This paper extends the real business cycle model with fiscal policy to allow for endogenous government expenditures and taxes. Fiscal policy in the model is determined by a government that seeks to maximize the welfare of a representative household under the assumption of commitment. On the revenue side, the government chooses an optimal program of distortionary taxes and borrowing in a dynamic version of the Ramsey (1927) optimal tax problem. On the expenditure side, government spending is disaggregated into an investment component that is productive and a consumption component that yields current period utility. The objective is to study the model's predictions for the behavior of the policy variables themselves. In particular, I try to account for the following empirical observations based on detrended post-war U.S. data: 1. Investment in the public-sector is less variable than private-sector investment. 2. Public consumption is more variable than private consumption. 3. The components of public-sector expenditures exhibit low correlations with output, in contrast to the highly procyclical nature of their private-sector counterparts. 4. The tax rate on capital income appears to be more variable than the tax rate on labor income. 5. Tax rates are weakly correlated with output. 6. The government debt-to-output ratio has a high standard deviation relative to output. 7. The government debt ratio exhibits a weak negative correlation with output. I find that a version of the model with multiple stochastic shocks (to technology and preferences) can broadly account for observations 1, 2, 4, 6, and 7. The model partially captures observation 3 but not observation 5.