Financial Times vs. Piketty on U.S.: Smoke, No Fire
This article originally appeared in Forbes.
Last week I began critiquing the critique by the Financial Times of Thomas Piketty’s wealth inequality estimates in his Capital in the Twenty-First Century.
In that column, I enumerated the claims, by economics editor Chris Giles, that just fall flat, and I discussed at length some claims he made specific to the United Kingdom trends.
My conclusion was that while Piketty picked a probably-unjustifiable 2010 data point for the U.K. that shows wealth inequality rising after 2000, the increase from 1980 to 2000 that Piketty shows is solid. The 2010 issue—even if I am right that it is an issue—is probably not that consequential. The point Piketty picked may show too big an increase, but it may show too small an increase. Either way, his argument does not hinge on what happened between 2000 and 2010.
Part Two of my critique was always intended to focus on the United States trends Piketty estimates, but Piketty had the nerve to preempt my response to Giles with his own. Piketty’s response, while generally strong in my view, made no mention of one important issue with his time series. It is worth discussing it here, not for purposes of criticizing Piketty, but to clarify whether the trend he shows is likely to reflect reality. Assessing reality will require a third column to explore what, in fact, has happened to wealth inequality.
Giles has some nits to pick regarding Piketty’s two nineteenth century data points, saying in his spreadsheet that they appear not to come from the source Piketty references. I’m not going to bother to look into that because it just doesn’t matter. Here is Giles’s U.S. chart (with some color added to distinguish the various series). You should be comparing the purple and blue lines and should not be all that wowed.
Giles also criticizes Piketty for essentially making up the top ten percent share data from 1910 to 1950 and in 1970 by adding a constant 36 percentage points to the top one percent estimates. That’s entirely fair. This is definitely a case where I would have just left those data points off of the chart. But for Giles’s project of refuting the increase in wealth inequality and the U.S.-versus-“Europe” comparison, it’s basically irrelevant. And Piketty’s made-up numbers actually track new estimates released after Capital, as we’ll see below.
Piketty also constructs the top one percent estimates in opaque ways from 1910 to 1970, averaging estimates within a decade from a paper by Wojciech Kopczuk and frequent Piketty collaborator Emmanuel Saez to get the initial point and then adjusting the estimates upward by non-constant ratios. He presumably has his reasons. Should we worry? Here is a chart of my own that covers 1920 through 1970 (ignoring “1910,” which averages estimates from 1916, 1917, and 1918).
This chart includes the Piketty and Kopczuk-Saez series and adds a series in that was unavailable when Piketty wrote his book, from preliminary research by Gabriel Zucman (another collaborator of his) and Saez. Note that Piketty’s made-up 1920-1970 trend for the top ten percent is probably quite reasonable, since the Zucman-Saez trend for the top ten percent is similar to that for the top one percent. Indeed, even the levels of Piketty’s top ten percent share estimates are quite close to those of Zucman and Saez. Note, too, that whatever the basis for adjusting the Kopczuk-Saez top-one-percent figures, the adjustment only has the effect of raising inequality levels, not altering the trend. I believe Piketty adjusts the trend line upward to make it more consistent with the post-2000 estimates he has, which (in the absence of estate tax return estimates after 2000) rely on household survey data. Through 1970, I can’t see how this is problematic.
Still, it’s a big adjustment to raise the estimates over multiple decades in order to get one more (2010) data point into the trend. Here’s another chart, where I’ve enhanced Giles’s in several ways:
I’ve used New York University economist Edward Wolff’s latest estimates, which extend to 2010, rather than the estimates cited by Giles, and I’ve included new 2010 estimates from Federal Reserve Board analysts to update Giles’s “Kennickell” trend for the top ten percent share. (Arthur Kennickell is a top statistical economist at the Fed.) I’ve also added the Zucman-Saez series by eyeballing Zucman’s PowerPoint charts.
The most important issue to note is that Piketty has found increases in wealth inequality during the 1970s that neither Kopczuk-Saez nor Zucman-Saez found. As Giles emphasizes, Piketty switched in 1980 from adjusted estate-tax-return-based estimates through most of the twentieth century (except in 1960) to estimates based on household surveys. Presumably, Piketty wanted a consistent series from 1980 through 2010, but the choice may have produced an increase in wealth inequality in the 1970s where there was actually a decline. Another problem is that the points labeled “1980” in the Piketty series come from the 1989 Survey of Consumer Finances estimate. They are more accurately characterized as 1990 estimates.
Meanwhile, the 1998 SCF estimate goes into the “1990” data point (averaged with 1992 and 1995 estimates) instead of the “2000” one while the 2007 estimate is withheld from both the “2000” and the “2010” point. It is hard to see a consistent rationale across these decisions.
Piketty also finds an increase in the top ten percent’s share of wealth during the “1980s” (i.e., between 1989 and 1992-98), whereas Zucman and Saez find a decline. Wolff finds an increase in wealth inequality between 1983 and 1989, but the Fed economists do not believe the 1983 survey is sufficiently comparable to those from 1989 forward.
Piketty’s response to these issues concedes at first that the 1970 data point is uncertain and so the resulting trends are as well:
In particular, the estimate for year 1970 tries to combine the estimates available for top 10% and top 1% wealth shares for years 1960 and 1980 and the evolution of very top wealth shares between 1960, 1970 and 1980. This has little impact on the overall long-run pattern, but I agree that this is relatively uncertain, and that this could have been explained more clearly.
He goes on to cite the Zucman-Saez estimates to say, essentially, no worries—I still got it right, and in fact I underestimated the rise in wealth inequality. I’m not so sure, but more on that in my next column.
Ultimately, where I am less agitated than others about these issues is that I do not believe that the most unflattering (and unfair) interpretation of Piketty’s estimates—that they were cherry-picked in some way—changes the way we should think about his data. His estimates show the top one percent’s wealth share rising just six points, from 28 to 34 percent, over the forty years from 1970 to 2010. They show that wealth inequality basically stopped increasing in 1990. I just don’t think that anyone should have been particularly persuaded or alarmed by Piketty’s results in the first place.
Scott Winship is the Walter B. Wriston Fellow at the Manhattan Institute for Policy Research. You can follow him on Twitter here.
Interested in real economic insights? Want to stay ahead of the competition? Each weekday morning, e21 delivers a short email that includes e21 exclusive commentaries and the latest market news and updates from Washington. Sign up for the e21 Morning eBrief.