Skip Navigation

Effective Federal Tax Rates, 1979-1997 October 2001

The Congressional Budget Office (CBO) uses a multistep methodology to produce its estimates of the distribution of incomes and taxes. The Current Population Survey (CPS) is the primary source of data for CBO's distributional analyses, providing initial estimates of population and incomes. CBO adjusts the incomes reported in the survey to bring them in line with the totals reported by the Internal Revenue Service in its Statistics of Income file. Next, CBO estimates federal taxes for each family in the CPS on the basis of family income, demographic characteristics, and existing laws in the relevant year. CBO then groups the families in the CPS into quintiles, or fifths, on the basis of their incomes and tabulates the incomes and taxes for each quintile.
 

The Current Population Survey

Conducted by the Bureau of the Census, the CPS is a monthly survey of approximately 47,000 households, containing roughly 122,000 individuals. The sample used for the survey is statistically designed to represent the entire U.S. population and many demographic subpopulations. Every March, the CPS adds questions about family income during the previous calendar year, which form the basis for CBO's income estimates for that year. (For example, CBO's 1997 estimates are based on the CPS conducted in March 1998.)⁽¹⁾

The CPS data have three problems that must be corrected before they can be used to estimate the distribution of taxes and incomes. First, the CPS does not collect enough information about individual income taxes to allow CBO to estimate tax liabilities accurately; in particular, the CPS lacks data on itemized deductions and capital gains. Second, high incomes in the CPS are top-coded (that is, capped at given levels) to maintain confidentiality for the families included in the sample. Third, incomes in the survey differ substantially from the amounts reported on tax returns. Tax-return data are generally believed to be more accurate than the incomes reported on surveys.

CBO's adjustments to the CPS income data are based on data in the Statistics of Income (SOI) file created annually by the Internal Revenue Service (IRS). The file is based on a nationally representative sample of more than 100,000 individual income tax returns; it contains detailed information on incomes and taxes. High-income tax returns are oversampled in the SOI file (that is, the proportion of high-income people surveyed is greater than their share of the population), enhancing the richness of the data at the top of the income distribution. CBO adjusts incomes for each year using the relevant SOI file; for example, the adjustments for 1997 come from the SOI file for the 1997 tax year.⁽²⁾

CBO imputes information missing from the CPS data, such as deductions and capital gains, by using the SOI. For example, to impute deductions, CBO first splits families from the CPS into tax-filing units comparable with those in the SOI. Data from the SOI file are then used to determine the probability that a tax unit with particular income and demographic characteristics claims deductions and the ratio of those deductions to income. The tax units developed from the CPS data are then assigned deductions on the basis of those probabilities. A similar procedure is used to impute other missing information.

To adjust for top-coding, CBO uses the SOI to construct a distribution of each type of income among high-income families. It then replaces top-coded incomes from the CPS with a randomly selected amount from the appropriate SOI high-income distribution.

A more complex procedure is necessary to adjust incomes that have been incorrectly reported on the CPS to match incomes in the SOI. Because low-income families are not required to file tax returns, the CPS includes many more tax units than the SOI does. Therefore, CBO compares only incomes from CPS units that are estimated to owe taxes with incomes from the SOI. In addition, for each source of income, it compares data from the CPS with data from the SOI and adjusts the CPS data until the level and distribution of income on the two files are approximately the same. For example, interest and dividends are generally increased for families reporting such income, and new recipients are created; rental income and income from sole proprietorships are generally adjusted downward. As a final step, CBO revises the weights and incomes of taxpaying families in the top 1 percent of the income distribution so that the number of such families and their average income are identical in the two files.
 

Estimating Federal Taxes

Once incomes in the CPS have been adjusted to match the SOI's totals, CBO estimates individual income taxes for each CPS tax-filing unit, using the income tax law for the relevant year. However, for high-income taxpayers, CBO uses data from the SOI to impute taxes to those filing units in the CPS.

CBO estimates payroll taxes--covering Social Security, Medicare, and federal unemployment insurance--for each unit on the basis of two sets of factors: earnings and self-employment income in the income-adjusted CPS, and the relevant tax rates and maximum amounts of income subject to tax for each year. It assumes that both the employee's and employer's shares of payroll taxes are borne by the employee.

