CASE STUDY METHODOLOGY

Given the proliferation of scholarship on case study methodology (Brady and Collier, 2004; George and Bennett, 2005; Zartman, 2005; Goertz, 2006; Bennett andElman, 2006; Gerring, 2007; Blatter and Blume, 2008), a complete review is not necessary here.⁴ I do not deal with the question of the definition of case study and case study method, other than to say that the increasing emphasis on theory has led to the widespread shift away from the definition of a case as a temporally and spatially bounded series of events, to the conception of a case as an instance of something else, of a theoretically defined class of events.

The theoretical orientation of contemporary case study research leads to discussions of the various roles of case studies in theory development, and to case selection criteria that advance those theoretical aims. I focus on these issues in the next two sections.

Types of case studies

One can identify several types of case studies, and also a variety of ways of classifying them. Let me suggest the following classification, which is based on the theoretical purpose or function of case studies. It represents a combination of categories proposed by Lijphart (1971) and Eckstein (1975), which have been quite influential.⁵ These are ideal types, and in practice many case studies combine several of these aims

(1) Atheoretical/idiographic case studies are descriptive studies of individual cases that aim to understand and interpret a single case as an end in itself rather than to develop broader theoretical generalizations.

(2) Theory-guided/idiographic case studies also aim to explain and/or interpret a single case, but that interpretation is explicitly structured by a theory or well-developed conceptual framework that focuses attention on some theoretically specified aspects of reality and neglects others.⁶ This is analytic history rather than total history.

As the social sciences have shifted in a more theoretical direction in the last few decades, theory-guided explanations of individual cases - what Van Evera (1997) calls “case explaining” case studies - have replaced atheoretical explanations of individual cases. Even more common and more influential, however, are case studies that are explicitly designed to contribute to the construction and testing of theoretical generalizations about a broader class of behavior.

(3) Hypothesis generating case studies have more explicit theoretical purposes. They examine a particular case or perhaps several cases for the purpose of developing more general theoretical propositions, which can then be tested through other methods, including large-n methods. By permitting an intensive examination of individual historical episodes, case studies can contribute to theory development by suggesting additional causal variables, causal mechanisms, and interaction effects. They also help to sug­gest important contextual variables, thus to identify the scope conditions under which particular theories are valid. These are all important steps in the theory-building process.

The role of case studies in generating hypotheses, and particularly in refining and sharpening them, is enhanced by the close interaction of theory and data in case study analysis (and in some other forms of inquiry as well).

A good example is George and Smoke's (1974) analysis of deterrence in American foreign policy, which is organized sequen­tially in terms of theory specification, appli­cation of the theory to historical cases, and reformulation of the theory based on the cases. Another example, involving multiple authors, is Zartman's (1995) collection of comparative case studies on the ending of civil wars. The theoretical lessons generated by the study were the product of an ongoing dialogue between the conceptual framework guiding the project and evidence from specific cases.

(4) Deviant case studies focus on empirical anomalies in established theoretical gener­alizations in order to explain those anoma­lies and in the process refine the existing theory that failed to predict the anomaly, by identifying omitted variables, interaction effects, or alternative causal paths, or by specifying the scope conditions under which a particular theory is valid. Thus, deviant cases aid in the hypothesis generating func­tion of case studies. A deviant case strategy can be usefully combined with statistical methods, in that some of the most significant deviations from the regression line in a statistical analysis are ideal cases for selection for more thorough examination by case studies.

(5) Theory-testing case studies. In addition to their essential role in the explanation of individual historical episodes and a contribu­tory role in the generation of hypotheses, case studies can also be used to test hypotheses and theories.

As early work in case study methodology acknowledged (Campbell, 1975; Lijphart, 1971), a potentially serious problem con­fronting the theory-testing role of case studies is the combination of many variables and a relatively small number of cases - the low N/V ratio. This often makes it difficult if not impossible to attribute changes in the dependent variable to changes in the theoretically specified independent variables and not to the effects of extraneous variables. The problem of how to make causal inferences in small-n research, when the number of variables generally exceeds the number of cases, is a central question in case study methodology.

