In both
Italy and Japan late industrialization was accompanied by a groping
ex-authoritarianism. And they paid the same price—devastation in the
Second World War and subordinate roles in the new American world
order. (Samuels, 2003, p. xii)

Both countries built centralized prefectoral systems of territorial governance;

both would pass through authoritarian interludes that placed them in armed
conflict with the democratic states of the West; and both democratized under
the not-always-gentle hand of U.S. hegemony after defeat in World War II.
Following brief periods of postwar cooperation between left and right, both
were led for most of the succeeding half-century by anti-Communist catch-all
parties of the center-right whose interests were deeply imbricated with the state
(Pempel, 1990). Yet for most of this period, strong left-wing opposition parties
persisted in both countries.
The two countries also followed similar economic trajectories. Both made
astonishing economic recoveries from World War II, based on the exploita-
tion of initially cheap labor, advantageous foreign trade positions, and
242 Comparative Political Studies 43(2)

entrepreneurial innovation. And after catching up, both suffered recurring

economic crises and a pattern of stop-and-go economic prosperity.
Samuels identifies “Machiavelli’s children”—the innovative leaders in mod­
ern Japanese and Italian history—with three elite “mobilizing mechanisms”—
inspiring, buying, and bullying. By inspiring, he means the use of symbolic
politics to control the way the collectivity is represented to itself; by buying he
means the strategies of accommodation, trading, sharing, brokering, balanc-
ing, and co-optation that leaders adopt to purchase support; and by bullying he
means using power to weaken, surround, and squeeze adversaries (Samuels,
2003, p. 9). These mobilizing mechanisms were used in different combina-
tions to control the political agenda.
Samuels’ book shows how these three mechanisms combined in modern
Italian and Japanese history in different ways to produce processes of political
mobilization and control. After detailing the parallels in their 19th and early
20th century trajectories, he turns to their postwar developmental paths.
These offered even more interesting intersections—especially in the “dom-
inant party systems” that each developed (Pempel, 1990). Although both
postwar ruling parties—the Japanese Liberal Democrats and the Italian
Christian Democrats—were “first, and above all else, anticommunist,” nei-
ther party was especially “inspiring”: in fact, their anticommunism was
compounded of a mixture of pleasing their hegemonic ally, fear for property,
and out-maneuvering domestic opponents. Of “bullying” there was less than
in the earlier periods, although the Italian police at first showed no hesitation
in mowing down demonstrators and Japanese nationalistic and right-wing
gangs were mobilized by the U.S. Occupation Forces and later by the LDP to
suppress left-wing groups (Samuels, 2003, p. 189).
The major mobilizing strategy of both postwar elites centered around
Samuels’ third mechanism—“buying.” He writes;

each “digested” its opponents with consummate political skill, gener-

ously distributing material benefits, enticing the extremes to abandon
anti-system views, and establishing themselves a popular image as
indispensably sturdy democrats and friends of the American protector.
(Samuels, 2003, p. 186)

Samuels’ elite mechanisms avoid the cultural determinism that has dogged
accounts of both Italian and Japanese politics since the end of World War I.
But the continued dominance of the Liberal Democrats until 2009—
compared with the collapse of Christian Democracy in Italy—is hard to
understand in the absence of the form of the two states.
Tarrow 243

The attractions of Samuels’ mechanism-based account may also have left

Samuel unimpressed with the differences in the Japanese and Italian states. If
Japanese leaders deployed a uniform “developmental ideology,” can this be
detached from the more centralized, more cohesive, and more disciplined
Japanese state—one which became the model for the “developmental state”
(Johnson, 1982)? And if Italian leaders failed to tame the forces that shaped
economic development, can that be independent of the fact that their state
was less centralized, less cohesive, and less disciplined than the Japanese one
(Pempel, 1998, pp. 37-41)?
The variations in these four paired comparisons are legion. To summarize,
Tocqueville un-self-consciously used a most-different system design but
could not erase French centralization and revolution from his mind as he
traveled through the early United States. Putnam moved to a most-similar
systems design to uncover the key independent variable that could explain
the different outcomes of the regional reform in the two main regions of Italy.
Skocpol chose three cases of social revolution in which the same or very
similar antecedent variables combined and showed that they had very similar
outcomes. Her paired comparison indicated that the logic of correlation can
strengthen arguments from similar cases. Finally, Samuels used the similarities
in Japan’s and Italy’s historical development to provide a common canvas on
which to draw out the differences in the mix of mechanisms he found in their
political processes.
As I hope to have shown in this brief overview, there is no one best way
of carrying out a paired comparison, just as there is no one best way of doing
social science in general. Paired comparison is not simply a degenerate form
of large-N analysis, nor is it only a “case study plus one.” It is a distinct
analytical strategy for working through complex empirical and historical
materials using the leverage afforded by the differences and similarities of
comparable cases. But what is distinct about it? And what are its advantages
and pitfalls?

The Distinctiveness of Paired Comparison

Some of the reasons for the widespread use of paired comparison are properties
it shares with case studies in general. First, it provides an intimacy of analysis
that is almost never available to large-N analysis. Second, it draws on—and
indeed insists on—deep background knowledge of the countries being exam-
ined. Third, it facilitates what Brady and Collier call “causal-process analysis,”
in contrast to the “data-set observations” that are the basis of correlational
and regression analysis (Brady & Collier, 2004, p. 277).
244 Comparative Political Studies 43(2)

But paired comparison is distinct from single-case studies in several ways.

