Testing Risk Dominance and Payoff Dominance in Repeated Global Stag Hunt Games

Abstract

In any 22 global game, Carlsson and van Damme (1993b) showed that the game has a unique dominance solvable equilibrium that corresponds to the risk dominant equilibrium of the related common knowledge game with multiple strict equilibria. We test this prediction in repeated global stag hunt games. Under private information, a few cohorts coordinate on thresholds close to the global games prediction, but many cohorts coordinate on thresholds close to the ecient threshold. We argue that initial conditions and adaptive behavior play a key role in forming mutually consistent expectations in this game. We also investigate why the iterated dominance argument used to get uniqueness in the private informationtreatment is not salient.