Commentary | Wages, Incomes, and Wealth

Economic Opportunity and Poverty in America


Economic Opportunity and Poverty in America

By  Jared Bernstein

Chairman McDermott and members of this subcommittee, I thank you for the opportunity to testify.  I sincerely applaud your willingness to examine these issues of great importance to those on this panel and to our most economically vulnerable families.
It is a symbol of a just society that we engage in an accurate assessment of the extent of material need among our population.  Such an assessment serves multiple purposes. 

First, we want a measurement tool that will tell us, given what we know about human needs and prevailing living standards, how many people lack the resources to meet those needs.  Policy makers may and do have different ideas about what should be done about such deficits, but all would presumably like an accurate count.  Note that this framing of the concept introduces both an absolute (meeting basic needs) and a relative dimension (prevailing standards) to the question of poverty measurement.

Second, since it implies underinvestment in the economic well-being of adults and, in particular, children, poverty can cause long-term harm to our economy and society.   One recent estimate suggested that child poverty ultimately costs society half-a-trillion dollars in sacrificed productivity and ancillary costs each year (Holzer, 2007).1 It is thus very much in our national interest to measure poverty’s extent as accurately as we can.

Third, we want to be able to assess the anti-poverty effectiveness of market forces and non-market interventions.  When policy makers undertake initiatives to reduce the extent of economic deprivation, an accurate accounting of the effectiveness of such interventions is critical.  All of us, whether we’re members of this panel, taxpayers, voters, or the targets of these programs themselves, have a vested interest in their cost effectiveness.  Are they accomplishing their goals?  Are they doing so without creating unintended consequences that threaten to offset the gains?  Are we getting the best possible “bang for the buck?”

It is widely agreed upon that the current poverty measure fails to meet these criteria, and does so by a long shot.  It does not provide an accurate picture of the extent of material deprivation, it does not tell us how far the poor are falling behind relative to the rest of us, and it does not enable us to gauge the effectiveness of our antipoverty initiatives.

As a British analyst who reviewed a quarter-century of our poverty debate summarized, “The United States got itself the worst of all worlds—an increasingly mean measure of poverty that also suggested that U.S. social programs were not making a difference when they were” 2 (Glennerster, 2002).Key points in this testimony are:

• Our current approach to measuring poverty is far outdated and fails to provide an accurate count of the extent of need in America.
• Newer methods that correct many of the problems with the official measure show more people in poverty than the 37 million officially poor (12.6% of the population), including 13 million children.  These methods should be adopted to replace the current, official measure.

• The fact that the current measure is adjusted only for price changes and not for income growth, in tandem with rising income inequality, has led to large and growing gaps between the officially poor and the rest of society.  Even while today’s poor have some goods that were out of reach of the poor in decades past, in relative terms, today’s poor are increasingly left behind the mainstream.

• Efforts to gauge the true cost of meeting an adequate, basic living standard in today’s economy yield income thresholds that are about twice that of the official poverty lines.

• Relative to prior years, a significantly larger share of poor children are living in families with working parents.  The income constraints faced by these working parents underscore the need for increased work supports, including subsidies for wages, health care, child care, housing, and transportation.

Critique of the Official Poverty Measure

The shortcomings of our poverty measure have been amply documented and I will only briefly review these critiques (see Bernstein, 2001, for a thorough review3 ).

• The official thresholds were developed in the mid-1960s based on data from the mid-1950s.  Since then they have largely been adjusted only for price changes but not for improvements in general living standards.

The original poverty thresholds were derived by poverty analyst Mollie Orshansky, who based the measure on research on food consumption of low-income families in the mid-1950s.  Surveys from the mid-1950s also revealed that families spent about a third of their income on food, so she simply tripled the value of the “economy food plan” for a given family size.  