Its distinctiveness can be understood through an analogy with experimental
design. It is similar to experimentation in its ability to compare the impact of
a single variable or mechanism on outcomes of interest. Of course, an experi-
menter has the ability to randomly assign subjects to treatment and control
groups, making both groups probabilistically equal on all variables other than
the experimental one. All the pairing comparativist can do this respect is to
attempt to carefully match the confounding variables that he or she knows
about (but not ones that do not come to mind). In addition (see below), given
the small number of cases used it is difficult or impossible to control for more
than a few factors.
However, paired comparison does have an analytical baseline, eliminating
the possibility that the dependent variable can have occurred even in the absence
of the independent variable, thus significantly increasing the inferential power
of the design over the single-case study. And by permitting dual-process tracing,
it reduces the possibility that a supposed determining variable is as critical as it
might seem from a single-case study alone. Finally, as Becker (1968) argued,
paired comparison can add confidence in a “building-block strategy” that moves
from a single-case to a multicase analysis. Katzenstein’s (1984) comparison of
Switzerland with Austria illustrates all three of these properties.

Katzenstein, Corporatism, and Compensation

In the late 1970s a number of scholars who had noticed the frequency of
political economic cooperation at the elite level of European systems coined
the terms neocorporatism and liberal corporatism (Lehmbruch & Schmitter,
1982). Because most of the cases involved were either social democratic
(e.g., Sweden) or were governed by alternately by Social and Christian
democratic parties (e.g., the Netherlands), it seemed logical to see neocorpo-
ratism as a product of domestic partisanship. Katzenstein’s dual comparison
of Switzerland and Austria broadened the scope conditions of the theory and
pointed to the importance of an internal mechanism that was shared among a
number of smaller European democracies regardless of their internal politics.
This pointed to their deep involvement in international trade and to an inter-
nal mechanism that won the support of all major actors for the compromises
necessary for economic concertation.
At first glance Austria and Switzerland looked very different. Austria
looked like a social democratic–oriented welfare state, whereas Switzerland
resembled model capitalist system with a small welfare state and a conservative-
leaning political elite. Both were deeply involved in and dependent on the
Tarrow 245

international economy, which made them comparable. But what was the inter-
nal mechanism that translated international involvement into concertation?
From his close duel-process tracing, Katzenstein observed that regardless of
their internal partisan tendencies, each provided more resources to the politi-
cally less powerful interest group in each system (e.g., business in Austria and
labor in Switzerland) than that group would have received based on its political
clout alone. Compensation kept the minority player in the game. Process tracing
made it possible to go beyond the simple bivariate correlation that others had
traced between international trade and domestic corporatism.
I see three additional reasons for the widespread attraction of paired
comparison in comparative politics:

• Correcting generalization from single cases: Many scholars use

paired comparison as an analytical wedge to complement or contrast
a case they know well or think they do. (In fact, one of its serendipi-
tous virtues is that it helps country experts understand their primary
cases better.) For example, when Chalmers Johnson discovered a
pattern of peasant nationalism behind the Chinese struggle against
Japanese invaders, he turned to the Yugoslav partisan movement to
determine how far his findings could travel (Johnson, 1962, chap. 6).
• Assessing the influence of institutions: The paired comparison of
different political systems allows analysts to use differences in insti-
tutional form as a variable to demonstrate the sources of intrasystemic
behaviors. Just as Italy’s common institutional reform was the criti-
cal hinge on which Putnam’s comparison of northern and southern
Italy turned, institutional contrasts were central to Juan Linz’s and
Arturo Valenzuela’s comparisons of presidential and parliamentary
systems (1994).
• Creating an intermediate step in theory building: A productive use
of paired comparison is as an intermediate step between a single-
case study, which suggests a general relationship, and a multicase
analysis that tests or refines a theory. This use of paired comparison
seems to fit with what Evan Lieberman had in mind with his con-
cept of “nested comparison” (2005). A good example in which
quantitative and qualitative methods triangulate can be found in the
literature on postconflict peacemaking. Large-N studies show that
the power-sharing arrangements that international organizations
and Western states often press upon countries emerging from civil
war are often good at ending internal conflicts but do not produce
durable democratization (Jung, 2008). But statistical correlations on
246 Comparative Political Studies 43(2)

their own tell us little about the mechanisms that produce demo-
cratic reversals and require detailed process tracing, as in the paired
comparison of Bosnia and Mozambique carried out by Jai Kwan
Jung (2008, chap. 5).

Why Two?
If paired comparisons have so many advantages, the sympathetic critic
may ask, why stop at examining two cases? Why not three…four…or x
cases? The answer, I think, is that the move from single-case to paired
comparison offers a balanced combination of descriptive depth and ana-
lytical challenge that progressively declines as more cases are added. The
moment we go from one case to two, I would argue, we are in the realm of
hypothesis-generating comparative study, while also enabling ourselves to
examine how common mechanisms are influenced by the particular fea-
tures of each case; as we increase the number of cases, however, the
leverage afforded by paired comparison becomes weaker, because the
number of unmeasured variables increases.8 This is essentially what
Howard Becker (1968) argued in his building-block strategy. But it is
paired comparison that gets Becker to the theoretical argument, with addi-
tional cases either providing additional evidence or pointing to a more
refined explanation of the findings from the original pairing. This is even
the case in some multicase comparisons, in which the analytical work is
done by paired comparison of 2 × 2 cases, for example, as in Kathleen
Thelen’s How Institutions Evolve (2004).

