Essays on bounded rationality in auctions and bargaining

This dissertation is a theoretical and experimental analysis of boundedly rational behavior in dynamic auctions and bargaining.; The first chapter develops a novel model of boundedly rational learning of an agent's private type during a dynamic auction. This is a comprehensive auction model that explains empirical regularities from eBay auctions. In this auction, the current price equals the second highest bid received so far at any time. One bidder knows her private valuation while the other bidder does not know his private valuation. Upon seeing a posted price, he learns whether his valuation is above that price. In any equilibrium, the informed bidder bids only in the last period or snipes with positive probability. The uninformed bidder bids in every period nibbles to optimally change the price unless the price is above his valuation or he is the high bidder.; The second chapter (joint with John Morgan), tests the effect of different price "frames" using field experiments on eBay. We auctioned off pairs of popular music CDs and Xbox game cartridges with the same reserve but different compositions of the opening price and the shipping charge. Theory predicts that the seller should make the same revenue, on average, in each auction. In the low reserve treatments, auctions with lower opening price consistently earned significantly higher revenue. However, there was no systematic difference in revenues in the high reserve treatments. This suggests that some buyers view shipping costs differently from "the price." They somewhat discount the shipping charge as long as it is not excessively high relative to some reference point.; Finally, the third chapter uses the learning model developed in the first chapter in a continuous-time bargaining game. Under two-sided incomplete information with no time discounting, no trade occurs when the buyer and the seller know their private valuations. When a player does not know her own valuation, but learns about it during the bargaining process, there exists an ex-post efficient equilibrium where trade occurs with positive probability. Thus, a very simple boundedly rational model of learning circumvents the Myerson-Satterthwaite theorem and an efficient incentive compatible mechanism exists.