Debating the Direction of Comparative Politics Comparative Political Studies Volume 40 Number 1 January 2007 5-31 © 2007 Sage Publications 10.1177/0010414006294815 http://cps.sagepub.com hosted at http://online.sagepub.com An Analysis of Leading Journals Gerardo L. Munck University of Southern California, Los Angeles Richard Snyder Brown University, Providence, Rhode Island This article contributes to ongoing debates about the direction of comparative politics through an analysis of new data on the scope, objectives, and methods of research in the field. The results of the analysis are as follows. Comparative politics is a rich and diverse field that cannot be accurately characterized on the basis of just one dimension or even summarized in simple terms. In turn, the tendency to frame choices about the direction of the field in terms of a stark alternative between an old area studies approach and a new economic approach relies on largely unsupported assumptions. It is therefore advisable to focus on problematic methodological practices that, as this study shows, are widespread in comparative research and thus pose serious impediments to the production of knowledge. Keywords: comparative politics; paradigm; scope; methods; area studies; rational choice Since the late 1980s, scholars of comparative politics based in the United States have debated many fundamental issues concerning three broad dimensions of the research process: the scope and objectives of research, the methods of theory generation, and the methods of empirical analysis. In these discussions, conventional research practices and proposed alternatives have Authors' Note: We thank David Laitin for suggesting that we write this article and for organizing a panel with the editors of Comparative Political Studies, Comparative Politics, and World Politics at the 2005 APSA Annual Convention in Washington, DC. We are grateful to Angela Hawken for her advice on the construction of the data set and the analysis of the data and to Matthew Lieber for his assistance with data collection. In addition, we appreciate comments from Yitzhak Brudny, James Caporaso, John Londregan, Sebastian Mazzuca, David Samuels, Saika Uno, and three anonymous reviewers. 5 6 Comparative Political Studies been challenged. Indeed, it has become readily apparent during the past 15 years that comparativists hold widely divergent views about the nature of the problems facing their field and the solutions required to realize more fully the potential of comparative politics to produce knowledge about politics around the globe. With regard to the scope and objectives of comparative politics, though some authors claim that comparativists are addressing big, substantive questions about politics (Pierson & Skocpol, 2002, pp. 695-698), others express doubts about this. For example, Robert Dahl states that "very few people study power today" and that "perhaps we've not only failed to progress in the study of power, we've actually gone into reverse" (Munck & Snyder, 2004, pp. 26-27). Juan Linz and Samuel Huntington highlight the lack of attention to political leadership and political elites (Munck & Snyder, 2004, pp. 28-29). Others raise questions about the policy relevance of comparative politics research. Thus, Giovanni Sartori (2004) argues that comparative politics places great emphasis on the theory-research nexus yet largely disregards the theory-praxis nexus or, as Adam Przeworski puts it, in less complex language, "we do not speak about politics to people outside academia" (Munck & Snyder, 2004, p. 31; see also Laitin, 2004, p. 16). The methods of theory generation in comparative politics have also been a contentious topic. One hotly debated issue concerns the role of paradigms. Despite widespread agreement that the field currently lacks a single dominant paradigm, scholars offer sharply divergent proposals for how the field should respond to this situation. Some see it as a chance to move forward by placing bets on a new, economics-inspired rational choice paradigm (Geddes, 2003, chap. 5; Levi, 1997; Weingast, 2002), whereas others call for pluralistic competition that avoids the hegemony of any single paradigm, such as rational choice theory (Lichbach, 1997, 2003; see also Almond, 1990, chap. 4). And others perceive an opportunity to avoid paradigms altogether—and allegedly sterile paradigm disputes—and shift instead to midrange theorizing (Portes, 2005, pp. 34-38; see also Huntington in Munck & Snyder, 2004, p. 29; Laitin, 2004, pp. 35-36; Pierson & Skocpol, 2002, p. 698). Scholars have also counterposed, sometimes quite starkly, deductive and inductive methods of theorizing. Some defend the traditionally inductive methods used by comparativists (Brady, Collier, & Seawright, 2004, p. 13; Collier, 1999, pp. 4-5), whereas others advocate greater reliance on deductive and also formal methods of theorizing as part of a broader proposal to incorporate tools and theories from economics into comparative politics (Keech, Bates, & Lange, 1991; Levi, 2000). The methods of empirical analysis in comparative politics have also been subject to great scrutiny and divergent assessments. Much of the discussion Munck, Snyder / Direction of Comparative Politics 7 focuses on the methodological underpinnings of the qualitative research that is very common in comparative politics. This debate was initially launched by quite critical assessments of the qualitative tradition that argued that the area studies literature—seen as largely qualitative—lacked methodological sophistication (Bates, 1996), unwittingly introduced biases because of flawed case selection (Geddes, 1991; 2003, chap. 3), and was poorly equipped to test hypotheses because of the limitations of qualitative methods (King, Keohane, & Verba, 1994). Recently, however, the basis for a more pluralistic view of methods has been emerging, as scholars emphasize that qualitative and quantitative researchers alike face serious hurdles in making descriptive and causal inferences (Brady et al., 2004, pp. 10-11), draw attention to how qualitative research can potentially contribute to hypothesis testing (Brady & Collier, 2004), and propose a mixed-method approach that combines quantitative and qualitative components (Laitin, 2002, pp. 630-631; 2003). Taken together, these controversies constitute a debate about the direction of comparative politics. This debate is closely linked with the broader disciplinary debate, raised largely by the Perestroika insurgency, about the future of political science (Mansfield & Sisson, 2004a, 2004b; Monroe, 2005; Shapiro, Smith, & Masoud, 2004). What is at stake is how we think about the goals and means of a core field of political science. Knowledge about what? What kind of knowledge? Knowledge for what? How do we generate knowledge? Thus, this is a debate that should be addressed responsibly, on the basis of explicit criteria and arguments, systematic evidence about actual research practices, and clear thinking about how best to proceed from where we stand today. This article seeks to contribute to the debate about the direction of comparative politics through an analysis of articles in the three leading U.S. journals dedicated fully or largely to comparative politics—Comparative Political Studies, Comparative Politics, and World Politics—during the 1989 to 2004 period.1 This is, admittedly, a limited sample of the universe of work on comparative politics. It covers neither the main discipline-wide journals in political science nor the area-specific journals. It does not include books. Moreover, it does not encompass publications in other disciplines and countries. The comparative politics research published in these other outlets could diverge systematically from the kind of material published in the three journals on which this article focuses. Still, these three journals are widely seen as the leading ones in the field, where much of the best research in comparative politics appears and where collective standards get set. Analyzing articles in these journals is thus a fruitful strategy for providing a stronger empirical foundation for debating the direction of comparative politics.2 The first section presents an overview of the scope, objectives, and methods of research in comparative politics. The next section challenges widely 8 Comparative Political Studies accepted depictions of the old area studies and the new economics-inspired approaches. The third section refocuses the debate by drawing attention to key methodological problems. Finally, the conclusions are summarized. Scope, Objectives, and Methods: An Overview This section maps the contours of the field of comparative politics by considering how scholars handle three broad aspects of the research process. It first addresses the scope of research, touching on the issues of subject