Pitfalls of Paired Comparison

These advantages have made paired comparison a favored tool of many
comparativists, but it is not without its problems. Depending on how it is
used, it carries a number of pitfalls. The most important are insufficient
degrees of freedom, nonrepresentativeness, atheoretical case selection, and
ignoring scope conditions.

Insufficient Degrees of Freedom

The most common critique of paired comparison is similar to the criticism
made of case studies in general: the absence of enough degrees of freedom to
reliably choose among alternative explanations for outcomes of interest.
Mill’s “method of difference” is the major technique used to limit unmeasured
Tarrow 247

variance in paired comparison. As it is laid out in Gerring’s (2007) book on

case studies, Mill’s method assumes that observable variables will exhaust all
the possible causes of an outcome of interest and that a single difference
among those variables can be identified as the sole cause of that outcome. But
this assumes that all relevant variables can be observed, that there is only one
difference among these variables, and that the correlation between this differ-
ence and differences in outcome can confidently be regarded as causal. These
assumptions are heroic, to say the least (Little, 1998).

Nonrepresentativeness
Nonrepresentativity is a problem for all small-N comparative analysis but is
particularly worrisome in paired comparison. In traditional area studies
comparisons, cases were often chosen because of their familiarity or geo-
graphic proximity to one another, ignoring potential comparisons of cases
that are either unfamiliar or geographically “outside of one’s area.” That left
many theoretically interesting comparisons unexplored; some of them, like
Johnson’s comparison of China and Yugoslavia, have already been men-
tioned. Two others, Bunce’s comparisons of leadership succession in the
USSR and the United States, and Samuel’s of Italy’s and Japan’s political
development, were examined earlier.
The pitfall of nonrepresentativity was most famously pointed to by Bar-
bara Geddes in her critique of Skocpol’s States and Social Revolutions
(Geddes, 1990). Geddes argued that with a few cases on the extreme end of a
continuum from nonrevolution to revolution, Skocpol could not legitimately
claim that her independent variables of interest were in fact connected to the
social revolutions she had studied (Geddes, 1990, pp. 142-143).

Atheoretical case selection

Atheoretical case selection was also common in traditional area studies.
Countries that have gone through similar experiences (e.g., the losers in
World War II), which go through similar phases at the same time (e.g., as
parts of “waves of democratization”), or are simply in the same part of the
world are obvious subjects for paired comparison. “Find a second case” is
sometimes the casual advice we hand out to graduate students when they go
to the field; less often, we provide them with advice about how to select cases
so that their findings cannot be dismissed by critics as no more than the result
of the noncomparability of their cases.
248 Comparative Political Studies 43(2)

Ignoring Scope Conditions

Waves of theoretical innovation in the social sciences often produce efforts
to expand the scope conditions of a theory. This is a fruitful exercise but it
can easily produce the extension of theories beyond their feasible range.9
For example, when Ronald Inglehart developed his important theory of
postmaterialism, it was seen as an effect at the individual level of changes
common to Western countries that had enjoyed a high level of prosperity
during the economic “miracle” of the 1950s and 1960s (Inglehart, 1977).
Armed with the world values study, it was natural for Inglehart and others
to apply the methods and the model to other parts of the world. When they
did so, it emerged that “postmaterial” attitudes are widespread in countries
that it would be difficult to see as advanced industrial societies (Inglehart,
Basánez, Díez-Medrano, Halmann, & Luijkx, 2004). This is not to question
the importance of Inglehart’s findings; it is only to point out that extending
the original theory to interpret them goes well beyond the scope conditions
of the theory.

Digging Out of the Pit

These four pitfalls are not unique to paired comparison. As in most method-
ological discussions, plausible responses can be offered to each of them:

Degrees of Freedom
Although some scholars regard the degrees of freedom problem as fatal, I do
not agree. Intensive case studies produce multiple observations and paired
comparison doubly useful ones. Consider Tarrow’s (1977) analysis of Italian
and French local elites: There were multiple observations in that study con-
verging on the finding that French mayors are primarily part of the administrative
pyramid, whereas Italian ones are fundamentally partisan actors. But of course,
this places a heavy burden on case selection and calls for vigorous attention to
problems of unmeasured variables and the limited diversity of most political
variables (Lijphart, 1975; Ragin & Sonnett, 2005).

Nonrepresentativity
Of course, single paired comparisons lack the capacity to impose controls
by adding new units to the comparison. But as Harry Eckstein long ago
wrote (1975), the function of case studies is not always to represent larger
Tarrow 249

universes; they may be “crucial cases” to demonstrate a theory, or they may

be deliberately chosen because they are outliers. The paired comparison of a
case that lies on a regression line with one that is far off that line can fre-
quently teach us more about both mainstream and deviant cases than a
comparison of many cases scattered across a regression matrix.

Atheoretical Case Selection

Much of area study is based on countries that are close to one another or
familiar to the researcher, but this type of area study is no longer as common
as it once was (Hall & Tarrow, 1998). More typically today, area studies
scholars choose countries for paired comparison because they are compara-
ble but exhibit different processes or outcomes, for example, as in Bunce’s
(1999) comparison of violent and nonviolent routes out of state socialism or
in McAdam and colleagues’ comparison of revolution, democratization, and
nationalism in widely spaced systems (2001, chapters 8-10).

