The Congressional Budget Office regularly analyzes the distribution of income in the United States and how it has changed over time. This report presents the distributions of household income, means-tested transfers, and federal taxes between 1979 and 2019 (the most recent year for which tax data were available when this analysis was conducted).
- Income. Households at the top of the income distribution received significantly more income than households at the bottom. Between 1979 and 2019, average income, both before and after means-tested transfers and federal taxes, grew for all quintiles (or fifths) of the income distribution, but it increased most among households in the highest quintile.
- Means-Tested Transfers. Means-tested transfers are cash payments and in-kind benefits from federal, state, and local governments that are designed to assist individuals and families who have low income and few assets. Between 1979 and 2019, households in the lowest quintile received more than half of all means-tested transfers. As a percentage of income before transfers and taxes, means-tested transfers rose over the 41-year period, primarily driven by an increase in Medicaid spending.
- Federal Taxes. Higher-income households typically paid a higher average federal tax rate than lower-income households. Average federal tax rates fell between 1979 and 2019 across the income distribution, with the sharpest decline in the lowest quintile.
- Income Inequality. Income inequality, as measured by the Gini coefficients for income both before and after transfers and taxes, rose between 1979 and 2019. (The Gini coefficient is a standard measure of income inequality that summarizes an entire distribution in a single number that ranges from zero to one.) The degree to which transfers and taxes reduced income inequality increased over that same period.
Numbers in the text, figure, table, and exhibits may not add up to totals because of rounding.
Unless this report indicates otherwise, all years referred to are calendar years.
All dollar amounts are expressed in 2019 dollars and are rounded to the nearest hundred. To convert dollar amounts to 2019 dollars, the Congressional Budget Office used the price index for personal consumption expenditures from the Bureau of Economic Analysis.
Unless this report indicates otherwise, “income” refers to household income before accounting for means-tested transfers and federal taxes, “transfers” refers to means-tested transfers, and “taxes” refers to federal taxes. For additional definitions, see Appendix B.
Specific colors have been used to represent certain income concepts in the exhibits and Figure S-1: Green denotes income before transfers and taxes, blue denotes means-tested transfers, orange denotes federal taxes, and purple denotes income after transfers and taxes.
Supplemental data, additional data for researchers, and a table builder are posted along with this report on CBO’s website (). The supplemental data and the additional data for researchers present detailed information on income, means-tested transfers, federal taxes, and household types.
In 2019, household income was unevenly distributed among the roughly 128 million households in the United States, which received a total of about $15.4 trillion in annual income, the Congressional Budget Office estimates. The agency estimates that the average income among households in the highest quintile (or fifth) of the income distribution was 14 times the average income of households in the lowest quintile:
- Average income before means-tested transfers and federal taxes among households in the lowest quintile of the income distribution was about $23,800.
- Average income before transfers and taxes among households in the highest quintile was about $332,100.
Furthermore, income within the highest quintile was skewed toward the very top of the distribution: Average income before transfers and taxes among households in the bottom half of the highest quintile (the 81st to 90th percentiles) was about $180,100; average income among the 1.2 million households in the top 1 percent of the distribution was about $2.0 million; and average income among the approximately 12,000 households in the top 0.01 percent of the distribution was about $43.0 million.
Income before transfers and taxes consists of market income and social insurance benefits (such as benefits from Social Security and Medicare) and excludes means-tested transfers and federal taxes. Means-tested transfers are cash payments and in-kind benefits from federal, state, and local governments that are designed to assist individuals and families who have low income and few assets. They include benefits from government programs such as Medicaid and the Children’s Health Insurance Program (CHIP), the Supplemental Nutrition Assistance Program (SNAP, formerly known as the Food Stamp program), and Supplemental Security Income (SSI). Federal taxes consist of individual income taxes (net of refundable tax credits, such as the earned income tax credit and the child tax credit), payroll taxes, corporate income taxes, and excise taxes.
For this report, CBO focused on the distribution of household income in 2019, the most recent year for which relevant data from tax returns were available. In addition, CBO assessed trends in household income, means-tested transfers, federal taxes, and income inequality over the 41-year period beginning in 1979 and ending in 2019.
Many households experience changes in their income, transfers, taxes, or household composition from year to year. As a result, the households in any given group of the income distribution in 2019 do not represent the same households that were in that group in prior years. Therefore, this analysis focuses on the changes in the overall distribution of household income rather than the experiences of particular households.
Federal fiscal policies have significant effects on the economic resources available to U.S. households. Before means-tested transfers and federal taxes are taken into account, average income among all households in 2019 was $119,600, CBO estimates. Means-tested transfers provided households with an additional $5,900 in income, on average, in that year. Federal taxes amounted to $23,200 per household, on average. The net effect of means-tested transfers and federal taxes was to decrease household income by $17,300, on average, bringing average household income after transfers and taxes to $102,400 in 2019.
Those averages, however, obscure a significant amount of variation in household income and in how means-tested transfers and federal taxes affect income. In 2019, means-tested transfers and federal taxes caused household income to be more evenly distributed (see Figure S-1, upper panel). On average, households in the lower quintiles received more in transfers than they paid in taxes, while households at the higher end of the income distribution paid more in taxes than they received in transfers. Those transfers and taxes had the following effects:
- They increased income among households in the lowest quintile by $15,100 (or 64 percent), on average, to $38,900; and
- They decreased income among households in the highest quintile by $80,000 (or 24 percent), on average, to $252,100.
Furthermore, within that highest quintile, income after transfers and taxes was significantly higher among households at the top of the distribution. Among households in the 81st to 90th percentiles, transfers and taxes reduced income by $34,800, on average, to $145,300. They decreased income by about $600,000, on average, in the top 1 percent of the distribution, to $1.4 million. Among households in the top 0.01 percent of the distribution, they reduced income by $13.0 million, on average, to $30.0 million. Although transfers and taxes reduce income inequality, income after transfers and taxes remains skewed toward the top of the income distribution.
