Crisis Bargaining, Escalation, and MAD

Although incomplete information is recognized to be an essential feature of crises, game-theoretic formulations have not generally modeled this explicitly. This paper models a mutually assured destruction (MAD) crisis as a game of sequential bargaining with incomplete information, sufficiently simple that its equibria may be found. These provide better game-theoretic foundations for the notions of resolve and critical risk and their role in crises and also make it possible to compare the bargaining dynamics of this model with those of descriptively richer, but incompletely specified models, revealing several inconsistencies: several analyses of MAD conclude that the state with the greatest resolve in this contest of resolve will prevail. Many models based on critical risks suggest that a state is less likely to escalate, the greater its adversary's resolve. In our model, however, the state with the weakest resolve sometimes prevails, and some states prove more likely to escalate if their adversaries' resolve is greater.