Ignoring Scope Conditions

Expanding the scope conditions of a theory too far is certainly a problem for
paired comparison, but it is a problem for general comparison as well. The
solution for large-N comparisons is to examine outliers within the data set,
whereas the solution for paired comparison is the addition of third, fourth, or
nth cases that deliberately reverse the value of the independent variable that
appeared to explain the outcomes of the original pair. Paired comparison is a
method that can easily be extended, examining whether the scope conditions
within which a theory is developed appear to work in other research sites. A
classical example was how Juan Linz developed the concept of authoritarian-
ism (1975). Linz was unhappy with the Cold War typology of democratic and
totalitarian states. By adding Spain to the nondemocratic spectrum of states, he
pinpointed its differences from the classical totalitarian model—mainly its lim-
ited pluralism—and developed the concept of authoritarianism that gave rise to
a rich literature of case studies, paired comparisons, and large-N studies.

Conclusions: No Magic Bullet

This has not been a research monograph or even a research agenda, and one
should hesitate before offering “rules” for paired comparison to the wide
range of scholars who use it. So I will limit myself to suggesting four ways
through which paired comparison can be systematized and improved: increasing
250 Comparative Political Studies 43(2)

methodological plurality, progressively testing scope conditions, localizing

theory, and combining correlational with process-tracing methods.

Expanding Methodological Plurality

Because the case study tradition mainly grew up through the use of qualitative
and historical methods, it has sometimes been associated only with “soft”
methods of analysis. That has allowed advocates of “hard” methods to dis-
miss findings that result from paired comparison as mere story-telling and
led those who use historical or narrative forms of analysis to embrace it to the
exclusion of other methods. But paired comparison can gain analytical lever-
age from a variety of methods that allow analysts to triangulate on the same
research questions from different angles (Tarrow, 2004). Paired comparison
is methodologically rich and diverse.
A number of pairing comparativists have effectively combined quantitative
and qualitative methods of comparison. Cleary and Stokes (2006), Putnam
et al. (1993), Tarrow (1967, 1977), Weinstein (2006), and Wood (2000) have
deliberately turned a battery of methods on their research questions. Wood’s
comparative study of two regions in El Salvador is a good example; she used
surveys, archival materials, interviews, ecological analysis, and agrarian prop-
erty mapping to show how and where insurgency emerged (Wood, 2002).
Comparativists are even experimenting with quasi-experimental designs, in
which survey respondents are divided into a subject group exposed to a stimu-
lus and a control group which is not (Gibson, 2008).
In a real departure for paired comparison, we are beginning to see creative
uses of matching in concert with statistical methods. For example, in biosta-
tistics, Rosenbaum and Silber argue, “matching facilitates thick description of
a handful of cases, and such scrutiny of cases benefits statistical studies”
(2001). Abadie and Gardeazabal use “synthetic controlled” matching methods
to estimate the effects of such factors as terrorism, public policy innovations,
and other changes on outcomes. They created a simulated Spanish region
without terrorism and estimated that GDP declined about 10 percentage points
in the Basque country relative to a simulated region without terrorism (Abadie
& Gardeazabal, 2003). More recently, Abadie and his collaborators have
applied similar methods to the effect of a tobacco control program in Califor-
nia and to the economic impact of German unification (Abadie, Diamond, &
Hainmueller, 2007). As the traditional wall between quantitative and qualita-
tive methods breaks down we should see more and more use of paired
comparison as a tool together with statistical work in comparative politics
(Brady, Collier, & Seawright, 2006; Caporaso, 2009).
Tarrow 251

Progressively Testing Scope Conditions

Paired comparison sometimes appears to be an iron cage in which the cases
studied become the determining sites to define and limit the variables
inducted for analysis. We saw an example of the perils of such a procedure in
Tocqueville’s determination to see the United States as the mirror image of
France. But no single pair of cases exhausts all potential variables of interest.
The adventurous researcher will extend generalizations emerging from an
initial paired comparison to additional research sites—what Richard Samuels
aptly calls the 3-C strategy (1999).
“It ain’t easy,” as Samuels eloquently shows. To imagine how such a
procedure might affect theoretical conclusions, imagine a Tocqueville who
returned to earth today and, with the aid of statistical methods not available
to him in the 1830s, added a third case to his paired comparison: let us say
the France of the Fifth Republic. He would find a state that is still quite
centralized but in which associational capacity is high—by most accounts,
comparable to that of other European countries. Yet France today is as estab-
lished a liberal democracy as the United States, so the causal links Tocqueville
imputed from centralization to associational capacity and from there to
democracy would need to be questioned.
A successful example of the testing of scope conditions through progres-
sive paired comparisons can be seen in the ongoing research of Valerie Bunce
and Sharon Wolchik on the electoral “revolutions” in east central Europe, the
Caucasus, and central Asia (2006a, 2000b). The strategy of attempting to
overthrow semi-authoritarian rulers by exposing and then contesting corrupt
electoral practices through mass protest was first used in Slovakia in the late
1990s; it was then used as the cornerstone of the overthrow of Milosevic in
Serbia in 2000, before diffusing to the Caucasus and central Asia. As the
strategy diffused from its origins in the more advanced nations of east central
Europe to the Balkans, the Caucasus, and then central Asia, it was less and
less successful. This extension of scope conditions permitted Bunce and Wol-
chik to zero in on the values that appeared in the early risers but are missing
in the latecomers.

