Information-Robust Optimal Auctions

TL;DR

This paper characterizes a robust auction design that maximizes worst-case revenue when bidders have private signals and the seller lacks knowledge of signal distribution, using a second-price auction with a random reserve.

Contribution

It introduces a robust auction mechanism optimal under unknown signal distributions, extending auction theory to settings with incomplete information about signals.

Findings

Second-price auction with random reserve maximizes worst-case revenue.

The mechanism is optimal across all equilibria in undominated strategies.

The approach accounts for unknown signal distributions in auction design.

Abstract

A single unit of a good is sold to one of two bidders. Each bidder has either a high prior valuation or a low prior valuation for the good. Their prior valuations are independently and identically distributed. Each bidder may observe an independently and identically distributed signal about her prior valuation. The seller knows the distribution of the prior valuation profile and knows that signals are independently and identically distributed, but does not know the signal distribution. In addition, the seller knows that bidders play undominated strategies. I find that a second-price auction with a random reserve maximizes the worst-case expected revenue over all possible signal distributions and all equilibria in undominated strategies.