ESTIMATING BID LEVELS DURING ONLINE AUCTIONS

Sellers in eBay are often small-business owners whose livelihood depends on fast turnaround of their cash flows. Unlike in traditional auctions, these sellers are content to sell as soon as some target price is reached. While a wealth of literature exists on the final rent of the various stakeholders in a traditional auction setting, what is of interest here is to estimate current bid levels in ongoing auctions. The question that we analyze here is this - how does a seller estimate the time required to reach a target price in an auction? To this end, we model the bidding process as an infinite quasi-birth-death process, characterized by bursts of rapid bidding and subsequent lulls. We obtain closed-form solutions for the transient probability distribution in the frequency domain of the bid prices in an ongoing auction. We use transform inversion algorithms to numerically compute the transient probability distributions in the time domain. Experienced auctioneers can use these results to estimate expected ending times for their auctions. Sample observations from online auctions indicate that there might be significant room for improvement for sellers in setting their auction ending times. Subject classifications: Probability: stochastic model applications, Markov processes, random walk. Area of review: Stochastic models.