Case Studies and Theory Development in the Social Sciences - Alexander L. George [18]
Methodologists are working to reduce this limitation, however. Douglas Dion, for example, has focused on the role of case studies in testing theoretical claims that a variable is a necessary or sufficient condition for a certain outcome.55 Dion convincingly argues that selection bias is not a problem in tests of necessity or sufficiency, that single counterexamples can falsify deterministic claims of necessity or sufficiency (if measurement error can be ruled out), and that only small numbers of cases are required to test even probabilistic claims that a condition is almost always necessary or sufficient for an outcome.56 These factors make case studies a powerful means of assessing claims of necessity or sufficiency.
It is important to distinguish carefully, however, among three kinds of claims of necessity or sufficiency. The most general claim would be that a single variable is necessary or sufficient for an outcome with respect to an entire population of cases. Unfortunately, few nontrivial single-variable relationships of necessity or sufficiency have been found to hold for large populations or wide-scope conditions in the social world. A second kind of claim is that a variable was either necessary or suffi-cient in a particular historical context or case for a specific historical outcome to have occurred. This kind of claim can only be tested counterfactually, and there is no infallible means of making such counterfactual tests.
The third and in our view most useful kind of assertion of necessity or sufficiency concerns the relationship of a variable to conjunctions of variables that are themselves necessary and/or sufficient for an outcome. Consider the following example. Let us assume that the variable A causes Y only in conjunction with B and C. Assume further that the conjunction ABC is sufficient for Y, and the conjunction BC cannot cause Y in the absence of A. In this instance, A is a necessary part of a conjunction that is sufficient for the outcome Y. Many different possible combinations of conjunctive necessity and sufficiency are possible. If equifinality is present, for example, the conjunction ABC itself may not be necessary for the outcome, which might arise through other causal paths that have little or nothing in common with ABC.57
Three caveats are in order regarding inferences of necessity or sufficiency. First, it is often not possible to resolve whether a causal condition identified as contributing to the explanation of a case is a necessary condition for that case, for the type of case that it represents, or for the outcome in general. It is often more appropriate to settle for a defensible claim that the presence of a variable “favors” an outcome, or is what historians often term a “contributing cause,” which may or may not be a necessary condition. When a complex explanation identifies a number of contributing causes, it may be difficult, even with the help of counterfactual analysis, to offer a convincing argument that one condition or another was necessary to the outcome.
Second, whether a factor is necessary to an outcome in a case is a separate issue from how much it contributed to the magnitude of the outcome. One “last straw” may be necessary to break a camel’s back, but it does not contribute as much to the outcome as the bales of straw that preceded it. As noted above, determining such relative causal weights for variables can be difficult to do with any precision in a single case or a small number of cases, but process-tracing evidence and congruence tests