VoxEU Column Taxation

Income taxation of US households: Facts and parametric estimates

Income tax is always a controversial issue in the US, particularly in election year. But how progressive is the US income tax system? This column looks at data on over 100,000 taxpayers from the Internal Revenue Service 2000 Public Use Tax File. It finds the question has more than one answer.

Income distribution in the US is subject to intense policy interest and cutting edge research (Piketty et al. 2011). Furthermore, empirical properties of the distributions of earnings, income and wealth are key ingredients for a substantial and growing literature in macroeconomics that uses dynamic, equilibrium models with household heterogeneity (see Heathcote, Storesletten and Violante 2009 for a survey). In contrast, the distribution of taxes paid by households has received much less attention. How do federal income taxes paid by a cross-section of US households vary according to their income, marital status, and number of dependent children? What is the relation between household income and effective taxes paid? How progressive is the US income tax system?

In our recent work, we use data from the Internal Revenue Service 2000 Public Use Tax File to answer these questions, and provide ready-to-use parametric estimates of tax functions for applied work (Guner et al. 2012a). With 145,663 records, it is a representative subsample of the universe of US taxpayers who filed taxes in the year 2000. Since this data effectively contains no restrictions on income, either at the bottom or at the top, it allows for a comprehensive representation of income and tax liabilities. The notion of income that we use is standard in cross-sectional studies and encompasses all income flows accruing to households; labour income, asset income from different sources, and transfers. Our notion of federal income taxes is comprehensive as well. It corresponds to total income taxes owed after credits (including the Earned-Income Tax credit). From this notion of tax liabilities, we calculate effective average tax rates. We find that:

1. Taxes and income. Effective average rates increase substantially as income increases. Increasing household income from the central quintile (40% to 60%) to the top quintile, increases the mean, effectively-paid average tax rate, from about 3.9% to 14.0% for married households, and from 8.2% to 16.7% for unmarried households. In terms of statutory marginal rates, the increase goes from 13.5% to 27.7% for married households, and from 16.2% to 28.6% for unmarried ones.1  For households at the top 1%, average (marginal) rates are 23.1% (36.3%) for married households and 21.7% (34.6%) for unmarried ones.

2. Taxes and marital status. The US tax system does not treat individuals with different marital status symmetrically – the household, not the individual is the unit subject to taxation in the law. Not surprisingly, there are significant differences in average rates between married and unmarried households. At low levels of income, effective rates are substantially higher for unmarried households, while these rates subsequently converge as income increases, and eventually become higher for married households at top levels of income. These differences are due to a host of factors; differences in the levels of income concentration between married and unmarried households, differences in standard deductions and differences in the width of tax brackets, as well as the presence of children. These latter factors are arguably more important in reducing average rates at lower levels of income. For instance, children are concentrated in married households and they lead to higher personal exemptions and tax credits, thereby reducing average rates for these households. Figure 1 summarises the relationship between taxes and income, conditional on marital status.

Figure 1. Average tax rates

3. Taxes and children. For married households, children reduce effective rates although the overall effect varies across income levels. Households with income at the top 20%, face an effective average rate of about 15.0% when no children are present, a rate of about 13.2% when two children are present, and a rate of 11.7% when more than two children are present. For unmarried households, the patterns just described above are similar but more pronounced; households at the top 20%, face an effective average rate of about 17.3% when no children are present, a rate of about 13.5% when two children are present, and a rate of 12.2% when more than two children are present.

4. State and local taxes. On average, state and local taxes amount to about 4%-5%. Moreover, the overall structure of state and local taxes is rather flat as a function of income.

5. The distribution of taxes paid. Who pays? A substantial fraction of households has no tax liabilities: this occurs for about 14.5% of the married group and for about 31.8% of the unmarried one. Median and mean effective tax rates are on the low side for both groups, with a median rate for married households of about 8.5% and a mean rate for married households of about 8.8%. For unmarried households, the median rate is of about 6.1% whereas the mean rate amounts to 6.4%. As documented in Table 1, the top 20% of households earns about 61.3% of total income and pays more than three quarters of total taxes (79.1%). Similarly, the top 1% earns about 20.9% of total income, yet it accounts for about 35.8% of total tax collections.

