This document is available here as an Adobe Acrobat PDF.

A version of this article appeared in the Fourth Quarter 2008 Milken Institute Review. 

November 2008

The proportion of people living in poverty has increased slightly since the 1970s, according to the official poverty measure. Since the "poverty threshold" used for counting the poor is fixed and unchanging, those numbers suggest a disturbing rise in absolute want. But data on household spending show consumption growth even for those with low incomes. The official poverty rate is seriously flawed, and many analysts are ready to scrap it.

Washington regularly collects vast amounts of data for hundreds upon hundreds of social and economic indicators bearing on poverty. But within that compendium, a single number is widely taken to be more important than the others--the so-called official poverty rate (OPR), which is based on the federal poverty measure established in the 1960s. For four decades, that rate has served as the benchmark for both policy analysis and public discourse regarding the national struggle to reduce the deprivation in our midst. Yet even a casual examination shows that this metric is deeply flawed and increasingly biased toward the overestimation of material poverty.
 
While the OPR numbers say that the proportion of the American population living in poverty has changed little--indeed, has slightly increased--since the early 1970s, data on household spending show substantial and continuing growth in consumption among those reporting very low incomes. Indeed, it is becoming increasingly clear that the OPR is of no help in figuring where we are today or even where we have come from. Signs are finally on the horizon that analysts on both the left and the right are prepared to scrap the official rate in favor of more realistic ways to track poverty.

A Little History

The poverty rate measure was introduced in 1965 in a landmark study by Mollie Orshansky, an economist and statistician at the Social Security Administration. Drawing on her own research, in which she had experimented with using household income thresholds to identify children living in impoverished conditions, Orshansky proposed a set of income criteria for setting a poverty threshold and determining who lived below it.

Orshansky's threshold was essentially a multiple of the cost of a nutritionally adequate--though humble--diet. For the food-budget anchor, Orshansky used the U.S. Department of Agriculture's "economy food plan," the lower of its two budgets for nonfarm families of modest means. She then applied a multiplier of roughly three--the number varied with family size--to calculate household poverty thresholds. The multiplier itself, incidentally, came from statistics on the ratio of after-tax income to food budgets for all Americans in the 1950s.

Using these new poverty thresholds, along with census data on income, Orshansky calculated the total population living below the poverty line for the United States as a whole, as well as for demographic subgroups, for 1963. Although Orshansky's study did not employ the term "poverty rate," talking instead about the "incidence of poverty," the term quickly came to mean the proportion of people or families below the poverty line.

Where's the Beef?

Little has changed since 1965 in the way the federal government measures poverty. The OPR is still calculated annually on the basis of poverty thresholds adjusted for inflation. The rate is and always has been a measure of absolute material poverty--one that intentionally ignores changes in the culture and the economy that influence popular perceptions of what constitutes deprivation.

Estimates of the OPR for the United States are thus available for the past forty-eight years. At first, they gratifyingly tracked the expectations of those who assumed that the rising tide of economic growth would carry all boats. The rate fell by roughly half between the late 1950s and the late 1960s for both families and individuals.

Strikingly, however, the numbers suggest virtually no improvement since then. The lowest OPR yet recorded was for 1973, when the index bottomed out at 11.1 percent. The OPR has since declined for older Americans, for people living alone, and for African Americans. But for most demographic slices--children under eighteen, families, and non-Hispanic whites--the OPR was higher at the start of the new century than it had been in the early 1970s. Low-income Hispanics were somewhat better off in 2006 than in 1973, but the difference is distressingly modest.

To go by the OPR, then, America, through three decades of both Democratic and Republican administrations, has utterly failed to improve the material lot of the more vulnerable elements of society--to raise them above the income line where, according to the author of the federal poverty measure, "everyday living implied choosing between an adequate diet of the most economical sort and some other necessity, because there was not money enough to have both."

Remember the all-boats-rising thesis? Experts on poverty long held that the OPR is largely driven by macroeconomic conditions--employment opportunities and wage rates. In a series of influential publications in the mid-1980s, for example, David Ellwood and Lawrence Summers of Harvard found that "almost all of the variation in the measured poverty rate is tracked by movements in median family income."

But their work only covered 1959–83. Since the early 1970s--the long decades of stagnation in the OPR--the correspondence between the statistic and median family income appears to have broken down altogether. In fact, over the past three-plus decades, data on the median income for American families have provided no clue to the OPR.

