Bennett Elman2006

download Bennett Elman2006

of 18

Transcript of Bennett Elman2006

  • 8/13/2019 Bennett Elman2006

    1/18

    Complex Causal Relations and Case Study Methods:The Example of Path Dependence

    Andrew Bennett

    Department of Government, Georgetown University, Washington, DC 20057

    e-mail: [email protected]

    Colin Elman

    Department of Political Science, Arizona State University,

    Box 873902, Tempe, AZ 85287-3902

    e-mail: [email protected] (corresponding author)

    This article discusses the application of qualitative methods in analyzing causal complexity.

    In particular, the essay reviews how process tracing and systematic case comparisons can

    address path-dependent explanations. The article unpacks the concept of path depen-

    dence and its component elements of causal possibility, contingency, closure of alterna-

    tives, and constraints to the current path. The article then reviews four strengths that case

    studies bring to the study of path dependence: offering a detailed and holistic analysis of

    sequences in historical cases, being suitable for the study of rare events, facilitating the

    search for omitted variables that might lie behind contingent events, and allowing for the

    study of interaction effects within one or a few cases.

    Because life is complicated.

    T-shirt epigram for the 2006 Arizona State University Institute

    on Qualitative Research Methods

    1 Introduction

    Methodological choices must take into account the characteristics of the phenomena we

    seek to understand.1 We need to heed Peter Halls warning that as we have sought to

    understand and explain complexity in social and political life, our ontologies have outrunboth our methodologies and standard views of explanation (Hall 2003, 387). Hall argues

    that political scientists have moved toward theories, such as those based on path de-

    pendence or strategic interaction, whose conceptions of the causal structures underlying

    outcomes are at odds with the assumptions required for standard regression techniques and

    conventional comparative method to provide valid causal inferences (Hall 2003, 375).

    Authors note: We thank John Gerring, Gary Goertz, James Mahoney, Paul Pierson, and three anonymousreviewers, for helpful comments on earlier versions of this essay.1Although this implies that ontology should precede methodology, we realize that the relationship is likely to be

    more complicated. Epistemological and methodological choices make it much more likely that scholars willsee the social world in a particular way (see Pierson 2004, 910, citing R. L. Jepperson, unpublished data).

    The Author 2006. Published by Oxford University Press on behalf of the Society for Political Methodology.All rights reserved. For Permissions, please email: [email protected]

    250

    Advance Access publication June 7, 2006 Political Analysis(2006) 14:250267doi:10.1093/pan/mpj020

  • 8/13/2019 Bennett Elman2006

    2/18

    This requires that we adapt and develop our methods, whether formal, statistical, or

    qualitative, to address the kinds of complexity that our theories increasingly entail.

    This essay deals with the methodological challenges that complexity raises for the

    study of politics and the means through which qualitative methods can help address these

    challenges even if they cannot fully resolve them. There are several different phenomena

    that exhibit causal complexity, including tipping points, high-order interaction effects,

    strategic interaction, two-directional causality or feedback loops, equifinality (many differ-

    ent paths to the same outcome), and multifinality (many different outcomes from the same

    value of an independent variable depending on context). The possible presence of these

    kinds of complexity affects how knowledge statements can be most usefully constructed and

    verified. We conclude, as Hall does, that qualitative methods, particularly the combination

    of within-case analysis and cross-case comparisons, are useful approaches toward assessing

    these kinds of complex causation even when scholars study only one or a few cases.

    We do not attempt to review case study methods relevant to all these kinds of com-

    plexity or all the possible combinations of different kinds of complexity. Rather, we focus

    on the contributions that process tracing and systematic case comparisons can make,especially together, to address issues of path dependence. Arthur (1994), an economist

    who helped to develop and popularize the concept of path dependence before its impor-

    tation into political science, suggested that his work was driven by the view that the world

    is messy, organic, and complicated. In the context of path dependence, these complex-

    ities mean that, as Pierson (2004, 189) has argued:

    specific patterns of timing and sequence matter; starting from similar conditions a range of social

    outcomes is often possible; large consequences may result from relatively small or contingent

    events; particular courses of action, once introduced, can be virtually impossible to reverse; and,

    consequently, political development is often punctuated by critical moments or junctures that

    shape the basic contours of social life.

    In the next section of the essay we unpack the concept of path dependence before pro-

    ceeding to discuss how case study methods provide leverage over explaining and under-

    standing this particular characteristic of a complex social world. There are several

    excellent discussions of path dependence in the social sciences (including Thelen 1999;

    Mahoney 2000, 2006; Pierson 2000, 2004; I. Greener 2005; Kay 2005; Page 2006;

    Mahoney and Schensul 2006; Boas forthcoming), and our purpose is not to reproduce

    these efforts.2 Rather we provide a brief overview of some of the central elements of path

    dependence in order to facilitate a discussion of the benefits of using case study methods to

    analyze them. These benefits include the detailed study of particular cases with sensitivity

    to sequencing, the use of process tracing to gain inferential leverage on rare or unique

    events, the opportunity to study cases inductively to help identify omitted variables, andthe ability to study interaction effects in the context of particular cases.

    2 Path Dependence

    The concept of path dependence has been readily imported into the different subfields of

    political science (see, e.g., Krasner 1988; Kato 1996; Ikenberry 1999; Thelen 1999). There

    is, however, substantial disagreement among political scientists on how best to define and

    apply path dependence (I. Greener 2005; Mahoney and Schensul 2006). Part of the dis-

    agreement arises from the different combinations of literatures on which political scien-

    tists have relied. Some have drawn on literature in economics, which focuses on functional

    2See also Aminzade (1992) and Abbott (2001).

    Complex Causal Relations and Case Study Methods 251

  • 8/13/2019 Bennett Elman2006

    3/18

    considerations such as the efficiency of technologies or institutions in achieving public or

    private goods (see David 1985, 2001; Arthur 1989, 1994).3 Scholars have also drawn on

    other sources, including institutional sociology, which places emphasis on how institutions

    interact with the ideas and identities of actors and on how the distributional consequences

    of institutions affect the power and legitimacy of actors hoping to maintain the status quo

    and those seeking change (Thelen 1999, 386, referencing Meyer and Rowan 1991; Powell

    and DiMaggio 199l; Zucker 1991). Some of the variations in how political scientists

    approach path dependence may also be due to subfield divisions in the discipline. As

    Jervis (2000) notes, scholars from different subfields often work in relative isolation from

    each other. As path dependence has been imported and integrated, this tendency is likely to

    have produced some redundancy but also some differences.

    We argue that dissimilarities among political scientists on the concept of path depen-

    dence can be represented in terms of the different content scholars give to, and emphasis

    they place on, four elements common to most accounts: causal possibility, contingency,

    closure, and constraint.

    First, causal possibility suggests that more than one path might have been taken. Overthe length of the entire history, there has to be some space for different possible outcomes.