Average Income, Means-Tested Transfers, and Federal Taxes, 2019, and Cumulative Growth in Average Income, 1979 to 2019
In 2019, the average means-tested transfer rate among all households was about 5 percent, CBO estimates—that is, in total, means-tested transfers received by households were equal to 5 percent of all income before transfers and taxes. However, the average rate varied significantly by income group. Among households in the lowest quintile of the income distribution (ranked by income before transfers and taxes), the average means-tested transfer rate was about 64 percent; among households in the middle quintile, the average rate was about 4 percent; and among households in the highest quintile, the average rate was less than one-half of one percent.
In 2019, the average federal tax rate (based on tax liabilities incurred during that calendar year) also varied significantly by income group. Among all households it was about 19 percent, CBO estimates. Among households in the lowest quintile, the average rate was 0.5 percent, on net; in the middle quintile it was about 13 percent; and in the highest quintile it was about 24 percent. The average federal tax rate among households in the top 1 percent of the income distribution in 2019 was about 30 percent.
Thus, both means-tested transfers and federal taxes are progressive—that is, low-income households receive a larger share of their income as means-tested transfers than high-income households do, and high-income households pay a larger share of their income in federal taxes than low-income households do. In 2019, means-tested transfers went overwhelmingly to low-income households—just over half of such transfers went to households in the lowest income quintile, and nearly a quarter went to households in the second quintile.
Not all households receive means-tested transfers, but virtually all households pay federal taxes in some form (individual income taxes, payroll taxes, corporate taxes, or excise taxes). However, some households near the lower end of the income distribution have net negative average federal tax rates—that is, refundable tax credits exceed the payroll taxes, corporate taxes, and excise taxes paid by those households.
Households at the top of the income distribution pay the majority of federal taxes. Households in the highest income quintile, which received about 55 percent of all income, paid more than two-thirds of all federal taxes in 2019, CBO estimates. In contrast, households in the lowest quintile, which received about 4 percent of all income, paid about 0.1 percent of federal taxes, on net, in that year.
Because of the progressive structure of means-tested transfers and federal taxes, the distribution of income after transfers and taxes was more even than the distribution of income before transfers and taxes. In 2019, those transfers and taxes boosted the lowest quintile’s share of total income by nearly 4 percentage points, CBO estimates. In contrast, among households in the highest quintile, the share of income after transfers and taxes was roughly 6 percentage points lower than the share of income before transfers and taxes.
According to CBO’s estimates, between 1979 and 2019, average household income before transfers and taxes grew more among households at the top of the income distribution than among those at the bottom. Among households in the highest quintile, average real income (that is, income adjusted to remove the effects of inflation) in 2019 was 114 percent higher than it was in 1979. In contrast, among households in the lowest quintile, average income before transfers and taxes was 45 percent greater in 2019 than in 1979, and among households in the middle three quintiles, it was 43 percent greater in 2019 than in 1979 (see Figure S-1, lower panel). Because of those differences in cumulative growth rates, income inequality was greater in 2019 than it was in 1979.
From 1979 to 2019, among households in the lowest income quintile, cumulative growth in income after transfers and taxes was greater than cumulative growth in income before transfers and taxes—94 percent versus 45 percent. That faster growth is attributable both to an increase in spending on Medicaid and CHIP and to a reduction in federal taxes—the latter largely the result of increased refundable tax credits provided through the individual income tax.
Among households in the middle three income quintiles (the 21st to 80th percentiles), the expansion of means-tested transfers and generally declining average federal tax rates had a similar effect. Specifically, cumulative growth in income after transfers and taxes was larger for those groups than it was before transfers and taxes—59 percent versus 43 percent.
In the highest quintile, cumulative growth in income after transfers and taxes was also greater than growth in income before transfers and taxes—123 percent versus 114 percent. Households in the top 1 percent of the income distribution experienced the largest cumulative growth in income after transfers and taxes. In 2019, real income after transfers and taxes for that income group was 262 percent greater than it was in 1979, CBO estimates.
Overall, transfer programs and the tax system reduced income inequality by more in 2019 than they did in 1979. Consequently, inequality of income after transfers and taxes increased by less than inequality of income before transfers and taxes.
This report incorporates new tax data for 2019—the year preceding the onset of the coronavirus pandemic. Because those data are based on tax returns filed in 2020, they may have been affected by the pandemic and by subsequent delays to tax-filing deadlines.
The report does not show the effects of federal programs, such as economic impact payments and expanded unemployment payments, that were enacted in response to the pandemic. Those programs will be addressed in The Distribution of Household Income, 2020, which CBO will publish once the relevant tax data are available.
In 2019, income growth was strong across the five quintiles, with each quintile reaching record levels of income before transfers and taxes. Income gains were larger for lower-income households than for higher-income households: Specifically, the lowest quintile’s income grew by 7 percent between 2018 and 2019. By contrast, income in the middle quintile grew by 4 percent and, in the top quintile, by 1 percent. Among households in the top 1 percent of the income distribution, however, income before transfers and taxes declined for the second straight year, remaining below its 2007 peak.
As a result of strong income growth, means-tested transfer rates declined in the lowest quintiles, which receive most of such transfer income. Average tax rates, which continued to reflect the effects of Public Law 115-97 (referred to as the 2017 tax act throughout this report), rose among the bottom three quintiles by 0.3 percentage points to 0.7 percentage points. Those increases were driven by increases in individual income tax rates, reflecting higher levels of income. Meanwhile, tax rates were constant in the top two quintiles.
Higher average tax rates and lower transfer rates tempered some of the growth in income for the bottom two quintiles. Growth in income after transfers and taxes ranged from 3 percent to 4 percent for each of the bottom four quintiles, whereas the top quintile’s income after transfers and taxes grew by 2 percent.
Income before transfers and taxes consists of market income plus social insurance benefits. Market income comprises wages and other forms of labor income (including cash wages, employers’ contributions for health insurance premiums, and payroll taxes paid by employers), business income, capital income (including capital gains), and other income sources. Social insurance benefits include Social Security and Medicare benefits, unemployment insurance, and workers’ compensation. Notably, income before transfers and taxes excludes the effects of government policies carried out through means-tested transfer programs or the federal tax system.
Income before transfers and taxes is skewed toward households at the top of the income distribution. As a result, those households receive a substantial share of income before transfers and taxes.
