Ulric B. and Evelyn L. Bray Social Sciences Seminar

Abstract: We study information aggregation in a fully rational model of social learning on networks with endogenous action timing. In the model, each agent receives a single weak private signal about the value of making a one-time, irreversible investment decision. In each period, every agent who has not made the irreversible investment observes the actions of her neighbors in the previous period, then decides based on the whole history of her observations whether to make the irreversible investment.

We study information diffusion and aggregation in this model by studying how likely agents are to eventually make the right investment decision. Using information theoretic tools and a partial characterization of equilibria, we show that information aggregation does not occur in equilibrium in ``low-complexity'' trees, even when agents are very patient, whereas there is no such obstruction in social networks with high-degree vertices.