Speeding Up Ascending-Bid Auctions

Abstract

In recent years auctions have grown in interest within the AI community as innovative mechanisms for resource allocation. The primary contribution of this paper is to identify a family of hybrid auctions, called survival auctions, which combine the benefits of both sealed-bid auctions (namely, quick and pr edictable termination time) and ascendingbid auctions (namely, more information revelation often leading, among other things, to better allocations and greater expected revenue). Survival auctions are multi-round sealed-bid auctions with an information-revelation component, in which some bidders are eliminated from the auction from one round to the next. These auctions are intuitive, easy to implement, and - most importantly - provably optimal. More precisely, we show that (a) the survival auction in which all but the lowest bidder make it into the next round (the auction lasts for (n- 1) rounds when there are n bidders) is strategically equivalent to the Japanese ascending-bid auction, which itself has been proven to be optimal in many settings, and that (b) under certain symmetry conditions, ~ven a survival auction in which only the two h1ghest bidders make it into the next round (the auction lasts only two rounds) is Nash outcome equivalent to the Japanese auction.