The composition of income before transfers and taxes varies throughout the distribution. For most households, labor income is the majority of income before transfers and taxes. But among households at the top of the income distribution, capital income constitutes a greater portion of income before transfers and taxes than it does for the rest of households. Additionally, as income rises, social insurance benefits tend to decline as a share of income.
Between 1979 and 2019, income before transfers and taxes grew faster in real terms among households in the highest quintile of the distribution than among households in the lower quintiles. As a result, the share of income before transfers and taxes received by the highest income quintile increased over that 41-year period.
Cumulative Growth in Income Before Transfers and Taxes Among Households in the Highest Quintile, 1979 to 2019
Composition of Income Before Transfers and Taxes Among Households in the Top 1 Percent, 1979 to 2019
Millions of Dollars
Means-tested transfers are cash payments and in-kind benefits from federal, state, and local governments that are designed to assist individuals and families who have low income and few assets. This analysis focuses on the average means-tested transfer rate, which is the ratio of average means-tested transfers to average income before transfers and taxes in a given income group.
Means-tested transfers go largely to households near the bottom of the income distribution. In 2019, more than half of means-tested transfers went to households in the lowest quintile. Between 1979 and 2019, means-tested transfer rates doubled among households in that quintile—growth that is attributable both to increases in the number of people receiving benefits and increases in the average cost of those benefits per recipient.
Eligibility for some means-tested transfer programs has expanded since 1979. Consequently, means-tested transfers provided to individuals and families in the second and the middle income quintiles increased over the 1979–2019 period.
Over that 41-year period, growth in means-tested transfer rates was primarily driven by spending on Medicaid, which was the largest—and fastest growing—means-tested transfer program. During that time, the number of people enrolled in Medicaid or CHIP increased almost fivefold, from about 20 million in 1979 to 95 million in 2019. Furthermore, the average benefit per recipient (in 2019 dollars) increased from $1,800 in 1979 to $5,900 in 2019. Meanwhile, total spending on SNAP, SSI, and other transfers was roughly constant as a share of income between 1979 and 2019.
Average Means-Tested Transfer Rates Among Households in the Lowest Quintile, by Type of Transfer, 1979 to 2019
In this analysis, federal taxes consist of individual income taxes, payroll taxes, corporate income taxes, and excise taxes. The taxes allocated to households in the analysis account for approximately 93 percent of all federal revenues collected in 2019. Individual income taxes and payroll taxes are the largest sources of tax revenues, followed by corporate taxes and excise taxes. CBO’s examination of household income focuses on the average federal tax rate, which is calculated by dividing total federal taxes in an income group by total income before transfers and taxes in that group.
Average federal tax rates generally rise with income. Households in the highest income quintile, which received about 55 percent of all income in 2019, paid more than two-thirds of federal taxes that year. In contrast, households in the lowest quintile, which received about 4 percent of all income, paid about 0.1 percent of federal taxes, on net, that year. Among households in the lowest two quintiles, individual income taxes are negative, on average, because they include refundable tax credits, which can result in net payments from the government.
Year-to-year fluctuations in average federal tax rates are caused both by underlying changes in the income distribution and by legislative changes to federal tax rules. For most income groups, the average federal tax rate fell over the 41-year period analyzed here; the lowest income quintile experienced the sharpest decrease. The average federal tax rate among households in the middle of the income distribution also decreased but not as much as it did among households in the lowest quintile. In contrast, the average federal tax rate for households in the 81st to 99th percentiles of the income distribution was relatively stable over the 1979–2019 period. The average rate for the top 1 percent of the distribution was significantly more volatile than the average rate for other income groups.
Income after transfers and taxes is income before transfers and taxes plus means-tested transfers minus federal taxes. Because of the progressivity of means-tested transfers and federal taxes (driven primarily by the size and structure of the individual income tax), income after transfers and taxes is less skewed toward households at the top of the distribution than income before transfers and taxes. From 1979 to 2019, income after transfers and taxes grew more evenly across the income distribution than income before transfers and taxes.
The average income after transfers and taxes of households in different income groups grew at different rates because of changes in means-tested transfer programs, federal tax laws, and economic conditions. Income grew significantly faster among households in the highest quintile than among all other income groups, mainly because of changes in income before transfers and taxes.
Cumulative Growth in Income After Transfers and Taxes Among Households in the Highest Quintile, 1979 to 2019
As the distribution of income shifted in the United States between 1979 and 2019, so did the degree of income inequality. A standard measure of income inequality is the Gini coefficient, which summarizes an entire distribution in a single number that ranges from zero to one. At the theoretical extremes, a value of zero means that income is distributed equally among all income groups, whereas a value of one indicates that all income is received by the highest income group, and none is received by any of the lower income groups.
The Gini coefficient can also be interpreted as a measure of one-half of the average difference in income between every pair of households in the population, divided by the average income of the total population. For example, the Gini coefficient based on income before transfers and taxes—which was 0.516 in 2019—indicates that the average difference in income before transfers and taxes between pairs of households in that year was equal to 103.2 percent (twice 0.516) of average household income, or about $80,800 (adjusted to account for differences in household size).
CBO’s analysis compares Gini coefficients based on four different income measures: market income, income before transfers and taxes, income after transfers but before taxes, and income after transfers and taxes. Social insurance benefits, transfers, and taxes tend to reduce income inequality as measured by the Gini coefficient. Still, the Gini coefficients based on each of the four income measures indicate a rise in income inequality between 1979 and 2019; changes in the distribution of market income caused much of that increase.
The degree to which means-tested transfers and federal taxes reduce income inequality can be measured by the difference between the Gini coefficient for income before transfers and taxes and the Gini coefficient for income after transfers and taxes. That difference has fluctuated over time, as average federal tax rates and means-tested transfer rates have changed. But overall, the degree to which income inequality was reduced by transfers and taxes increased between 1979 and 2019.
Change in Gini Coefficient
1. In this report, CBO estimates that 315 million people lived in those households. The agency’s estimate of the U.S. population excludes members of the armed forces on active duty and people in institutions such as prisons or nursing homes.
2. Each quintile of the income distribution contains approximately the same number of people but slightly different numbers of households.
3. Market income comprises labor income (including cash wages, employers’ contributions for health insurance premiums, and payroll taxes paid by employers), business income, capital income (including realized capital gains), and income from other nongovernmental sources.