Amazingly, with very few changes, and with adjustments for inflation, the Orshansky measure remains the official poverty measure to this day.  Food consumption represents a much smaller share of family budgets than was the case 50 years ago (its average share has fallen by about half),4 while housing, transportation, and health care, for example, comprise larger shares.  Simply updating the official thresholds for this change alone would lead poverty thresholds (and poverty rates) to be much higher today.
One problem with the official approach is that as living standards rise for the rest of society, those deemed poor by an absolute threshold adjusted solely for price changes will fall behind the rest of us (this would not be the case with a relative measure, such as 50% of median income).  Back in 1960, the official poverty threshold for a family of four was about half the median income for a four-person family.  Today, at about $20,000 for a family of four with two children, it’s around 30% of the four-person median.

In an era with sharply growing income inequality, it is worth contemplating the importance of this development.  Why should we be concerned if our poverty thresholds drift further below the income of the median household?

The answer is that the concept of deprivation is not solely an absolute concept; it is a relative one as well.  Economists since Adam Smith have recognized that even if the poor are able to meet their fundamental needs for food and shelter in such a way to sustain their lives, they can, by dint of the economic and social distance between themselves and the rest of us, still experience deprivation that is harmful to society.

As Smith put it, over two-hundred years ago:

“By necessaries I understand not only the commodities which are indispensably necessary for the support of life, but what ever the customs of the country renders it indecent for creditable people, even the lowest order, to be without. A linen shirt, for example, is, strictly speaking, not a necessary of life. The Greeks and Romans lived, I suppose, very comfortably, though they had no linen. But in the present times, through the greater part of Europe, a creditable day-laborer would be ashamed to appear in public without a linen shirt, the want of which would be supposed to denote that disgraceful degree of poverty which, it is presumed, nobody can well fall into, without extreme bad conduct. Custom, in the same manner, has rendered leather shoes a necessary of life in Eng
land.” 5
To this day, some poverty analysts overlook this point, citing material gains made by today’s poor relative to those of the past.  Two such analysts, for example, writing in 1999 noted that “By the standards of 1971, many of today’s poor families might be considered members of the middle class.”6

Poverty analyst Doug Besharov notes that “…poor people’s physical and material well-being is considerably better now than in the late ’60s.  How else to explain why so many poor now have color TV (93%) and air conditioning (50%), and own their own homes (46%)?”7

Such comparisons are misleading.  They implicitly freeze the well-being of the poor at a point-in-time, ignoring progress in technology, consumption, relative prices, and opportunities.  In short, to ignore the relative economic distance between the poor and everyone else is to ensure that they will remain outside the mainstream.  Yes, they will not starve, many will be housed, and a large majority will watch TV in color.  But they will still be separate and unequal relative to the majority.

Interestingly, as Fisher points out (2005), subjective measures—responses from the public as to what it takes to make ends meet—clearly support a relative component to measuring poverty.  For each 1% increase in national income, these subjective measures grow by 0.6%-1%.8  Much as Adam Smith recognized hundreds of years ago, when thinking about what constitutes a fair poverty threshold, we instinctively add a strong relative component.  Implicitly, we want to prevent a growing gap between ourselves and the least well off among us.  Our official poverty measure, however, allows this gap to grow.

• The official measure ignores the value of some publicly-provided benefits that should be counted as income to their recipients.

As measured by the Census Bureau, under rules established by the Office of Management and Budget, the official income measure in our poverty accounts is pretax, post-cash transfer.  Thus, it includes the cash value of government transfers like welfare payments and Social Security, but omits, for example, the market value of food stamps or tax benefits like the Earned Income Tax Credit.

These are salient omissions.  By excluding such resources, we create two problems.  First, we underestimate the actual resources accruing to low-income families, and second, we prevent ourselves from observing the anti-poverty impact of these initiatives.

For these reasons, the omission of these benefits is widely agreed to be a significant problem with the current measure.  There is, however, some disagreement about how to value of economic resources.  For example, some analysts argue that we should also consider wealth and service flows from investments, such of the value of housing consumed by homeowners.  Another controversial area, one of some magnitude, is whether to include the value of publicly provided health care, and if so, how to calculate it.

• The official thresholds fail to account for necessary expenses associated with work and medical care.

