Structural Analysis of First-Price Auction Data: Insights from the Laboratory

We use laboratory data from first-price auctions and manipulate the quantity and the quality of information available to assess the robustness of structural inferences (i.e., estimates, revenue predictions and optimal reserve price recommendations). We show that the latter are sensitive to the quantity of information when quality is low such as in field settings, and that improving quality in such settings dampens the effect of quantity and unveils out-of-equilibrium bidding patterns. Yet, a counterfactual analysis of the seller's revenues and optimal reserve prices indicates that behavior is best explained by the usual Nash equilibrium model with either risk aversion or a power form of probability misperception. When the information available is of the highest quality, as in laboratory settings, this model is typically rejected because of a nonlinear bidding behavior. We consider two rationales for such behavior and find that they leave revenue predictions and optimal price recommendations hardly affected.