ABSTRACT
This article provides an overview of recent developments in historical institutionalism. First, it reviews some distinctions that are commonly drawn between the “historical” and the “rational choice” variants of institutionalism and shows that there are more points of tangency than typically assumed. However, differences remain in how scholars in the two traditions approach empirical problems. The contrast of rational choice’s emphasis on institutions as coordination mechanisms that generate or sustain equilibria versus historical institutionalism’s emphasis on how institutions emerge from and are embedded in concrete temporal processes serves as the foundation for the second half of the essay, which assesses our progress in understanding institutional formation and change. Drawing on insights from recent historical institutional work on “critical junctures” and on “policy feedbacks,” the article proposes a way of thinking about institutional evolution and path dependency that provides an alternative to equilibrium and other approaches that separate the analysis of institutional stability from that of institutional change.

INTRODUCTION

Institutional analysis has a distinguished pedigree in comparative politics, and the “new” institutionalist literature of the past two decades has both sustained this venerable tradition and deepened our understanding of the role of institutions in political life. At the same time, recent work has given rise to new debates. It is now conventional to distinguish three different varieties of institutionalism: rational choice institutionalism, historical institutionalism, and sociological institutionalism. Each of these three schools in fact represents a sprawling literature characterized by tremendous internal diversity, and it is often also difficult to draw hard and fast lines between them. The differences that have been identified amount to tendencies that apply unevenly across particular authors within each school of thought (Hall & Taylor 1996). The walls dividing the three perspectives have also been eroded by “border crossers” who have resisted the tendencies toward cordoning these schools off from each other and who borrow liberally (and often fruitfully) where they can, in order to answer specific empirical questions.Afew examples will suffice to illustrate this point.

First, a group of prominent theorists working out of a rational choice perspective have become proponents of a more eclectic approach that combines elements of deductive theory “the hallmark of rational choice” with an explicit attempt to contextualize the analysis in ways that historical institutionalists have long advocated (Bates et al 1998b). This strategy, which they call analytic narratives, represents an attempt to construct explanations of empirical events through analyses that “respect the specifics of time and place but within a framework that both disciplines the detail and appropriates it for purposes that transcend the particular story” (Levi 1999). The analyses offered thus incorporate elements of deduction and induction in ways that overcome traditional distinctions between historical institutionalism’s characteristic focus on specific contextual conditions and rational choice’s characteristic search for generalizable features of political behavior rooted in the incentive structures that individuals face.

A second illustration of border crossing is the equally important impact of rational choice on the work of many historical institutionalists. One development has been an enhanced appreciation of the need for explanations that rest on firm micro foundations. Although much macro-historical work was already implicitly sensitive to these issues, articulating the micro-foundational logic of the arguments offered was not always a top priority, and recent work reflects a heightened appreciation for such issues. Many historical institutionalists have also taken on board the notion that institutions that solve collective action problems are particularly important in understanding political outcomes (Rothstein 1996:159). This has long been a central concern of rational choice theory, and it has set an important agenda for historical institutionalists as well. Thus, an increasing number of historical studies focus precisely on explaining the emergence and persistence of institutions that do (or do not) facilitate coordination among employers and other groups (see e.g. Hall 1994, Thelen&Kume 1999).

Third, there has been some important borrowing and cross-fertilization between historical institutionalism and sociological institutionalism. The works that lie at this intersection often embrace a more expansive view of institutions, not just as strategic context but as a set of shared understandings that affect the way problems are perceived and solutions are sought. Katzenstein’s analysis of the evolution of Japanese security policy, for example, shows how collectively held norms define appropriate conduct, shape actor identities, and influence actor interests (1996:23), and in doing so, “inform how political actors define what they want to accomplish” (1996:ix). Katzenstein roots his cultural approach in a political analysis of how some norms (and not others) came to be institutionalized, and so his perspective resonates especially with those versions of institutional sociology that specifically incorporate considerations of power and/or legitimacy in explaining how institutions emerge and are reproduced (e.g. Fligstein 1991, DiMaggio 1988, Stinchcombe 1997). In sum, there have been some rather fruitful developments at the intersection of these various schools of institutionalism, and in my view historical institutionalism has been enriched by encounters with alternative perspectives.

This article provides an overview of recent developments within the historical institutional tradition. Given the vast scholarship in this area, this tour is necessarily selective. I attempt to capture the current state of the literature in two passes. First, I review some distinctions that are commonly drawn between the “historical” and the “rational” choice variants of institutionalism. This exercise reveals that there are more points of tangency than commonly assumed. However, differences remain, and here I focus on the difference between rational choice’s emphasis on the coordinating functions of institutions (generating or maintaining equilibria) versus historical institutionalism’s emphasis on how institutions emerge from and are embedded in concrete temporal processes. This discussion serves as the foundation for the second half of the essay, in which I revisit one of the key frontiers in historical institutionalism (identified in Thelen & Steinmo 1992) and assess the progress that has been made in our understanding of institutional formation and change. Taking up an important challenge by Orren & Skowronek (1994), and drawing on insights from recent historical institutional work on critical junctures and on policy feedbacks, I propose a way of thinking about institutional evolution and path dependency that provides an alternative to equilibrium and other approaches that separate the analysis of institutional stability from that of institutional change.

