clock menu more-arrow no yes mobile

A new study says much of the rise in inequality is an illusion. Should you believe it?

Everyone agrees inequality grew. The question is how much.

Human manifestations of income and wealth inequality, at a charity event in New York City.
Andrew Lichtenstein/Corbis via Getty Images
Dylan Matthews is a senior correspondent and head writer for Vox's Future Perfect section and has worked at Vox since 2014. He is particularly interested in global health and pandemic prevention, anti-poverty efforts, economic policy and theory, and conflicts about the right way to do philanthropy.

Few economics findings have penetrated the public consciousness in recent years as much as this one: Income inequality has exploded in recent decades, and the top 1 percent in particular have made out like bandits.

Economists Emmanuel Saez of UC Berkeley and Thomas Piketty of the Paris School of Economics have been documenting a massive rise in income inequality since 2003 using hyper-detailed IRS records. According to their latest data, compiled with Berkeley’s Gabriel Zucman, the top 1 percent’s share of national income, after taxes are taken into account, rose from 9.1 percent in 1979 to 15.7 percent in 2014.

It’s hard to overstate the influence of this line of research. It won Saez the John Bates Clark medal, America’s most prestigious prize for academic economists, made Piketty’s Capital in the 21st Century an international best-seller, and helped frame the debate over inequality coming out of Occupy Wall Street and the Obama White House’s proposals. President Obama’s budget director, Peter Orszag, wrote in 2009 that Saez’s work “had no small influence on the President’s Budget.”

But another paper released recently suggests the spike in inequality Piketty and Saez have documented is a dramatic overestimate.

Gerald Auten and David Splinter, economists at Congress’s Joint Committee on Taxation and the Treasury Department’s Office of Tax Analysis, used the same IRS tax data as Piketty, Saez, and Zucman. They found that the top 1 percent’s share of after-tax income rose from 8.4 percent in 1979 to 10.1 percent in 2015 — an increase less than a third as large.

What looks on paper like a big increase in inequality in the 1980s and onward, Auten and Splinter argue, is really just money being shuffled around in response to Ronald Reagan-era changes to tax law. In 1980, the top individual income tax rate was 69.13 percent; by 1989, it had fallen by more than half, to 28 percent.

In the 1960s and 1970s, companies usually reinvested their profits rather than giving raises to executives — the high tax rates meant those salaries would be largely taxed away. Reinvesting the money ultimately benefited shareholders in the company by increasing the company’s value, and benefiting shareholders means benefiting rich people. Owning corporate shares was much rarer for middle-class people in the ‘60s and ‘70s before the rise of 401(k)s and IRAs.

After the tax cuts, companies started directing more money to raises. Rather than exploding actual inequality, Auten and Splinter write, the Reagan tax changes mostly shifted money that used to go to rich people through stocks so that it instead went to rich people in the form of salary.

That looks like a big increase in the rich’s slice of the pie on paper, because the higher salaries show up on tax returns, but the increasing value to shareholders doesn’t, at least until the shares are sold.

Auten and Splinter argue that the salary boost is largely an illusion. These compensation changes and other measurement issues, they find, account for 85 percent of the apparent rise in the top 1 percent’s share of after-tax income since 1960.

For their part, Piketty, Saez, and Zucman argue that the Auten and Splinter data is incomplete, and relies on an unrealistic way of treating corporate taxes. They insist that the finding that inequality has dramatically increased is robust, and holds up even after considering these issues.

The literature on income inequality is growing rapidly, and is fraught with political implications. Auten and Splinter are serious, nonpartisan researchers, but you could easily imagine conservative politicians latching onto their findings to argue that inequality isn’t that big of a deal. Their work is not the last word on the subject, and there’s plenty of analysis left to do. But it illustrates just how tricky it is to get a complete picture of what’s happening with inequality.

The case that we’re exaggerating the rise in inequality

Auten and Splinter adjust for the incentive the high tax rates of the 1960s and 1970s created for corporations to hoard money by treating all profits earned by corporations as income going to people who own corporations. They use dividends paid out and capital gains tax paid to estimate how much corporate stock individual taxpayers own, and then distributed the profits to those taxpayers accordingly.

This isn’t the only adjustment they make. Marriage rates, Auten and Splinter note, have fallen substantially over the last half-century. About 69 percent of tax filers were married in 1960, compared to 39 percent in 2015. But there’s been basically no decline among the richest Americans: Ninety percent of the top 1 percent of filers were married in 1960, and 86 percent were in 2015.

That can bias results when you’re comparing tax filers — which can either be single people, or married couples (including married couples where both people work) — to each other. If rich people are likelier to be married, that artificially increases the incomes of rich tax units, by making it likelier those units will include two people rather than just one.

“In order to control for these declining marriage rates, our analysis defines income groups based on the number of adults, rather than the number of tax units,” they write. “This means that each percentile includes the same number of adults instead of the same number of tax units.” This adjustment alone reduces the top one percent’s share of income by about 10 percent, across the board.

To get a full picture of incomes, Auten and Splinter do many of the same adjustments that Piketty, Saez, and Zucman do in their most recent work, including incorporating untaxed benefits like health insurance, adjusting for income that’s underreported to the IRS, adding in government transfers, and so forth.

