We study Pareto optimal tax and education policies when human capital upon labor market entry is endogenous and individuals face wage uncertainty. Though optimal labor distortions are history-dependent, i.e. depend on income and education, simple policy instruments can yield the desired distortions: a single nonlinear labor income tax schedule combined with income-contingent loans. To take themodel to the (US) data, we simplify the model to a binary education decision (graduating from college or not). We find that for lowand intermediate incomes the labor supply decision of college graduates should be distorted more heavily than for individuals without a college degree. As a consequence, the optimal student loan repayment schedule increases in income for this range. This result holds along the Pareto frontier. We compare the second best to a situation where loan repayment is restricted to be independent from income and find significant welfare gains.