Case Studies and Theory Development in the Social Sciences - Alexander L. George [121]
One may recognize that in principle any historical explanation implies a counterfactual in the sense that the historical outcome would not have occurred had the causal variables adduced in support of the explanation been different. Such a counterfactual can be said to serve the purpose of a second case and, if so, the real and counterfactual cases together might constitute a controlled comparison. However, such a claim rests on the supposition that the causal variable in question was a necessary condition for the occurrences of that outcome, at least in the particular case in question. It also assumes that the causal variable identified operated independently of other causal variables. Such assumptions are often difficult to substantiate, a fact that makes the use of a counterfactual problematic.
Thus, one must recognize that a plausible, useful counterfactual case is often not possible and, if attempted, does not add much, if anything, in support of a within-case historical explanation. It is very difficult if not impossible to conduct a plausible, useful counterfactual when the explanation for a historical event is very complex. “Complexity” can take several different forms, for example:
When many variables, though independent of each other, are part of the historical explanation (as is often the case), it is difficult to formulate a plausible counterfactual.
When the historical explanation is in the form of a sequential development over time, and not a single variable or cluster of variables at a given point in time—i.e., when the explanation is not derived from a simple “before-after” comparison—then it is very difficult to formulate a plausible counterfactual case.
When the causal variables in the historical explanation are not independent of each other but interdependent, then formulation of a plausible counterfactual case is exceedingly difficult, since it requires varying a number of causal variables and runs into the difficulty of weighing the precise weight of each variable.
For these reasons, we believe that the burden of supporting a historical explanation must be met not by using a counterfactual but by employing the process-tracing method in order to infer and construct a causal chain account of how various conditions and variables interacted over time to produce the historical outcome. In any case, counterfactual support for the explanation of a historical outcome is not needed if that explanation is supported by a strong theory or generalization; or if the causal chain is highly plausible, consistent with the evidence, and survives comparison with alternative explanations.
This is not to discourage investigators from trying to develop plausible, useful counterfactual cases but to alert them to the difficulties that stand in the way. While we believe that in principle a counterfactual is not needed to support any historical explanation, we recognize that opinions on this question differ and are content to rest our argument on the ground that plausible counterfactuals are generally infeasible, for the reason indicated here and in Chapter 8. This is not to deny the possibility that forcing oneself to attempt counterfactual analysis—even under such adverse conditions—may be useful in clarifying the process-tracing basis for the explanation.
There is another, quite different question that needs to be recognized and discussed. The preceding discussion focused on getting a good explanation for a given historical outcome. But the investigator may want to undertake a different task—namely, to address the question of whether an outcome other than the historical outcome would have been possible if some of its causes could have been different. This question is often raised when observers are dissatisfied with the historical outcome and argue that policymakers could have achieved a better outcome if they had acted