The American Prospect
In 1992, economist Paul Krugman, now a New York Times columnist, published this article in the Fall issue of The American Prospect. Today, his assertions hold up, especially in answer to the conservative critics of Thomas Piketty's Capital in the Twenty-First Century .
By Paul Krugman
Editor's note: This article originally appeared in the Fall 1992 issue of The American Prospect. The author, in his New York Times column, cited this piece on June 1, 2014, in refutation of the assertions of the Financial Times'Chris Giles that Thomas Piketty got his data wrong in the bestselling book,Capital in the 21st Century.
During the mid-1980s, economists became aware that something unexpected was happening to the distribution of income in the United States. After three decades during which the income distribution had remained relatively stable, wages and incomes rapidly became more unequal. Academic researchers soon began arguing vigorously about the causes of the growth in inequality: was it global competition, government policy, changing technology, or some other factor? What nobody, whatever his or her political stripe, questioned was the fact that there had been a dramatic change in income distribution.
During 1992 this genteel academic discussion gave way to a public debate, carried out in the pages of the New York Times, the Wall Street Journal, and assorted popular magazines. This public debate was remarkable in two ways. First, the conservative side displayed great ferocity in presenting its case and attacking its opponents. Second, conservatives chose to take an odd, and ultimately indefensible, position. They could legitimately have challenged those who have called attention to the growing dispersion of income on the grounds that nothing can, or at any rate should, be done about it. But with only a few exceptions they chose instead to make their stand on the facts to deny that the massive increase in inequality had happened. Since the facts were not on their side, they were forced into an extraordinary series of attempts at statistical distortion.
The whole episode teaches us two lessons. At one level, it is a sort of textbook demonstration of the uses and abuses of statistics. This article reviews that lesson, tracing out how conservatives tried to distort the record and why they were wrong. But the combination of mendacity and sheer incompetence displayed by theWall Street Journal, the U.S. Treasury Department, and a number of supposed economic experts demonstrates something else: the extent of the moral and intellectual decline of American conservatism.
I begin with a review of the basic data, followed by an assessment of the three kinds of conservative attacks on the simple facts about growing inequality: (i) efforts to deny the facts, through a mixture of confused statistical arguments; (ii) claims that the growth record of the Reagan years outweighs or negates any apparent increase in inequality; (iii) claims that income mobility makes comparisons of the income distribution at a point in time meaningless. A final section tries to put some perspective on the whole debate.
Some Basic Facts
There are some non-official sources that provide evidence for growing inequality of income in the United States. For example, Fortune has long carried out annual surveys of executive compensation; and since the mid-1970s compensation of top executives has risen far faster than average or typical wages, a process entertainingly discussed by Graef Crystal in his In Search of Excess. Surveys carried out by the University of Michigan have also shed useful light on income distribution, in particular on the dynamics of income over time. There is also anecdotal evidence: Tom Wolfe noted the soaring demand for apartments in Manhattan's "Good Buildings" well before academics had started to take the growing concentration of wealth seriously, and indeed his Bonfire of the Vanitiesarguably tells you all you need to know about the subject.
What the Census Shows
Most academic studies on the distribution of income in the United States rely on Census data, compiled from the Current Population Survey. These data have certain limitations, to which I will turn in a moment. But as a starting point, the Census numbers have one great advantage: they are not controversial. In all the mud-slinging of the income distribution debate, nobody has yet accused the Census of bias or distortion (although that may come next).
Figure 1 shows a picture that ought to be part of the consciousness of anyone who thinks about trends in the U.S. economy since the 1970s. The figure shows the rate of growth of income at selected points in the income distribution over several different periods.
Figure 1: Distribution of Income Gains, 1947-89
Percentile/years % Annual Increase (to nearest tenth)
20
1947-73 2.6%
1973-79 0.4
1979-89 -0.3
40
1947-73 2.7
1973-79 0.4
1979-89 0.3
60
1947-73 2.8
1973-79 0.7
1979-89 0.6
80
1947-73 2.7
1973-79 0.6
1979-89 1.1
95
1947-73 2.5
1973-79 1.1
1979-89 1.6
The income distribution is measured in percentiles. For example, the first set of bars shows the rate of growth of income of the family at the 20th percentile (the top of the bottom quintile). The choice of percentiles ranging from 20 to 95 means excluding the real extremes. Some very important developments are missed by these exclusions, especially at the top. But this picture still gives us a useful baseline.
