Tax Compliance and Rank Dependent Expected Utility

Abstract

Formulating the classic Allingham and Sandmo [1972] tax compliance problem under Rank Dependent Expected Utility (RDEU) provides a simple explanation for the “excess” level of full compliance observed in empirical studies, which standard Expected Utility (EU) theory is unable to explain. RDEU provides a compelling answer to this puzzle, without the need for the moral sentiments or stigma arguments that have recently been advanced in the literature. Formally, we show that the threshold audit probability or penalty rate at which full compliance becomes optimal for the decisionmaker are significantly lower under RDEU axiomatics than in the EU case, and that the optimal level of underreporting is lower under RDEU. Numerical simulations using various parameterizations of the probability weighting function illustrate the large quantitative differences between the two models, while a simulation of underreporting rates in the US over the past 50 years shows how RDEU can go some way towards explaining the tax-compliance puzzle.