They find that adding corporate profits, including ones taxed away by the federal government, substantially increases the top 1 percent’s share of income in 1960 and 1979, but only mildly increases its share in 2015. The effect of adding them in is to shrink the growth in the top 1 percent’s share considerably:

How including corporate earnings changes measures of inequality Auten and Splinter 2017

But other adjustments, like adding employer payroll taxes back into employees’ pre-tax income and including health insurance, also make a difference. The big change isn’t a result of just a handful of changes, but the cumulative effect of all these adjustments. Before the adjustments, the baseline Piketty-Saez data found an 11.3 percentage point boost in the top 1 percent’s share of income from 1979 to 2015. After, the boost shrank to five points.

Adding in transfer income from government programs, and then applying taxes to get post-tax income, shrinks the scale of the increase still further. Compare the Piketty-Saez numbers below (with the 11.3 point increase) to the after-tax income measure that Auten and Splinter devise (with only a 1.7 point increase):

Comparing Auten/Splinter to Piketty-Saez Auten and Splinter

Piketty, Saez, and Zucman respond

Piketty, Saez, and their coauthor Zucman aren’t as far away from the Auten and Splinter analysis as you might think. In recent work, they’ve moved beyond merely measuring pre-tax “market income” using tax returns to trying to use tax returns, survey data, and other sources to construct what they call “distributive national accounts.” This research tries to take every dollar in the American economy in a given year and assign it to specific individuals, and then look at the distribution of that income.

This approach yields a more modest increase in the top 1 percent’s share than the doubling from 9 percent to 20.3 percent that the market income data shows. Instead, the national accounts data shows that the top 1 percent’s share of after-tax national income grew from 9.1 percent in 1979 to 15.7 percent in 2014.

But that’s still a bigger boost than the one Auten and Splinter document. So what accounts for the difference?

Each set of researchers has their own answer. Auten and Splinter note that they’re not tracking total national income the way that Piketty, Saez, and Zucman are. But once they adjust their data to do so, they find that “after-tax national income shares [are] essentially unchanged: remaining at 8.5 percent in 1960 and increasing slightly from 10.1 to 10.2 percent in 2015. In 2014, top one percent after-tax national income shares were estimated by PSZ to be 15.7 percent, as compared to our estimate of 10.2 percent—a 5.5 percentage point difference.”

They find that about half the difference is explained by how the two teams distribute underreported income.

Piketty, Saez, and Zucman assume that underreported business income is distributed roughly the same way as normal business income: so business owners who don’t fess up on their full income to the IRS have the same income distribution as fully honest business owners.

Auten and Splinter don’t think that’s right; they argue that lower-income business owners are likelier to underreport. The rest of the gap is accounted for by differences in how to handle married couples, how to deal with distributing government deficits to individuals, and treatment of retirement income.

Piketty, Saez, and Zucman responded with a detailed data appendix defending their numbers. They argue that, while random audit data suggests pretty even rates of tax evasion across income, IRS audits aren’t as effective at unraveling more intricate attempts at tax evasion undertaken by rich people. If you consider that rich people’s efforts at evasion are likelier to be successful, then their approach of assigning more underreported income to the rich looks more reasonable.

They also disagree with the pre-tax income data that Auten and Splinter produced. “They have an unorthodox treatment of the corporate tax, which they assume falls a lot on current retirees (which are typically relatively low-income),” Zucman wrote in an email. “As a result the top 1% only pays 17% of the corporate tax in their series in 2013. This is too low. CBO, JCT, Treasury, and ourselves have a more justified treatment where the corporate tax falls primarily on capital owners, and the top 1% pays 40%-45% of the corporate tax.” But, as Saez adds, “This is relevant for the pre-tax income series comparison, not the after-tax one.”

A lot of questions about inequality come down to seemingly technical details

Those disagreements probably sound rather technical and nit-picky. But they make a huge difference to the overall numbers. Again, Piketty, Saez, and Zucman document a big boost in top-end income inequality, while Auten and Splinter show a much milder increase in the top 1 percent’s share since 1979 or so.

That being said, it’s important to remember that Piketty, Saez, and Zucman are hardly the only people documenting a major increase in inequality. The Congressional Budget Office, for instance, finds that the top 1 percent’s share of after-tax income increased from 7.4 percent in 1979 to 15.1 percent in 2012; it declined to 12.4 percent in 2013 as a result of Obama’s tax hikes.

And signs of inequality’s pernicious effects outside of the raw income data are clear. Just look at the relative stagnation of life expectancy for poor Americans, even as rich Americans are living longer and longer.

Further, the Auten/Splinter data doesn’t touch on middle-class wage stagnation, another major theme of the Piketty/Saez/Zucman work. The Auten/Splinter paper might reasonably make you question the scale of the increase in inequality, but the fact that it’s increased is hard to dispute.

The Auten/Splinter data does suggest that we might be underestimating how well the rich were doing in the 1960s and 1970s, and thus overestimating how different today’s age of inequality is from that period, a conclusion shared by some other recent research.

Among other things, that would suggest that the high individual tax rates of that period might have been less effective at tempering inequality than commonly thought. They certainly did something, but it appears rich people of that period found ways to collect the money tax-free regardless, specifically by parking it in corporations.