Disentangling risk aversion and loss aversion in first-price auctions: An empirical approach

We develop a model that combines risk averse preferences with anticipated loss aversion to explain bidding in induced value first-price auctions. In particular, we allow bidders to be heterogeneous in risk aversion and loss aversion. We first show that risk aversion and loss aversion both induce ‘overbidding’ with respect to the standard expected utility model without risk and loss aversion. We then identify the distribution of risk aversion and loss aversion coefficients and develop a Bayesian method to estimate the model primitives, augmenting bidder-specific risk and loss coefficients. Our method predicts the data well, and the counterfactual analysis shows that loss aversion explains 65 ∼ 86 percent of overbidding in the data.