Additionally, since 1973, the behavior of the OPR looks increasingly aberrant when compared to other indices widely thought to bear on the risk of poverty in a modern urbanized society. In 1973, nearly 40 percent of adults over the age of twenty-five lacked a high school degree; by 2001, the figure was under 16 percent. Or consider trends in means-tested benefit programs--food stamps, housing subsidies, Medicaid, the Earned Income Tax Credit, and other programs that benefit the poor. Between the 1973 and 2001 fiscal years, spending on those programs more than tripled from $163 billion to $507 billion (in 2004 dollars) and increased by over 130 percent in real, per-capita terms.

The simplest and most plausible explanation for these seeming contradictions is that something is seriously wrong with the way the OPR is calculated. A variety of minor and major technical problems have been noted by specialists over the years--among them, the method for adjusting for inflation and the use of the food budget as the sole benchmark for income sufficiency. One other defect, however, fundamentally flaws the current approach.

The Income-Consumption Mystery

The rate calculation implicitly assumes that consumption by low-income Americans is accurately tracked by their reported incomes. In fact, there is good evidence that, for the lowest fifth of Americans on the income ladder, reported expenditures are almost twice their incomes.

Correcting for changes in household size, real expenditures per person for all Americans were 110 percent higher in 2005 than in 1960–61 for the country as a whole. We do not have precisely comparable figures for poor households, but we do know that the real expenditures of the poorest fifth of households were 112 percent higher in 2005 than the expenditures of the poorest fourth in 1960–61.

Put simply, consumption in low-income households has grown even faster than that of average American households. Moreover, other statistical evidence confirms that lower-income Americans are doing far better than the stagnation in the OPR suggests.

Food and Nutrition. In the early 1960s, inadequate caloric intake was hardly unusual among the officially defined poor. By the end of the century, however, the proportion of the adult population between twenty and seventy-four who were underweight (defined as a body mass index below 18.5) dropped from 4 percent to 1.9 percent.

By the same token, nutritional deprivation among children has been declining. According to the Centers for Disease Control and Prevention, the percentage of low-income children younger than five who were underweight dropped from 8 percent in 1973 to under 5 percent in 2005. (In the same period, the OPR for children rose from 14.4 percent to 17.6 percent.)

Housing and Home Appliances. In 1970, about 14 percent of poverty-level households were officially deemed "overcrowded," with more people than rooms to live in. By 2001, just 6 percent of poor households were overcrowded--a proportion lower than for nonpoor households as recently as 1970. Moreover, between 1980 and 2001, heated floor space per person in the homes of the officially poor increased by 27 percent. And in 2001, just 2.5 percent of poverty-level households lacked plumbing facilities--a lower share than for nonpoor households in 1970.

Trends in furnishings and appurtenances tell the same story: poor households' possession of modern conveniences has been growing rapidly. For many of these items--telephones, television sets, central air conditioning, and microwave ovens--prevalence in poverty-level households in 2001 exceeded that of median-income households in 1980.

Personal Transportation. In 1973, almost three-fifths of the households in the lowest income quintile lacked a car. In 2003, by contrast, over three-fifths of poverty-level households owned one or more cars. In that same year, moreover, 14 percent of households below the poverty line owned two or more cars, and 7 percent had two or more trucks.

Health Care. Between 1970 and 2004, the infant mortality rate fell by a remarkable two-thirds. And it continued its almost uninterrupted decline after 1973, even as the OPR for children began to rise. The disconnect is particularly striking for white infants. Between 1974 and 2004, their mortality rate fell by three-fifths, from 14.8 deaths per thousand to 5.7 deaths per thousand. Yet the OPR for white children rose from 11.2 percent to 14.3 percent.

The gains in access to medical care for infants extend to older children. The proportion of children who did not report a visit to a physician was significantly lower for the poor population in 2004 (12 percent) than it had been for the nonpoor population twenty-two years earlier (17.6 percent).

Mystery Solved?

The more striking anomaly in the OPR that cries out for explanation is that the gap between reported income and personal spending has widened sharply. Between 1960–61 and 2005, the ratio of income to expenditures for all households remained fairly stable, with expenditures exceeding income by a significant amount. The subgroup of poorer households also seemingly overspent. But in contrast to households in general, the margin by which the spending of the poor exceeded reported income has moved steadily upward. In the early 1960s, the ratio was 1.12 for the lowest income quartile. By 1972–73, that ratio had reached 1.4 for the lowest income quintile. (Note again the lack of data on precisely comparable groups.) By 2005, the ratio for the bottom fifth had reached 1.98.

