Did Inequality Rob Middle-Class Households Of $18,000?

By Scott Winship.

"Part 1 in the series, “Inequality and the Middle Class.” Also available: Part 2.

Last week, the Economic Policy Institute released a report on trends in wages, earnings, and incomes called, “Wage Stagnation in Nine Charts.” The report, by EPI president Larry Mishel, Elise Gould, and Josh Bivens is a greatest-hits compilation of EPI’s attempts to show that the American middle class is in trouble because of rising inequality. It arrived just in time for Congressman Chris Van Hollen to appropriate one of the charts for a presentation at the Center for American Progress laying out a new Democratic agenda to reduce income disparities.

The left appears ready to bank its 2016 hopes on the idea that there is a “middle class squeeze” caused by inequality, though at least one potential Democratic nominee for president will likely find it difficult to inveigh against “the top one percent.” The next two years are sure to feature myriad claims that rising income inequality has hurt the middle class, necessitating liberal policies to pull down the top. Invariably, these claims convey overly negative impressions of how the middle class is doing, omit important contextual details, exaggerate the extent to which inequality has risen, or otherwise present an inaccurate case that inequality is the source of all our troubles.

In the interest of clearing up some complicated data issues—and in some cases, in the interest of usefully complicating some apparently clear results—I intend to devote time this winter to a series that will critique these claims from EPI, CAP, and other sources.


Some of my critiques will benefit from EPI’s commendable practice of making its data available on its website. In general, EPI’s State of Working America website is often a great place to check first if you are looking for trend data on wages, income, or numerous other economic topics. Unfortunately, using data is often far from straightforward and requires specialized knowledge of datasets, measurement issues, and even history. It’s my hope that the series will contribute to making this knowledge a little less specialized.

The first chart in EPI’s report is presented under the headline, “The cost of inequality to middle-class households.” It is intended to show how much greater the incomes of middle-class households would be if inequality had not risen, an income loss EPI dubs “the inequality tax.” I’ve included the chart below, which is based on EPI’s analysis of income data from the Congressional Budget Office. The chart shows that in 2007, the average American in the middle 60 percent of household income had $17,867 less in household income than it would have “because of the growth of inequality.” That’s an inequality tax of around 20 percent.

Have you been robbed?
Calling for Inequality Reduction without Counting Inequality Reduction

To construct the “Income with no change in inequality” trend—the dashed line in the chart—EPI assumed that the average household income within the middle three-fifths of Americans grew at the same rate as the average income across all households between 1979 and 2007. The increase in the average for all households (67 percent) was higher than it was for the middle class (35 percent) because the dramatic income gains at the top have had the effect of pulling the overall average more sharply upwards. If income had grown at the same rate among the bottom fifth, the middle three-fifths, and the top fifth of Americans, that would have left the shares of aggregate income received by each group at their 1979 levels.

The EPI estimate is as high as it is partly because it is based on CBO’s pre-tax income figures. When social critics decry inequality, they generally advocate an increase in top tax rates to redistribute income. Thomas Piketty, for instance, in his surprise best-seller, Capital in the Twenty-First Century, advocated a top tax rate of 80 percent on all income over $500,000.

Let us keep assuming for now that middle-class households in 2007 were worse off by $17,867 before taxes as a consequence of rising pre-tax inequality. In that case, it would have been possible in principle to lower middle-class taxes and to raise taxes at the top by enough that after-tax inequality remained at its 1979 levels, wiping out EPI’s inequality tax. The point is that what middle-class Americans probably care most about is whether after-tax incomes are lower today because of rising inequality. The claim that the “middle class had $17,867 less income in 2007” because inequality rose loses its force if redistribution whittles that amount down significantly.

Conducting the EPI exercise using CBO’s after-tax income figures, the inequality tax falls to $12,802. In other words, through actual redistributive taxes, we manage to reduce the supposed “inequality tax” on middle-class households by 28 percent. Still, an additional thirteen grand under the mattress would be great, right? (I’m going to hold you to that later in the series when we get into the question of how robust middle-class income growth has been.)

The Size of the Pie

Well, you probably wouldn’t have seen that extra $12,802. EPI’s analysis has a hidden assumption—that aggregate income growth would have been no slower if we had prevented the top from pulling away. If holding the top fifth at its 1979 share of income would have slowed overall growth, then the middle class would have ended up with a bigger share in 2007 than what it actually got, but it would have come out of a smaller pile of money. If growth would have slowed enough, the middle class would have ended up worse off in 2007 for thwarting inequality.