The comparative method⁹ is often defined as a strategy for conducting research of natu­rally occurring phenomena in a way that aims to control for potential confounding variables through careful case selection and matching rather than through experimental manipula­tion or partial correlations (Frendreis, 1983: 255). The problem, of course, is finding sufficiently matched cases to justify this inference. I return to this issue, along with a discussion of alternative research designs for theory testing in small-n research, in the next section, after first completing this typology of case studies.

(6) Plausibility probes. Many qualita­tive methodologists have followed Eckstein (1975) in suggesting plausibility probes as a distinct category of case studies. The aim is nomothetic, since presumably what is being probed is the match between the details of a particular case and some broader theoretical proposition.

I have suggested several different objec­tives of case studies. It is important to be explicit about the specific objective of a specific case study, because different theoretical (or descriptive) purposes require different kinds of case study research designs. If the aim is the idiographic one of explain­ing a particular case, whether through an inductive analysis or an analysis driven by an explicit theoretical framework, then case selection becomes somewhat less critical, since theoretical imagination becomes more important than establishing scientific control over extraneous variables. A hypothesis­testing case study, however, has a different set of requirements, including a carefully matched set of cases to maximize control over extraneous variables.

Case study designs for theory testing

Variations on Mill's methods

The comparable-cases strategy is closely related to John Stuart Mill's (1875/1970) method of difference, which focuses on cases that have different values on the dependent variable and similar values on all but one of the possible causal variables. Since the values of alternative causal variables are constant, they cannot explain the variation in the dependent variable and hence can be eliminated as sources of causation, leaving the one independent variable that co-varies with the dependent variable.

Przeworski and Tuene's (1970) description of “most different” and “most similar” systems designs (see also Meckstroth, 1975) follows similar inferential logic. A most dif­ferent systems design identifies cases that are different on a wide range of explanatory vari­ables but similar on the dependent variable, while a most similar systems design identifies cases that are similar on a wide range of explanatory variables but different on the value of the dependent variable.¹⁰

As many critics have pointed out, a potentially serious problem in the application of Mill's methods and of most similar and most different systems designs is the difficulty of identifying cases that are truly comparable - identical or different in all respects but one. The less perfectly matched a set of cases, the weaker the causal inferences that can be drawn. Even if well-matched cases can be found, however, another problem remains. Mill's methods work fine for bivari­ate hypotheses involving a single explanatory variable, if the researcher can find matched cases and if she can assume that measurement error is low. Those methods work well enough for additive models in which there are multiple causes but no interaction effects, though the number of cases required to incorporate sufficient controls may become cumbersome. Mill's methods are much more problematic, however, in situations involving complex causation involving interaction effects, and particularly if there are several different sets of conditions that may lead to the same outcome (Ragin, 1987; Lieberson, 1992).

It is useful to distinguish between cross­case designs and within case designs, and between longitudinal designs and various forms of cross-sectional designs. Longitudi­nal, within-case comparisons of hypothesized relationships at different points in time within the same case are particularly powerful. They are essentially “most similar” systems designs because the dependent variable of interest changes over time; hypothesized explanatory variables change, but many other variables are constant, including political history, culture, institutions, geography, and other variables that change only slowly (if at all) over time. This facilitates the identification of the small number of variables that vary with the dependent variable of interest.

Good examples of such longitudinal desi­gns include Touval's (1982) study of nine attempts at mediation in the Arab-Israeli conflict and Stedman's (1991) comparative study of mediation in the war in Zimbabwe. One can also combine cross-case and within- case designs, in order to impose more controls. Snyder's (1991) study of imperial overex­tension, for example, combines comparisons among the behaviors of different states, different individuals within the same state but in different bureaucratic roles, and the same individuals within a given state over time.

Process tracing

Mill's methods and related varieties of matching case selection strategies are basi­cally correlational, and examine whether a particular set of conditions is associated with hypothesized outcomes while holding constant as many other factors as possible. George and Bennett (2005), building on George (1979), refer to within-case com­parisons of hypothesized relationships at different points in time within the same case as the “congruence method” and include it within the methodology of structured, focused comparison.