4. Although data from tax returns include information on tax filers’ family structure and age, they do not include information about people’s race, ethnicity, or education. The supplemental data posted along with this report include additional distributional data for three types of households: households headed by elderly people; households with children; and nonelderly, childless households. The additional data, broken out by household type, are reported for each income group. The supplemental data are available at .
5. Annual income is only one measure of economic well-being. In this report, CBO does not assess trends in the distributions of other measures of economic well-being, such as household income measured over a longer period, household consumption, or household wealth. Nor does this report analyze the considerable variation in income, taxes paid, and tax rates within each income group, which cannot be captured by calculating averages alone.
6. Much research has been conducted on the related topic of economic mobility. For a comprehensive overview of that research, see Federal Reserve Bank of St. Louis and the Board of Governors of the Federal Reserve System, Economic Mobility: Research and Ideas on Strengthening Families, Communities, and the Economy (2016), . See also Katharine Bradbury, Family Characteristics and Macroeconomic Factors in U.S. Intragenerational Family Income Mobility, 1978–2014, Opportunity and Inclusive Growth Institute System Working Paper 19-08 (Federal Reserve Bank of Minneapolis, October 2019), (PDF).
7. In addition to the federal government’s fiscal (tax and spending) policies, its monetary, regulatory, and trade policies affect the distribution of household income. The direct distributional effects of those other federal policies, however, are not examined in this report. Although some state-level means-tested transfers are included in this analysis, most state and local fiscal policies are also not examined here.
8. In this analysis, CBO classified means-tested transfers in four categories: Medicaid and the Children’s Health Insurance Program, the Supplemental Nutrition Assistance Program, Supplemental Security Income, and other means-tested transfers. The other means-tested transfers that are analyzed in this report are housing assistance programs, low-income subsidies for Part D of Medicare (which covers prescription drugs), Temporary Assistance for Needy Families, child nutrition programs, cost-sharing reductions under the Affordable Care Act, the Low Income Home Energy Assistance Program, and state and local government general assistance programs.
9. Although means-tested transfers are designed to assist people with low income, the data indicate that some high-income households receive benefits from the transfer programs. That may happen for several reasons. For example, some people have income that varies during the year and may therefore qualify for benefits on the basis of low monthly income even though their annual income is high. In addition, some people who qualify for benefits because their own income is low live in high-income households. Finally, a portion of the benefits reported as going to higher-income households probably reflects some misreporting of income, program participation, and benefit amounts in the survey data that underlie CBO’s estimates.
10. CBO’s estimates represent the number of recipients who were ever enrolled in Medicaid or CHIP in a given calendar year. Furthermore, the estimates apply to the noninstitutionalized population; they do not include recipients living in nursing homes and other long-term care facilities. The CHIP program began in 1998.
11. The value of Medicaid and CHIP benefits allocated to households is based on the average cost to the government of providing those benefits. CBO did not attempt to estimate the value that households place on those benefits. Although sick people enrolled in federal health programs that provide assistance to low-income families may value those benefits more than the average cost to the government of providing them, some empirical evidence suggests that, on average, Medicaid recipients value the benefits at less than the average cost to the government of providing those benefits. See Amy Finkelstein, Nathaniel Hendren, and Erzo F. P. Luttmer, “The Value of Medicaid: Interpreting Results From the Oregon Health Insurance Experiment,” Journal of Political Economy, vol. 127, no. 6 (December 2019), pp. 2836–2874, .
12. The remaining federal revenue sources not allocated to U.S. households include states’ deposits for unemployment insurance, estate and gift taxes, net income earned by the Federal Reserve System, customs duties, and miscellaneous fees and fines. Because of the complexity of estimating state and local taxes for individual households, this report considers federal taxes only. Researchers differ about whether state and local taxes are, on net, regressive, proportional, or slightly progressive, but most agree that state and local taxes are less progressive than federal taxes. For estimates of the distribution of state and local taxes, see Meg Wiehe and others, Who Pays? A Distributional Analysis of the Tax Systems in All 50 States, 6th ed. (Institute on Taxation and Economic Policy, October 2018), ; and Gerald Prante and Scott Hodge, The Distribution of Tax and Spending Policies in the United States, Special Report 211 (Tax Foundation, November 2013), (PDF).
13. Federal taxes allocated to households in this analysis are based on tax liabilities incurred in calendar year 2019.
14. For information about the distribution of tax expenditures in 2019, see Congressional Budget Office, The Distribution of Major Tax Expenditures in 2019 (October 2021), .
15. In the federal budget, the portion of refundable credits that reduces the amount of taxes owed is counted as a reduction in revenues, and the portion that exceeds a filer’s tax liability is treated as an outlay. In the analysis presented here, CBO treated the refundable and nonrefundable portions of the credit jointly. For more details about the history and economic effects of refundable tax credits, see Congressional Budget Office, Refundable Tax Credits (January 2013), .
16. For information about how the 2017 tax act affected the distribution of federal taxes in 2018, see “The Distributional Effects of the 2017 Tax Act in 2018” in Congressional Budget Office, The Distribution of Household Income, 2018 (August 2021), pp. 24–29,.
17. A significant body of research has examined changes in U.S. income inequality over time using various data sources and measures of income. For recent examples, see Thomas Piketty, Emmanuel Saez, and Gabriel Zucman, “Distributional National Accounts: Methods and Estimates for the United States,” Quarterly Journal of Economics, vol. 133, no. 2 (May 2018), pp. 553–609, ; and Gerald Auten and David Splinter, “Income Inequality in the United States: Using Tax Data to Measure Long-Term Trends” (February 2022), (PDF).
The Congressional Budget Office has released its analyses of the distribution of household income and federal taxes on a regular basis for more than 30 years. This appendix provides additional details about CBO’s analytic method and the most important assumptions underlying its analyses. The estimates in this report were produced using the agency’s framework for analyzing the distributional effects of both means-tested transfers and federal taxes. That framework uses income before transfers and taxes, which consists of market income plus social insurance benefits. That measure is used to rank households when creating income groups and serves as the denominator when calculating average means-tested transfer rates and average federal tax rates.
