Report | Wages, Incomes, and Wealth

Any way you cut it: Income inequality on the rise regardless of how it’s measured

Download PDF

The Economic Policy Institute and the Center for Budget and Policy Priorities recently published a joint report called Pulling Apart: A State-by-State Analysis of Income Trends (Bernstein et al. 2000), which showed that income inequality-the gap between those at the top, middle, and bottom of the income scale-has grown significantly throughout the past two decades and remains higher than at any other time in the post-war era.

This report was by no means the first to make these observations. Since inequality began climbing sharply in the 1980s, it has become the focus of much research in economics.[1] Nevertheless, the report, which also examined the growth of inequality at the state level, reminded those interested in this issue that inequality continued to grow, albeit more slowly, in the much-praised “new economy” of the 1990s. While the methods and findings of the report were widely accepted, some criticisms were raised that need to be addressed. Some of the report’s critics simply missed the central point of the study – that income inequality has increased significantly over the past two decades. The weakest of these critiques, such as the one by Alan Reynolds from the Hudson Institute, ignored a key issue-that the analysis was measuring a trend-and simply argued that there are different ways to measure inequality at a certain point in time (Reynolds 2000). For example, Reynolds’ article points out that families with more workers earn more than those with fewer, and that those with more education and experience earn more than those with less. Similarly, an analysis by the Heritage Foundation shows that different income definitions, such as pre- versus post-tax, lead to different levels of inequality at a point in time (Rector and Hederman 1999). Again, these types of analyses reveal nothing about the key question of whether inequality has increased over time. As we show below, no matter which income definition is applied, the finding of increased inequality stands. A second group of critics accepts that inequality has risen over time but argues that its increase has been offset by income mobility in the U.S. economy. An assertion in this argument holds that, while the income gap may be wider now than in the past, the poor are better off than ever before. Neither claim withstands close scrutiny. A third and final category of critiques more or less accepts the findings but suggests that increased income inequality is simply unavoidable. These critics claim that increased inequality is simply the inevitable outcome of a growing economy in which globalization and technology lead to higher rewards for the winners and bigger losses for the rest. We address this interpretation in the conclusion.

Do different income definitions yield different answers?

Studies in income distribution, as in all areas of empirical work, have to be careful and complete in how the items being analyzed are defined. More specifically, it is important to ensure that any conclusions drawn from a data analysis are not sensitive to particular measurement choices. In both the recent Pulling Apart and in The State of Working America 2000-01 (Mishel et al. forthcoming), we have employed the same definition of income used by the Census Bureau when it analyzes the primary source on income distribution, the income supplement to the March Current Population Survey. The Census Bureau itself produces measures of poverty and income distribution using a variety of income definitions so that analysts can assess the sensitivity of measured trends to changes in definitions. We use these Census Bureau data as well as Congressional Budget Office income distribution data to assess whether any of our conclusions are determined by our decision to use the official Census Bureau measure of income. Before elaborating on why the measurement choices did not affect the conclusions we reached, it is important to be clear about the questions we asked in our research. As discussed above, our research has measured the changes in income for the range of income groups (low, middle, and high income) as well as the change in income inequality. Our conclusions, therefore, were about trends in income and income inequality. So the measurement question is whether differing definitions of income yield different trends in income growth and inequality. We show below that they do not. We agree (and document) that analyses using different income measures will result in different levels of inequality at a specific point in time (e.g., examining a single year). What is important to note, though, is that studies using comparable measures of income inequality demonstrate that income inequality in the U.S. is now historically high, and high relative to other advanced countries. [2]

A review of a wide array of income measures confirms that:

  • Income inequality grew quickly in the 1980s and continued to increase at roughly a third to half as fast in the 1990s.
  • Income growth at the bottom of the income distribution was modest in the 1990s and, for some income definitions, even fell. Middle-class family income growth in the 1990s, as a whole, was also modest, notwithstanding recent rapid growth. Despite recent improvements, a longer-term view of the economy reveals that income inequality has grown substantially from 1979 to the present. This increase has been accompanied by only modest income growth for most families, especially relative to the overall productivity gains over these 20 years.

Census Bureau Alternative Income Definitions

We first turn to an analysis of the various household income measures produced by the Census Bureau. We have selected four of the 15 available measures for analysis so that we can identify how results change as incomes become more comprehensively measured and as changes in the tax system are integrated (see box below). The first measure is the official Census Bureau money income definition, which includes all labor and self-employment income, government cash assistance, pensions, interest, dividends, and other money income. The limitation of this definition is what it excludes: any effect of the tax system, employer-provided health benefits, realized capital gains (i.e., gains from selling assets such as stocks), and the value of government non-cash programs (housing subsidies, Medicaid/Medicare, food stamps, etc.). The second definition we have selected is pre-tax market income, obtained by subtracting government cash assistance from money income and adding in realized capital gains and employer-provided health insurance. The third definition measures after-tax market incomes. The final measure, comprehensive income, is the broadest and adds the value of government cash transfers, health programs (Medicare/Medicaid), food stamps, and school lunch programs. [3] All of the published Census data, regardless of definition, suffer from the problem of “top-coding, which results from the fact that income above a certain level (affecting about 1% of families) is not reported by the Census Bureau because of confidentiality concerns. Consequently, the income of the top income group is under-reported and reported to a different degree in different years. In Pulling Apart and in The State of Wo
rking America
we present data that adjust for top-coding so as to have a consistent income series.

Alternative income definitions

Money income: The Census Bureau’s official definition of income used to compute income and poverty trends. This definition combines all labor income (wage and salary and self-employment), all government cash transfers (unemployment insurance, Temporary Assistance for Needy Families, Social Security), pensions, alimony, rent, interest, dividends, and other money income. This definition does not take account of non-cash government assistance (e.g., Medicaid), taxation, and capital gains.