Many social critics make EPI’s assumption—that the American economy is essentially a pie of fixed size, so that gains at the top must come at the expense of others. But higher inequality—at least to a point—can enlarge the economic pie, so that even if the slice enjoyed by the middle class becomes skinnier, the middle can still end up with more pie.

For the top fifth of Americans to have been kept to its 1979 share of pre-tax household income in 2007, it would have been necessary to hold its average income to a level 20 percent lower than its actual 2007 average. Since people in the 80^th through 90^th percentiles actually saw their share of income fall between 1979 and 2007 and those in the 90^th through 95^th percentiles saw essentially no change in their share, that 20 percent drop would have to have come entirely from the top five percent. That would have required that their incomes end up 39 percent lower on average in 2007 than what the top five percent actually received.

Would those who made it to the top have worked as hard, taken as many innovation-promoting risks, or invested as much if we had taxed them enough to reduce their payoff by 39 percent? Would we have been able to attract the best people to the most important positions in executive suites, in Silicon Valley, and on Wall Street? If not, then economic growth might have lagged, and the middle class might have suffered because of it.

Of course, it is also possible that reducing inequality would have increased growth, as Congressman Van Hollen maintained in his presentation, in which case the inequality tax would have been greater than $13,000. In a recent paper on cross-national differences in inequality and living standards, I concisely review the sizable and inconclusive literature on inequality and economic growth. If anything, the research offers more support for the conclusion that inequality strengthens growth than the claim that it hurts growth, at least in developed countries, but the effect is probably small either way.

However, “economic growth” is not the only consideration here. Much of the rise in income at the top in recent decades has come from realizations of capital gains. In the CBO data, 27 percent of the increase in the average pre-tax income of the top five percent came from realized capital gains. Capital gains are not counted as income in our national accounts because they do not represent income from new production; therefore, they are not included in GDP. Reducing the magnitude of capital gains realizations or taxing them away would not have affected measured economic growth directly.

That means that even if GDP growth would not have been harmed by successful efforts to hold inequality in check, income growth could still have diminished if it were defined to include realizations of capital gains (as in the CBO data, where taxable realized gains are counted as income). If so, then EPI is overstating the inequality tax because 2007 aggregate income in a world with fewer and smaller capital gains realizations would not have been as large as it actually was that year.

We can use the CBO data to see how sensitive EPI’s estimated inequality tax is to its assumption that income growth would not have been harmed by reining in inequality. Imagine, for instance that, as a consequence of capping inequality in 1979 and holding the shares of different income groups fixed, total after-tax income in 2007 (including realized capital gains) would have been 6 percent lower than the actual 2007 income we saw. Equivalently, we are imagining that capping inequality would have reduced the annual rate of income growth by 12 percent, or 0.23 points. In that case, the middle three-fifths of Americans would have been worse off for the increase in inequality by only $8,308 rather than $12,802. That’s 54 percent smaller than EPI’s headline estimate. (It would have taken a reduction of 0.68 points in the annual income growth rate to reduce the inequality tax to $0.)

Getting Real

Thought experiments may or may not be useful as a guide to policy. EPI’s estimation of how much better off the middle class would be if incomes had grown at the same rate across the board (and economic growth had not suffered) is illustrative, but how relevant is it? As noted above, even Americans between the 80^th and 90^th percentiles would have had higher incomes in 2007 if their share had remained at its 1979 level. The income share between the 90^th and 95^th percentiles was unchanged. So EPI’s thought experiment amounts to asking how much better off the middle class would be if the income growth of the top five percent had diminished enough to “fund” the additional income growth in the bottom 90 percent necessary for keeping income shares fixed.

If we assume that overall income growth would not have been harmed (or helped), then in 2007 around $194,000 per household in the top five percent would have had to be redistributed to the bottom 90 percent to get back to the 1979 income shares. For the middle class to have maintained its 1979 share of income, it would have needed just under $13,000 per household in 2007. That could have been accomplished by transferring 76 percent of the proceeds from the top five percent’s per-household $194,000 to the middle three-fifths. Another 8 percent would have been required to prop up the bottom fifth, and 10 percent would have gone to households between the 80^th and 90^th percentiles. (The remaining 6 percent is a residual from the imprecision in multiplying quite a few numbers, many of them rounded in the CBO data.)

This is a very specific thought experiment that is fine as far as it goes, but as a practical matter, it is unlikely that the proceeds from slowing income growth in the top five percent—however achieved—would have been distributed in this fashion. Capping the gains at the top likely would have shifted a disproportionate amount of income to knowledge workers and professionals between the 80^th and 95^th percentiles rather than sending enough gains to the middle class to maintain its 1979 income share.