What accounts for the gap? One possible hypothesis is that low-income Americans are overspending at an increasing rate--that is, going ever deeper into debt. By this reasoning, the widening gap represents an unsustainable binge that must eventually come to an end, with doleful consequences for future living standards of the disadvantaged.

The overspending hypothesis, on its face, seems plausible. Luckily, it is confuted by evidence from U.S. Census Bureau and Federal Reserve surveys showing that the average net worth of households in the bottom fifth has actually grown in the last decade. Additionally, the gains in wealth have been broadly shared, with the portion of bottom fifth households reporting no assets whatever falling from 21 percent in 1989 to just 8 percent in 2004.

If the poor are not overspending, is it possible they are underreporting income? There is little doubt. For one thing, the OPR measure of income ignores tens of billions in tax rebates delivered by the Earned Income Tax Credit. By the same token, the poor surely supplement their incomes off the books. But to use the underreporting phenomenon to explain why the gap between spending and income has widened so much, one would also need to explain why underreporting was increasing rapidly. Hence, the explanation for the widening gap more likely lies elsewhere.

Might the growing disparity be explained by another big change in modern America--namely, the rise of illegal immigration? The argument would go like this: there has been a surge of undocumented immigration over the past generation, and illegal immigrants are understandably inclined to underreport their incomes. Yet they have no similar incentives to underreport their consumption in interview-based surveys. Thus, all other things equal, as the undocumented become an ever greater proportion of the lower-income population, the gap between spending and reported income should grow. (Immigrants, furthermore, tend to be savers--think remittance flows--a fact that could also help account for the reported increases in wealth among the poor in recent decades.)

The argument is plausible, but the actual magnitude of the effect is likely to be small. Undocumented immigrants are believed to comprise less than 3 percent of all U.S. residents. Moreover, they are probably undersampled by the survey techniques used by the Census Bureau to estimate the poverty rate.

But even if all illegal immigrants were fully represented in our income and expenditure surveys, if there had been no illegal immigrants in the country in the early 1970s, if all illegal immigrants now fall in the lowest quintile of the U.S. income distribution, and if this entire group reported no income at all, the illegal immigration effect could account for less than half of the rise in the ratio of spending to income that was actually reported by the bottom fifth of American households between 1972 and 2005. In reality, the impact is probably much smaller. If we want to understand the uncanny continuing divergence between reported spending levels and reported income levels for the lowest fifth of American households, then we are going to have to look into the economic circumstances of legal American residents--the overwhelming majority of whom are native-born.

To see where we are heading, note that poverty status is not a fixed, long-term condition for the overwhelming majority of Americans who are ever designated as poor. Quite the contrary: long-term poverty appears to be the lot of only a tiny minority counted as poor in any particular year by the OPR. For example, the Census Bureau found that from 1996 to 1999, fully 34 percent of all households spent two months or more below the poverty line--but only 2 percent stayed below the line in all forty-eight of those months. Both economic theory and common sense suggest that the temporarily poor would try to maintain their living standards in lean times by spending more than they earn.

But this alone would not explain why the spending-income gap increased so much in recent decades. What is needed is a reason to believe that household incomes are more variable than they used to be, sharply increasing the portion of the materially disadvantaged that are only temporarily poor. Here, the accumulating evidence is intriguing. For example, Jacob Hacker of Yale University found that, for households headed by people of working age, the odds of seeing income fall by half or more in the coming year rose from 7 percent in 1970 to 16 percent in 2002.

This unintuitive explanation for the growing income-consumption gap meshes neatly with another surprising bit of survey data. According to the Federal Reserve, differences in net worth for the bottom quintile and the next highest quintile of American families narrowed between 1989 and 2004--hardly what one would expect in light of evidence of growing income inequality since the 1960s. If year-to-year income volatility were on the rise, however, we would expect an increasing share of families permanently lodged in the second quintile to register temporarily in the bottom quintile in any given year--and conversely, we would also expect a rising share of lowest-quintile families to bounce up to the second quintile in any given year. Rising income volatility, in short, could be a key to explaining the seemingly paradoxical behavior of lower-income households with respect to both spending and the accumulation of wealth.

