Marginal Tax Rates by Income Type

Average Marginal US Income Tax Rates by Income Type

These are dollar weighted average marginal income tax (and deduction subsidy) rates for the US (and/or state if the headline so indicates) Individual Income Tax as calculated by the NBER TAXSIM model from micro data for a sample of US taxpayers. No allowance is made for tax deferred accounts or any other ownership not reported on an individual 1040 form.

Here we provide the average rate for each dollar of income or deduction, not the avererage rate per taxpayer. Because higher income taxpayers will generally have higher rates, these rates will be higher than taxpayer averages but may be preferred for estimating effects that should add up to national totals.

Using each year's data

These tables use the SOI micro data from each year for the calculation of each year's tax rates.
1. Federal rate by year or by state and year(.csv)
2. State rate by year or by state and year(.csv)
3. Federal plus State by year or by state and year(.csv)

   Using 1984 data in/deflated to law year

   These next 3 tables are useful in characterizing the change in the tax law, without conflating changes in the distribution of income.

4. Federal rate by year or by state and year (.csv)
5. State rate by year or by state and year (.csv)
6. Federal plus State by year or by state and year (.csv)

   Using 1984 national data for each state and year

   These next 2 tables are useful in characterizing differences in the state laws across states and times, without conflating changes in the distribution of income through time or differences across states.

7. Federal Rate by year or by state and year (.csv)
8. State Rate by year or by state and year (.csv)
9. Federal plus State by year or by state and year (.csv)

   Using national data for each state

   These next 2 tables are useful in characterizing differences in the state laws across states and times, while allowing for changes in the distribution of income through time.

10. Federal Rate by year or by state and year (.csv)
11. State Rate by year or by state and year (.csv)
12. Federal plus State by year or by state and year (.csv)

This is the 20th version of this table posted at https://www.nber.org/taxsim/marginal-tax-rates, showing results calculated in December of 2024.

The rates are calculated by the method of finite difference. First the tax liability of each eligible return is calculated, then each income or deduction type is increased by 1% of its value and the tax is recalculating under the assumption that other incomes and expenses are constant. The difference in aggregate tax divided by the difference in aggregate income or deduction is the marginal tax rate on the average dollar of that income or deduction type. For some individuals with low or negative AGI the measured marginal rate may be zero even negative (EIC) for a large income on a particular item.

The rates take account of most features of the tax code including the maximum tax, minimum tax, alternative taxes, partial inclusion of social security, earned income credit, phaseouts of the standard deduction and lowest bracket rate, etc. If included, state tax liabilities are calculated to the best of our ability using the data from the federal return. The major missing items at the state level are municipal bond interest, federal interest, income splitting between husbands and wives, itemized deductions for taxpayers who itemize only on the state return etc. Because state of residence for taxpayers with AGI>200,000 is not given in the data, high income taxpayers are assigned randomly to states in proportion to the number of high income taxpayers listed in the SOI annual volumes. For years before 2003 one-third of dividends are assumed to be qualified.

Only taxpayers with positive net capital gains (short plus long) are used in the calculation of the tax rates on gains. This is a change from early versions of this table.

Proprietors income includes Schedules C and F, and partnership income from Schedule E. Taxpayers with non-positive proprietors income are ignored in the calculation of this marginal rate. Because losses are so common, this is more than a detail.

For years when any income item is not broken out in the data the tax rate is shown as ``na''. The subsidy for deduction items is shown as a negative tax.

For charitable gifts, mortgage interest and (real) property tax deductions the result is biased by the absence of any information on non-itemizers. It would be a mistake to assume the bias could be ignored for any particular purpose, as a great many homeowners are not itemizers.

Differences in marginal rates reflect both differences in the tax treatment of different types of income and differences in the (possibly endogenous) functional distribution of income. These ``dollar weighted'' marginal rates are typically higher than ``person weighted'' tax rates would be, but are more appropriate for most analysis of changes in the tax structure.

The micro data are from the Individual Income Tax Models available from the Statistics of Income Division of the Internal Revenue Service. The last year of real data is 2010 but after 2007 there is no state id at all. Therefore we have no state tax calculation for 2008-9. We no longer supply extrapolations beyond the actual data. Sample sizes range from 80,000 to 200,000 actual tax returns, with weighted oversampling of high income returns. Our state tax calculator begins in 1977 but our state data begins in 1979. State ID is imputed for taxpayers with income above 200K.

Federal tax law includes EGGTRA, JGTRRA,TIPRA, ARRA and ATRA and is correct through January 2013, at least (to the best of our knowledge).

For another perspective on calculations of marginal tax rates, you may wish to read the Congressional Budget Office report "Effective Marginal Tax Rates on Labor Income" which, however, covers only 2005. Simplified calculations of marginal tax rates are available in the Tax Policy Center note "Marginal Tax Rates 1955-2005".

Please cite as Feenberg, Daniel, and Elizabeth Coutts,''An Introduction to the Taxsim Model'' Journal of Policy Analysis and Management Vol 12, Number 1, Winter 1993, and this web site (https://www.nber.org/taxsim).

Suggestions and comments are welcome. If you plan to use these numbers in a paper, it would be good to discuss the application with me first. I am not too busy to do this.

Daniel Feenberg
National Bureau of Economic Research
1050 Mass Ave
Cambridge MA 02138
feenberg@nber.org