Social learning with coordination motives

We study observational learning among agents with coordination motives. On a discrete time line, communities of agents receive private information about an uncertain state, observe some predecessors' actions, and then take their own action. An agent's payoff is both state-dependent and increasing in the number of her peers taking the same action as hers. We find that connectivity between observations is the key determinant of the equilibrium pattern of information aggregation. When observations are connected, coordination motives inevitably lead to herding even when private beliefs are unbounded; when observations are separated, asymptotic learning becomes achievable. Herding can also be eliminated by making private information public within each community, and reduced by making observation endogenous and costly.