RED HERRINGS AND REAL ISSUES

Theoretical Versus Empirical Work

One of the lines that is frequently drawn between historical institutionalism and rational choice institutionalism is between “theoretical” and “empirical” work. In a well-known critique, Green & Shapiro (1994) charge that rational choice has produced elegant theories but has generated little to explain real observed events. From the other side, rational choice theorists have often argued that historical institutionalists are engaged in something less than theory building; they are stringing details together, “merely telling stories.” Even where the distinction is not drawn so starkly, the assertion is that the difference is fundamental. Thus, for example, Levi argues (in opposition presumably to historical institutionalism) that rational choice is “almost always willing to sacrifice nuance for generalizability [and] detail for logic” (1997a:21).

This dichotomy is often exaggerated, and in my view it is misplaced, for the best work in both perspectives is concerned with generating hypotheses that are then brought to bear on empirical phenomena. For example, Luebbert’s (1991) analysis of the origins of fascism, social democracy, and liberal democracy is based on a comparative analysis of class relations in the interwar period in Europe. Combining comparative method with close historical process tracing in individual cases, Luebbert singles out as the crucial explanatory variable the issue of how the landed peasantry was mobilized politically. Where working-class-based parties allied themselves with the landed peasantry, this produced the mass base necessary for the establishment of social democratic regimes. By contrast, where social democrats failed to forge this alliance, the landed peasantry turned against the working class and provided the mass base on which fascism grew. One can disagree with his findings, but Luebbert’s work is an exemplary model of the testing, through the comparative method, of a strong and clear hypothesis that is obviously capable of being falsified.

Another example is Collier & Collier’s (1991) impressive study of regime transformation in Latin America. Based on structured comparisons and historical process tracing for individual cases, they found that differences in patterns of labor incorporation were key to the type of regime that emerged. Like Luebbert’s work, this study was not meant to capture every detail; on the contrary, Collier&Collier’s account is precisely designed to probe the plausibility of alternative hypotheses. Much is thus lost in terms of a comprehensive history of each country, but the payoff is a set of general propositions about the way in which labor incorporation affected subsequent regime outcomes across a number of countries.

Many other works could be invoked that use the comparative historical method to sort out the causal mechanisms behind observed empirical patterns
(see the discussion, below, of the critical junctures literature). All of them “go beyond conventional history’s preoccupation with historical particularity and aim for theoretical generalization” (Rueschemeyer et al 1992:4). In this, they share much with rational choice analyses as characterized by Levi, sacrificing much detail in order to identify general causal patterns that hold across a number of countries.

Conversely, rational choice work in comparative politics arguably has become more empirical over time. For example, the authors of the “analytic Narratives” project mentioned above specifically emphasize that the papers they present are “problem driven, not theory driven; they are motivated by a desire to account for particular events or outcomes. They are devoted to the exploration of cases, not to the elaboration of theory...” (Bates et al 1998b:11). Some of the analyses are comparative, others focus on single cases; however, like good single case studies by historical institutionalists, the latter use close analysis of critical cases to illuminate important general issues.

Although the difference between the historical and rational choice variants of institutionalism cannot be summed up accurately in a strict dichotomy of “empirically” versus “theoretically” oriented work, there do appear to be some differences between the two traditions’ approaches to theory building. First, most historical institutionalists are working at the level of mid-range theory of the sort that Bendix and others have advocated. Often, though not always (e.g. Rueschemeyer et al 1992, Karl 1997), this involves focusing on a limited range of cases that are unified in space and/or time, for example, explaining transitions to democracy in Eastern Europe (Stark & Bruszt 1998) or the effects of new international pressures on the political economies of the advanced industrial democracies (Hall 1994). Rational choice theorists, by contrast, sometimes aspire to produce more general (even universal) theoretical claims or use historical examples not so much for their intrinsic importance (e.g. as “critical cases”) but to demonstrate how widely applicable are the theoretical claims (e.g. Knight 1992, Levi 1988, Tsebelis 1990). Still, the object of both is to test theoretical propositions against observed phenomena, in order not only to explain the cases at hand but also to refine the theory.

A second difference with respect to theory building lies in the ways that historical institutionalists and rational choice institutionalists approach hypothesis formation. Very frequently, historical institutionalists begin with empirical puzzles that emerge from observed events or comparisons: Why did the policies of the advanced industrial countries differ so much in response to the oil shock of 1973 (Katzenstein 1978)? Why have some industrial relations systems proved more stable than others in the face of globalization pressures (Thelen 1993, 1994)? Why do some countries tax and spend more than others (Steinmo 1993)? The analyst then uses the comparisons to test hypotheses that can account for the observed differences. Rational choice theorists often proceed somewhat differently, deriving their puzzles from situations in which observed behavior appears to deviate from what the general theory predicts: Why, given free rider problems, do workers join unions (Wallerstein 1989)? Why would unions lead workers into hopeless battles (Golden 1997)? Why would citizens ever volunteer for war (Levi 1997b)? This difference in the analytic point of departure accounts for the extensive use of counterfactuals in rational choice research (e.g. Bates et al 1998a), which function much as crossnational comparisons often do in historical institutional research. The question is, what is “off the equilibrium path” (Levi 1997a:31)? The theory gives us prior expectations that we can then hold against the observed outcomes (see also Scharpf 1997:29).