Market income, pretax: This definition adjusts money income by subtracting government cash assistance and by adding market incomes excluded from the official definition: employer-provided health insurance and realized capital gains (gains from selling assets such as stock). Thus, this definition includes only income generated by the market.

Market income, aftertax: This definition adjusts market income to an “after-tax” basis by subtracting estimates of federal income and payroll taxes, the EITC, and state income taxes. There is no adjustment for other federal (corporate, excise), other state (sales) taxes, or any local taxes.

Comprehensive income: This definition adds the value of government assistance to income. It includes both cash assistance (Social Security, unemployment insurance, etc.) and the value of various subsidies and programs, such as housing subsidies, food stamps, school lunch programs, and health programs (Medicare/Medicaid). This definition is thus the most comprehensive in including both market income and government assistance as well as adjusting for most taxes.

There are reasons to favor or reject particular income definitions. However, the important point is that, although differing definitions yield a wide variation in the amount of inequality measured in a given year, all the definitions show a similar income inequality trend over the 1980s and the 1990s. Table 1 presents several measures of inequality for each definition of income for the years 1979, 1989, and 1998, and the percent growth over time. The top panel employs the Gini coefficient, a widely used measure of income inequality. There is a persistent and comparable growth of inequality since 1979 for every income definition, with the growth of income inequality ranging from an increase of 30% to 40% in the 1990s in comparison to the 1980s. For instance, over the 1979-98 period the Gini coefficient grew 10.7% for the first two income definitions and 12.8% for the last two income definitions. Plus, the Gini coefficient grew about 4% in the 1990s. This is incontrovertible evidence that, regardless of definition, income inequality has grown in each of the last two decades.

Table 1

The conclusions in Pulling Apart regarding income inequality trends were based on the trends in two income ratios: the incomes of the highest fifth in relation to the lowest fifth of families and the incomes of the highest fifth to the middle fifth. We also use the various Census Bureau definitions of income to compare the sensitivity of the original findings. As the bottom half (family income ratios) of Table 1 shows, income inequality both between the top and the bottom and between the top and the middle grew persistently over the 1980s and 1990s, regardless of how income is measured. In fact, each income inequality ratio grew roughly the same amount-around 25% on average over the 1979-98 period for nearly every income definition examined in Table 1. The inequality of market-based incomes grew the fastest in the 1990s for both income ratios. Contrary to some analysts’ contentions, the income gap between the top and the bottom grew as quickly with the official Census measure as it did with an after-tax income measure, at least in the 1990s (11.7% versus 12.7%). Clearly there is no basis to the claim that the Census Bureau’s money income measurement somehow distorts the trend in income inequality. The alternative Census Bureau income definitions also can be used to examine the income growth of groups along the income scale, as presented in Table 2. As the table shows, the income growth over the 1989-98 period is roughly 2% for the lowest fifth, regardless of income definition. Growth over the 1980s varied more widely, particularly for money income, which grew less than market income for the bottom fifth (reflecting the reduction in government cash assistance). The key comparison-between Census money income and comprehensive income-shows that, over the entire 1979-98 period, the difference between the two measures for households in the the lowest fifth was a growth rate of 4.9% (money income) and 9.8% (comprehensive income). Note, however, that these differences are spread over a 19-year period, yielding a difference in annual growth rates of only 0.25% and 0.49%. Needless to say, either rate represents very modest income growth.

Table 2

With respect to inequality, the difference in growth rates between the top and the bottom using these two measures was almost identical. For Census money income, the top fifth grew 33.8%, or 28.9% faster than the bottom fifth over the full period; when using the comprehensive income definition, the difference between the lowest and highest fifths was nearly identical at 28.4%. Growth in the middle income group does not vary much across definitions for the 1980s but does vary by roughly seven percentage points in the 1990s (a 0.6% decline for Census money income versus a 6.7% gain for comprehensive income). Nonetheless, middle-class incomes relative to the top or bottom grew comparably across definitions (since each group shows somewhat more growth when comprehensive income measurements are used).

Congressional Budget Office (CBO) data

The trends in the CBO’s income distribution data (CBO 1998) provide another opportunity to check how conclusions are affected by alternate income definitions. Although the CBO’s data are based on the Census Bureau’s Current Population Survey (CPS), the CBO excludes those with negative income from the lowest fifth, corrects for top-coding problems, adds realized capital gains, and arranges the data so that 20% of the persons are in each fifth (rather than 20% of the households). The particular strengths of the CBO data are that they allow an examination of the impact of the federal tax system, including all federal taxes (the Census Bureau estimates only income and payroll taxes and the earned income tax credit), and within the upper fifth, the CBO includes the income trends for the upper l%, upper 5%, and upper 10%. [4] Table 3 presents the CBO data on the growth of pre-tax and after-tax household income. The latest year of data is the CBO’s projection for 1999. These data (pre-tax household income at the top of Table 3) reveal that the groups at the top of the income scale (the highest fifth) have done far better than any income group in the bottom 80%. This was especially true in the 1977-89 period when the real income of the top 1% grew 63.2%, while the income of the bottom 60% declined, and the income of the upper-middle fifth grew only 1%. In the 1989-99 period, the income growth at the very top was considerably slower (13.2%) but still stronger than the mostly flat or falling incomes of the bottom 60%.

