Recent works on optimal income tax found that for unbounded distributions of earnings the optimal tax rate at the top is relatively high (around 60 percent). This finding is puzzling in light of the well-known result for bounded distributions of a zero optimal tax rate at the top. Our paper shows that the more recent papers have used assumptions that favor a high asymptotic tax rate: Pareto instead of Log-normal distribution and linear instead of non-linear utility of consumption. Using these two assumptions along with a logarithmic utility of leisure leads to an optimal rate of 100%, a result that is avoided in recent literature by assuming a constant compensated elasticity of labor. We find that even when using a Pareto distribution of earnings the optimal asymptotic tax rate is about a half compared to recent literature.

Keywords: Optimal Income Tax

JEL Classifications: H21

The full article in PDF file