Network formation and anti-coordination games

We study a setting in which individual players choose their partners as well as a mode of behavior in 2 x 2 anti-coordination games -- games where a player's best response is to behave differently than the opponent. We characterize the nature of equilibrium networks as well as study the effects of network structure on individual behavior. Our analysis shows that both the network architecture and the induced behavior are crucially dependent on the value of the cost of forming links. In general, the equilibrium configurations are found to be neither unique nor efficient. This conclusion continues to hold if the population game is embedded in a standard evolutionary model of learning, since all equilibria turn out to be stochastically stable