The other day, Scott Greenberg of the Tax Foundation claimed that “Tax Rates on the Rich Were Not That Much Higher in the 1950s.” His idea? That despite a statutory top marginal income tax rate of 91% in that era, the rich actually paid a much lower effective tax rate, because they were able to recategorize their income as coming from investments, and thus subject it to a relatively favorable rate. The upshot, according to Greenberg, is that progressives should stop pointing to the 1950s as a high-tax, high-growth era, since taxes on the rich weren’t actually that much higher than they are now. Jordan Weissman expressed some justified—but excessively limited—skepticism here.
The data Greenberg draws on couldn’t come from a more reputable source, the Distributional National Accounts compiled by Thomas Piketty, Emmanuel Saez, and Gabriel Zucman. Greenberg points to their data series on effective top tax rates by income quantile. It shows that the effective tax rate for the top 1% of households (by income) was 42% in the 1950s, versus 36.4% today.
Unfortunately, Greenberg commits some basic errors in formulating his conclusion that “the tax burden on high-income households today is only slightly lower than what these households faced in the 1950s.” The total national income share earned by the top 1% and top 0.1% in that era was far lower than it is now, and consequently, the income thresholds required for entry into the ranks of the top 1% or the top 0.1% were lower. By today’s standards, there were many fewer rich households in the 1950s than there are now—in fact, almost none. The rich people from the 1950s that Greenberg is comparing to the rich of today were what we would now call the upper middle class—thus, not an apples-to-apples comparison. Had there been any 2017-style rich people in those days, they would likely have faced an effective tax rate near that confiscatory statutory rate of 91%.
It’s not a coincidence that the rich are so much richer now than they were in the 50s: it’s precisely because effective tax rates on the rich have gone down so much that it’s worthwhile to become rich in the first place. After all, when the government was going to tax away 91% of your income, what’s the point in bargaining for so large a slice of the pie?
You don’t have to agree with me that the ‘bargaining elasticity,’ to use Piketty, Saez, and their coauthor Stefanie Stantcheva’s term for it, is the operative mechanism explaining why the 1950s rich weren’t as rich as the 2017 rich. Instead you can agree with…. the Tax Foundation, which is well-known for its maintained assumption that high marginal tax rates on the rich cause them to supply less labor and acquire less education since the return to doing so is lower. Greenberg’s inference about effective tax rates in the 1950s is erroneous no matter what impact you think the effective tax rate has on individual behavior, provided it has some effect—which is why it’s strange for an organization like the Tax Foundation to be promulgating analysis that assumes, at least behind the scenes, that taxes don’t affect the decisions that economic agents make.
Greenberg’s mistake is a basic example of the bias that comes from mishandling a selected dataset. There aren’t any households earning $30 million, or $300 million for that matter, in the tax records of the 1950s, so they don’t enter into Greenberg’s analysis. From his selected sample, he draws erroneous conclusions about what their effective tax rate would have been had they been present in the data—that it would have been equal to the effective tax rate of the richest households he does observe. In fact, if they had been in the data, he would have observed a much higher effective tax rate than the one he inferred for them.
The great economist James Heckman taught us that selection bias is a frequent, indeed overwhelming, concern in economics precisely because economic agents are rational and hence their presence in a dataset reflects optimization decisions that they make. In Greenberg’s case, the 2017-style super-rich aren’t in the 1950s data, because they made the decision not to earn 2017-level top incomes, and they made that decision thanks to the exact policy that Greenberg imagines himself to be investigating: their effective tax rate.
So what can we learn from all this? Don’t trust the Tax Foundation, of course, but beyond that: even the best and most comprehensive data sources, like Piketty, Saez, and Zucman, cannot by themselves answer policy questions without some degree of economic interpretation and reasoning to go with them. And yes: effective tax rates on the rich were much higher in the 1950s than they are now. Exactly how much higher? That’s a tough question to answer—since there weren’t any rich people, we’ll never know for sure.