Table 3

For the entire 1977-99 period, the CBO data show the income growth for the top 1% was 84.8% and 44.6% for the top 10%, while the upper middle (the “fourth fifth”) had 5.2% growth, equivalent to a minimal 0.25% annual income gain. The entire bottom 60% of the income scale lost ground over the last 20 years, with the income of the poorest fifth falling 12.5% as income declined steadily in both decades. With these disparities in income growth, it is not surprising that the income gaps between the top and every other group have risen in both the 1980s and the 1990s, according to the CBO. Whereas the ratio of the incomes of the highest fifth to the lowest fifth was 10.2 in 1977, it rose to 13.1 in 1989 and to 15.7 in 1999. The gap between the highest 5% and the bottom fifth grew even more, from 18.6 in 1979 to 32.9 in 1999. The gap between the top and the middle also grew substantially, with the top earning twice as much as the middle in 1977 but two-and-a-half times as much in 1999. Do these patterns hold up if we take into account changes in the federal tax system? We can look directly at this by examining the data at the bottom of Table 3, which shows the changes in the effective federal tax rates for each income group. Over the 1977-89 period, federal tax rates grew modestly for the bottom 40%, fell slightly for the next 40% (the middle and fourth fifths), and were reduced significantly for the upper fifth, especially the top 1% (whose effective tax rates were cut by a quarter, from 37.3% to 28.1%). These changes to the tax structure exacerbated, if only slightly, the income declines of the bottom 60%. But these tax revisions substantially raised the income growth of the upper 1%, shifting the 63.2% pre-tax income growth rate to 87.2% after taxes. As a result, the top/bottom income ratio grew 28% on a pre-tax basis but 33% on an after-tax basis. Thus, tax changes in the 1980s, by reinforcing the inequality generated by the pre-tax market distribution, led to an even greater increase in inequality. Changes in taxation in the 1990s, on the other hand, were progressive, as taxes were lowered on the bottom 40% and raised for the highest fifth, especially the upper 1% and 5%, with other income groups seeing only slight shifts in taxes. In the 1990s, therefore, the income inequality generated by the market, as seen in pre-tax income trends, was partially offset by the increased progressivity of taxes. Note, for example, that in the 1990s, the three income ratios all grew about half as fast for after-tax income as they did for pre-tax. Over the entire 1977-99 period, tax rates were reduced for those both at the very top of the income scale (by 2.9% for the top 1%) and for those in the bottom 40%. These changes served to offset each other, and, as shown by the similar percent changes in the income ratios in the 1977-99 period, overall inequality growth over the last 20 years was much the same whether measured on a pre-tax or after-tax basis. The reduction of market inequality by progressive tax shifts in the 1990s has meant that the federal government has had to increase its redistributive efforts in order to offset market forces. If one wants to downplay the growing inequality of pre-tax incomes in the 1990s because after-tax income inequality grew less, then one must be supportive of increased tax progressivity. However, many of those who take comfort in after-tax trends actually opposed the increased tax progressivity. Moreover, if market forces continue to widen pre-tax income inequality, then even further increases in tax progressivity are required to forestall a growth in after-tax income inequality. For instance, a larger, broader Earned Income Tax Credit (EITC) would be needed. Needless to say, such an agenda is not prominent among the critics who downplay growing inequality. Table 4 presents an additional tabulation of income trends from the CBO that adjusts family income for family size. [5] This adjustment is made so as to measure income relative to “needs,” on the basis that smaller families need less income than larger families, and visa versa. Some analysts have suggested that the trend toward smaller families means that income declines are less onerous. This line of reasoning can be specious, though, since some households may not increase their size (by marrying or having children) precisely because of disappointing income growth.

Table 4

Nevertheless, it is worth assessing whether adjusting family income trends for changes in family size affects our conclusions regarding trends in incomes and income inequality. What we find is that family-size adjustment does not change any of the results. Using this measure, the incomes of the bottom 40% have been flat or falling over the last 20 years, while incomes for the upper 1% grew 78.9% on a pre-tax and 89.4% on an after-tax basis. Income inequality, when measured with family-size-adjusted data, increased sharply in the 1980s and continued to grow (albeit at a slower rate) in the 1990s. The ratio of the size-adjusted incomes of the upper 5% to those in the bottom fifth grew about 70%, both before and after taxes, over the last two decades. Whether or not we adjust by family size, examine pre-tax or post-tax income, or add in the value of various other income sources-including capital gains and food stamps-there is nothing in either the CBO or Census Bureau data that would alter any of the findings in the Pulling Apart report or The State of Working America 2000-01. The CBO data, in fact, shows a larger growth of inequality and a broader and more severe decline in income at the bottom than that shown by our analysis of CPS data in these other studies.

Does income mobility counteract the inequality problem?

