Bidding in first-price and second-price interdependent-values auctions: A laboratory experiment∗

We report a laboratory experiment on first-price and second-price auctions in settings with independent signals and interdependent values. The environment includes independent private values and the common-value “wallet game” as limiting cases. We manipulate the degree of interdependence of values across sessions, while maintaining the same Bayes-Nash equilibrium bidding function. In contrast, cursed equilibrium predicts bids will be raised for lower signals. We find some support for cursed equilibrium, in that bids change as the degree of value interdependence changes. Contrary to both Bayes-Nash and cursed equilibrium, auction revenues are largest for intermediate levels of interdependence. We construct a model combining cursedness with an underweighting of the opportunity costs of higher bids, and find substantial bidder heterogeneity. A majority of bidders are either fully cursed and disregard completely the bad news that winning the auction entails, or are not cursed at all. We also find evidence for some systematic procedural differences in bidding between first-price and second-price auctions.