Rarely has a technical issue in economics generated so much passion or consumed so much ink as the dispute over changes in income distribution in the 1980s. Polemics flew last year in The Wall Street Journal between former President Bush's top economic adviser, Michael Boskin, and MIT economist and Clinton adviser Paul Krugman.
Any number of liberal politicians, including President Clinton, have decried the alleged demise of the middle class and the runaway gains of the rich. Yet this "fairness" debate, so important in the 1992 presidential campaign, has been fundamentally misguided by arguments over the wrong data.
Caught up in analyzing numbers, economists sometimes forget what they are really trying to measure. The fairness question has centered entirely on comparisons of income, an easily quantified and understood concept. But income is really only a proxy for something more important: material welfare. And in this regard, income proves to have been a highly misleading indicator.
With this realization, a growing number of social scientists are taking another look at the data and coming to surprising conclusions. The traditional view that inequality followed a U-shaped curve—falling in the 1960s and '70s, only to rise sharply in the '80s—may be flat wrong.
"The corrected statistics show that the standard of living is rising, inequality is falling, and poverty is disappearing," asserts the distinguished Harvard economist Dale Jorgenson. 'The 'fairness' issue that has dominated the recent debate over tax policy is based on faulty statistics."
Economists have long understood that consumption is a better measure than income of welfare and that the two often diverge. A telling indicator is the fact that, according to government surveys, the bottom fifth of households in 1990 actually spent 80 percent more than their reported income. A study by MIT economist James Poterba of 1985 data shows that only 61 percent of people in the bottom income tenth were also at the bottom in terms of consumption.
How could this be? The poor are usually eligible for substantial noncash benefits, including medical care, food stamps, and housing allowances, which do not show up in official income statistics. And many people, including those on welfare, work in the informal or underground sectors of the economy without reporting their full incomes. (Northwestern University sociologist Christopher Jencks devotes a chapter of his recent book, Rethinking Social Policy, to documenting this finding.)
People may also "consume" more than they earn if they own substantial assets, such as houses. A surprising 40 percent of all poor households own their own homes, the overwhelming majority of which are in good condition, according to surveys by the Census Bureau and the Department of Housing and Urban Development. These homes provide an important stream of valuable services (shelter being the most obvious) even when family income is low.
Another major reason for the wide discrepancy between income and consumption is that people save or borrow to smooth out transitory income gains or losses. They try to maintain a relatively level degree of consumption that reflects their long-term resources, or "permanent" income. Consumption at any given time may thus be higher than actual earnings.
For example, a medical school student may appear impoverished by normal measures of income yet live reasonably well by borrowing against his or her future income. Someone who receives a one-time bonus or large capital gain may show up in that year's statistics as rich yet save the windfall and maintain a modest standard of living.
Imagine a society in which every citizen becomes unemployed at age 40 for one year but otherwise enjoys exactly the same salary. Everyone is therefore equal. Yet an income-distribution chart would show a huge gap between individuals earning nothing and those earning the standard salary—a false sign of inequality. Measures of consumption, however, would more accurately show everyone nearly equal—since most people would save to tide them through their year of unemployment.
In the real world, individual incomes do bounce around from year to year. A one-year snapshot thus gives a misleading picture of long-term welfare. Nearly half of all poverty spells last no more than a year, for example. And in any given year, nearly a quarter of all people in the bottom fifth of income move up and out. At the same time, many of the temporarily rich move down a notch or two. The result is that the distribution of resources over time is much more equal than income measures suggest.
Even critics of the Reagan era grant these observations. For example, Harvard economists David Cutler and Lawrence Katz, now chief economist at the Department of Labor, state in a recent working paper for the National Bureau of Economic Research that "economic theory suggests that permanent income or consumption is a more accurate measure of the distribution of resources than is current money income."
Measures of consumption are not only better indicators of well-being at any one time but also better indicators of changes over time. Christopher Jencks and Susan Mayer, a sociologist at the University of Chicago, argue in a new paper that "the distribution of reporting errors, taxes, saving and borrowing, noncash benefits, physical and financial assets, consumer efficiency, household size, age, physical and mental health, and work-related expenses have all changed over the past generation. As a result, changes in the distribution of income need not imply parallel changes in the distribution of material well-being."
Why, then, do the vast majority of economists persist in haggling over census data on personal and household income? The reason may simply be intellectual inertia: The numbers are easy to get and familiar to work with. "We are wedded in tradition," says University of Texas economist Daniel Slesnick, a leading revisionist. "I don't think it is a matter of dispute that it is better looking at consumption rather than income, but there is a large investment in this income data."
What difference does this make to the fairness debate? It turns out that consumption can be measured—thanks to Census Bureau surveys of consumption expenditures—and that the data throw an entirely new light on changes in material welfare over the decades.
Recent unpublished work by Slesnick, who pioneered some of the analytical techniques with Jorgenson, suggests that "the level and trend of inequality" when measured in terms of consumption "is dramatically different from the income inequality index." By 1989, the level of inequality was 23 percent lower than in 1947, he found. Most of the trend toward greater equality took place between 1958 and 1973, but no reversal took place thereafter. The infamous U-turn in inequality, he declares, is simply "illusory."
Slesnick adjusts the consumption data according to sophisticated econometric estimates of individual needs in different household types based on actual expenditure patterns. He accounts for such variables as family size, type of residence (farm or nonfarm), region of residence, and the age, sex, and race of the household's head.
Some of his adjustments are not particularly controversial. Most economists agree that bigger families need more resources than smaller ones to maintain the same standard of living, for example. (This adjustment can be significant, since families in the highest income fifth are 80 percent larger, on average, than those in the bottom quintile.) On the other hand, Slesnick admits that there is no consensus behind his adjustments for sex and race.
But even scholars who part company with Slesnick on some of these points come up with similar results. Mayer and Jencks, for example, calculate that the ratio of the top quintile's consumption to that of the bottom quintile changed only microscopically from 1972-73 to 1988-89, from 4.78 to 4.81. Comparable ratios for income, on the other hand, were much higher and rose faster, from 7.95 to 9.79. "My conclusion is that there was really not an increase in inequality to make any fuss over," Mayer says of her findings.
Using different assumptions of what adjustments are appropriate between different kinds of households, Cutler and Katz produce more pessimistic conclusions from the consumption data. But even they acknowledge that income measures make matters look worse than they are. Their top-to-bottom consumption ratio rises from 3.9 in 1972-73 to 5.1 in 1988-89, but even in the latter period the level of inequality remains only half that for income. Using their consumption measures, moreover, brings the poverty rate in 1988 down from 12 percent (as measured by the census) to 8.6 percent.
What should one make of all these data? These new numbers should recast the empirical terms of the political debate, but they will not and cannot end the controversy.
It is still perfectly possible to conclude from this new picture of inequality in America that resources are distributed unfairly. It is true that earnings have grown more unequal over time. One can still argue that Presidents Reagan and Bush should have done more to help the poor or soak the rich (although most economists agree that their tax and spending policies were not the primary cause of widening earnings and income inequality during the '80s). Social critics may argue that some Americans went too far into debt to maintain levels of consumption when their incomes fell. On the other side, defenders of the status quo may take comfort in the consumption data and defend rising inequality of earnings as reflecting appropriate returns to acquired skills and education.
But these are primarily matters of values and ideology. The debate over fairness should rest there—not on misleading data and faulty methodology.
Contributing Editor Jonathan Marshall is economics editor of the San Francisco Chronicle.
This article originally appeared in print under the headline "Economics: Inconspicuous Consumption".