CBO uses households as the unit of analysis in its distributional reports. A household consists of the people who share a housing unit regardless of their relationship. The data used in CBO’s analyses come from two primary sources: One provides data on tax-filing units, and the other provides household-level data. A household can consist of more than one tax-filing unit, such as a married couple and their adult child.
To incorporate data on tax-filing units into the analysis, the agency creates tax-filing units from the household-level data on the basis of the relationship and income information collected by household surveys. After both data sources are organized using the same unit of analysis (tax-filing units), they are statistically matched to create a database with information from both sources (see the next section in this appendix for details about the statistical matching method). For the final presentation of distributional results, data for those statistically matched tax-filing units are combined and represented at the household level.
The core data used in CBO’s distributional analyses come from the Statistics of Income (SOI), a nationally representative sample of individual income tax returns collected by the Internal Revenue Service (IRS). The number of returns sampled grew over the period studied—1979 to 2019—rising from roughly 90,000 in some of the early years to more than 350,000 in later years. That sample of tax returns becomes available to CBO approximately two years after the returns are filed. The sample of tax returns for 2019 was delayed because of disruptions stemming from the coronavirus pandemic. The filing deadline for 2019 tax returns was extended from April 15, 2020, to July 15, 2020. In addition, the IRS experienced delays in processing returns that were filed on paper.
Information from tax returns is supplemented with data from the Annual Social and Economic Supplement of the Census Bureau’s Current Population Survey (CPS), which contains survey data on the demographic characteristics and income of a large sample of households. The two sources are combined by statistically matching each SOI record to a corresponding CPS record on the basis of demographic characteristics and income. Each pairing results in a new record that takes on some characteristics of the CPS record and some characteristics of the SOI record.
The first step in the statistical matching process is to align the unit of analysis by constructing tax-filing units from CPS households. A tax-filing unit is a single person or a married couple plus any dependents. In CBO’s analysis, the heads of CPS households (and their spouses, if present) are designated as tax-filing units. Tax rules are used to determine whether other members of the household can be claimed as dependents on the basis of their age, relationship with the primary tax-filing unit, and income. People who meet those criteria are classified as dependents; those who do not are classified as separate tax-filing units within the household. When multiple people could claim one member of a household as a dependent, CBO assumes that the household chooses the arrangement that results in the most advantageous tax situation—for example, two unmarried, cohabitating partners with two children might each claim one child and file as a head of household if doing so lowers their combined taxes.
Next, the agency divides the tax-filing unit records in each file into 15 demographic groups on the basis of marital status (married or single); number of dependents (zero, one, or two or more); whether the tax-filing unit can be claimed as a dependent (yes or no); and whether the tax filer and his or her spouse (if applicable) are 65 or older (neither, one, or both). Records from the two files are matched within the same demographic groups, with certain exceptions. Because the CPS file contains fewer head-of-household tax-filing units (single parents with dependent children) than the SOI file does, some SOI head-of-household tax-filing units are matched with single tax-filing units without children and married tax-filing units from the CPS. The deficit in head-of-household filers in the CPS data probably reflects some combination of misreporting of filing status in the SOI and a failure of the algorithm that creates tax units for the CPS to account for complex living arrangements.
Within each demographic group, CBO estimates an ordinary least squares (OLS) regression model of total income as a function of all the sources of income that are common to both the SOI and the CPS—such as wages, interest, dividends, rental income, business income and losses, pension income, and unemployment insurance. The OLS models are estimated using the SOI data. CBO applies the coefficients estimated from the regression models to the records in both files to construct a predicted total income variable. Tax-unit records in both files (independently within each demographic cell) are then sorted in descending order by predicted total income.
The SOI data and the CPS data come from samples, and therefore each record from both files has a sample weight associated with it. The sum of all the sample weights in the SOI file represents the total number of tax units that filed taxes in a given year. The sum of all the weights in the CPS file represents all of the tax units in the United States—both those that filed a tax return and those that did not. The SOI file contains many more records than the CPS file yet represents fewer total tax units. Therefore, the average sample weight in the SOI file is lower than the average sample weight in the CPS file.
Because of those differences in sample weights, SOI and CPS records are not matched on a one-to-one basis. Within each demographic group, matching begins with the record from each file that represents the highest predicted total income. Of the two records, the one with the lower sample weight is matched to only one corresponding record from the other file. The record with the higher weight is “split” and is available (with its weight reduced) to be matched to the next record in the other file. (In practice, the highest-income SOI records have very low sample weights, so the matching algorithm matches the top CPS record to many SOI records.)
That process is repeated until all the SOI records are exhausted. Each matched pairing results in a new record with the demographic characteristics of the CPS record and the income reported in the SOI. Some types of income, such as certain types of transfer payments and in-kind benefits, appear only in the CPS records; values for those items are drawn directly from that survey. Income values for CPS records that represent nonfiling tax units are taken directly from the CPS. Residual CPS records (those with the lowest predicted income) are assumed to represent tax-filing units that did not file a tax return.
Finally, households are reconstructed from tax-filing units on the basis of relationships reported in the CPS. In general, CPS tax-filing units will have been matched to multiple SOI tax-filing units. When CPS tax-filing units are combined at the household level, multiple replications of a given household are created to cover all possible combinations of the matched SOI–CPS tax units. Each household replication is appropriately weighted so that the sum of all the replications equals the original CPS household-level sample weight.
Most distributional analyses rely on a measure of annual income as the metric for ranking households. In CBO’s analyses, information on taxable income sources for tax-filing units that file individual income tax returns comes from the SOI, whereas information on nontaxable income sources and income for tax-filing units that do not file individual income tax returns comes from the CPS. Among households at the top of the distribution, the majority of income data are drawn from the SOI. In contrast, among households in the lower and middle quintiles, a larger portion of income data is drawn from the CPS (see Table A-1).
Weighted and Unweighted Sample Sizes, 2019
Most measures of income are drawn from federal tax returns, and those income measures are not adjusted to match the Bureau of Economic Analysis’s (BEA’s) national income and product accounts. This analysis does not capture income that is underreported or misreported to the Internal Revenue Service as a result of tax noncompliance. Underreported income that is excluded from this analysis may affect the distribution of income.
