Does America Still Work?


Challenge, (1-800-541-6563) Volume 39, Number 3, May-June 1996.
Jeffrey Madrick's interview with Stephen J. Rose
The Truth About Social Mobility
Copyright 1996 by M.E. Sharpe, Inc.

Q. When did you first become interested in mobility, wages, and income growth?

A. As a graduate student in the mid-1970s. What we'd always been told was that income distribution was constant throughout the twentieth century. So it didn't seem to be an issue worth studying. But nonetheless, a lot of people talked about how different people lived. So in 1978, I put together the Social Stratification in the U.S. poster and fact book.

It was a very large poster that showed the distribution of income, occupations, and family status in one graphic. It was used by a lot of people to understand the basic data. It was sometimes called Sociology 101 on a poster.

Q. What did you discover?

A. When I did the first update in 1983, I found some interesting new data that showed for the first time that the middle class was shrinking. Robert Kuttner first used the term in August 1983 in Atlantic magazine. But he didn't have any numbers. In December 1983, Bruce Steinberg, who was then at Fortune, and I independently used CPS data to show polarizing incomes. Since then it has obviously been a very hot issue.

There was a movement from the middle to both extremes. Nineteen eighty-three was right after a fairly serious recession, remember. What I found was an 8 percent decline in the middle. Five and one-half percent went down to the lower income group and two and one-half percent went up.

Q. How did you define middle class?

A. Well, that becomes very interesting. If you go to a couple that's making $20,000 a year--let's say the husband's a sanitation worker; the wife just works part time; they have one kid; and they live in the city--they'd say they were middle class. Or they might say lower middle class. In the suburbs, if you have a dentist husband and a social worker wife and they make $120,000, and they have three cars and one kid in college and another in high school, they might define themselves as upper middle class. Back in 1983, I thought the right definition of middle class was family income between $15,000 and $50,000 a year. Today, it's about the equivalent of $25,000 to $75,000 a year.Q So in 1983 that's how you found the shrinking middle class. But then you came to another conclusion. You also found that taking a snapshot of America's income profile at any particular moment could be misleading.

A. Right. Everyone who's used longitudinal data--that is, when you follow the income movements of the same people over many years--has been surprised at the amount of volatility that exists.

Q. Volatility of what? Movement up and down?

A. Let's look at the number of people having a single bad year, which can be defined as being 25 percent below the average of the three adjoining years. To get 25 percent below an average, that means income would have to drop, for example, from $50,000 to $33,000 and then bounce back up to $50,000. That's how you get a 25 percent decline in the middle year from the three-year average. Ten percent of the population have that experience in any given year. Of the 10 percent that are having a bad year, two-thirds fall all the way to the bottom quintile. These people who are having the one bad year tend to jump back immediately the next year. So fully one-third of the bottom quintile in any given year are there temporarily.

Q. That can be misleading because it implies that they are climbing the mobility ladder when in fact they only had a bad year.

A. Right. If you just look at the stratification and movement forward from that year, you find very strong growth. This is a central problem with some social mobility studies.

Q. You mean that if you do a longitudinal study in too simple a fashion it will show a great deal of upward mobility almost automatically?

A. Yes. They might have been laid off, and then they jump right back. But that doesn't imply more social mobility.

Q. There have been a couple of studies about social mobility that suggested that there was considerable social mobility. Did they make this adjustment you're talking about?

A. No. Another problem of longitudinal data is the need to be very careful with the age group, because if you start with very young people, what you'll see is very strong growth simply as they enter the labor force and start their careers. Similarly, if you include older people--for instance, if you track people from 55 to 65--you will find that they will have retired or cut back a lot so their income will have declined. So if you don't choose your ages carefully, you'll bias the study by starting with young people in the bottom quintile and old people in the top quintile; then when you look at them ten years later, you will see that young people move up and people in the top quintile move down, and really all you've done is seen the entry of young people into the labor force full time and the exit, partial or full, of old people.

A 1992 Treasury study using IRS data argued that the poor did better than the rich in the 1980s. But they failed to account for these age effects and ended up with a bottom quintile with only 7 percent of the survey and a top quintile with 28 percent. Not surprisingly, they thought they found considerable mobility out of the bottom.

Q. Now, you tried to correct for these biases. How did do you go about doing that?

A. Well, I focused simply on the same age group and followed them over ten years. Specifically, I took a sample of people who were 24 to 48 in the 1960s and followed them until they were 34 to 58 in the 1970s. Between the 1970s and the 1980s, I followed another group with similar age characteristics. By starting at 24, you basically have everybody in the labor force who's going to get there, and so you're not seeing the movement into the labor force from school, which biases the data. And, similarly, by making the maximum age 48 to 58, you haven't run into the early retirement years. If you track people who age from 50 to 60, what you find is that 60 percent of them have lower earnings at 60 than they had at 50. That is, they started to cut back. I didn't find that up to age 58.