Other critics accept the fact that inequality has grown over time but argue that reports like Pulling Apart, which take snapshots of the income distribution at different points in time, miss the extent to which families move up and down that distribution over the course of their lives. Essentially, this critique agrees that the distance from the basement to the penthouse has indeed grown further over time. But these critics argue that a family that starts out in the basement has a better chance these days of making it to the top floor than they used to. In other words, they implicitly argue that an increase in income mobility has served to offset the increase in income inequality. In fact, the “mobilitists” fail to either articulate or substantiate this part of their argument. Instead, they simply show evidence of economic mobility, and leave it at that, as if mobility in and of itself should lessen our concern over increased inequality. But unless the rate of mobility is increasing relative to that of earlier decades, families are no more likely today to span the now wider income gap. Unfortunately, we show below, there has been no such increase in mobility. The economist Joseph Schumpeter came up with a useful analogy to explain the concept of income mobility. He suggested that the income scale could be thought of as a hotel where the quality of rooms improves as you move up to higher floors. If everyone simply ended up in the same room they started out in, society would be totally immobile, with the poor stuck in the basement and the rich ensconced in the penthouse. The reality, of course, is that some stay where they start while others move up and down. How does this analogy help us to understand the interplay between increased cross-sectional inequality and income mobility? The fact that, as we show in our report, the income gap between those at the top, middle, and bottom of the income scale has expanded over time means that the quality of life is worse now for a resident of the basement relative to his neighbor in the penthouse than it was two decades ago. The mobilitists acknowledge this, but argue that this family won’t always be in the basement. This is true, but unless their chance of making it to the higher floors has increased over time, the increase in cross-sectional inequality means that they are certain to experience more inequality over the course of their lives. The wider income gap means that the higher floors are “further away,” and the chance of reaching them has not increased. Cox and Alm are the most vocal advocates of the mobility argument. In a series of articles, they claim that Americans’ “remarkable ability to propel themselves upward” obviate concerns about the “red herring” of increased inequality (Cox and Alm 2000). But as we show below, they fail to address the crucial question of whether the rate of mobility is increasing, and thus their findings cannot be taken to imply that mobility can offset increasing inequality. As mobility expert Peter Gottschalk (1996) has noted, Cox and Alm “ask the wrong question and give a misleading answer to the question they ask.” Instead of examining whether the rate of mobility has increased, Cox and Alm use longitudinal data to follow a group of families over one 17-year period, and show that many of their members do indeed move up the income scale. This is not a surprising finding. In fact, given the way they set up the study, it was the inevitable result of the fact that one would expect most families to see their incomes increase as they age. A significant flaw in their study is the use of individual, as opposed to family income. Thus, a teenager from a rich family, but with a minimum wage job, will be classified in their study as starting out in the bottom quintile. As a result, the average income (in 1997 dollars) of those in the bottom quintile in their study was $1,263; clearly, such a person is very unlikely to stay in the lowest fifth. All other income studies of which we are aware use family or household income. [6] The next mistake in the study was to follow a sample that is getting older and compare the income changes of this sample to that of the entire population, an analytical setup that virtually ensures the finding of substantial mobility. [7] The only way these authors could have gotten a different result would be if most individuals in their sample actually lost ground as they got older, a very unlikely scenario. Due to this problem, we learn little about even the extent of mobility from their study. But it should again be stressed that the extent of mobility is not the key issue when trying to understand the impact of mobility on inequality. Even if their research had approached the question correctly, the fact of mobility should not assuage concerns over increased inequality, unless it can be shown that workers do indeed have more economic mobility than before. The better way to measure mobility is to control for age, by comparing the income of the sample only to itself as it ages, not to the larger population. Table 5, based on the same data set used by Cox and Alm, applies this method. [8] When done this way, mobility occurs because one family’s income grows faster than that of another in the same age cohort. This is by no means inevitable and is therefore more interesting than the rather obvious Cox and Alm finding that families do better as they age. Table 5

In fact, the bottom of Table 5 shows that the rate of mobility over the 1980s, when inequality was increasing most quickly, was no faster than in the 1970s. In both decades, about 61% of those who started in the bottom fifth were still there 10 years later. Another 24% made it to the second fifth. Thus, in each decade, about 85% of those who began the period in the basement either stayed there or moved up one floor. Of those who started out at the other end of the income scale, about 80% in both periods (slightly more in the latter period) either remained in the top fifth or moved down to the next highest fifth. The table provides no evidence of an increase in the rate of mobility. [9]

Are low-income families better off than before?

Some critics argue that a family’s well-being should not be judged by their income, but by what they are able to buy. These critics argue that if we look at what today’s poor are able to consume, they are much better off than even the middle class of previous decades. There are two fundamental flaws to this argument. First, these critics focus on what the poor are able to afford, not on what goods and services the poor can’t afford but the more affluent can. Just as incomes have grown more unequal, inequality has also grown in regards to consumption. This is what we term “relative hardship,” or the gap between the standard of living of the poor and non-poor. Second, these kinds of arguments tend to center on the fact that today’s poor have access to more non-essential consumer goods, like TVs and VCRs, than the poor of previous generations. At the same time, these critics ignore the cost of basic necessities, such as housing and health care, that are less affordable today than in the past. Thus, the argument ignores what we call “absolute hardship,” or the difficulties low-income families have in meeting their basic needs, regardless of the standard of living enjoyed by the non-poor.

Relative hardship

The fact that the poor can now afford some things formerly beyond their means does not negate that there is a large consumption gap. In terms of inequality, what matters is the relative distance between income classes. The disparities between how much of their income the poor and non-poor spend on basic necessities today is evidence of this inequality. In 1992, households living below the poverty line spent 71% of their expenditures on food, clothing, and shelter, while non-poor households spent 46% on these items (Federman et al. 1996). Thus, the poor have a smaller share of their income to spend on other necessities like child care, transportation, and health care. They also have less to spend on things that will help them get ahead, like education, and are less likely to be able to save for future needs or make investments that would raise their future earnings capacity. Furthermore, when comparing poverty rates over time, the definition of “poor” must relate to the overall growth of living standards. If not, virtually all people in the U.S. today would be considered non-poor because they have indoor plumbing, electricity, and modes of transportation not available 100 years ago. Because the current official poverty line has not been updated to reflect improvements in our standard of living that have occurred since it was developed, it is still set relative to income in the mid 1950s. Yet output per hour has more than doubled since that time. Comparing consumption patterns over a 40-year (or even longer) period using a fixed yardstick of this sort is obviously inappropriate.

