Looking for audience costs

The methodological issues that arise in testing Fearon's argument about domestic political audience costs and signaling in international crises are examined, in particular the difficulty of finding direct evidence (1) that escalating a crisis and then backing down jeopardizes a leader's tenure in office, and (2) that democratic leaders are more vulnerable to removal in this event than are nondemocratic leaders. Tests that seek to measure the existence and magnitude of audience costs encounter severe problems of partial observability and strategic selection: the effect of audience costs on a leader's political survival can only be detected by looking at cases in which the costs are actually incurred, but strategic choice implies that the probability of incurring audience costs is a function of their value. A formal model, brief case studies. and Monte Carlo simulations are used to show that these problems bias direct tests against supporting either of the audience cost propositions. Tests based on observed audience costs understate both the mean level of audience costs in the full population and the difference in means across regime types.