I do two other things to deal with the volatility and the problems of what statisticians call reversion to the mean. That's when you have a bad year and you tend to jump back. Or when you have a good year and you tend to fall back. What I do is stratify people into rich and poor, not on the basis of their early years, not on the basis of their late years, but on the basis of all ten years. So I take their ten-year average earnings or income. That determines who is in the top and who is in the bottom quintiles. The other thing I do is, I never use single years. All of my comparisons are actually three-year averages. I compare business cycle peaks in 1969, 1979, and 1989. I compare the average income of 1967, 1968, 1969 to the average of 1977, 1978, 1979. And in the 1980s I compare the average of 1977, 1978, 1979 to 1987, 1988, 1989. The basis of the study, then, is to compare the experience of the 1970s, which is actually 1967 to 1979, to the experience of the 1980s, which is 1977 to 1989. So, it's a reasonably fair test.

Q. Let's talk about some of the conclusions.

A. In terms of family income in the 1970s, what I found is that 21 percent of prime-age adults had lower incomes at the end of the decade versus the beginning. And in terms of male earners, 24 percent had lower earnings at the end of the decade compared to the beginning.

In the 1980s, instead of 21 percent being income losers you now had 33 percent being losers. A significant jump. In other words, one out of three prime-age adults had lower income at the end of the 1980s than they did at the beginning. And for male earners, 36 percent were losers in the 1980s versus 24 percent in the 1970s.

Q. So, for one out of three people, income actually declined in the 1980s. This helps us get around the problems with dealing with average wages only.

A. Correct. Because an average does not compare the same people, as I do here. This does compare the same people, and you can make a strong statement about how many people are individually better or worse off.

Q. Now, did this surprise you?

A. I didn't have a preconception. I think what this does is provide very strong evidence that things changed for the worse.

Q. Between the 1970s and the 1980s?

A. Yes. Income inequality did indeed go up and it became harder for people to succeed.

Q. How did you measure income in these studies?

A. There are two measures. One is family income in inflation-adjusted terms. And the other is individual annual earnings in inflation-adjusted terms. So one is a family measure, which includes earnings from all family members, property income, and transfer payments. The only money flows excluded are employer-provided health care and pension benefits.Q There's often an argument made that by not including corporate benefits in these measures we are underestimating the growth of income.

A. That's partially true. But I think most of the growth in corporate benefits occurred early on, when wages were also growing.

Q. You mean in the 1970s?

A. The 1960s. Benefits really started jumping early on. By about the mid-1970s the real value of the benefits didn't change that much, even though it cost corporations more as health care expenditures grew at a high rate. The inflation index for health care expenditures outpaced the consumer price index in general by a lot so employers had to pay more to get the same benefits.

Q. You also found a change in the social mobility between the two decades.

A. In the 1970s people on the points of the income scale had about the same chance to move ahead, even though they started off at different points. A person in the lowest quintile had as good a chance to move up as one in the second quintile. But in the 1980s two things are worth noting. One is that your chances of success became much better the higher up you started off. So instead of everyone basically moving up together, the people who had more education, who started out being higher earners moved ahead faster than those who had less education and started off at a lower point. That's number one. But, number two, even in the top quintile there were more losers in the 1980s than in the 1970s. That is, the jump from 21 to 33 percent for income and from 26 to 36 percent for male income was so great that it affected all people up and down the income scale, even though it affected those at the bottom of the scale more dramatically than those at the top.

So, for example, in the 1970s, of people in the bottom quintile, 33 percent were losers compared to 21 percent for society as a whole. In the top quintile, only 13 percent were losers over the course of the decade. If we look at the bottom quintile in the 1980s, however, 53 percent lost ground over the course of the decade. In other words, more than one out of every two in the bottom quintile at the beginning of the 1980s made less by the end of the decade. And for the top quintile, that ratio went up to 18 percent.

So what this showed is that the 1980s were worse for everybody. But it affected those at the lower end much more than those at the top.

Q. You've also done a lot of work on job stability.

A. Job stability is difficult to measure because it's measuring a multiyear phenomenon. Most researchers in this subject area have relied on a single question that the census bureau asks every four years. And that is employer tenure. How long have you been with the same company? Surprisingly, if we track the answers to those questions from 1973 to 1993, we don't see much change in the distribution. If you adjust for age and you adjust for education, there's very little change. Given how much discussion there is of downsizing, this result was not anticipated. But I would argue that this is not a good question because there is a lot of

recall bias.

