Annual Statistical Supplement, 2003


Appendix C: Poverty Data

Table 3.E1 presents weighted average poverty thresholds for families in the United States for 1959–2001. Table 3.E2 presents data on the number and percentage of people in poverty in the United States for 1959–2000. Tables 3.E3, 3.E4, and 3.E6 present more detailed data on people and families in poverty in the United States during 2000. (Both poverty thresholds and poverty population data are issued by the U.S. Census Bureau.) Table 3.E8 presents poverty guidelines by family size for the contiguous United States, Alaska, and Hawaii for 1965–2002. (Poverty guidelines are issued by the U.S. Department of Health and Human Services.)

Poverty thresholds are used primarily for statistical purposes—producing statistics on the number of Americans in poverty. Poverty guidelines are used for administrative purposes—for example, to determine whether a person or family is financially eligible for assistance or services under certain federal government programs (not including cash public assistance). Both thresholds and guidelines are sets of dollar figures that vary by family size and (in the case of the thresholds) family composition.

The poverty thresholds were developed in 1963–1964 by Mollie Orshansky of the Social Security Administration as a measure of income inadequacy. The poverty definition was modified in 1969 and 1981 by federal interagency committees. The thresholds were based on food expenditure/income patterns (from the Department of Agriculture's 1955 Household Food Consumption Survey) and the costs of the Department of Agriculture's economy food plan for families of different sizes and compositions. (See Joseph Dalaker, U.S. Census Bureau, "Poverty in the United States: 1998," Current Population Reports: Consumer Income, Series P60-207, September 1999, Appendix A, for an explanation of the poverty definition.) While the poverty thresholds in some sense represent families' needs, the official poverty measure should be interpreted as a statistical yardstick rather than as a complete description of what people and families need to live.

The poverty thresholds have become the basis for the official statistics on the extent of poverty in the United States, which are issued annually by the Census Bureau in the Current Population Reports series. The thresholds are adjusted annually for price changes using the annual average consumer price index (CPI-U). (See U.S. Bureau of the Census, "Revision in Poverty Statistics, 1959 to 1968," Current Population Reports: Special Studies, Series P-23, No. 28, August 1969; and Directive No. 14, "Definition of Poverty for Statistical Purposes," Statistical Policy Handbook, U.S. Department of Commerce, Office of Federal Statistical Policy and Standards, 1978.)

The poverty guidelines are a simplified version of the poverty thresholds; there are separate sets of guidelines for the two noncontiguous states (Hawaii and Alaska). The guidelines are used for determining whether a person or family is financially eligible for assistance or services under certain federal programs. Authorizing legislation or regulations for specific programs indicate whether a program uses the poverty guidelines as one of several eligibility criteria, uses a modification of the guidelines (for example, 125 percent or 185 percent of the guidelines), or uses them for the purpose of setting priorities in providing assistance or services.

Since 1973, the guidelines have been computed from the poverty thresholds by increasing the most recently published weighted average poverty thresholds by the percentage change in the CPI-U over the past year (more precisely, from the next most recent calendar year to the most recent calendar year) and rounding the figure for a family of four up to the next highest multiple of $50. Figures for all family sizes over and under four persons are computed by adding or subtracting equal dollar amounts derived from the average difference between poverty lines (rounded to the nearest multiple of $20).

The poverty thresholds were calculated using data (the 1955 Household Food Consumption Survey) that defined income as after-tax money income; accordingly, the thresholds were intended to be applied to data on money income. The National Research Council's Panel on Poverty and Family Assistance put great emphasis on the principle of consistency in poverty measurement—that the definition of family resources (income) used should be consistent with the concept underlying the poverty thresholds. (For a discussion of this principle, see Measuring Poverty: A New Approach, Constance F. Citro and Robert T. Michael (eds.), Washington, D.C., National Academy Press, 1995, pp. 4, 9–10, 37–40, 65–66, 98, 203–206, and 227–231. This important report proposed a new approach for developing an official poverty measure for the United States.) According to the Panel's principle of consistency, it would be inappropriate to apply the current poverty thresholds (calculated using an income definition of after-tax money income) to an income distribution that used an income definition of money income plus selected noncash benefits.

