On the Relevance of Capital Income Taxation

In a standard model of the intertemporal labour supply/consumption/saving decision, with weak separability between consumption and leisure in each period; identical preferences across households; and a Mirrlees optimal nonlinear tax on labour earnings; the Atkinson-Stiglitz Theorem (AS) implies that there is no case for taxing future consumption and therefore the return to saving. An extension of AS by Konishi (1995), Laroque (2005) and Kaplow (2006), (KLK), replaces the assumption of an optimal nonlinear tax system with that of the planner being able to choose any smooth function from gross to net income, and shows that an allocation with both direct and indirect taxation can always be Pareto-dominated by one with direct taxation alone. This in the intertemporal context again implies no taxation of the return to saving. Thus taxation of capital income is purported to be irrelevant, superfluous or non-optimal. This paper is intended to contribute to a recent body of literature1 that contests this view. The assumptions of weak separability, identical preferences and optimal or unconstrained nonlinear taxation are of course strong, and a number of studies have shown that the result, which in policy-related discussions is often expressed2 as "capital income ought to be untaxed", is not robust to their relaxation.3 This paper contributes to this literature by analysing the implications for the form of the optimal taxes on labour earnings and capital income first of introducing household production as a form of time use, along with market work and leisure; secondly, of imposing the constraint that taxation must take the