[Background Material and Data on Programs within the Jurisdiction of the Committee on Ways and Means (Green Book)] [Appendices] [Appendix H. Data on Poverty] [From the U.S. Government Printing Office, www.gpo.gov] APPENDIX H-DATA ON POVERTY CONTENTS Measuring Poverty Trends in the Overall Poverty Rate Poverty Rates for Individuals in Selected Subgroups of the Population Poverty Rates for Families Poverty Under Alternative Measures of Income and Price Inflation Poverty by Metro Area and State Trends in Family Composition and Income, 1967-2002 Definitions and Methods Income Measure Income Shares Trends in Pretax Cash Incomes by Type of Family Pretax Adjusted Family Income Average Family Cash Income By Family Type Antipoverty Effectiveness of Various Cash and Noncash Transfers References MEASURING POVERTY When the Federal Government began measuring poverty in the early 1960s, the continued existence of poor people in a time of the "Affluent Society" seemed anomalous. Official concern soon translated into efforts to measure the size of the poverty population, and the search began for programmatic ways to alleviate poverty. The first rough estimates of the incidence of poverty were based on survey data indicating that families generally spent about one-third of their income on food. A poverty level income was then calculated by using as a yardstick the amount of money necessary to purchase the lowest cost "nutritionally adequate" diet calculated by the Department of Agriculture (roughly equivalent to the current Thrifty Food Plan). This price tag was multiplied by three to produce a poverty threshold. The assumption underlying this procedure is that if a family did not have enough income to buy the lowest cost nutritionally adequate diet, and twice that amount to buy other goods and services, it was "poor." Adjustments were made for the size of the family, the sex of the family head, and for whether the family lived on a farm. Farm families were assumed to need less cash income because their needs could be met partially by farm products, particularly food. The adjustments for sex of the family head and for farm-nonfarm residence were abolished in 1981. Policy officials made a major change to the basic approach for calculating the poverty threshold in 1969. Rather than multiplying the cost of the Thrifty Food Plan by three to establish the poverty threshold, officials decided to simply increase the previous year's threshold by the change in the Consumer Price Index (CPI). In addition to this major change, the Census Bureau made minor revisions in its method of estimating the poverty threshold four times-in 1966, 1974, 1979, and 1981. These revisions changed the estimate of the poverty rate. The first two revisions slightly reduced the estimated number of poor, while the more recent revisions slightly increased the number. In 1984, the Census Bureau also revised its method of imputing missing values for interest income, which slightly lowered the estimated poverty rate. Data on income and poverty after 1987 may not be comparable to data in earlier years because of changes in the methods used by the Census Bureau to process survey results. This new processing system was applied to 1987 data so that 1988 and 1987 data are comparable. Revised 1987 data are denoted as 1987R. The new processing system increased aggregate income by 0.9 percent and lowered the poverty rate for 1987 by 0.1 percent. The tables in this subsection provide poverty data calculated using the official Census definition of poverty. The Census definition of poverty has remained fairly standard over time and is useful for measuring progress against poverty. Under this definition, poverty is determined by comparing pretax cash income with the poverty threshold. Table H-1 shows the population, number of persons in poverty, and the poverty rate in 2002 by age, race, region and family type. In 2002, 12.1 percent (34.6 million persons) of the total U.S. population lived in poverty. Of all demographic groups shown, poverty was second highest among female-headed families with children (33.1 percent). Among children under age 18, 16.7 percent, or 12.1 million children, lived in poverty in 2002. The weighted average poverty thresholds for families of various sizes for selected years between 1959 and 2002 are presented in Table H-2. TRENDS IN THE OVERALL POVERTY RATE In 1959, the overall poverty rate for individuals in the United States was 22 percent, representing 39.5 million poor persons (Tables H-3 and H-4). Between 1959 and 1969, the poverty rate declined dramatically and steadily to 12.1 percent. As a result of a sluggish economy, the rate increased slightly to 12.5 percent by 1971. In 1972 and 1973, however, it began to decrease again. The lowest rate over the entire 24-year period occurred in 1973, when the poverty rate was 11.1 percent. At that time roughly 23 million people were poor, 42 percent less than were poor in 1959. TABLE H-1 -- POVERTY STATUS OF PERSONS BY AGE, ETHNICITY, REGION, AND FAMILY TYPE, 2002 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] TABLE H-2 -- WEIGHTED AVERAGE POVERTY THRESHOLD FOR NONFARM FAMILIES OF SPECIFIED SIZE, SELECTED YEARS 1960-2002 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] TABLE H-3 -- NUMBER OF PERSONS IN POVERTY BY DEMOGRAPHIC GROUPS, SELECTED YEARS 1959-2002 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] The poverty rate increased by 1975 to 12.3 percent, and after 1978 the poverty rate rose steadily, reaching 15.2 percent in 1983. Between 1983 and 1993, the poverty rate moved up and down within a narrow range of about 2.5 percentage points, declining somewhat during economic recoveries and rising somewhat during economic downturns. However, poverty declined every year between 1993 and 2000, reaching a low of 11.3 percent, the lowest rate since 1974. The rate rebounded slightly in 2001 and 2002, rising to 12.1 percent, which absent recent years would have been