matter and empirical range. Next, it considers the objectives of research, exploring the kinds of knowledge produced by comparativists. Finally, it focuses on the methods employed in comparative research, distinguishing between the methods of theory generation and the methods of empirical analysis. Scope: Knowledge About What? The scope of comparative politics—derived from a widely accepted definition of the field as the study of politics and political power around the world—entails a substantive and empirical dimension. The substantive dimension pertains to the subject matter and research questions addressed by comparativists, whereas the empirical dimension pertains to the spatial and temporal range of comparative analysis. Students of comparative politics have focused on a diverse array of subject matters and research questions centrally related to politics (see Table 1). The largest number of articles focuses on democratic and state institutions, addressing aspects such as their nature and the process of decision making within them. Living as we do in a democratic age, the importance of studying such questions is obvious and central to the mission of comparative politics.3 Comparative research has also addressed many other key issues. A substantial amount of research focuses on political order and regimes, addressing questions about states, wars, revolutions, ethnicity, regimes, and democratization and thus exploring the processes that render the institutionalized decision making of democracies possible, or impossible, in the first place. Other research focuses on actors, such as social movements, interest groups, and citizens, going beyond the voters and politicians that are essential to democratic politics; on substantive outcomes, such as economic development, the welfare state, the developmental state, capitalism and neoliberalism, that are linked closely with the well-being of most of the world's population; and on processes, such as globalization and supranational integration, that have a Munck, Snyder / Direction of Comparative Politics 9 Table 1 The Substantive Scope of Comparative Research Subject Matter Overarching % Articles Subject Matter % Articles State formation and state collapse War Revolutions Nationalism Civil wars and violence Ethnicity and ethnic conflict Varieties of political regimes Democratization and democratic breakdowns Political order 17.9 Political regimes 26.6 Social movements and civil society (includes 7.2 social capital, strikes, and protests) Interest groups (includes business 11.0 and labor studies) Citizen attitudes and political culture 11.0 Religion 1.9 Clientelism 1.6 >■ Social actors 32.6 Elections, voting, and electoral rules Political parties Democratic institutions (executive and legislative branches of government) Federalism and decentralization Judiciary Bureaucracy Military and police Policy making in general 10.3 12.9 3.4 1.9 5.6 2.5 8.2 J > Democratic and state institutions 51.4 Economic policy and reform (includes the welfare state, the developmental state, neoliberalism, and varieties of capitalism) Economic development Globalization Supranational integration and processes 27.0 4.1 4.7 5.6 Economic and extranational processes 41.4 Other 0.9 Total 170.8 Note: N = 319. The total percentage for the % Articles column exceeds 100% because individual articles frequently address multiple subjects. The data are drawn from the variable question of the Munck-Snyder Comparative Politics Articles Data Set. 10 Comparative Political Studies growing impact on politics and even challenge the assumption that the nation-state is the dominant site of political power. It is hard to judge whether this overview fully and adequately reflects the substantive scope of comparative politics. The doubts noted at the outset that senior scholars voice about the field's efforts to study power and leadership cannot be dismissed easily. Moreover, even research that ostensibly addresses big, substantive questions can do so in a superficial way, and there are certain topics that seem to hold great importance in the world of politics—for example, religion and clientelism—that receive very little attention. Still, the questions that are addressed deal with key actors and institutions involved in making political decisions and are quite obviously connected to the issue of political power. Turning to the empirical dimension of the scope of comparative politics, it is easier to assess how much the pretense of comparative politics to study politics around the world is realized. The data show that Western Europe is studied in 41.0% of articles, Latin America in 27.2%, East Asia in 20.3%, North America (Canada and the United States) in 17.0%, sub-Saharan Africa in 12.4%, the Soviet Union or post-Soviet republics in 11.8%, the Middle East and North Africa in 11.5%, Eastern Europe in 10.8%, Oceania in 8.2%, Southeast Asia in 6.9%, South Asia in 5.9%, and the Caribbean in 5.5%.4 Comparativists thus do a good job providing broad coverage of the world's regions and have also made important strides to incorporate the study of the United States as part of comparative politics. Still, the unevenness of regional coverage—the heavy coverage of Western Europe in contrast to the strikingly few articles on populous regions such as Southeast Asia and South Asia— bears emphasis. With regard to the temporal range of research, comparativists have employed varying time spans, as would be expected in light of the diverse questions they address—some of which focus on contemporary citizen attitudes and electoral processes, others that involve the historical origins of states, regimes, and institutions. Still, a considerable amount of the articles published in the leading journals adopt a long-term perspective, and a majority (52.4%) analyze a time span of more than 20 years.5 Objectives: What Kind of Knowledge? Knowledge for What? The field of comparative politics is heavily oriented toward empirics, with very few studies focusing exclusively on generating theory, that is, propositions about how or why the world is as it is (see Table 2). Indeed, virtually all articles (95.6%) are empirically oriented, and only 4.4% aim to contribute solely to theory. This does not mean that theory generation has been neglected. Munck, Snyder / Direction of Comparative Politics 11 Table 2 The Objectives of Comparative Research Aggregate Objectives Options % Articles Options % Articles Theory and Theory generation empirics Theory generation and empirical analysis Empirical analysis Total Description Descriptive and causation Descriptive and causal, but primarily descriptive Descriptive and causal, but primarily causal Causal Total Note: N = 319. A theory is understood to consist of a proposition or set of propositions about how or why the world is as it is. An empirical analysis is understood to consist of an inquiry based on observable manifestations of a concept or concepts. Thus, empirical analysis is not restricted to causal hypothesis testing. In turn, the term descriptive is not used, as is common, in a critical fashion, as when a work is characterized as being merely descriptive. Here, the term is used in a positive manner, as referring to accounts about what the state of the world is, that are differentiated from causal accounts that seek to explain why the state of the world is as it is. The data are drawn from the variables theory_empirical and descriptive_causal of the Munck-Snyder Comparative Politics Articles Data Set. 49.8 100.0 16 35 4.4 H-^ Theory 50.2 35.4 12.5 100.0 Empirical 95.6 analysis Mainly 52.0 descriptive Mainly causal 48.0 After all, more than half the articles (50.2%) seek to generate theory. But it does mean that theory generation in comparative research is almost always tightly linked with empirical analysis. Though David Laitin's (1993) warning in 1993 that "the specter of pure theory ... is haunting comparative politics" (p. 3) may have reflected the feelings of many comparativists at the time, the research produced during the past 15 years does not validate this assessment.6 In contrast to their overwhelming preference for empirical analysis over pure theory, comparativists have responded in a more balanced manner to the option of producing descriptions, that is, accounts about what the state of the world is, or causal analysis, that is, accounts that seek to explain why the world is as it is. Indeed, researchers have produced articles in roughly an equal proportion that aim at advancing these two fundamental and related goals of social science, more often than not combining the two objectives (see Table 2). Given the demanding nature of the enterprise of empirically supporting causal arguments (a point discussed below), the decision to place a strong emphasis 12 Comparative Political Studies on descriptive knowledge is sound. It shows that comparativists recognize the value of description, a point made by methodologists (King et al., 1994, pp. 7-8, chaps. 