    That is, for path dependence to be pertinent, there need to be some different feasible

    histories. It would not be applicable if, regardless of how many times or in how many

    ways we started from the beginning, the same outcome would inevitably be reached. For

    most scholars who use path-dependent arguments, this causal possibility occurs earlier

    rather than later in the history.

    Second, contingency implies that the causal story is influenced by a random or un-

    accounted factor. There needs to be some contingent element that intervenes in the causal

    narrative: one or more of the factors that influences the direction of events needs to be

    random or (by some readings) exogenous to the main theory of interest. In most accounts,

    the path that results must be not only contingent but also highly contingent. It may be one

    among many possible outcomes or an unlikely or inefficient outcome compared to one or

    more alternative paths.

    Third, closure connotes that as a result of that influence, some causal paths become less

    possible or impossible. There needs to be some degree of narrowing, a closure of some

    previously feasible paths. As a result, some causal pathways become more and some

    less likely.

    Finally, once a path is selected, there needs to be some degree of constraint, some

    processes that operate to keep actors on it. Constraint suggests that the actors are tied to the

    path that is chosen or would face high costs in moving off this path once it is established.

    Although the same mechanism may be implicated in both closing alternative paths andconstraining actors to the contemporary path, this need not be so. For example, as noted

    below, an institution may be selected over other feasible alternatives for reasons that are

    unrelated to the factors that explain the subsequent longevity of the chosen institution.

    As shown in Fig. 1 (adapted from Mahoney and Schensul forthcoming), the four

    elements can be represented pictorially: a time during which there are a number of

    plausible alternatives (t0t1), followed by a critical juncture, where contingent events

    lead one of these alternatives to emerge (t1t2), after which actors are constrained to

    remain on that path (t2tn).

    3

    See Liebowitz and Margolis (1990, 1995) for skeptical readings of whether (and if so how often) pathdependence conflicts with neoclassical economics. See Mahoney (2006) for the argument that the critique ofLiebowitz and Margolis is less compelling in nonmarket contexts.

    252 Andrew Bennett and Colin Elman

  • 8/13/2019 Bennett Elman2006

    4/18

    We use as an example here Gregory Luebberts theory on the origins of liberal de-

    mocracy, fascism, social democracy, and traditional authoritarianism in interwar Europe.

    In Luebberts view, European countries all faced the challenge of reconciling political

    institutions and coalitions with the new class structures that emerged as a result of in-

    dustrialization. In Britain, France, and Switzerland, middle classes emerged early on and

    established political dominance prior to World War I, allowing a strong liberal-labor

    coalition that forestalled radicalization of the working classes. In late-industrializing

    countries, however, preindustrial cleavages such as urban-rural and religious tensions

    prevented liberal communities from becoming powerful partners of workers undergoing

    mobilization in the 1920s and 1930s. This made middle peasants a key coalition partner in

    struggles among liberal, socialist, and reactionary groups. Socialists, Luebbert argues,were able to make a red-green coalition with middle peasants wherever the agrarian pro-

    letariat had first been mobilized by other groups, leading to social democracy in Norway,

    Sweden, and Denmark. Otherwise, socialists were tempted to mobilize the agrarian

    proletariat themselves for their own more radical ends, thereby alienating the middle

    peasants and leading to a brown green coalition and fascism, as happened in Germany,

    Italy, and Spain. Finally, in cases where liberal groups allied with middle peasants, tradi-

    tional dictatorships, lacking the institutional innovations of fascism, emerged.

    Luebberts argument represents one particular combination of contingency and con-

    straint. In his view, the necessity of reformulating political coalitions in the face of

    industrialization was common to all European states, the possible coalitions that might

    emerge were limited, and the particular preindustrial cleavages that led to one kind of

    coalition or another were highly contingent. Luebbert does not focus on the mechanisms

    Fig. 1 Path dependency: the example of Luebberts theory of regime type.

    Complex Causal Relations and Case Study Methods 253

  • 8/13/2019 Bennett Elman2006

    5/18

    that constrained each of these polities on their selected paths through the interwar years,

    but he seems to share Piersons sentiment that once established, basic outlooks on

    politics, ranging from ideologies to . . .orientations toward political groups or parties, are

    generally tenacious (Pierson 2000, 260). Of course, a subsequent contingent event

    the outcome of World War IIbrought an end to the fascist and authoritarian regimes

    of Western Europe but allowed for continuity in the governance of the liberal and social

    democracies Luebbert studied.

    As noted above, the different elements of path dependence have been the subject of

    considerable disagreement and debate. We revisit below the discussion of two of them:

    contingency and constraint.

    2.1 Contingency and Path Dependence

    The narrowest reading of contingency is that it involves a stochastic process, something

    that is irreducibly unexplainable. Depending on the unfolding of chance events, a different

    path could have been selected.A different reading of contingency is that it requires something out of left field,

    something unexpected in the context of the main theory under investigation. This

    understanding of contingency is included in the broad approach taken by Mahoney

    (2000, 513), who suggests that it is the inability of theory to predict or explain, either

    deterministically or probabilistically, the occurrence of a specific outcome. Mahoney

    argues that:

    In the actual practice of research, social analysts will consider an event to be contingent when its

    explanation appears to fall outside of existing scientific theory. For example, most sociologists will

    treat as contingent both small events that are too specific to be accommodated by prevailing social

    theories, such as the assassination of a political leader or the specific choices and agency of

    particular individuals, and large, seemingly random processes such as natural disasters or suddenmarket fluctuations.

    The phenomenon could be potentially explainable but nevertheless exogenous to the main

    causal story up to that point. One possible example of an exogenous but explainable

    argument is the thesis that the K-T boundary extinction was caused by a meteor (Dessler

    2003, 395). Astronomers have a fairly good understanding of the factors influencing the

    trajectories of heavenly bodies and of the amounts of energy released when they occa-

    sionally collide. Until Luis and Walter Alvarez suggested the connection in 1980, however,

    it is unlikely that these would have been closely followed by paleontologists investigating

    the disappearance of the dinosaurs.

    An alternative category of arguments that are exogenous to the main explanation arethose that cannot currently be addressed from this or other disciplines. This may be simply

    because no theories have yet been developed that cover the gaps. For example, for much of

    the history of the field of evolutionary biology, genetic mutations were treated as exoge-

    nous and contingent events because the instruments and theories necessary to observe and

    understand mutations at the microbiological level did not exist until relatively recently. In

    addition, we may have good theories covering the gaps but either no data or insufficiently

    sophisticated methodologies to handle them. The meteors we cannot yet observe with

    extant instruments are in some sense contingent, even though we could explain their

    trajectories fairly well if we could observe them.

