More neighbors, more efficiency

This paper considers a finite population of agents located within an arbitrary fixed network. Every agent plays a coordination game with his neighbors. If one neighbor's payoff from a specific interaction exceeds his average payoff per interaction, the neighbor is perceived as better performing. Over time agents imitate the strategies of their better performing neighbors; occasionally they make mistakes. Sufficient conditions for emergence of Pareto efficient and risk dominant conventions are provided. The paper also illustrates the main results through relevant examples.