On the Robustness of Second-Price Auctions in Prior-Independent Mechanism Design

Abstract

Market designers often only have partial information about the environment and prefer simple mechanisms that are robust to the underlying uncertainty. In “On the Robustness of Second-Price Auctions in Prior-Independent Mechanism Design,” Anunrojwong, Balseiro, and Besbes study the fundamental problem of selling an item to n buyers, when only an upper bound on values is known and the seller minimizes worst-case regret. They establish that the same mechanism is robustly optimal across a wide range of environments for distributions of values: i.i.d., mixtures of i.i.d., exchangeable, and affiliated (the last two capturing positive dependence). Moreover, this robust mechanism is a second-price auction with random reserve price depending on n. Without positive dependence, the problem reduces to the one-buyer case. This result supports the wide use of second-price auctions in practice and allows them to quantify the robust value of competition in auctions.