Simple versus Optimal Mechanisms JASON D. HARTLINE Northwestern University and TIM ROUGHGARDEN Stanford University 1. INTRODUCTION A striking theme in the theory of single-item auctions is that simple auctions are optimal. Foremost, Myerson [Myerson 1981] showed that practically prevalent reserve-price-based auctions are indeed expected revenue-maximizing in natural, though stylized, models. This is fortunate as these auctions are simple and easy to optimize (just set the reserve price). That practitioners widely employ reserveprice-based auctions in settings much more complex that those where they are provably optimal motivates our rst question: When are reserve-price-based mechanisms approximately optimal? Bulow and Klemperer [Bulow and Klemperer 1996] up the ante further by showing that the reserve price is unnecessary, in the sense that a seller of a single item earns more revenue from the Vickrey auction with one extra bidder than from the optimal auction with the original bidders. This guarantee for an auction even simpler than Vickrey with an optimal reserve motivates our second question: How generally does a Bulow-Klemperer-style result approximately hold? Our recent paper [Hartline and Roughgarden 2009] provides tight approximation guarantees for the expected revenue of simple auctions, relative to that of an optimal auction, in environments
/lp/association-for-computing-machinery/simple-versus-optimal-mechanisms-Mx4K5eaFdV