“Optimal Discriminatory Disclosure” by Dr. Xianwen Shi
University of Toronto
A seller of an indivisible good designs a selling mechanism for a buyer who knows which of two distributions that the valuation of the good is drawn from but not the realization of the valuation. The seller can choose how much private information about the valuation that the two types may access. Under the assumption that the two distributions are ranked by likelihood ratio dominance, we show that optimal disclosure is discriminatory and has the structure of nested intervals. Moreover, the revenue under optimal discriminatory disclosure may not be attainable by non-discriminatory disclosure. A natural generalization of binary partition is shown to be optimal among all disclosure policies consistent with the given information environment.