Measuring the middle class: America's lesson
Despite impressions, the US middle class is doing better today than it was 40 years ago. This economic truth is being stymied by subjective and relative evaluations of people's wealth.
You could be forgiven for answering (d), given the angst-producing state of discourse on the economy, but the truth is that none of these claims about middle-class decline – made most recently by the Pew Research Center – are supported by the best evidence. Like other analyses before it, PRC’s recent report "The Lost Decade of the Middle Class” badly misrepresents the economic standing of the middle class. Pew generally does very good work, and I am a fan particularly of its Economic Mobility Project, the research portfolio of which I managed for three years. But 'declinist' analyses are both wrong and insidious, and since PRC’s work is especially influential, its errors deserve a careful examination.
The PRC report is cheerily subtitled Fewer, Poorer, Gloomier. Let’s take those claims up one by one.
PRC argues that the middle class has been shrinking over the near and the long term. The near-term claim relies in part on comparing the class designations Americans chose for themselves in surveys from early 2008 and again this past July. The share of Americans self-identifying as "middle class” declined from 53 per cent to 49 per cent over the four years.
It isn’t clear why we should care whether the middle class, so defined, shrinks. It could shrink despite an increase in the number of people who choose "middle class” rather than "lower middle class” or "lower class” if the number who choose "upper middle class” or "upper class” grows even more. What is important is whether the share choosing "lower middle class” or "lower class” grows.
PRC reports that it has done so, rising from 25 per cent in 2008 to 32 per cent this past July. Other pollsters asked this self-identification question in 1976, 1984, 1987, and 2004, and like PRC in 2008 they all found the lower-than-middle share to be between 24 per cent and 27 per cent. The question has been asked six times since 2010 (five times by PRC), and the share ranged from 29 per cent to 40 per cent in those surveys. The venerable General Social Survey has asked people to self-identify as lower, working, middle, or upper class since 1972, and it also found that the share of Americans self-identifying as lower than middle class was relatively high in 2008 and 2010, although the 2008 percentage (53) was lower than the one in 1972, and the 2010 percentage (55) was lower than the one in 1982. The share fell between 1996 and 2004, with the 2004 percentage (48) matching the General Social Survey’s 1989 low.
In short, historical data suggest that the Great Recession has increased the share of the population that considers itself lower than middle class but offers no evidence of a long-term increase or even a steady rise over the "lost decade". The same conclusion holds if we stick to the narrow question of whether the middle class has shrunk.
The PRC’s second set of analyses, purporting to show that the middle class has shrunk over the long run, are based on the Census Bureau’s Current Population Survey. PRC defines "middle income” in relation to the median income (the income of the household smack in the middle). A household is middle income, according to PRC, if its income is at least two-thirds of, but no more than twice, the median. By this definition, and adjusting incomes for inflation and differences in household size, middle-income households fell steadily from 61 per cent of the total in 1970 to 51 per cent in 2010.
When Alan Krueger, chairman of the White House Council of Economic Advisors, presented similar figures in January, he too asserted that the middle class had shrunk, claiming a decline from 50 per cent to 42 per cent between 1970 and 2010. But, as I argued in these pages ("The President’s Depressing Statistics", February 6), the reason the middle class "shrank” was almost entirely that Americans grew richer over time. There was no statistically discernible increase in the share of Americans poorer than middle class.
The PRC study, by contrast, did find an increase in the number of households below the middle-income boundary. In 1970, 25 per cent of adults were lower income, compared with 29 per cent in 2010. This finding was not particularly robust, however, when I replicated it. When I compared 1969 to 2007 – business-cycle peaks, both – the increase in the poorer-than-middle-class share was just 3 percentage points, and when I excluded Hispanics from the 2007 data (to control for the large influx of immigrants from Mexico over the period) the increase was just one point – not statistically meaningful. Nor is it clear that choosing different but still plausible boundaries for the middle class – say, half to 1.5 times the median, which is the definition that Krueger used – would result in the same conclusions. (PRC appears to have chosen its definition because it matches the share of adults who self-identified as middle class in July.)
