Efficient dynamic coordination with individual learning

We study how the presence of multiple participation opportunities coupled with individual learning about payoffs affects the ability of agents to coordinate efficiently in global coordination games. Two players face the option to invest irreversibly in a project in one of many rounds. The project succeeds if some underlying state variable è is positive and both players invest, possibly asynchronously. In each round they receive informative private signals about è, and asymptotically learn the true value of è. Players choose in each period whether to invest or to wait for more precise information about è. We show that with sufficiently many rounds, both players invest with arbitrarily high probability whenever investment is socially efficient, and delays in investment disappear when signals are precise. This result stands in sharp contrast to the usual static global game outcome in which players coordinate on the risk-dominant action. We provide a foundation for these results in terms of higher order beliefs.