The point is that it is not the case that one perspective’s analysis is guided by clear hypotheses and the other’s is not. The issue is where these hypotheses come from. Moreover, the difference in sources of hypotheses is not a hard and fast rule. As Scharpf points out, rational choice theories seek “to explicate what the authors of ‘good’ case studies always have in the back of their minds: a ‘framework’ that organizes our prior (scientific and prescientific) knowledge about what to expect in the province of the world that is of interest to us, that emphasizes the questions that are worthwhile asking, the factors that are likely to have high explanatory potential, and the type of data that would generally be useful in supporting or invalidating specific explanations” (1997:29_30).

Moreover, there is much overlap between the two perspectives when it comes to testing hypotheses against empirical cases, for this invariably involves contextualizing the theory (assumptions and propositions) and demonstrating that the hypothesized processes are actually at work (Pierson 1996: 158). As Scharpf puts it, “We need to have hypotheses that specify a causal model showing why and how a given constellation of factors could bring about the effect in question,” but equally, “we need to have empirical evidence that the effect predicted by the hypothesis is in fact being produced” (1997:28). There is, in other words, no dichotomy between theoretical and empirical work because good analyses have to be both. The generation of the hypotheses is not the analysis, although it is the vital starting point for engaging the empirical material. The utility of a theory, after all, cannot be assessed apart from the empirical material it is meant to explain.

Preferences: Problematical or Not

In 1992, Steinmo and I argued that “one, perhaps the core, difference between rational choice institutionalism and historical institutionalism lies in the question of preference formation, whether treated as exogenous (rational choice) or endogenous (historical institutionalism)” (Thelen & Steinmo 1992:9, emphasis in original). I now think this may be a less stark difference than before. Setting aside some important ambiguities on this issue in both the rational choice and the historical institutionalist literature, one of the core claims of historical institutionalism from the beginning was that institutions do more than channel policy and structure political conflict; rather, “the definition of interests and objectives is created in institutional contexts and is not separable from them” (Zysman 1994:244). In almost any version this is quite different from strong versions of rational choice theory, which begin with a (universal, not context-specific) rationality assumption.

Or is it? As Levi has suggested, in rational choice “the trick is in defining the preferences in general, ex ante to a particular application” (1997a:24). She notes that this is frequently difficult; for example, in the case of citizens, “given the range of interests [they] might have; there is nothing comparable to the economics dictum of getting the most for the least for one’s money in the marketplace” (1997a:24). In a candid discussion of the problem of imputing preferences, she notes the danger that assuming utility maximization “can...produce tautology: Whatever people do becomes a ‘revealed preference‘” (1997a:24).

Even the assumptions traditionally considered “safe” may be trickier than we thought, the most common perhaps being that politicians are seeking reelection. This cannot be valid for Mexico, where the rules of the game (read: the specific institutional configuration) rule out, by law, reelection of legislators. (The same is true for presidents across a much larger number of countries.) In such cases, an analysis that starts with the reelection-seeking assumption will be widely off the mark. How do rational choice analysts deal with this situation? By and large they try to figure out, within a given context, what would make sense for a politician to seek, which is to say that they contextualize the preferences based on the particular institutional incentives that these politicians face in different settings. This is not so different from how historical institutionalists often proceed. The point is that the move from general propositions about what political actors are seeking to maximize inevitably brings the theorist face to face with the question of what it means to, say, maximize power within a given context. Until this step is complete, the analysis cannot begin. This seems to be what Bates et al (1998a) mean when they note that the “’cultural
‘knowledge required to complete a rational choice explanation reveals the complementarity of the [“rational” and “interpretive”] approaches. Game theorists often fail to acknowledge that their approach requires a complete political Anthropology”. Game-theoretic accounts require detailed and finegrained knowledge of the precise features of the political and social environment within which individuals make choices and devise political strategies
“ (1998a:628).

