EMPIRICAL INDUSTRIAL ORGANIZATION Identification and Estimation of Bidders' Risk Aversion in First-Price Auctions

Risk aversion is a fundamental concept in economics used to explain agents' behavior under uncertainty. Risk aversion in auctions has been justified through the many uncertainties faced by bidders and through the large value of bids relative to bidders' assets. In first-price auctions, risk aversion renders more aggress­ sive bidding, while bidding in ascending auc­ tions is not affected, leading to the dominance of the sealed-bid mechanism over the ascending one. Risk aversion has been tested extensively on experimental data, as overbidding relative to the Nash equilibrium is frequently observed. In view of recent developments in the structural estimation of auction models, Patrick Bajari and Ali Hortacsu (2005) show that the risk-aversion model provides the best fit for experimental data. Given larger financial stakes, it is likely that bidders' risk aversion is present in field auction data as welL The theoretical auction literature does not provide simple implications that can be tested on bidding data. Consequently, detecting risk aversion in auctions is difficult if not impos­ sible. More generally with microeconomic data, risk aversion can be detected only when diver­ sification occurs, such as in portfolio manage­ ment and in auctions with diversification across species, as in Susan Athey and Jonathan Levin (2001). This calls for the necessity of structural modeling to evaluate bidders' risk aversion. In the structural approach, observed bids are assumed to be the outcome of the Bayesian-Nash equi­ librium of a particular model. Though a tight structure is imposed to explain observed bids,