The National Academy of Sciences (NAS), in their work seminal work on how we might improve our poverty measure, concluded that it made sense to subtract from income costs associated with work, largely child care and transportation.  This adjustment is particularly germane in an era when anti-poverty policy is predicated on work in the paid labor market.  Imagine, for example, a single parent who works full time, with earnings that lift her family above the poverty line.  Yet, once we net out her child-care expenditures, she falls below that line.  Such an example shows that the costs associated with climbing out of poverty can make the climb that much steeper.  Similarly, if her out-of-pocket medical expenditures pushed her back below poverty, we would want to account for that spending as well, subtracting it from income before comparing her income to the poverty threshold (this too was a NAS recommendation).

• The official measure makes no adjustments for geographical variation in the cost of living.

Though prices differ considerably by region, the official poverty measure makes no adjustments for the fact that the same level of income has greater buying power in one area of the country relative to another.  Part of this omission stems from the lack of official inter-area price deflators, though exciting progress is being made in this area (see Aten, 2006).  Aten finds, for example, that prices in New York City in 2003 were about the same as those in San Diego, but about 50% higher than those in St. Louis.

Political constraints are in play here too.  Adjusting for inter-area price differences, areas with relatively lower prices will find their poverty rates decrease compared to current measures, and this could lower anti-poverty benefits received by such areas (and vice-versa, of course, for areas with higher prices), as a range of federal programs allocate their benefits based on formulae that depend on calculating numbers of people in poverty.

There are numerous other concerns of a technical nature regarding the official measure.  Poverty analysts have found arguably better equivalence scales—adjustments for the needs of families of different sizes and composition—than those used in the current measure.  Also, and this one makes a big difference, many analysts argue that thresholds should be adjusted for prices using a different version of the consumer price index than the CPI-U, the current deflator used by Census.  The alternative deflator most often referenced in these discussion is the CPI-RS, which incorporates in an historically consistent manner (back to 1978) all of the advances made by BLS in measuring inflation.

In sum, I can firmly assert that a consensus exists among social scientists regarding the inadequacy of the current measure.  To the extent we depend on it, we unnecessarily limit our knowledge of the magnitude and composition of the poor population, the impact of our programs, and our ability to reach those truly in need.

An Improved Measure

An improved measure would correct these shortcomings.  What’s needed is a set of thresholds and an income measure that designates as poor those families whose members cannot adequately meet their basic needs, given what we know about human needs and prevailing living standards. 

It is critical in this measurement endeavor to avoid a piecemeal approach: our new measure must deal as comprehensively as possible with both a complete accounting of available resources on the income side, as well as expenses on the threshold side.  Some analysts, for example, add near-cash benefits like food stamps, or tax benefits, like the EITC, to income, and show how this reduces poverty.  This may be analytically useful way to isolate the impact of a particular program, but it is not an improved measure of poverty.  Adjustments to the income side of the equation must be matched by adjustments to the thresholds.

What would such a measure show?  Since any poverty measure invariably involves normative decisions, there are lots of different measures.  However, it is again widely agreed upon that the NAS recommendations deal successfully with many of the concerns raised above.  We also benefit from the fact that poverty analysts at the Census Bureau have operationalized the NAS recommendations, creating numerous variants based on the NAS suggestions.9

The NAS measures have these advantages over the official measure:

• The NAS thresholds are based on actual expenditures on food, clothing, and shelter and thus reflect increases in living standards (though not to the extent of family budgets, as discussed below).

• The NAS income measure is after-tax, and thu
s reflects the poverty reduction effects of tax credits.

• They include non-cash benefits in income (though they do not include the value of publicly provided health care).

• They deduct some work expenses, like child care expenditures for working families, from income and subtract out-of-pocket medical expenses, including premium payments.

• They factor in regional differences in cost-of-living.

As noted, there are many variants to these measures, and the Census Bureau has generated a consistent time series back to 1999 of 12 different NAS-based approaches.  For example, some measures account for geographical differences while others do not. 

A fundamental question for this committee to consider is, relative to the official measure, do these improved measures generate lower or higher poverty rates?  The answer, shown in Figure 1, is clear: the NAS measures are uniformly higher than the official measure.10 The Figure shows the range of the 12 measures, which is almost always above the official.  On average over the period covered by the graph, the NAS rates are about one percentage point above the official rate, implying about 2.5 million more persons on the poverty rolls.