This is, arguably, both good news and bad. On the one hand, it suggests that lower-income Americans are not spending themselves into oblivion. On the other, it implies that income volatility is a large and growing concern for ever more Americans at the short end of the income stick.

What It All Means

We would surely discard a statistical measure that showed life expectancy was falling during a time of ever-increasing longevity, or one that suggested our national finances were balanced in a period of rising budget deficits. Central as the OPR has become to antipoverty policies--or, more precisely, especially because of its central role in such policies--it should likewise be discarded in favor of a more accurate way (or ways) of describing trends in material deprivation.

Do not misinterpret this dismissal of the OPR. Nothing about the analysis here leads me to conclude that poverty is a thing of the past--or even that the general plight of the poor in America is markedly better today than in 1965 when Lyndon Johnson's "War on Poverty" was ramping up. To the contrary, I think that in many tragic respects, the misery and degradation suffered by America's most disadvantaged may well be more acute today than it was forty years ago.

For example, no matter how you measure it, family structure is far more frayed than it was in 1965. While the consequences of family breakdown are seldom auspicious, they tend to be most severe for the poor. By the same token, despite the past decade and a half of decline in major urban areas, crime rates in America remain far higher today than in 1965. And it is no secret that the greatest burden of crime falls directly on the very poorest. The corollary of that crime explosion--today's historically unprecedented levels of prison incarceration for the country's young men--not only reflects on misery in modern America, but contributes to it.

Nor does this study suggest that America's long war on poverty has been a failure. It does not even attempt an overall assessment of the material impact of those policies. At the risk of beating a dead horse, the only issue here is the reliability of the OPR as a measure of poverty. And, whereas the rate shows no progress in reducing poverty over the past three and a half decades, practically every other available statistical indicator points to major improvements in material living standards.

Accommodating such findings will require a fairly major recasting of the conventional narrative about long-term progress against poverty. Yet rethinking what has happened is hardly likely to end disputes over the value of the welfare reform plan of the 1990s or of the efficiency of the government's antipoverty programs. Our contentious, ongoing national debate about the adequacy and efficacy of the social safety net is at its core a dispute over first principles and underlying premises--not the simple facts of how many people go hungry or lack access to a telephone.

New Directions

This study points to some promising avenues for inquiry in the years ahead. Consumption is, without doubt, a more faithful measure of material deprivation than income. Further, the complex (and, for our purposes, crucial) interplay between consumption and income can be much better captured by surveys that track specific households or individuals over time than by "snapshot" measures at a single date. Yet the government's capacity to follow the long-term dynamics of household income and consumption in America is woefully limited--a curious oversight for an information-rich society.

Equally curious is the fact that the first efforts to create an alternative poverty metric are not coming from federal antipoverty agencies, but from New York City, whose mayor, Michael R. Bloomberg, decided that the OPR was all but useless in determining how to spend the city's limited antipoverty resources. So New York City created a new poverty measure. It does not tackle the inherently difficult problem of accounting for year-to-year variability in individual household resources, but it is plainly a step up because it focuses on consumption rather than income.

The first results, released in mid-July, are provocative. The new measure implies that 23 percent of New York's population is poor, as opposed to the OPR's 19 percent. But a much smaller proportion of the recorded poor is shown to be in extreme poverty because the new measure takes into account food stamps and housing subsidies. Conversely, the number of elderly poor is much higher (32 percent) than previously recognized, apparently because of increases in the cost of medical services not covered by government insurance. For now, the New York City metric offers only a snapshot of poverty--a picture of a single point in time--but it should be possible to use this framework to estimate long-term poverty trends.

On the same day New York released its first estimates, the House Ways and Means Subcommittee on Income Security and Family Support held hearings to explore the idea of modernizing the federal measurement of poverty. The subcommittee's chairman, Jim McDermott (D-Wash.), has submitted a bill calling for a new consumption-based metric for poverty derived from decade-old recommendations from the National Academy of Sciences. As with the New York City approach, it does not account for the role of year-to-year income variability in poverty. But it is a start: the process of rethinking the federal government's obsolete official poverty measure has begun.

Nicholas Eberstadt is the Henry Wendt Scholar in Political Economy at AEI.

 This document is available here as an Adobe Acrobat PDF.