lower than any year since the 1970s. The poverty rate for children followed a similar path, falling prior to 1969, fluctuating between about 15 and 17 percent in the 1970s, and then remaining between about 20 and 23 percent during the 1981 to 1996 period. Since 1996, the children's poverty rate fell from 20.5 percent in 1996 to 16.2 percent in 2000 - a 21 percent decline. The rate rose slightly in 2001 and 2002, but remained in the 16 percent range last seen in 1979. TABLE H-4-- POVERTY RATES FOR DEMOGRAPHIC GROUPS, SELECTED YEARS 1959-2002 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] POVERTY RATES FOR INDIVIDUALS IN SELECTED SUBGROUPS OF THE POPULATION As Table H-4 illustrates, there are substantial differences between the overall poverty rate and the poverty rates of individuals in certain demographic subgroups. Most notably, blacks, individuals in female-headed households, and Hispanics have poverty rates that greatly exceed the average. The poverty rates for individuals in female-headed households remained above 35 percent over the 1959-97 period. However, it declined every year after 1991 and in 2000 reached its lowest level ever at 28.5. The poverty rate for blacks and Hispanics has remained near 30 percent during the 1980s and mid 1990s. However, both rates declined every year after the early 1990s and for blacks it reached its lowest level ever in 2000 at 22.5 percent, and for Hispanics reached a record low of 21.4 percent in 2001. The poverty rate for the aged, which exceeded the overall poverty rate in 1959, fell quickly beginning in the 1960s. By 1999 it had reached the remarkably low level of 9.7, a decline of over 70 percent since 1959. The poverty rate for whites was below the overall poverty rate throughout the entire 1959-2002 period. It was 10.4 percent in 2002. The poverty rate for children exceeded the overall poverty rate every year between 1959 and 2002. POVERTY RATES FOR FAMILIES Table H-5 shows the composition of the poverty population for various demographic groups for selected years between 1959 and 2002. Table H-6 presents poverty data for families and unrelated individuals (individuals living alone). Female-headed families with children and unrelated individuals are more likely to be poor than other families with children or families with aged members. In 2002, 33.6 percent of female-headed families with children were poor, compared with 7.6 percent of male-present families. Although only 6.7 percent of all families with an aged member were poor, 19.5 percent of all aged unrelated individuals were poor. About 20.7 percent of nonaged unrelated individuals were poor. POVERTY UNDER ALTERNATIVE MEASURES OF INCOME AND PRICE INFLATION The Census Bureau publishes data that reflect two adjustments in the official definition of poverty. The first of these is an alternative inflation adjustment. The official poverty line is based on a procedure developed in 1965 with yearly adjustments for inflation using the CPI. The CPI, in turn, is based on the yearly change in prices of goods used by most Americans. Prior to 1983, the CPI measured housing prices using a procedure that included changes in the asset value of owned homes. Because the asset value of houses was growing so much faster than the consumption value, the inflation rate that included asset values was excessive. In 1983 the Bureau of Labor Statistics began using a rental equivalence approach to measure the value of housing. The official CPI-U inflation rate is based on the asset value of housing prior to 1983 and rental equivalence in 1983 and later. To provide a consistent time series, the Bureau constructed an experimental series, the CPI-U-X1, for 1967-82 based on rental equivalence. The general effect of using the CPI-U-X1 is to lower inflation in past years which in turn has the effect of lowering poverty thresholds for those years. A lower threshold means that fewer people are poor. As can be seen by comparing the first two columns in Table H-7, adjusting the poverty threshold using the CPI-U-X1 reduced the official poverty rate by between 1.3 and 1.5 percentage points in most years between 1979 and 2002. In 2002, the CPI-U-X1 reduced the poverty rate by 1.3 percentage points (11 percent or 3.8 million persons). The second adjustment in the official poverty rate made by the Census Bureau is to expand the definition of income to take into account some noncash income, including government benefits. Under the procedures by which the official poverty rate is calculated, only cash is counted in determining whether a family is poor; income from cash welfare programs counts, but benefits from food programs, medical care, social services, education and training, and housing are not included in the calculation. Moreover, because government spending on means-tested noncash benefits has increased more rapidly than spending on means-tested cash benefits over the years, ignoring noncash benefits may be an increasingly serious omission if we want a broad picture of the impact of government programs on poverty. The question of how to value noncash benefits raises a variety of substantive and technical issues. The Census Bureau has been working on these issues, consulting with academic experts, sponsoring conferences, and issuing technical reports for many years. In 1997, the Bureau published a consistent historical data series, covering the years 1979-91, to trace the impact of a variety of taxes and noncash benefits on poverty and income. The measurement of noncash benefits extended beyond government spending for the poor to include government spending programs such as Medicare that are not means tested as well as to employer contributions to employee health plans. To examine the impact on income and poverty of various State and Federal taxes, government noncash programs, employer-provided benefits, and so forth, the Bureau has adopted a framework that includes 15 definitions of income. By comparing income under