Absolute hardship

Some critics argue that the poor’s access to consumer goods is evidence that families classified as poor based on income are not facing real hardship. They point to data showing that high percentages of families living below the poverty line now own consumer durables that previously only the middle class could afford. Rector (1998), for example, notes that the number of poor who have access to TVs, VCRs, washers, dryers, and automobiles has increased substantially since 1984. The poor are able to afford goods like TVs and VCRs because these goods are now less expensive relative to other goods, like housing. As an economy grows, the prices of some goods fall relative to others-items that were once affordable only to the elite become commonplace. The fact that everyone, including the poor, can buy more of selected goods and services is not a very telling observation. In fact, it is one that could be made at any time in our economic history, since as the economy grows, certain goods and services become affordable to all. In an important way, low-income families’ access to TVs is just a reflection of the relatively faster growth of productivity in manufacturing that leads the prices of goods to fall relative to services over time, a phenomenon known in economics as the “Baumol effect.” Ironically, critics who dismiss income inequality by pointing to upward income mobility ignore downward mobility. The existence of downward mobility is another reason to question whether the durable goods ownership (TVs, VCRs, etc.) of low-income families is an appropriate indicator of their living standards. Some families observed in the bottom fifth were, in fact, previously middle-class families who have suffered a change in family circumstances (death of an adult, divorce, or separation), a layoff, or a health emergency that has temporarily (at least for a few years) lowered their income. [10] Yet these families still retain the house (and other durable goods) obtained while enjoying a middle-class income. In this light, durable goods ownership at the bottom is partially a reflection of the temporary drops in income that our social safety nets (unemployment insurance, disability insurance) are meant to offset. It seems strange to argue that safety net programs are not necessary anymore because low-income families may have TVs and VCRs. [11] The critics also argue that the poor are now better off because the share spent on basic necessities-which they define as food, clothing, and housing-has fallen since the 1930s. As Figure 1 shows, the relative shares of income spent on these three items combined has decreased since 1935. But who honestly doubted that the poor today are better off than they were 65 years ago in the middle of the Great Depression? The relevant time period for comparison is the last 20 years.

Figure 1

Figure 1 shows that the share of household consumption spent on “necessities” has remained relatively constant over the past 20 years. Looking at the period from 1972-73 to 1996-97 shows that the consumption shares did fall for apparel (from 8% to 5%) and food (from 21% to 16%). On the other hand, the consumption shares rose for housing (from 31% to 37%). Health care and transportation expenditures have remained comparable over this time period. If one considers these five items-food, apparel, housing, health care, and transportation-as comprising “necessities,” then the share of consumption represented by necessities has not changed much over the last 25 years, remaining at about 85% overall. The share spent on child care for families with working mothers also remained constant at around 7% of income between 1985 and 1991 (Casper et. al. 1994). [12] Thus, it is only by selectively looking at food and apparel costs and ignoring other costs that critics can claim that necessities comprise a smaller share of spending. Focusing on durable goods ownership and the share of income spent on necessities does not answer the central question of whether the poor are able to meet their basic needs. In one of the richest countries in the world, there is evidence in this time of economic prosperity that the poor still face hardships. So-called “hardship measures” from the Census Bureau show that in 1995, 38% of households in the bottom income fifth had trouble paying for at least one basic necessity (food, housing, healthcare). Below we further discuss difficulties low-income families face in meeting their basic needs.

Absolute and relative difficulties in meeting basic needs

Food: While the share of a family’s income devoted to food has declined, food insecurity persists. Even in the context of the booming late 1990s economy, 46% of the non-elderly in families below 200% of the poverty level experienced either food shortages or worries about food shortages (Staveteig and Wigton 2000). This suggests that, while the cost of food relative to other goods has declined, families still often have to decide between paying other bills, such as rent or utilities, and purchasing food.

Housing: Critics like Robert Rector point out that, in 1992, 41% of households below the poverty line owned their homes (1998). But this share is just about half of the 78% of non-poor households that own their homes. It also masks differences in home-ownership rates between different types of poor families; the poor elderly are much more likely to own their homes (63% in 1992) than are poor single-parent families (24%) (Federman et al. 1996). Focusing on home ownership also ignores the crisis in affordable rental housing. In 1997, there were only 36 affordable [13] and available rental units for every 100 households making 30% of the median income (HUD 1999). Both owners and renters face increased housing costs. In 1995, 28% of households with children paid more than 30% of their income for housing, up from 15% in 1978 (Federal Interagency Forum on Child and Family Statistics 1999). Many poor families are unable to meet their housing costs; in 1995, 14% of households in the lowest income quintile were unable to pay their rent or mortgage, and 20% were unable to afford their utilities bill (Bauman 1999). Given the steep rise in housing costs since the mid 1990s, it is likely that these conditions have worsened.

Health care: Critics who argue that the prices of basic necessities have decreased relative to luxury goods downplay the rising cost of health care. Although recognizing that health care costs have increased dramatically, they dismiss this as the consequence of improved health care technology. However, improvements in health care only benefit those who are able to afford them. The percentage of non-elderly uninsured Americans increased from 14.8% in 1987 to 17.7% in 1996 (Kaiser Commission on Medicaid and the Uninsured 1998). Much of the increase in the uninsured stems from a decline in the number of employers offering health insurance benefits. While there has been a downward trend in employer-provided health insurance coverage for all workers, this trend has been greatest among the bottom fifth of wage earners. By 1996, less than a third of these workers (32%) received employer-provided health insurance, a decline of 8.8 percentage points since 1979 (Mishel et al. 1999). In contrast, 82.4% of those in the top fifth received health coverage through work in 1996. As a consequence of inadequate insurance coverage, many families in the lowest income quintile report going without medical care because they can’t afford it. In 1995, 11% of these households reported that in the past 12 months, one of the household members needed to see a doctor but couldn’t afford to go. For those without health insurance, 15% needed to see a doctor but didn’t go, regardless of their income level. In fact, people without health insurance in at least one out of the past four months were twice as likely to live in households that had trouble meeting other basic needs as were families with continuous health care coverage (Bauman 1999). Such a finding clearly demonstrates the burden health care costs place on the uninsured.

Child care: The cost of child care is completely ignored by many of those who claim that the poor are better off. The family budget literature shows that child care costs can comprise up to 30% of a working family’s budget, meaning that, for working families, child care can take as large a share of the family’s budget as housing (Bernstein et al. forthcoming). In 1991, families below the poverty level spent 27% of their income on child care, almost four times the share spent by higher income families (Casper et al. 1994). In 1997, the cost of quality, center-based child care ranged from an average of $8,268 annually in a city like Chicago to $3,640 annually in rural Arkansas (Children’s Defense Fund 1998). While more child care subsidies and tax credits are becoming available, many working families are not helped by these programs. [14] Transportation: High car-ownership rates by the poor reflect an increased reliance on automobiles to get to work. The spatial mismatch between poor urban workers and suburban jobs that are not accessible by public transportation has been well documented (Lacombe 1998). As for rural areas, only 40% have a public transportation system (Dewees 1998).

