Working Paper, Brown University, Department of Economics 2007-01
We apply stochastic stability to study the evolution of bidding behavior in private-values second-price, first-price and k-double auctions. The learning process has a strong component of inertia but with a small probability, the bids are modified in the direction of ex-post regrets. We identify essentially a unique bid that will be used by each type in the long run. In the second-price auction, this is the truthful bid. In the first-price auction, bidding half of one’s valuation is stable. The stable bid in the k-double auction is a toughening of the Chatterjee-Samuelson linear equilibrium strategy. If we add a friction in changing one’s bid, then truth-telling behavior is also obtained in the first-price and k-double auctions. Intuitively, the stochastically stable bid minimizes the maximal regret.