these multiple definitions, it is possible to estimate the impact of the various income sources on the average income and the poverty rates of individuals and families. Income definition 14 is of interest to those concerned with the impact of government means-tested, noncash benefits on poverty rates. Unlike the official poverty rate, which includes only cash government benefits, definition 14 includes the effects of State and Federal taxes, employer-provided benefits, non-means-tested government benefits, and means-tested noncash benefits including food stamps, housing, school lunch, and the fungible value of Medicaid. TABLE H-5--COMPOSITION OF POVERTY POPULATION FOR SELECTED DEMOGRAPHIC GROUPS, SELECTED YEARS 1959-2002 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] TABLE H-6--POVERTY RATES BY FAMILY TYPE, SELECTED YEARS 1987-2002, AND PERCENTAGE OF FAMILIES AND UNRELATED INDIVIDUALS BY RATIO OF TOTAL INCOME TO POVERTY THRESHOLD, 2002 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] TABLE H-7 -- POVERTY UNDER ALTERNATIVE MEASURES OF PRICE INCOME AND INFLATION, SELECTED YEARS 1980-2002 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] The question of whether to include medical benefits when measuring poverty has great implications on poverty rates. The valuation of medical benefits is particularly difficult. Most poverty experts believe that medical coverage should not by itself raise poor individuals above the poverty line or constitute a major portion of the poverty threshold. The development of the poverty thresholds did not take into account medical costs. Although poor persons are clearly better off with medical coverage, such benefits cannot be used by recipients to meet other needs of daily living. Also, since health insurance costs are not imputed to the incomes of those above poverty, it seems inappropriate to count health benefits as income for those below the poverty line. POVERTY BY METRO AREA AND STATE Tables H-8 and H-9 present poverty rates for non-metro and metro areas and by race in non-metro and metro areas respectively. Table H-8 shows that over the period depicted poverty rates in nonmetro areas have consistently been several percentage points higher than in metro areas, but several percentage points lower than in central cities only, which consistently have had the highest poverty rates. For all three areas, poverty rates in 2002 are well below their most recent 1993 peak rates in all three areas. For nonmetro areas, the 2002 poverty rate was 17 percent below its 1993 value; for metro areas, 21 percent below; and for central cities only, 22 percent below. TABLE H-8 -- POVERTY RATES IN NONMETRO AND METRO AREAS, SELECTED YEARS 1980-2002 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] Table H-9 shows that despite recent progress, poverty among blacks and Hispanics remains much higher than poverty among whites in metro areas, non-metro areas, and inner cities. Table H-10 presents poverty rates by State for 1988-2002, based on 3-year averages. The data are shown as 3-year averages due to poor statistical reliability of State poverty rates in a single year, resulting from small sample sizes. TABLE H-9 -- PERCENTAGE OF PERSONS IN POVERTY BY RACE, BY METRO AND NONMETRO RESIDENCE, 2002 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] TRENDS IN FAMILY COMPOSITION AND INCOME, 1967-2002/1/ In the past 30 years, the level of and inequality among family incomes has changed significantly according to all income measures. Between 1967 and 1973, income increased for all quintiles, and income inequality went down. As measured by the Congressional Budget Office, over this time period the lowest quintile experienced an increase in mean adjusted family income (AFI; family income divided by the poverty threshold for the appropriate family size) of 30 percent, while income for the highest quintile grew by 21 percent. Since 1973, income of the bottom quintile declined through the remainder of the 1970s and the 1980s, and rose modestly in the 1990s. Income for the highest quintile has risen through this period. While the general trends in families' economic well-being are similar regardless of how measured, varying results for the distribution of family incomes are obtained depending on which income measure is used. Three commonly used income measures (all adjusted for inflation) are family cash income, family cash income per capita, and AFI. While no measure perfectly captures the economic well-being of families, AFI most accurately accounts for differences in family size by incorporating the scale implicit in the official Federal poverty thresholds. Family composition in the United States has undergone pronounced changes since 1973 (Table H-11). The number of married couples with children has been almost flat since 1973. By contrast, the number of families headed by a single mother grew by 93 percent over the entire 1973-2000 period, the number of non-elderly childless units grew by 105 percent, and the number of elderly childless units grew by 61 percent. Changes in family composition also are reflected in the number of persons and earners per family. The average family has become smaller, reflecting in part relatively fewer families with children (and fewer children in those families). The average family also had fewer earners in 2000 than in 1973. TABLE H-10 -- STATE POVERTY RATES: 3-YEAR AVERAGES, 1988-2002 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] DEFINITIONS AND METHODS Analyzing trends in the distribution of family incomes over time requires making decisions about a number of variables: How should variation in incomes be measured? What is the appropriate timeframe over which to examine changes? How should inflation be taken into account? And, finally, what is the appropriate measure of income to use? Measuring Variation Most of the data in this section are presented for income quintiles, each of which represents one-fifth of the income