Is Wealth Inequality Rising? (On Paul Krugman's Ambiguity Denial)
This article originally appeared in Forbes.
Paul Krugman has a typically fair-minded and even-handed column on “inequality denial” in the New York Times this week.
The “usual suspects,” he says, “continue to deny the obvious,” that is, “the sharp rise in U.S. inequality.” “There are so many independent indicators pointing to sharply rising inequality,” he says, “that any claim that inequality isn’t rising almost has to be based on faulty data analysis.”
The certainty of Krugman in his claims, shared by Thomas Piketty, and other economists I’ll discuss below, stands in contrast to the arguments made by actual inequality moderates not living in Krugman’s head. While Krugman is sure that “inequality” has sharply risen, inequality moderates are open to the idea that some forms of inequality—say, income concentration at the top—have risen while others—say, the ratio of annual earnings between middle-class and low-earning male workers—have not. (If, that is, you can believe University of Wisconsin head and economist Rebecca Blank, who was in charge of the Census Bureau at the start of Obama’s presidency.)
Inequality moderates do not assume that trends in annual earnings inequality must follow, say, trends in wealth inequality. We can concede that the ratio of household income between the middle and bottom fifth rose during, say, the 1980s, but then we might emphasize that it has not grown much or at all since then. (If, that is, you can believe the nonpartisan Congressional Budget Office.)We can ask whether the most widely cited estimates of income concentration overstate a smaller rise in inequality.
In short, what distinguishes inequality moderates from Paul Krugman and other inequality alarmists is an open-mindedness that is willing to tolerate the messiness and ambiguity of empirical data. That open-mindedness extends to other important questions around inequality. While willing to consider new evidence, we find the existing research on the supposed harms done by inequality weak, mixed, or flat-out unsupportive. We question whether there’s any basis for deeming inequality “the defining challenge of our time.” (That link is to the eminently dispassionate sociologist Lane Kenworthy, a self-described Social Democrat.) We take into account economic mobility and the living standards of the middle class and poor in evaluating whether rising gains at the top are worrying. We acknowledge that while reducing inequality could have important benefits in the short-run, it might have substantial costs in the long-run. And, it is true, we do not instinctively deem mind-blowing compensation at the top to be undeserved.
Tomorrow, the Manhattan Institute’s Economics21 research center, where I am a senior fellow, will release an inequality primer that reflects the nuance and moderation that many of us strive to bring to inequality debates. It features brief essays exploring the ambiguity in the most widely-cited income concentration estimates, on economic mobility (generally and in and out of the top), exploring the lack of strong evidence that rising inequality has had great costs, on the importance of demographics in interpreting inequality trends, on trends in wages and productivity, on risk-taking, and on the differences between consumption and income inequality. It includes think tank, academic, and government economists. We will all be declared conservative by folks like Krugman for simply embracing the ambiguity and contradictions of the available evidence.
It’s worth assessing who the close-minded and anti-empirical in inequality debates really are, and the Piketty-FT kerfuffle over wealth inequality trends provides an opportunity. (To review: the FT questioned the analysis of Piketty, whose Capital in the Twenty-First Century has single-handedly put inequality back at the center of policy and political debate in Washington.)
Consider the chart (below) from my last column assessing Piketty’s analyses. The chart includes the existing estimates of the share of wealth received by the top ten percent and top one percent since 1970. It includes not only the Piketty estimates but the ones from the sources on which he drew. Those include Federal Reserve Board estimates from its Survey of Consumer Finances (extended through 2010 for the top ten percent share). They also include New York University economist Edward Wolff’s estimates from the same survey, also extended to 2010. In addition, the chart shows the series from Wojciech Kopczuk and Piketty collaborator Emmanuel Saez, based on estate tax data, which only goes through 2000. Finally, I’ve added a preliminary series that came out after the publication of Piketty’s book, to which he has pointed in defending his own estimates. That series comes from two Piketty collaborators, Gabriel Zucman and Emmanuel Saez, and I’ve eyeballed their PowerPoint charts to add them below.
