Modeling Bidder Risk Preferences to Optimize Pricing

Name-your-own-price selling is a tractable laboratory paradigm for studying bidding behavior because it involves only one bidder per transaction. Using data from an incentive-compatible laboratory experiment that implemented a name-your-own-price seller who charges entry fees, we estimate a flexible model of risk preferences at the individual level, and we find substantial population heterogeneity. The behavior of three quarters of our subjects is more consistent with prospect theory than with expected utility theory in that their estimated utility functions are convex in the loss domain. In several counterfactual simulations, we measure the impact of accounting for the heterogeneity in risk preferences on setting entry and reserve prices in three market institutions that involve bidding, including first-price auctions. We find that when the seller cannot discriminate based on risk preferences, prices set using the simpler homogeneous model (that assumes everyone has the same risk preferences) achieve over 99% of the optimal profit. By contrast, a discriminating seller can benefit from using the heterogeneous model in several situations that we characterize as depending on both the institution and the dispersion in the distribution of valuations.