Figure 1: Poverty rates, official compared to NAS alternatives, 1999 to 2004

These measures also tend to change the composition of who is poor.11 Under the NAS because work expenses are subtracted from income, working poverty rises, especially among single parents facing child care costs.  A similar treatment of out-of-pocket medical spending leads to higher poverty rates among the elderly.  Because more transfers are counted as family resources, African-American poverty rates are lower under the NAS measures, though still much higher than those of whites.

A Relative Measure

The NAS measure is a vast improvement in all the ways noted above, but it too is limited in the extent to which it captures relative differences between the poor and the rest of society.  Another way to measure poverty—one with great intuitive appeal—tracks the poor while accounting for changes in prevailing income levels among the non-poor.  Such measures are called “relative,” in that they set the poverty threshold as a percent of the median income, which moves each year, typically rising in nominal terms.  (The NAS measure has a relative component, as the thresholds are keyed to changes in median consumption expenditures).

The utility of this measure—and note that it is the norm in international comparisons—is that it shows how the poor or faring relative to middle income families, and thus speaks directly to the concept of “relative deprivation.” 

The 1990s are a good example of the importance of this approach to poverty measurement.  The tight job market, in tandem with a large expansion of the Earned Income Tax Credit, helped lead to significant reductions in a comprehensive poverty measure (i.e., one that includes such transfers as the EITC).  But because median family income also grew quickly over this period, much less relative than absolute progress was achieved. 

Table 1 compares relative poverty to absolute poverty, using adjusted income measures much like those recommended by the NAS (the absolute poverty measure here is from unpublished tabulations provided by Wendell Primus).  Absolute poverty fell fairly steeply in the 1990s, from 15.5% to 10% by this measure.  But relative poverty fell only slightly, from 18.4% to 17.7%.

Table: Relative and absolute poverty measures, 1989-2004

The table shows that the poor made a great deal of ground in absolute terms: over the 1990s, as low-incomes rose in real terms, more families made it over the threshold.  But the relative measure shows that low incomes grew at about the same rate as middle incomes, so the share of poor below half the median changed little over these years.

In other words, the relative measure tracks social/economic distance between the poor and the middle-class in a way that absolute measures do not.  As such, they quite directly reveal the impact of changes in inequality on poverty.  The share of the population that is poor in relative terms has hovered around 18% since the mid-1980s, showing that by this benchmark, many more persons are poor in relative terms—their income is less than half the median—than in absolute terms.  The fact that such a significant share of our population remains relatively distant from the mainstream is an important dimension of the poverty problem.

Family Budgets

Though the official Orshansky poverty measure has gotten by far the most attention in this debate, budget analysts have a long history of measuring the amount of income needed to meet a basic standard.  This work, under the rubric of family budgets, has generally been underutilized in the poverty debate, yet there is much we can learn from it about the income constraints facing American families today.

In this work, economists (along with nutritionists, health care experts, etc.) have set out to tally the amount of income needed to meet a basic living standard, one where a generally accepted set of material needs is met.  As Johnson et al noted, “most budget standards have been calculated by building up a budget that would provide families with a modest, fair, or sufficient income.”12 In our own work on basic needs budgets for working families, these needs included decent housing, an adequate diet, child care (when no parental caretaker is available), health care, transportation, and the money needed to pay taxes.13  

Obviously, criteria like “modest,” “fair,” and even “sufficient,” are normative judgments, although, as noted above, family budgets are often based on expert opinion, such as when nutritionists recommend an adequate diet.  But the committee should recognize that there is simply no “right” way to measure such concepts, including poverty.  When we engage in this exercise, we balance a variety of needs, sensibilities, and political, if not existential considerations.  We recognize that there is a distribution of well-being, and that it would be unreflective of realistic outcomes in a market economy to designate, say, everyone below the 80th, or even the 50th percentile of the income scale as “poor.”  Yet, it would be unjust in an affluent, highly productive economy to label only  those facing the most severe material deprivation as poor. 

