Inequality Explorer

What this is built on

CBO publishes The Distribution of Household Income once a year, for the calendar year three years earlier. Its tables give, for ten income groups and every year since 1979, the average household’s income and each piece of it (wages, business income, capital gains, dividends and interest, Social Security, Medicare, unemployment insurance), each means-tested transfer (Medicaid and CHIP, SNAP, SSI, other programs), each federal tax (individual income, payroll, corporate, excise), the number of households and people, and the income needed to enter each group. Its figures add the Gini coefficient under four income definitions, the finer slices at the very top, and refundable tax credits.

Four official income definitions. Market income is labor income (including employers’ health premiums and the employer share of payroll taxes), business income, capital income and gains, and other income. Income before transfers and taxes adds social insurance (Social Security, Medicare, unemployment insurance, workers’ compensation). After transfers, before taxes adds means-tested transfers, with Medicaid and CHIP counted at cost. Income after transfers and taxes subtracts the four federal taxes. State and local taxes are not in CBO’s accounting and are not here. CBO counts Medicare and Medicaid and CHIP at what the government spends on them. The page adds two alternatives built by Econvitals: health transfers at 40 cents on the dollar (“After, health at 40¢”), CBO’s after-tax income with Medicare and Medicaid and CHIP counted at 40 percent of their cost, and excluding health transfers (“After, ex health”), with both left out. Employers’ health insurance premiums stay in at full value in both, as part of pay. Their averages are exact arithmetic on CBO’s; their shares and Gini are approximations, described below. A custom definition can also count the two programs at 40 cents. The 40 cents is consistent with 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 127, no. 6 (2019): 2836–2874, who estimate that Medicaid is worth about 20 to 40 cents to the people who get it for each dollar the government spends.

Groups and rankings. Each quintile holds a fifth of people, not households, ranked by size-adjusted income. CBO’s headline ranking is by income before transfers and taxes; the page can rank by market income or by after-tax income instead, which changes which households are “the top 1 percent.” Households with negative income are in the totals but in no quintile, so the groups do not quite add to all households. Bar charts show the four lower fifths and CBO’s four slices of the top fifth, with the top fifth as a whole in a note; the growth bars can show the five fifths instead.

Household types. CBO publishes every table for all households and for three types that between them cover everyone: households with children, elderly-headed households (head or spouse 65 or older) and nonelderly childless households. “All nonelderly” on this page combines the first and third, weighted by households — an Econvitals combination, not a CBO table. (CBO does not say whether its “with children” type includes households headed by someone 65 or older, so a few such households may be in it.) The income groups are the same for every type: they are set on all households.

Econvitals’ calculations, and where they are approximate. A custom definition (say, income after transfers and taxes but without Medicare) is built from CBO’s per-household averages. Its income shares weight each group by the number of households CBO’s own published shares imply (the printed counts are rounded to 0.1 million, too coarse for the top 1 percent); rebuilt that way, CBO’s printed shares come back within 0.15 point for every group — close, but not CBO’s, which is why market income, income before transfers and taxes and income after transfers and taxes always show CBO’s printed share. CBO prints no shares for income after transfers but before taxes, so those are rebuilt. On the step charts the single-program steps are spread so they add up to CBO’s change between one official definition and the next. The grouped Lorenz curves run through the eight groups and are only an approximation. CBO computes the Gini on its microdata; for a custom definition the page starts from CBO’s Gini for the nearest official definition and moves it by the difference the ten groups show, and says so on the chart. Growth and percent changes, ratios of group averages, tax and transfer rates and net positions are plain arithmetic on the averages. Dollars are the report’s dollars (), deflated by CBO with the PCE price index.

The data are CBO’s; the views, the custom definitions and the approximations are Econvitals’ and have not been reviewed by CBO. Source: Congressional Budget Office, The Distribution of Household Income.