distribution (either families or persons, as indicated). Quintiles are calculated by ordering all relevant family units from those with the lowest income to those with the highest. For the analysis of changes in incomes among different types of families, quintiles are defined separately for each family type. The analysis of changes in the distribution of family incomes over time is done by examining average incomes, adjusted for inflation, by income quintile for specific types of families. Timeframe Most of the analysis focuses on data for 4 years: 1973, 1979, 1989, and 2000. The first 3 years reflect peaks in the business cycle, and allow comparisons to be made across time periods in which general economic conditions were broadly similar. Information also is presented for 2000, the most recent year for which data are available. Income data provided by the Census Bureau to outside researchers are frequently limited in certain ways both to protect confidentiality and to reduce the impact of reporting and coding errors on statistical calculations. Beginning with information for 1995, the Census Bureau substantially increased the maximum earnings it reports for individuals on public-use computer files. As a result, comparisons of incomes for high-income individuals and families in years before and after 1995 may reflect actual differences in their economic circumstances, differences in the way their income is coded, or both. To account for this reporting change, income data for 2000 are presented here in two ways. First, individuals' earnings for 2000 are limited to (or top-coded at) the same inflation-adjusted value they were limited to in 1989 ($99,999 in 1989; $138,870 in 2000.) Second, individuals' earnings in 2000 are presented the same way they are reported on the Census Bureau's public-use files ($1 million upper limit). Adjustment for inflation To examine changes in family income over time, the dollar amounts must be adjusted for inflation to compare actual buying power. Adjustment for inflation is done here using the CPI-U-X1, a revised version of the official Consumer Price Index that provides a consistent treatment of the costs of home ownership over the years examined. The CPI-U-X1 is an index of the cost of a market basket of goods and services representing the average consumption of the urban population (Table H-7). INCOME MEASURE The purpose of examining the distribution of family incomes over time is to analyze changes in family economic well-being. Two important issues in choosing an appropriate income measure are how to adjust for differences in family size and what to include as income. One measure is real family cash income, which is the sum of wage, salary, and self-employment earnings, private pension and retirement income, interest and dividends, and government cash transfers received by each family member. By this measure, which takes inflation into account but not changes in family size, noncash transfers, or taxes, the average income of families increased throughout the 1973-2000 period (Table H-12, top panel). However, the increases were uneven over time and among families with different levels of income. Regarding the former, the period from 1973 to 1979 was one of relatively slow growth in family income while the period from 1979 to 1989 saw more rapid growth. The period from 1989 to 2000 saw growth roughly comparable to the prior decade under one measure and more robust growth under the income measure that allows more income in the top quintile. It is notable that for the 60 percent of American families in the middle- and upper-income quintiles, average income growth over the decade of the 1980s is stronger than growth during the preceding period, when a similar method of computing income in the upper quintile is used for both periods. Further, average income growth during the period 1989-2000 exceeded growth during the 1980s across all income quintiles, with the exception of the highest, when using this comparable measure. These figures for mean family growth over the three periods mask very large differences in the patterns of growth in the five income quintiles. The table shows clearly that progress in family income over the 1973-89 period was negative for the bottom two income quintiles. This was reversed by modest increases in income for these two quintiles from 1989 through 2000. By contrast, growth was consistently substantial for the upper two income quintiles, especially after 1979. TABLE H-11-FAMILY COMPOSITION AND NUMBER OF EARNERS PER FAMILY, SELECTED YEARS 1973-2000 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] TABLE H-12 -- ALTERNATIVE MEASURES OF FAMILY INCOME BY INCOME QUINTILE AND CHANGE OVER TIME, SELECTED YEARS 1967-2000 FOR ALL FAMILIES [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] Examining the income data by quintiles also shows why the two measures of computing family income for the 1989 through 2000 period yield such different estimates of income growth; namely, $45,797 to $48,955 or 6.9 percent under one definition versus $45,797 to $51,701 or 12.9 percent under the other. Not surprisingly, the decision to allow more income at the top of the distribution has an impact only on the top income quintile (see the last two columns of the top panel). More specifically, income growth in the top quintile under the more restricted income definition is only from $107,925 to $118,760 or 10.0 percent, whereas growth under the broader income definition used by the Congressional Budget Office starting in 1995 is from $107,925 to $132,487 or 22.8 percent. Thus, the difference in the two measures of average family income growth over the 1989-2000 period is accounted for entirely by the top quintile. Family cash income has several shortcomings as a measure of change in economic well-being. Most notably, it fails to take into account change in family size and composition: a family of one with $30,000 in income is treated as being as well