2 and 3). And it means that efforts at causal theorizing can draw on a wealth of descriptive research that establishes variation in real-world politics worth addressing in causal theories. Still, despite the advocacy by some scholars of an engagement with real-world political issues (Skocpol, 2003), there is little evidence that comparativists aim to produce knowledge of direct relevance to policy decisions.7 Methods: How Do We Produce Knowledge? Despite the vigorous advocacy and equally forceful critique of deductive, formal methods, the data show that very little work in comparative politics actually uses these methods (see Table 3).8 In comparative politics, deductive theorizing is of the soft, semiformal, or informal varieties. Moreover, the overwhelming majority of theorizing in comparative politics is inductive, which highlights again that comparativists tend to work close to the empirical level. In turn, the most common method of empirical analysis is qualitative, understood here simply as research that relies on words as opposed to numbers.9 Still, a sizable proportion of research is quantitative, and a considerable amount combines qualitative and quantitative methods of empirical analysis. In sum, this overview shows that comparative politics is a diverse field that (a) addresses substantial and pressing issues in all regions of the world, (b) is heavily oriented toward empirics, (c) aims to produce descriptive and causal knowledge in a roughly equal balance, (d) lacks much concern with policy, (e) generates theory mainly through inductive methods, and (f) relies largely on qualitative methods of empirical analysis. The data cast doubt on simple characterizations of the scope, objectives, and methods of comparative research that gloss over the diversity of options routinely employed by comparativists. Indeed, the data show that any facile characterization is likely to be just a partial description or, more problematically, an inaccurate caricature of what is really going on in comparative politics. Old and New Approaches: Demystifying Common Assumptions Our analysis of articles in the leading journals sheds light on the two most prominent alternatives currently facing comparativists: the old area studies Munck, Snyder / Direction of Comparative Politics 13 Table 3 The Methods of Comparative Research Aim of Method Options % Articles Aggregate Options % Articles Methods of Inductive, qualitative 80.0 theorizing Inductive, quantitative 22.5 Deductive, semiformal 32.5 or informal Deductive, formal 4.4_ Total 139.4 Methods of Qualitative 44.3 empirical Mixed method, 19.0 analysis dominantly qualitative Mixed method, 13.1 dominantly quantitative Quantitative 23.6 Total 100.0 Inductive • Deductive J-1 Mainly qualitative Mainly quantitative 102.5 36.9 63.3 36.7 Note: The N for the data on methods of theorizing is 160, the N for the data on methods of empirical analysis is 305. The total percentage exceeds 100% because individual articles frequently use multiple methods of theorizing. The data are drawn from the variables method_theory and method_analysis of the Munck-Snyder Comparative Politics Articles Data Set. approach, understood here simply as research focused on one geographical region of the world, and the new economics-inspired approach, defined as research that uses formal and quantitative methods. The data show that much of the recent debate about these approaches is based on false or misleading assumptions that poorly correspond to actual research practices. The Question of Area Studies Area studies, which emerged out of the formal-legal tradition of research prior to World War II, was critiqued originally in the 1950s by pioneers of the behavioral revolution in comparative politics (Almond & Coleman, 1960, pp. vii; Macridis & Cox, 1953, pp. 653-655).10 Subsequently, during the 1960s and 1970s, area studies research underwent a great expansion, partly in reaction to the perception that the structural-functional approach—one of the key fruits of the behavioral revolution in comparative politics—was overly abstract 14 Comparative Political Studies and used categories that were not grounded in the realities of third world societies (Schmitter, 1993, pp. 172-173). Despite a second round of challenges from different quarters that criticized area studies for being atheoretical and lacking methodological sophistication (Bates, 1996; Skocpol, 2003, pp. 412-413; see, however, Laitin, 1998, p. 434), most research in comparative politics during the past 15 years still focuses exclusively on one geographical region.11 Indeed, single-region studies are by far the most common type of article, accounting for 70.2% of all articles.12 The most notable difference between area studies and non-area studies research is that area studies articles are far more likely to aim solely or mainly at producing descriptions—understood, as noted above, as an important goal of the social sciences—instead of causal accounts (see Table 4). Other differences, though not as large, between area studies and non-area studies research concern the weaker orientation of area studies research toward theory generation and its more frequent use of qualitative methods of empirical analysis.13 However, these differences do not mean that area studies research constitutes a distinctive approach. The data show that area studies research has a range of objectives that are fundamental to the social sciences and employs a variety of established methods. Moreover, the methods of theorizing employed in area studies research are statistically indistinguishable from those employed in non-area studies research. Indeed, like the field of comparative politics as a whole, area studies research does not constitute a homogenous approach and cannot be neatly set apart from other bodies of work. Although blanket characterizations of area studies research as atheoretical and "merely descriptive" may have been true in the past, they provide misleading caricatures of the "new area studies" of recent years. An Economic Turn in Comparative Politics? Since the early 1990s, there has been much heated discussion of an "economic turn" in comparative politics and, more pointedly, of the promise of research that combines formal and deductive methods of theorizing with quantitative methods of empirical analysis.14 This proposal was seen by many as a hegemonic project of rational choice theorists that threatened the diversity of the field (Lustick, 1994), and various shortcomings—including the failure to deal with big, substantive questions—of what was viewed as a "methods-driven" agenda were criticized (Green & Shapiro, 1994).15 It is instructive to consider how these expectations and critiques match up against the literature published during the 1989 to 2004 period. Munck, Snyder / Direction of Comparative Politics 15 Table 4 Objectives and Methods in Area Studies Research Ratio of Options per Objectives and Individual Area to Non- Methods Options Area Studies Article Theory and empirics Theory generation and empirical analysis 0.77* Empirical analysis 1.31* Description and Descriptive 1.84** causation Descriptive and causal, but primarily descriptive 1.41** Descriptive and causal, 0.80** but primarily causal Causal 0.43** Methods of Inductive, qualitative 1.09 theorizing Inductive, quantitative 0.76 Deductive, semiformal 0.93 or informal Deductive, formal 2.23 Methods of empirical Qualitative 1.21* analysis Mixed method, dominantly qualitative 1.47* Mixed method, 0.79* dominantly quantitative Quantitative 0.63* Note: The N for the data on theory and empirics, description and causation, and methods of empirical analysis is 305; the N for the data on methods of theorizing is 146. The data are drawn from the variables region, theory_empirical, descriptive_causal, method_theory, and method_analysis of the Munck-Snyder Comparative Politics Articles Data Set. Chi-square tests were performed on the data on theory and empirics and on methods of theorizing (in a value-by-value fashion). Wilcoxon-Mann-Whitney tests were performed on the data on description and causation and on methods of empirical analysis. *Significant at 5% level. **Significant at 1% level. The most important finding is that little research in comparative politics actually uses deductive, formal methods: Only 4.4% of all articles do. Though far from dominant, quantitative methods are more common, accounting for 23.6% of articles that include empirical analysis (see Table 3). Moreover, there are practically no signs of an alliance among deductive, 16 Comparative Political Studies formal, and quantitative methods. Among articles that aim both at theory generation and empirical testing, only 4.8% combine