Corporate taxes are apportioned to each family in proportion to their adjusted capital income (interest, dividend, capital gains, and rental income), under the assumption that the burden of corporate taxes falls on capital income. For example, a family with 0.001 percent of total capital income is assigned 0.001 percent of overall corporate tax liability.

CBO estimates excise taxes on the basis of a family's spending on the goods subject to those taxes, using data from the Consumer Expenditure Survey (CEX). The CEX, which is a quarterly survey of family income and expenditures conducted by the Bureau of Labor Statistics, divides families into categories on the basis of their income and demographic characteristics.⁽³⁾ For four consecutive quarters, the survey asks families detailed questions about their expenditures on particular items over the previous three months. Those data are then used to compute average spending by each category of family on the goods subject to excise taxes. CBO assumes that each family in the CPS has the same average spending as do similar families with comparable levels of income in the CEX.

CBO further assumes that excise taxes have two distinct effects on consumers: they raise the prices of the commodities subject to tax, and they cause prices in general to increase. Consequently, for each good on which an excise tax is levied, CBO apportions to families (in proportion to their spending on the taxed commodity) the part of excise tax revenues collected in each year that reflects a rise in the price of that good. The remaining revenue--the part that shows up in higher prices overall--is distributed among families in proportion to their total spending.
 

Summarizing the Results

In a series of steps, CBO combines the incomes and taxes of families for each year to create tables showing distributions of incomes and taxes among income groups and types of households. First, it combines families that share a residence into one household record. Second, it groups households by one of several measures of income. For example, cash income is defined on a cash-receipts basis and comprises wages and salaries, net income from a business or farm, taxable and nontaxable interest, dividends, rental income, realized capital gains, cash transfers from the government, and retirement benefits. Some definitions of income also include the imputed values of taxes paid by businesses--the employer's contribution to payroll taxes and federal corporate income taxes--as well as contributions to 401(k) plans. For some analyses, CBO adds the estimated cash value of in-kind benefits, as reported in the CPS, to the income definition. (Those in-kind benefits are food stamps, school lunches and breakfasts, Medicare and Medicaid health insurance, housing assistance, energy assistance, and employer-provided health insurance.) Some definitions of income exclude health-related in-kind benefits (Medicare, Medicaid, and employer-provided health insurance).

To take account of the differing needs of larger and smaller households, CBO adjusts some of the income measures it uses to group households. Many of the tables CBO constructs to display its distributional analyses use a measure of household income that is adjusted for household size; the principal measure divides household income by the square root of family size. Another measure adjusts income according to the equivalence scale implied by the federal poverty thresholds (adjusted income equals household income divided by the relevant threshold).

In the third step, CBO ranks households by the (adjusted) incomes and groups them in quintiles (fifths of the distribution) containing equal numbers of people. CBO then tabulates overall income and taxes for each quintile as well as for smaller groupings at the top of the distribution.
 

Specific Changes from Earlier Estimates

Changes in methodology and databases mean that the estimates presented in this study cannot be directly compared with estimates from CBO's previous analyses of tax rates and incomes. Five differences are most important:

• Income measures now include in-kind income from government transfer programs, such as Food Stamps, rent subsidies, Medicare, and Medicaid, as well as health insurance premiums paid by employers.
  
  
• Both cash income and the payroll tax base include imputed pretax contributions made by families to 401(k)-type retirement funds.
  
  
• Households, rather than families, are the primary unit of analysis.
  
  
• The methodology for estimating excise taxes has changed.
  
  
• Dollar amounts are adjusted for inflation using the Research Series Consumer Price Index, a consistent-methodology index developed by the Bureau of Labor Statistics.

1. For more information about the March supplement to the CPS, see the section on 1999 methodology and documentation, available at www.bls.census.gov/cps/ads/1999/smethdoc.htm.

2. For more information about the SOI, see Internal Revenue Service, Statistics of Income, Individual Income Tax Returns 1997, IRS Pub. 1304 (December 1999).

3. For more information about the Consumer Expenditure Survey, see Bureau of Labor Statistics, Consumer Expenditure Survey, Interview Survey, 1997 (October 1999).

---