off as a family of four with $30,000 in income. This assumption is inappropriate, however, as a family of four requires more income to attain the same standard of living as a single person. An alternative approach to measuring family economic well- being is to take advantage of the family size adjustment implicit in the official Federal poverty thresholds. This scale assumes, for example, that a family of four needs about twice as much income as a single person to attain an equivalent standard of living (Table H-13). The equivalence scale implicit in the poverty thresholds may not perfectly capture the disparate needs of families of different sizes, but it yields a better assessment of relative economic well-being than making no adjustment (mean family cash income) or assuming no economies of scale (mean family cash income per capita). The AFI measure shown in the second panel of Table H-12 incorporates the equivalence scale underlying the poverty thresholds. Each family's pretax cash income is divided by its poverty threshold, yielding family income as a multiple of poverty. Thus, for example, the average family in the middle quintile in 2000 had an income of 3.49 times its poverty threshold. By taking family size into account, the AFI measure greatly reduces the income losses in the bottom two quintiles over the 1973-89 period. In fact, it completely eliminates income losses in the second quintile. It also increases the income gains experienced by the top three income quintiles. The obvious conclusion to be drawn from the comparison of the two income definitions is that taking family size into account substantially improves the picture of family income changes over the years since 1973. However, as Chart H-1 shows, the difference in income between the top and bottom quintiles, even under the AFI measure, grew substantially throughout the 1973-2000 period. CHART H-1--RATIO OF AVERAGE ADJUSTED FAMILY INCOME OF HIGHEST QUINTILE TO AVERAGE ADJUSTED FAMILY INCOME OF LOWEST QUINTILE, 1973-2000 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] INCOME SHARES Another way of tracking income trends is to look at changes in the percentage share of income received by families in each quintile. Income shares measure whether families have gained or lost in relative terms. That is, a given quintile may receive a smaller share of real income even as its average income has increased. All three income measures (family cash income, AFI, family income per capita) show broadly similar trends in the share of income received by each quintile (Table H-14). In general, between 1973 and 2000, the shares of the lowest four quintiles fell, and the share of the top quintile rose. The measures show somewhat different patterns of shares at any point in time, however. For example, in 2000 the top quintile had 48.5 percent of income under the family cash income definition, but 45.5 percent under the AFI definition. In that same year, the bottom quintile had 3.4 percent under the family cash income definition, but 4.2 percent under the AFI definition. Even so, the income shares analysis, like the other analyses in this section, generally shows that the top quintile had an increasing percentage of the income pie over the period 1973-2000. TABLE H-13 -- POVERTY THRESHOLDS AND EQUIVALENCE VALUES FOR DIFFERENT FAMILY SIZES, 2000 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] TABLE H-14 -- SHARES OF FAMILY INCOME BY INCOME QUINTILE FOR ALL FAMILIES, SELECTED YEARS 1967-2000 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] TRENDS IN PRETAX CASH INCOMES BY TYPE OF FAMILY As we have seen (Table H-11), the composition of the typical family has changed over time. Compared with 1973 and 1979, there were fewer persons in each family in 2000, on average, and married couples with children made up a smaller fraction of all families (Table H-15). Additional insights can therefore be gained by looking at changes in incomes for specific family types. This analysis distinguishes six types of family units: 1. Married couples with children, which are families composed of a married couple living only with their own or related children, at least one of whom is under age 18; 2. Single mothers with children, which are families composed of unmarried, divorced, separated, or widowed mothers living only with their own or related children, at least one of whom is under age 18; 3. Non-elderly childless families, which are families composed of two or more related people living together, in which the family head and the spouse of the head are both under age 65 and there are no children under age 18; 4. Non-elderly unrelated individuals, which are people over age 17 and under age 65 who are not living with relatives; 5. Elderly childless families, which are families composed of two or more related people living together, in which either the family head or the spouse of the head is 65 or older and there are no children under age 18; and 6. Elderly unrelated individuals, which are people 65 or older who are not living with relatives. TABLE H-15--AVERAGE FAMILY SIZE AND NUMBER OF FAMILIES BY FAMILY TYPE, WEIGHTED BY FAMILIES, SELECTED YEARS 1973-2000 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] In addition, results also are presented for four aggregates: 1. All families with children, which comprises married couples, single mothers, and other families with children; 2. Nonelderly childless units, which comprises nonelderly childless families and non-elderly unrelated individuals; 3. Elderly childless units, which comprises elderly childless families and elderly unrelated individuals; and 4. All families, which comprises all families and unrelated individuals (i.e., the noninstitutional U.S. population). Unless otherwise noted, the analysis of changes in income for each family type listed above is based on quintiles computed for that family type. This procedure permits comparisons within, but not across, family types; the