Share |

Has Income Inequality Really Risen?

Scott Winship | October 9, 2013

Few serious scholars believe that middle class and poor households have seen the income growth experienced by the top in recent decades. Both the ubiquitous estimates from economists Thomas Piketty and Emmanuel Saez and figures from the Congressional Budget Office show dramatic increases in the share of income received by the richest 1 percent of Americans. Between 1979 and 2007, the Piketty-Saez numbers rise from 10 percent to 24 percent, and the CBO share increases from 7 to 17 percent. Attempts to deny that the top has pulled away generally have been wholly unpersuasive and in more than a few cases conducted with minimal regard for the truth of the matter.

When conventional wisdom coheres around some accepted truth and most of the non-adherents are easily debunked, it becomes that much easier to casually dismiss any challenge as unserious and unimportant. However, a commitment to empiricism means not only refuting sketchy claims but taking seriously well-supported ones.

Scholars, journalists, and policymakers are now confronted with such a responsibility in the inequality research of Cornell economist Richard Burkhauser and his colleagues. If their latest research holds up—and there are good reasons to think that it will—we will have to rethink whether inequality has really increased since the 1980s.

It is difficult to overstate the potential importance of Burkhauser’s latest paper with regular coauthor and former student Jeff Larrimore and current student Phillip Armour. The study may force a rethinking of entire research literatures that presume inequality to be rising and that seek to explain it or to assess how it affects other outcomes. If inequality has not risen, that would also require policymakers to reassess the nature of various problems that are thought to be connected to it. If correct, their findings would affect debates over taxation, spending, deficits—even macroeconomic policy. The paper could even affect the political balance of power.

Burkhauser, Armour, and Larrimore walk the reader step-by-step through uncharted territory. They consider an income measure that accounts not only for shrinking household size, but for declining federal and state income and payroll taxes and for the rising importance of federal benefits and employer-provided health insurance. Using the Census Bureau’s Current Population Survey, they find that among the bottom fifth of Americans, household income rose just 4 percent from 1979 to 1989—both peak years in the business cycle. Income rose 12 percent for the middle fifth and 45 percent for the top 5 percent of Americans. Inequality grew. From 1989 to 2007, however (again comparing peak years), the bottom fifth, middle fifth, and top 5 percent saw gains of 26 percent, 20 percent, and 17 percent, respectively, indicating a decline in inequality.

A problem with these estimates, however, is that they do not include any income from capital gains. Capital gains derive from assets that appreciate in value, and since assets are especially unequally distributed, inequality estimates that leave capital gains out are potentially problematic. Starting in 1989, Burkhauser and his colleagues can incorporate capital gains into their income measure by imputing amounts to households using information from other datasets.

To facilitate comparability with the CBO and Piketty-Saez figures, they first add in capital gains that are both taxable and realized—that is, which result in taxable income directly received as a consequence of selling assets in a given year. After doing so, they show the top 5 percent pulling away from everyone else even from 1989 to 2007. The poorest fifth of Americans saw their incomes rise by 28 percent, the middle fifth by 22 percent, and the top 5 percent by 52 percent.

Even many economists are likely to be surprised by the finding that inequality between the poor and the middle class did not grow over these years. They should not be. Others have found the same result without taking into account taxes, health insurance, or capital gains (see Figures 3a and 3b of this paper by Christopher Jencks and colleagues, for instance). Rebecca Blank, current head of the University of Wisconsin-Madison and former acting Secretary of the U.S. Department of Commerce, wrote a book, Changing Inequality, that indicates little change in poor-versus-middle-class inequality over the same period either in terms of men’s hourly wages or men’s or women’s annual earnings.

CBO’s estimates confirm both the decline in poor-middle inequality and the disproportionate rise in income at the top. After applying the same cost-of-living adjustment as in the Burkhauser paper, CBO income figures show gains for the bottom fifth, middle fifth, and top 5 percent of Americans of 31 percent, 23 percent, and 81 percent from 1989 to 2007. While the CBO figures combine tax return and CPS data, Piketty and Saez rely entirely on tax return data. They ignore the bottom and middle fifth but show an 87 percent increase in the income of the top 5 percent of “tax units” (essentially, tax returns). CBO and Piketty and Saez both show bigger income gains for the top 1 percent (116 percent and 123 percent, respectively). The survey used by Burkhauser cannot reliably capture changes in income in the top 1 percent, but it is safe to say that if it could, it would show a rise similar to that in the CBO figures.

