Optimal Audit Policies With Heterogeneous Agents

This paper considers a tax evasion game where the tax authority intends to prevent income under-reporting within a network of heterogeneous taxpayers who are engaged in social interactions and exchange information. I propose a two-step game-theoretic optimal audit strategy from the point of view of the tax authority, which consists of a credible threat-to-audit message followed by a network-based audit policy. Subsequently, the tax authority targets taxpayers in function of their individual productivity and their position inside the network, triggering a series of spillover effects which eventually maximize the mean perceived subjective probability of being audited among all taxpayers. Moreover, the optimal audit strategy is robust to expected and non-expected utility theories, and it is invariant for any taxpayer utility function. Additionally, computer simulations determined that the proposed enforcement regime is robust to an ample range of parameter specifications and settings.