Data sources

The data in this report are derived from the decennial census and the American Community Survey (ACS), both conducted by the U.S. Census Bureau. The decennial census (long form) is the source for the data for 1970 (1% Metro forms 1 and 2 sample), 1980 (5% state sample), 1990 (5% state sample) and 2000 (5% sample). The ACS (1-year sample) is the source for the 2010 and 2016 data.

The combined forms 1 and 2 samples of the 1970 decennial census have about 4 million observations. The 1980, 1990 and 2000 decennial census samples each have more than 11 million observations. The 2010 and 2016 1-year ACS files each have about 3 million observations.

The specific versions of the data used in this report are the Integrated Public Use Microdata Series (IPUMS) provided by the University of Minnesota.33 The IPUMS assigns uniform codes, to the extent possible, to data collected over the years. More information about the IPUMS, including variable definitions and sampling error, is available at

Changes in data collection methods, revisions to survey questions and other factors may affect the comparability of demographic and income data over time. For example, changes to income codes or to the components of income that are measured (see below) may have an impact on estimates of changes in the Gini coefficient over time. The key issue for this analysis is whether there is a differential impact by race and ethnicity.

Timing of data collection

Decennial censuses are conducted in April of the census year and ask about income received in the preceding calendar year. For example, the 2000 decennial census asks about income received during 1999.

Unlike the decennial census, the ACS is conducted year-round. Respondents are asked to report their income received in the 12 months before the survey date. For example, a respondent completing the ACS questionnaire on Jan. 15, 2010, is expected to report income received from Jan. 15, 2009, to Jan. 15, 2010, while a respondent completing the questionnaire on Dec. 15, 2010, would report income received from Dec. 15, 2009, to Dec. 15, 2010. Thus, in principle, the 2010 ACS includes income data from a total of 24 months, from January 2009 to December 2010.

Income data for 1969 (1970 census), 1979 (1980 census), 1989 (1990 census) and 1999 (2000 census) refer to years close to the peak of business cycles (as dated by the National Bureau of Economic Research). Income data from the 2010 ACS represent a period (January 2009 to December 2010) that overlaps with the Great Recession (December 2007 to June 2009). Thus, data from the 2010 ACS reflect the effects of the recession on household incomes. The income data from the 2016 ACS (spanning January 2015 to December 2016) were collected several years into an economic expansion.

Race and ethnicity

Where mentioned, whites, blacks and Asians include non-Hispanics only; Asians include Pacific Islanders; and Hispanics are of any race. In 1970, 1980 and 1990, respondents had the option of reporting a single race only. In a revision in 2000, respondents were given the option of selecting one or more race categories to indicate their racial identifications. In this analysis, data for whites, blacks and Asians in 2000, 2010 and 2016 include single-race, non-Hispanics only.

In 2000, some 2.4% of the U.S. population chose to identify with two or more races. The share inched up to 2.9% in 2010. Another notable change in 2000 was the introduction of the “some other race” category. Those who identify multiple races in 2000, 2010 and 2016 and other racial and ethnic groups are included in all totals but are not shown separately.

Other revisions to the race and Hispanic origin questions may also affect the comparability of data over time. A key issue regarding Hispanic origin is that the identification in 1970 is not based on the question asked in the 1970 decennial census (Form 1). This is because of the cautions reported by the Census Bureau regarding the comparability of the 1970 Hispanic origin question with later years (see page B-13 of the linked report). In this analysis, Hispanic origin for 1970 is as imputed by IPUMS.

People are assigned to a racial or ethnic group based on their personal identification, not the identification of the head of the household. In 2016, about 98% of whites, 95% of blacks, 90% of Hispanics and 90% Asians lived with a household head of the same race or ethnicity.

Households in census data

The analysis in this report is based on a sample of the adult, civilian household population. The Census Bureau defines a household as the entire group of persons who live in a single dwelling unit. A household may consist of several persons living together or one person living alone. It includes the household head and all of his or her relatives living in the dwelling unit and also any lodgers, live-in housekeepers, nannies and other residents not related to the head of the household.


Household income is defined as the sum of total personal income for all members of the household of a certain age. In 1970, all household members ages 14 and older are included. In all other samples, all household members ages 15 and older are included. Household income may be zero or negative. The share of adults living in households with zero or negative income is about 1% across the years analyzed.

Personal income, or “money income,” as per the Census Bureau, is the income received on a regular basis (exclusive of certain money receipts such as capital gains and lump-sum payments) before payments for personal income taxes, Social Security and Medicare taxes, union dues, etc. It includes income received from wages, salary, commissions, bonuses, and tips; self-employment income from own nonfarm or farm businesses, including proprietorships and partnerships; interest, dividends, net rental income, royalty income, or income from estates and trusts; Social Security or Railroad Retirement income; Supplemental Security Income (SSI); any cash public assistance or welfare payments from the state or local welfare office; retirement, survivor, or disability benefits; and any other sources of income received regularly such as Veterans’ (VA) payments, unemployment and/or worker’s compensation, child support, and alimony.

The Census Bureau’s current definition of income is broader than in the past. Data for all years include income received from wages or salary before taxes and deductions, net earnings from own businesses or farms, income from Social Security or Railroad Retirement, other forms of public assistance or welfare, and other sources of income regularly received. Investment income (including dividend payments and net rental income) was added starting with the 1980 sample. Retirement income other than Social Security was added starting in the 1990 sample. Supplemental Security Income (SSI) was added starting in the 2000 sample.