But this is not the last word either. The CBO and Piketty-Saez income figures are only able to account for capital gains that are both taxable and realized. Burkhauser, Armour, and Larrimore point out two big problems with this restriction. First, tax-exempt realized capital gains are ignored, including those from the sale of homes. These constitute a large share of capital gains received by the non-rich, so ignoring them overstates the rise in inequality. Another issue related to tax exemption is that savvy taxpayers at the top can alter their asset allocations so that more or fewer of their realized gains are taxable in response to tax law changes.

Second, and more important, there is a conceptual problem including realized capital gains in “income” but not the gains that accrue on assets that are held rather than sold. For one, the distinction is immaterial. Gains that accrue each year add to the resources available for consumption or saving whether they are realized or not. No less than realized gains, accrued gains not realized constitute part of the annual “flow” of resources properly conceived as “income” (as distinguished from the “stock” of accumulated resources properly thought of as “wealth”). In addition, investors strategically choose to realize capital gains depending on the state of asset markets and on changes in the tax treatment of different assets. Realization of gains accrued over many years tends to show up in tax return data in lumpy ways, as Cato Institute scholar Alan Reynolds has argued. A sizable share of the capital gains accruing to middle class households builds up over adulthood in accounts such as IRAs and 401(k)s and is not realized until retirement.

When Burkhauser and his colleagues incorporate into their income figures estimates of all capital gains accrued by a household over the past year—taxable or tax-exempt, realized or not—from investments in public companies and housing, a shocking result emerges. From 1989 to 2007, the incomes of the bottom and middle fifth rise (by 13 percent and 6 percent), but the income of the top 5 percent declines by 5 percent. Inequality—even between the top and everyone else—falls.

Burkhauser, Armour, and Larrimore also include estimates in their paper that account for accrued gains from investments in private companies, though they are less confident of these results. Nevertheless, those figures indicate that the bottom fifth’s income rose by 15 percent while the incomes of the middle fifth fell by under 1 percent and those of the top 5 percent by 25 percent.

When New York Times columnist Thomas Edsall highlighted the Burkhauser results in June, a chorus of big-name experts dismissed the results. Some pointed to the difficulty of valuing health insurance. Others criticized the imputation procedures used to assign accrued gains to households. Saez himself said that unless tax return data are used, nothing reliable can be said about top incomes.

Most interestingly, Edsall cited the work of Jeffrey P. Thompson and Timothy Smeeding, who used a similar approach in an earlier paper and concluded, “inequality measures peaked in 2007 at their highest levels in 20 years.” By incorporating results from an earlier draft of the Thompson-Smeeding paper, it is possible to compare the 1989 to 2007 change in inequality they found with the Burkhauser results. Instead of relying primarily on the Current Population Survey and supplementing it with results from the Survey of Consumer Finances, as Burkhauser and colleagues do, Thompson and Smeeding confine themselves to the latter survey, which interviews an expanded sample of rich households in order to produce reliable estimates for top earners. Consequently, they are able to give results for the 99th percentile of income—the household richer than all but the top 1 percent. Adjusting the estimates consistently for cost-of-living changes to make them comparable to the Burkhauser figures, the incomes of the 10th, 50th, and 99th percentiles rose by roughly 100 percent, 75 percent, and 190 percent. Inequality between the poor and middle once again declines, but the top pulls away from both groups.

However, to impute an accrued gain to assets held over the past year, Thompson and Smeeding use the average performance of asset markets over the previous three decades. This assumption is much stronger than any in the Burkhauser paper. It defies common sense—in assessing how much wealthier you are this year versus last, would you assume that your assets had the average rate of return since 1983 or the average rate in the past year?

Using results from the earlier Thompson-Smeeding draft, it is possible to estimate trends in inequality when accrued gains are imputed to assets based on just the preceding three years’ average returns, which brings the estimates closer to those of Burkhauser and his colleagues. These results indicate the same income increases for the 10th and 50th percentiles as when the 30-year average returns are used (about 100 percent and 75 percent), but switching to the three-year average returns lowers the increase in the 99th percentile to 140 percent (from 190 percent using the 30-year average returns).

These increases in income levels are much higher—across the board—than in the Burkhauser data, where no one sees much of an increase over the 18-year period. A big part of that must reflect the averaging of investment returns over three years so that losses after the bursting of the housing bubble are balanced out by gains in the preceding years. Had Thompson and Smeeding used just the preceding year’s average returns in estimating accrued gains, the discrepancy between their results and the Burkhauser ones would be narrower—quite possibly eliminated entirely. That is, using the preceding year’s average returns to estimate accrued gains would have captured the housing losses experienced in 2006 and 2007, producing a smaller rise in income (or a decline) from 1989 to 2007 for rich and poor alike. Whether or not that would have produced a decline in inequality depends on the extent to which using average returns over the previous year instead of the previous three years would have affected rich or poor more.