The three periods chosen are 1947-73, 1973-79, and 1979-89. The first period represents what Alice Rivlin has called the "good years" the great postwar boom generation. The remaining two periods show the "seventies" the period from the business cycle peak of 1973 to that of 1979 and the "eighties" from the 1979 peak to the 1989 peak.
What do we see in the figure? First, the 1947-73 numbers show what real, broad-based prosperity looks like. Over that period incomes of all groups rose at roughly the same rapid clip, more than 2.5 percent annually. Between 1973 and 1979, as the economy was battered by slow productivity growth and oil shocks, income growth became both much slower and more uneven. Finally, a new pattern emerged after 1979: generally slower income growth, but in particular a strong tilt in the growth pattern, with incomes rising much faster at the top end of the distribution than in the middle, and actually declining at the bottom.
In some of the conservative critiques I will describe below, apologists claim that the 1980s represented a normal process, that there was nothing unusual or distressing about the rise in inequality. As the discussion gets a bit complicated, it will be useful to retain the basic image of Figure 1: "good" growth looks like an all-American picket fence; growth in the 1980s looked like a staircase, with the well-off on the top step.
The CBO Numbers
The Census numbers shown in Figure 1 tell a pretty clear story. Nonetheless, it has been apparent for some time that the story is incomplete, because it fails to give a full picture of gains among families with very high incomes.
Census numbers are of little use in studying high-income families, for two reasons, one major, one minor. The main problem is the arcane technical issue of "top-coding." The questionnaires on which the Current Population Survey is based do not ask for precise incomes; instead, families are asked to place their income within a series of categories, of which the highest is "over x," currently $250,000. This means, of course, that the Census data give no information about changes in the fortunes of families with incomes high enough to be above that top number. The minor problem is that Census data do not count one important source of income for high-income families: capital gains.
It is precisely because Census data are weak when it comes to very high incomes that those who use that data usually look no higher than the 95th percentile; that is, the bottom of the top 5 percent. Over the period 1947-73, when everyone's income went up at about the same rate, the weakness of Census data at the top end didn't matter much. But it became obvious during the 1980s that incomes were rising even faster among the very well off than at the 95th percentile.
One might have guessed this simply from Figure 1: Since the available data show that the higher you go in the income distribution, the bigger the gains, one might reasonably suppose that the same is true for the unavailable data. One might well expect to find that inequality within the top 5 percent has risen, implying larger gains at, say, the 99th percentile than at the 95th.
One could also guess that income was growing especially rapidly at the top from less formal evidence. Notably, Graef Crystal's executive compensation numbers suggested a tripling of the compensation of the compensation of CEOs relatively to ordinary workers, and virtually every social observer has noted an apparent explosion of affluence at the top. All that was lacking was hard statistical evidence.
Work by the Congressional Budget Office fills the gap. The CBO is charged by the House Ways and Means Committee with estimating changes in the incidence of federal taxation, to provide the supporting appendices for that committee's mammoth annual publication, the Green Book. To do this, the CBO has developed a model that pools Census and IRS data. This model allows the CBO to bypass the problem of top-coding, and also allows incorporation of taxable capital gains.
Figure 2 shows the CBO estimates for the gains in income at different parts of the income distribution over the period 1977-1989. (Ideally, we would use 1979-89. Unfortunately, for reasons having to do with its original mandate to focus on tax incidence, the CBO did not do an estimate for 1979.) Here, the data are presented a little differently from those in Figure 1. We are shown changes in, say, average income for families in the bottom quintile, rather than for the individual family at the top of that quintile, and the numbers show the percentage change over the period as a whole, rather than annual rates of change. But the picture is clear: there were truly huge income gains at the very top. In particular, the top 1 percent of families saw their incomes roughly double over a twelve-year period. That's a 6 percent rate of growth, which means that for the very well-off the 1980s really were a very good decade not only compared with the slow growth lower down in the distribution, but even compared with the postwar boom years.