Q. What do you mean by that?

A. We should remember that these are just surveys. We know from the census that some questions are answered better than others. For example, wage income is basically reported correctly. If we compare the answers of the census of all wages to what we know from the national product accounts numbers, they come within 1 percent of each other. With welfare, interest payments, rents, dividends, these are 50 percent underreported. That is, the answers in the census don't add up to the answers of the national accounts. Similarly, I would argue that a value-laden question, such as how long have you been with your employer, is the kind of question that people will get wrong. If you're 45, you don't want to admit that you've been on the job only one year.

Q. So how did you try to design a study of job stability?

A. The approach I used was again to look at ten years' worth of answers. Each year in the data set people were asked, _ _Is the job you have now the same job you've had for the previous twelve months?___ That's a kind of simpler question to answer. If you said you had a different job, they then asked what was the industry and occupation of your previous employment. So I summed up the answers for all ten years in the 1970s and all ten years in the 1980s. Strong stability was defined as changing jobs at most once in ten years. So to be strongly stable, you either never changed jobs or you changed once. Workers were medium stable if they changed two or three times in ten years. Workers were defined as unstable if they changed jobs in four or more years out of ten.

Let me say a few words about working women. The ten-year slices show that women worked much less than the single-year slices. In the 1970s women averaged only 880 hours of work a year, whereas prime-age men averaged slightly over 2,200 hours. In the 1980s, women averaged 1,250 hours a year, while men averaged 2,180. So women increased their participation dramatically, but they still remained very far behind men in terms of hours worked.

Q. So they were not really comparable?

A. Right. Because women have fewer hours, the only data on stability I looked at were for men. It turned out that in the 1970s, 67 percent had strong employment stability. Some two out of three changed jobs at most once. Only 12 percent had weak stability or changed jobs four or more times. In the 1980s, only 52 percent were strongly stable, compared to 67 percent in the 1970s. Those with weak job stability rose from 12 percent to 24 percent.

Q. That suggests that there has been a considerable rise in job instability in the 1980s.

A. Yes, this supports the notion that downsizing has increased and something has seriously changed in workplace relations.Q In general, then, would you conclude that social mobility has declined in America?

A. Yes, and it especially declined for lower-income individuals markedly in the 1980s. That is not to say that there wasn't a lot of short-term ups and downs. Nearly two-thirds of Americans made it into the top 40 percent of income in at least one year in the 1980s. Forty-two percent of Americans spent at least one year among the top 20 percent of earners. This may be why Americans are so unwilling to tax the well-off. But dropping into the bottom quintile was also common. I asked how many people would qualify at least once in ten years for the Earned Income Tax Credit--that is, their earnings were so low that annual income fell below $28,000 for a family with one child in a single year. The number came to 39 percent. That is, two out of five people would have made so little income that they would have qualified for the EITC at least once in the decade. In a given year, about 70 percent of EITC-eligible families have low ten-year incomes. The rest are having only a single bad year and are in effect using EITC as wage insurance.

Only a small group of people are in the top quintile year in and year out. According to my calculations, only 6 percent of Americans were in the top quintile in ten out of ten years. By the way, the cutoff point for the top quintile for a family is $75,000 to $80,000 of annual income. The break point for the two top quintiles is $60,000. The lowest quintile were those families that made $25,000 and less.

Q. Despite the life-cycle effect, one out of three actually made less money after ten years even as they aged and became more experienced at their work.

A. Correct. Let me conclude with one number that is particularly striking. Take young men who are more or less just starting out. These are typically the years in which their earnings grow the fastest on a percentage basis. It's the takeoff point in their careers. I am talking about men 24 to 28. Some 62 percent saw their incomes grow by more than 40 percent in the 1970s, and only 9 percent had declining earnings over time. But in the 1980s, 26 percent had lower incomes at the end of the decade--not 9 percent, but 26 percent. That's a threefold increase. And instead of 62 percent having gains of 40 percent or more, only 42 percent did.

Q. Sum up the state of American incomes for us.

A. Inequality is growing. People at the top are doing better than the rest. And a higher proportion of people are actually losing ground. Finally, while we don't yet have the data, there is no reason to think that this situation has improved in the 1990s, and there are many reasons to think that the experience of the 1950s and 1960s was much better.

Challenge, the Magazine of Economic Affairs

Published by M.E. Sharpe, Inc., 80 Business Park Drive, Armonk, NY 10504. Phone: 914-273-1800. Toll Free Phone: 1-800-541-6563. Fax: 914-273-1800.

FRONTLINE / WGBH Educational Foundation
web site copyright 1995-2014 WGBH educational foundation