André Sztutman

How are optimal taxes affected by reputation building and imperfect information in labor markets? In this paper, I build a model of labor markets with incomplete and asymmetric information where job histories play a crucial role in transmitting information about workers’ productivity, which allows us to better understand the efficiency and distributive consequences of imperfect monitoring and screening in labor markets, and the tradeoffs the government faces when setting taxes. Optimal taxes are described by generalized versions of standard redistributive and corrective taxation formulas, which depend crucially on labor wedges: the ratios of the marginal contribution to output over the increases in lifetime earnings that result from supplying one extra unit of labor at each period. Combining estimates from the literature and new estimates using data from the Health and Retirement Study, I find that, the corrective component of taxes is large, especially at the top of the income distribution.

Suppose an income tax schedule is (constrained) Pareto efficient. We show it may still be suboptimal for utilitarian welfare under all cardinalizations of utilities that admit an upper bound on the curvature of household utility with respect to consumption. Taxes are optimal for some such cardinalization if and only if tax revenues are decreasing and concave with respect to a class of narrowly targeted reforms. We reformulate this condition as a test on sufficient statistics. The test
fails whenever elasticities of taxable income vary enough within some income level. We evaluate our test empirically and find welfare-improving reforms exist.