Empirical Relevance of Ambiguity in First Price Auction Models

We study the identification and estimation of first-price auction models where bidders have ambiguity about the valuation distribution and their preferences are represented by maxmin expected utility. When entry is exogenous, the distribution and ambiguity structure are nonparametrically identified, separately from risk aversion (CRRA). We propose a flexible Bayesian method based on Bernstein polynomials. Monte Carlo experiments show that our method estimates parameters precisely, and chooses reserve prices with (nearly) optimal revenues, whether there is ambiguity or not. Furthermore, if the model is misspecified -- incorrectly assuming no ambiguity among bidders -- it may induce estimation bias with a substantial revenue loss.