Family budgets attempt to balance these extremes by recognizing that families who are unable to meet basic needs—and again, as Adam Smith pointed out, needs that derive in part from societal standards—face a material disadvantage that government should recognize and address.  In fact, such budgets continue to be used by the Department of Labor to set eligibility criteria for job training programs (Johnson et al, 2001). 

It is instructive that these budgets are well above poverty thresholds, usually in the range of two-times their value.  For example, Johnson et al report a family budget for a married couple with two children of $36,550 in 1998; Allegretto (2005) reports a family budget for the same family type of just under $40,000 for 2004.14   In both cases, these budget levels are about twice the official poverty threshold for that family type.  In fact, Allegretto’s work shows that while about 9% of the family types she examines are officially poor, about 30% are below the family
budget thresholds.

Does this finding imply that 30% are poor, or materially deprived in the sense that has been discussed in this testimony?  No, for a number of reasons.  First, the family budget standard is higher than the poverty standard.  For example, some of the family budget assumptions would likely be considered too generous for the poverty debate.  Much of this work uses HUD Fair Market Rents for housing costs, and these typically give the 40th percentile rent for currently available rentals.  Child care costs are often based on qualified center-based care; health care includes some measure of non-group premium costs.15 The distinction between these two standards—poverty and family budgets—recalls the views of poverty measurement pioneer Mollie Orshansky, who viewed her original poverty thresholds as a measure of income inadequacy, not of income adequacy.16    Family budgets are closer to the latter.

Second, this research tends to deal only with the threshold side of the question, and not with the resource side.  As such, it lacks the holistic quality of the NAS work.

But it does provide a common sense benchmark that has had some considerable impact on the poverty analysis community.  The logic of the family budget work is straightforward and commonsensical: if we take an objective look at what things cost, it takes an income well above the poverty threshold to make ends meet.  This, and the fact that family budgets often correspond to roughly twice poverty have led many analysts to use twice-poverty as a benchmark. 

Moreover, there are important public programs that recognize this, including SCHIP, the public health insurance program for children.  The vast majority of children living in families with incomes below twice poverty are eligible for the program.  A moment’s reflection suggests that this is a stark repudiation of the official poverty threshold.  Our government itself, to our credit, obviously recognizes the inadequacy of the official measure as a criterion for setting eligibility for families in need.

Working Poverty

In moving towards a more accurate approach to measuring poverty, the committee also needs to consider the increasingly important role of work among the poor and near poor.  Though the share of the poor in the job market has not changed much over time, its composition has changed a great deal, with many more parents, especially single parents at work.  It is also the case that low-income persons (family income below twice poverty) are working more hours than in the past, and, most importantly, a much larger share of their income derives from the labor market, including wage subsidies.  Because more low-income parents (especially mothers) are working now than in the past, the share of children in working but poor families has climbed significantly.

Most poor families have at least one worker: 61% of poor families had at least one worker in 2005 (4.6 million families); 71% of twice-poor families have at least one worker (14.4 million).17  Figure 2  shows that income from work (earnings plus the EITC) for low-income single mothers with at least two children rose from 45% of income in 1979 to 72% in 2000 (comparison made at business cycle peaks; Mishel et al, 2006, Table 6.11).  This increase is a function of the steep growth in both the share of single mothers at work in the paid labor market, and in their annual hours worked.

Figure 2: Earnings, including EITC, as a share of income, low-income single-mother families

As noted, these trends have meant that more low-income children have working parents.  This is evident in Figure 3 which shows the percent of poor children in families with a working parent, 1979-2005.18 This share shot up in the 1990s, particularly among poor families with a single mother.  Between the economic peaks of 1989 and 2000, the share of poor children in homes with a working parent increased from 59 to 71 percent; for kids of single moms, the increase was from 42 to 61 percent, i.e., from a minority to a solid majority.

Note also the steady decline in the 2000s, driven again by employment trends of single mothers (married mothers’ employment rates also fell over this period).  The long jobless recovery and the weaker labor demand over the current recovery has been particularly damaging to these economically vulnerable families.  As their labor market opportunities have diminished, their family poverty rates have gone up 3.2 points, 2000-05, compared to 1.2 points for the overall family poverty rate.  The problem for them relates closely to the observation that earnings, and benefits tied to earnings, have become much more important to the economic well-being of single-mother families.

