Using computational methods to perform counterfactual analyses of formal theories

Recently there has been an increase in the number of researchers who use rational choice models to explain single cases and rare events. Because of the small number of cases under study, these researchers must rely either explicitly or implicitly on counterfactual reasoning. This paper argues that computational methods provide a profitable means of carrying out rigorous counterfactual analysis. The authors advocate robustness analysis as one important part of the counterfactual analysis of formal theories. Specifically, they evaluate the robustness of the behavioral assumptions of two formal models using various heuristic search algorithms and Markov chains. They find that Kuran's (1989) threshold model of mass protest and Ingberman's (1985) model of direct-democracy referenda are robust to perturbations in their behavioral assumptions. These findings increase the plausibility of causal claims made by scholars who use these models to explain specific events.