Distribution of Income

Whether government should redistribute income is a normative question, and each person’s answer will depend on his or her values. But for many people, answering the normative question requires understanding the facts about the current income distribution. The term “income distribution” is a statistical concept. No one person is distributing income. Rather, the income distribution arises from people’s decisions about work, saving, and investment as they interact through markets and are affected by the tax system. The 1990s and early 2000s witnessed the establishment of a growing body of work, increasingly precise, describing how the income distribution has changed. This work can be summarized in three points:

The most frequently cited statistics come from the U.S. Census Bureau’s Current Population Survey (CPS), the monthly household survey best known as the source of the official unemployment rate. Since 1948, the March edition of the CPS has collected household income information for the previous year, as well as the personal characteristics of household members—their age, education, occupation, and industry (if they work), and other data that help give insight into changing income patterns. Although this makes the CPS an indispensable statistical source, it has disadvantages as well. The CPS uses a restricted income definition: pretax money receipts excluding capital gains. This definition is further restricted by a “cap,” currently $999,999, imposed on reported annual earnings for reasons of confidentiality.1 Together, these problems mean that CPS estimates of inequality omit the effects of taxes, nonmoney income such as government and private health insurance, and the portion of individual earnings that exceeds the cap.

A second source of inequality statistics is the U.S. Treasury’s Statistics of Income (SOI), which summarizes income reported on federal income tax returns. SOI data contain no personal data on taxpayers such as age or education, and they cannot describe the precise shape of the lower part of the income distribution.2 The strengths of SOI data are their ability to accurately describe the upper part of the distribution—SOI income data are not “capped”—and to extend this description back to 1917, thirty years before CPS statistics begin.

Table 1 contains selected information on CPS measures of family and household income inequality since World War II.3 The upper panel describes income patterns across families: living units occupied by two or more related persons. The lower panel describes income patterns across households: all occupied living units including families, persons who live alone, unrelated roommates, and so on. To form each distribution, the sample of families (or households) is listed in order of increasing income. The Census then calculates the fraction of all family income going to the quintile (one-fifth) of families with the lowest incomes, the quintile of families with the second-lowest incomes, and so on, as well as the share going to the highest 5 percent of families (who also are included in the top quintile).

The CPS data in Table 1 trace a J-shaped evolution of post–World War II inequality. In 1947, the top quintile of families received $8.60 for every dollar of income received by the bottom quintile. This ratio fell gradually through the 1950s and 1960s until 1969, when it reached $7.25 to $1.00—the low point of inequality. Beginning in the late 1970s, the ratio began to rise again until, by 2002, it had increased to $11.36 to $1.00, significantly greater than in 1947. Household data tell a similar story.

To make these trends more concrete, Table 1 includes the 1947, 1979, and 2001 income levels that divide each quintile from the next. Similar data are presented for households, and all income levels are expressed in 2003 dollars. Between 1947 and the mid-1970s, income grew rapidly at all points in the distribution, resulting in both rising living standards and moderating inequality. After the mid-1970s, average income grew much more slowly, and the growth that did occur was concentrated in the distributions’ upper half. Between 1979 and 2001, the income dividing the first and second family quintiles grew slightly, from $22,280 to $24,000 (7.7 percent), while the income dividing the fourth and fifth quintiles grew from $74,470 to $94,150 (26.4 percent). Now, as in the 1970s, a majority of families would describe themselves as middle class, but the “middle class” is now a larger, more diverse concept than it once was.4

Inequality estimates based on the U.S. Treasury’s SOI data expand on this picture. At the outset, SOI data do not “cap” high incomes, so household income inequality as reported in the SOI is significantly larger.5 Using “capped” statistics, the CPS reports that the top one-fifth of households receives 49 percent of all pretax money income. The SOI estimates, more accurately, that the top one-tenth of households, with average annual income of about $200,000, receives 42 percent of total pretax money income. The top 1 percent of households with average annual incomes of about $800,000 receives 15 percent of all pretax income. With their longer historical perspective, SOI statistics also show that inequality in the 1920s and 1930s was as high as it is today. Beginning in 1938, the income share of the top one-tenth of households fell from 43 percent to about 32 percent, where it remained until the deep blue-collar recession of the early 1980s. At that point, inequality began its return to the levels of the 1920s and early 1930s.6

In one sense, the growth of inequality in the last part of the twentieth century comes as a surprise. In the 1950s, the bottom part of the income distribution contained large concentrations of two kinds of families: farm families whose in-kind income was not counted in Census data, and elderly families, many of whom were ineligible for the new Social Security program. Over subsequent decades, farm families declined as a proportion of the population while increased Social Security benefits and an expanding private pension system lifted elderly incomes. Both trends favored greater income equality but were outweighed by four main factors.

These factors, however, can explain only part of the increase in inequality. One other factor that explains the particularly high incomes of the highest-paid people is that between 1982 and 2004, the ratio of pay of chief executive officers to pay of the average worker rose from 42:1 to 301:1, and pay of other high-level managers, lawyers, and people in other fields rose substantially also.

As noted above, both CPS and SOI statistics measure pretax money income. These measurements are deficient for three reasons. First, increases in governmental aid to the poor have been concentrated in nonmoney benefits such as Medicaid and food stamps and through tax credits under the Earned Income Tax Credit (EITC). Nonmoney benefits are excluded from standard statistics, and EITC tax credits are typically underreported. Second, an increasing proportion of wage-earners’ total compensation goes to health insurance and pension benefits—which are not counted in standard statistics. Third, taxes themselves modify the income distribution.

The U.S. Census has attempted to correct these definition problems for recent years by estimating the household income distribution under alternative income definitions. Table 2 shows the effect in 2001 of moving from the standard Census definition (pretax money excluding capital gains) to an adjusted definition that includes the estimated effects of capital gains, taxes, the EITC, and the monetary value of private and governmental nonmoney benefits. The result is a substantial reduction in inequality, with the ratio between incomes in the top and bottom quintiles falling from $14.31:$1.00 to $10.40:$1.00. Similar adjustments for selected earlier years indicate that better income measurement reduces inequality in any single year. Even under the adjusted definition, though, the trend toward increasing inequality in the 1980s and 1990s remains, but at a slower pace.

A second offset to estimated inequality is economic mobility. Because most family incomes increase as people’s careers develop, long-run incomes are more equal than standard single-year statistics suggest. Table 3 summarizes the results of one study of recent family income mobility.8 Among families in the bottom quintile in 1988, half were in the bottom quintile ten years later, a quarter had moved up to the second quintile, and a quarter had moved to the third or higher quintiles. Families in the fifth quintile (highest incomes) show a similar mobility over time.