quintile in which a particular family is found says nothing about its place among all families, but measures its position in relation to families of the same type. For example, individuals in the middle quintile of single mothers with children may be in the lowest quintile of the all-families grouping. Comparisons over time show how the incomes of families of a given type compare with similar families at another time, not how incomes have changed for a particular type of family. Families may move among income quintiles as their incomes-or the incomes of other families-rise or fall; they also may change types as their members grow older, have children, marry, or divorce. In addition, the average number of members and earners within a given type of family may change over time, as may the characteristics of those persons. PRETAX ADJUSTED FAMILY INCOME Trends in incomes for different family types show more variation than trends for families overall. Between 1973 and 1979, adjusted family income (AFI) grew 12.2 percent, on average, for all families with children (Table H-16). This compares with an income gain of only 7.9 percent for all families. For families with children, average AFI fell 4.5 percent during this period for the lowest quintile, from 88 percent of poverty to 84 percent of poverty. For the highest quintile, average AFI rose 7.3 percent, compared with 7.6 percent for all families. During the 1979-89 period, the bottom two quintiles of families with children experienced reduced income, by 11.7 percent and 4.1 percent respectively for the lowest and second quintiles; meanwhile, the highest quintile had an income increase of 17.0 percent. These losses at the bottom were greater for families with children than for all families. Most of the divergence in incomes among families with children reflects compositional change, as families of single mothers with children became increasingly common (Table H-11). The lowest quintile of married couples with children had a 3.0-percent decline in average AFI between 1979 and 1989; the lowest quintile of single mothers with children fared much worse, with a 22.0-percent decline during the same period. These two family types as a whole, however, showed income gains over the period: 11.2 percent for married couples with children and 3.3 percent for single mothers with children. More recently, during the 1989-2000 period, all quintiles of both family types have experienced rising incomes. Single mothers in the bottom experienced the greatest increases in income, far exceeding increases among married families with children during this period. These developments in the bottom quintiles are almost certainly due to increased work by poor and low-income mothers in general and by mothers leaving welfare in particular (see Appendix L). TABLE H-16 -- AVERAGE PRETAX ADJUSTED FAMILY INCOME (INCOME AS A MULTIPLE OF POVERTY) BY FAMILY TYPE AND INCOME QUINTILE, WEIGHTED BY PERSONS, SELECTED YEARS 1967-2000 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] Elderly persons experienced income gains across the board between 1973 and 2000. For elderly childless units, which include both single persons and married couples, average AFI rose 10.5, 13.4, and 1.8 percent respectively for the lowest quintile across the three periods shown in the last panel of table H-16 and 4.7, 26.0, and 12.7 percent respectively over the same periods for the highest quintile (using the new method of income coding). Despite their gains, the elderly generally had much lower incomes than the non- elderly. In 2000, for example, the average income of elderly childless units was about 3.9 times poverty; the average income of non-elderly childless units, by comparison, was about 5.6 times poverty (not shown in table). AVERAGE FAMILY CASH INCOME BY FAMILY TYPE For all families, average cash income grew more slowly than average pretax AFI between 1973 and 2000. This was also generally true for specific family types. At the same time, those groups of families whose average cash incomes declined had more pronounced decreases than occurred in pretax AFI. Average family cash income grew throughout the 1973-2000 period for families with children (Table H-17, second panel). However, families at the bottom of the income distribution lost ground during the 1973-89 period, with income declines of 11.0 percent during the 1973-79 period and 17.7 percent during the 1979-89 period. The decline stopped between 1989 and 2000 when the income of families with children in the bottom quintile increased at a faster pace than all quintiles except for the highest. As was the case with all the measures we have examined, average family cash income of families in the top two quintiles improved substantially throughout the entire period after 1973. As compared with the cash family income losses in the bottom quintile for all families, the pattern of losses in the bottom quintile was even greater for single mothers with children before 1989 (Table H-17, fourth panel). From 1979 to 1989, for example, these mothers lost almost a quarter of their income. However, between 1989 and 2000 this group made up for at least some of the lost ground as their income increased by 28.0 percent. During this period, which included strengthened efforts to encourage and support low-income parents, many of them single mothers, in work, income gains by single mothers with children in the lowest, second, and middle quintiles far exceeded gains at the top of the income spectrum for this group. It is also interesting that during both the 1973-79 and the 1989-2000 periods, income gains in the second, third, and fourth income quintiles of single mothers with children were usually greater than income gains in the top quintile. Because the change in family size among elderly persons was almost negligible over the period, their trend in average family cash incomes is almost identical to the trend in average pretax AFI. Elderly childless units and