If we start with the top ten percent, Zucman and Saez find that wealth concentration began rising after 1986. The SCF studies by the Fed’s economists and by Wolff are in rough agreement, and they both agree with Zucman and Saez. The top ten percent’s share of wealth rose by six to eight percentage points between 1989 and 2010. It should be noted, though, that Wolff finds this increase only began after 2004, and the Fed economists find only a small increase before then. It is possible that the increase in the top ten percent’s share is only a decade-old trend.
What about the top one percent? Zucman and Saez report a strong increase in wealth inequality, with the top one percent share rising from 28 to 38 percent between 1970 and 2010 and from 25 to 38 percent between 1980 and 2010. (While not apparent in the chart above, they actually show an increase of about 16 points between 1978 and 2012.) In contrast, the Kopczuk-Saez estimates show the top one percent’s share falling from 23 to 21 percent over the 28 years between 1972 and 2000 (compared with an increase from 27 or 28 percent to 34 percent in the Zucman-Saez data).
While Zucman and Saez find a seven-point increase in wealth inequality between 1982 and 2000, Kopczuk and Saez report only a two-point increase. Both sets of estimates agree that wealth inequality in 1990 looked more or less like it did in 1970, but the Zucman and Saez results show that by 1990 wealth inequality had been rising for 12 years, while Kopczuk and Saez find it had barely budged over that interval. And in contrast to their top-one-percent estimates, Zucman and Saez find a decline in the top ten percent’s share through 1986.
Using the SCF, the Fed analysts find the top one percent’s share rising from 30 to 34 percent between 1989 and 2007—four points in 18 years compared with the seven-point increase found by Zucman and Saez. Wolff, using the same SCF data but somewhat different methods and definitions of wealth, finds a decline from 37 to 35 percent over the same period. And he finds no change between 2007 and 2010.
There are three different non-comparable sets of estimates here—estate-tax-based figures from Kopczuk and Saez, estimates from Zucman and Saez based primarily on capital income data, and numbers from the SCF (which is a survey of households). But the trends produced by each set of estimates ought to tell the same story if they are all valid. For the top one percent share, they do not (even the SCF analyses do not agree with each other).
Piketty, in his response to the FT, calls the Zucman-Saez trend “more reliable” than his own and says it “should be used as the reference series for wealth inequality in the United States.” In a forthcoming reference article he and Zucman do just that.
Zucman certainly believes that the new approach is superior to the others, but since their figures are preliminary and there is no paper written up, it is difficult to evaluate that claim. Having seen Saez present the new results, he seems to me inclined to throw out his results with Kopczuk, though they were published in a top economics journal. Kopczuk’s comments on Twitter indicate that he is decidedly not ready to throw out his results.
It should be reiterated that while Piketty, Zucman, and many journalists and liberal pundits seem ready to embrace the new Zucman-Saez results as the final word, there is, as of yet, no academic paper describing what they did to obtain their results. Up to last year, the Kopczuk-Saez paper was the most definitive work on wealth inequality trends. It was the equivalent of the Piketty-Saez income concentration series. Who is to say the charts in Zucman’s PowerPoint will hold up as long?
Perhaps they will, but for Piketty or Zucman to claim they are more reliable than the estate-tax-based estimates in Capital is not only premature, it calls into question all of the wealth inequality trends Piketty presents, which are based primarily on estate tax data. Piketty can be right that we know wealth inequality is rising in the U.S., or he can be right that we know his European trends to be correct, but it seems to me he can’t be right in both instances (and I’d venture he’s wrong about the certainty of our knowledge in any of these countries).
Zucman’s criticisms of the SCF estimates on Twitter have not, to my mind, been all that compelling. He believes that the SCF does not interview enough of the top one percent to produce valid estimates of its share, though the Kopczuk and Saez paper says that the survey interviews enough people to estimate the share of wealth owned by the top one-half of one percent. Zucman says that the SCF estimates are too imprecise to detect increases in wealth inequality, but if the true increase from 1989 to 2010 in the top one percent’s share were ten points, as suggested by the Zucman-Saez figures, the SCF would be exceedingly unlikely to indicate a two-point drop as it does in Wolff’s figures. (For the wonks, the Fed researchers have estimated that the top-one-percent-share estimates have a standard error of 1.5 points.) The problem here is that it’s very hard to measure wealth, especially at the top, and we really don’t know who is doing it better.
