Bayesian Comparison of Private and Common Values in Structural Second-Price Auctions
(English)Manuscript (preprint) (Other academic)
We compare the performance of the Gaussian second-price common value (CV) model in Wegmann and Villani (2011) to a comparable independent private value (IPV) version of that model. The two models are contrasted on a dataset from $1050$ Internet coin auctions at eBay. The models are evaluated along several dimensions, such as parameter inference, in-sample fit, and accuracy of out-of-sample predictive density forecasts. Both models fit the eBay data well with a slight edge for the more robust CV model. We do not find any evidence of a winner's curse effect in the eBay data, which speaks in favor of the IPV model. However, the optimal minimum bids in the CV model are clearly closer to the actual minimum bids in the eBay data than the optimal choice of no minimum bid in the IPV model. The IPV model predicts auction prices slightly better in most auctions, while the CV model is much better at predicting auction prices in more unusual auctions. The robustness of the CV model is also supported by a small simulation study, where the CV model performs relatively better on simulated data from the IPV model than the IPV model fitted to CV data.
Probability Theory and Statistics
Research subject Statistics
IdentifiersURN: urn:nbn:se:su:diva-57275OAI: oai:DiVA.org:su-57275DiVA: diva2:415130