Optimal Information Disclosure in Contests with Stochastic Prize Valuations
We study optimal information design in static contests where contestants do not know their values of winning. The designer aims at maximizing the total expected effort. Before the contest begins, she commits to the information technology that includes (1) a signal distribution conditional on each values profile (state) and (2) the type of signal disclosure to contestants -- public, private or none at all. Upon observing the signal, contestants simultaneously choose effort that maximizes their expected payoff in an all-pay auction game. We find that the optimal information technology involves private signals, which are slightly positively correlated and never reveal the true state precisely if the contestants' values of winning are different. In settings where public disclosure must be used, the optimal signal distribution generates symmetric beliefs about the values profile, so that, for example, a complete information concealment is optimal, while public and precise disclosure of each state is not.