When are Nash equilibria self-enforcing? An experimental analysis

We investigate the effect of non-binding pre-play communication in experiments with simple two-player coordination games. We reproduce the results of other studies in which play converges to a Pareto-dominated equilibrium in the absence of communication, and communication moves outcomes in the direction of the Pareto-dominant equilibrium. However, we provide new results which show that the effectiveness of communication is sensitive to the structure of payoffs. Our results support an argument put forward by Aumann: agreements to play a Nash equilibrium are fragile when players have a strict preference over their opponent's strategy choice. We also find that informative communication does not necessarily lead to the Pareto-dominant equilibrium.