As noted above, the poverty thresholds were developed in 1963–1964 by Mollie Orshansky of the Social Security Administration. In May 1965—just over a year after the Johnson Administration had initiated the War on Poverty—the Office of Economic Opportunity adopted Orshansky's thresholds as a working or quasi-official definition of poverty. At that time, the thresholds comprised a matrix of 124 detailed poverty thresholds based on the total number of family members, the number of family members who were related children under age 18, the sex of the family householder, the age of the individual or family householder (for one- and two-person units only), and whether the family lived on a farm.

In 1969, a federal interagency committee made two changes in the poverty definition: the thresholds would be annually updated by the consumer price index instead of by the per capita cost of the economy food plan, and farm poverty thresholds would be set at 85 percent rather than 70 percent of corresponding nonfarm thresholds. (Poverty threshold and poverty population figures for prior years were retabulated retrospectively on this basis.) In August 1969, the Bureau of the Budget designated the poverty thresholds with these revisions as the federal government's official statistical definition of poverty.

In 1981, several other changes recommended by another interagency committee were made in the poverty definition: (1) elimination of separate thresholds for farm families, (2) elimination (through appropriate averaging) of separate thresholds for female-householder and "all other" families, and (3) extension of the poverty matrix to make the largest family size category "nine persons or more" rather than "seven or more persons." (See U.S. Bureau of the Census, "Characteristics of the Population Below the Poverty Level: 1980," Current Population Reports: Consumer Income, Series P-60, No. 133, July 1982, pp. 2–5, 9, and 186.) As a result of these changes, the current matrix of poverty thresholds used by the Census Bureau to determine the poverty status of families and unrelated individuals consists of a set of 48 detailed thresholds arranged in a two-dimensional matrix by family size (number of family members, ranging from one person, that is, an unrelated individual, to nine or more persons) cross-classified by the presence and number of family members who are related children under age 18 (from no children to eight or more children present). Unrelated individuals and two-person families are further differentiated by the age of the individual or family householder (under age 65 and aged 65 or older).

The current official definition of poverty is over 30 years old. In 1990, Congress requested a study of the official U.S. poverty measure by the National Research Council (NRC) to provide a basis for a possible revision of the poverty measure. In 1992, the NRC's Committee on National Statistics appointed a Panel on Poverty and Family Assistance to conduct this study. In 1995, the Panel published its report of the study, Measuring Poverty: A New Approach, Constance F. Citro and Robert T. Michael (eds.), Washington, D.C., National Academy Press, 1995. In the report, the Panel proposed a new approach for developing an official poverty measure for the United States—although it did not propose a specific set of dollar figures. The Panel's proposed approach focused on three major areas: new poverty thresholds, a new and consistent definition of family resources (income), and data sources.

In July 1999, the Census Bureau, in collaboration with the Bureau of Labor Statistics, released a report, Experimental Poverty Measures: 1990 to 1997 (P60-205) that examined the effects of different resource definitions and thresholds on poverty and which estimated several experimental poverty rates based on the NRC panel's recommendations. That report and subsequent updates are available on the Census Bureau poverty measurement Web page at

Data on the poverty population and on family and personal income are collected in the Current Population Survey Annual Social and Economic Supplement (CPS ASEC), formerly known as the March Supplement. Following the standard Census Bureau definition, the family is defined as two or more persons related by birth, marriage, or adoption and residing together. "Income" refers to money income before federal, state, or local personal income taxes and excludes capital gains and lump-sum payments; however, public income transfers are included. Money income does not reflect that many families receive noncash benefits such as employee use of business transportation and facilities, employer-paid health insurance and other employer-supported fringe benefits, Medicare, Medicaid, food stamps, and housing assistance. Many farm families receive benefits in the form of rent-free housing or goods produced and consumed by the family.