deductive, formal methods of theorizing and quantitative methods of empirical analysis.16 In sum, there is scant evidence that comparative politics has taken an economic turn.17 The data also do not support the criticism that the choice of method limits the kind of substantive questions a researcher can pursue (see Table 5). Concerning the methods of theorizing, the data show that a wide range of methods is used to address diverse questions. Importantly, the only statistically significant differences are that inductive methods that rely on quantitative data are used more frequently in the study of political order—generally understood as one of the broadest questions in comparative politics—and that deductive methods of a nonmathematical nature are used more often in the study of democratic and state institutions—generally seen as involving narrower questions. Still, there is no evidence that the use of deductive, formal methods steers researchers away from broad questions. Concerning the methods of empirical analysis, the data show that the study of political regimes—a large literature addressing issues such as democratization—relies overwhelmingly on qualitative research and rarely uses mixed methods, whereas the study of social actors is more likely to rely on quantitative methods, whether used alone or in conjunction with qualitative methods, a result that largely reflects the use of survey research to study citizen attitudes and political culture. Still, there is no evidence that the methods researchers choose impose constraints on the questions they address or that the substantive agendas of comparativists have been driven by their methodological predilections. The ability of researchers to address big, substantive questions about politics does not appear to depend on their methodological choices. Though strong criticisms of old and new approaches have been voiced, the data show that some of the central assumptions underpinning these criticisms are not valid. Area studies research does not constitute a distinctive approach in terms of research objectives and methods, and critiques of this research as atheoretical and "merely descriptive" are unfounded. Moreover, there is no evidence of an economic turn, and deductive, formal methods of theorizing and quantitative methods of empirical analysis are routinely used to address big questions about politics, thus invalidating the charge that these methods cause a narrowing of the substantive compass of comparative research. As the next section argues, the debate about the direction of comparative politics will be more productive if it focuses not on a confrontation among broadly conceived approaches that hold false assumptions about their rivals but on a set of key methodological challenges that pose serious impediments to the production of knowledge. Table 5 The Subject Matter and Methods of Comparative Research Methods of Theorizing Methods of Empirical Analysis (% Articles on a Given Subject Matter) (% Articles on a Given Subject Matter) Mixed Mixed Overarching Subject Matter Inductive, Qualitative Inductive, Quantitative Deductive, Semiformal or Informal Deductive, Formal Total Qualitative Method, Dominantly Qualitative Method, Dominantly Quantitative Quantitative Total Political order 54.6 21.0* 18.4 6.0 100 52.9 7.5 16.6 22.9 100 Political regimes 59.8 14.3 23.8 2.1 100 66.7* 8.4* 5.78* 19.1* 100 Social actors 60.2 20.5 17.2 2.0 100 35.7** 15.5** 25.4** 23.4** 100 Democratic and 53.0 9.6 33.8** 3.5 100 53.0 13.7 12.4 20.8 100 state institutions Economic and 57.5 13.8 25.1 3.6 100 35.2 30.1 12.7 22.0 100 extranational processes Note: The N for the data on methods of theorizing is 160; the N for the data on methods of empirical analysis is 305. The specific subjects included under each overarching category of subject matter are presented in Table 1. The figures are averages of the percentages, based on the total number of studies on a given subject matter, for all the subject matters included in each overarching category. The data are drawn from the variables question, method_theory, and method_analysis of the Munck-Snyder Comparative Politics Articles Data Set. Chi-square tests were performed on the data on methods of theorizing by comparing articles on one subject matter to all the other articles, in a value-by-value fashion. Wilcoxon-Mann-Whitney tests were performed on the data on methods of empirical analysis by comparing articles on one subject matter to all the other articles. *Significant at 5% level. **Significant at 1% level. 18 Comparative Political Studies Current Research Practices and Challenges: Taking Methods Seriously Methods are just a means and thus cannot substitute for substantive inquiry. But methods do constitute a necessary aspect of research, and a focus on methods becomes all the more important as comparativists increasingly make claims about causality and knowledge accumulation. Indeed, when comparativists make bold claims about the knowledge they have produced,18 and especially when they seek to use these claims as a basis for intervening in policy debates, a responsible view about knowledge, one that takes methods seriously, is imperative. Our data on current research practices in comparative politics suggest that the danger of ignoring basic methodological issues has probably not been stressed enough in recent discussions about the future of the field.19 Linking Theory and Empirics Comparativists commonly combine theory generation and empirical analysis in the same study, a practice that potentially exposes researchers to the risk of using the same data to generate and assess theories. This risk affects qualitative and quantitative researchers alike, and in both cases the solution is the same: ensure that the data used to assess theories is distinct from the data used to generate them. It is hard to establish how widespread this practice is because it may routinely go unreported. Still, there is evidence that the problem of using the same data to generate and assess theories may be more serious for qualitative studies (see Table 6). Comparativists who employ qualitative and quantitative methods of empirical analysis are equally likely to combine theory generation and empirical analysis—the difference is statistically indistinguishable. However, qualitative researchers are far more likely than their quantitative counterparts to combine inductive theorizing and empirical analysis that relies on data of a similar nature—77.5% versus 37.5%. Qualitative researchers are also less likely than quantitative researchers to use multiple methods of theorizing—23.4% versus 54.8%; hence, they miss an important opportunity to decrease the chances of "contamination" between theorizing and empirical analysis. Moreover, this problem is exacerbated because, as will be shown below, qualitative researchers tend to use countries as their unit of observation and usually do not take advantage of the opportunity to multiply the number of observations by formalizing the use of information about units other than countries. The Munck, Snyder / Direction of Comparative Politics 19 Table 6 The Link Between Theory and Empirics I. Issues of Method Methods of Empirical Analysis (% Articles Using Each Method of Empirical Analysis) Mixed Method, Mixed Method, Objective Dominantly Dominantly and Methods Qualitative Qualitative Quantitative Quantitative Total Theory and empirics A. Theory generation and empirical analysis B. Empirical analysis Methods of theorizing A. Inductive, qualitative B. Inductive, quantitative C. Deductive, semiformal or informal D. Deductive, formal Total Multimethods (% of all articles that combine theorizing and empirical analysis) 44.5 19. .9 44.7 18. .2 77.5** 61. .5** 1.3** 7. 