    As an example of contingency arising from methodological limitations, high-order

    interaction effects may make the events that precede the selection of a constrained path

    appear inexplicable. Interaction effects take place when the effect of one independent

    254 Andrew Bennett and Colin Elman

  • 8/13/2019 Bennett Elman2006

    6/18

    variable on the outcome of interest varies depending on the value of one or more other

    independent variables. The forms of such interactions can be as many and varied as causal

    relations in general; they can be linear (i.e., the effect of a given variable increases or

    decreases in a linear fashion as the value of the variable with which it interacts increases)

    or nonlinear. Interactions can also be sideways: the interaction of two variables can

    produce a qualitatively different outcome rather than an increase or decrease of one

    dimension of the outcome of interest (Jervis 1997, 1289). Moreover, interactions

    can involve only two interacting variables or they may involve interactions among a

    configuration of many variables (Ragin [1987] terms the latter multiple conjunctural

    causation). The various kinds of contingency are summarized in Fig. 2.

    The use of contingent events in case study narratives places them at risk of being viewed

    as just-so stories. We should also acknowledge that, to the extent that the relevant con-

    tingencies are truly random and not just current lacunae for known or potential theories and/

    or methods, social sciences confront an irreducible gap in the causal narrative. Even under

    these circumstances, however, case studies can still help maximize our understanding of

    events. By process tracing the causal narrative up through the random contingencies and byshowing how these (albeit unexplainable) events interact with other more tractable parts of

    the account, analysts can clarify which parts of the account are contingent and which are

    explicable and the respective roles of each in subsequent events. Also, even when the

    contingent period remains largely inexplicable, case studies can help identify and test

    the mechanisms that maintain an equilibrium in the constrained period that follows.

    Fig. 2 Types of contingency.

    Complex Causal Relations and Case Study Methods 255

  • 8/13/2019 Bennett Elman2006

    7/18

    Circumstances are somewhat different where contingencies are potentially knowable

    rather than stochastic, even though explanations for them are either exogenous or un-

    developed. There is a much stronger argument in these situations for researchers either

    to trace out the missing links or to justify why they do not. Here case studies can guide

    analysts to look for theories in other fields or to try to develop explanations where no such

    theories are found. On a practical note, we acknowledge that it will often be unclear which

    types of contingency analysts are dealing with. It may only be after the first iteration of the

    case study has been completed that the researcher will have a better view on whether, and

    if so how, to explicate the first, as-yet unexplained development that defines the critical

    juncture in question.4

    Scholars vary in the emphasis they place on contingency as a necessary element of

    path-dependent explanations and on its meaning. Pierson (2004) argues that contingency is

    a possible feature but stops short of including it as a required element for path dependence

    (see Collier and Collier 1991; Thelen 1999, 2003, for similar positions). By contrast,

    Mahoney and Schensul (2006) argue that only by including some form of contingency

    in the definition of path dependency will some of the most interesting and unexpectedphenomena associated with this type of explanation be captured.

    2.2 Constraints and Path Dependence

    The nature of the constraints that keep events to the single path are also understood

    differently by different authors. For example, one reading, adopted by Pierson (2004,

    20), is that the crucial feature of a historical process that generates path dependence is

    positive feedback (or self-reinforcement).5 Each successive step down the path increases

    the likelihood that a particular event or choice will be repeated and/or the magnitude of its

    subsequent manifestations. Positive feedback is often associated with a tipping point,

    where the causal pathway becomes fixed after the causal variable increases past a givenpoint.6 Prominent examples of technological choices that were reinforced by their

    increasingly dominant market share (as opposed to their intrinsic superiority) include

    the VHS (and not Betamax) taping systems, the Intel/Microsoft (and not Motorola/IBM/

    Apple) computer chip-software combinations, and the widespread adoption of the

    QWERTY keyboard (Jervis 1997, 158, 165). Pierson (2004, 2244) suggests that these

    types of explanation for path dependence travel easily to the study of politics. Perhaps the

    best-known example in international relations is the domino effect argument that as

    a state increases its territory and strength; other states are more likely to acquiesce in its

    rise or to be defeated if they do not (Jervis 1997, 16575).7

    For economists, much of what is interesting about path dependence is its ability to help

    explain the persistence of suboptimal or inefficient outcomes like the QWERTY key-

    board.8 In political life, however, the question of efficiency is less clear cut, as distri-

    butional issues loom large and the relevant question is often not whether an institution is

    efficient but for whom it is efficient or beneficial. In addition, although positive returns to

    scale are an important mechanism behind one form of path dependence, we draw upon

    4See Gorges (2001, 141) for a critique of the role of exogenous factors in institutionalist accounts. See also M. J.Greener (2002, 164) on the implications for policy limitations of an analytical approach that requires unexpectedor uncontrollable factors to explain change.

    5For a discussion of some problems with Piersons use of increasing returns see Gains, John, and Stoker(2005, 279).

    6

    On tipping points, see Schelling (1978, 1012); Lohmann (1994); Jervis (1997, 1502).7See also Jervis (1991).8Though see Liebowitz and Margolis (1990, 1995) for a different view of the paradigmatic examples.

    256 Andrew Bennett and Colin Elman

  • 8/13/2019 Bennett Elman2006

    8/18

    a range of political scientists and economists who include a wider set of mechanisms

    behind path dependence. These include negative feedback and positive and negative

    externalities as well as positive feedback. When these mechanisms operate in various

    combinations, the result can be not only one stable long-run path but also cyclical se-

    quences and sequential but nonrepeating sequences (Mahoney terms the last of these

    reactive sequences). Although this group of mechanisms broadens out the notion of

    path dependence, they are nevertheless specific enough to distinguish path dependence

    from overly vague arguments that history matters.

    In the view taken by Kathleen Thelen, for example, the QWERTY case is both too

    contingent and too deterministic to be a general model of path dependence (Thelen 1999,

    385). It is too contingent in that the initial choice is wide open and easily tipped, whereas

    few political processes start with a tabula rasa of open options, and it is too deterministic in

    that once the initial choice is made it is quickly and irrevocably locked-in. In politics, as

    Thelen notes, those who lose one round of a battle often adapt and bide their time instead of

    disappearing in the manner of Betamax. In Thelens (1999, 387) view, the key is to focus not

    only on the mechanisms through which institutions are created but also on the ongoingmechanisms of reproduction that sustain institutions, as the latter provide insights into

    how institutions change and break down after a period of being seemingly locked-in.

    Similarly, Crouch and Farrell (2004) argue that sophisticated versions of path dependence

    eschew deterministic notions of lockin, and they point out that social actors can use past

    or redundant institutions, transfer experiences or ideas from other social units, or tap into the

    capabilities of other agents or networks to break out of established equilibria.