elderly childless families experienced income gains in every quintile during every period between 1973 and 2000. TABLE H-17--AVERAGE FAMILY CASH INCOME BY FAMILY TYPE AND INCOME QUINTILE, SELECTED YEARS 1973-2000 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] TABLE H-18 -- FAMILY CASH INCOME LIMITS1 BY QUINTILE AND FAMILY TYPE, SELECTED YEARS 1973-2000 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] Table H-18 shows family cash income limits (the income cutoffs between quintiles) by quintile and family type. Between 1973 and 1989, income limits among families with children declined or grew slowly while those for the elderly increased, in some cases significantly. This pattern reversed itself in the 1989-2000 period, as income limits for families with children grew at roughly twice the pace as among elderly childless units. In general during the 1973-2000 period, income limits among the higher quintiles increased more than among the lower quintiles. In fact, income limits for the lower quintiles have decreased for several family types during several periods. A notable exception involves the limits among single mothers with children. Following declines in the 1979-1989 period, income limits for this group rose sharply in the 1990s, with especially steep increases at the bottom of the income ladder. For example, the 36.8 percent increase noted for the lowest quintile of single mothers with children was the greatest for any group during any period from 1973-2000. ANTIPOVERTY EFFECTIVENESS OF VARIOUS CASH AND NONCASH TRANSFERS Tables H-19 through H-21 provide estimates of the number and percentage of individuals removed from poverty by market income and by social insurance programs (Social Security, unemployment compensation, and workers' compensation), means-tested cash programs (Aid to Families with Dependent Children, Supplemental Security Income, and general assistance), means-tested noncash programs (food stamps, housing benefits, and school lunch), and Federal payroll and income taxes and the earned income credit (EIC). Tables are provided separately for elderly persons, for children, and for persons in units with an unmarried head and children under age 18, for selected years between 1979 and 2002. The tables present alternative measures of poverty to the official measure. They include counts of the number of people below the poverty line before any government benefits are taken into account, after each type of benefit is added to income, and after the government cash and noncash benefits and Federal taxes and the EIC are added to (or subtracted from) income. The tables also measure the effect of these government programs on the "poverty gap" - the gap between a poor family's income and the poverty line. The poverty gap represents the degree of poverty by showing the amount of money that would be needed to lift every poor person exactly to the poverty line. TABLE H-19 -- ANTIPOVERTY EFFECTIVENESS OF CASH AND NON-CASH TRANSFERS (INCLUDING FEDERAL INCOME AND PAYROLL TAXES) FOR ALL PERSONS AGE 65 AND OLDER, SELECTED YEARS 1979-2002 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] Table H-19 shows the anti-poverty effectiveness of market income and government programs for the elderly. Based both on cash income before transfers and on post-transfer income, the poverty rates among the elderly in 2002 were among the lowest on record. As compared with 1979, when over 54 percent of the elderly were poor before transfers, by 2002 only about 50 percent of the elderly were poor before transfers. The comparable figures for the percentage of the elderly in poverty after transfers were 13.5 in 1979 and 9.0 in 2002. The impact of Social Security transfers is by the far the greatest reason so many of the poor are removed from poverty by government transfers. In 1979 the poverty rate was dropped from 54.2 to 17.4 by Social Security payments; in 2002 the comparable figures were 49.9 to 11.5 percent. In 1979, a total of 8.9 million elderly persons were removed from poverty by Social Security; in 2002, the number had jumped to 13.1 million. The figures for the poverty gap for the elderly are not quite as impressive as the overall figures. Both the total number of dollars required to close the poverty gap and the size of the poverty gap per person in poverty have been almost stagnant in recent years. Even so, in 2002 the poverty gap is only $7.0 billion or $2,284 per person in poverty. As we will see, no other government program has as huge an impact on poverty among any group as does Social Security among the elderly. The impact of market income and the safety net on children's poverty are shown in Table H-20. The poverty rate among children before transfers was 19.7 percent in 2002, among the lowest levels since 1979 and more than 6 percentage points lower than in 1993. Similarly, the child poverty rate after transfers in 2002 was 12.6, its lowest level since 1979 and 7.4 percentage points or 37 percent below its level in 1993. These figures show substantial progress against children's poverty, both before and after government transfers. That the pre-transfer level is so low suggests that the substantial increase in work by former welfare mothers after the 1996 welfare reform legislation (see Appendix L) has played an important role in poverty reduction among children. The important role of work by single mothers in reducing child poverty is also shown by the data on percentage of children removed from poverty due to Federal taxes. The row of figures for taxes in all the panels of Table H-20 show that Federal tax policy is having a major and growing effect in reducing child poverty. In 1983, Federal taxes actually increased the poverty level among children by 5.1 percent. However, as the Federal Government reduced taxes and increased the EIC for low-income families with children by enacting reform legislation in 1986, 1990, 1993, and 2001, the impact of taxes actually became positive. By 