Preserving Bidder Privacy in Assignment Auctions: Design and Measurement

• 成果类型: Article
• 署名作者: Liu, De; Bagh, Adib
• 署名单位: University of Minnesota System; University of Minnesota Twin Cities; University of Kentucky
• 刊物名称: MANAGEMENT SCIENCE
• ISSN/ISSBN: 0025-1909
• DOI: 10.1287/mnsc.2019.3349
• 发表日期: 2020
• 页码: 3162-3182
• 关键词: assignment problem; ascending auctions; privacy preservation; entropy; quasi-Monte Carlo

摘要

Motivated by bidders' interests in concealing their private information in auctions, we propose an ascending clock auction for unit-demand assignment problems that economizes on bidder information revelation, together with a new general-purpose measure of information revelation. Our auction uses an iterative partial reporting design such that for a given set of prices, not all bidders are required to report their demands, and when they are, they reveal a single preferred item at a time instead of all. Our design can better preserve bidder privacy while maintaining several good properties: sincere bidding is an ex post Nash equilibrium, ending prices are path independent, and efficiency is achieved if the auction starts with the auctioneer's reservation values. Our measurement of information revelation is based on Shannon's entropy and can be used to compare a wide variety of auction and nonauction mechanisms. We propose a hybrid quasi-Monte Carlo procedure for computing this measure. Our numerical simulations show that our auction consistently outperforms a full-reporting benchmark with up to 18% less entropy reduction and scales to problems of over 100,000 variables.