Optimal Nonlinear Income Taxation with Learning-by-Doing

This paper examines a two-period model of optimal nonlinear income taxation with learning-by-doing, in which second-period wages are an increasing function of first-period labour supply. We consider the cases when the government can and cannot commit to its second-period tax policy. In both cases, the canonical Mirrlees/Stiglitz results regarding optimal marginal tax rates no longer apply. In particular, if the government cannot commit and skill-type information is revealed, it is optimal to distort the high-skill consumer's labour supply downwards through a positive marginal tax rate to relax the incentive-compatibility constraint. Alternatively, if the government cannot commit and skill-type information is concealed, it is optimal to distort the high-skill consumer's labour supply upwards to relax the incentive-compatibility constraint, but due to some other factors at work the high-skill consumer's marginal tax rate cannot be signed. Our analysis therefore identifies a setting in which a positive marginal tax rate on the highest-skill individual can be justified, despite its depressing effect on labour supply and wages.