Incentive Compatible Information Disclosure

This paper studies the optimal disclosure of information about an agent's quality when it is a combination of a component privately observed by the agent and another latent component. Upon soliciting a report from the agent about his private observation, a principal performs a test which reveals the latent component. The principal then discloses information to the market/public which rewards the agent with compensation equal to the agent's expected quality. We study incentive compatible disclosure rules that minimize the mismatch between the agent's true and expected qualities while inducing truth-telling from the agent. The optimal rule entails full disclosure when the agent's quality is a supermodular function of the two components, but entails partial pooling when it is submodular. We express the optimization problem as a linear transformation of the mean dual-belief, which describes the joint distribution of prior and mean posterior beliefs under disclosure, and obtain the optimal disclosure rule as a corner solution to this linear problem. We identify the number of messages required under the optimal rule and relate it to the agent's incentive compatibility conditions.