Another approach to within-case analysis, one that is quite common in the practice of case study research but that was often neglected in early efforts to formally describe case study methodology, is process-tracing (George, 1979; George and Bennett, 2005). Process tracing follows a different logic than correlational methods. It tries to uncover the intervening causal mechanisms between conditions and outcomes through an intensive analysis of the development of a sequence of events over time.¹¹ Process tracing is particularly useful for exploring the percep­tions, expectations, and political interactions of actors inside the “black box” of decision- making.¹²

Process tracing can be useful for several different theoretical tasks. It is often essential for a complete description and explanation of a particular historical episode. It can also contribute to hypothesis construction. Many propositions about bureaucratic poli­tics, for example, originate in Allison's (1971) intensive study of the Cuban missile crisis. Process tracing can also contribute to the testing of certain theoretical propositions. One of the implications of the democratic peace proposition, for example, is that poli­tical leaders differ in their perceptions of democracies and autocracies, and that these differences have a significant impact on behavior. Such perceptions are usually better explored through small-n case study methods than through large-n statistical methods. To take another example, validations of prospect theory propositions on loss aversion and risk behavior, which are central to conflict resolution behavior, require the identification of actors' reference points. Case studies provide the most efficient way of determining reference points.¹³

Crucial case studies¹

Eckstein (1975: 113-23) suggested that crucial case studies can be useful for the purposes of testing certain types of theo­retical arguments - where predictions were precise and where measurement error was low. Two important types of crucial case studies are most-likely or least-likely case research designs. Each implicitly adopts a Bayesian perspective, and basically weights the evidence from a particular case as a function of prior theoretical expectations (McKeown, 1999). If one's theoretical priors suggest that a particular case is quite unlikely to be consistent with a theory's predictions - either because the theory's assumptions and scope conditions are not fully satisfied or because the values of many of the theory's key variables point in the other direction - and if the data supports the hypothesis, the evidence from the case provides a great deal of leverage for increasing our confidence in the validity of the theory. Similarly, if one's theoretical priors suggest that a theory is highly likely to be confirmed, and if the data do not support the theory, that result can be quite damaging to a theory. The logic of least-likely case design is based on the “Sinatra inference” - if I can make it there, I can make it anywhere. Similarly, the logic of most-likely case design is based on the inverse Sinatra inference - if I cannot make it there, I cannot make it anywhere (Levy, 2002: 442).

Allison's (1971) application of his three models of foreign policy decision-making to the Cuban missile crisis provides a good example, though he did not explicitly use the language of most-least-likely case analysis. Allison argued that the conventional wisdom in foreign policy analysis held that politics stopped “at the water's edge,” particularly in acute international crises involving serious and immediate threats to vital national interests. As perhaps the most extreme threat to US national security interests, the Cuban missile crisis was a most-likely case for the “rational unity actor model” of foreign policy decision-making, and simultaneously a least-likely case for alternative organizational process and governmental politics models of decision-making. By showing that the evidence contradicted many predictions of the rational unitary actor model but fit many of the predictions of the organizational process and governmental process models, Allison (1971) made a strong argument for the limitations of Model I and the potential applicability of Model II and Model III.¹⁵

This discussion of most- and least-likely case study research designs, in conjunction with our earlier discussion of the role of case studies in testing hypotheses positing necessary or sufficient conditions, demon­strates that a small number of case studies, and possibly even a single case study, can be quite valuable for the purposes of testing certain types of theoretical propositions, if the theory takes a particular form, and if cases are selected in a way that maximizes leverage on the theory (Dion, 1998). A hypothesis positing a necessary condition for a particular outcome can be severely damaged by a single case in which the condition is absent but the outcome nonetheless occurs, and a hypothesis positing a sufficient condition for a particular outcome can be severely damaged by a single case in which the condition is present but the hypothesized outcome is absent. In addition, a theory can be severely damaged if it can be demonstrated that the theory is invalid in a case where theoretical expectations lead us to think it is an easy case for the theory.

Having briefly surveyed the literature on types of case studies and types of research designs for hypothesis testing, let me turn to the substantive questions of crisis man­agement, a subset of conflict resolution, and explore the contributions of case studies to the development and testing of a theory of crisis management.