Figure 2: Increases in Income, 1977-89
Percentile % increase, 1977-89
0-20 -9%
20-40 -2
60-80 8
80-90 13
90-95 18
95-99 24
100 103
There is one other important point to be learned from the CBO numbers: how well-off the well-off actually are. The usual story still told by conservatives is that the so-called "rich" are not really all that rich. Conservatives often point out that, according to Census numbers, in 1989 it required an income of only $59,550 to put a family in the top quintile, an income of only $98,963 to put it in the top 5 percent. The implication is that we are essentially a middle-class society, with only an insignificant handful of people rich enough to excite any concern about ill-gotten gains.
But the CBO numbers paint a different picture, because they let us look higher up the scale. According to the CBO, to be classified in the top 1 percent a family of four needed a pre-tax income (in 1993 dollars) of at least $330,000. The average income of four-person families in the top 1 percent was about $800,000. We are no longer talking about the middle class.
The "Krugman Calculation"
It is a remarkable fact that incomes have soared so much at the top of the U.S. income distribution. But is it important? Until recently, most economists thought not; growing poverty might be an important social issue, but the fact that some people are very rich was only a social curiosity.
My own contribution to this discussion was to point out that there is a sense in which the rise in incomes at the top is in fact a major economic issue, and to offer a shorthand way of conveying that point: the now infamous "Krugman calculation" that 70 percent of the rise in average family income has gone to the top 1 percent of families.
Let's begin by recalling that typical incomes grew very slowly during the 1980s. For example, even if one uses a revised consumer price index that shows lower inflation than the standard index, one finds that median family income the income of the family at the midpoint of the income distribution in 1989 was only 4.2 percent higher than in 1979. That is, median family income rose at only about a 0.4 percent annual rate. And many measures of real wages for typical workers show a decline during the 1980s.
Now one would have expected incomes in the U.S. to grow more slowly than in the good years before 1973, because of the productivity slowdown. Productivity growth in the U.S. economy fell from about 3 percent annually during the postwar boom to about 1 percent annually after 1973; and ordinarily productivity growth determines real income growth.
But although productivity growth is slow, it is not negligible. We are a substantially more productive country now than we were in 1979. So why isn't the typical family significantly better off? Where did the productivity growth go?
The proximate answer is that average incomes went up relative to the medianincome. Figure 3 shows average versus median family income from 1979 to 1990. It turns out that from 1979 to 1989, average family income rose 11 percent, just about exactly what one would have expected given 1 percent productivity growth. So there is no problem with the accounting.
Figure 3: Average vs. Median Income, 1979=100
1979
Average 100
Median 100
1980
Average 97
Median 97
1981
Average 95
Median 94
1982
Average 95
Median 93
1983
Average 96
Median 94
1984
Average 99
Median 96
1985
Average 102
Median 97
1986
Average 106
Median 102
1987
Average 107
Median 103
1988
Average 108
Median 103
1989
Average 111
Median 104
1990
Average 108
Median 102
The rise in average income relative to median should not be a surprise, given Figures 1 and 2. That is exactly what one would expect to see when incomes become more unequal, because when incomes at the top of the scale are rising faster than the average, incomes farther down must correspondingly grow less rapidly than the average. In an arithmetic sense, we can say that most of the growth in productivity was "siphoned off" to high-income brackets, leaving little room for income growth lower down. I emphasize that this is only an arithmetic point: it says nothing about the economic forces at work, in particular whether something else could or should have happened.
When I say that growth was "siphoned off" to high-income families, however, who am I talking about? Are we talking about two married schoolteachers, whose $65,000 income is enough to put them into the top quintile? Or are we talking about Donald Trump?
Figure 2 ought to suggest to you that we are not talking about those schoolteachers: the really big income gains were not near the bottom of the top quintile, but at its top. Indeed, according to the CBO's numbers the share of after-tax income going to the ninth decile families between the 81st and 90th percentiles actually fell slightly between 1977 and 1989. So all of the siphoning went to families in the top 5 or 10 percent. And given Figure 2, one might well suspect that the bulk went to the top 1 percent.
