JMP: "Persuasion with Unknown Beliefs: On the Usefulness of Knowledge in Persuading Regulatory Agencies”.
When trying to persuade someone, one finds it useful to know the beliefs the target of persuasion holds. Yet often such beliefs are unknown to the persuader. How should persuasion be designed when knowledge of prior beliefs is limited? I ask this question in an environment where the sender does not know the prior belief of the receiver, and thus evaluates each signal structure using the receiver's prior that is the worst for the sender. The model applies to lobbying settings such as one in which a pharmaceutical company would like to persuade the FDA to approve a drug, and the FDA only wants to approve drugs that are sufficiently safe.
I show that the sender's lack of knowledge can lead to particularly pernicious outcomes: the FDA approves even the least safe drugs with a positive probability. If the sender is maximally ignorant about the receiver's prior, then the optimal signal recommends approval with a strictly positive probability in every state. The optimal signal structure when the sender is ignorant is qualitatively different from the optimal signal structure when the sender is knowledgeable. In particular, the sender hedges her bets: the optimal signal induces the high action in more states than in the standard model, albeit with a lower probability. Turning to the welfare consequences of the sender's ignorance, I show that the receiver strictly prefers to face an ignorant sender rather than a sender that is perfectly informed about the receiver's prior and that the sender's lack of knowledge can lower the payoff of both the sender and the receiver. These results have important implications for the optimal transparency requirements for the FDA, showing that full transparency is never optimal, and that the optimal transparency level may be interior.
Read the paper here.
Learn more about Svetlana here.