This document is available here as an Adobe Acrobat PDF.

September 2007

Each year, the Census Bureau reports on the nation's poverty rate, based on the number of people with incomes below the official poverty line, as adjusted annually for inflation. In 2006, the poverty line, which varies by family size, was $16,079 for a family of three, and $20,614 for a family of four.

According to this year's report on Income, Poverty, and Health Insurance Coverage in the United States (issued on August 28), poverty fell from 12.6 percent to 12.3 percent[1] (about 36 million people). This was the first statistically significant decline since 2000, largely because of the economic slowdown between March and November 2001 and the relatively slow recovery since then.

As usual, there was much media commentary about poverty's intractability. And, in fact, today's official poverty rate is hardly lower than the 1968 rate, when it was about 12.8 percent, reinforcing the feeling that little progress has been made. But a closer look at forty years of poverty data, made possible by a new data series that goes beyond the official poverty measure, tells a more mixed story.

A Rising Tide

As figure 1 shows, since the 1960s, year-to-year changes in the poverty rate have been roughly correlated (rising or falling) with unemployment. Underlying this blunt indicator are more direct factors such as changes in employment and wage rates.

Last year's poverty decline followed the expected pattern. Between 2005 and 2006, unemployment fell from 5.1 percent to 4.6 percent. More significantly, the single-mother unemployment rate fell from 9.2 percent to 8.5 percent. As a result, the reported poverty rate declined from 12.6 percent to 12.3 percent.

This is good news, and it demonstrates, once again, the wisdom of the aphorism repeated by John F. Kennedy: "A rising tide lifts all boats."[2] This has not been true for all groups, however, as discussed below.

One wishes that the decline had been as large as some of those in the 1990s, and that it had benefited as many demographic groups. It appears that the decline in poverty was due to more people working more hours--not to higher wages. It is not necessarily bad that the poor are doing better because they are working more, but sustained declines in poverty require both more work and higher wages.

Long-Term Poverty Trends

Long-term trends are ordinarily much more important than year-to-year comparisons, although small yearly increases (or decreases) in poverty can quickly add up to significant changes in real poverty. Over the long term, besides trends in employment and wages, the poverty rate is also determined by the following:

• Government cash transfers (mainly Social Security because non-cash benefits and the Earned Income Tax Credit are not counted in the official poverty measure)
  
• Demographic factors such as changes in divorce and nonmarital birth rates (that produce more single-mother families with high poverty rates), and increases in the number of immigrants (higher poverty) or the elderly (now lower poverty)

The following briefly sketches the impact of both factors. Neither, however, is fully understood using only the official poverty measure. To do so, one needs a more nuanced view of poverty, disaggregated by key demographic groups.

Money Income vs. Disposable Income

The poverty measure has been widely criticized by analysts from the left and the right for under- or overstating poverty. But--like many observers--I despair of ever achieving wide agreement over a reformed poverty measure and have grown to believe that, instead, it would be preferable to develop various additional measures that broaden and enrich our understanding of contemporary poverty.

That is what is so important about recent Census Bureau efforts to develop four additional definitions of income: money income, market income, post-social insurance income, and disposable income. In this analysis, poverty measured by money income is contrasted with poverty measured by disposable income.

• Money income is the definition of money in the official poverty measure, which is essentially pre-tax, cash income (including wages, Social Security, and welfare). 
  
• Disposable income adds to money income the value of non-cash government benefits (including food stamps, public or subsidized housing, and free or reduced price school lunches), as well as the Earned Income Tax Credit (EITC). It deducts federal payroll taxes, federal and state income taxes, and property taxes for owner-occupied homes. Also added are imputed net realized capital gains and imputed rental income.

Elderly Poverty

As figure 2 shows, material poverty among the elderly has been all but eliminated--with the exception of some of the oldest and frailest. In 1968, about 25 percent of the elderly were poor (using either money- or disposable-income definitions). By 2005, their money-income poverty rate had fallen to 10.1 percent, and last year it declined further to 9.4 percent.

Using the more informative disposable-income definition (which includes the income value of housing), the 2004 poverty rate of the elderly (more recent data are not available[3]) was only 6.4 percent, one-third lower. And that does not consider the benefits they receive from Medicare, the new prescription drug plan, and various other programs. (Some analysts would subtract their out-of-pocket medical expenses from disposable income, which would raise elderly poverty to 9.4 percent.)

Child Poverty

Child poverty, however, continues to be a serious problem (as figure 2 also shows). Using money-income poverty (the official poverty rate), the number of children in poverty has changed little from its 1980 level--twenty-five years ago. (Last year's rate of 17.4 percent was not statistically different from the 17.6 percent rate the year before.)

