Working Paper: CEPR ID: DP18662
Authors: Ran Eilat; Kfir Eliaz; Xiaosheng Mu
Abstract: In many auction settings the auctioneer must disclose the identity of the winner and the price he pays. We characterize the auction that minimizes the winner's privacy loss among those that maximize total surplus or the seller's revenue, and are strategy-proof. Privacy loss is measured with respect to what an outside observer learns from the disclosed price, and is quantified by the mutual information between the price and the winner's willingness to pay. When only interim individual-rationality is required, the most privacy preserving auction involves stochastic ex-post payments. Under ex-post individual rationality, it is the second-price auction with deterministic payments.
Keywords: Bayesian; Privacy; Mutual Information; Auction Theory
JEL Codes: D44; D82
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Auction Mechanisms (D44) | Privacy Loss (K24) |
| Stochastic Payments (G19) | Privacy Loss (K24) |
| Deterministic Payments (C69) | Privacy Loss (K24) |
| Second-Price Auction (SPA) (D44) | Privacy Loss (K24) |
| Post-Bid Randomization (C90) | Privacy Loss (K24) |
| Maximum Payment Type-Dependent (E42) | Deterministic Payments (C69) |