The Effects of Auction Parameters on Price Dispersion and Bidder Entry on eBay: A Conditional Logit Analysis

This paper studies price dispersion and bidder entry in sequential online auctions of indistinguishable items, where goods and shipping costs are identical. Contrary to intuition and the “law of one price,” previous empirical analyses show large price swings among auctions of identical products. To study this phenomenon, this paper performs empirical analysis on a dataset of 3,164 sequential eBay auctions between May and June of 2004, all of which advertised invitations to beta-test Google’s invite-only Gmail email service. This novel dataset, unlike the auctions of coins and trading cards studied by previous papers, facilitates accurate analysis because there are no shipping costs and each auction supplies exactly the same product. In spite of this obvious homogeneity, however, a huge spread was observed in the ending prices of these auctions—prices ranged from $0.99 to $217.50 per item. This paper studies existing theories of price formation and investigates how different parameters of an auction can influence its ending price. It also applies McFadden’s choice model, which uses conditional logit regressions, as part of an in-depth study of the effects of auction characteristics on bidder entry. Since this paper is the first to apply McFadden’s model to online auctions, it significantly extends existing literature on bidder choice. The paper then uses these results to develop a general theory of bidder behavior for private-value auctions on eBay. Finally, to conclude the study of price dispersion in sequential auctions, the paper empirically verifies the predictions of Weber’s martingale theorem; that is, previously ended auctions will influence the ending price of future auctions.