In addition, norms and culture, which for a long time were of concern mainly to historical institutionalists and institutional sociologists, appear to be assuming an increasingly important role in rational choice analysis. Ferejohn’s recent work, for example, argues that “culturally shared understandings and meanings” are crucial to selecting among the many possible strategic equilibria (1991:285). He argues that

in social action, human agents make strategic or allocative choices while simultaneously enacting (ontologically) prior understandings about the nature of the strategic situation in which they find themselves, the characteristics or identities of the players (including themselves), and the common understandings or expectations as to how the game will be played. Thus, when it comes to explaining action, rational accounts, no less than interpretive ones, must appeal to principles external to the individual agents. (Ferejohn 1991: 285)

Many rational choice theorists follow North in his argument that norms constitute informal institutions- another set of rules that create incentives or constraints on behavior. However, theorists differ in explaining how norms and culture fit into the analysis of political outcomes. Some authors see cultural symbols and norms as resources invoked in strategic interactions (e.g. Johnson 1997); others view them as signaling devices in games of incomplete information (e.g. Bates et al 1998a); still others as focal points that affect which of a number of possible equilibria prevails (Rogowski 1997, Greif 1994). Here it is sufficient to note that in order to operate in any of these ways, norms must exert some independent power over individual behavior, and in this sense most of these works appear to go well beyond traditional “instrumental rationality” assumptions. In light of such developments, the issue of preferences (exogenous/ endogenous) no longer seems to provide a clear line of demarcation between the different approaches.

Micro-Foundational Versus Macro-Historical Research

The issue of micro foundations is a third one that is typically cited as distinguishing historical institutional research from rational choice. The idea is that aggregate outcomes need to be understood in terms of the actions and behavior of individuals behaving strategically. This is frequently contrasted to broad macro-historical research that either sees interests as structurally generated in one way or another or stays at the level of aggregations (such as class) without regard for the way in which the strategic actions of individuals figure into the aggregation process.

But this too is a false dichotomy. The issue of the embeddedness of interests has been discussed above; however, there is the additional question of how to think about aggregate behavior. “Micro-foundational” obviously does not preclude dealing with collectivities, since most rational choice work deals with collective actors. The analysis remains, nonetheless, “actor-centered” in the sense that the players are defined as “any individual or composite actor that is assumed to be capable of making purposeful choices among alternative courses of action” (Scharpf 1997:7). This may be unproblematical, as for example in Scharpf’s case, since unions (or at least the union leadership) and not individual workers are the relevant players who actually engage in the strategic bargaining that generates the policy outcomes of interest. In other cases, however, it is worthwhile questioning whether the collectivities to which strategic action is attributed in fact constitute players in the sense that Scharpf identifies.

Take the example of Rogowski’s (1987) analysis of coalitional alignments and realignments in the late nineteenth century. Rogowski’s analysis hinges on the interaction of land (agricultural interests), labor, and capital, aggregations that are viewed as acting purposefully and strategically in the face of changing conditions in international trade. To be persuasive, we need to do more than impute (actor-centered) motives and strategies to these aggregations; we have to demonstrate that these actors were in fact players in Scharpf’s sense, i.e. that these aggregations were cohesive and strategic. But we know from historical Work-and indeed from the logic of Rogowski’s own analysis-that these aggregations were not coherent players, each being riven by internal tensions that derive from precisely the changes in international trade that are the focus of Rogowski’s analysis. Rogowski’s argument itself suggests that business interests would have been divided, depending on whether firms were oriented toward the domestic market or the international market. And we know from historical work that this was in fact the case. Similarly, among labor, workers’ interests frequently followed the divisions within business, opening the door for “cross-class alliances” that were crucial to the outcomes in which Rogowski is interested (Gourevitch 1986, Swenson 1989). Swenson’s (1989) deeply inductive, historical approach is clearly more micro-foundational than Rogowski’s rational choice approach in the sense that Swenson gets much closer to the actual players whose strategic interactions produced the outcomes at issue.

There is no dichotomy because taking micro foundations seriously means that we cannot be content to impute coherence to actors identified by the analyst; we must do the empirical work to make sure that the actors to whom we attribute certain strategic behaviors are in fact “players” in the first place. In other words, the fact that a particular analysis employs a rational choice perspective does not necessarily mean that it stands on strong micro foundations. Conversely, historical institutional research is not necessarily not microfoundational -quite the contrary, as we have seen.

Functional Versus Historical View of Institutions

It is frequently noted that, unlike most historical institutionalism, a good deal of rational choice theory embraces a functional view of institutions (e.g. Hall & Taylor 1996:943-44, Pierson 1996). This distinction probably originated in the two schools’ differing approaches to the issue of institutions. Zysman notes succinctly that “rational choice institutionalists start with individuals and ask where institutions came from, whereas historical institutionalists start with institutions and ask how they affect individuals’ behaviors” (1992, comments at Conference of Europeanists, Chicago). Indeed, much of the early and pathbreaking work in rational choice did pose the question in the way Zysman suggests, and the question of why institutions emerge and are sustained has been answered in terms of the functions that they perform. One example is the literature on the US Congress and the way its rules eliminate “cycling”; another is the rational choice literature on institutions in international relations, which are viewed as mechanisms by which states can reduce transaction costs and achieve joint gains in an anarchic world (e.g. Shepsle & Weingast 1981,
Shepsle 1986, Moravcsik 1993).

However, there has been much work within rational choice that, like historical institutionalism, embraces a non-functionalist, more historical view of institutions (Pierson 1996:131). North’s later work (e.g. 1990), for example, is concerned with tracing, historically, the emergence of different kinds of institutional arrangements that either promote or distort development. Knight too has criticized a good deal of rational choice literature for embracing either an evolutionary or a spontaneous-creation view of institutions (1992:ch. 1). Against more functionalist accounts, Knight sets his own model of institutional formation and change, which places issues of distributional conflict at the center of the analysis.