7** 21.3 30. .8 0.0** 0.0** 100 100 23.4** 34.5** 14.4 21 .2 11.3 25 .8 40.5** 33 3** 32.4** 37 .5** 21.6 22. .9 5.4** 6.3** 100 100 66.7** 54.8** Note: The figures are the percentage of articles unless otherwise noted. The data on theory and empirics are drawn from the variables theory_empirical and method_analysis of the Munck-Snyder Comparative Politics Articles Data Set (N = 305). The overall total for the methods of theorizing exceeds 100% because individual articles frequently use more than one method of theorizing. The data on methods of theorizing are drawn from the variables method_theory and method_analysis of the Munck-Snyder Comparative Politics Articles Data Set (N = 146). Wilcoxon-Mann-Whitney tests were performed on the data on theory and empirics and on methods of theorizing (in a value-by-value fashion for the four values). **Significant at 1% level. problem, then, is that a considerable number of studies seem not to distinguish clearly between theory generation and empirical analysis as two distinct steps in the research process; they thus offer illustrations of theory or plausibility probes rather than real tests of theory. Testing theories properly also requires an appropriate fit between the concepts used in theories and the data used in empirical analyses. This fit appears to be achieved better by qualitative research, partly because it tends to use a broader variety of types of data than does quantitative research (see Table 7).20 20 Comparative Political Studies Table 7 The Link Between Theory and Empirics II. Issues of Data Methods of Empirical Analysis (% Articles Using Each Method of Empirical Analysis) Mixed Method, Mixed Method, Method of Dominantly Dominantly Data Collection Qualitative Qualitative Quantitative Quantitative Average Secondary sources 87.4 82.8 72.5 51.4 73.5 Newspapers and 55.5 43.1 20.0 6.9 31.4 news sources Government sources 39.3 72.4 67.5 52.8 58.0 and official documents Nonofficial documents 17.8 20.7 10.0 4.2 13.2 Interviews 40.7 32.8 17.5 2.8 23.4 Targeted surveys and 0.7 5.2 7.5 2.8 4.0 questionnaires Mass surveys and 0.7 12.1 20.0 33.3 16.5 questionnaires Other 0.7 3.4 0.0 1.4 1.4 Total 242.9 272.4 215.0 155.6 221.5 Note: N = 305. The overall total exceeds 100% because individual articles frequently use more than one method of data collection. The data are drawn from the variable method_data of the Munck-Snyder Comparative Politics Articles Data Set. Moreover, qualitative research draws far more on "hard-earned" data not accessible via the Internet, such as information collected through interviews, nonofficial documents, and newspapers.21 Thus, qualitative research tends to offer rich accounts of actors, choices, events, and processes, something that, as methodologists have argued, makes qualitative studies especially well suited for testing hypotheses about causal mechanisms (Collier, Brady, & Seawright, 2004, pp. 252-264). In contrast, quantitative studies tend to focus on actors mainly through individual-level data on attitudes and thus have little to say about collective actors and their decision-making processes. Quantitative studies also tend to use data that offer only one observation per year even when analyzing processes that require constant monitoring of events and that are not well summarized by annual data series. The need to generate data that are linked more closely to the concepts used in theorizing is a challenge for comparativists, especially those who rely on quantitative methods of analysis. Munck, Snyder / Direction of Comparative Politics 21 Table 8 Hypothesis Formulation and Methods Objectives and Methods Formulation and Use of a Testable Hypothesis (% Articles With a Given Objective and Using Each Method of Empirical Analysis) Yes Partial No Total Theory and empirics Theory generation Theory generation and empirical analysis Methods of empirical analysis Qualitative Mixed method, dominantly qualitative Mixed method, dominantly quantitative Quantitative 7.1* 28.1* 9 2** 13.8** 47.6** 67.7** 57.1* 58.2* 70.8** 62.1** 52.4** 32.3** 35.7* 13.7* 20.0** 24.1** 0.0** 0.0** 100 100 100 100 100 100 Note: The data are drawn from the variable testable_hypothesis of the Munck-Snyder Comparative Politics Articles Data Set. The N for the data on theory and empirics is 160, the N for the data on methods of empirical analysis is 146. A Wilcoxon-Mann-Whitney test was performed on the data on theory and empirics, and a Kruskal Wallis test was performed on the data on methods of empirical analysis. *Significant at 5% level. **Significant at 1% level. Formalizing Hypothesis Formulation and Data Collection A different kind of methodological challenge concerns the formalization of hypothesis formulation and data collection. It is critical to address these central aspects of the research process in a formalized manner so that knowledge can be tested in a transparent fashion that is open to inspection by the scholarly community. Yet the deficiencies of comparative research in this regard are patent. Only slightly more than one fourth of all articles that combine theory generation and empirical analysis offer testable hypotheses, that is, hypotheses that explicitly specify the variables and the relationship among the variables used in a causal model (see Table 8). Among the rest, only a partial sense of hypotheses can be gained because fundamental issues such as the posited associations among variables are unclear or, even worse, because the variables analyzed in the study are themselves unclear. These problems, it bears noting, are far more common in qualitative work. 22 Comparative Political Studies Table 9 Data Collection and Methods New Data (% Articles Formal Data (% Articles Using Each Method of Using Each Method of Empirical Analysis) Empirical Analysis) Methods of Empirical - - Analysis Yes No Total Yes Partial No Total Qualitative 60.7* Mixed method, 62.1* dominantly qualitative Mixed method, 57.5* dominantly quantitative Quantitative 43.1 * 39.3* 100 2.2** 37.9* 100 8.6** 42.5* 100 52.5** 56.9* 100 76.4** 23.7** 74.1** 100 51.7** 39.7** 100 42.5** 5.0** 100 11.1** 12.5** 100 Note: N = 305. The data are drawn from the variables data_new and data_formal of the Munck-Snyder Comparative Politics Articles Data Set. A Wilcoxon-Mann-Whitney test was performed on the data on new data, and a Kruskal Wallis test was performed on the data on formal data. *Significant at 5% level. **Significant at 1% level. A similar pattern exists with regard to data collection (see Table 9). Though qualitative research introduces new data more often than quantitative research, it is much less likely to analyze data that are formally organized in a data set showing values for all the variables and units. In most articles (74.1%) using qualitative methods of empirical analysis, it is either not possible to readily understand the values assigned to variables or the data presented consist only of values on select units and variables. Quantitative research, though having room for improvement, fares considerably better in this respect. Increasing Coverage and Sample Size A final methodological challenge concerns the coverage of countries and the size of samples. Though comparativists have done fairly well at covering most corners of the world, the old question, "How much research by comparativists is really comparative?" is still germane.22 Indeed, inasmuch as the term comparative refers to comparison across countries, it is striking that nearly half (45.7%) the articles published in journals dedicated to comparative politics are single-country studies.23 Furthermore, inasmuch as the goal of comparative research is to produce knowledge about politics around the world, it is telling that only one fourth of the articles cover more than five countries. Most comparativists do not attempt to provide generalizations of even moderate scope. Munck, Snyder / Direction of Comparative Politics 23 Table 10 Time and Methods Number of Years Methods of (% Articles Using Each Method of Empirical Analysis) Empirical - Analysis n < 1 Year 1 < «< 5 5 < «< 20 20 < «< 50 « > 50 Years Qualitative 2.2* 8.9* 33.3* 32.6* 23.0* Mixed method, 6.9* 5.2* 37.9* 36.2* 13.8* dominantly qualitative Mixed method, 5.0* 7.5* 30.0* 37.5* 20.0* dominantly quantitative Quantitative 23.6* 9.7* 20.8* 34.7* 11.1* Note: N = 305. The data are drawn from the variables time and method_analysis of the Munck-Snyder Comparative Politics Articles Data Set. Kruskal Wallis tests were performed on the data on number of countries and on number of years. *Significant at 5% level. The narrow empirical scope of comparative research has implications for the possibility of testing theories. Much has been written recently about how qualitative researchers who produce case studies or small-Af comparative analyses may be able to surmount the "many variables, small N" problem highlighted by Arend Lijphart (1971, p. 686) more than three decades ago. The core argument is that, despite studying one or a small number of countries, qualitative researchers can multiply the number of observations through within-case analysis. Yet qualitative comparativists do not, in fact, appear to be doing this, at least in articles. Qualitative studies often cover long periods (see Table 10).24 Thus, much potential exists in qualitative research to increase the number of observations through within-case studies over time. Yet qualitative research tends to take the country as the unit of observation, rarely using even time to multiply observations, and thus largely missing the opportunity to increase the number of observations (see Table 11). By contrast, quantitative research has done a much better job exploiting such opportunities because it far more frequently moves away from the country as the unit of observation and thus increases the N. Still, as pointed out above, in the effort to increase the N, quantitative research relies heavily on individual-level data