There are five main categories of income before transfers and taxes: labor income, business income, capital income, other income sources, and social insurance (for more information about each category of income, see Appendix B). Those categories largely reflect how income is reported on individual income tax returns. Over time, the division of income among labor income, business income, and capital income has been shaped by compensation trends and changes in tax law. For example, since the 1980s, corporations have increasingly used stock options to compensate employees for their labor. Some of those stock options are reported as capital gains on tax returns and are therefore classified as capital income in this report rather than as labor income. Another example illustrates the effects of the Tax Reform Act of 1986, which lowered the top statutory tax rate on individual income below the top statutory tax rate on corporate income. After its enactment, many C corporations (which are taxed under the corporate income tax) were converted to S corporations (which pass corporate income through to their shareholders, who then pay individual income taxes on that income) or other types of entities not subject to the corporate income tax. As a result, some income previously reported as capital gains and dividends (from C corporations) or as labor income was instead reported as business income (from S corporations or other pass-through entities).
In this report, CBO’s measures of federal taxes are based on tax liabilities incurred in a calendar year, regardless of when those liabilities are paid; by contrast, federal receipts measure taxes paid to the government in that year, regardless of when those liabilities are incurred. The measures of individual income taxes (including taxes on pass-through business income) and payroll taxes are calculated on the basis of the income and characteristics of each tax-filing unit in the underlying data set. Those calculated values align closely with the reported values. The measure of excise taxes is drawn from data on tax liabilities and collections from the IRS. The measure of corporate taxes comes from BEA’s estimate of taxes on corporate income plus CBO’s estimate of repatriation tax payments due.
The measures of transfers used in this report are mostly drawn from the agencies that administer the relevant programs. For example, the measure of benefits from the Supplemental Nutrition Assistance Program comes from the Food and Nutrition Service in the Department of Agriculture. CBO makes some adjustments to those data to align them with the sampling frame and reporting period of the CPS.
CBO allocates individual income taxes and the employee’s share of payroll taxes to the households paying those taxes directly. CBO also allocates the employer’s share of payroll taxes to employees because employers appear to pass on their share of payroll taxes to employees by paying lower wages than they otherwise would. However, the research literature suggests that many factors could cause the incidence to differ from CBO’s allocation, especially in the short term.
CBO allocates excise taxes to households according to their consumption of taxed goods and services. Excise taxes on intermediate goods, which are paid by businesses, are allocated to households in proportion to their overall consumption. CBO assumes that household spending patterns among income and demographic groups in the CPS are similar to those observed in the Bureau of Labor Statistics’ Consumer Expenditure Survey.
There is less consensus among researchers about how to allocate corporate income taxes (and taxes on capital income generally). CBO allocates 75 percent of corporate income taxes to owners of capital in proportion to their income from interest, dividends, rents, and adjusted capital gains. That measure excludes some forms of capital income that are more difficult to measure, such as investment earnings in tax preferred retirement accounts and unrealized capital gains. The agency adjusts capital gains by scaling them to their long-term historical level given the size of the economy and the applicable tax rate; that method reduces the effects of large year-to-year variations in the total amount of gains realized. The remaining 25 percent of corporate income taxes is allocated to workers in proportion to their labor income.
Households with identical income can differ in ways that affect their economic status. For example, a larger household generally needs more income to support a given standard of living than a smaller one does. However, economies of scale in some types of consumption—housing, in particular—can mean that two people generally do not need twice the income to live as well as one person who lives alone. Because of those known economies of scale, household income is an imperfect measure of economic status.
To better rank households by their relative economic status, CBO adjusts the income measure, dividing household income by an adjustment factor known as an equivalence scale. Various equivalence scales are in use today, and a significant, if somewhat dated (though still useful), body of literature explores why and how alternative equivalence scales should be calculated for the purpose of setting public policy parameters—specifically, those related to measuring poverty and means-tested programs.
To account for household economies of scale, the equivalence scale should have a value between one and the number of people in the household. An equivalence scale equal to one would make no change to the income measure and would not account for the greater needs of larger households. At the other end of the spectrum, an equivalence scale equal to the number of people in the household would imply that each person requires the same resources, which would not capture the benefits of shared consumption—most significantly, housing expenses—within the household.
A generalized formula for calculating an equivalence scale can be expressed as follows:
where n is the number of people in the household and e is an elasticity parameter for household size that ranges from zero to one, with larger values implying smaller economies of scale. To adjust household income for differences in household size, CBO uses an equivalence scale known as the square root scale. Under that method, adjusted household income is calculated as household income divided by the square root of the number of people in the household.
Calculating the equivalence scale as the square root of the number of people in the household is the same as setting the elasticity parameter for household size to 0.5 because . Using 0.5 as the elasticity parameter for household size is convenient for several reasons:
- It is the midpoint in the range of possible values for the parameter (n0 < n0.5 < n1).
- It implies that each additional person increases the household’s needs but at a decreasing rate.
- The resulting household-size adjustment is similar to the family-size adjustments the Census Bureau uses in setting U.S. poverty thresholds.
- It is transparent and relatively easy to understand.
Applying the square root equivalence scale to adjust income for differences in household size means that some households with higher income (but more people living in them) may be considered equivalent in income to households with lower income (but fewer people living in them).
CBO adjusts income for household size using the square root equivalence scale only for the purpose of ranking households and assigning them to income groups. All other income measures presented in the agency’s distributional analyses are unadjusted.
CBO presents households in adjusted household income quintiles and provides additional detail for smaller, percentile-based groupings of households within the highest income quintile (the 81st through 90th percentiles, the 91st through 95th percentiles, the 96th through 99th percentiles, the 99th to 99.9th percentiles, the 99.9th to 99.99th percentiles, and the top 0.01 percent). Each quintile contains approximately 20 percent of the civilian noninstitutionalized U.S. population, and each full percentile (that is, a percentile expressed as a whole number) contains approximately 1 percent of the population. However, because household sizes vary, the adjusted household income quintiles contain slightly different numbers of households (see Table A-1).
1. For links to reports in this series going back to 2001, see Congressional Budget Office, “Major Recurring Reports,” .
2. For more details about CBO’s current framework and how it differs from the agency’s previous approach to distributional analyses, see Kevin Perese, CBO’s New Framework for Analyzing the Effects of Means-Tested Transfers and Federal Taxes on the Distribution of Household Income, Working Paper 2017-09 (Congressional Budget Office, December 2017), .
