Optimal Income Taxation with Spillovers from Employer Learning

I study optimal income taxation when human capital investment is imperfectly observable by employers. In my model, Bayesian employer inference about worker productivity drives a wedge between the private and social returns to human capital investment by compressing the wage distribution. The resulting positive externality from worker investment, all else being equal, calls for lower marginal tax rates. To quantify the significance of this externality for optimal taxation, I calibrate my model to match empirical moments from the United States. To inform my calibration, I provide new evidence on how the speed of employer learning about new labor market entrants varies over the worker productivity distribution. Taking into account the spillover from human capital investment introduced by employer inference reduces optimal marginal tax rates by up to 13 percentage points and produces a welfare gain equivalent to raising every worker's consumption by one percent.