By Phillip W. Magness
Such claims fit with a growing tendency to depict Saez and Zucman as anti-establishment outsiders and victims of viewpoint discrimination against their work. Warren Gunnels, an economic adviser to the Bernie Sanders campaign, suggested as much to the Times, arguing that economists “that disagree with Saez and Zucman…are funded by the powers that be, the establishment, the billionaire class.” Zucman has personally encouraged and cultivated a similar narrative on his Twitter feed, even though the evidence speaks to a very different reality.
As to the matter of their disputed statistics, a much simpler explanation than viewpoint discrimination lies at the heart of the matter. Rather, the vetoing of Zucman’s hire by Harvard represented a rare display of academic integrity in a politicized academy that all too often elevates and celebrates subpar research on account of ideological agreement with its message.
The issue with Zucman’s work revolves around a stunning statistical claim that he made last fall. According to his own proprietary calculations, the overall effective tax rate paid by the ultra-rich in the United States had dipped below that paid by the bottom 50 percent of earners for the first time in 2018.
The weeks that followed their release also revealed something far worse than failing to adequately vet this seemingly stunning empirical claim.
Instead of objectively reporting the latest findings from tax statistics, Zucman was placing his finger on the scale. He appeared to be bending his results to conform to the political narrative of Warren’s campaign, which he was also advising at the time. Through a series of highly opaque and empirically suspect adjustments, Zucman had artificially inflated the tax rate paid by the poorest earners while simultaneously suppressing the tax rate paid by the rich.
Whereas Zucman now claimed to show the ultra-wealthy paid just slightly north of 20 percent of their earnings in taxes, the most recently available year of his previously published numbers (2014) places the rate at 41 percent. I called attention to this discrepancy with a tweet, as did Columbia’s Wojtek Kopczuk and the University of Central Arkansas’s Jeremy Horpedahl. Then the floodgates of scrutiny opened.
At the bottom of the income ladder, he was artificially raising the depicted rate faced by the poorest earners. He did so by excluding federal tax programs that are intentionally designed to alleviate the tax burden on the poor, such as the Earned Income Tax Credit and the Child Tax Credit. By leaving out these programs, Zucman not only broke from decades of statistical conventions – he also created the illusion that the tax rate paid by the bottom quintile was nearly twice its actual level.
Zucman’s handling of the very top of the distribution ventured even more aggressively into the territory of intentional data manipulation. The biggest discrepancy here came from his handling of how to assign corporate tax incidence across earnings. When economists examine corporate tax incidence, they usually distribute it across a variety of affected parties according to fairly standard assumptions about the portion that falls onto shareholders, onto other forms of capital, and onto the noncorporate sector of the economy due to various pass-through effects.
Indeed, Zucman followed these conventional assumptions in his aforementioned academic article from 2018, coauthored with Saez and Thomas Piketty. In his new statistics, however, he jettisoned all conventional literature on corporate tax incidence and adopted his own heterodox approach that effectively assigns 100 percent of actual incidence to its statutory incidence, namely shareholders.
To cap off his data-bending exercise, Zucman undertook a final suspicious step. He claimed to report tax rates for both groups in 2018, citing this as the first time in history that the rich paid a lower rate than the poor. Except there’s one glaring problem: at the time Zucman released his headline-grabbing new stats to the press, the IRS had not yet published any statistical reports from its 2018 returns (these only become available in the early spring of 2020). Closer examination revealed that Zucman simply imputed his “new” numbers from 2016 stats based on his own highly imprecise assumptions about the effects of the 2017 Trump tax cut. While these figures are sure to be scrutinized when the actual numbers come out, suffice it to say that Zucman was presenting estimates for years where he does not actually possess data and, furthermore, staking one of the most audacious claims of his argument on the same.
Within a few weeks of the earliest scrutiny of Zucman’s statistics, the wheels began to fall off of his narrative.
There’s a good reason that most of the profession has looked on his most recent work with a skeptical eye. It is the product of political advocacy, open electioneering on behalf of Warren, and statistics that were constructed primarily to fit with the narratives of her campaign. In bending his data and eschewing the process of peer review, Zucman shed any pretense of doing social science – including his own previously published work where it contradicted his new political narrative.
Harvard’s administration, to its credit, recognized this inappropriate breach of scholarly standards and exercised a rare but necessary check upon their ongoing erosion in the academy. Naturally, that has left other activist academics and journalists who support his claims for political rather than scientific reasons in a fit of rage.
Phil Magness is a Senior Research Fellow at the American Institute for Economic Research.