Case Studies and Theory Development in the Social Sciences - Alexander L. George [114]
ASSESSING THE CAUSAL POWER OF AN INDEPENDENT VARIABLE
Most case studies are outcome-oriented; they focus on explaining variance in the dependent variable. But when researchers or policymakers wish to assess the causal power of a particular factor—such as an independent variable that policymakers can manipulate—they have an interest in exploring the contingent conditions under which similarity or variance in the independent variable leads to different outcomes.438 Research on the strategy of coercive diplomacy, for example, treats it as an independent variable and develops a typology of such strategies to investigate variations in outcome of these strategies.
We differ with many methodologists in that we argue that a theory can be derived or modified based on the evidence within a case, and still be tested against new facts or new evidence within the same case, as well as against other cases. Detectives do this all the time—clues lead them to develop a new theory about a case, which leads them to expect some evidence that in the absence of the new theory would have been wildly unexpected, and the corroboration of this evidence is seen as strong confirmation of the theory.
This process relies on Bayesian logic—the more unique and unexpected the new evidence, the greater its corroborative power. For example, in The Limits of Safety, Scott Sagan made process-tracing predictions on particular kinds of evidence regarding nuclear accidents that would be true if his theory were true, but that would have been highly unlikely if the alternative explanations were true.439 Another example comes from research on schizophrenia. When researchers looking at brain chemistry proposed a chemical mechanism that might help explain schizophrenia, they unexpectedly found that this same chemical mechanism was involved in the brain’s reaction to the inhalation of cigarette smoke. The proposed mechanism thus appeared to explain the long-known but unexplained fact that some schizophrenics tend to be chain-smokers. In other words, schizophrenics may have unconsciously been using chain-smoking to ameliorate the brain chemistry abnormalities that caused their schizophrenia. As the researchers were not looking for or expecting an explanation of schizophrenic’s chain-smoking, this finding is a heuristically independent confirmation. Although the study involved many schizophrenics, the logic of this kind of confirmation does not derive from sample size and it applies in single cases of the kind that historians often investigate.440
VALIDITY OF CONCLUSIONS BASED ON SINGLE CASE STUDIES
Some political scientists argue that causal explanation requires case comparisons and that single-case studies have limited uses in theory building. James Lee Ray, for example, has argued that causal linkages cannot be identified within the context of one case.441 Similarly, the authors of Designing Social Inquiry (DSI) argue that the single observation is not a useful technique for testing hypotheses or theories unless it can be compared to other observations by other researchers. They add that single cases cannot exclude alternative theories, and that their findings are limited by the possibility of measurement error, probabilistic causal mechanisms, and omitted variables.442
Indeed, the conclusions of single case studies are much stronger if they can be compared to other studies, but we suspect that most historians would join us in arguing that the limitations attributed to single case studies are not categorical. As DSI acknowledges, its view of the limits of single case studies is based in part on its definition of a case having only one observation on the dependent variable, and it notes that “since one case may actually contain many potential observations, pessimism is actually unjustified.” Thus, while process-tracing may not be able to exclude all but one of