To get a sense of this and, to be honest, to help attract attention to a trend that I thought had been neglected I proposed the following thought experiment. Imagine two villages, each composed of 100 families representing the percentiles of the family income distribution in a given year in particular, a 1977 village and a 1989 village. According to the CBO numbers, the total income of the 1989 village is about 10 percent higher than that of the 1977 village; but it is not true that the whole distribution is shifted up by 10 percent. Instead, the richest family in the 1989 village has twice the income of its counterpart in the 1977 village, while the bottom forty 1989 families actually have lower incomes than their 1977 counterparts.
Now ask: how much of the difference in the incomes of the two villages is accounted for by the difference in the incomes of the richest family? Equivalently, how much of the rise in average American family income went to the top 1 percent of families? By looking at this measure we get a sense of who was "siphoning off" the growth in average incomes, accounting for the fact that median income went up so little.
The answer is quite startling: 70 percent of the rise in average family income went to the top 1 percent.
What does this tell us? Since the 1970s median income has failed to keep up with average income or, to put it differently, the typical American family has seen little gain in spite of rising productivity. So when we speak of "high income" families, we mean really high income: not garden-variety yuppies, but Tom Wolfe's Masters of the Universe.
Wealth distribution. Wealth the assets that families own and income are different though related things. Wealth is typically much more concentrated than income: current estimates are that the 1 percent of families with the highest incomes receive about 12 percent of overall pretax income, while the wealthiest 1 percent of families has some 37 percent of net worth. Precisely because wealth is so concentrated, it is difficult to measure accurately from sample surveys: a random survey of a few hundred or even a few thousand people will contain only a handful of really wealthy people. Nonetheless, researchers at the Federal Reserve Board have tried to use sophisticated sampling procedures to deal with this problem. For some time their surveys have shown that average wealth was rising much faster than median as in the case of income distribution, a sure sign of growing inequality. In March, 1992 they released a working paper that showed a sharp increase in the concentration of wealth even since 1983, with the share of the top 1 percent of families rising from 31 to 37 percent.
Recently, several academic researchers (Claudia Goldin and Brad DeLong of Harvard, together with Edward Wolff of New York University) have put together long-range historical estimates on wealth distribution. They suggest that the concentration of wealth in the U.S. reached a trough in the late 1970s at a level not seen since the nineteenth century, then surged rapidly back to 1920s levels. The point is that the wealth numbers confirm the general picture of a dramatic and rapid increase in economic inequality in the U.S.
Political Implications
Rising inequality need not have any policy implications. Even if you would prefer to have a flatter distribution, other things equal (and not everyone even shares that goal) what should we do about it? Few people in America would currently support a policy of wage and salary controls (although Claudia Goldin has noted that World War II wage controls seem to have produced a long-term narrowing of wage differentials). One might use growing inequality as an argument for restoring some of the progressivity of the tax system; but most of the growth in inequality has come from changes in pre-tax income, not from regressive tax policies. An honest conservative like Herbert Stein is willing to say "Yes, inequality has increased, but I don't think that calls for any policy response."
Nonetheless, many conservatives were furious when the income distribution story surfaced in early 1992. Above all, the story made the editors of the Wall Street Journal and the Bush administration see red.
The reason was pretty clear. Supply-siders like Robert Bartley, the Journal's editorial page editor, believe that their ideology has been justified by what they perceive as the huge economic successes of the Reagan years. The suggestion that these years were not very successful for most people, that most of the gains went to a few well-off families, is a political body blow. And indeed the belated attention to inequality during the spring of 1992 clearly helped the Clinton campaign find a new focus and a new target for public anger: instead of blaming their woes on welfare queens in their Cadillacs, middle-class voters could be urged to blame government policies that favored the wealthy.
So the dismay and anger of conservatives was understandable. The response from the administration, the Journal, and other conservative voices was, however, inexcusable: instead of facing up to the fact of rapidly growing inequality under conservative rule, they tried to deny the facts and shoot the messengers.
