Counterfactuals, Causal Inference, and Historical Analysis

I focus primarily on the utility of counterfactual analysis for helping to validate causal inferences in historical analysis. How can we use what did not happen but which easily could have happened to understand what did happen? With an infinite number of things that might have happened, and with temptations to construct “counterfactuals of convenience” to bolster one's preferred historical interpretations or political preferences, we need a set of rules or best practices for evaluating the scientific legitimacy of counterfactuals. Building on earlier work in several disciplines, I develop a set of criteria for the conduct of counterfactual analysis in historical case studies. The best counterfactuals begin with clearly specified plausible worlds involving small and easily imaginable changes from the real world. They make relatively short-term predictions based on empirically validated theoretical generalizations and on secondary counterfactuals that are mutually consistent. These counterfactuals are also sensitive to strategic behavior that might return history to its original course, and they are explicitly tested against competing counterfactuals.