Detection and Deterrence in the Economics of Corruption: a Game Theoretic Analysis and some Experimental Evidence

This thesis contributes to our understanding of corruption deterrence for a specific class of game-theoretic corruption models, in which we assume that inspection of corrupt be- haviour happens through randomisation. Three models are explored theoretically and one experimentally. All models are three-player variations of the inspection game, and their typically unusual insights result from mixed-strategy equilibrium solutions. The first model examines an inspection game between an inspector and two potentially col- laborating offenders (a corrupt client and an official). Strikingly, its comparative statics suggest that higher penalties on corrupt clients increase the probability of corruption in the mixed equilibrium. The second model compares two states of the world, one where corrupt officials merely reject bribes (if they do not accept them), and one where corrupt officials report bribes (the latter leading to definite punishment of clients). The surprising result here is that, when officials prefer to report bribes (instead of merely rejecting them), the probability of corruption is again higher in equilibrium. The third model takes into account three different types of officials, a reporting type, a rejecting type, and a corruptible type. Its results show that e.g. an increase in the proportion of the reporting type increases the probability of corruption. To compare our theoretical results with data, we test a simple version of this game in the laboratory. Results of this pilot experiment were mixed, suggesting that three-player mixed-equilibrium behaviour is only in part and only qualitatively true on the aggregate, but not quantitatively or for individual play. An epilogue describes developments of a new, much improved experimental design and software, intended for future experiments.