2002, EIC payments to families reduced the child poverty rate from 14.8 percent to 12.6 percent. It seems reasonable to conclude that the effectiveness of the EIC in fighting poverty can be attributed to two factors-the increasing generosity of EIC policy itself and the increase in work by low-income families with children, especially families headed by mothers. TABLE H-20 -- ANTIPOVERTY EFFECTIVENESS OF CASH AND NON-CASH TRANSFERS (INCLUDING FEDERAL INCOME AND PAYROLL TAXES) FOR ALL CHILDREN UNDER 18, SELECTED YEARS 1979-2002 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] TABLE H -21 -- ANTIPOVERTY EFFECTIVENESS OF CASH AND NON-CASH TRANSFERS (INCLUDING FEDERAL INCOME AND PAYROLL TAXES) FOR PERSONS IN UNITS WITH AN UNMARRIED HEAD AND RELATED CHILDREN UNDER 18, SELECTED YEARS 1979-2002 [GRAPHICS NOT AVAILABLE IN TIFF FORMAT] Data on the poverty gap for children are somewhat mixed. Data on the pre-transfer poverty gap are uniformly positive. Despite the fact that the number of children grew by over 10 million or nearly 17 percent between 1983 and 2002, the poverty gap before transfers nonetheless fell from almost $49 billion to about $37 billion, in constant 2002 dollars, a real decline of 23 percent. Similarly, the pre-transfer poverty gap per poor child in 2002 continued to fall to $2,615, its lowest level during this period. However, the post-tax, post-transfer poverty gap per poor child rose during this period. This period saw significant declines in receipt of cash and other welfare benefits, and means-tested cash (and, especially in recent years, non-cash) benefits have become generally less effective in removing children from poverty; as shown in the middle two rows of the last panel of Table H-20, taken together these policies reduced by only 29.5 percent in 2002 as compared with 46.3 percent in 1979, 45.3 percent in 1989, 43.0 percent in 1996, and 34.3 percent in 1999. The effect of the EIC in reducing the poverty gap, however, remained potent; in fact, at 6.2 percent it was greater than in any previous year shown other than 1999. Despite the effectiveness of the EIC, the overall impact of government programs reduced the poverty gap less than in any previous year. The major reason for the reduced effectiveness of government programs in reducing the poverty gap seems to be a decline in the impact of means-tested cash benefits. In 1979 these benefits reduced the poverty gap by 28.8 percent. By contrast, in 2002 they reduced the poverty gap by only 10.9 percent. Undoubtedly, the decline in the welfare rolls and in cash benefits from the Temporary Assistance for Needy Families Program play an important role here. Poverty data for persons in units headed by single parents is generally consistent with the data for children. The first point to emphasize with these data (see the top row of Table H-21), which simply reinforces the conclusion from the data on single mothers in Table H-15, is that there has been a very large increase in the number of persons in families with unmarried heads. The number jumped from 23.5 million in 1979 to 38.4 million in 2002, an increase of more than 60 percent. By contrast, the number of persons in married couple families increased from 101.3 million to only 104.7 million in 2000 (see Table H-11), an increase of about 3 percent. Thus, the family type with the highest poverty rate has been increasing more than 10 times as fast as the family type with the lowest poverty rate. These demographic developments make progress against poverty somewhat difficult. Even so, the pre-transfer poverty rate among persons in families with an unmarried head nearly matched its lowest level ever in 2002 at 37.6 percent. Compared with the 53.8 percent pre- transfer rate in 1983, that's a drop of over 30 percent. Progress against pre-transfer poverty among these families was continuous and rapid during the 1990s economic expansion, with a drop of 21 percent from 49.5 percent in 1990 to 39.3 percent in 1999. Despite the 2001 recession, this figure continued to fall, to 37.6 percent in 2002. Again, as we saw in the case of children, progress against pre- transfer poverty has been substantial in recent years, in all likelihood due to the increase in work by single mothers. On the other hand, again as was the case with children, progress against poverty as measured by the poverty gap has been uneven. Although the pre-transfer poverty gap at $42.8 billion for these families nearly matches its low since 1979, and although the gap has fallen 22 percent just since 1995, means-tested cash programs have been increasingly less effective in reducing the poverty gap. By contrast, as with children, Federal tax policy has been more effective at reducing the poverty gap. Even so, the combination of the reduced pre-transfer poverty gap and the increased effectiveness of the EIC in reducing the poverty gap failed to outweigh the declining effectiveness of means-tested cash and noncash transfers in reducing the poverty gap. As a result, the reduction in the post- tax, post-transfer poverty gap for these families was lower than in the past as measured either by dollars or percentage reduction. REFERENCES U.S. Census Bureau. (1986). Money income and poverty status of families and persons in the United States: 1985. Current Population Reports (Series P-60, No. 154). Washington, DC: U.S. Government Printing Office. U.S. Census Bureau. (1993). Money income of households, families, and persons in the United States: 1991. Current Population Reports (Series P-60, No. 180). Washington, DC: U.S. Government Printing Office. U.S. Census Bureau. (1996 and various years). Income, poverty, and evaluation of noncash benefits, 1994. Current Population Reports (Series P-60, No. 189). Washington, DC: U.S. Government Printing Office. [Several publications from the P-60 Series were used in preparing the tables for this chapter. These include numbers 124, 140, 145, 149, 154, 157, 161, 166, 168, 174, and 180.]