Abstract: This dissertation studies optimal fiscal policy in an economy with idiosyncratic labor productivity shocks and endogenous borrowing constraints. In contrast with the well-known result of zero capital tax by Chamley (1996), our main finding is that optimal capital tax rate should be positive regardless of market completeness. Alvarez and Jermann (2000) show that in an economy with complete market and borrowing constraints, the market discount rate is higher than preference discount rate. If the government is an unconstrained investor, as in Aiyagari (1996), then the golden rule holds in the long run. As a consequence, the positive capital income taxation is needed to bring the pre-tax capital return back to the rate of time preference. Moreover, by the introduction of endogenous borrowing constraints, the government can use a labor tax as a policy tool to relax the constraints and increase social welfare. In our model, the autarky value is composed of the discounted utilities from the labor incomes. The higher the labor tax, the lower the autarky value. Hence it slacks the endogenous borrowing constraints, Although the increase in labor taxation leads to distortion in production, it effectively helps to relax the borrowing constraints. The Ramsey planner takes this additional benefit of labor taxation into account in a way that balances the social cost and benefit. We interpret this additional benefit as a new role for labor taxation. The later part of this dissertation considers a model of a limited role of the government. We prove that the optimal tax rate on capital income should be positive in steady state and should be increasing over time provided that full risk-sharing is not feasible. Moreover, we show that the optimal tax rate are potentially much bigger than one might expect.