3. Social insurance benefits consist of benefits provided through Social Security (Old Age, Survivors, and Disability Insurance); Medicare (measured as the average cost to the government of providing those benefits, net of offsetting receipts); unemployment insurance; and workers’ compensation. Although those social insurance benefits are often considered forms of government transfers, they are included in the base measure of income CBO used to rank households; however, the distributional effects of those benefit programs are not directly examined in this report. Social Security and Medicare, in particular, provide substantial resources to retirees and significantly affect the distribution of household income. In CBO’s estimation, when analyzing the distributional effects of those programs, it is more appropriate to use lifetime measures of income earned, payroll taxes paid, and benefits received. The framework used for analyzing the distribution of household income in this report is based on annual income data and, therefore, is less suitable for analyzing the distributional effects of those retirement benefit programs.
4. The Treasury’s Office of Tax Analysis uses family units in its distributional analyses. Family units are similar to household units but exclude unrelated persons living together. The Internal Revenue Service, the Joint Committee on Taxation, and the Urban–Brookings Tax Policy Center all use tax-filing units as the unit of analysis in their distributional analyses.
5. The CPS sampling frame seeks to represent the civilian noninstitutionalized population of the United States. The scope of CBO’s analysis is therefore limited to that target population. People living in correctional facilities, nursing homes, and on military bases are not included in this analysis. However, members of the armed forces living in civilian housing units on a military base or in a household not on a military base are included. In 2014, the Census Bureau split the CPS sample into two groups to test new income and health insurance questions on a smaller subsample. For this report, CBO used the data corresponding to survey questions that were consistent with those used in prior years.
6. For a general description and evaluation of statistical matching, see Marcello D’Orazio, Marco Di Zio, and Mauro Scanu, Statistical Matching: Theory and Practice (John Wiley & Sons, 2006), ; and Michael L. Cohen, “Statistical Matching and Microsimulation Models,” in Constance F. Citro and Eric A. Hanushek, eds., Improving Information for Social Policy Decisions: The Uses of Microsimulation Modeling—Volume II: Technical Papers (National Academies Press, 1991), pp. 62–86, .
7. A dependent may be considered a tax-filing unit if he or she received income above a certain threshold in a given tax year.
8. For a graphical presentation of the statistical matching algorithm, see Kevin Perese, “Statistically Matching Administrative Tax Data With Household Survey Data” (presentation at a Washington Center for Equitable Growth workshop on distributional national accounts, July 21, 2017), .
9. For a description of tax noncompliance, see Internal Revenue Service, Federal Tax Compliance Research: Tax Gap Estimates for Tax Years 2011–2013, Publication 1415 (September 2019), .
10. Other researchers have found that as a result of tax noncompliance, tax data may understate income. See Andrew Johns and Joel Slemrod, “The Distribution of Income Tax Noncompliance,” National Tax Journal, vol. 63, no. 3 (September 2010), pp. 397–418, ; and John Sabelhaus and Somin Park, U.S. Income Inequality Is Worse and Rising Faster Than Policymakers Probably Realize (Washington Center for Equitable Growth, May 2020), .
11. See Congressional Budget Office, The Taxation of Capital and Labor Through the Self-Employment Tax (September 2012), .
12. See Andrea L. Eisfeldt, Antonio Falato, and Mindy Z. Xiaolan, “Human Capitalists,” in Martin Eichenbaum, Erik Hurst, and Valerie A. Ramey, eds., NBER Macroeconomics Annual 2022, vol. 37 (University of Chicago Press, 2022), Chapter 1, . See also Eric Zwick, “Comments on ‘Human Capitalists’ by Eisfeldt, Falato, Xiaolan,” in Martin Eichenbaum, Erik Hurst, and Valerie A. Ramey, eds., NBER Macroeconomics Annual 2022, vol. 37 (University of Chicago Press, 2022), Chapter 1, .
13. See Matthew Smith and others, “The Rise of Pass-Throughs and the Decline of the Labor Share,” American Economic Review: Insights, vol. 4, no. 3 (September 2022), pp. 323–340, (PDF).
14. CBO uses the Bureau of Economic Analysis’s series from the national income and product accounts, Table 3.2, Federal Government Current Receipts and Expenditures, line 8 (Taxes on Corporate Income). Repatriation tax payments reflect a provision of the 2017 tax act that imposed a onetime tax on foreign profits that had not been previously taxed by the United States. Those payments can be made in installments over an eight-year period.
15. For more details about how CBO develops administrative totals for transfer programs, see Bilal Habib, How CBO Adjusts for Survey Underreporting of Transfer Income in Its Distributional Analyses, Working Paper 2018-07 (Congressional Budget Office, July 2018), .
16. In theory, if payroll taxes did not exist, an employee’s salary or wages would be higher by approximately the amount of the employer’s share of the payroll tax. Therefore, CBO adds the employer’s share of payroll taxes to a household’s earnings when calculating income before transfers and taxes.
17. See Dorian Carloni, Revisiting the Extent to Which Payroll Taxes Are Passed Through to Employees, Working Paper 2021-06 (Congressional Budget Office, June 2021), .
18. For a discussion of alternative methods for allocating corporate income to individuals, see the online appendix to Matthew Smith and others, “Capitalists in the Twenty-First Century,” The Quarterly Journal of Economics, vol. 134, no. 4 (November 2019), pp. 1675–1745, .
19. For a more detailed discussion about how CBO allocates corporate taxes, see Congressional Budget Office, The Distribution of Household Income and Federal Taxes, 2008 and 2009 (July 2012), .
20. See, for example, OECD Project on Income Distribution and Poverty, “What Are Equivalence Scales?” (accessed April 27, 2021), (PDF); Constance F. Citro and Robert T. Michael, eds., Measuring Poverty: A New Approach (National Academies Press, 1995), ; and Patricia Ruggles, Drawing the Line: Alternative Poverty Measures and Their Implications for Public Policy (Urban Institute Press, 1990).
21. Some equivalence scales have additional parameters to differentiate between the needs of additional adults and additional children, in which case the formula would be , where and are weights between zero and one applied to the additional number of adults and children (na and nc) in the household, respectively.