Table 1. Distribution of income and tax liabilities

Parametric estimates

Using our data, we estimate different functional forms for effective tax rates, with the aim of providing researchers a workable representation of taxes in cross section. In each case, we report estimates for all households, as well as for married and unmarried households, with different numbers of dependent children. We first estimate a two-parameter specification, used by Guner et al. (2012-b, 2012-c), which we refer to as the log-specification. Our second set of estimates, another two-parameter specification, pertains to the functional form in Heathcote et al. (2011), which we refer to as the HSV specification. Our third case is a three-parameter specification, which we refer to as the Power specification (a version of power function is used by Guvenen et al. 2009). Finally, we estimate the same functional form used in Gouveia and Strauss (1994), who provided estimates of tax functions for all taxpayers using the US tax structure prevailing in 1989. Figure 2 plots the resulting average tax rates under all specifications for the universe of married households, alongside with data. The figure shows that the resulting shape of average tax rates are similar under all cases; all track the shape of average rates at most income levels. Indeed, the fit is good even at high levels of income.

Figure 2. Average tax rates for married households (data and the parametric estimates)

How progressive is the US tax system?

There are probably two answers to this question. One the one hand, a clear picture emerges from our findings. Effective tax rates on most households are relatively low (below 10%) and differ substantially from those at the top. For instance, married household around median income experience tax rates around 4%, while those at the top 1% face tax rates of around 23%. Furthermore, taxes paid are concentrated at the top. In a nutshell, the provisions in the law, in conjunction with the observed dispersion in income lead to the finding that the bulk of tax payments are concentrated in upper income households and that a large fraction of US households have effectively no tax liabilities. From this perspective, the answer to the question above is that there is substantial progressivity in the tax burden as measured by effective, average tax rates. Put differently and in plain terms, moving a hypothetical household along the income ladder implies substantial increases in average tax rates.

On the other hand, tax rates at the top of the income distribution are essentially constant as income changes. Once high income levels are reached, effective tax rates do not change. We calculate that at ten times the level of household income in the sample (about $530,630 in 2000 dollars), the average tax rate for a married household was around 24%. At fifteen times the level of mean income (about $795,945) the rate was about 25%, while at twenty times mean household income (about $1,061,260), the rate effectively unchanged at 25%. In other words, income taxes at the top are flat and do not approach the statutory marginal rate on income that these households faced (39.6%). Why this occurs should motivate debates and proposals on tax reform.


Gouvieia, Miguel and Robert Strauss (1994), “Effective Federal Individual Income Tax Functions: An Exploratory Analysis”, National Tax Journal, 47(2):317-339.

Guner, Nezih, Remzi Kaygusuz, and Gustavo Ventura (2012a), “Income Taxation of U.S. Households: Facts and Parametric Estimates”, CEPR Discussion Paper 9078.

Guner, Nezih, Remzi Kaygusuz, and Gustavo Ventura (2012b), “Taxing Women: A Macroeconomic Analysis”, Journal of Monetary Economics, 59(1):111-128.

Guner, Nezih, Remzi Kaygusuz, and Gustavo Ventura (2012c), “Taxation and Household Labor Supply”, Review of Economic Studies, 79(3):1113-1149.

Guvenen, Fatih, Burhanettin Kuruscu, and Serdar Ozkan (2009), “Taxation of Human Capita and Wage Inequality: A Cross-Country Analysis”, mimeo.

Heathcote, Jonathan, Kjetil Storesletten, and Giovanni Violante (2009), “Quantitative Macroeconomics with Heterogeneous Households”, Annual Review of Economics, 1:319-354.

Heathcote, Jonathan, Kjetil Storesletten, and Giovanni Violante (2011), “Redistributive Taxation in a Partial-Insurance Economy”, mimeo.

Piketty, Thomas, Emmanuel Saez, and Stefanie Stantcheva (2011), “Taxing the 1%: Why the top tax rate could be over 80%”, VoxEU.org, 8 December.

1 By marginal rates, we mean the average, statutory marginal rate encountered by households in their tax filing.

5,039 Reads