All-pay auctions—an experimental study

This paper reports the results of a repeated all-pay auction game. The auction form used is the simplest possible, complete information, perfect recall and common value. Our main findings are that in such an auction, over-bidding is quite drastic, and the seller's revenue depends strongly on the number of bidders in early stages. However, after a few rounds of play, this dependence completely disappears and the seller's revenue becomes independent of the number of participants. The results are confronted with two solution concepts of economic theory, the Nash-equilibrium and the symmetric Logit equilibrium.