Efficient Equilibrium Selection in Evolutionary Games with Random Matching

This paper investigates the sensitivity of recent evolutionary models of learning to the specification of the matching mechanism. We study a literally random matching mechanism, combined with a process of strategy adjustment based on the realized average performance of each strategy. In the key class of symmetric 2×2 coordination games, the Pareto-efficient equilibrium, per se, is selected, rather than the risk-dominant equilibrium, as the probability of experimentation (or “mutation”) goes to zero. Furthermore, convergence to the equilibrium is relatively fast. We extend these results, for example, to games of common interest.