The costs and benefits of coordinating with a different group.

We consider a setup where agents care about i) taking actions that are close to their preferences, and ii) coordinating with others. The preferences of agents in the same group are drawn from the same distribution. Each individual is exogenously matched with other agents randomly selected from the population. Starting from an environment where everyone belongs to the same group, we show that introducing agents from a different group (whose preferences are uncorrelated with those of each of the incumbents) generates costs but may also (surprisingly) generate benefits in the form of enhanced coordination.