Revealed types and beliefs in bayesian games

Abstract This paper examines players’ behavior in Bayesian games when they may have incorrect beliefs about their opponents’ types. Their behavior can be characterized by a type distribution function and a belief function if their actions across games satisfy four conditions: Dominance, Monotonicity, Continuity, and Linearity. Both the type distribution function and players’ shared belief can be identified from their aggregate behavior across games. The analyst can also test if players’ beliefs are accurate with action data only.