Localizing Theory
Paired comparison affords researchers the possibility of what Tilly calls “indi-
vidualizing comparison”—for example, clarifying what is distinctive about a
particular case’s dynamics and trajectory (1984). Tilly used paired comparison
in this way in his book on the Vendée rebellion in the west of France during the
252 Comparative Political Studies 43(2)

revolution (1964). In contrast to most historians, who had seen the rebellion as
the result of religion and legitimacy, Tilly saw urbanization as the fulcrum on
which the counterrevolution turned. He used a paired comparison of two
adjoining areas in western France to gain analytical leverage.
Tilly had to use mainly archival methods to localize his theory of the
connection between urbanization and republicanism in revolutionary
France. But Putnam and his collaborators (1993) were able to use statistical
methods to carry out a similar localizing exercise. When they disaggregated
the large regions of northern and southern Italy into smaller statistical units,
they found support for their theory at the sub-regional level that civil com-
petence is related to policy performance (chap. 4). The replication of large
scale hypotheses at smaller scales through paired comparison added strength
to the findings of both Tilly and Putnam. It localized theory as well as
strengthening it.

Combining Correlational with Process-Tracing Methods

There is an unfortunate tendency in comparative politics to choose a particular
way of doing research because it seems the One Best Way to do research. As
readers will already have concluded from the thrust of this article, this is not my
view. But much more difficult than calling for methodological pluralism is to
solve the problem of how to combine different logics of comparative research.
Consider the association between membership in social networks and the
willingness to participate in collective action (Diani & McAdam, 2003;
McAdam & Paulsen, 1993). It is a correlation that has been observed in so
many different correlational studies that it comes close to achieving the status
of a sociological law. But no matter how high the correlation between social
network membership and the willingness to mobilize, without understanding
the mechanisms that produce it, its outcomes will be indeterminate. For exam-
ple, it may be the result of social control, of physical propinquity, of solidarity,
or of ideology. Only by going inside the process of mobilization to specify its
connective mechanisms can we understand how the chain of causation oper-
ates. Yet advocates of statistical correlation and enthusiasts for process tracing
are often indifferent to one another’s contributions.
Methodological pluralists need not despair. Consider the civil war litera-
ture of the last decade. As is well known, there was a surge of large-N
quantitative studies of civil conflict from the early 1990s to the turn of the
century (Collier & Hoeffler, 1998; Fearon & Laitin, 2003; Gurr, Harff, Mar-
shall, & Scaritt, 1993). Statistical data arrayed from official sources for
literally hundreds of civil war cases demonstrated a number of significant
Tarrow 253

correlations for civil war outbreak (e.g., the presence of extractable resources
and the existence of mountainous territory). But the large-N tradition left
open a number of questions that were not easily susceptible to quantitative
analysis (Tarrow, 2005). The quantitative civil war literature appeared to be
playing itself out.
But just at that moment, Paul Collier and Nicolas Sambanis, who had done
some of the best quantitative work, brought together a number of country and
regional specialists to test some of the main conclusions that had come out
of the earlier wave of studies with a series of case studies, including the
method of process-tracing paired comparisons (Collier & Sambanis, 2005;
Humphreys & ag Mohamed, 2005). Their findings are paving the way for
the next wave of large-N studies of civil wars, including more refined mea-
surement of the escalation of contention from social protest to organized
violence (Sambanis & Zinn, 2005).
Process-based and correlational studies are best seen as complements, not
as competitors, on the long hard road to scientific knowledge (Caporaso, 2009).
So are most-different and most-similar system designs. Only by exploiting to
the full and varied potential of paired comparison will the strategy of paired
comparison, which has been much used but little thought about, take its rightful
place in the armamentarium of comparative politics.

Acknowledgments
This article draws on a talk given at the Princeton University Department of Politics in
November 2007 and on a short article from the APSA/CP (Tarrow, 1999). I am grateful
to Alberto Abadie, Mark Beissinger, Andy Bennett, Valerie Bunce, Jim Caporaso, David
Collier, John Gerring, Gary Goertz, Jai Kwan Jung, Peter Katzenstein. James Mahoney,
Jonas Pontusson, Tom Pepinsky, Charles Ragin, Benoît Rihout, Richard Samuels, and
the late Charles Tilly for comments on an earlier version of this text. Jennifer Hadden
worked hard to keep me from straying too far from the topic and read every iteration,
for which I thank her. Kathleen Jercich cut out as much of the unnecessary verbiage
as she could.

Declaration of Conflicting Interests

The author declared no potential conflicts of interests with respect to the authorship
and/or publication of this article.

