Implications of the bidders' arrival process on the design of online auctions

Abstract

We have recently seen a tremendous number of auctions conducted over the Internet. This form of electronic commerce is rapidly growing, and it is projected to account for 30% of all e-commerce by 2002. Using actual bidding transaction data from 324 business-to-consumer online auctions, we analyzed the bidders' arrival process during each auction. We found that most bidders like to sign on early in the auction; typically, 70% of the bidders sign on during the first half. Our statistical analysis reveals that the minimum initial bid is negatively correlated with the number of bidders per auction, while the number of units offered and the length of the auction are positively correlated with the number of bidders. We also present a model for estimating the expected price as a function of the number of bidders, the mean and variance of the private valuation distribution, and the number of units to be sold in the auction. Our analysis shows that increased dispersion in the bidders' values may either increase or decrease the auction price, depending on the bidders' overall arrival process, the length of the auction and the number of units. We calculate the optimal auction length and show that an auction's profit is a unimodal function of its duration and the number of units. This paper also addresses several other economic tradeoffs that are relevant for the optimal design of Internet auctions.