Tax hikes targeting the wealthy are the rage in Washington, with Rep. Alexandria Ocasio-Cortex proposing a top marginal rate of 70 percent and now Sen. Elizabeth Warren endorsing a new 2 percent “wealth tax” on the total assets of multi-millionaires. Both proposals purport to address allegedly skyrocketing inequality in the United States.
The claim that inequality is on the rise invariably traces back to the statistical work of economists Emmanuel Saez and Gabriel Zucman, along with their frequent collaborator Thomas Piketty. Piketty, Saez, and Zucman’s (hereafter PSZ) studies do depict exploding inequality in the past 40 years, a pattern they attribute to the repercussions of Reagan-era tax cuts that ended over 40 years of top marginal rates exceeding 70 percent.
Eager to boost the viability of the Ocasio-Cortez and Warren proposals, journalists have credulously repeated the PSZ numbers as if they were a stylized fact of the modern era. Unfortunately, this repetition gives a very misleading picture of inequality in the United States.
For all the attention they have received, the PSZ statistics are actually an outlier compared to other attempts to measure top income and wealth concentrations. They consistently depict higher levels of inequality than virtually all other metrics, and by substantial margins. As a recent comparative analysis from the Urban Institute showed, the PSZ numbers report more than twice the income share growth for the top 10 percent of earners than other studies depict during the same period.
We may see this effect by looking at both their income and wealth numbers directly. The first chart depicts the concentration of the top 10% income share over time, comparing the results of a 2003 study by Piketty and Saez (on which the current PSZ numbers are based) with subsequent revisions to adjust for data problems.
As the adjustments reveal, the PSZ numbers consistently overstate the income concentration in two periods. First, their income concentration figures are consistently too high in the period before World War II, and largely miss the decline in inequality that came about as a result of the 1929 stock market crash. The difference here is largely the result of unrealistic assumptions and accounting errors they make while adjusting their data for changes in how the IRS reported deductions and certain exempt income categories prior to the war.
Second, their stats show a much more rapid acceleration in income concentration after 1980 than a more conservative measure by economists Gerald Auten and David Splinter. The Auten-Spinter numbers adjust for the well-documented phenomenon of income shifting around the Tax Reform Act of 1986, where wealthy individuals took advantage of differences in individual, corporate, and capital gains rates to legally shelter some of their earnings from high rates of taxation. The PSZ numbers neglect this practice, which causes what economist James Galbraith calls a “data break” in 1986. The resulting illusion of an inequality spike largely disappears with Auten and Splinter’s adjustments to ensure consistency in the definition of market income before and after the change.
Turning to the broader measure of personal wealth – the target of Warren’s proposed tax - we find an almost identical situation at play among the wealthiest 0.1 percent of earners. While PSZ depict a sharp rise in wealth concentrations since 1980 using estimates they derive from the same problematic IRS records, two independent measures tell a very different story.
The first alternative, calculated by economist Wojciech Kopczuk in collaboration with Saez (before the latter switched to using IRS data), uses a more consistent set of records from the federal estate tax. Combined with mortality statistics, they estimate the top wealth shares on an annual basis. Here we see a similar pattern across the century to the adjusted income concentration figures. The estate records depict a Great Depression era decline in wealth concentration that PSZ’s preferred data does not fully capture until World War II. The estate records also show that wealth concentration has either remained flat, or only slightly risen, from 1980 to the present.
A second alternative measure derives from the Federal Reserve’s triennial Survey of Consumer Finances (SCF), which uses statistical sampling of wealthy households to assemble a robust estimate of top level net worth. Except for a few predecessor surveys, the SCF only dates to the late 1980s, but still represents an independent check on tax-derived statistics for the overlapping period. It too shows a top wealth share that is more prone to fluctuation than PSZ, and that only appears to be rising at a modest rate.
So why do these measurement gaps exist for both income and wealth shares? Part of the reason has to do with imperfect data sources, including historical records that become substantially less complete the further one goes back in time. Some of the differences also result from assumptions that each study bakes into its estimation method, such as Auten and Splinter’s adjustment for income shifting around the 1986 Tax Reform Act and PSZ’s omission of the same.
While economists continue to refine and improve their measurement techniques, one thing is certain. The PSZ statistics represent the upper extreme of studies claiming a sharp rise in inequality, whereas alternative approaches – including corrections to oversights in the PSZ statistics – reveal substantially subdued patterns. Given the high levels of uncertainty surrounding inequality measurement and ongoing improvements to each technique, common sense should caution us against adopting a sweeping tax code overhaul based on such an ambiguous and disputed statistical record.