    Thelen (1999, 3929) adds that with regard to political institutions, mechanisms

    involving power and legitimacy as well as functionality become important sources of

    institutional creation and change. New institutions, for example, are reinforced when they

    have distributional consequences that enhance the power of actors interested in maintain-

    ing these institutions. Mahoney notes that the mechanisms behind the creation of institu-

    tions may be different from the mechanisms of its reproduction (Mahoney 2000), and it is

    important to add that one kind of mechanism can be undermined by another. Institutions

    based on power, for example, can be undermined by a drop in the power or legitimacy of

    the actors that maintain them or in a reduction of these institutions functionality in the

    context of other changing institutions. Thus, whether an institution becomes locked-in

    and whether it is impervious or vulnerable to particular kinds of shocks depend on a

    whole constellation of mechanisms that support or undermine the institution and its

    alternatives, rather than just relying upon positive feedback regarding the institution itself.

    Along similar lines, Scott Page, a professor of complex systems, political science, and

    economics, is concerned that the paradigmatic QWERTY case has been both misunder-stood and overextended. Page focuses on issues of functionality and argues that increasing

    returns is only one of several mechanisms that can lead to path dependency. In Pages

    (2006, 24) view, increasing returns are neither necessary nor sufficient to bring about

    path dependency. Page argues that the dominance of the QWERTY keyboard was driven

    not only by its own positive externalities and increasing returns but also by the negative

    externalities that the QWERTY keyboard imposed on the users of alternative keyboards,

    whose typing skills became less valuable as the QWERTY keyboard gained in market

    share (Page 2006, 236). Accordingly, the QWERTY case combines several different,

    but by no means all of the possible, mechanisms by which history can matter. Self-

    reinforcement or the creation of complementary institutions (such as the development

    of gas stations and roadways to complement gas-powered automobiles) can also contribute

    to lock-in.

    Complex Causal Relations and Case Study Methods 257

  • 8/13/2019 Bennett Elman2006

    9/18

    Page identifies and differentiates several different history-dependent processes that are

    not path dependent. Some of these are phat rather than path dependent, in that their

    equilibria depend on what happened in preceding periods but not on the order of those

    events. In other processes, long-run equilibria are not dependent on earlier outcomes or

    their order. The ultimate triumph of the internal combustion engine, for example, was

    probably not dependent on whether the horse and buggy or the steam engine came first or

    whether they existed at all. In some processes, the early moves are critical to subsequent

    equilibria, whereas in others, more recent events are key.

    Of the dynamics Page discusses, one of the most important for political analysis is

    balancing processes, where after a period of increasing returns negative feedback effects

    work to bring a system back to equilibrium. The dynamic here is not the amplification of

    what comes before but reactions against it. One example often mentioned in the interna-

    tional relations literature is the dominance of the nation-state. Spruyt (1994) argues that

    one of the factors that allowed the nation-state to win out over other plausible institutional

    forms, such as city-states, was their comparative advantage in co-opting military power.

    But scholars have also often argued that the persistence of the anarchic Westphalian statesystem necessary for this continuing dominance is due to balancing against attempts at

    hegemony (Vasquez and Elman 2003). That is, the state system is sustained by reactions

    against attempts to overturn it (Jervis 1997, 1314).

    A related dynamic is that a critical juncture may result in cycling between two (or more)

    alternatives. As Page (2006, 13) notes, successes by one constituency may result in the

    mustering of greater political forces by the other. The Roe v. Wade decision by the

    U.S. Supreme Court, for example, was a critical juncture that institutionally locked-in

    a right to abortion but politically it provided mobilizational resources to the opponents of

    abortion, allowing them to successively narrow abortion rights through subsequent legis-

    lation (Rosenberg 1991). Should Roe v. Wade be overturned, it might well mobilize

    proponents of a right to choose, leading to a turn back toward legislative rather than

    judicial protections of abortion rights. Similar long-term policy cycles are evident in

    the symmetrical versus asymmetrical approaches to containment (Gaddis 1982), limited

    war versus overwhelming force military doctrines (Gacek 1994), national versus

    territorial conceptions of sovereignty (Barkin and Cronin 1994), and rehabilitation

    versus criminalization approaches to abuses of drugs and alcohol.

    Finally, some authors allow for the possibility of path dependence without the kinds of

    feedback mentioned above. Mahoney (2000, 5267) differentiates systemic dynamics from

    reactive sequences, a series of closely bound causal links. As Mahoney (2006) notes each

    event in the sequence is both a reaction to antecedent events and a cause of subsequent

    events. Mahoney gives the example of how the assassination of Martin Luther King Jr. ledto the failure of the Poor Peoples Campaign, which in turn contributed to massive riots,

    which heightened welfare militancy and increased Aid to Families with Dependent Children

    (AFDC) applications, bringing about court rulings that led to an explosion in AFDC rolls in

    the 1960s (citing Isaac, Street, and Knapp 1994). Although the links between the events in

    this kind of path dependence must have some special characteristics (otherwise any casual

    story would qualify), it is not yet clear what those features must be. One possibility is that the

    causal links are characterized by a high degree of sufficiency, that is, once the first step on the

    path is taken the final outcome is very likely to happen. This would capture the why actors

    stay on the path part of the story. Another possibility would be to characterize the links as

    necessary conditions, without which the next step in the chain would not have been possible.

    Mahoneys notion of reactive sequences is similar to stage theories common in

    developmental psychology and international political economy, in which individuals or

    258 Andrew Bennett and Colin Elman

  • 8/13/2019 Bennett Elman2006

    10/18

    states pass through stages of development in a set sequence at different rates (Abbott

    1995). The difference is that Mahoney envisions a contingent period that determines which

    of several possible developmental sequences will be followed, whereas many stage theo-ries, such as Marxism, have a more deterministic quality, in which all units converge at

    different rates and to different degrees on one single-staged process.

    The mechanisms and processes of path dependence discussed above can be dia-

    grammed as in Fig. 3.

    3 Value Added from Using Case Studies To Study Path Dependence

    Because path dependence invokes causal possibility, contingency, closure, and constraint,

    case study methods are well suited to analyze these kinds of arguments. Methodological

    choices involve trade-offs among criteria that are individually desirable but that often

    conflict with one another (Gerring 2001). Our argument is not that statistical methods

    are incapable of addressing path dependency or other kinds of complexity. Indeed, several

    statistical methodologists have risen to the ontological challenge Hall outlines by refining

    statistical techniques for addressing issues such as strategic interaction and selection

    effects (Signorino 1998), interaction effects (Brambor, Clark, and Golder 2006), and

    equifinality (Braumoeller 2003). Although our impression is that there are as yet few

    statistical studies of path-dependent processes in political science, we in no way rule

    out statistical methods as one means of investigating such processes.9 Rather, our argu-

    ment is that qualitative methods face different trade-offs from those of statistical methods

    and thus provide some relative advantages in addressing path dependency and high-order

    interactions that may be one source of contingency. Case studies can utilize within-caseanalysis and/or cross-case comparison of the detailed sequential events within one or a few

    cases to provide inferential leverage on complex causation even when only a few relevant

    cases are available for analysis.

    In particular, case studies offer four advantages for the analysis of path dependencies

    and interactions: they allow for detailed and holistic analysis of sequences in historical

    cases, they are suited to the study of rare events, they can facilitate the search for omitted

    variables that might lie behind contingent events, and they allow for the study of interac-

    tion effects within one or a few cases.