These works share much with the more historical view of institutions embraced by historical institutionalists. Zysman writes of political-economic institutions, “The institutional approach begins with the observation that markets, embedded in political and social institutions, are the creation of governments and politics” (1994:244). This is close to Knight’s view, and for that matter March & Olsen’s (1989); none of these works makes any assumptions about the social efficiency of institutional arrangements, and all of them allow for suboptimality and inefficiency. Differences remain, of course. Here I simply wish to point out that these differences do not boil down to the commonly cited divide between rational choice’s “functionalist” and historical institutionalist’s more “historical” approaches to institutions.

Synthesis or Creative Borrowing?

Should we conclude from the above that there has been a blending of these different approaches? In my view, no. What we see is a partial convergence of the issues at stake as, for example, historical institutionalists have come to a deeper appreciation of micro foundations and problems of collective action, and rational choice theorists have come to treat preferences, norms, and beliefs as a more central (also more complicated) issue than heretofore. But, as pointed out above, differences remain-in how theorists working out of these different traditions approach these issues, in how they generate the hypotheses that guide their work, and in the level at which they attempt to build theory, for example.

Rather than a full-fledged synthesis, we might instead strive for creative combinations that recognize and attempt to harness the strengths of each approach. This seems to be the strategy advocated by prominent proponents within each school, Zysman from a historical institutionalist perspective and Scharpf from a rational choice perspective. Thus, for example, Zysman (1994:277) argues that

institutions and broad processes of social change certainly have microfoundations.
The “naked” institution emerging from a state of nature by rational choice and the “socially embedded” institution are one and the same, but they represent two different narratives whose perspectives highlight different processes within a common story. That is, the arguments built around institutions and historical dynamics should be consistent with notions of the “rational” dynamics of individual behavior. Inconsistencies are instructive to both those who would build micro-foundations and macro-theories.

He likens the difference between the two perspectives to that between “highlevel computer languages (historical narrative) and the bit-level machine language of the computer (microeconomic narrative)” and maintains that “inherently they must work together, they must be consistent” (Zysman 1994:277) and that “issues must be segmented to make appropriate use of the perspectives, not to reject the insight of one or the other as part of an ideological quarrel” (1994:278; see also Ostrom 1995, esp. pp. 177-78).

Scharpf (1997) elaborates in some detail how this might be achieved. He advocates the use of rational choice and game theory as a way of generating hypotheses that can discipline empirical analysis, but he acknowledges that “even when we can rely on models with high predictive power, they are likely to be of limited scope and will only represent certain subsets of the complex, multiarena and multilevel interactions that are characteristic of real-world processes” (1997:31). This being the case, he argues that “it is usually nec essary to combine several such modules into a more complete explanation” (1997:31, emphasis in original). The composite explanation of particular processes

is likely to be unique for each country but...the modules employed in constructing it may reappear more frequently in other cases as well and thus are more likely to achieve the status of empirically tested theoretical statements. Even then, however, the linkages between these modules remain problematical....Thus we will often depend on narrative, rather than analytical, connections between partial theories that have analytical as well as empirical support-which also means that the composite explanation itself remains vulnerable to charges of being ad hoc. (Scharpf 1997:32)

Equilibrium Order Versus Historical Process

One area does seem to distinguish these two analytic approaches, at least in emphasis. This distinction can be characterized in terms of the relative centrality of “equilibrium order” versus “historical process” in the analysis of political phenomena. One of the defining features of rational choice institutionalism is its assumption of equilibria and its view of institutions as coordinating mechanisms sustaining these equilibria (Levi 1997a:27, Scharpf 1997:10, Shepsle 1989:145). As Levi (1997a) has emphasized, this does not imply efficient equilibria; indeed, a major concern of rational choice has been to sort out the reasons why individual actors, behaving rationally, often produce suboptimal or inefficient collective outcomes. Nor does the equilibrium assumption imply that there exists a single, unique equilibrium outcome; another central problem in rational choice is to understand the process through which one equilibrium rather than another is reached. And finally, this assumption does not mean that rational choice theorists are uninterested in political change-merely that they tend to treat it as involving a transition between equilibrium orders. As Orren & Skowronek put it, “institutional politics appears as ‘normal,’ as politics as usual, explicitly or implicitly opposed to an extraordinary politics, in which equilibria are upset, norms break down, and new institutions are generated” (1994:316). Bates et al (1998a) appear to concur, also with the implicit critique: “The greatest achievement of rational choice theory has been to provide tools for studying political outcomes in stable institutional settings… Political transitions seem to defy rational forms of analysis” (1998a:604-5). (Bates et al address this problem by incorporating elements of an “interpretivist” approach into the analysis).

