"Dark rate estimation of (illegal) conduct is inherently estimating population sizes of unknown populations. This thesis reversely engineers prediction of the population of collusion in an economy from a non-randomly sampled set of observed cases. If conducted successfully, our method applies in various situations where the population size is in doubt."
Title: Estimating Dark Rates of Collusion
Keywords: Applied Statistics, Microeconometrics, Industrial Organization
Planned Duration: Fall 2019 – Spring 2024
Abstract: To evaluate substantive law systems, enforcement regimes and their changes it is crucial to understand their impact on detection and deterrence rates of illegal behaviour. Detection and deterrence rates jointly constitute the dark rate of cases, i.e. unknown population. Evaluations based on the number of detected cases neglect the impact of undetected cases. Therefore, estimating dark figures of unknown criminal activities is crucial in multiple areas like criminology, biology, epidemiology, law, economics. The economic literature on estimating dark rates is grounded in empirical industrial economic research starting off with hazard rate models.
A more recent method infers population sizes of animals from samples. This method, capture-recapture analysis, offers the potential to analyze and compare the impact of detection and deterrence of different law enforcement systems without relying on the assumption of independence. Since capture-recapture analysis is a parametric test (of zero parameters), we use it as an inspiration for a game theoretical model on formation and termination of structural collusion, i.e. illegal cartels. On the resulting data we test existing micro-econometric models in terms of predictive power as we understand the data generating process, and aim to develop new estimation methods based on Markov chain transition probabilities to get a more precise estimate of the population of illegal cartel offenses.
We hope to contribute to the ongoing discussion of undetected illegal behavior by agents (not limited to economic incidences) or population sizes, where one might retrieve a subsample having a potential sample selection bias.