Robust Benchmarks in Optimal Taxation: From Many People to Many Worlds

• In Event: Social Objectives in Optimal Taxation

Abstract

How should tax policy be designed when there is uncertainty about the key parameters that inform optimal policy rules? We greatly expand the literature on this topic by pointing out that, for many optimal tax problems interest, models where agents exhibit multidimensional heterogeneity can alternatively be interpreted as models of parameter uncertainty. Building on this basic insight, we show that in some cases, parameter uncertainty will lead the planner to gravitate towards simple benchmark results ex ante, even in circumstances where the assumptions used to derive the benchmark are violated ex post. For instance, we show that even if taste-based heterogeneity provides a tagging rationale for differential commodity taxation, as uncertainty about the nature of this heterogeneity grows, the ex ante optimal commodity tax system gets closer to the classic Atkinson-Stiglitz prescription of uniformity. By contrast, in other cases, we show that uncertainty generically complicates ex ante optimal policy, even if ex post the economy satisfies assumptions that would normally yield a simple benchmark result. For example, in a world with parameter uncertainty, externalities which are ex post atmospheric are ex ante non-atmospheric, implying that ex ante optimal corrective taxation deviates from the Pigouvian benchmark. Throughout the paper, we consider both expected welfare and robust (maximin) criteria, showing that the latter can sometimes perfectly rationalize standard benchmark results in the optimal tax literature. Ultimately, the paper helps to clarify which optimal tax benchmarks are reinforced by parameter uncertainty, and which are broken by it.

Authors

• Dylan Moore, dtmoore@hawaii.edu

• Ashley Craig, Australian National University

• Thomas Lloyd, University of Michigan