Financial Disclosure/Funding
The author received no financial support for the research and/or authorship of this
article.
254 Comparative Political Studies 43(2)

Notes

1. For truth in advertising, I confess that through four single and jointly authored
books, I failed to develop a theory of paired comparison. See Tarrow (1967), Tarrow
(1977), Blackmer and Tarrow (1972), and McAdam et al. (2001).
2. For a large number of references to the use of paired comparison on which this
article is based, visit http://falcon.arts.cornell.edu/Govt/faculty/Tarrow.html.
3. In an earlier book, Gerring gives more attention to paired comparisons in the
context of a discussion of most-similar and most-different case analysis (2001,
pp. 208-214).
4. I use Gerring’s lucid definition of most-different systems analysis rather than
Przeworski and Teune’s more opaque one because it will be understandable to the
ordinary reader (cf. Przeworski & Teune, 1970, pp. 34-39).
5. Although there is a superficial resemblance between Mill’s methods of agree-
ment and difference and Przeworski and Teune’s most-similar and most-different
systems, they are actually somewhat different. Mill was trying to devise methods
to look at similarities and their outcomes regardless of the proximity or distance
between the objects of analysis; Przeworski and Teune were distinguishing the
distance between the objects being compared. For an innovative application of
most-similar systems analysis to European public policy, see De Meur, Bursens,
and Gottcheiner (2006).
6. Skocpol (1979) writes, “The [purposive] image strongly suggests that revolu-
tionary processes and outcomes can be understood in terms of the activity and
intentions or interests of the key group(s) who launched the revolution in the
first place” (p. 17).
7. This analysis is based on Samuels’ book, Machiavelli’s Children (2003), and on an
earlier article, “Tracking Democracies: Italy and Japan in Historical Perspective”
(1998). For his methodological reflections, see Samuels (1999).
8. This argument is close to that of Peter Hall (2003), who has argued for a closer fit
between the ontology and the methodology of paired comparison. I am grateful to
an anonymous reviewer for pointing out this convergence between Hall’s position
and my own.
9. I am grateful to Gary Goertz for pointing this out in his comments on an earlier
version of this article.

References
Abadie, A., Diamond A., & Hainmueller, J. (2007). Synthetic control methods for
comparative case studies: Estimating the effect of California’s tobacco control
program. NBER Working Paper 12831. Cambridge, MA: National Bureau of
Economic Research.
Tarrow 255

Abadie, A., & Javier Gardeazabal. (2003). The economic costs of conflict: A case
study of the Basque country. American Economic Review, 93, 113-132.
Becker, H. (1968). Social observation and social case studies. In International Ency-
clopedia of the Social Sciences (2nd ed., pp. 232-238). New York: Macmillan.
Bennett, A. (2008). Process tracing: A Bayesian perspective. In J. M. Box-Steffensmeier,
H. E. Brady, & D. Collier (Eds.), The Oxford handbook of political methodology
(pp. 702-721). New York: Oxford University Press.
Blackmer, D. L.M., & Tarrow, S. (Eds.). (1975). Communism in Italy and France.
Princeton NJ: Princeton University Press.
Brady, H. E., & Collier, D. (Eds.). (2004). Rethinking social inquiry: Diverse tools,
shared standards. Lanham, MD: Rowman & Littlefield.
Brady, H. E., Collier D, & Seawright, J. (2006). Towards a pluralistic vision of
methodology. Political Analysis, 14, 353-368.
Bunce, V. (1981). Do new leaders make a difference? Executive succession and public
policy under capitalism and socialism. Princeton, NJ: Princeton University Press.
Bunce, V. (1999). Subversive institutions: The design and the destruction of socialism
and the state. Cambridge, UK: Cambridge University Press.
Bunce, V. J., & Wolchik, S. L. (2006a). Favorable Conditions and Electoral Revolutions.
Journal of Democracy, 17, 5-18.
Bunce, V. J., & Wolchik, S. L. (2006b). International diffusion and postcommunist
electoral revolutions. Communist and Post-Communist Studies, 39, 283-304.
Bunge, M. (1997). Mechanism and explanation. Philosophy of the Social Sciences,
27, 410-465.
Caporaso, J. (2009). Is there a quantitative-qualitative divide in comparative politics?
The case of process tracing. In T. Landman & N. Robinson (Eds.), Sage handbook
of comparative politics (pp. 67-83). Thousand Oaks, CA: Sage.
Cleary, M. R., & Stokes, S. C. (2006). Democracy and the culture of skepticism:
Political trust in Argentina and Mexico. New York: Russell Sage.
Collier, D., & Collier, R. (1991). Shaping the political arena: Critical junctures, the
labor movement, and regime dynamics in Latin America. Princeton, NJ: Princeton
University Press.
Collier, P., & Hoeffler, A. (1998). On the economic causes of civil war. Oxford
Economic Papers, 50, 563-573.
Collier, D., & Levitsky, S. (1997). Democracy with adjectives: Conceptual innovation
in comparative research. World Politics, 49, 430-451.
Collier, P., & Sambanis, N. (Eds.). (2005). Understanding civil war (2 vols.).
Washington, DC: The World Bank.
Converse, P., & Dupeux, G. (1962). Politicization of the electorate in France and the
United States. Public Opinion Quarterly, 26, 1-23.
256 Comparative Political Studies 43(2)

