[An updated version of this article can be found at Distribution of Income in the 2nd edition.]
The distribution of income is central to one of the most enduring issues in political economics. On one extreme are those who argue that all incomes should be the same, or as nearly so as possible, and that a principal function of government should be to redistribute income from the haves to the have-nots. On the other extreme are those who argue that any income redistribution by government is bad.
Whether government should redistribute more or less income is, of course, a normative question. Each person's answer depends on his values. But for many people, answering the normative question requires an understanding of just how income is distributed now, and how—and why—the distribution has changed over the decades. To start, here is what the basic numbers tell us.
A statistical summary of U.S. family income distribution since World War II shows the following:
2. The degree of income inequality is not much greater today than it was at the end of World War II.
3. Family income inequality declined slowly from 1946 through 1969, increased slowly from 1970 through 1979, and has increased somewhat faster since then.
Data for the summary comes from the U.S. Census Bureau's Current Population Survey (CPS), a monthly survey of sixty-five thousand households that includes both families and unrelated individuals. Every March the CPS collects data on household income in the previous year. To keep interviews simple, the questions focus on gross money income (excluding capital gains but including interest and dividends). This means that census statistics—the standard source of income data—measure income before taxes and do not count nonmoney income like Medicare coverage and employer-paid health insurance.
The census constructs the family income distribution by listing CPS sample families in order of increasing income. The distribution is described by computing the share of total family income going to the poorest one-fifth of families, the second fifth, and so on. Part A of table 1 contains these data for selected years since World War II.
In 1929, on the eve of the Great Depression, the richest one-fifth of families received over half of all income going to families. The depression and World War II reduced that figure so that in 1949, the richest one-fifth of families received 42.7 percent of all family income while the poorest one-fifth received 4.5 percent. Put differently, in the late forties the top fifth of families received about $9.50 in income for every $1.00 received by the bottom fifth. Inequality continued to decline slowly through the fifties and sixties. By 1969 the top fifth of families received $7.25 for every $1.00 received by the bottom fifth.
But 1968 and 1969, when unemployment averaged 3.5 percent, marked the high point of family income equality. Beginning in 1970, inequality began to grow again at a moderate rate. Through much of the seventies, the slow rise of inequality seemed to reflect the economy's general weakness. Yet contrary to expectations, inequality increased even more rapidly in the post-1982 recovery. In 1989 the top fifth of families received $9.69 in gross money income for every $1.00 received at the bottom, roughly the same as in the late forties.
Family income inequality is now high not only by our own post-World War II standards, but also when compared to other industrialized countries. Detailed comparisons from the Luxembourg Income Study show that in the United States, West Germany, and Israel, the richest fifth of families receives about 45 percent of all family income, compared to 39 percent in Sweden and 41 percent in Canada, the United Kingdom, and Norway. One reason for greater U.S. inequality is the large number of female-headed families, which comprise about two-fifths of the bottom quintile. These female-headed families, in turn, reflect the nation's high divorce rate and high rate of out-of-wedlock births. Partially because of these single-parent families, there are only eight earners for every ten families in the distribution's bottom fifth, while there are twenty-three earners for every ten families in the top fifth. In recent years U.S. income inequality has also been driven by falling wages for less skilled workers and relatively limited cash benefits for the poor.
A Vanishing Middle Class?
Although many people have claimed that the United States is losing its middle class, the 1989 distribution of income was not radically less equal than the distribution of 1949 or 1959, a period when the middle class was perceived to be growing rapidly. Moreover, the whole concept of a middle class is vague. In 1989 the top fifth of families (with 44.6 percent of all income) included every family with income above $59,500, many of whom saw themselves as middle class. Because the concept is vague, it follows that inequality statistics cannot, by themselves, say whether the middle class is vanishing. They must be supplemented with data on both economic growth and demographics.
Begin with economic growth. In 1947, median family income—the midpoint of the income distribution—stood at $16,712 (all income figures are in 1990 dollars). By 1973 it had doubled to $33,398. During these years income inequality had declined modestly. But more important, the whole income distribution had moved to much higher real incomes. The poor and the rich were both getting richer and an increasing proportion of all families could afford a middle-class life, including a single-family home, two cars, and so forth.
