April 16, 2021, 11:00–12:30
We study the optimal taxation of risky and non-risky capital income when the government imposes an optimal nonlinear earnings tax on heterogeneous households. Households can hold three assets: one risk-free, one risky but diversifiable, and one a private investment with idiosyncratic risk whose expected return differs among households. Contrary to expectations, the optimal tax on excess returns to risky assets is ineffective for redistribution due to the Domar-Musgrave effect. It assumes only an insurance role, and is positive. The optimal tax on risk-free returns fulfills a redistributive role. Its sign depends on the information content of the risk-free returns about the investors' types, beyond what is revealed by the earnings tax base. The optimal nonlinear earnings tax takes the standard Mirrleesian form amended to take account of the stochasticity of capital income tax revenue. (joint with Robin Boadway).