22. The most recent distributional analyses by the Treasury and the Organisation for Economic Co-operation and Development (OECD) also adjust for household or family size using the square root equivalence scale. By contrast, recent studies by government agencies in the United Kingdom and Australia use a more complex adjustment called the modified OECD equivalence scale (although it is no longer used by the OECD), which gives a full weight to the first adult in a household, a half weight to the second adult, and a 0.3 weight to each child. The Urban–Brookings Tax Policy Center, the Internal Revenue Service, the Joint Committee on Taxation, and economists Thomas Piketty and Emmanuel Saez all use tax units as their units of analysis and do not make any adjustments for differences in tax-unit size.
Household income, unless otherwise indicated, refers to income before accounting for the effects of means-tested transfers and federal taxes. Throughout this report, that income concept is called income before transfers and taxes. It consists of market income plus social insurance benefits.
Market income consists of the following:
- Labor income. Wages and salaries, including those allocated by employees to 401(k) and other employment-based retirement plans; employer-paid health insurance premiums (as measured by the Census Bureau’s Current Population Survey); the employer’s share of Social Security, Medicare, and federal unemployment insurance payroll taxes; and the share of corporate income taxes borne by workers.
- Business income. Net income from businesses and farms operated solely by their owners, partnership income, and income from S corporations.
- Capital income (including capital gains). Net profits realized from the sale of assets (but not increases in the value of assets that have not been realized through sales); taxable and tax-exempt interest; dividends paid by corporations (but not dividends from S corporations, which are considered part of business income); positive rental income; and the share of corporate income taxes borne by capital owners.
- Other income sources. Income received in retirement for past services and other nongovernmental sources of income.
Social insurance benefits consist of benefits from Social Security (Old Age, Survivors, and Disability Insurance), Medicare (measured by the average cost to the government of providing those benefits), unemployment insurance, and workers’ compensation.
Means-tested transfers are cash payments and in-kind services provided through federal, state, and local government assistance programs. Eligibility to receive such transfers is determined primarily on the basis of income, which must be below certain thresholds. Means-tested transfers are provided through the following programs: Medicaid and the Children’s Health Insurance Program (measured by the average cost to the government of providing those benefits); the Supplemental Nutrition Assistance Program (formerly known as the Food Stamp program); housing assistance programs; Supplemental Security Income; Temporary Assistance for Needy Families and its predecessor, Aid to Families With Dependent Children; child nutrition programs; the Low Income Home Energy Assistance Program; and state and local government general assistance programs.
Average means-tested transfer rates are calculated as means-tested transfers divided by income before transfers and taxes.
Federal taxes consist of individual income taxes, payroll (or social insurance) taxes, corporate income taxes, and excise taxes. Those four sources accounted for 93 percent of federal revenues in fiscal year 2019. Revenue sources not examined in this report include states’ deposits for unemployment insurance, estate and gift taxes, net income of the Federal Reserve System that is remitted to the Treasury, customs duties, and miscellaneous fees and fines.
In this analysis, taxes for a given year are the amount a household owes on the basis of income received that year, regardless of when the taxes are paid. Those taxes comprise the following:
- Individual income taxes. Individual income taxes are paid by U.S. citizens and residents on their income from all sources, except those sources exempted under the law. Individual income taxes can be negative because they include the effects of refundable tax credits, which can result in net payments from the government. Specifically, if the amount of a refundable tax credit exceeds a filer’s tax liability before the credit is applied, the government pays that excess to the filer. Statutory marginal individual income tax rates are the rates set in law that apply to the last dollar of income.
- Payroll taxes. Payroll taxes are levied primarily on wages and salaries and generally have a single rate and few exclusions, deductions, or credits. Payroll taxes include those that fund the Social Security trust funds, the Medicare trust fund, and unemployment insurance trust funds. The federal portion of the unemployment insurance payroll tax covers only administrative costs for the program; state-collected unemployment insurance payroll taxes are not included in the Congressional Budget Office’s measure of federal taxes (even though they are recorded as revenues in the federal budget). Households can be entitled to future social insurance benefits, including Social Security, Medicare, and unemployment insurance, as a result of paying payroll taxes. In this analysis, average payroll tax rates capture the taxes paid in a given year and do not capture the benefits households may receive in the future.
- Corporate income taxes. Corporate income taxes are levied on the profits of U.S.-based corporations organized as C corporations. In its analysis, CBO allocated 75 percent of corporate income taxes in proportion to each household’s share of total capital income (including capital gains) and 25 percent to households in proportion to their share of labor income.
- Excise taxes. Sales of a wide variety of goods and services are subject to federal excise taxes. Most revenues from excise taxes are attributable to the sale of motor fuels (gasoline and diesel fuel), tobacco products, alcoholic beverages, and aviation-related goods and services (such as aviation fuel and airline tickets).
Average federal tax rates are calculated as federal taxes divided by income before transfers and taxes.
Income after transfers and taxes is income before transfers and taxes plus means-tested transfers minus federal taxes.
Income groups are created by ranking households by their size-adjusted income before transfers and taxes. A household consists of people sharing a housing unit, regardless of their relationships. The income quintiles (fifths) contain approximately the same number of people but slightly different numbers of households (see Appendix A). Similarly, each full percentile (hundredth) contains approximately the same number of people but a different number of households. If a household has negative income (that is, if its business or investment losses are larger than its other income), it is excluded from the lowest income group but included in totals.
This report was prepared at the request of the Chairman of the Senate Committee on Finance. In keeping with the Congressional Budget Office’s mandate to provide objective, impartial analysis, the report makes no recommendations.
Ellen Steele wrote the report with guidance from Edward Harris, John McClelland, and Joseph Rosenberg. Bilal Habib, Kurt Seibert, and Naveen Singhal contributed to the analysis. Omar Morales and Tess Prendergast fact-checked the report.
The staff of the Joint Committee on Taxation provided helpful comments. The assistance of external reviewers implies no responsibility for the final product; that responsibility rests solely with CBO.
Mark Doms, Jeffrey Kling, and Robert Sunshine reviewed the report. Scott Craver and Loretta Lettner edited it, and Jorge Salazar created the graphics and prepared the text for publication. The report is available at .
Phillip L. Swagel