    Fig. 3 Mechanisms and types of path dependency.

    9One possible approach for applying statistical analysis to path-dependent processes would be to include a time-

    bound interaction term. A variable X might have a causal impact from time periods 1 to 3, for example, but zeroimpact thereafter. Similarly, sequencing hypotheses suggest that Xt* Zt1is not the same as Xt1* Zt. We thankGary Goertz for these suggestions.

    Complex Causal Relations and Case Study Methods 259

  • 8/13/2019 Bennett Elman2006

    11/18

    Before considering these strengths, however, we should note that, as suggested by the

    term trade-offs, there are costs as well as benefits to using case studies. First, case studies

    are unsuited to drawing strong conclusions about how much an increment in a particular

    cause will affect the outcome (George and Bennett 2005, 25). Second, where there are

    reasons to believe unit homogeneity holds, case studies will be less able than statistical

    methods to generalize with confidence to broader populations. Finally, case study methods

    are weaker than statistical methods at systematically identifying which cases are outliers

    or deviant cases with respect to a specified population.10

    Our list of potential drawbacks is perhaps shorter than some readers might expect. We

    do not, for example, believe that case studies have an inherent degrees of freedom

    problem. Cases may provide a variety of different evidence of the operation of causal

    mechanisms, none of which is directly comparable, some of which may be more important

    than other pieces, and all of which taken together may allow analysts to draw conclusions

    about the (in)adequacy of an explanation (Bennett and Elman 2006). For similar reasons,

    to the extent that case studies rely on process tracing or process observations, small-ncase

    studies are not susceptible to standard selection bias critiques derived from a frequentisttemplate. Selection bias critiques do not apply in the same way to inferences drawn from

    within-case process tracing or causal process observations (Collier and Mahoney 1996).11

    As George and Bennett (2005, 207) note, process tracing is fundamentally different from

    methods that rely on covariation. The methods contribution to causal inference arises

    from its evidence that a process connects the cause and the outcome.12 Because the method

    does not rely on intuitive regression, it is not susceptible to selection bias (Collier,

    Mahoney, and Seawright 2004, 96).

    We now turn to consider the advantages of using case studies to analyze path

    dependence: holistic and detailed analysis, the analysis of rare or singular events, dis-

    covering left-out variables, and investigating high-order interactions.

    3.1 Case Study Methods Offer a Holistic and Detailed Analysis and Help Elucidate

    How Causal Mechanisms Operate in Context

    Perhaps one of the most important contributions of case study methods to the study of path

    dependence is that they allow for both a holistic view of the story and a detailed view of

    events. As Arrow (1994,x) notes in his preface to a volume of Arthurs collected works on

    path dependence, the object of study is a history. Taking a holistic view allows for

    appropriate distinctions to be made between different parts of the story. For example, it

    will often be the case in path-dependent accounts that constraints increase as time goes by.

    This means that events are more likely to shift a sequence off of a path the earlier in thestory they happen. The further into a constrained period, ceteris paribus, the larger a change

    is needed to move off of that particular path (Pierson 2004, 19). By the same token, the

    nature of explanations in open and contingent time periods will differ from those in con-

    strained periods. For example, often the critical juncture period is explicable in terms of

    agent-centered theories, whereas the equilibrium period is more amenable to structural

    explanations. Thus, theorists often differentiate between accounts of how institutions are

    10Note that we expect the second and third concerns to be less pressing here, since path-dependent cases are likelyto have highly individualized trajectories.

    11On process tracing or causal process observations as a distinct form of drawing causal inferences, see Collier,

    Brady, and Seawright (2004, 2525) and George and Bennett (2005, 20532).12Strictly speaking, the process itself is not directly observable. The process and its associated mechanisms arebelieved to exist because of the observable implications of their operations.

    260 Andrew Bennett and Colin Elman

  • 8/13/2019 Bennett Elman2006

    12/18

    created and those of how they are sustained, and case studies can help to identify and

    explain the mechanisms involved in both periods (Mahoney 2000, 5112).

    James Mahoneys (2001) study of the rise of liberal elites to power in Central America in

    the 19th century illustrates these points. In Mahoneys view, liberal elites throughout the

    region sought to weaken the role of the Church and other traditional groups and expand the

    role of the state, but they did so in different ways that had profound and lasting effects. In

    Guatemala and El Salvador, liberals undercut traditional communal landholding and encour-

    aged capitalist commercial agriculture but provided no protection to peasants and small

    producers. This led to agrarian polarization and militarization of the state. In contrast, in

    Costa Rica, liberals worked to preserve small farms even as they commercialized agriculture,

    limiting the polarization of the countryside and the role of the military.

    What makes Mahoneys explanation compelling is the within-case analysis that

    supplements his cross-case comparisons. Mahoney demonstrates in detail, for example,

    how the Guatemalan leader Justo Rufino Barrios bureaucratized the state and profes-

    sionalized the military to keep himself in power while pursuing land reforms that had

    the effect (though not necessarily the intent) of empowering large coffee producers at theexpense of the peasantry. This set in motion mutually reinforcing dynamics in which

    a heavy military presence stimulated rural polarization, and that polarization then

    perpetuated leaders felt need for a coercive military apparatus. By 1890, the military

    oversaw a forced labor market in the countryside and consumed 60% of the coffee-

    dependent state budget (Mahoney 2001, 131). This pattern of politics persisted for nearly

    a century, culminating with a succession of military governments from 1954 to 1986 that

    brutally repressed a rural guerilla movement before finally relinquishing some of the

    militarys power to an elected civilian government. In contrast, Costa Ricas liberal re-

    formers, facing fewer security threats and lacking a large preexisting commercial agricul-

    tural sector, pursued reforms in the 19th century that encouraged small farmers and thus

    did not need a large military sector. A century later, Costa Rica still lacked any institu-

    tionalized military force.

    3.2 Case Study Methods, and Especially Process Tracing, Are Suited To

    Explanations of Rare Events

    There are likely to be only a small number of cases, perhaps only a single case, which

    follows any given particular path. As Goldstone (1998, 843) notes:

    if one is concerned to explain a particular unique event that has occurred only once and then

    perhaps diffused or spread but did not repeat, despite similar initial conditions being found

    elsewhere, then one has most likely identified a path-dependent system in which the uniqueoutcome was produced by some contingent conditions or choices that separated the outcome in

    that particular system from outcomes in other systems that started from similar conditions.

    Because there are only one or a few cases, and many plausible outcomes, there may not be

    sufficient data to allow for inferences drawn from the standard effects-of-causes template

    followed by conventional quantitative approaches (Bennett and Elman 2006). Users of

    quantitative methods commonly direct their investigations to inferring systematically how

    much a cause contributes on average to an outcome within a given population. This is also

    the template that animates the well-known interpretation of qualitative methods of King,

    Keohane, and Verba (1994), which gives priority to identifying causal effects rather than

    causal mechanisms.