Figure 3: Share of poor children with at least one working parent

In sum, recent experience has shown that these low-income working families depend on two forces to ensure that their living standards are rising.  First, relative to more economically secure populations, they depend on tight labor markets and strong labor demand, compelling employers to provide the jobs and wage advancement they need. 

The second point links back to the measurement themes explored in this testimony.  Even in the best of times, gaps will exist between what any working poor families can earn in the low-wage labor market and what they need to meet their basic needs, as discussed above.  This implies an important role for work supports. 

Work supports are any publicly-provided resource that either boosts the earnings of low-income workers (like the EITC), or helps offset the cost of a family budget component, including subsidies for health care, child care, housing, and transportation.  These supports play an important role in helping to close the needs gap, but their provision is not always guaranteed, and in the face of budget constraints, many states have cut back.  At the federal level, the most recent budget offered by the president includes significant cuts in access to SCHIP—the health coverage program for low-income children and an important work support for low-income workers whose jobs often fail to provide family coverage.

Given evidence provided in this testimony regarding the extent of low-income work, and the material needs of these families, strengthening the nation’s system of work supports would be a highly useful anti-poverty strategy.


Ours is a nation with one of the strongest, most productive economies in the world.  Yet considerable poverty exists amid the plenty.  By the official measure, one that most consider inadequate to the task of accurately measure material need, 12.6% of our population, 37 million persons, are poor.  As I have argued, a more accurate measure would show a greater share of persons in need.  While one should not be dismissive of the political constraints pushing back against changing the official measure, its time has passed, and I urge the committee to begin taking steps to replace it with a better alternative.

As I have shown, such alternatives have been developed by a team of researchers at both the Census Bureau and BLS, implementing the seminal work of the National Academy of Sciences.  The advances in poverty measurement made by these analysts have the potential to vastly improve our knowledge and understanding of who is poor.

Of course, measurement is a means to an end, and this committee has shown great interest is taking steps to address poverty amid plenty.  Given
the sharp rise in the number of children in working poor families, I have stressed the importance of ensuring that these families have enough to not simply pass the poverty threshold, but to meet their basic needs, as shown in the family budget literature.  To this end, work supports, including wage subsidies along with subsidies for other basic needs such as housing, health and child care, and transportation, have proved vital in closing the gap between what low-income workers earn and what they need. 

I urge the committee to examine and strengthen this system.  I urge members to fight back when components of the system are attacked, as with the inadequate funding of SCHIP in the president’s most recent budget proposal, a change that could lead to lost health care coverage for over 600,000 children.19

By updating our measurement tools and strengthening our system of supports for working but poor families, we can make important progress toward reconnecting the economic lives of the most vulnerable among us to that of the mainstream, a laudable goal indeed.

The author thanks Ross Eisenbrey, Danielle Gao, Mark Greenberg, and James Lin for helpful comments and research support (Lin).  Any mistakes are my own.


Allegretto, Sylvia. “Basic Family Budgets: Working families’ incomes often fail to meet living expenses around the U.S.” EPI Briefing Paper. 1 Sept. 2005. < http://www.epi. org/content.cfm/bp165>.

Bernstein, Jared. “Let the War on the Poverty Line Commence.” The Foundation for Child Development Working Paper Series. June 2001.

Bernstein, Jared and Arloc Sherman.  “Poor measurement: New Census Report on Measuring Poverty Raises Concerns,” Economic Policy Institute, 2006.

Besharov, Doug. “Poor America.” Wall Street Journal. 24, March 2006.

Bureau of Labor Statistics. “At Issue: Tracking Changes in Consumers’ Spending Habits.” Monthly Labor Review. September 1999.

Cassidy, John. “Relatively Deprived: How Poor is Poor?” The New Yorker. 3 April 2006. <>.

Fisher, Gordon M. “Relative or Absolute: New Light on the Behavior of Poverty Lines Over Time.” Newsletter of the Government Statistics Section and the Social Statistics Section of the American Statistical Association. Summer 1996: pp. 10-12.