that offer only one observation per year, thereby creating a blind spot to collective actors and to processes that are not captured by annual observations. If qualitative research can get beyond the fixation on the country as the unit of observation, it would appear well poised to generate Table 11 Unit of Observation, Sample Size, and Methods Methods of Empirical Analysis (% Articles Using Each Method of Empirical Analysis) Unit of Observation % Articles per Unit of Observation n on Unit of Observation (Average) Qualitative Mixed Method, Dominantly Qualitative Mixed Method, Dominantly Quantitative Quantitative Country 55.7 6 80.0** 75.9** 27.5** 9 7** Country-time period 16.4 168 9.6** 6.9** 27.5** 30.6** Subnational jurisdiction 6.2 13 5.9 8.6 10.0 2.8 Subnational jurisdiction- 2.0 1,599 0.0** 0.0** 5.0** 5.6** time period Group or organization 4.9 116 3.7 6.9 2.5 6.9 Group-year or organization- 1.3 257 0.0* 0.0* 2.5* 4.2* time period Individual 9.2 1,279 0.0** I 7** 15.0** 29.2** Individual-time period 3.6 12,736 0.0** 0.0** 5.0** 12.5** Other 3.3 1,249 0.7* 1.7* 12.5* 4.2* Total 102.6 100.0 101.7 107.5 105.6 Note: N = 305. The n on unit of observation for "other" excludes an article with an n of 396,167. The totals for the columns on percentage articles per unit of observation and on methods of empirical analysis exceed 100% because individual articles occasionally use more than one unit of observation. The data are drawn from the variables unit_observation, «_observations, «_countries, and method_analysis of the Munck-Snyder Comparative Politics Articles Data Set. Wilcoxon-Mann-Whitney tests were performed on the data on methods of empirical analysis, comparing one unit of observation to all others at a time. *Significant at 5% level. **Significant at 1% level. Munck, Snyder / Direction of Comparative Politics 25 precisely such "within-year" and group-level data. In any event, the "many variables, small N" problem in comparative research remains extremely relevant today. Conclusions Although contentious debates about the direction of comparative politics have raged since the 1980s, they are rarely grounded in systematic evidence about how research in the field is actually done. Hence, this analysis serves as a counterbalance to the tendency to discuss what comparative politics should be in the absence of an empirically valid understanding of what comparative politics is. By analyzing the field's leading journals during the past 15 years, this article provides a much-needed empirical foundation for debating the direction of comparative politics.25 The main conclusions of our analysis are the following. With regard to scope, objectives, and methods, comparative politics is a field that (a) addresses humanly important questions about political order, political regimes, societal forces, democratic and state institutions, economic processes, and extranational processes, in all regions of the world; (b) is far more oriented toward empirical analysis than theory generation; (c) aims to produce descriptive and causal knowledge in nearly equal balance; (d) does not aim to provide knowledge with direct relevance for public policy; (e) generates theory mainly through induction; and (f) relies mostly on qualitative methods of empirical analysis. Comparative politics is thus a rich and diverse field that cannot be accurately characterized on the basis of just one dimension or even summarized in simple terms. Our analysis shows that the tendency to frame choices about the direction of the field in terms of a stark alternative between old and new approaches relies on unsupported assumptions about actual research practices in comparative politics. First, despite two rounds of sharp criticism, from behavioralists in the 1950s and then from a variety of camps in the 1990s, area studies is still the dominant form of research in comparative politics. Most comparative research focuses on a single region of the world. Moreover, despite some differences between area studies and non-area studies work in terms of research objectives and methods, area studies does not constitute a distinctive approach, and blanket characterizations of area studies research as atheoretical or "merely descriptive" are simply misleading. Both the bemoaning of the death of area studies and the broad critiques lodged against them poorly correspond to actual practice in comparative research. Second, despite all the ink spilled 26 Comparative Political Studies recently concerning the merits and demerits of emulating economics, there is scant evidence that comparative politics has, in fact, taken an economic turn. Formal, deductive methods of theorizing are seldom used in comparative research, and deductive theorizing, when it does occur, is of a soft, informal or semiformal, variety. Moreover, there is no evidence that the use of deductive, formal methods is associated with narrow research questions, a charge made by critics of rational choice theory and formal methods. Nor, more broadly, is there any evidence of a strong relationship between "method selection" and "problem selection": Scholars of all theoretical and methodological stripes routinely address important, substantive questions about politics.26 Thus, a key conclusion concerning the debate about the direction of comparative politics is that the field should get beyond a confrontation among broadly conceived approaches that hold false assumptions about their rivals. Comparativists should focus instead on tackling key methodological problems that, as this analysis shows, are widespread in current research and thus seriously hinder the production of knowledge. In this regard, five specific methodological desiderata are emphasized: (a) the use of data to test theories that are distinct from the data used to generate them, (b) the use of data to test theories that are closely linked to the concepts used in theorizing, (c) the formulation of hypotheses that explicitly specify the variables and the relationship among the variables used in a causal model, (d) the analysis of data that consist of values for all the variables and units, and (e) the analysis of a number of observations large enough to assess theories. Addressing methodological challenges such as these will provide a far stronger foundation for producing knowledge about politics around the world. Notes 1. The data set codes 319 articles from three journals—Comparative Political Studies (150 articles), Comparative Politics (112 articles), and World Politics (57 articles)—on 29 variables (a description of the variables can be accessed at http://www-rcf.usc.edu/~munck/ or at http:// brown.edu/polisci/people/snyder/). The articles were drawn from issues published in 1989, 1992, 1995, 1998, 2001, and 2004 (except for the last issue of World Politics, which was not in print when the data set was constructed). In the case of World Politics, a journal that includes articles on international relations, the coders determined which articles belonged to the field of comparative politics. Book reviews, research notes, introductions to special issues, and articles on methodology were not included. The analysis was done using Stata 9.1. The coding was carried out by the two authors of this article. For most variables, each author was responsible for three noncon-secutive years of each journal. In addition, Munck coded all articles on the variables unit_obser-vation and «_observations. Snyder coded all articles on the five variables related to the attributes of authors and the variables foreign_language and funding. See Munck and Snyder (in press-b) for an analysis of the data set that focuses on the attributes of the authors of comparative research. Munck, Snyder / Direction of Comparative Politics 27 2. Another, complementary strategy for analyzing cutting-edge research is to focus on the leading scholars in the field who largely set the research standards and define the substantive agendas that we detect through our statistical analysis of articles in journals. We adopt this second strategy in another work, Passion, Craft, and Method in Comparative Politics (Munck & Snyder, in press-a), which contains in-depth interviews with 15 leading scholars of comparative politics and touches on many of the issues discussed in this article. That book further supplements the current analysis because it has a broader temporal scope, spanning most of the 20th century. 3. In their inventory of articles published in comparative journals during the 1968 to 1981 period, Sigelman and Gadbois (1983) found that government institutions were a neglected topic, a situation they attributed to the attack by behavioralists against formal institutional research. Our data indicate that the study of government institutions has made a comeback. 