Higher education: While not necessary for basic survival, access to higher education is important for an equitable and economically mobile society. The rising cost of higher education is also ignored by those who argue that today’s poor are better off than the middle class of the past. College costs doubled between 1980 and 1998, while the median income of families with college-aged children only increased 22% over this period. And since those at the bottom of the income scale have lost ground in real terms, the share of income required to pay college costs has increased faster among these lower-income families. Financial aid has not kept pace with increased tuition levels during this time period, and between 1989 and 1999, there has been an increased reliance on loans versus need-based grants as a means of financing college (College Board 1999). In the end, it is obvious that, in some respects, today’s poor are better off materially than the poor of many years ago. But in relative terms, the consumption gap mirrors the income and wealth gaps. And even in absolute terms, many poor families continue to face difficulties meeting their basic needs.

Is rising inequality a serious concern or merely a necessary tradeoff for a growing economy?

Unlike the critiques discussed thus far, the final critique we examine takes the increase in inequality as a given, arguing that it is simply the inevitable outcome of economic progress, the natural result of a free market economy. According to this view, the faster growth of inequality over the past few decades is due to the fact that markets became less regulated, more global, and more technologically advanced over this period. In this regard, critics have argued that concerns about inequality’s growth are “a distraction,” a “red-herring,” or a “sideshow.” [15] Far from a distraction, we believe that this problem has already hurt our society in numerous ways, and has the potential to do considerably more harm. First of all, the argument that free markets generate ever-increasing inequality growth is belied by both historical evidence and by comparing the U.S. economy with that of other industrialized countries. In the 1950s through the early 1970s, the U.S. macro-economy was even more impressive than today’s, with lower unemployment and faster productivity growth. True, the structure of employment was very different then, with a much larger share of the blue-collar workforce in the expanding manufacturing sector, and fewer workers in the lower-paying services. Cash-transfer programs also became more generous over this era. Due to these and other trends (e.g., a higher real minimum wage and a more unionized workforce), this was a period of highly equalizing income growth: the average income of the bottom fifth more than doubled in real terms between 1947 and 1973, while that of the top fifth grew by 85%. Since the late 1970s-as we stressed in the earlier inequality report, Pulling Apart-income growth has been highly unequal, with the bottom fifth actually losing ground in real terms, the middle remaining relatively flat, and the top growing consistently. History clearly demonstrates that a fast-growing economy does not have to result in increased wage inequality. Furthermore, the current economy also challenges the connection between growth and inequality. Over the last few years of the 1990s, when unemployment finally began to fall to the range typical of the late 1960s, the growth of inequality attenuated significantly. If anything, the end of the last decade revealed that faster growth is associated with less, not more, inequality growth. We can also turn to other countries, such as those in Europe, to investigate whether growth and inequality should be expected to accompany one another in advanced market economies. In fact, this comparison has some advantages over the historical one just made because these economies face essentially the same global and technological environments as the U.S. does. As recent research on a number of OECD countries by Schmitt and Mishel (2000) shows, although the growth rates of GDP per capita and productivity were faster in OECD countries in the 1980s and 1990s, the levels and growth rates of earnings inequality in these countries were almost all lower than those of the U.S. Moreover, several European nations (Germany, France, Belgium, and the Netherlands) have now achieved, or surpassed, U.S. levels of productivity. [16] According to the survey by Gottschalk and Smeeding (1997), these countries all have substantially lower levels of income inequality and have experienced less growth in inequality than the U.S. Clearly, the OECD countries have been able to achieve GDP growth and efficiency gains comparable to the U.S. but without the high costs of inequality. [17] Nor can the growth of inequality be fully attributed to the growth in the returns to education. Over the 1980s, the pay premium for college-educated workers relative to those with less education grew steeply, and this certainly contributed to the growth in inequality. This trend led many commentators to fully attribute the growth of inequality to what appeared to be increased demand for college graduates relative to those with less education. [18] But as various inequality researchers have pointed out, the growth of “education differentials” explains about half of the increase in wage or income inequality since the late 1970s. [19] The rest of the growth has taken place within groups of workers with similar levels of education, and thus cannot be attributed to education. This pattern is most apparent in the 1990s, when the education pay premium has hardly grown, yet wage and income inequality continued to grow in the 1990s (even if at a slower rate). If educational pay differences fully explained growing inequality, the growth should have ceased in the 1990s. Instead, other factors not related to education continued to drive inequality’s climb. [20] A related argument attributes the rise of inequality to technology, with claims that inequality is an inevitable outcome of the shift to the new computer-driven economy. But as numerous analysts have shown, technological change did not arrive in the economy with the personal computer. [21] This research shows that technology has been an ongoing phenomenon; we have seen many “new economies” over the course of our economic history. Most importantly, technology’s impact on the wage structure, and thus on inequality, was no greater in the 1980s or 1990s than in earlier periods. [22]

In Pulling Apart, we go through the factors we think are most responsible for the growth in inequality. We will not review them here other than to say that they mostly relate to structural changes in the U.S. economy, such as the shift from manufacturing employment to lower-wage service jobs, increased trade imbalances, the decline in union power, the long-term fall in the minimum wage, and monetary policy that has unnecessarily kept the economy from reaping the benefits of full employment. We also provide a thorough analysis of these points in The State of Working America 2000-01 (Mishel et al. forthcoming).