The Conservative Response: 1 Denial
The Number of Families: When the New York Times published a story reporting my estimates on the impact of rising inequality, the initial response of a number of conservative economists (including staffers at the Council of Economic Advisers) was to do a different calculation: to ask what share of the growth in total rather than average income went to the top 1 percent. Let's call it the "CEA calculation." This is a very different number, because the number of families in the U.S. grew substantially between 1977 and 1989, as the last of the baby boomers grew up. So total income went up, not by 10 percent, but by about 35 percent. Naturally, the share of this much larger rise that accrued to the richest was a good deal smaller, 25 versus 70 percent. This revised number was widely circulated in Washington as a refutation of the number published in the New York Times; indeed, I have been told that one major news magazine almost ran a gleeful story on the Times's blunder, but at the last minute was warned that I was right and the CEA was wrong.
What's wrong with the CEA calculation? Remember the questions we are trying to answer: why didn't the typical American family see much increase in income even though productivity rose substantially, and who was reaping the benefits of rising productivity? If you think about it for a minute, you'll see that using income growth numbers that include sheer growth in working-age population gets us completely away from those questions. Consider, for example, what happened to the bottom 20 percent of the income distribution. Average income among these families fell 10 percent over the CBO period but their numbers went up about 25 percent, and their total income therefore rose about 15 percent. So the CEA calculation has the bottom quintile sharing in economic growth, even though average family income in that group went down!
The CEA also distributed a memo presenting a hypothetical numerical example, too complicated to reproduce here. Its point was that if the labor force were to receive a large influx of inexperienced workers, the experience of the median worker might decline; in that case, stagnation in median income might mask rising wages for a worker with any given degree of experience, and a "Krugman calculation" would erroneously suggest that only the very well-off had gained. The memo was right in principle. As anyone who has looked at labor force and wage data knows, however, the real facts look nothing like the contrived example: the growing inequality of wages represents increased dispersion in wages for workers with given characteristics, not a change in the mix.
The Size of Families: The next issue fits awkwardly into this scheme, since it involves an honest difference of opinion between myself and the CBO, and does not in the end make much difference. This is the question of how, if at all, to deal with the declining size of families in the U.S.
As noted above, the CBO likes to measure, not raw family income, but "adjusted" family income (AFI), measured in multiples of the poverty line. Adjusted family income has been rising faster than income itself, because families have been getting smaller. From 1977 to 1989, AFI grew by 15 percent compared with 10 percent for the raw number.
When you do a Krugman calculation using AFI instead of raw income, the result looks a little bit less extreme: the top 1 percent get 44 instead of 70 percent of the increase. This is still pretty impressive; but is the correction appropriate?
The CBO likes to use adjusted family income because they view it as a better measure of the material standard of living: a family with one child will be able to afford things that a family with three children and the same income cannot. Fair enough. But it seems to be stretching the usefulness of the concept when the decision of U.S. families to have fewer children is considered to be a form of income growth (what would the Republican platform committee say?).
Indeed, the number of hours worked by the typical American family actually rose during the 1980s. So if we are asking why family incomes didn't rise along with productivity, we should if anything be discounting the slight rise in income for the fact that families were working harder to get it. The CBO's adjustment goes in the opposite direction. Or to put it another way: the adjusted family income measure helps explain why so many families are able to afford VCRs, but misses the reason why they feel worse off than their parents.
All this is relatively minor, however. With or without the family size adjustment, the data confirm a radical shift of income to the top 1 percent.
Capital Gains: Many conservative commentators including Paul Craig Roberts, Alan Reynolds, Representative Richard Armey, and the editorial page of the Wall Street Journal have bitterly attacked the CBO for including capital gains in its estimates of income. They charge that this inclusion overstates the income of the rich in several ways: it includes one-time sales as if they were persistent income; it counts capital gains on assets held by the rich, but ignores the non-taxable gains of middle-class families on their houses; it counts as income the inflation component of capital gains. And all of these commentators have claimed that the CBO's capital gains estimates are the basis of the conclusion that the rich have done better than you or me.