    Mainstream qualitative methodologists in political science, by contrast, tend to marry a

    complex view of the social world with a mechanisms and capacities approach to

    Complex Causal Relations and Case Study Methods 261

  • 8/13/2019 Bennett Elman2006

    13/18

    causation. This nexus of commitments results in a coherent and distinct set of methodolog-

    ical choices, nicely captured by the notion of causes-of-effects (Brady 2003; Goertz and

    Mahoney 2006; though cf. King, Keohane, and Verba 1994). Qualitative methodologists do

    not look for the net effect of a cause over a large number of cases but rather for how causes

    interact in the context of a particular case or a few cases to produce an outcome.

    In particular, process tracing (George and Bennett 2005) and causal process observa-

    tions (Collier, Brady, and Seawright 2004, 2525) allow inferences about causal mecha-

    nisms within the confines of a single case or a few cases. Causation is not established

    through small-n comparison alone (what Collier, Mahoney, and Seawright [2004, 945]

    call intuitive regression) but through uncovering traces of a hypothesized casual mecha-

    nism within the context of a historical case or cases. As a consequence, within-case

    methods may provide evidence that bears on multiple testable implications of a theory

    within a single case (George and Bennett 2005, 289; see also Campbell 1975). What is

    indeterminate for an effects-of-causes quantitative design will not necessarily be indeter-

    minate for a causes-of-effects process-tracing research design. A single smoking gun

    piece of evidence may strongly validate one explanation and rule out many others. Con-versely, numerous within-case observations may fail to identify which of two incommen-

    surable explanations is more accurate if there is no evidence on key steps in the

    hypothesized processes on which they differ. In this sense, process tracing can be char-

    acterized as following a Bayesian logic rather than a frequentist logic (Bennett 2006).

    3.3 Case Study Methods, and Especially Process Tracing, Are Suited To

    Discover Left-out Variables

    As noted above, part of the causal sequence in path dependence is contingent occurrences

    that cannot be explained on the basis of prior historical conditions (Mahoney 2000, 5078).

    Causal possibilities may well be closed off by unlikely and hitherto unlooked for factors.

    Open-ended case studies are well suited to uncover these, though as we note above, perhaps

    not explain them. In a sense, the contingency period in a path-dependent argument is

    analogous to a deviant or outlier case, or a case that does not fit expectations or that has

    a large unexplained error term. The contingent outcome could be the result of stochastic

    processes (or measurement error), but a combination of inductive and deductive process

    tracing can explore the possibility that the outcome is instead the result of an omitted

    variable (and/or interaction effects, discussed below). This is analogous to detective work,

    in which the researcher identifies suspects (omitted variables) that might explain the con-

    tingency and clues (observable process-tracing implications of the hypothesized variable)

    and examines the evidence to rule out some variables and to instantiate other variables orinteractions as explanations of the contingent event.

    For example, Thomas Ertman uses process tracing to test and substantiate his argument

    that differences in the timing of the onset of geopolitical competition help explain much of

    the variation in European state institutions. Ertman theorizes that polities that were pushed

    by early (pre-1450) geopolitical competition to undertake state building became locked-in

    to patrimonial institutions, whereas those not pressed by competitors until later periods

    were able to draw upon more advanced models of governance and more educated elites to

    develop professionalized rather than patrimonial bureaucracies. Ertman uses process trac-

    ing to uncover explanations for the anomalous cases that do not fit his framework. In the

    case of England, the early development of strong representative institutions (parliament)

    accompanied the onset of geopolitical competition and counteracted the tendency toward

    proprietary officeholding. In Hungary and Poland, meanwhile, parliamentary bodies allied

    262 Andrew Bennett and Colin Elman

  • 8/13/2019 Bennett Elman2006

    14/18

    themselves with the nobility before foreign competition created a stimulus for moderni-

    zation, enabling them to block professionalization of the state even to the point of making

    these states vulnerable to neighbors with more modernized military organizations (Ertman

    1997, 312).

    Ertman (1997, 33) also uses process-tracing evidence to argue that powerful contin-

    gent events intervened to prevent the cases of Denmark and Sweden from fitting his

    general argument. There is of course always a danger of regressively explaining away such

    anomalies through just-so stories. Ertmans account gains credibility, however, through

    his detailed demonstration that his hypothesized mechanisms were indeed in operation in

    the many states that fit his theory, as well as his ability to identify plausible mechanisms

    that explain his anomalies. In addition, Ertman is careful to distinguish between the

    mechanisms and interactions that he considers generalizable and those which, as in the

    cases of Denmark and Sweden, he considers to be one of a kind.

    Similarly, in his explanation of interwar regimes in Europe, Luebbert combines his case

    comparisons with process tracing of how the general phenomena of class mobilization and

    coalition making played out in individual cases. In the case of Czechoslovakia, for exam-ple, Luebbert notes that a series of crosscutting cleavages catapulted the Communist Party

    to 13% of the vote in 1925. The communists failed to mobilize agrarian workers, however,

    because the middle peasants had already assuaged some of these workers demands with

    land reform. Thus, the working class movement remained urban and had the flexibility to

    collaborate with middle peasants (Luebbert 1991).

    3.4 Case Study Methods Can Help Study Interaction Effects (if any) in

    the Contingent Period

    As noted above, one possible form of contingency is high-order interaction effects. Our

    interest here is in the particular value added that qualitative methods can offer in un-derstanding interaction effects in the context of a path-dependent explanation.

    Qualitative methods address possible interaction effects in a very different way from

    quantitative approaches, resulting in a different set of methodological and theoretical

    trade-offs. As Goertz and Mahoney (2006) note, qualitative researchers seek explanations

    of individual cases and of potentially recurring types of cases or configurations of variables

    (see also Ragin 1987). They often proceed from a working assumption that homogeneity

    holds only for particular configurations of variables rather than for broad populations and

    that multivariate interaction effects are common and important.

    The inductive use of process tracing is particularly effective when a researcher antici-

    pates that the potential interaction terms would be difficult or impossible to theorize upon

    deductively before developing intimate knowledge of a critical juncture. In such instances,

    the researcher can begin with a preliminary understanding of the possible causes that

    selected one path over another, but as the researcher proceeds, whether through general

    soaking and poking in secondary accounts or primary research through interviews and

    archival work, she/he should work to be attentive to variables or interactions that she/he

    had failed to anticipate.13

    13Other qualitative methods for exploring interaction effects, though less relevant to interactions in the contingentperiod of path-dependent arguments, include deductive process tracing and typological theorizing (Elman 2005;George and Bennett 2005). Like all methods of causal inference, these tools are imperfect and open to possibleinferential errors, but they offer a different set of methodological trade-offs from that involved in statistical

    analysis. Generally, this combination of qualitative methods for addressing interaction effects gives up consider-able parsimony relative to statistical models, but it reduces the risks of omitting relevant variables or interactionterms.