While the problem of high unemployment has been ameliorated over the past few years, over the long term these factors threaten to undermine many working Americans’ faith in the economy. In a recent Business Week/Harris Poll, 75% of respondents “felt that the benefits of the ‘new economy’ are unevenly distributed” (Business Week 1999). This sentiment may very well have arisen from the fact that productivity—the best single indicator of the impact of new technologies on overall growth—grew by 18% in the 1990s while the real median family income grew only 4%. The typical American family is working harder than ever and spending more hours in the labor market. Only over the past few years has it begun to see some returns from its effort and sacrifices. But even so, most families’ fortunes are clearly lagging well behind those of the better off.

The repercussions of increasing income inequality are manifold. Some polls and studies reveal the public’s great concern that the wealthy are disproportionately determining the outcome of the political process, associating increased inequality with political disengagement (Lewis 2000). And new “growth theory” literature in economics is starting to build an empirical case that the economies in countries with higher levels of inequality grow more slowly than those of economies with more equal distributions (Aghinon et al. 1999). Finally, new medical research suggests a connection between increasing inequality and worse health outcomes (Boston Review 2000).


There are many reasons to be concerned about the phenomenon of growing inequality, and it is irresponsible to try to explain it away with misleading arguments that ignore the undeniable trend toward larger gaps between those at the top, the middle, and the bottom of the income or wealth scales. It is also wrongheaded to argue that increased consumption or the mobility of low-income families counteracts the growth of income inequality. Perhaps worst of all would be to dismiss inequality’s growth as an inevitability or distraction. If we want working Americans to have a stake in the economy, they must receive their fair share of its growth.

The authors would like to thank Elizabeth McNichol, Gary Burtless, Frank Levy, and John Schmitt for helpful comments. We are also grateful to The John D. and Catherine T. MacArthur Foundation; The Joyce Foundation; The Rockefeller Foundation; and the Charles Stewart Mott Foundation for their support of this work.


1. For examples of this literature, see Danziger and Gottschalk (1995), Karoly and Burtless (1995), Ryscavage (1995); for an international perspective, see Gottschalk and Smeeding (1997).

2. See analysis in Chapter 7 in The State of Working America 2000–01 (Mishel et al. forthcoming).

3. The Census Bureau has an additional measure that converts home equity into an income-equivalent annuity. We do not analyze this measure because it shifts from an income to a consumption measure. We also examined a measure comparable to “comprehensive income,” except that government medical programs are excluded. Data with this definition go back only to 1991. Our analysis shows that government medical programs disproportionately help the bottom fifth but do not alter the trend in inequality relative to the comprehensive measure that includes medical programs.

4. The CBO is able to analyze trends in the upper 1% in a way that cannot be done with the Census Bureau data because the CBO corrects for top-coding problems.

5. This adjustment is made by dividing a family’s income by their poverty threshold. Since the poverty lines are adjusted for family size, this serves to adjust incomes for this same factor.

6. In Census Bureau terminology, families are units of two or more persons related by blood, marriage, or adoption, while households include one-person units. Our Pulling Apart study uses family income.

7. This framework leads to the curious result of having more than 20% of their sample in the top fifth by the end of the period analyzed.

8. These tables, provided by Gottschalk, appear on p. 88 of Mishel et al. (forthcoming).

9. Figure 1K of Mishel et al. (forthcoming) cites other work by Gottschalk showing that the percent of families staying on the same fifth in adjacent years has trended up since the late 1960s, evidence of declining rates of mobility.

10. Analysis using longitudinal data shows that only one-third of those who were poor in an average month in 1994 were poor throughout both that year and 1993 (Naifeh 1998).

11. Another group with low incomes and high assets are the elderly, a group that has accumulated durable goods but now have low incomes.

12. Child care is not included in Figure 1 because only families with children in which all adults work outside the home face this expense.

13. Housing is considered “affordable” if it costs less than 30% of a family’s income.

14. The Child and Dependent Care Tax Credit is non-refundable, and therefore low-income families with no tax burden cannot receive this credit. Furthermore, many eligible families are turned away from assistance through the Child Care and Development Block Grant (Greenberg 1998).

15. The first two comments were in Cox and Alm (2000), and the latter was made by John Weiker of the Hudson Institute during a debate on this topic with Jared Bernstein.

16. As Schmitt and Mishel (2000) show, this would be true even if we adjusted the data to account for the greater unemployment in these other countries.  17. For evidence of the slower growth in income inequality in these countries compared to the U.S., see Gottschalk and Smeeding (1997, Table 4). The fact that unemployment rates are higher in Europe than they are in the U.S. has led some critics to suggest that there exists a tradeoff between unemployment and inequality. There is, however, little evidence for this claim, and a number of recent articles have challenged this connection. See Glyn and Salverda (2000), Schmitt and Mishel (2000), and OECD (1996).

18. Recent research has challenged the notion that the demand for college-educated workers accelerated in the 1980s over the 1970s (see Card and Lemieux (200) and Mishel et al. forthcoming, chapter 2)). This work attributes the faster growth in the college premium to either a deceleration in the supply of college-educated workers or to the negative impact of structural changes (discussed below in the text) of the wages of non-college graduates.

19. See Burtless (1995) and Mishel et al. (forthcoming, Table 3.24).

20. Of course, education differentials grow for reasons other than changes in skill demand. For instance, institutional changes in the labor markets, such as the decline in the real value of the minimum wage and the rate of unionization (both of which disproportionately affect the wages of non-college-educated workers) also led to higher education premiums over the 1980s.

21. See discussion of this point in Bernstein and Mishel (forthcoming).

22. While inequality continued to grow in the 1990s, which was a period of accelerated computer investment, education differentials grew much more slowly than they did in the 1980s. This pattern contradicts the technology argument, which closely associates increased computer usage with the increase in the college premium.


