INFORMATION DESIGN IN ALLOCATION WITH COSTLY VERIFICATION

Abstract We study information design in a single‐agent allocation problem with costly verification, where the principal has a positive reservation value. The agent learns privately a signal about the principal's allocation value, drawn from a distribution controlled by an information designer. Given the signal distribution, the principal designs a mechanism to maximize her net value. We show that the agent‐optimal information pools high values at a signal just worth verification, whereas the principal‐optimal information features full disclosure. Moreover, any agent‐optimal information is principal‐worst, despite their partially aligned interests and the absence of transfers.