Distributionally Robust Auction Design with Deferred Inspection

Abstract: Mechanism design with inspection has received increasing attention due to its applications in the field. For example, large warehouses have started to auction scarce capacity. This capacity shall be allocated in a way that maximizes the seller's revenue. In such mechanism design problems, the seller can inspect the true value of a buyer and his realized sales in the next period without cost. Prior work on mechanism design with deferred inspection is based on the assumption of a common prior distribution. We design a mechanism with a deferred inspection that is (distributionally) robustly optimal either when the ambiguity-averse mechanism designer wants to maximize her worst-case expected payoff or when she wants to minimize her worst-case expected regret. It is a relatively simple mechanism with a concave allocation and linear payment rules. We also propose another robustly optimal mechanism that has the same concave allocation function but extracts the maximal payment from all the types of the agent, which can have a strictly higher expected payoff under non-worst-case distributions compared to the robustly optimal mechanism with the linear payment rule. We show that multi-bidder monotonous mechanisms might not exist.