Krugman has not only said that it “can’t be right” that wealth inequality has not risen, he claims that any finding that it has not is an indicator that the analyst is “doing something wrong.” He cites Zucman-Saez, but also Congressional Budget Office data showing that capital income inequality rose between 1979 and 2007. How can the income derived from wealth grow more unequal unless wealth itself is growing more unequal?
It should be noted first, that Krugman’s CBO chart is only for capital income excluding capital gains, so in theory inequality in capital gains could have declined. CBO says it did not, but Richard Burkhauser’s research throws that finding seriously into question. In income tax return data, taxable capital gains are counted all at once in the year they are realized. Gains from a business purchased in 1987 and sold in 2007, reflecting twenty years of annual gains that accrued and added to wealth over time, show up as income received in 2007, even if most of those gains actually came, say, in the late 1980s and 1990s. Similarly, income from 401(k)-type retirement plans shows up when someone receives a distribution from the plan. If that distribution comes in retirement, income that may have accrued over decades shows up as having been received only in the tax year. For that matter, the income tax data omit non-taxable gains, including most of the gains earned by middle-class households (for instance, through the sale of a home). When the data exaggerates annual capital gains at the top and omits annual capital gains below the top, that’s a problem for inequality measurement.
Burkhauser and his colleagues find that when capital gains are measured appropriately, as best they can, household income concentration actually fell between 1989 and 2007 (the period in which Wolff’s SCF analyses suggest wealth concentration fell). As I have said elsewhere, this study is by no means the final word on the issue, but it should be a four-alarm warning that income concentration estimates based on tax return data may have serious shortcomings.
At any rate, Krugman should call out Edward Wolff publicly for “doing something wrong” if he really thinks that analyses showing flat to declining wealth inequality must be flawed.
Atif Mian and Amir Sufi, economists at Princeton and the University of Chicago, also should consult with Wolff. They say “It seems crazy to have such a heated debate on facts that are so easily measurable,” before using the SCF to show that the ratio of the top twenty percent’s wealth to the middle twenty percent’s rose strongly between 1998 and 2010. Wolff’s results confirm their finding, yet he also finds the top one percent’s share of wealth falling from 38 percent to 35 percent. It is eminently plausible that gains to the top twenty percent accrued entirely to households below the top one percent.
Krugman, Piketty, and Zucman recently have become big fans of the Forbes400 list as evidence of increases in wealth inequality, though there are well recognized challenges in compiling these estimates too. Arguably, this group’s share of wealth is too sensitive for trend estimates of national wealth inequality due to mis-reporting, evasion and avoidance, and the exclusivity of the group. Even if wealth has become more concentrated among the top 400 Americans, that is no indicator that wealth concentration has become more concentrated among the top 3 million Americans (i.e., the top one percent).
At this point, the fair-minded take-away is that we don’t know whether or not wealth concentration in the hands of the top one percent has risen, though it is safe to say that it has risen in the rest of the top ten percent. Zucman and Saez may prove right that the top one percent’s share has increased, but even in that case their estimates may overstate the rise. If you think I am reaching a tortured conclusion, again consider that the most careful analysis of wealth inequality trends up until 2014 (the Kopczuk-Saez study) showed wealth inequality flat to declining through 2000. It received a small fraction of the attention that the Zucman-Saez results are getting, for reasons that are useful to ponder. When I searched the New York Times website for the search terms “Krugman” and “Kopczuk,” here’s what Google produced:
Scott Winship is the Walter B. Wriston Fellow at the Manhattan Institute for Policy Research. You can follow him on Twitter here.
Interested in real economic insights? Want to stay ahead of the competition? Each weekday morning, e21 delivers a short email that includes e21 exclusive commentaries and the latest market news and updates from Washington. Sign up for the e21 Morning eBrief.