But the biggest shortcoming of the PRC analysis is that, like the Krueger figures that preceded it, it conceives the middle class in relative terms: meaning that, as Americans become wealthier, the price of admission to the middle class increases over and above the increase in the cost of living. If inflation-adjusted median household income rises by 20 per cent, then the entry point to the middle class does, too. In the PRC study, it took $9500 more in income – after adjusting for increases in the cost of living – for a family of three to make it into the middle-income group in 2010 than it did in 1970. In 100 years, Americans will likely be much richer than they are today. Yet the share of the population calling itself middle class will probably remain large, because people will define themselves relative to their peers. If the share of Americans who consider themselves lower or lower middle class ends up somewhat higher than it is today but everyone is much better off in absolute terms, will that be a problem?
It is possible to defend PRC’s definition of the middle class by noting that relative status matters to people – one wants to "keep up with the Joneses.” Economist Robert Frank has catalogued studies showing that people’s happiness depends not just on their absolute level of material well-being but also on their standing compared with others. But it is far from clear that relative inequality matters as much to people as their absolute level of income, and in any case PRC makes no effort to defend its definition.
The definition had the odd consequence that "middle-income” adults became scarcer, and "lower-income” adults more prevalent, even though the median income within each group rose by around 30 per cent over the 40-year study period. When I reran the PRC analyses but kept the lower boundary of the middle class at its inflation-adjusted 1970 level, I found that the share of adults poorer than "middle income” declined from 25 per cent in 1970 to 18 per cent in 2007. That is, even as the fraction of American households making less than two-thirds of the current-year median held steady or rose, the share making less than two-thirds of the 1970 median fell significantly.
So much for "fewer". Is the middle class "poorer”? The PRC report uses the Current Population Survey to examine trends in median income. By viewing the past 60 years as a collection of six decades, PRC is able to describe the 2000s as a "lost decade". According to PRC, from 2000 to 2010, incomes fell among rich and poor alike for the first time since World War II. Median income in 2010 was at its 1997 level. Over a longer period, the report finds that income growth was strongest in the 1950s and 1960s, with the 1990s looking better than the 1970s and 1980s, and the Aughts bringing up the rear. There are three problems with these analyses.
First, the bulk of the evidence suggests that our measures of inflation overstate growth in the cost of living, which means that real-income growth is understated.
Second, the Current Population Survey understates median-income growth relative to data sources that measure income more comprehensively. In the PRC analyses, median household income rose 22 per cent from 1979 to 2007 (business-cycle peaks), but the CPS figures do not take into account non-cash transfers such as food stamps, Medicaid, and Medicare; employer-provided health insurance; or changes in taxes.
Congressional Budget Office estimates, which remedy those omissions and use a different cost-of-living index, find a 46 per cent increase in the median over the same period. That’s right: The typical household in 2007 was half-again richer than the typical household in 1979.
The third problem with the PRC analyses is the assumption that decades are appropriate periods by which to assess economic conditions. When rigorous economists consider income trends, they compare years that represent similar points in the business cycle: They do not compare a recession year to a boom year, for instance. As it turns out, the 1950s, 1960s, 1970s, 1980s, and 1990s can be roughly represented by comparing business-cycle peaks in 1948, 1960, 1969, 1979, 1989, and 2000. Looking at trends between 2000 and 2010, on the other hand, compares a boom year with a bust year; the proper comparison would be of 2000 and 2007.
Viewed this way, peak to peak, and using Congressional Budget Office income figures, the 2000-07 period saw faster annual median-income growth than the 1990s (1989-2000) or the 1980s (1979-89). CBO data do not go back farther than 1979, but census data show that the 1970s (1969-79) had slower household-income growth than the 1980s – so the period from 2000 to 2007 actually featured the fastest income growth since the 1960s. The Aughts were not a lost decade when measured this way. Rather, a period of solid gains was followed by a crash. And even so, the CBO figures indicate that median income was 12 per cent higher in 2009 than in 2000. Figures for 2010 are unavailable, but if they track the 2009-to-2010 trend in the CPS, the decade will show an increase of 10 per cent in median income rather than the 7 per cent decline PRC reports. So the middle class did not get "poorer".
To all of this, PRC researchers might suggest I’m asking people to believe me (to believe the CBO, actually) over their own lying eyes. After all, PRC’s survey respondents overwhelmingly agreed that "compared with ten years ago” it is harder for "middle-class people to maintain their standard of living". Fully 85 per cent endorsed the view that 2002 was better than 2012 in this regard. We have arrived at "gloomier".
