Optimal public information dissemination : introducing observational learning into a generalized beauty contest

We develop a dynamic two-period generalized beauty contest to study the optimal level of publicity when disclosed information is subject to multiplier effects inherent to social learning. We build upon the static case, where all agents receive a private signal about an unknown fundamental state and only a fraction of all agents receive an additional public signal. However, in our model, agents no longer act simultaneously; rather, agents informed by both signals act in the first period, while those uninformed about the public signal delay their action and learn about informed agents' actions. We show that in the unique equilibrium of our dynamic game, informed agents overreact more strongly to public signals. The optimal dissemination of public information is thus considerably lower than the static case suggests. If the social learning signal is reasonably precise, aggregate welfare is higher than in the static case. Our results hold relevance for the optimal information policy design of public authorities.