There are answers to each of these criticisms: asset sales must take place sometime; capital gains on houses are much smaller than the critics imagine; the inflation component has fallen with the rate of inflation, so that if anything the rate of growth of income at the top is understated. The main point, however, is that excluding capital gains from the CBO numbers makes very little difference. With capital gains included, the CBO shows the share of income ac- cruing to the top 1 percent rising from 7 to 12 percent between 1977 and 1989, and shows this group receiving 44 percent of the rise in adjusted family income. Without capital gains, the shift is from 6 to 10 percent, and the share of the rise is 38 percent. Although the CBO does not report this, we can guess that a "Krugman calculation" excluding capital gains would still yield a number in excess of 60 percent. In other words, the capital gains issue is a complete red herring.
Can You Be Too Rich? When the Federal Reserve wealth study came out, it was immediately attacked by Alan Reynolds in the Wall Street Journal, as well as by Republican Congressman Richard Armey. Reynolds's main argument was that the study, based on a survey of 3,000 families, could not be reliable about the top 1 percent, since thirty families is too small a sample. This was an interesting reaction, since the Fed study carefully explains that they used a two-stage procedure and that their estimates were based on over 400 families in the top 1 percent. In fact, the study is written in the form of a working paper on statistical methodology, and the issue of sample size is raised immediately. One can only conclude that Reynolds did not bother to read the study before attacking it.
Rep. Armey, whose results were reported by Reynolds and Paul Craig Roberts in several columns, took a different tack. By careful search through a previous Fed study, he found what he took to be a significant fact: the average wealth of families with incomes above $50,000 rose more slowly over the period 1983-89 than overall average wealth. He claimed that this fact showed that wealth distribution had actually become more, not less equal. He apparently failed to notice that the size of the "over $50,000" group had increased over the period, from 17 to 20 percent of the population. Suppose that I told you that the average SAT scores of the top 20 percent of students today are lower than those of the top 17 percent a few years ago. Would you worry, or would you simply point out that the extra students I added to the sample obviously dragged down the average?
The wealth dispute was a minor part of the distribution controversy, but it was revealing about the desperation, unscrupulousness, and sheer lack of competence of today's conservatives.
The Conservative Response 2: Taking Credit For Growth
The second line of conservative defense has become a familiar one: they claim that the growth record of the Reagan years shows that supply-side policies produce gains for everyone, and that it is destructive to worry about or even to notice the distribution of income.
Look again at Figure 3. It is clear that from the recession year of 1982 to the business cycle peak in 1989, median income rose substantially (12.5 percent, versus 16.8 percent for average income). If you use these years as the basis of comparison, the lag of median behind average income doesn't look very important. The question is whether these are really the right years to compare.
If there is one really solid contribution of macroeconomic theory to human knowledge, it is the distinction between the business cycle and long-term growth. Long-term growth is achieved by expanding the economy's productive capacity; recessions and recoveries represent fluctuations in the degree to which that capacity is being utilized. It is a bad thing to be in a recession, and a good thing to recover, but one should never confuse the rapid growth that takes place during a recovery with an improvement in the economy's long-term performance: once the economy is near capacity, growth is bound to slow down. Moreover, recessions and recoveries depend far more on the Federal Reserve than on the administration in power, and happen to Republicans and Democrats alike. That is why a sensible assessment of economic trends involves comparing business cycle peaks or, even better, asking what has happened to the level of income associated with any given unemployment rate.
It is therefore ironic that supply-side ideologues, who originally crusaded against the traditional Keynesian focus on the business cycle, now rest their claims for success entirely on the business cycle recovery from 1982 to 1989. But of course they must, for their program failed to produce any acceleration in long-term growth.
The rise in median income from 1982 to 1989, Robert Bartley's "seven fat years," represented almost entirely a transitory business cycle recovery, which reached its inevitable limit at a level only 4 percent above the 1979 peak. And the subsequent recession, which is no more George Bush's fault than the 1980 slump was Jimmy Carter's, has probably dropped median income back to within less than 4 percent above the 1980 level.
The basic proposition that the "Krugman calculation" was meant to convey is that income inequality has been increasing so rapidly that most families have failed to get much benefit out of long-term growth. This proposition stands. One need not take seriously the efforts by supply-siders to chop the past fifteen years into little slices, and claim the good ones while disclaiming the bad ones.
