Click to view a PDF (Adobe Acrobat PDF) file.
Revised April 2002
Economic theory suggests that if firms can successfully collude, they will be able to increase their own profits at the consumers� expense. Government antitrust policy aims to protect the public from such behavior. Enforcing such policy depends on a prosecutor�s ability to detect collusive behavior among firms. In this paper, we discuss an approach developed Bajari and Ye (2001a,b) for detecting collusion. First, we discuss two conditions: conditional independence and exchangeability, which the data must exhibit if it has been generated by non-collusive behavior. Second, by eliciting information from industry experts about costs across the industry under scrutiny, it is possible to compute the probability that the data was generated by competition or collusion. These tests appear to work well in empirical applications. In industries where competition is widely believed to exist, we find that our conditions are largely supported by the data. In industries where collusion has already been proven, we find that indeed, our tests point to obvious collusion. While we do not believe these empirical tests are flawless, we argue that they can be a useful first step in detecting suspicious bidding behavior.