Estimating consumer surplus in ebay computer monitor auctions

Abstract

In this study, using data from computer monitor auctions on eBay collected in 2000, bidding functions are estimated by maximum likelihood using five different assumptions about the underlying distribution of independent private values. It is assumed that these values come from the logistic, gamma, weibull, pareto and lognormal distributions. Next, consumer surplus in the market for computer monitors is estimated and its sensitivity to different distributional specifications is tested. Two types of consumer surplus estimates are provided. First, ex-post consumer surplus estimates are constructed and then a lower bound for consumer surplus is computed using a “rational reassignment” methodology. Median consumer surplus estimates vary from $39 with the logistic to $143 with the lognormal, or the consumers’ share of surplus from 30% to 61%. Expected consumer surplus estimates indicate a high sensitivity to the distribution specification, especially to the tails of the distribution. Lower bound estimates are more solid and more reliable since they are independent of the tails. Accordingly, these statistics, which do not vary with distribution, yield a median estimate of $41 and a consumer share of 32%. Finally, the last part of this study examines which distribution best fits the data. Information criteria favor the gamma distribution, tests against the empirical distribution of second and third highest values prefer the logistic distribution.