Two Reasons for not using Commodity Taxation in the Presence of an Optimal Income Tax

Abstract

This paper presents two arguments in favor of the sole use of an income tax for redistribution purposes. The first is that most findings on nonlinear commodity taxation such as Mirrlees’s formulae for Pareto efficiency, Seade’s results on zero tax rates at the endpoints of the scale, and the Corlett-Hague tax rule only become valid under the second-order approach for the unrealistic class of utility functions that are affine with respect to ability. The second is that, for the Atkinson and Stiglitz theorem to hold, it suffices with assuming that preferences must be weakly separable in consumption and ability, a result that also remains valid with linear commodity taxation. All this relegates the use of nonlinear commodity taxes to the unlikely scenario in which, apart from non-weakly separable preferences between consumption and ability, there is a solution to the first-order approach problem.