This article was originally published in National Review.
By Jon Hartley, December 28, 2023
Since the 2014 publication and viral reception of Thomas Piketty’s Capital in the Twenty-First Century, much of the Western world has been captured by the narrative that income inequality (measured by how much income is earned by the top 1 percent or top 10 percent) and wealth inequality (measured by how much wealth is owned by the same) have been rising to levels not seen since the early 20th century. Some like Piketty argued that this had become a resurgent phenomenon driven by the dynamics of the rate of return on assets being greater than the economic-growth rate. When the empirical and theoretical backing of that argument was questioned, there then arose a new narrative based simply on the idea that there were negative social and political ills caused by too much economic power at the top.
In the 2010s, there were also important debates about the nature of inequality as a measure of economic progress: Should we care if the top 1 percent have a greater share of overall income if lower-income individuals’ earnings and absolute level of well-being improve as well? What about pre-tax versus post-tax inequality and the degree to which the existing tax and transfer system is properly working to help those least well off among us.
For many years, authors such as Piketty, Emmanuel Saez, and Gabriel Zucman have had a sort of monopoly on the IRS microdata used to produce their top 1 percent income-share and wealth-share estimates (few researchers have access to the fixed number of IRS data seats). Many critics publicly and privately questioned whether their analysis was biased in ways designed to reinforce the case that inequality was rising sharply.
Now, ten years after the publication of Piketty’s best seller, a different research team — Gerald Auten and David Splinter (of the U.S. Treasury Department and the Joint Committee on Taxation, respectively) — with access to the same IRS data, has produced radically different results trying to measure the same top income shares, albeit with a slightly different methodology.
In their new paper, they have found that post-tax income inequality has hardly changed over recent decades. Auten and Splinter still find that pre-tax income for the top 1 percent has increased but not to the extent estimated by Piketty, Saez, and Zucman (in their most recently updated estimates).
The Auten and Splinter estimates have been around for several years, but now they’re being published in the Journal of Political Economy, one of the very top journals in the field of economics, an event which is causing great upheaval. Could the rising inequality narrative of the past ten years have been substantially overstated?
Divergent Income Estimates: Piketty, Saez, and Zucman Compared with Auten and Splinter
Source: Gerald Auten and David Splinter, 2023
How exactly do they get to such different inequality results with the same data? There are often value judgments that need to be made in the calculation of income, especially non-visible income. Auten and Splinter find that prior to the 1980 tax reforms, poorer income deciles had more underreported “informal” income, which Auten and Splinter include in their analysis based on existing tax-evasion data. There are also some differences between the two studies in their treatment of untaxed capital income and in the way they include taxes and transfers.
Piketty, Saez, and Zucman have recently responded to the Auten and Splinter results, and not kindly, resorting often to ad hominem attacks. Calling critics “inequality deniers” or, in a reply to Auten and Splinter earlier this month, saying their results should be “discarded” is strange when someone has simply used different assumptions and come to a different result.
I have been long skeptical of the Piketty, Saez, and Zucman results because of the opaqueness of their methods and some suspect choices, such as their exclusion of the earned-income tax credit (the largest anti-poverty measure in the U.S. tax code) from their analyses, and because of the inclusion of health insurance (a nontax item) in the calculation of effective tax rates in Saez and Zucman’s The Triumph of Injustice, which makes it appear as though the rich pay less than they do in taxes relative to the poor. But, without direct access to the IRS data and without using the same calculations as them, or as Auten and Splinter, I cannot know exactly what led to such different outcomes and what influenced them. Perhaps charitably, one can think of the Auten and Splinter results as lower-bound estimates of income inequality and the Piketty, Zucman, and Saez results as higher-bound estimates. This acknowledgement alone would significantly disrupt the rising-inequality narrative and foster some humility among those who emphasize it incessantly.
But beyond debates over the data, it’s worth getting back to the fundamental question of what inequality is and why we should care.
“Inequality” is an amorphous catch-all word that can be correlated with a series of related ills, principally poverty and the lack of economic mobility or dynamism. Inequality can rise because the poor get richer while the rich get even richer, or because the poor get poorer and the rich get richer. An analysis of the income share of the top 1 percent alone does not make such a distinction. If top 1 percent incomes can be lowered by a policy that makes everyone, including the poor, worse off, it would reduce inequality even though it is clearly not something desirable.
As a general rule, we should avoid using relative measures (such as the top 1 percent’s share of income) as criteria for social welfare. Rather, we should focus on absolute distributional measures such as the level of income of the poor and the percentage of the population that makes below a certain income threshold in real dollars.
If we want to help the poor, we need to get serious about how we measure their well-being. Improved data and different measures, such as the “distributional national accounts” pioneered by Piketty, Saez, and Zucman in 2018, that attempt to capture absolute incomes and wage growth across the income distribution on a more timely basis would be a good start. Believe it or not, measuring the distribution of income (for purposes other than analyzing past tax returns) is hard to do in real time.
Once we have a better consensus on the measures and proxies for the social ills we want to address, then we can study how regulations, taxes, spending, and other government policies affect them. In the meantime, economists should avoid reading their preferred policies in the data and dismissing their colleagues who reach different conclusions.
Jon Hartley is a senior fellow at the Macdonald-Laurier Institute, a Research Fellow at the Foundation for Research on Equal Opportunity, and an economics PhD student at Stanford University.