Honesty versus progressiveness in income tax enforcement problems

Abstract

We study an income tax enforcement problem using a principal agent model where the government sets the tax and inspection functions. These functions are announced to the agents and there is no commitment problem. The penalty function for dishonest taxpayers is given exogenously and must satisfy certain social norms. We give a partial characterization of the optimal policy and prove that for a large family of penalty functions this policy is such that honesty implies regressiveness. It is also shown that the result does not depend on the fact that agents know the true probability of inspection.