Case Studies and Theory Development in the Social Sciences - Alexander L. George [64]
While theories need to be developed into a testable form, a theory should not be forced into predictions beyond its scope; this leads to the creation of an easily discounted “straw man” version of the theory. A test could also be too tough if countervailing variables mask the causal effects of the variable under study.228 Of course, researchers frequently disagree on whether a theory is being forced to “stick its neck out” sufficiently far, or whether it is being pushed into predictions beyond its rightful scope.229 If an empirical test is beyond the domain of phenomena to which the theory has been applied, then findings inconsistent with the theory limit its scope rather than falsify it.
How can a researcher avoid too readily rejecting or narrowing the scope conditions of a theory that is in fact accurate, or accepting or broadening the scope conditions of a theory that is in fact false or inapplicable? There are no infallible criteria for addressing all of the complications of generalizing the results of a case study’s theory tests. A key consideration, however, is the issue of how tough an empirical test a case poses for a theory: How strongly do the variables predict the case’s outcome, and how unique are the predictions the theory makes for the case?230
TESTING COMPETING EXPLANATIONS OF CASES
An explanation of a case is more convincing if it is more unique, or if the outcome it predicts “could not have been expected from the best rival theory available.”231 If a phenomenon has not previously received wide study, a theory can only make a rather weak claim to being the “best” explanation. For closely studied phenomena, however, the finding that a case fits only one explanatory theory is powerful evidence that the theory best explains the case. Of the five hypotheses considered in the study of burden-sharing in the 1991 Gulf War noted above (balance of threat, alliance dependence, collective action, domestic politics, and policymaking institutions) only the alliance dependence hypothesis fit the outcome and process of the German and Japanese contributions to the coalition. This highlighted the power of alliance dependence, since the variables identified by all the other hypotheses militated against this outcome.
In testing competing historical explanations of a case, then, it is important to find instances where explanations make unique predictions about the process or outcome of the case. An excellent example of this is Scott Sagan’s work on the safety of nuclear weapons from accidental or unauthorized use.232 Sagan treats the safety of nuclear weapons as a subclass of the ability of complex organizations to manage hazardous technology. The latter problem has been addressed in two major theories: Charles Perrow’s normal accidents theory, and the high reliability theory developed by a group of Berkeley scholars.233 Neither of these two organizational theories had addressed the specific problem of nuclear weapons safety, but Sagan argues they each have implications for this issue.
Sagan notes that both theories often make ambiguous predictions.234 Neither theory excludes the possibility of a serious accident, though the normal accident theory is more pessimistic. There is considerable overlap between the two in their predictions on the nuclear weapons cases of interest to Sagan, but he finds the theories to be at odds in several important respects. Sagan notes that “many of the specific conditions that the high reliability theorists argue will promote safety will actually reduce safety according to the normal