Historical patterns show, however, that this widespread view does not reflect an objectively correct assessment of post-Great Recession American life so much as a persistently too-negative guess at how badly others are doing. In early 2008 – prior to the financial crisis – 79 per cent of those surveyed by PRC believed that things had gotten tougher since 2003. So the damage done by the Great Recession has little to do with the consensus that things are getting tougher for Americans; they were gloomy barely into it, well before the crisis. And before concluding that the strong majority taking a dim view was objectively correct in early 2008, consider that 65 per cent of respondents to an NBC/Wall Street Journal poll in 1986 said things had gotten tougher for the middle class over the preceding five years. That belief was almost surely not true, since the economy had double-dipped into a deep recession in 1981. The implication of these polls is that there is a general tendency for strong majorities of respondents to endorse the view that things are getting tougher, whether or not they actually are.
These figures, which PRC includes toward the back of its report, do suggest that the extent of economic pessimism increased between 1986 and 2008. But other polling evidence contradicts that conclusion. For more than 20 years, surveys have asked whether it has grown harder or easier over the past generation to attain "the American Dream.” According to polls by the Roper Center for Public Opinion Research, six in 10 Americans thought it had grown harder in 1990, as did six in 10 Americans in 1995. A 2010 Xavier University poll using a slightly different version of the question found this belief to be held by ... six in 10. A Xavier poll a year later, in 2011, saw the figure jump to seven in 10. That result was comparable, however, to what Roper found in 1992, another year in which the nation was recovering from recession.
More important, longstanding majority belief that it is getting harder to keep up should be considered along with another longstanding result of opinion polling on the economy: When asked about their own economic circumstances, Americans give much more optimistic responses than they do when asked about Americans in general. People’s impressions of how others are doing are more pessimistic than what they say about their own lives.
As an example, when I was at Pew, the Economic Mobility Project sponsored a poll independently of PRC in early 2009 asking about a host of mobility-related topics. We found that 73 per cent of Americans rated "economic conditions in this country today” as "poor". In contrast, just 25 per cent rated their "personal economic situation today” as "poor” when given the same response options. Three in four people said they were "very much” or "somewhat” in control of their personal economic situation, but only 43 per cent thought the same about "people in this country". Three in five parents thought their children would have a higher standard of living than they had, but just two in five non-parents thought the same would be true of "kids today".
This sort of result is pervasive in public-opinion research related to the economy. People feel good about their own situation but guess incorrectly that others must feel bad about theirs. Indeed, this "I’m okay, they’re not” syndrome – a phenomenon identified by the writer David Whitman – extends to non-economic matters as well. Parents like their kids’ schools but think schools in general are awful; voters like their representatives but think Congress is the pits.
PRC emphasises that 42 per cent of the middle class say they are "less financially secure” than 10 years ago. Forty-two per cent is obviously a large minority, but on the other hand this figure doesn’t tell us how many people feel significantly less secure. I feel less secure than 10 years ago, and I have some anxiety as a relatively new parent with a wife who is due to graduate into the labour force soon. But I can’t say that economic anxiety is anything like a central feature of my day-to-day existence.
The same is likely to be true of many other people who feel less secure than they did in 2002. Don’t take my word for it: According to PRC’s poll, 80 per cent of self-identified middle-class adults are "pretty” or "very” happy with their life, as are 77 per cent of all adults. Only 37 per cent of middle-class adults "frequently” experience stress (much of it presumably related to non-economic issues), and only 42 per cent of all adults. That is basically the same as in 1994, 2001, and 2002. Seventy-two per cent of the middle class and 64 per cent of all adults are "somewhat” or "very” satisfied with their personal financial situation. The corresponding satisfaction levels for respondents’ "present housing situation” are 90 per cent for the middle class and 86 per cent for everyone. In 1996, 87 per cent of adults were that satisfied. "Gloomier?” Sure – but we are barely clear of the worst recession in 70 years. "Gloomy?” Hard to see it.
Indeed, perhaps the biggest rejoinder to the PRC report’s "fewer, poorer, gloomier” conclusion is its own finding that, despite the Great Recession and molasses recovery, 57 per cent of all adults say their living standards exceed those of their parents at the same age, while just 17 per cent say they are worse off. The conventional wisdom is that they shouldn’t trust their lying eyes, and it’s true that in the aggregate the respondents’ assessment does not match the facts. In actuality, as Pew’s Economic Mobility Project has shown using 40 years of data on actual parents and children, 84 per cent of today’s adults are better off than their parents were at the same age.
Scott Winship if a fellow in economic studies at the Center on Children and Families. This story was first published at The Brookings Institution. Republished with permission.