Optimal Disclosure in Two-Sided Matching

Abstract

A matching platform elicits information about two customers' (quality) types and decides how to disclose that information. The platform intends to persuade the customers to form a match, knowing that adverse selection may occur (and undermine matchmaking) when the customers are uncertain about each other's type. We show that the optimal disclosure policy features upper censorship: It pools the customer types above a threshold and fully reveals the rest. This policy is optimal because it not only maximizes the chance that each customer is accepted by his partner but also eliminates the adverse selection problem. Methodologically, we introduce a novel approach to solve a persuasion problem where the sender's ex-post utility does not merely depend on the mean of the receiver's posterior belief about a state.

Keywords: two-sided matching, Bayesian persuasion, adverse selection, upper censorship.

JEL Classification: D47, D82, D83.