Aghion, Philippe, Eve Caroli, and Cecilia Gracia-Penalosa. 1999. Inequality and Economic Growth: The Perspective of New Growth Theories. Journal of Economic Literature. Vol. 37, No. 4. Baker, Dean, and John Schmitt. 1998. “The Macroeconomic Roots of High European Unemployment — The Impact of Foreign Growth.” Paper presented at “Creating Competitive Capacity” conference, Washington, D.C. Available online at

Bauman, Kurt J. 1999. Extended Measures of Well-Being: Meeting Basic Needs. (P70-67). Washington, D.C.: U.S. Census Bureau.

Bernstein, Jared, and Lawrence Mishel. Forthcoming. “Six Reasons for Skepticism About the Technology Story of U.S. Wage Inequality.” In Ivan Berg and Arne Kalleberg, eds., Sourcebook on Labor Markets. New York: Plenum Publishing.

Bernstein, Jared, Chauna Brocht, and Maggie Aguilar. 2000. How Much is Enough: Basic Family Budgets for Working Families. Washington, D.C.: Economic Policy Institute.

Boston Review. 2000. New Democracy Forum: Justice is Good for Our Health. Boston Review. Vol.25, No. 1.

Burtless, Gary. 1995. Widening U.S. Income Inequality and The Growth in World Trade. Washington, D.C.: Brookings Institution.

Business Week. 1999. Business Week/Harris Poll. December 27. pp. 52-4.

Card, David, and Thomas Lemieux. 2000. “Can Falling Supply Explain the Rising Return to College for Younger Men? A Cohort-Based Analysis.” NBER Working Paper No. W7655.

Casper, et al. 1994. Who’s Minding the Kids? (P70-36). Washington, D.C.: U.S. Census Bureau.

Children’s Defense Fund. 1998. Child Care Challenges. Washington, D.C.: CDF.

Citro, Constance, and Robert Michael. 1995. Measuring Poverty: A New Approach. Washington, D.C.: National Academy Press.

The College Board. 2000. Trends in Student Aid 1999. New York, N.Y.: The College Board.

Congressional Budget Office. 1998. Estimates of Federal Tax Liabilities for Individuals and Families by Income Category and Family Type for 1995 and 1999. Washington, D.C: CBO.

Cox, Michael, and Richard Alm. 2000. Why Decry the Wealth Gap? New York Times. January 24.Danziger, Sheldon, and Peter Gottschalk. 1995. America Unequal. Cambridge, Mass.:Harvard University Press.

Dewees, Sarah. 1998. The Drive to Work: Transportation Issues and Welfare Reform in Rural Areas. Mississippi State, Miss.: Southern Rural Development Center.

Federal Interagency Forum on Child and Family Statistics. 1999. America’s Children: Key National Indicators of Well-Being, 1999. Washington, D.C.: Federal Interagency Forum on Child and Family Statistics.

Federman, et al. 1996. What Does it Mean to Be Poor in America? Monthly Labor Review. May.

Gottschalk, Peter. 1996. “Notes on ‘By Our Own Bootstraps: Economic Opportunity and the Dynamics of Income Distribution’ by Cox and Alm.” Unpublished memo.

Gottschalk, Peter, and Timothy Smeeding. 1997. Cross-National Comparisons of Earnings and Income Inequality. Journal of Economic Literature. Vol 35, No. 2.

Greenberg, Mark. 1998. Child Care Policy Two Years Later. Washington, D.C.: Center for Law and Social Policy.

Glyn, Andrew, and Wiemer Salverda. 2000. Does Wage Flexibility Really Create Jobs? Challenge. Vol. 43, No.1.

Kaiser Commission on Medicaid and the Uninsured. 1998. Uninsured Chartbook. Washington, D.C.: Kaiser Family Foundation.

Karoly, Lynn, and Gary Burtless. 1995. Demographic Change, rising Earnings Inequality, and the Distribution of Personal Well-Being, 1959-89. Demography. Vol. 32, No. 3, pp. 379-406.

Lacombe, Annalyn. 1998. Welfare Reform and Access to Jobs in Boston. Washington, D.C.: U.S. Department of Transportation.

Lewis, Charles. 2000. The Buying of the President 2000. New York, N.Y.: Avon Books.

Madrick, Jeff. 1999. How New Is the New Economy? New York Review of Books. September 23.

Mishel, Lawrence, and Jared Bernstein. 1998. Technology and the Wage Structure: Has Technology’s Impact Accelerated Since the 1970s? Research in Labor Economics, Vol. 17.

Mishel, Lawrence, Jared Bernstein, and John Schmitt. Forthcoming. The State of Working America 2000-01. Ithaca, N.Y.: Cornell University Press.

Naifeh, Mary. 1998. Poverty, 1993 to 1994: Trap Door? Revolving Door? Or Both? Current Population Reports, Household Economic Studies, (P70-63). Washington, D.C.: U.S. Bureau of the Census.

Rector, Robert. 1998. America Has the World’s Richest Poor People. Wall Street Journal. September 24.

Rector, Robert, and Rea Hederman. 1999. Income Inequality: How Census Data Misrepresent Income Distribution. Washington, D.C.: Heritage Foundation.

Reynolds, Alan. 2000. The Great “Income Gap” Scam. New York Post, January 21.

Ryscavage, Paul. 1995. A Surge in Growing Income Inequality? Monthly Labor Review, Vol. 118, No. 8.

Schmitt, John, and Lawrence Mishel. Forthcoming. “The United States is not Ahead in Everything that Matters.” Challenge.

Staveteig, Sarah, and Alyssa Wigton. 2000. Racial and Ethnic Disparities: Key Findings from the National Survey of America’s Families. Washington, D.C.: Urban Institute.

U.S. Department of Housing and Urban Development. 1999. The Widening Gap: New Findings on Housing Affordability in America. Washington, D.C.: U.S. HUD.

Vanderheide, Walter. 1999. Tracking Changes in Consumers’ Spending Habits. Monthly Labor Review. Vol. 122, No. 9.