Casting Doubts on the Approach of Mirrlees to Optimal Taxation - a Technical Critique : An Abridged Version A Call for a Universal Second Order Approach

We focus exclusively on preferences characterized by the absence of income effects. We argue that the standard approach of Mirrlees to optimal taxation is faulty as it fails to identify the true optimum of the optimal taxation problem. Specifically, we present an elementary, feasible, and implementable mechanism that induces higher welfare than the level of welfare obtained under Mirrlees, Diamond, or Saez who implicitly/explicitly/operationally and willingly restrict themselves to the first order approach and focus on candidate tax functions that do not induce bunching or gaps. We establish our results by considering a tax function that does not compromise revenue, is globally IC compliant, but at the same time it allows for bunching or gaps in the income distribution. Formally, we simply extend the domain by including fully admissible functions that lead to bunching and gaps and by doing so we outperform the outcomes of Mirrlees, Diamond, and Saez. Secondly, we argue that the optimal tax function in the income space must be characterized by infinitely many points in any neighborhood of the highest type where the marginal tax is equal to zero. Furthermore, in light of the results of Hellwig, we conclude that the optimal tax function must induce universal bunching of types at different income levels, i.e., we argue that it is imperative to incorporate the second order conditions explicitly into the optimal taxation problem