    Complex Causal Relations and Case Study Methods 263

  • 8/13/2019 Bennett Elman2006

    15/18

    For example, Gosta Esping-Andersen used descriptive data and several OLS regressions

    (without interaction terms) to identify three types of welfare capitalism: the liberal,

    conservative, and social democratic regime types. These types represent different com-

    binations of social stratification, employment policies, and de-commodification strategies

    allowing workers a standard of living independent of their work. Esping-Andersen

    hypothesizes that the type of welfare state that emerges in a particular case is a function

    of the nature of class mobilization, class-political coalition structures, and the historical

    legacy of regime institutionalization. To examine the interactions among these factors,

    Esping-Andersen used case studies of the United States, Sweden, and Germany. He finds

    that in Sweden a red-green coalition of workers and farmers put in place generous

    benefits for workers and subsidies for farmers, whereas in the United States a similar

    coalition was less powerful because the labor-intensive South blocked more generous

    welfare-state benefits. In Germany, labor-intensive agriculture and conservative social

    and religious forces prevented the emergence of a red-green coalition, leading to highly

    stratified and only modestly redistributive welfare policies (Esping-Andersen 1990). It is

    hard to imagine that Esping-Andersen could have deductively theorized a priori all theparticular variables and interactions that constituted the contingencies leading to these

    path-dependent outcomes.

    In sum, as case study methods allow inferences on complex events and interactions

    even when only one or a few cases exist, they are useful tools for analyzing path de-

    pendency. This is particularly true for the contingent period in path-dependency argu-

    ments, as this period is by definition difficult to explain in terms of established theories.

    Case studies can also be valuable in providing a clearer understanding of the causal

    mechanisms that lie behind the creation and reproduction of institutions and of the

    interactions among these mechanisms that lead to either the lock-in or breakdown of

    these institutions.

    4 Conclusion

    Complex causal relations are difficult to study with traditional statistical and qualitative

    methods. We have focused in this essay on the contributions that case study methods can

    make to the analysis of complex causal relationships, building on Halls assertion that

    small-nresearch methods can help narrow the gap between our ontologies and our meth-

    odologies (Hall 2003, 375). But these are observations of a work in progress, and we

    anticipate that over time both qualitative and quantitative methodologists will further

    develop tools for addressing causal complexity.There are many kinds of potential complexity in social life, including path depen-

    dence, interaction effects, tipping points, strategic interaction, two-directional causality

    or feedback loops, equifinality, and multifinality. We have focused here on case study

    methods for addressing path dependence, arguing that process tracing and detailed

    comparisons of a small number of cases, especially when used together, can help to

    unravel these kinds of complexity. The challenge of developing and testing theories

    about complex phenomena is still more daunting when various combinations of kinds

    of complexity may be present, but even here case study methods, because they allow for

    detailed study of individual cases, can be helpful. As the social sciences attempt to

    grapple with ever more sophisticated forms of complexity, both statistical and qualitative

    methodologists will need to continue innovating so that our methodologies do not fall

    behind our ontologies.

    264 Andrew Bennett and Colin Elman

  • 8/13/2019 Bennett Elman2006

    16/18

    References

    Abbott, Andrew. 1995. Sequence analysis: New methods for old ideas. Annual Review of Sociology 21:93113.

    . 2001. Time matters: On theory and method. Chicago: University of Chicago Press.

    Aminzade, Ronald. 1992. Historical sociology and time. Sociological Methods and Research 20:45680.

    Arthur, W. Brian. 1989. Competing technologies and lock-in by historical small events. The Economic Journal

    99:11631.

    . 1994. Increasing returns and path dependency in the economy. Ann Arbor: University of Michigan

    Press.

    Arrow, Kenneth J. 1994. Foreword to Increasing Returns and Path Dependence in the Economy by W. Brian

    Arthur. Ann Arbor: University of Michigan Press.

    Barkin, Samuel, and Bruce Cronin. 1994. Changing norms and the rules of sovereignty. International Organi-

    zation48:10730.

    Bennett, Andrew. 2006. Stirring the frequentist pot with a dash of Bayes. Political Analysis 10.1093/pan/

    mpj011.

    Bennett, Andrew, and Colin Elman. 2006. Qualitative research: Recent developments in case study methods.

    Annual Review of Political Science,pp. 45576.

    Boas, Taylor C. Forthcoming. Conceptualizing Continuity and Change: The Composite-Standard Model of Path

    Dependence.Journal of Theoretical Politics.Brady, Henry E. 2003. Models of causal inference: Going beyond the Neymen-Rubin-Holland theory. Paper

    presented at the annual meeting of the Midwest Political Science Association, Chicago, IL.

    Brambor, Thomas, William Roberts Clark, and Matt Golder. 2006. Understanding interaction models: Improving

    empirical analyses. Political Analysis 14:6382.

    Braumoeller, Bear. 2003. Causal complexity and the study of politics. Political Analysis 11:20933.

    Campbell, Donald. 1975. Degrees of freedom and the case study. Comparative Political Studies 8:17893.

    Collier, David, Henry E. Brady, and Jason Seawright. 2004. Sources of leverage in causal inference: Toward an

    alternative view of methodology. In Rethinking social inquiry: Diverse tools, shared standards, ed. Henry E.

    Brady and David Collier, 22971. Lanham, MD: Rowman and Littlefield.

    Collier, David A., and Ruth Collier. 1991. Shaping the political arena: Critical junctures, the labor movement,

    and regime dynamics in Latin America. Princeton, NJ: Princeton University Press.

    Collier, David, and James Mahoney. 1996. Insights and pitfalls: Selection bias in qualitative research. World

    Politics 49:5691.

    Collier, David, James Mahoney, and Jason Seawright. 2004. Claiming too much: Warnings about selection bias.

    InRethinking social inquiry: Diverse tools, shared standards, ed. Henry E. Brady and David Collier, 85102.

    Lanham, MD: Rowman and Littlefield.

    Crouch, Colin, and Henry Farrell. 2004. Breaking the path of institutional development: Alternatives to the new

    determinism in political economy. Rationality and Society 16:543.

    David, Paul A. 1985. Clio and the economics of QWERTY. American Economic Review 75:3327.

    . 2001. Path dependence, its critics, and the quest for historical economics. In Evolution and

    path dependence in economic ideas: Past and present, ed. Pierre Garrouste and Stavros Ioannides, 1540.

    Northampton, MA: Edward Elgar.

    Dessler, David. 2003. Explanation and scientific progress. In Progress in international relations theory:

    Appraising the field, ed. Colin Elman and Miriam Fendius Elman, 381404. Cambridge: MIT Press.

    Elman, Colin. 2005. Explanatory typologies in qualitative studies of international politics. InternationalOrganization 59:293326.

