Welcome to the new version of CaltechAUTHORS. Login is currently restricted to library staff. If you notice any issues, please email coda@library.caltech.edu
Published June 2003 | public
Book Section - Chapter

Selling online versus offline: theory and evidences from Sotheby's

Abstract

We consider a recent business and policy question of "how and why does a firm use online markets versus traditional offline markets ?" using a unique dataset of more than 3000 auctions held by Sotheby's online at eBay and offline at New York in June-July2002. We find robust empirical regularities in our dataset about the use of online markets. First, the average transaction price is more than 10 times higher in offline markets. This fact strongly suggests that the seller is not simply randomly assigning assets between online and offline markets. Second, the higher the mean and spread of pre-auction estimates of an asset, the more likely seller is to sell the asset in offline markets. Third, the transaction rate is higher in offline markets. Next, we build a simple model of offline and online markets to identify the business logic behind these empirical regularities. We model offline markets as an auction with endogeneous entry a la McAfee and McMillan (1987) where the traders pay transaction costs to hold transactions. We model online markets as standard ascending auctions. In online markets, the seller can save transaction costs and entry by bidders is easy, but the seller cannot reveal much information, leading to higher valuation risk and severe winner's curse. The seller sells the asset with high valuation risk in offline markets to alleviate winner's curse. In order to compensate for the transaction costs, the expected value of the asset sold in offline markets is higher. Since the seller's profit is equal to social surplus in offline markets due to entry costs, the seller is more eager to sell assets. Finally we provide a simple maximum likelihood estimation of transaction costs and information revelation effects based on discrete choice models.

Additional Information

© 2003 ACM. This paper is based on a joint project with John McMillan. I thank Takeshi Ameiya, Susan Athey, Alan Sorensen, and Robert Wilson for helpful conversations. We thank seminars participants at Stanford GSB, Toulouse IDEI, and SWET for their useful feedbacks. We are grateful for Stanford GSB CEBC for financial assistances. The full version of the paper, with proofs, is available at http://www.hss.caltech.edu/kazumori

Additional details

Created:
August 19, 2023
Modified:
October 23, 2023