The Conservative Response 3: Income Mobility
America is not a static society. People who have high incomes one year may have lower incomes the next, and vice versa. In the two hypothetical villages that I described earlier, one would not necessarily suppose that the same people (or their children) occupied the same positions in 1977 and 1989. And economic welfare depends more on the average income you earn over a long period than on your income in any given year. So there are some risks in drawing too many conclusions about the distribution of economic welfare from statistics on the distribution of income in any one year.
There are two ways in which income mobility the shuffling of the economic deck that takes place as families move up or down the income ranking could offset the proposition that inequality has increased sharply. First, if income mobility were very high, the degree of inequality in any given year would be unimportant, because the distribution of lifetime income would be very even. I think of this as the blender model: whatever the current position of the bubbles in your Mixmaster, over the course of a few minutes each bubble will on average be halfway up.
Second, if income mobility had increased over time, this could offset the increased inequality at each point in time. An increase in income mobility tends to make the distribution of lifetime income more equal, since those who are rich have nowhere to go but down, while those who are poor have nowhere to go but up.
Unfortunately, neither of these possibilities actually characterizes the U.S. economy. There is considerable income mobility in the U.S., but by no means enough to make the distribution of income irrelevant. For example, Census data show that 81.6 percent of those families who were in the bottom quintile of the income distribution in 1985 were still in that bottom quintile the next year; for the top quintile the fraction was 76.3 percent. Over longer time periods, there is more mixing, but still not that much. Studies by the Urban Institute and the U.S. Treasury have both found that about half of the families who start in either the top or the bottom quintile of the income distribution are still there after a decade, and that only 3 to 6 percent rise from bottom to top or fall from top to bottom.
Even this overstates income mobility, since (i) those who slip out of the top quintile (say) are typically at the bottom of that category, and (ii) much of the movement up and down represents fluctuations around a fairly fixed long-term distribution. Joel Slemrod of the University of Michigan has provided a useful indicator that suggests how persistent high incomes tend to be: the average income of families whose income exceeded $100,000 in 1983 was $176,000 in that year; their average income over the seven-year period ending in 1985 was $153,000.
Nor is there any indication that income mobility increased significantly during the 1980s. Table 1 shows some evidence from a study by Greg Duncan of the University of Michigan on transitions over a five-year period into and out of a somewhat arbitrary but reasonable definition of the "middle class". This middle-class category shrank in the 1980s, so that middle-class families became more likely both to rise and to fall; but correspondingly fewer poor families moved up or rich families down into the middle class. (Vanishingly few poor fami- lies became rich or vice versa). The overall picture suggests little change in mobility.
Table 1: Percentages of families making transitions from:
Before 1980 After 1980
Middle income to low income 8.5% 9.8%
Middle income to high income 5.8 6.8
Low income to middle income 35.1 24.6
High income to middle income 30.8 27.6
Income mobility might in principle be an important offset to the growth in inequality, but in practice it turns out that it isn't. That did not stop conservatives from trying to use it as a debating point.
The Hubbard Study
In June [1992] the Treasury's Office of Tax Analysis, under the direction of Glen Hubbard, an economist on leave from Columbia, released a report claiming that there is actually huge upward mobility in the U.S. In particular, it claimed that 86 percent of individuals who started in the bottom quintile in 1979 had moved out by 1988, and indeed that an individual who started in the bottom quintile was more likely to end up in the top quintile than to stay where he was.
But this report was based on what we may charitably call a strange procedure. Here's what Hubbard's report did: it tracked a group of individuals who paid income taxes in all ten years from 1979 to 1988, and compared their incomes not with each other but with those of the population at large. The restriction to individuals who paid taxes in all years immediately introduced a strong bias toward including only the economically successful; only about half of families paid income taxes in all ten years. This bias toward the successful was apparent in the fact that by the end of the sample period the group contained very few poor people and a lot of affluent ones: indeed, only 7 percent of the sample were in the bottom quintile by the sample's end, while 28 percent were in the top quintile. More important, by comparing the sample with the population at large rather than with each other, the report essentially treated the normal tendency of earnings to rise with age as representing social mobility. The median age of those whom the study classified as being in the bottom quintile in 1979 was only twenty-two.