Whereas rational choice theorists tend to view institutions in terms of their coordinating functions, historical institutionalists see institutions as the legacy of concrete historical processes. In embracing this view, historical institutionalism
“brings questions of timing and temporality in politics [rather than equilibrium order] to the center of the analysis of how institutions matter” (Orren& Skowronek 1994:312). This does not mean that historical institutionalists are uninterested in regularities and continuities in politics;14 it just means that the emphasis tends to be on political development as a (structured) process and on the way institutions emerge from particular historical conflicts and constellations (e.g. Steinmo 1993, 1994). As Pierson puts it, historical institutionalism“stresses that many of the contemporary implications of…temporal processes are embedded in institutions-whether these be formal rules, policy structures, or norms” (1996:126; see also Skocpol 1992:58-59).

Orren & Skowronek emphasize two features of political life that have been central to historical institutional analyses. The first is that “institutions, both individually and collectively, juxtapose different logics of political order, each with their own temporal underpinnings” (1994:320). That is, the various institutional arrangements that make up a polity emerge at different times and out of different historical configurations. For this reason, the various “pieces” do not necessarily fit together into a coherent, self-reinforcing, let alone functional, whole. For example, some analysts treat the German political economy as a well-integrated “system” in which various institutional subsystems-vocational education and training, collective bargaining institutions, financial institutions, bank-industry links-form a mutually reinforcing whole. Historical institutionalists, by contrast, are likely to be concerned with the origins rather than the functions of the various pieces, and indeed, historically oriented research has demonstrated that the evolution of the German model was highly dyssynchronous and full of unintended consequences (Streeck 1992; P Manow, unpublished manuscript; Thelen & Kume 1999). This has important implications for our view of the operation of this system. Streeck (1997), for example, has drawn attention to the ways in which industrial-relations institutions actively generate problems and pressures in other parts of the system, especially vocational education and social welfare institutions. As in the more functionalist view, the interdependencies among the various parts of the system are central to the analysis; however, in Streeck’s work the frictions as well as the functional interdependencies come to the fore. The result is very much in the spirit of Orren & Skowronek’s characterization: “The single presumption abandoned is that institutions are synchronized in their operations or synthetic in their effects; the more basic idea, that institutions structure change in time, is retained” (1994:321).

The second claim, related to the first, is that one important source of change comes from the interactions of different institutional orders within a society, as “change along one time line affects order along the others” (Orren & Skowronek 1994:321), that is, as interactions and encounters among processes in different institutional realms open up possibilities for political change. Stark & Bruszt’s (1998) work on the transition to democracy and market economy in Eastern Europe provides one example of what this looks like in practice. In language that resonates with Orren&Skowronek’s, they argue that in Eastern Europe, “we see social change not as a transition from one order to another but as transformation ”rearrangements, reconfigurations, and recombinations that yield new interweavings of the multiple social logics...”(1998:7). Like Orren & Skowronek, they stress the incongruities among the multiple processes as they unfold: “...within any given country, we find...many[transitions] occurring in different domains” political, economic, and social -and the temporality of these processes is often asynchronous and their articulation seldom harmonious”(Stark & Bruszt 1998:81). Change in one arena affects other ongoing processes, which is what drives institutional evolution.

Pierson’s (1996) analysis of the evolution of social policy in the European Union provides further examples. In one instance (the case of EU policy on gender equality), he shows how provisions adopted by the EU member states in one period-largely symbolic and without much effect-were later picked up by emergent women’s groups, who were able to use these provisions to achieve gains at the EU level that had eluded them at the domestic level. Changes in the political and socioeconomic context brought new actors into the game (in Pierson’s case, women’s groups) who were able to use existing but previously latent institutions (in Pierson’s case, Article 119 of the Treaty of Rome), whose new salience had important implications for political outcomes (Thelen & Steinmo 1992:16).

These examples point to the importance of examining politics as a dynamic process that frequently produces unintended consequences as different, ongoing processes interact. Perspectives that conceive of change as the breakdown of one equilibrium and its replacement with another do not capture this well. Nor, however, do alternative conceptions, for example, some early versions of the new institutionalism in sociology, in which the definition of institutions as “shared cultural scripts” obscures political struggles among competing scripts and/or conceives of change as the displacement of one script by another.

In sum, although historical institutionalists are just as interested as “other” institutionalists in the regularities of politics over time, they tend to emphasize historical process over equilibrium order. Whereas alternative conceptions view institutions in terms of their coordinating functions, historical institutionalists see them as the product of concrete temporal processes. Thus, rather than conceiving of institutions as “holding together” a particular pattern of politics, historical institutionalists are more likely to reverse the causal arrows and argue that institutions emerge from and are sustained by features of the broader political and social context. In this approach to institutions, path dependency involves elements of both continuity and (structured) change; institutions are conceived in relational terms (Immergut 1992, Katznelson 1997:104); and institutional arrangements cannot be understood in isolation from the political and social setting in which they are embedded.

PATH DEPENDENCY

Two ways of thinking about path dependency- one from the literature on economics and technology, the other from the work of “new” institutional sociologists-have gained prominence, and a brief discussion of each provides a baseline for a discussion of the historical institutional approach to path dependency. I argue that both contain insights into the mechanisms that sustain particular patterns of politics, but some of the most prominent formulations tend to obscure the distributional consequences of political institutions and blend out important sources of dynamism in political life.

