A comparison of multiple-unit all-pay and winner-pay auctions under incomplete information

This paper examines the properties of independent-private-value all-pay and winner-pay auctions when there are multiple units sold. We study bidding behavior, efficiency and revenue in a set of nine experimental sessions, each with six bidders. All-pay auctions were played in six of the sessions, three sessions with four units and three sessions with two units auctioned. A four-unit winner-pay auction was played in three of the sessions. Our data show that the all-pay auction and the winner-pay auction are empirically revenue equivalent and yield higher revenue than the risk neutral Bayesian equilibrium. Revenue is higher in the allpay auction when K=2 than when K=4, despite the fact that Bayesian equilibrium revenues are identical for the two cases. Our evidence also suggests that the winner-pay auction is more likely than the all-pay auction to lead to a Pareto-efficient allocation.