Searching and Learning in Internet Auctions: The eBay Example

This dissertation develops a dynamic bidding model for Internet auctions, especially eBay. The model parameters, namely bidders’ valuations and bidding costs, are then structurally estimated using a data set of around 800 eBay auctions for a handheld computer. Identification of the parameters is before formally shown. Next, the basic model is extended by a Bayesian updating mechanism. Using well known results about asymptotic distributions of extreme values, it is shown that bidders can use past observed transaction prices to update their beliefs about the location parameter of the distribution of the highest bid. Panel data estimations finally show that this kind of learning provides an explanation for the observed behavior of bidders at eBay.