Technological Models of Path Dependency from Economics

The most widely invoked model of path dependency is the one that comes out of the work of economists seeking to understand technological trajectories. Most closely associated with the “QWERTY keyboard,” the argument developed by David (1985) and elaborated by Arthur (1989) holds that certain technologies, for idiosyncratic and unpredictable reasons, can achieve an initial advantage over alternative technologies and prevail even if in the long run the alternatives would have been more efficient. (Krasner 1988, Kato 1996a, and Pierson 1997 all review these arguments in detail.) What political scientists have taken from this is the intuitively attractive idea that technology, like politics, involves some elements of chance (agency, choice), but once a path is taken, then it can become “locked in,” as all the relevant actors adjust their strategies to accommodate the prevailing pattern.

Some features of politics are undoubtedly subject to the kinds of “positive feedback” effects to which the David/Arthur model of technological change draws attention, and the notion of increasing returns certainly has important applications to politics.18 But as a general guide to understanding political development, the QWERTY model is both too contingent and too deterministic. It is too contingent in that the initial choice (call it a “critical juncture”) is seen as rather open and capable of being “tipped” by small events or chance circumstances, whereas in politics this kind of blank slate is a rarity, to say the least. The openness implied in this model is belied by the vast literature on critical junctures (discussed below) that traces divergent trajectories back to systematic differences either in antecedent conditions or in the timing, sequencing, and interaction of specific political-economic processes, suggesting that not all options are equally viable at any given point in time.

The QWERTY model is also too deterministic in that once the initial choice is made, then the argument becomes mechanical. There is one fork in the road, and after that, the path only narrows. In this model, actors adapt to prevailing institutions by investing in them in ways that reinforce the institutions (e.g. people learn to type in a particular way, firms make products that fit with the standard). In other words, in the world of firms and users of technology, adapting to the standard means adopting it; those who do not adapt lose, And “importantly” the losers disappear (for example, as firms go out of business).

Politics is characterized by disagreement over goals and disparities in power, and in fact institutions often reinforce power disparities (Hall 1986, Knight 1992, Riker 1980:444-45). However, the losers do not necessarily disappear, and their adaptation can mean something very different from embracingand reproducing the institution, as in the technology model. For those whoare disadvantaged by prevailing institutions, adapting may mean biding their time until conditions shift, or it may mean working within the existing framework in pursuit of goals different from “even subversive to” those of the institution’s designers. Such considerations provide insights into the reasons why, in politics, increasing returns do not necessarily result in an irrevocably locked-in equilibrium; further choice points exist.
Path Dependency in Institutional Sociology

Another strong argument about path dependency emerges from the work of the new institutionalists in sociology. Whereas economic models start with individuals or firms in the market, sociological perspectives begin with society. Institutions, in this view, are collective outcomes, but not in the sense of being the product or even the sum of individual interests. Rather, institutions are socially constructed in the sense that they embody shared cultural understandings (“shared cognitions,” “interpretive frames”) of the way the world works (Meyer&Rowen 1991, Scott 1995:33, Zucker 1983:5). Specific organizations come and go, but emergent institutional forms will be “isomorphic” with (i.e. compatible with, resembling, and similar in logic to) existing ones because political actors extract causal designations from the world around them and these cause-and-effect understandings inform their approaches to new problems (DiMaggio&Powell 1991:11, Dobbin 1994). This means that even when policy makers set out to redesign institutions, they are constrained in what they can conceive of by these embedded, cultural constraints.

The strong emphasis on cognition in the new institutionalism in sociology gives us powerful insights into the persistence of particular patterns of politics over time, but as DiMaggio & Powell point out (1991, esp. pp. 1, 11-12), the early formulations (e.g. Meyer and colleagues) were less helpful in understanding change. Some versions of new institutionalism in sociology make it hard to see any forks in the road at all; for example, Zucker (1991:85) argues that

each actor fundamentally perceives and describes social reality by enacting it, and in this way transmitting it to the other actors in the social system....The young are enculturated by the previous generation, while they in turn enculturate the next generation. The grandparents do not have to be present to ensure adequate transmission of this general cultural meaning. Each generation simply believes it is describing objective reality.

The notion of institutions as shared scripts sometimes obscures conflicts among groups (because the scripts are by definition shared), and the notion of isomorphism emphasizes continuity across time and space (because new problems are solved using the same cultural template). Yet we know that dominant cultural norms emerge out of concrete political conflicts, in which different groups fight over which norms will prevail (Katzenstein 1996); we know that dominant policy paradigms can and do shift at times (Hall 1993); we know that organizational fields are often imposed by powerful actors (Fligstein 1991, DiMaggio 1991), and that legitimacy, not automaticity, explains why people follow scripts in the first place (Stinchcombe 1997). This is why recent formulations argue that the cognitive dimensions (though important) should not eclipse the strategic and political elements of action, and frequently find that in questions of institutionalization and institutional change, the political part of the story (and not the cognitive) is more important (DiMaggio & Powell 1991:27,31; see also Katzenstein 1996). At a minimum, then, much work remains to be done to sort out the relationship between the political (decision/ power) and the cognitive (script) aspects of institutional stability and change.