Using the disposable-income definition of poverty, however, tells a more positive story because it captures the impact of increased aid to the poor since the 1960s. In 1968, with little in the way of food stamps, housing assistance, the EITC, and so forth, there was no difference between money- and disposable-income poverty. But by 1980 and continuing through 2004, counting these benefits cuts the child poverty rate by roughly a quarter to 13 percent.

Child poverty is both the most politically salient form of poverty and the most difficult to address. It is largely the product of trends in family composition and immigration that cannot be cured by simple income transfers, which can have unintended consequences.

Single-Mother Poverty

Family structure and family income are highly correlated. If adjusted for family type, the income differences among racial and ethnic groups shrink substantially. Overall, white family incomes are almost twice that of African-American family incomes, but, for all three groups, the incomes of married-couple families with children and of female-headed families with children are much closer (see figure 3). The differences that remain are partly explained by education level, the age of parents, and the family's time in this country.

In fact, family composition, especially single motherhood, continues to be a major characteristic of poverty. As figure 4 shows, in 2006, single-mother families were almost six times more likely to be poor than their married-couple counterparts--36.5 percent versus 6.4 percent. Brookings scholars Isabel Sawhill and Adam Thomas have estimated that if single-mother rates had remained at their 1970 levels, poverty in 1998 would have been 24 percent lower.

Contrary to common assumption, there has been a forty-year decline in single-mother poverty--down by 20 percent--driven by both increases in government benefits and higher earnings (see figure 5).

Between 1968 and 1980, single-mother poverty declined largely because of the growth of means-tested programs. For both white and black single mothers (there are no data for Hispanic single mothers in 1968), money-income poverty was essentially unchanged (at very roughly about 30 and 60 percent, respectively). But for both groups, in 1980, disposable-income poverty was about a quarter lower--again, because many of these families received food stamps, housing assistance, the EITC, and so forth.

Between 1980 and 2004, things changed: the greatest impact on poverty came from earnings increases, especially for African Americans and Hispanics. The decline in their poverty was almost entirely driven by their money income (that is, higher earnings), with the disposable-income poverty rate lower simply because of the decline in the money-income poverty rate. Some of this rise in earnings was undoubtedly a product of welfare reform, but clearly broader economic and social forces were at work--and probably played a larger role.

Nevertheless, a single-mother poverty rate of over 36 percent is much too high and compromises the life chances of millions of children. Much more progress is needed against this deep-seated and complex form of poverty.

Poverty among Minorities

African-American men have, on average, made substantial earnings gains since the 1960s, but they remain a group doing poorly--much worse than white or Hispanic men (see figure 6). The official poverty rate does not capture the depth of their economic problems. African-American men are about two-and-a-half times more likely to have zero income than either white or Hispanic men (18 percent for blacks compared to 7 percent for whites and Hispanics).

Hispanic poverty cannot be considered without also weighing the impact of immigration. In the last thirty years, the United States has experienced a tremendous amount of immigration, largely Hispanic. In 1975, Hispanics made up 5.3 percent of the population (about 11.1 million people); they now make up 15.1 percent (about 44.8 million people). If the percentages of Hispanics in this country had remained the same, the U.S. population would be 29 million lower, 267 million instead of 296 million.

This influx has resulted in a higher U.S. poverty rate, mainly because so many immigrants are low-skilled workers or women with young children. If the proportion of Hispanics in the population in 2006 had been the same as it was in 1975 (and they had the same poverty rate as now), then the overall U.S. poverty rate in 2006 would have been 7 percent lower (11.4 percent rather than 12.6 percent). That would be 2.4 million fewer people in poverty.

On the other hand, the poverty rate of Hispanics has dropped by a third from its high twelve years ago, falling from 30.7 percent in 1994 to 20.6 percent in 2006. Last year's figure represented another 5.5 percent decline (see figure 7). Moreover, the poverty rate for foreign-born persons (many of whom are Hispanic), declined even more, falling from 16.5 percent to 15.2 percent, about an 8 percent decline.

Considering the large and continuing influx of Hispanic immigrants, one would expect that their poverty rate might have increased, or at least remained flat, as millions of new immigrants entered the country. That the poverty rate of Hispanics is falling is a testament to their work effort and the opportunities they have found in this country.

One reason Hispanics are making economic progress is their success in obtaining what might be considered skilled blue-collar jobs--apparently at the expense of African Americans. In the twelve year period from 1994 to 2006, for example, as the total number of workers in precision production, craft, and repair occupations grew, the percentage of these jobs held by whites fell, the percentage held by blacks remained constant, and the percentage held by Hispanics more than doubled (see figure 8). Thus, as whites have left these relatively well-paid jobs, Hispanics, rather than blacks, have moved into them.