4. The total percentage exceeds 100% because individual articles frequently cover multiple regions. The data are drawn from the variable region of the Munck-Snyder Comparative Politics Articles Data Set (N = 305). 5. Of the articles, 8.5% span less than a year, 8.2% between 1 year and less than 5 years, 30.1% between 5 years and less than 20 years, 34.4% between 20 years and less than 50 years, and 18.0% more than 50 years. The data are drawn from the variable time of the Munck-Snyder Comparative Politics Articles Data Set (/V = 305). 6. The centrality of empirics in comparative politics also suggests that despite all the discussion about the value of various paradigms and the dangers of research that does not include empirical analysis (Green & Shapiro, 1994), comparativists seldom seek to develop or selfconsciously apply paradigms. In constructing our data set, we initially sought to gauge whether authors worked within certain paradigms, such as a rationalist, culturalist, or structural paradigm (Lichbach, 1997). After surveying numerous articles and finding little evidence that research in these journals made significant use of these paradigms, we dropped that variable. 7. We initially tried to gauge whether articles sought to offer knowledge directly relevant to policy matters. Our review showed that virtually none addressed policy questions in a way that could be construed as talking to public policy makers or offering results relevant to debates in policy circles. Thus, we dropped that variable from our questionnaire. It may be that comparativists choose other outlets, such as Foreign Affairs and Foreign Policy, for their policy-oriented work. 8. Comparativists have also made virtually no use of simulations. 9. This is admittedly a minimalist approach to distinguishing between qualitative and quantitative research, one that is well suited to our goal of coding a large sample of articles. For a multidimensional approach, see Collier, Brady, and Seawright (2004, pp. 244-250). We analyze below several of the dimensions, such as the size of the N and the use of within-case analysis, that Collier et al. highlight to differentiate qualitative and quantitative research. 10. Almond (1990, pp. 244-246) later voiced a different view of area studies, arguing that they had made major contributions to theory development across many substantive questions. 11. See Sigelman and Gadbois (1983) and Hull (1999) on the dominance of area studies research in comparative politics during the 1970s and 1980s. 12. Moreover, the total share of area studies articles varies little across the 15-year period covered by our data set (i.e., 1989-2004). The data are drawn from the variable region of the Munck-Snyder Comparative Politics Articles Data Set (/V = 305). 13. Another difference involves interdisciplinary collaboration: Political scientists engaged in area studies research are more likely to collaborate with non-political scientists. Although non-area studies research involves non-political scientists in 12% of jointly authored articles, the corresponding percentage for area studies research is 26%. However, this difference is not 28 Comparative Political Studies statistically significant. The data are drawn from the variables «_authors, region, and discipline of the Munck-Snyder Comparative Politics Articles Data Set (N = 305). 14. The notion of an alliance between rational choice formal modelers and quantitative analysts was articulated most clearly by the British sociologist John Goldthorpe (1998). An alternative proposal was made by advocates of "analytical narratives" who aimed to combine economic theory—rational choice theory in particular—and formal theorizing with an emphasis on traditional qualitative methods (Bates, Greif, Levi, Rosenthal, & Weingast, 1998). 15. A failure to deal with big, substantive questions has been attributed to a reliance on what are sometimes labeled "theory-driven" or "methods-driven" agendas, usually contrasted with "problem-driven" research grounded in real-world puzzles (Huber & Dion, 2002; Pierson & Skocpol, 2002, pp. 696, 705; Skocpol, 2003, pp. 410-411; Weyland, 2002), and also to the increased technification of research methods (Collier et al., 2004, p. 266). 16. The data are drawn from the variables method_theory and method_analysis of the Munck-Snyder Comparative Politics Articles Data Set (/V = 146). 17. This conclusion, like all conclusions in this article, is based on the three journals we analyze. Formal and quantitative research on comparative politics is regularly published in the three leading "general" political science journals (i.e., The American Political Science Review, The American Journal of Political Science, and The Journal of Politics). Thus, it is possible that the proportion of formal and quantitative comparative research published in leading journals across political science is higher than reported here. 18. See, for example, the arguments that conventional approaches to research, as exemplified by the tradition of comparative historical analysis, have been quite successful at producing knowledge concerning big questions about politics (Kohli, 2002; Mahoney & Rueschemeyer, 2003, pp. 7-10; Pierson & Skocpol, 2002, pp. 696-698; Skocpol, 2003). 19. This emphasis on methods is certainly not new. Previous works that highlight major methodological problems that hinder the production of knowledge about politics include Bollen, Entwisle, and Alderson (1993), Esping-Andersen and Przeworski (2001), and Geddes (2003). 20. Neither qualitative nor quantitative studies use experiments or focus groups as a source of data. Also, both qualitative and, especially, quantitative research rely quite heavily on data from official government sources. The implications of this reliance on official data are unclear. 21. The use of foreign-language sources is also a fairly distinctive trait of qualitative research. Articles that use a qualitative method of empirical analysis use foreign-language sources 75% of the time, whereas the figure for quantitative articles is 50% (chi-square test, significant at 1% level). The data are drawn from the variable foreign_language of the Munck-Snyder Comparative Politics Articles Data Set (/V = 305). 22. For example, in 1963, Eckstein and Apter (1963) wrote that "by far the largest proportion of the writings in the so-called field of comparative politics is essentially not comparative at all" (p. vi). 23. Of the articles, 45.7% study 1 country, 15.9% 2 countries, 6.0% 3 countries, 4.6% 4 countries, 3.3% 5 countries, 6.3% between 6 and 10 countries, 9.6% between 11 and 20 countries, 4.0% between 21 and 30 countries, 2.6% between 31 and 50 countries, 0.7% between 51 and 100 countries, and 1.3% between 101 and 158 countries. The data are drawn from the variable «_countries of the Munck-Snyder Comparative Politics Articles Data Set (/V = 302). Still, our data support Hull's (1999) observation that "comparative politics has become more comparative" (p. 119). In his analysis of articles published in the three leading comparative journals between 1982 and 1996, Hull found that 53.8% of articles were single-country analyses. Among the articles published between 1968 and 1981 in Comparative Politics and Comparative Political Studies analyzed by Sigelman and Gadbois (1983, p. 283), 61.7% were single-country studies. Munck, Snyder / Direction of Comparative Politics 29 Hence, during the past 35 years, there has been a clear decline in the share of single-country studies in the leading comparative politics journals. Of course, single-country studies are not necessarily noncomparative; see Snyder (2001) on this point. 24. The data show that quantitative studies, too, routinely adopt long-range time frames. Hence, a long-range temporal focus appears to cut across any putative quantitative-qualitative divide. 25. As pointed out in the introduction, the articles that form the basis for this analysis, although important, are still only part of the research produced in the field of comparative politics. 26. On "problem selection" in comparative research, see Stepan (2001) and Munck and Snyder (in press-a). References Almond, G. A. (1990). A discipline divided. Schools and sects in political science. Newbury Park, CA: Sage. Almond, G. A., & Coleman, J. S. (Eds.). (1960). The politics of the developing areas. Princeton, NJ: Princeton University Press. Bates, R. H. (1996). Letter from the president: Area studies and the discipline. APSA-CP: Newsletter of the APSA Organized Section in Comparative Politics, 7(1), 1-2. Bates, R. H., Greif, A., Levi, M., Rosenthal, J.