Both the economic-technological and the sociological-institutional perspectives provide strong tools for understanding continuity, but by stipulating and privileging particular mechanisms of reproduction (coordination effects for the former, isomorphism for the latter) they have a hard time incorporating notions of conflict and power, and they are not particularly helpful in talking about change. Dynamism in both models has to come from some exogenous shock, or, as Orren & Skowronek (1994) argue for equilibrium models generally, these perspectives strongly imply that political change is not amenable to the same type of analysis we use to understand the operation of the institutions themselves.

PATH DEPENDENCY IN HISTORICAL INSTITUTIONALISM

Ikenberry captures the essence of a historical institutional approach to path dependency in his characterization of political development as involving “critical junctures and developmental pathways” (1994:16ff). As the phrase implies, this approach includes two related but analytically distinct claims. The first involves arguments about crucial founding moments of institutional formation that send countries along broadly different developmental paths; the second suggests that institutions continue to evolve in response to changing environmental conditions and ongoing political maneuvering but in ways that are constrained by past trajectories. These two lines of argument tend to be reflected in a bifurcation of the literature in this area. A number of important analyses of critical junctures explore the origins of diversity across nations (e.g. Skocpol 1979, Luebbert 1991, Ertman 1997); other studies focus on the logic and self-reinforcing properties of particular national trajectories over time, drawing out comparisons to other countries where relevant (e.g. Weir 1992, Skocpol 1992).

Although obviously related, these two literatures have characteristic strengths and weaknesses, and each would be enriched by a more sustained engagement of the other. The great strength of the critical junctures literature lies in the way in which scholars have incorporated issues of sequencing and timing into the analysis, looking specifically at the different patterns of interaction between ongoing political processes and at the effect of these interactions on institutional and other outcomes. Where this literature has generally been weaker is in specifying the mechanisms that translate critical junctures into lasting political legacies. Here the policy feedback literature, which has provided many insights into the mechanisms that account for continuity over time, is useful. However, in this second literature, strong tools for understanding continuity are not matched by equally sophisticated tools for understanding political and institutional change. In the next three sections of this essay, I argue that greater insight into the different types of reproduction mechanisms behind different institutional arrangements holds the key to understanding what particular kinds of external events and processes are likely to produce political openings that drive (path-dependent) institutional evolution and change.

Historical Institutional Analyses of Critical Junctures

As Katznelson (1997) suggests, the macro-historical analysis of critical junctures that set countries along different developmental paths has long been the bread and butter of historical institutionalism. Rejecting a functionalist view of institutions, historical institutionalists see institutions as enduring legacies of political struggles. The classics in this genre (ably reviewed by others, e.g. Ikenberry 1994) include Moore (1966), Gerschenkron (1962), Lipset & Rokkan (1967), and Shefter (1977). All of these works emphasize sequencing and timing and, related to these issues, different patterns of interaction between ongoing political and economic processes in the formation and evolution of institutional arrangements. These studies are “configurative,” as Katznelson (1997) puts it, in the sense that they do not view political processes in isolation but rather focus specifically on how the temporal ordering of, and interactions among, processes influence outcomes- in these cases, institutional outcomes.

This venerable tradition is alive and well in historical institutional research, as a review of a few select works demonstrates. Collier&Collier’s landmark study, Shaping the Political Arena, links differences in patterns of labor incorporation to variation in party and regime outcomes across a wide range of Latin American countries. As in the earlier works cited above, Collier & Collier emphasize the central importance of sequencing and timing; their study “confront[s] the interaction between a longitudinal and a cross-sectional perspective: between the unfolding over time within each country of phases of political change, and a sequence of international developments that influenced all the countries roughly in the same chronological time, but often at a different point in relation to these internal political phases” (Collier & Collier 1991:19-20). In fact, one of the book’s central themes is the way in which “common” international events or trends translate into different challenges in different countries as a result of their intersections and interactions with ongoing domestic processes (see also Locke & Thelen 1995, Collier 1993).

Another example of critical junctures work in historical institutional research is Ertman’s Birth of the Leviathan, which traces the origins of state institutions across a broad range of European countries from the twelfth to the eighteenth century. Employing a logic that parallels Gerschenkron’s, Ertman argues that “differences in the timing of the onset of sustained geopolitical competition go a long way toward explaining the character of state infrastructures found across the continent at the end of the 18th century” (1997:26). Like Collier&Collier, Ertman attends to variation in the ways in which common international forces intersect with ongoing domestic political developments. Where state-builders faced geopolitical competition early, they were forced into greater concessions to the financiers, merchants, and administrators who financed and staffed the bureaucracy, resulting in patrimonial systems. Where rulers confronted geopolitical pressures later, “they found themselves in a quite different world,” where developments in education and finance made these side payments unnecessary, resulting in greater bureaucratic autonomy (Ertman 1997:28).