Marginal Tax Rates and Reported Incomes: New Time Series Evidence

This paper estimates the effects of changes in marginal tax rates on reported income for different income groups in the postwar US. A large public finance literature focuses on net-of-tax rate elasticities of reported income because it is indicative of the distortionary effects of taxation. Based on static regressions of income on average marginal tax rates, several recent studies find relatively small elasticities for the top 1% income groups and zero elasticities for other income groups. My estimates are dynamic and account explicitly for the endogeneity of average marginal tax rates. The main findings are (i) that reported incomes respond elastically in the year of a change in tax rates, (ii) that incomes respond also outside the top 1% group and (iii) that the response is larger in the years following the change in marginal rates. These results are based on structural vector autoregressions (SVAR) that allow for dynamic interactions with real GDP, the government budget (debt, spending and revenues), inflation and monetary policy. Unanticipated shocks to net-of-tax rates are identified using a narrative measure of federal tax policy changes using the methodology in Mertens and Ravn (AER forthcoming). Using the SVAR measure of exogenous changes in tax rates as an instrument has a large effect on the results of the static regressions previously considered in the literature: elasticities of reported income rise above 1 and are statistically significant across different income groups, including those below the top 1%. I verify the results for different measures for marginal tax rates and different income concepts: I use the 1960-2000 dataset of Saez (2004), a new dataset constructed from the Statistics of Income that spans 1950-2008, and a recent dataset made available by the CBO that starts in 1979. My empirical findings indicate that marginal tax rate changes have considerable effects on behavior, which has important implications for fiscal policy.