The Bomb-Crater Effect of Tax Audits: Beyond Misperception of Chance

In this paper, we run a laboratory experiment where the information set is relatively rich, and, in particular, it includes audits on other taxpayers. At the same time, the implementation of the Bayesian updating process for the subjective probability to be audited is fairly simple. By doing so, we are able to elicit a range of consistent but heterogeneous probability beliefs and to distinguish between Bayesian and non-Bayesian subjects. We obtain two major results concerning Bayesian subjects. First, they exhibit strong and robust short-run BoCE. Second, they are seemingly not affected by audits on other taxpayers in their compliance decision. These results are robust to different definitions of Bayesianity and to different specifications. They conflict with the evidence that Bayesian agents do perceive correctly the chance to be audited. In turn, this suggests that existing explanations of the BoCE are not entirely satisfactory and that alternative theories, possibly based on the Duality approach, are needed.