Abstract: Dark rate estimation of (illegal) conduct is inherently estimating population sizes of unknown populations.
Numerous studies identify detected cartels to be profitable, and some hint at significant numbers of undetected cartels. A reliable estimation of the dark rate of collusive behavior supports competition authorities to evaluate antitrust law and enforcement strategies, necessary for successful detection and deterrence of cartels. Estimations about any population are based on draws of a random sample. Empirical research agrees that the sample of detected cartels is non-random and shows two main selection biases: (i) certain firms and industries are by structure prone to collude; and (ii) competition authorities selectively search markets that are either structurally prone to collusion or have a track record of detection. We model sample selection of detected cartels: for every cartel detected, how many cartels are left undetected? We answer this question by introducing a novel econometric approach disentangling the relevance of industry and firm characteristics from selection by the enforcer.
