Distributional Effects of the 2001 and 2003 Tax Cuts: How Do Financing and Behavioral Responses Matter?

Douglas W. Elmendorf,
Douglas W. Elmendorf Former Brookings Expert, Dean - Harvard Kennedy School
Jason Furman,
Jason Furman Aetna Professor of the Practice of Economic Policy - Harvard University, Nonresident Senior Fellow - Peterson Institute for International Economics, Former Brookings Expert
William G. Gale, and
William G. Gale The Arjay and Frances Fearing Miller Chair in Federal Economic Policy, Senior Fellow - Economic Studies, Co-Director - Urban-Brookings Tax Policy Center

Ben Harris

June 30, 2008


Distributional analysis has long been a central element in discussions of tax policy. However, standard methods of estimating the distributional effects of tax changes omit two potentially important factors: the financing of the tax changes, and the implications of behavioral responses for economic growth, incomes, and well-being. In this paper we reexamine the distributional effects of the 2001 and 2003 tax cuts incorporating these two factors. Compared with the standard analysis, this “dynamic distributional analysis” shows that the benefits of these tax cuts were much smaller, on average, and much more skewed toward people with higher incomes.

There is no doubt that the 2001 and 2003 tax cuts will need to be financed in some manner. As the Congressional Budget Office (CBO, 2007) recently reminded policymakers, “… under any plausible scenario, the federal budget is on an unsustainable path.” Therefore, the revenue loss from these tax cuts (after accounting for revenue increases due to feedback effects) will need to be offset by future tax increases or government spending reductions. Ignoring the burden of these financing choices overstates the aggregate benefits of the tax cuts and likely distorts the analysis of the distribution of those benefits as well.

There is also no doubt that the 2001 and 2003 tax cuts will affect economic behavior in ways that are not incorporated in the standard distributional analysis and revenue scoring of tax changes. Of course, there is considerable uncertainty about the magnitude of the behavioral responses and even, after allowing for higher near-term federal debt and future financing, whether the net effect is higher or lower total output and income. But ignoring these effects is not an adequate substitute for considering the sensitivity of estimated distributional effects to different plausible responses.

We present three sets of results regarding the distributional effects of the 2001 and 2003 tax cuts. First, we reproduce familiar tables based on the standard approach to distributional analysis. Using this approach, almost all households are at least as well off after the tax cut, and the biggest percentage and absolute increases in after-tax income go to households with the highest pre-tax incomes.

Second, we add the financing of the tax cuts under two alternative scenarios. In both scenarios, the total amount of financing exactly offsets the tax cuts when fully phased in, so the net effect on the budget is zero. The first scenario assumes that each household pays an equal dollar amount, while the second assumes that each household pays the same percentage of income. Under either scenario, about three-quarters of households are worse off because of the tax cuts, and after-tax income falls for the bottom four quintiles of the income distribution but increases for the top quintile. To be sure, if one assumes that the financing occurs entirely through spending reductions and that the foregone spending is worthless to individuals, then the standard distributional analysis applies. However, despite decades of stump speeches about unnecessary government spending, the political process has been persistently unable to identify significant outlays that voters will blithely forego.

Third, we incorporate behavioral responses, including not only the induced increase in after-tax incomes but also the opportunity cost of the income gains. The central issue is that behaviorally-induced increases in taxable incomes overstate welfare gains because the lost leisure, foregone fringe benefits, and other concomitants of the rise in taxable incomes all generate some decline in well-being. We develop a novel methodology for undertaking welfare- based distributional analysis, beginning with an illustrative example to demonstrate the logic of our approach and then generalizing the formula for broad application. Our results show that the 2001 and 2003 tax cuts raised the well-being of only one-third of households and that more than half of those better-off households are in the top quintile of the income distribution.

These results strongly confirm the importance of undertaking dynamic distributional analysis—that is, of extending the distributional analysis of tax changes to include both the financing of the changes and the welfare consequences of behavioral responses to the changes. These financing and behavioral responses are unavoidable consequences of tax changes, and their incorporation in distributional analyses can significantly alter the results: Excluding these factors, the 2001 and 2003 tax cuts appear to have made most U.S. households better off, albeit with the largest percentage and absolute increases in after-tax income at the top of the income distribution. Including these factors, the 2001 and 2003 tax cuts made most U.S. households worse off, although large percentage and absolute increases in after-tax income are still apparent at the top of the income distribution.

The next section of the paper describes our approach to conducting dynamic distributional analysis, and the following section presents our results. A final section briefly concludes.