Since 1973, median family income has grown very little. Income growth fell victim to oil price shocks and to the productivity slowdown (the slow growth of output per worker that has plagued most industrialized countries). This slow growth has affected our outlook on economic life. When incomes grow rapidly, more inequality means that the poor get richer but the rich get richer faster. But when inequality increased in the slow-growth eighties, some groups' incomes fell in real terms. Between the business cycle peak of 1979 and the next business cycle peak of 1989, the average income of the poorest fifth of families fell from $10,900 to $10,200, while the average income of the top fifth grew from $89,600 to $97,600. Moreover, the price of two key pieces of a middle-class life—a single-family home and a college education—grew faster than the general rate of inflation and faster than average incomes. For all of these reasons, slow income growth played a key role in perceptions of a vanishing middle class.
Turn next to demographics, where movements of families within the income distribution add to perceptions of a vanishing middle class. In popular culture the middle class usually appears as urban families with children. In the late forties these families were concentrated in the top four-fifths of the distribution. The poorest fifth contained mainly farm families, other rural families, and elderly families, most of whom were not yet eligible for Social Security benefits. Put differently, those families that "should have been" in the middle class were relatively unlikely to have incomes in the lowest quintile.
Today the situation is reversed. Farm and rural families are far fewer in number. Many elderly families have moved from the bottom of the distribution to the lower middle, the result of a mature Social Security program and better private pensions. Now the poorest fifth does contain urbanized families with children. Included in the group are a significant number of families headed by single women and, since the early eighties, husband-wife families hurt by a sharp drop in the wages of men with a high school education or less. A higher proportion of families that "should be" in the middle class are now in the lowest quintile. For urban families with children, the picture of a distribution with a shrinking middle has some validity.
In sum, slow income growth and movements within the income distribution have led to a sense of a vanishing middle class even though overall family income inequality has not increased very much.
Does Measurement Matter?
All the data on family income discussed so far are on a pretax, money-only basis. Would a different income definition lead to a different story?
Consider three reasons why it might. First, increases in aid to the poor over the last two decades have been concentrated in nonmoney benefits like food stamps and Medicaid, neither of which is counted in standard census statistics. Second, an increasing proportion of wage earners' total compensation over the same period is for health insurance and other nonmoney benefits (again not counted). Third, taxes modify the income distribution.
The Census Bureau has attempted to answer this question for recent years by estimating the household income distribution under a number of alternative income definitions. Households include both families and unrelated individuals. Table 2 shows the impact in 1989 of moving from the standard census income (pretax money only) to an adjusted census income that subtracts taxes paid from gross income and adds to income the cost of benefits provided by the government and the employers. Improving the measure of income in this way reduces the level of inequality by shrinking the numbers of households at the highest and lowest ends of the income spectrum. In 1988 the percent of households with income over $100,000 falls by almost half, and the proportion with incomes under $10,000 falls by almost one-fourth.
Better income measurement reduces the estimated level of income inequality, but it does not appear to alter the trend in inequality over time. While the Census Bureau has produced adjusted income distributions for only a few years, these data and such other evidence as exists suggest that the trend toward greater inequality in the eighties would exist under any plausible definition of income.
A different kind of adjustment changes our picture of income growth: an adjustment of family income for family size. Over the last two decades average family size has declined by 12 percent, largely reflecting fewer children per family. As a result, even though family income has been largely stagnant, income per family member, a rough measure of living standards, has increased by about 15 percent. If we look beyond families to all households—including the growing number of young singles who have only themselves to support—the Congressional Budget Office estimated that living standards since the early seventies have increased by 20 percent. There are limits to this type of adjustment, however. Most people eventually form families and most families have children. In the long run living standards cannot grow faster than the general rate of economic growth.
Frank Levy is the Rose Professor of Urban Economics at MIT's Department of Urban Studies and Planning.
Congressional Budget Office. Trends in Family Incomes: 1970-1986. Report prepared by Robertin Williams. 1988.
Levy, Frank. Dollars and Dreams: The Changing American Income Distribution. 1989.
Levy, Frank. "Incomes, Families and Living Standards." In American Living Standards: Threats and Challenges. edited by Robert Litan, Robert Lawrence, and Charles Schultze. 1988.
O'Higgins, Michael, Gunther Schmaus, and Geoffrey Stephenson. "Income Distribution and Redistribution: A Microdata Analysis for Seven Countries." In Poverty, Inequality and Income Distribution in Comparative Perspective, edited by Timothy M. Smeeding, Michael O'Higgins, and Lee Rainwater. 1990.
U.S. Bureau of the Census. "Measuring the Effect of Benefits and Taxes on Income and Poverty: 1989." Current Population Reports, series P-60, no. 169. September 1990.
U.S. Bureau of the Census. "Trends in Income by Selected Characteristics." Current Population Reports. series P-60, no. 167. April 1990.