    Ertmann, Thomas. 1997. Birth of the Leviathan: Building states and regimes in medieval and early modern

    Europe. Cambridge: Cambridge University Press.

    Esping-Andersen, Gosta. 1990. The three worlds of welfare capitalism. Cambridge: Polity Press.

    Gacek, Christopher. 1994.The logic of force: The dilemma of limited war in American foreign policy. New York:

    Columbia University Press.

    Gaddis, John Lewis. 1982.Strategies of containment: A critical appraisal of postwar American national security

    policy. New York: Oxford University Press.

    Gains, Francesca, Peter C. John, and Gerry Stoker. 2005. Path dependency and the reform of English local

    government.Public Administration 83:2545.

    George, Alexander L., and Andrew Bennett. 2005. Case studies and theory development in the social sciences.

    Cambridge, MA: MIT Press.Gerring, John. 2001. Social science methodology: A criterial framework. Cambridge: Cambridge University

    Press.

    Complex Causal Relations and Case Study Methods 265

  • 8/13/2019 Bennett Elman2006

    17/18

    Goertz, Gary, and James Mahoney. 2006. A tale of two cultures: Contrasting quantitative and qualitative

    research.Political Analysis 10.1093/pan/mpj017.

    Goldstone, Jack A. 1998. Initial conditions, general laws, path dependence, and explanation in historical

    sociology. The American Journal of Sociology 104:82945.

    Gorges, Michael J. 2001. New institutionalist explanations for institutional change: A note of caution. Politics

    21:13745.Greener, Ian. 2005. The potential of path dependence in political studies. Politics 25:6272.

    Greener, Michael J. 2002. Understanding NHS reform: The policy-transfer, social learning and path-dependency

    perspectives.Governance: An International Journal of Policy, Administration and Institutions 15:16183.

    Hall, Peter. 2003. Aligning ontology and methodology in comparative research. In Comparative-historical

    analysis in the social sciences, ed. James Mahoney and Dietrich Rueschemeyer, 373404. Cambridge:

    Cambridge University Press.

    Ikenberry, John. 1999. Institutions, strategic restraint, and the persistence of American postwar order. Interna-

    tional Security 23:4378.

    Isaac, Larry W, Debra A. Street, and Stan J. Knapp. 1994. Analyzing historical contingency with formal

    methods: The case of the relief explosion and 1968. Sociological Methods and Research 23:11441.

    Jervis, Robert. 1991. Domino beliefs and strategic behavior. InDominoes and bandwagons: Strategic beliefs and

    great power competition in the Eurasian rimland, ed. Robert Jervis and Jack Snyder, 2050. New York:

    Oxford University Press.. 1997. System effects: Complexity in political and social life . Princeton, NJ: Princeton University Press.

    . 2000. Timing and interaction in politics: A comment on Pierson. Studies in American Political

    Development14:1.

    Kato, Junko. 1996. Review article: Institutions and rationality in politics: Three varieties of neo-institutionalists.

    British Journal of Political Science 26:55382.

    Kay, Adrian. 2005. A critique of the use of path dependency in policy studies. Public Administration83:55371.

    King, Gary, Robert O. Keohane, and Sidney Verba. 1994. Designing social inquiry: Scientific inference in

    qualitative research. Princeton, NJ: Princeton University Press.

    Krasner, Stephen D. 1988. Sovereignty: An institutional perspective. Comparative Political Studies 21:6694.

    Liebowitz, S. J., and Stephen Margolis. 1990. The fable of the keys. Journal of Law and Economics 33:125.

    . 1995. Path dependence, lock-in and history. Journal of Law, Economics and Organization 11:

    20526.Lohmann, Suzanne. 1994. Dynamics of informational cascades: The Monday demonstrations in Leipzig, East

    Germany, 19891991. World Politics 47:42101.

    Luebbert, Gregory. 1991. Liberalism, fascism, or social democracy: Social classes and the political origins of

    regimes in interwar Europe. New York: Oxford University Press.

    Mahoney, James. 2000. Path dependence in historical sociology. Theory and Society 29:50748.

    . 2001.The legacies of liberalism: Path dependence and political regimes in Central America. Baltimore,

    MD: Johns Hopkins University Press.

    . 2006. Analyzing path dependence: Lessons from the social sciences. In Understanding

    change: Models, methodologies, and metaphors, ed. A. Wimmer and R. Kossler, 12939. Basingstoke,

    UK: Palgrave Macmillan.

    Mahoney, James, and Daniel Schensul. 2006. Historical context and path dependence. In Oxford handbook of

    contextual political analysis, ed. R. E. Goodin and C. Tilly, 45471. Oxford, UK: Oxford University Press.

    Meyer, John W., and Brian Rowan. 1991. Institutionalized organizations: Formal structure as myth andceremony. In The new institutionalism in organizational analysis, ed. W. Powell and P. DiMaggio, 4162.

    Chicago: University of Chicago Press.

    Page, Scott E. 2006. Path dependence. Quarterly Journal of Political Science 1:87115.

    Pierson, Paul. 2000. Increasing returns, path dependence, and the study of politics. American Political Science

    Review94:25167.

    . 2004. Politics in time: History, institutions, and social analysis . Princeton, NJ: Princeton University

    Press.

    Powell, Walter W., and Paul J. DiMaggio, eds. 1991. The new institutionalism in organizational analysis.

    Chicago: University of Chicago Press.

    Ragin, Charles. 1987. The comparative method: Moving beyond qualitative and quantitative strategies.

    Berkeley: University of California Press.

    Rosenberg, Gerald N. 1991. The hollow hope: Can courts bring about social change? Chicago: The Universityof Chicago Press.

    Schelling, Thomas. 1978. Micromotives and macrobehavior. New York: Norton.

    266 Andrew Bennett and Colin Elman

  • 8/13/2019 Bennett Elman2006

    18/18

    Signorino, Curtis. 1998. Strategic interaction and the statistical analysis of international conflict. American

    Political Science Review 93:27997.

    Spruyt, Hendrik. 1994. The sovereign state and its competitors: An analysis of systems change. Princeton, NJ:

    Princeton University Press.

    Thelen, Kathleen. 1999. Historical institutionalism in comparative politics. Annual Review of Political Science

    2:369404.

    . 2003. How institutions evolve: Insights from comparative-historical analysis. In Comparative-historical

    analysis in the social sciences, ed. James Mahoney and Dietrich Rueschemeyer, 20840. Cambridge:

    Cambridge University Press.

    Vasquez, John A., and Colin Elman, eds. 2003. Realism and the balancing of power: A new debate . Upper

    Saddle River, NJ: Prentice Hall.

    Zucker, Lynne G. 1991. The role of institutionalization in cultural persistence. In The new institutionalism

    in organizational analysis, ed. Walter W. Powell and Paul J. DiMaggio, 83107. Chicago: University of

    Chicago Press.

    Complex Causal Relations and Case Study Methods 267