If Burkhauser and his colleagues had been able to choose, say, 2003 at their end year—which featured relatively high returns to both equity and housing—they might have found bigger increases in income for poor, middle class, and rich households, and they might have also found inequality higher than in 1989. Other starting and ending points might show other trends in inequality. But across the seven years for which they can estimate income, the share received by the top 5 percent is lower in 2001, 2004, and 2007 than in 1989, 1995, and 1998. The share is higher in 2004 and 2007 than in 1992, but the trend over the 18 years is clearly flat to declining. Nor is the 1989-to-2007 decline driven by the housing market crash; inequality declines when accrued gains from public investments alone are included. Once gains from public investments have been incorporated, adding accrued gains from housing investments has virtually no impact, year-by-year, on the share of income received by the top 5 percent.

The Thompson-Smeeding paper, for all its strengths, has other weaknesses. If Thompson and Smeeding had been able to account for the value of health insurance, that would also have narrowed inequality growth. And while they did not publish trends in their after-tax income measure, had they done so, inequality would have widened at least somewhat less than their pre-tax measures suggest. Finally, it seems pretty clear that Thompson and Smeeding double-counted realized capital gains unjustifiably. Though they claim that their approach “has a negligible effect on the results,” it would be surprising if that were true for income levels in each year, for trends, and for rich, middle class, and poor alike. After all, using 30-year average returns or using three-year average returns in estimating accrued gains leads to the same qualitative conclusion, though the magnitude of the increase in inequality differs quite a bit depending on which option is used.

Should the Burkhauser results be discounted because they cannot capture the incomes of the very rich, as Saez argues? This is surely a relevant question. Note, however, that the top 5 percent’s income growth taking only realized capital gains into account is eliminated by taking into account how much smaller total accrued gains were in 2007 than in 1989. At the very least, then, the income growth of the top 1 percent or the top 1 percent of the top 1 percent also would be expected to be significantly lower after accounting for accrued gains. Furthermore, Burkhauser’s imputation of accrued gains draws from the Survey of Consumer Finances, which is designed to give reliable estimates for the top 1 percent.

Is the Burkhauser research the final word on inequality trends? Is it perfect? Absolutely not (nor is the Thompson-Smeeding paper). Both papers, of necessity, use non-ideal imputation strategies to assign accrued gains to people, and neither fully accounts for all sources of income. Neither can tell us very reliably what happened to incomes at, say, the 99.9th percentile. There is little to suggest, however, that the ideal set of estimates would look qualitatively different from the Burkhauser results; they might, but there is no obvious way in which his estimates are simply a function of the decisions he and his coauthors made. For that matter, the tax-return-based estimates that constitute the basis for the conventional wisdom suffer from their own flaws, including some not mentioned above. (The Piketty-Saez estimates, for example, exclude most government transfers and all health insurance benefits from income, and they fail to account for declining taxes.)

Indeed, in one sense, even the Burkhauser results understate the extent to which the conventional wisdom may be wrong. Investment returns represent a reward for foregone consumption; investors make their income available to firms and other people at risk of getting back less than they initially committed. Arguably, the value of capital income from an investment (including accrued gains) should be discounted. Because of inflation, the initial amount invested (the foregone consumption) is worth less at the time returns are realized or accrue than it was at the time the investment was made. In fact, gains that accrue to that initial investment are also worth less over time if they are not realized. Failing to discount capital income makes it look more valuable to its recipients than it is, and since capital income is a bigger deal for the rich, it overstates inequality. Since capital income is a bigger deal over time, failing to discount it overstates any rise in inequality (or understates any decline).

That Burkhauser’s work offers up a radical refutation of the conventional wisdom and has dramatic policy and political implications are not reasons for rejecting it. A variety of approaches show that inequality between the bottom and middle has not grown since the 1980s. Those who read Burkhauser and remain certain the top 1 percent has really pulled away from “the 99 percent”—or who deem it unnecessary to engage his research—are not as empirically-minded as they believe.

 

Disclosure: In a previous position, I was research manager of the Pew Economic Mobility Project, which funded a paper on income inequality written by Burkhauser and one of his colleagues.

 

Scott Winship is the Walter B. Wriston Fellow at the Manhattan Institute for Policy Research


e21 PROJECTS & PARTNERSHIPS