Testing Theories of Strategic Choice: The Example of Crisis Escalation

Abstract

Strategic-choice explanations of international crisis escalation imply that, because actors anticipate the likely responses of others and choose actions that avoid undesirable responses, our observations of nations' decisions to use force are interdependent and censored. For example, suppose a weak nation (B) is threatened by a powerful neighbor (A). If B believes A's threat, then it should back down. Unfortunately, this censors our ability to observe whether A would actually use violence. Many traditional methods ignore this censoring and mis-specify the extent to which variables influence crisis behavior. To ameliorate the problems of interdependence and censoring, I develop the Strategically Censored Discrete Choice model. Estimating this model using Bayesian Markov chain Monte Carlo simulation methods, I analyze Bueno de Mesquita and Lalman's (1992) dyadically-coded version of the Militarized Interstate Dispute data (Gochmann and Maoz 1984). Based on Bayesian-model comparison statistics, I conclude that strategic explanations outperform nonstrategic ones in explaining crisis behavior.