De Meur, G., Bursens, P., & Gottcheiner, A. (2006). MSDO/MDSO revisited for
public policy analysis. In B. Rihoux & H. Grimm (Eds.), Innovative comparative
methods for policy analysis (pp. 67-94). New York: Springer.
Diani, M., & McAdam, D. (Eds.). (2003). Social movements and networks: Rela-
tional approaches to collective action. Oxford, UK, and New York: Oxford
University Press.
Dogan, M. (1967). Political cleavage and social stratification in France and Italy. In S.
M. Lipset & S. Rokkan (Eds.), Party systems and voter alignments: Cross-national
perspectives (pp. 129-195). New York: Free Press.
Eckstein, H. (1975). Case study and theory in political science. In F. I. Greenstein &
N. W. Polsby (Eds.), Handbook of political science: Vol. 7. Strategies of inquiry
(pp. 79-138). Reading MA: Addison-Wesley.
Fearon, J., & Laitin, D. (2003). Ethnicity, insurgency and civil war. American Politi-
cal Science Review, 97, 75-90.
Foran, J. (1994). The Iranian revolution of 1977-99: A challenge to social theory. In
J. Foran (Ed.), A century of revolution: Social movements in Iran (pp. 1-32). Min-
neapolis: University of Minnesota Press.
Geddes, B. (1990). How the cases you choose affect the answers you get. Political
Analysis, 2, 131-150.
George, A. L., & Bennett, A. (2005). Case studies and theory development in the
social sciences. Cambridge, MA, and London: MIT Press.
Gerring, J. (2001). Social science methodology: A critical framework. Cambridge,
UK, and New York: Cambridge University Press.
Gerring, J. (2007). Case study research: Principles and practices. New York and
Cambridge, UK: Cambridge University Press.
Gibson, J. L. (2008). Group identities and theories of justice: An experimental
investigation into the justice and injustice of land squatting in south Africa.
Journal of Politics, 70, 700-716.
Goldstone, J. (1998). Social movements or revolutions: Evolution and outcomes of
collective action. In M. Giugni, D. McAdam, & C. Tilly (Eds.), From contention
to democracy (pp. 125-145). Lanham, MD: Rowman & Littlefield.
Gurr, T. R., with Harff, B., Marshall, M., & Scaritt, J. M. (1993). Minorities at risk.
Washington, DC: United States Institute of Peace.
Hall, P. A. (2003). Aligning ontology and methodology in comparative research. In
J. Mahoney & D. Rueschemeyer (Eds.), Comparative historical analysis: New
approaches and methods (pp. 373-406). New York and Cambridge, UK: Cambridge
University Press.
Hall, P., & Tarrow, S. (1998, January 23). Globalization and area studies: When is too
broad too narrow? Chronicle of Higher Education, B4-B5.
Tarrow 257

Humphreys, M., & ag Mohamed, H. (2005). Senegal and Mali. In P. Collier & N.
Sambanis (Eds.), Understanding civil war: Vol. 1. Africa (pp. 247-302). Washington,
DC: The World Bank.
Inglehart, R. (1977). The silent revolution: Changing values and political styles
among Western publics. Princeton, NJ: Princeton University Press.
Inglehart, R., Basánez, M., Díez-Medrano, J., Halmann, L, & Luijkx, R. (Eds.).
(2004). Human beliefs and values: A cross-cultural sourcebook based on the
1999-2002 values surveys. Coyoacan: Siglo Veintiuno Editores.
Johnson, C. A. (1962). Peasant nationalism and communist power. Stanford, CA:
Stanford University Press.
Jung, J. K. (2008). The paradox of institution building after civil war: A trade-off
between short-term peacemaking and long-term democracy building. Unpublished
doctoral thesis, Cornell University, Ithaca, New York.
Katzenstein, P. J. (1984). Corporatism and change: Austria, Switzerland and the
politics of industry. Ithaca, NY, and London: Cornell University Press.
King, G., Keohane, R. O., & Verba, S. (1994). Designing social inquiry: Scientific
inference in qualitative research. Princeton, NJ: Princeton University Press.
Lehmbruch, G., Schmitter, P. (Eds.). (1982). Patterns of corporatist policy making.
Beverly Hills, CA, and London: Sage.
Lieberman, E. (2005). Nested analysis as a mixed-method strategy for comparative
research. American Political Science Review, 99, 435-452.
Lijphart, A. (1975). The comparable-cases strategy in comparative research. Com-
parative Political Studies 8, 169-171.
Little, D. E. (1998). Microfoundations, methods, and causation. New Brunswick, NJ:
Transaction Press.
Linz, J. (1975). Totalitarian and authoritarian regimes. In F. I. Greenstein & N. W.
Polsby (Eds.), Handbook of political science: Vol. 3. Macropolitical theory (pp.
175-357). Reading, MA: Addison-Wesley.
Linz, J., & Valenzuela, A. (Eds.). (1994). The failure of presidential democracy.
Baltimore and London: Johns Hopkins University Press.
McAdam, D., & Paulson, R. (1993). Specifying the relationship between social ties
and activism. American Journal of Sociology, 99, 640-667.
McAdam, D., Tarrow, S., & Tilly, D. (2001). Dynamics of contention. New York and
Cambridge, UK: Cambridge University Press.
Mill, J. S. (1973). A system of logic ratiocinative and inductive In J. M. Robson, Ed.,
Collected works of John Stuart Mill (Vols. 7-8).Toronto: University of Toronto
Press.
Pempel, T. J. (Ed.). (1990). Uncommon democracies: The one-party dominant regime.
Ithaca, NY, and New York: Cornell University Press.
Pempel, T. J. (1998). Regime shift: Comparative dynamics of the Japanese political
economy. Ithaca, NY, and London: Cornell University Press.
258 Comparative Political Studies 43(2)

