Optimal Income Taxation: Recent Empirical Applications

In this paper we present a critical survey of two "generations" of studies addressing the empirical computation of optimal income taxes. The first generation, from 1971 up to the late '90s, mainly consists of illustrative numerical exercises rather than of empirical applications. The optimal tax-transfer regime that most commonly emerges from these efforts is a negative income tax + a (almost) flat tax. The second generation is characterized by a more definite focus on policy implementations and, relying on extensions or reformulations of Mirrlees's model, attempts to establish a closer connection between theory and data or econometric estimates. We argue that both generations of studies suffer from taking for granted that the solution to the optimal taxation problem must be an analytical one (a "formula"), to be fed with numerical guesses or estimates. As a first consequence, the theoretical models must adopt very restrictive assumptions in order to generate analytical solutions. As a second consequence, the theoretical results are potentially inconsistent with the empirical estimates that are typically generated under very different (much less restrictive) assumptions. A different approach would consist in avoiding restrictive theoretical assumptions and obtaining the solution computationally by iteratively simulating a microeconometric model. This approach is exemplified here with the computation of optima taxes in Norway and with the evaluation of alternative basic income mechanisms in some European countries.