DP12276-2 | Optimal income taxation with composition effects

Publication Date


JEL Code(s)


Programme Area(s)


We study the optimal nonlinear income tax problem with multidimensional individual characteristics on which taxes cannot be conditioned. We obtain an optimal tax formula that generalizes the standard one by averaging the sufficient statistics of individuals who earn the same income. Dependence on the tax schedule is a well-known limitation of sufficient statistics: they have distinct values in the actual and optimal economies, which can bias the recommended tax schedule. This bias is generally considered to be negligible, but, as a first main contribution, we show that multidimensional unobserved individual heterogeneity actually makes it quite substantial. Multidimensional heterogeneity brings a new source of endogeneity to the sufficient statistics (due to changes in average behavioral responses) that we call composition effects. Using U.S. data, we highlight that composition effects substantially affect optimal marginal tax rates. Our results put the stress on the need for more empirical studies on the distribution of sufficient statistics and not only on their means conditional on income. As a second main contribution, we show the equivalence between the tax perturbation and mechanism design approaches which bridges the gap between both methods that have, so far, been used separately in the literature.