Prospect theory in a dynamic game: Theory and evidence from online pay-per-bid auctions

Abstract

Abundant evidence exists that expected utility theory does not adequately describe decision making under risk. Although prospect theory is a popular alternative, it is rarely applied in strategic situations in which risk arises through individual interactions. This study fills this research gap by incorporating prospect theory preferences into a dynamic game theoretic model. Using a large field data set from multiple online pay-per-bid auction sites, the authors empirically show that their proposed model with prospect theory preferences makes a better out-of-sample prediction than a corresponding expected utility model. Prospect theory also provides a unified explanation for two behavioral anomalies: average auctioneer revenues above current retail prices and the sunk cost fallacy. The empirical results indicate that bidders are loss averse and overweight small probabilities, such that the expected revenue of the auction exceeds the current retail price by 25.46%. The authors illustrate and empirically confirm a managerial implication for how an auctioneer can increase revenue by changing the details of the auction design.