Consider the case of Mark Zuckerberg. The Facebook founder and CEO reportedly earned $2.3 billion in 2012 exercising stock options. Imagine he had never been granted those options or that they ended up being worth less than they were. In that event, the likely beneficiaries from Zuckerberg’s loss would have been not middle-class families, but Facebook’s other investors (who saw their own Facebook shares decline in value when Facebook issued the additional shares to compensate Zuckerberg). According to economist Edward Wolff, the stocks and mutual funds held by the wealthiest 10 percent of households amount to 91 percent of the total value of those securities. Zuckerberg’s money would have mostly gone into the pockets of the rest of the top fifth.

There has been very little research on whether lower inequality results in better-off middle class households, but my own look at the cross-national correlation indicates that developed countries with less pre-tax and -transfer inequality actually have poorer middle classes. (I review the few other studies that have been conducted in that paper too; the evidence is fairly inconclusive.)

According to the CBO data, if we had kept the pre-tax income share of the top five percent at its 1979 level but the resulting proceeds going to the middle class would have been half as large as EPI assumes, then the inequality tax in 2007 would have been $8,590. That assumes no change in income growth from keeping inequality capped. Do the same exercise with after-tax income and the middle-class inequality tax would have been $6,401.

What if capping inequality also would have reduced income growth? Now it’s time to put all the pieces together. Imagine if in 2007 we had successfully held the top five percent’s share of after-tax income at its 1979 level of 18 percent instead of letting it rise to 29 percent. Imagine the size of the 2007 economic pie would have shrunk by 6 percent as a result. Further, imagine that the proceeds to the middle class from reducing inequality would have been half the amount needed to maintain its 1979 income share, so that 38 percent of the proceeds would have gone to the middle 60 percent of Americans instead of 76 percent. That would still have left the middle-class share of income at 48 percent in 2007—higher than the 44 percent it actually saw.

One way that might have occurred is if people between the 80^th and 95^th percentiles would have added that 38 percent to the 10 percent of proceeds they would have gotten in the EPI scenario, so that about half the proceeds from reducing the top five percent’s share would have gone to the rest of the top fifth. If this scenario is what would have happened had we capped top incomes, then the middle class in 2007 was just $1,775 poorer than it would have been had income concentration in the top five percent not risen. Keep everything in this setup the same but assume that capping the top five percent’s income share would have reduced total income in 2007 by 8 percent instead of 6 percent and the inequality tax disappears entirely.

Beware Correlations

When EPI and others show modest income growth since 1979 or 2000 relative to the 1950s and 1960s, they rarely make much effort to show that the slowdown in income was caused by the rise in inequality. They simply contrast the low inequality and strong income growth of the Golden Age with the high-and-rising inequality and sluggish income growth we’ve experienced in recent decades. But these sorts of comparisons are easily abused. Other datasets show that while the slowdown in income growth below the top began in the 1970s—before the years EPI analyzes—incomes at the top did not take off until after 1980. Apparently, rising inequality after 1980 caused middle-class income growth to slow during the prior decade.

Or we can compare the post-1979 experience in the U.S. to other countries. The World Top Incomes Database shows that in a number of countries, average income within the bottom 90 percent grew much slower after 1979 than it did during the 1950s and 1960s. Comparing two periods—1950-69 and 1979-2007—the annual income growth rate fell by 2.8 points in the U.S. according to this data. (To be clear, this data—primarily because it is based on tax returns—understates recent income growth in the U.S. compared with household surveys.) The growth rate fell by nearly as much (2.5 points) in Denmark, where the top ten percent’s share of income was stable over the period.
The top ten percent’s share increased much less in France and Switzerland than in the U.S., but annual income growth rates for the bottom 90 percent fell by 4.0 points or more in both countries. The rate also fell by 3.2 points in Canada, by 1.4 points in the United Kingdom, and by 1.2 points in Norway. In part, the lower income growth across all of these countries simply reflects a slowdown in productivity growth, which was lower between 1979 and 2007 than it was between 1950 and 1969 in nearly all of our peer nations in Europe and the English-speaking world.

Productivity growth features in EPI’s second chart, which made it into Congressman Van Hollen’s presentation, and which I’ll address next. But before leaving this first chart, let me emphasize that apart from analytics there is a values question lurking behind EPI’s depiction of the inequality tax—should the incomes of the middle class have grown as fast as or faster than those of the rich? To answer the question, EPI needs some benchmark for determining what a just rate of income growth for the middle class would have been. EPI believes that middle class incomes should grow as fast as productivity and that they have not. I will take up the empirical claim in my assessment of EPI’s second chart, coming soon."