In this report, all dollar amounts are reported in 2016 dollars, unless otherwise noted. Nominal figures are adjusted using the CPI-U-RS adjustments in the table provided on page 22 of the Census Bureau report, Income and Poverty in the United States: 2016.

Accounting for income allocation

Item nonresponse – the share of respondents who either do not provide or provide an invalid answer for an item – is relatively low for most items on the decennial census and American Community Survey. This is not true for the various components of income used in calculating household income, however.

To account for nonresponse on the income components, the Census Bureau uses statistical procedures to allocate missing values by looking at the values reported by similar respondents. These additional respondents may be the original respondent’s nearest neighbor or others with similar but relatively rare characteristics. Income allocation is not necessarily exact, and some studies have found that the inclusion of allocated values of income may lead to biased results in certain situations. For that reason, researchers often omit individuals from the sample if their earnings are allocated.

In this study, households are excluded from the sample if half or more of their household income is allocated. Allocation flags are provided for each household member’s individual components of income. For each member, the total amount of allocated income, identified via allocation flags, and total personal income are calculated. Then these two amounts are summed across all earners in the household. The share of allocated income for the household is then calculated as:

A table in Appendix B shows the shares of respondents in the civilian household sample living in households that had half or more of their income allocated by the Census Bureau. In 2016, these shares were as follows: whites – 17.4%, blacks – 31.2%, Hispanics – 26.5%, and Asians – 19.4%. The shares with allocated incomes are similarly high from 1980 to 2010. The allocation rates for 1970, only 0.1% or less, appear surprisingly low compared with other years.

The impact of removing households if half or more of their income is allocated is generally modest. Compared with a sample of the overall adult, civilian household population, a sample without households with half or more of their income allocated has the following effects:

  • There is no change of import to the estimates for 1970, unsurprising in view of the very small share of households with allocated income.
  • From 1980 to 2016, removing households with half or more of their income allocated raises income levels from about 1 to 10% among all adults in the sample, depending on the year and income percentile. The impact is greater in 1980, 1990 and 2000; greater at lower-income percentiles; and more notable among blacks. For example, in 2000, the 10th percentile income for blacks is raised 9% and the 90th percentile income is raised 4%.
  • Removing households with half or more of their income allocated tends to raise the estimated income growth from 1970 to 2016. The increase is in the range of 1% to 5%, varying across percentiles and racial and ethnic groups. One exception to this rule is that income growth for lower-income blacks and Hispanics is about 1% to 2% lower if households with allocated income are removed from the sample.
  • The removal of households with half or more of their income allocated reduces or leaves unchanged the estimates of income inequality among whites in all years, by either the 90/10 ratio or the Gini coefficient. For blacks, Hispanics and Asians, the results are mixed, with income inequality estimated to be higher in some years and lower in other years. The differences are within 5% in most years. Blacks are estimated to experience a higher level of inequality than other groups from 1970 to 2010 and Asians the highest level in 2016 in either sample.
  • Estimated changes in inequality are similar under both samples. In the sample without households with half or more of their income allocated, the 90/10 ratio increases as follows from 1970 to 2016: whites – 24%; blacks – 7%; Hispanics – 15%; and Asians – 77%. In the sample including all households, the 90/10 ratio increases as follows from 1970 to 2016: whites – 26%; blacks – 3%; Hispanics – 11%; and Asians – 72%. Inequality increases the most among Asians and the least among blacks in both samples.
  • Regarding the gaps in income across racial and ethnic groups, the removal of households with half or more of their income allocated has modest but mixed results. The income gap is about as likely to go up as to go down, by no more than 3 percentage points. The income gaps trend similarly over time in the two samples.

Adjustments to income for household size

In this report, “income” refers to what is afforded a person by the combined resources of his or her household, whether the person had personal earnings or not. Thus, people’s incomes are represented by their household’s income adjusted for household size. All members of a household have the same income.

Incomes are adjusted for household size because a four-person household with an income of, say, $50,000 faces a tighter budget constraint than a two-person household with the same income. In addition to comparisons across households at a given point in time, this adjustment is useful for measuring changes in income over time. That is because average household size fell in the U.S. from 3.1 persons in 1970 to 2.5 persons in 2016. Ignoring this demographic change would mean ignoring a commensurate loosening of the household budget constraint.

At its simplest, adjusting for household size means converting household income into per capita income. Thus, a two-person household with an income of $50,000 would have a per capita income of $25,000, double the per capita income of a four-person household with the same total income.

A more sophisticated adjustment for household size recognizes that there are economies of scale in consumer expenditures. For example, a two-bedroom apartment may not cost twice as much to rent as a one-bedroom apartment. A household of two, compared with a household of one, would likely not need a second refrigerator or internet subscription, and so on. For that reason, most researchers make adjustments for household size using the method of “equivalence scales.”

A commonly used equivalence-scale adjustment is follows:

This is also the approach used in this report: Household income is divided by the square root of the number of people (of any age) in the household. In practical terms, this means that household income is divided by 1.41 for a two-person household, 1.73 for a three-person household, 2.00 for a four-person household and so on.34 The size-adjusted household income is then assigned to each adult member of the household.