Every year between February and April, the sample of U.S. households interviewed in the monthly CPS is asked to provide information on household members' incomes during the preceding calendar year. Survey experience indicates that respondents tend to underreport their income in household surveys. Underreporting is most pronounced for dividends, interest, and workers' compensation; less pronounced for veterans' payments, public assistance, and private pensions; and modest for Social Security and other federal retirement programs. The proportion of nonresponses to CPS income questions is greater among middle income and higher income families than among lower income families.

Because the CPS sample size, content, and procedures changed several times, some differences in the data over time may in part reflect these methodological changes, not just changes in the population. Important changes to the CPS methodology took place for data years 1976, 1979, 1980, 1987, 1988, 1992, 1994, 1995, and 2000. A summary of these changes and references for more information about them appear below. The report numbers, "P60-#" refer to the Current Population Reports, Consumer Income series.

Year Methodological change Reference
1959 First year for which poverty data are available. The poverty measure was developed in 1963-1964. P60-210, Appendixes.
1969 Modification of poverty measure. Bureau of the Budget prescribed the new measure as the official poverty measure for statistical use in federal agencies. P60-68, pp. 11–12.
1976–1987 For wide income intervals, Pareto interpolation was used to compute median income. Before 1976 and after 1987, linear interpolation was used. P60-166, p. 93.
1979 "Householder" concept first implemented, replacing "head" concept—the husband is no longer automatically the family reference person. Secondary families (people who are related to each other but not the householder) are tallied separately from "families." P60-130, pp. 6–10.
Change in universe—income questions no longer asked of 14-year-olds. Poverty universe changed—people under age 15 not living with any family members are excluded. (Previously, unrelated individuals under age 14 were excluded.)
1980 Modification of poverty measure (final approval in 1981). Estimates weighted using 1980 Census results. P60-133, pp. 2–7.
1987 New CPS processing system. P60-166, pp. 1, 14–17; Welniak, 1990 ASA proceedings.
1988 Linear interpolation used to compute median incomes. P60-166, p. 93.
1992 Estimates first weighted using 1990 Census results. P60-188, p. vii.
1994 Computer-assisted interviewing replaced pencil-and-paper interviewing. P60-189, p. vii.
1994–1995 New CPS sample design. P60-189, p. vii and Table D-3; P60-194, pp. v and xiii and Table D-3.
1995 Revised edit and allocation procedures for race groups. P60-194, pp. v and xiii.
2000 Expansion of CPS sample interviewed from about 50,000 households to about 78,000 households nationwide. Poverty data for 2000 and 2001 from sample expansion first published in P60-219, Poverty in the United States: 2001. In the same report, data were weighted using Census 2000 results. P60-219, Appendix B.
2002 For the first time, the 2003 CPS ASEC recorded multiple categories for respondents who identified themselves with more than one race. We do not know how people who reported more than one race in 2002 previously reported their race. Therefore, there is no single way to compare changes to poverty by race between 2001 and 2002. The "Asian or Pacific Islander" race category was divided into two groups: "Asians" and "Native Hawaiians and Other Pacific Islanders." P60-222, pp. 2–3, 5.

For details about how questionnaire changes and file processing changes affected the data, see Welniak, Edward, "Effects of the March Current Population Survey's New Processing System on Estimates of Income and Poverty," Proceedings of the American Statistical Association, 1990.

Further details about CPS methodology may be found in Technical Paper 63RV, available at

If you need to find additional poverty data, you may browse the U.S. Census Bureau poverty Web site at, or contact the U.S. Census Bureau's Housing and Household Economic Statistics Division Information Staff at (301) 763-3242 or e-mail

For further information about technical changes to the poverty measure, contact Joe Dalaker at (301) 763-3213 or e-mail

For further information about the statistical reliability of the CPS estimates, contact dsmd_s&

CONTACT: Gordon Fisher (202) 690-5880 or Joe Dalaker (301) 763-3213.