Pizzorno, A. (1971). Amoral familism and historical marginality. In M. Dogan &

R. Rose (Eds.), European politics: A reader (pp. 87-98). Boston: Little Brown.
Przeworski, A., & Teune, H. (1970). The logic of comparative social inquiry. New
York: Wiley.
Putnam, R. D, Leonardi, R., & Nanetti, R. (1993). Making democracy work: Civil
traditions in modern Italy. Princeton, NJ: Princeton University Press.
Ragin, C. C. (1987). The comparative method: Moving beyond qualitative and
quantitative strategies. Berkeley: University of California Press.
Ragin, C., & Sonnett, J. (2005). Between complexity and parsimony: Limited diver-
sity, counterfactual cases, and comparative analysis. In S. Kropp & M. Minkenberg
(Eds.), Vergleichen in der Politikwissenschaft (pp. 180-197). Wiesbaden, Germany:
VS Verlag fur Sozialwissenschaften.
Rosenberg, R., & Silber, J. H. (2001). Matching and thick description in an observa-
tional study of mortality after surgery. Biostatistics, 2, 217-232.
Sambanis, N., & Zinn, A. (2005). From protest to violence: An analysis of conflict
escalation with an application to self-determination movements. Yale University
Department of Political Science working paper.
Samuels, R. J. (1998). Tracking democracies: Italy and Japan in comparative perspec-
tive. Journal of Modern Italian Studies, 2, 283-320.
Samuels, R. J. (1999). Nudging serendipity: The 3-C problem in comparative politics.
APSA/CP, Newsletter of the Comparative Politics Section of the APSA, 10, 13-16.
Samuels, R. J. (2003). Machiavelli’s children: Leaders and their legacies in Italy and
Japan. Ithaca, NY, and London: Cornell University Press.
Skocpol, T. (1979). States and social revolutions. Cambridge, UK: Cambridge Univer-
sity Press.
Tarrow, S. (1967). Peasant communism in southern Italy. New Haven, CT, and London:
Yale University Press.
Tarrow, S. (1977). Between center and periphery: Grassroots elites in Italy and
France. New Haven, CT, and London: Yale University Press.
Tarrow, S. (1999, Summer). Expanding paired comparison: A modest proposal. Ameri-
can Political Science Association—Comparative Politics Newsletter, 10, 9-12.
Tarrow, S. (2004). Bridging the quantitative-qualitative divide. In H. E. Brady &
D. Collier (Eds.), Rethinking social inquiry: Diverse tools, shared standards
(pp. 171-179). Lanham, MD: Rowman & Littlefield.
Tarrow, S. (2005). Inside insurgencies: Politics and violence in an age of civil war.
Perspectives on Politics, 5, 587-60.
Thelen, K. (2004). How institutions evolve. New York and Cambridge, UK: Cambridge
University Press.
Tilly, C. (1964). The Vendée. Cambridge, MA: Harvard University Press.
Tilly, C. (1984). Big structures, large processes, huge comparisons. New York:
Russell Sage.
Tarrow 259

Tilly, C. (1990). Capital, coercion and European states, AD 990-1992. Oxford,

UK: Blackwell.
Tilly, C. (1993). European revolutions, 1492-1992. Oxford, UK: Blackwell.
Tilly, C. (2000): Mechanisms in political processes. Annual Review of Political
Science, 4, 21-41.
Tilly, C. (2002). Stories, identities and political change. Lanham, MD: Rowman
& Littlefield.
Tilly, C. (2004). Contention and democracy in Europe, 1650-2000. New York and
Cambridge, UK: Cambridge University Press.
Tilly, C. (2008). Explaining social processes. Boulder: Paradigm.
Tocqueville, A., de. (1954). Democracy in America (2 vols.) (H. Reve, Trans.; P. Brady,
Ed). New York: Vintage.
Tocqueville, A., de. (1955). The old regime and the French revolution (S. Gilbert,
Trans.). Garden City, NJ: Doubleday.
Weinstein, J. (2006). Inside rebellion: The political economy of rebel organization. Cam-
bridge, UK: Cambridge University Press.
Wood, E. J. (2000). Forging democracy from below: Insurgent transitions in South
Africa and El Salvador. New York and Cambridge, UK: Cambridge University Press.
Wood, E. J. (2002). Insurgent collective action and civil war in El Salvador. New
York and Cambridge, UK: Cambridge University Press.

Bio
Sidney Tarrow (PhD, Berkeley, 1965) is Maxwell M. Upson Professor of Government
and Professor of Sociology at Cornell University. Tarrow’s first book was Peasant
Communism in Southern Italy (Yale, 1967). In the 1980s, after a foray into comparative
local politics (Between Center and Periphery, Yale, 1978), he turned to a reconstruction
of Italian protest cycle of the late 1960s and early 1970s, Democracy and Disorder
(Oxford, 1989). His recent books are Power in Movement (Cambridge, 1994, 1998),
Dynamics of Contention (with Doug McAdam and Charles Tilly, Cambridge, 2001),
Contentious Europeans (with Doug Imig, Rowman and Littlefield, 2001), Transnational
Protest and Global Activism (ed., with Donatella della Porta, Rowman and Littlefield,
2004), The New Transnational Activism (Cambridge, 2005), and (with the late Charles
Tilly) Contentious Politics (Paradigm, 2006). He is a fellow of the American Academy
of Arts and Sciences and has served as program co-chair of the American Political Sci-
ence Association Annual Convention, president of the Conference Group on Italian
Politics, and president of the APSA Section on Comparative Politics. He is currently
working on war-making, state building, and human rights.