Testing for Collusion in Asymmetric First-Price Auctions

This paper proposes fully nonparametric tests to detect possible collusion in first-price procurement (auctions). The aim of the tests is to detect possible collusion before knowing whether or not bidders are colluding. Thus we do not rely on data on anti-competitive hearing, and in that sense is ’ex-ante’. We propose a two steps (model selection) procedure: First, we use a reduced form test of independence and symmetry to shortlist bidders whose bidding behavior is at-odds with competitive bidding, and Second, the recovered (latent) cost for these bidders must be higher under collusion than under competition, because collusion dwarfs competition, hence detecting collusion boils down to testing if the estimated cost distribution under collusion first order stochastically dominates that under competition. We propose rank based and Kolmogorov-Smirnov (K-S) tests. We implement the tests for Highway Procurement data in California and conclude that there is no evidence of collusion even though the reduced form test supports collusion.