ESI Brown Bag Lectures
Jingfeng Lu, Ph.D. - "Optimal Disclosure of Value Distribution Information in All-pay Auctions"
Abstract: In this paper, we follow a Bayesian Persuasion approach to study the auction organizer’s optimal disclosure of information about players’ value distribution in a two-player all-pay auction setting. Players’ private values (either high vh or low vl) are independently and identically distributed. There are two possible value distributions (i.e., two possible states), and none of the players knows the actual distribution. Before the auction starts, the organizer pre-commits to a public signal to reveal information about the prevailing value distribution. We find that there exists a cutofff for value ratio v = vh/vl, above which a monotone equilibrium arises under any prior belief about the state. In this circumstance, no disclosure is optimal. When value ratio v is below the cutoff, there exist exactly two threshold beliefs about the state that separate prior beliefs generating monotone and non-monotone equilibria. A prior belief would lead to a non-monotone equilibrium if and only if it lies in between. If the original prior μ0 leads to a monotone equilibrium, then no disclosure is optimal; otherwise, a partial disclosure, which generates a posterior distribution over the two threshold beliefs, is optimal.