Selection Criteria in Coordination Games: Some Experimental Results

We study the selection of an equilibrium for coordination games: symmetric, simultaneous move, complete information games which have multiple, Pareto-ranked Nash equilibria. We design and experiment to explore regularities in the observed outcomes for this class of games. With replication, we find that the Nash equilibrium concept accurately predicts the strategies chosen by players in these games. However, the equilibrium outcome is not always the Pareto-dominant equilibrium so that coordination failures can arise. Moreover, we find that altering the payoffs of a dominated strategy can influence the selection of a Nash equilibrium. Our results are consistent with a modified version of Harsanyi's tracing procedure in which players initially place some positive probability that their opponent is a cooperative player even though the cooperative strategy may be dominated by another strategy.