Kevin Murphy, a labor economist at the University of Chicago, neatly summed up what the Treasury study had found: "This isn't your classic income mobility. This is the guy who works in the college bookstore and has a real job by his early thirties."
Income Gains
We have finally come to the last, and perhaps most effectively confusing, conservative argument.
Let's give the fact first: families who start out with high income on average have low or negative income growth over the next decade, while families who start out with low income on average see their incomes rise rapidly. This is true in both the Urban Institute and the Treasury data. In the Urban Institute's numbers, families in the bottom quintile in 1977 saw their income rise 77 percent by 1986, while families in the top quintile saw their income rise only 5 percent. The editorial page of the Wall Street Journal, Paul Craig Roberts, and others have seized upon this kind of number as evidence that the poor actually did better than the rich in the 1980s. Let me call this the "WSJ calculation."
The WSJ calculation seems striking; but on reflection it is completely consistent with the conclusion that the U.S. has rapidly growing inequality. It shows only that there is indeed some income mobility but nobody denied that. And it is no more a sign that supply-side policies helped the poor than the fact that very few people win the lottery several years in a row.
Unfortunately, it is hard to explain this without a numerical example: Imagine an economy in which in any given year half of the families earn $100,000 and the other half earn $200,000. And imagine also that this economy fits the blender model, so that a family that starts in the bottom half has a 50 percent chance of being in the top half ten years later, and conversely.
Now do the WSJ calculation. Families that start in the bottom half begin with $100,000; ten years later, on average they have $150,000, so they gain 50 percent. Families that start in the top half begin with $200,000; ten years later, on average they also have $150,000, so they lose 33 percent.
But has the distribution of income gotten more equal? No: it is unchanged. All that we see is the familiar statistical phenomenon of "regression toward the mean." Essentially, the initially rich have nowhere to go but down, the initially poor nowhere to go but up. So if the income distribution were stable, any income mobility would inevitably produce the WSJ result; and it is not surprising that we still get it even when income inequality is rising.
If income mobility were as high as in this example, of course, the income distribution at a point in time wouldn't matter very much. But as we have already seen, income mobility isn't that high: most poor or rich people stay that way. So we have enough income mobility to make the WSJ calculation seem right, but not enough to change the real story that inequality is rising.
If you want a more concrete image, think of it this way. In any given year, some of the people with low incomes are just having a bad year. They are workers on temporary layoff, small businessmen taking writeoffs, farmers hit by bad weather. These people will be doing much better in a few years, so that the average income of people who are currently low-income will rise a lot looking forward. But that does not mean that people who are persistently poor have rising incomes; they don't. Perhaps the most revealing way to show what is wrong with the WSJ calculation is to do it in reverse, as Isabel Sawhill of the Urban Institute did. In her data, families who were in the top quintile in 1977 had experienced an 11 percent fall in income by 1986. But when she instead looked at families who were in the top quintile in 1986, she found that they had experienced a 65 percent gain! Conservatives like to emphasize income mobility, because they can evoke the historical image of America as a land of opportunity, an image that has always been partially if not completely true. But when all is said and done, mobility in the 1980s was neither increasing, nor high enough to make any difference to the overwhelming picture of growing inequality.
The growth in income inequality in the United States since the 1970s is hardly an inconspicuous part of the economic landscape. On the contrary, it is apparent in virtually every economic statistic, and colors nearly everything about our national life. You may accept this trend or deplore it, but one might have thought that nobody could seriously deny it.
The surprise lesson of the income distribution controversy, then, is what it says about today's conservative mind-set. It turns out that many conservatives, for all their anti-totalitarian rhetoric, have Orwellian instincts: if the record doesn't say what you wish it did, hide it or fudge it.
There are substantive issues about income distribution. Nobody really knows all the reasons why incomes at the top have soared while those at the bottom have plunged. Still less is there a consensus about what kinds of policies might limit or reverse the trend. But it seems that many conservatives not only don't want to discuss substance: they prefer not to face reality, and to live in a fantasy world in which the 1980s turned out the way they were supposed to, not the way they did.
140 COMMENTS