Fisher, Gordon M. “The Development and History of the U.S. Poverty Thresholds: A Brief Overview.” Newsletter of the Government Statistics Section and the Social Statistics Section of the American Statistical Association. Winter 1997: pp. 6-7. < hptgssiv.htm>.

Glennerster, Howard. “United States Poverty Studies and Poverty Measurement: The Past Twenty-Five Years.” Social Science Review. March 2002.

Holzer, Henry. “The Economic Costs of Child Poverty.” Testimony before the U.S. House Committee on Ways and Means. U.S. House of Representatives. Washington, D.C. 24 January 2007. <>.

Johnson, David S., John M. Rogers, and Lucilla Tan. “A Century of Family Budgets in the United States.” Monthly Labor Review, May 2001.

Mishel, Lawrence, Jared Bernstein, and Sylvia Allegretto. The State of Working America. Ithaca: Cornell University Press, 2007.

Smith, Adam. Wealth of Nations. [1776] New York: Prometheus Books, 1991.

Weinstein, Deborah. “What’s in the President’s Budget for Human Needs?” Coalition on Human Needs. 8 Feb 2007. < frame.htm>.


[1] Henry Holzer, “The Economic Costs of Child Poverty,” Testimony before the U.S. House Committee on Ways and Means. U.S. House of Representatives, Washington, D.C. 24 January 2007, Accessed 8 Feb 2007, <>.
[2] Howard Glennerster, “United States Poverty Studies and Poverty Measurement: The Past Twenty-Five Years,” Social Science Review, March 2002.
[3] Jared Bernstein, “Let the War on the Poverty Line Commence,” The Foundation for Child Development Working Paper Series, June 2001.
[4] Bureau of Labor Statistics, “At Issue: Tracking Changes in Consumers’ Spending Habits,” Monthly Labor Review, September 1999.
[5] Adam Smith, 1776, “Wealth of Nations.”  See,, for context.  John Cassidy provides this quote in his New Yorker article, Relative Deprivation, 4/3/06 (
[6] W. Michael Cox and Richard Alm, as quoted by Cassidy, ibid.
[7] Doug Besharov, “Poor America,” Wall Street Journal, 24, March 2006.
[8] Gordon M. Fisher, “Relative or Absolute: New Light on the Behavior of Poverty Lines Over Time,” Newsletter of the Government Statistics Section and the Social Statistics Section of the American Statistical Association, Summer 1996: pp. 10-12.
[9] Some of this material was in Mishel, Bernstein, and Allegretto, 2006.  For a good discussion of the ongoing work at Census and BLS on these issues, see < conferen.html>.
[10] This figure also appears in Bernstein and Sherman, 2006.
[11] To compare the composition shift, rates must be standardized, as in this table: < hhes/www/povmeas/exppov/suexppov.html>.
[12] David S. Johnson et al, “A Century of Family Budgets in the United States,” Monthly Labor Review, May 2001.
[13] Bernstein et al, 1999.
[14] Sylvia Allegretto, “Basic Family Budgets: Working families’ incomes often fail to meet living expenses around the U.S.,” EPI Briefing Paper, 1 Sept. 2005. Accessed 8 Feb 2007, < cfm/bp165>.
[15] Ibid.
[16] Fisher, Gordon M. “The Development and History of the U.S. Poverty Thresholds: A Brief Overview.” Newsletter of the Government Statistics Section and the Social Statistics Section of the American Statistical Association. Winter 1997: pp. 6-7. Accessed 8 Feb 2007, < hptgssiv.htm>.
[17] Please refer to BLS tables, “Families by Number of Working Family Members and Family Structure:  2005,” <>.
[18] Though this is my analysis of CPS data, the figure is based on work done by the Congressional Research Service.
[19] Deborah Weinstein, “What’s in the President’s Budget for Human Needs?” Coalition on Human Needs, 8 Feb 2007. Accessed 8 Feb 2007, <>.


Jared Bernstein is an economist at the Economic Policy Institute in Washington, D.C.



See related work on Income and wages | Wages, Incomes, and Wealth

See more work by Jared Bernstein