Accounting for Unobserved Heterogeneity in Ascending Auctions

Abstract

We study identification of ascending auctions with additively separable auction-level unobserved heterogeneity. Usual deconvolution approaches are inapplicable due to the lack of the highest bid; both unobserved heterogeneity and incomplete bid data contribute to the correlation among observed bids. We propose an identification strategy exploiting "within" independence of unobserved heterogeneity and private value. First, the ratio of two observed order statistics' characteristic functions identifies the private value distribution. Second, standard deconvolution with a known error distribution identifies the unobserved heterogeneity distribution.