Income Inequality

Income inequality (or, more neutrally, income dispersion) continues to grow, at least by the way the Census Bureau counts it. Those at the top seem to be doing vastly better than the rest of America.

A common method of measuring income inequality is the Gini coefficient, which is an index that varies from zero (perfect equality) to one (perfect inequality). The higher the Gini coefficient, the greater the degree of income inequality. The Gini coefficient for family income, for example, was 0.376 in 1947, and then moved up and down for several periods until falling to 0.348 in 1968. Since then, it has risen steadily to 0.440 in 2005, indicating a higher degree of income dispersion. As of this writing, the Census Bureau has not released updated 2006 income quintile data for families, but the data it has released for households suggest that the trend continues.

But the Gini coefficient is an abstract concept, especially for laypersons. Figure 9 shows family income by quintile in 1968, 1980, and 2005. While incomes in the bottom two quintiles remained essentially flat for almost forty years, those in the top three rose by two-thirds, and those in the highest quintile almost doubled. Public (and political) attention has naturally focused on this difference--and the apparently phenomenal success of those in the highest quintile, where incomes have almost doubled since 1986 and are now apparently almost ten times those in the bottom quintile.[4]

These are real differences, but probably not as great as they appear. First, changes in the "top coding" of income have exaggerated the top quintile's income rise. In 1985 and again in 1993, the Census Bureau raised the highest earnings counted in the CPS from $99,999 to $299,999 and then from $299,999 to $9,999,999, although the highest amounts are effectively limited to $1 million when the Census Bureau processes the data; the change in 1993 raised the levels for several other income categories as well. The result was a one-year increase in reported income of the top quintile of about 10 percent. We very roughly estimate that the result is a more than $20,000 rise in the top quintile's average income.

Second, the quintiles are not rigid demarcations of economic or social situation. Many people move up or down, from one quintile to another--as people lose and gain jobs, have children, or retire (or slow down in preparation for retirement). For example, according to the Survey of Income and Program Participation (SIPP), roughly one-third of the households that were in the top income quintile in 1996 were not there three years later.[5] Likewise, roughly 38 percent of those in the bottom quintile were not there three years later.[6] Nevertheless, many live on these low incomes for a very long time.

Such large disparities in incomes are of increasing social significance and political import, but it is difficult to see them narrowing very much without dramatic changes in society and the economy because the underlying trend is the product of major economic and social factors: increasing returns to education and skills (including greater dispersion of incomes within and among job categories), changing family composition (especially so many single mothers and unmarried people living alone), and differences in the number of hours worked.

Conclusion 

Disaggregating poverty data by demographic group and using disposable income to measure poverty tells a much more promising story about long-term poverty trends. Liberals will see these figures as affirming their calls for massive government transfers. Conservatives will see them as affirming the value of education and work. But the larger message is that--looking behind the simple metric of the official poverty measure--substantial (albeit uneven) progress has been made against poverty. That is good news for us all.

Douglas J. Besharov (dbesharov@aei.org) is the Joseph J. and Violet Jacobs Scholar at AEI and a professor at the University of Maryland School of Public Policy. This analysis was prepared with the assistance of Gordon Green, a research associate at AEI.

Notes

1. Percentages and figures used throughout this essay are approximate.

2. Although little noted, President Kennedy was actually speaking about how a rise in the income of one state would raise the incomes of other states. He was using the argument to justify a public works project in Arkansas, but his aphorism is now widely applied to the poverty of individuals. See John T. Woolley and Gerhard Peters, "John F. Kennedy: Remarks in Heber Springs, Arkansas, at the Dedication of Greers Ferry Dam," The American Presidency Project, available at www.presidency.ucsb.edu/ws/index.php?pid=9455 (accessed September 26, 2007).

3. This analysis compares three years: 1968 (the first year for which we have data), 2004 (the last year for which we have data), and 1980 (a year after most social welfare expansions took place but before the increases in poverty because of the recession).

4. Using a disposable-income definition, so important for understanding poverty, does not raise incomes at the bottom enough to change this picture.

5. U.S. Census Bureau, "Dynamics of Economic Well-Being: Movements in the U.S. Income Distribution, 1996-1999," Current Population Reports, Series P70-95, July 2004, 3, available at www.sipp.census.gov/sipp/p70s/p7095.pdf (accessed September 5, 2007).

6. Ibid.

 This document is available here as an Adobe Acrobat PDF.