-L., & Weingast, B. (1998). Analytic narratives. Princeton, NJ: Princeton University Press. Bollen, K. A., Entwisle, B., & Alderson, A. S. (1993). Macrocomparative research methods. Annual Review of Sociology, 19, 321-351. Brady, H. E., & Collier, D. (Eds.). (2004). Rethinking social inquiry: Diverse tools, shared standards. Lanham, MD: Rowman & Littlefield, Berkeley Public Policy. Brady, H. E., Collier, D., & Seawright, J. (2004). Refocusing the discussion of methodology. In H. E. Brady & D. Collier (Eds.), Rethinking social inquiry: Diverse tools, shared standards (pp. 3-20). Lanham, MD: Rowman & Littlefield, Berkeley Public Policy. Collier, D. (1999). Data, field work and extracting new ideas at close range. APSA-CP. Newsletter of the APSA Organized Section in Comparative Politics, 10(1), 1-2, 4-6. Collier, D., Brady, H. E., & Seawright, J. (2004). Sources of leverage in causal inference: Toward an alternative view of methodology. In H. E. Brady & D. Collier (Eds.), Rethinking social inquiry: Diverse tools, shared standards (pp. 229-266). Lanham, MD: Rowman & Littlefield, Berkeley Public Policy. Eckstein, H., & Apter, D. (Eds.). (1963). Comparative politics. New York: Free Press. Esping-Andersen, G., & Przeworski, A. (2001). Quantitative cross-national research methods. In N. J. Smelser & P. B. Bates (Eds.), International encyclopedia of the social and behavioral sciences (pp. 12649-12655). New York: Elsevier Science. Geddes, B. (1991). How the cases you choose affect the answers you get: Selection bias in comparative politics. In J. A. Stimson (Ed.), Political analysis, vol. 2 1990 (pp. 131-149). Ann Arbor: University of Michigan Press. Geddes, B. (2003). Paradigms and sand castles: Theory building and research design in comparative politics. Ann Arbor: University of Michigan Press. Goldthorpe, J. H. (1998). The quantitative analysis of large-scale data-sets and rational action theory: For a sociological alliance. In H.-P. Blossfeld & G. Prein (Eds.), Rational choice theory and large-scale data analysis (pp. 31-53). Boulder, CO: Westview. Green, D., & Shapiro, I. (1994). Pathologies of rational choice. A critique of applications in political science. New Haven, CT: Yale University Press. 30 Comparative Political Studies Huber, E., & Dion, M. (2002). Revolution or contribution? Rational choice approaches in the study of Latin American politics. Latin American Politics and Society, 44(3), 1-28. Hull, A. P. (1999). Comparative political science: An inventory and assessment since the 1980's. PS: Political Science and Politics, 32(1), 117-124. Keech, W., Bates, R., & Lange, P. (1991). Political economy within nations. In W. Crotty (Ed.), Political science: Looking to the future, vol. 2. Comparative politics, policy, and international relations (pp. 219-263). Evanston, IL: Northwestern University Press. King, G., Keohane, R. O., & Verba, S. (1994). Designing social inquiry. Scientific inference in qualitative research. Princeton, NJ: Princeton University Press. Kohli, A. (2002). State, society, and development. In I. Katznelson & H. V. Milner (Eds.), Political science: The state of the discipline (pp. 84-117). New York: Norton. Laitin, D. D. (1993). Letter from the incoming president. APSA-CP. Newsletter of the APSA Organized Section in Comparative Politics, 4(2), 3, 18. Laitin, D. D. (1998). Toward a political science discipline: Authority patterns revisited. Comparative Political Studies, 31, 423-443. Laitin, D. D. (2002). Comparative politics: The state of the subdiscipline. In I. Katznelson & H. V. Milner (Eds.), Political science: The state of the discipline (pp. 630-659). New York: Norton. Laitin, D. D. (2003). The perestroikan challenge to social science. Politics and Society, 31(1), 163-184. Laitin, D. D. (2004). The political science discipline. In E. Mansfield & R. Sisson (Eds.), The evolution of political knowledge: Democracy, autonomy, and conflict in comparative and international politics (pp. 11-40). Columbus: Ohio State University Press. Levi, M. (1997). A model, a method, and a map: Rational choice in comparative and historical analysis. In M. I. Lichbach & A. S. Zuckerman (Eds.), Comparative politics: Rationality, culture and structure (pp. 19-41). New York: Cambridge University Press. Levi, M. (2000). The economic turn in comparative politics. Comparative Political Studies, 33(611), 822-844. Lichbach, M. I. (1997). Social theory and comparative politics. In M. I. Lichbach & A. S. Zuckerman (Eds.), Comparative politics: Rationality, culture and structure (pp. 239-276). New York: Cambridge University Press. Lichbach, M. I. (2003). Is rational choice theory all of social science? Ann Arbor: University of Michigan Press. Lijphart, A. (1971). Comparative politics and the comparative method. American Political Science Review, 65(3), 682-693. Lustick, I. S. (1994). Rational choice as a hegemonic project, and the capture of comparative politics. APSA-CP. Newsletter of the APSA Organized Section in Comparative Politics, 5(2), 7, 31. Macridis, R., & Cox, R. (1953). Research in comparative politics. Seminar report. American Political Science Review, 47(3), 641-657. Mahoney, J., & Rueschemeyer, D. (2003). Comparative historical analysis: Achievements and agendas. In J. Mahoney & D. Rueschemeyer (Eds.), Comparative historical analysis in the social sciences (pp. 3-38). New York: Cambridge University Press. Mansfield, E., & Sisson, R. (Eds.). (2004a). The evolution of political knowledge: Democracy, autonomy, and conflict in comparative and international politics. Columbus: Ohio State University Press. Mansfield, E., & Sisson, R. (Eds.). (2004b). The evolution of political knowledge: Theory and inquiry in American politics. Columbus: Ohio State University Press. Munck, Snyder / Direction of Comparative Politics 31 Monroe, K. R. (Ed.). (2005). Perestroika! The raucous rebellion in political science. New Haven, CT: Yale University Press. Munck, G. L., & Snyder, R. (2004). What has comparative politics accomplished? APSA-CP. Newsletter of the APSA Organized Section in Comparative Politics, 15(2), 26-31. Munck, G. L., & Snyder, R. (in press-a). Passion, craft, and method in comparative politics. Baltimore, MD: Johns Hopkins University Press. Munck, G. L., & Snyder, R. (in press-b). Who publishes in comparative politics? Studying the world from the United States. PS: Political Science & Politics. Pierson, P., & Skocpol, T. (2002). Historical institutionalism in contemporary political science. In I. Katznelson & H. V. Milner (Eds.), Political science: The state of the discipline (pp. 693-721). New York: Norton. Portes, A. (2005). Sociology in the hemisphere: Past convergencies and a new middle-range agenda. In C. H. Wood & B. R. Roberts (Eds.), Rethinking development in Latin America (pp. 27-52). University Park: Pennsylvania State University Press. Sartori, G. (2004). Where is political science going? PS: Political Science & Politics, 37(4), 785-786. Schmitter, P. C. (1993). Comparative politics. In J. Krieger (Ed.), The Oxford companion to the politics of the world (pp. 171-177). New York: Oxford University Press. Shapiro, I., Smith, R., & Masoud, T. (Eds.). (2004). Problems and methods in the study of politics. New York: Cambridge University Press. Sigelman, L., & Gadbois, G. H. (1983). Contemporary comparative politics: An inventory and assessment. Comparative Political Studies, 16(3), 275-305. Skocpol, T. (2003). Doubly engaged social science: The promise of comparative historical analysis. In J. Mahoney & D. Rueschemeyer (Eds.), Comparative historical analysis in the social sciences (pp. 407-428). New York: Cambridge University Press. Snyder, R. (2001). Scaling down: The subnational comparative method. Studies in Comparative International Development, 36(1), 93-110. Stepan, A. (2001). Reflections on "problem selection" in comparative politics. In A. Stepan (Ed.), Arguing comparative politics (pp. 1-10). New York: Oxford University Press. Weingast, B. R. (2002). Rational choice institutionalism. In I. Katznelson & H. V. Milner (Eds.), Political science: The state of the discipline (pp. 660-692). New York: Norton. Weyland, K. (2002). Limitations of rational-choice institutionalism for the study of Latin American politics. Studies in Comparative International Development, 37(1), 57-85. Gerardo L. Munck, Argentinian by birth, teaches in the School of International Relations at the University of Southern California and does research on political regimes and democracy, methodology, and Latin America. He is author of Authoritarianism and Democratization. Soldiers and Workers in Argentina, 1976-83; coauthor, with Snyder, of Passion, Craft, and Method in Comparative Politics; and editor of Regimes and Democracy in Latin America. Theories and Methods. E-mail: munck@usc.edu. Richard Snyder is an associate professor of political science at Brown University. He studies the political economy of development, political regimes, and Latin American politics. He is author of Politics after Neoliberalism: Reregulation in Mexico and coauthor, with Munck, of Passion, Craft, and Method in Comparative Politics. E-mail: Richard_Snyder@brown.edu.