Case Studies and Theory Development in the Social Sciences - Alexander L. George [60]
Case study findings can have implications for theory development and testing on three levels. First, they may establish, strengthen, or weaken historical explanations of a case. This is where within-case methods like process-tracing come into play. If a theory posits particular causal mechanisms as an explanation of a particular case, but these prove to be demonstrably absent, then the theory is greatly weakened as an explanation for this case, though there is still the possibility of measurement error or omitted variables.
Yet a modified historical explanation of a case may not add to explanations of other cases that are dissimilar in some respects. Establishing the general applicability of a new or modified explanation of a case requires showing that it accurately explains other cases. Conversely, invalidating an existing theory as an explanation of one case does not necessarily imply that the theory poorly explains other, dissimilar cases; indeed, the existing theory may have earlier demonstrated a strong ability to explain cases.216 Whereas some earlier approaches assumed or demanded that a new theory subsume or explain all of the phenomena explained by its predecessors, we do not require that this always be so. A new theory may be superior in explaining only some of the cases explained by its predecessor, or even only one case, while being inapplicable to others.
Second, and more generally, the finding that a theory does or does not explain a case may be generalized to the type or class of cases (e.g., deterrence) of which this case is a member. Here, the generalization depends on the precision and completeness with which the class of cases has been defined and the degree to which the case exemplifies the class. Generalization to cases not studied always entails some risk of mistaken inferences because they may differ from the case or cases studied in the values of potentially causal variables omitted from the theoretical framework.
Third and most broadly, case study findings may in some circumstances be generalized to neighboring cells in a typology, to the role of a particular variable in dissimilar cases, or even to all cases of a phenomenon. Here overgeneralization is a risk, since the analyst is generalizing cases that differ in the value of variables that have been already identified as causally related to the outcome. This is why case study researchers usually limit themselves to narrow and well-specified contingent generalizations about a type.217 Still, some cases may constitute particularly strong tests of theories, allowing generalization beyond the particular cases studied.
This chapter looks at each of these kinds of generalization, first in theory development and then in theory testing. It concludes that improved historical explanations of individual cases are the foundation for drawing wider implications from case studies, as they are a necessary condition for any generalizations beyond the case. Contingent or typological generalizations are often the most useful kind of theoretical conclusions from case studies, as they build on and go beyond improved historical explanations but present limited risks of extending these conclusions to causally dissimilar cases. Findings that can be extended to different types of cases are less common, and often must be stated as only loose generalizations. However, they can be important turning points in research programs, drawing attention toward avenues for future research.
Theory Development
The development of theory via