Information, Matching and Outcome Selection

We consider a finite population of agents who exchange information and are paired every period to play a game with tension between risk dominance and Pareto efficiency. Agents sample past plays and corresponding payoffs from their information neighborhood, and choose one of two possible actions using either best response or imitation. Information exchanges and possible matchings each constitutes a network. We first provide a complete description of the medium run outcomes and show that in the medium run only information matters. We then identify the conditions whereby either the risk dominant or the Pareto efficient convention is stochastically stable, and show how efficiency in the long run depends on the matching network. JEL Classification C73, D85.