Usually, the Friday afternoon before Memorial Day is the perfect time for a political news dump. The Financial Times used it to drop a major investigation into the data behind Thomas Piketty’s hit book, Capital in the Twenty-First Century. For those on Twitter who hadn’t left yet for vacation, it was the top story. “Not so fast with that Nobel,” Clive Crook tweeted. From Quartz economics writer Tim Fernholz: “This is big, if true: Piketty's data flawed?”
But the errors that the FT finds, while significant, do not materially change Piketty’s conclusions or disprove the economic theory behind his work. Chris Giles, the economics editor of the FT, reviewed Piketty’s data set and found multiple occasions where the French economist had miscopied numbers from one data set into his published spreadsheets. Piketty also averaged data for France, Sweden and Britain without making any population adjustments. Effectively, that makes every Swedish citizen carry the same weight as seven British and French citizens, according to Giles. In his spreadsheets, Piketty also makes adjustments to the numbers that seem arbitrary. “In the US data, Prof Piketty simply adds 2 percentage points to the top 1 per cent wealth share for his estimate of 1970,” Giles writes, providing a screenshot to prove it. It’s unclear what to make of these adjustments.
“[O]ne needs to make a number of adjustments to the raw data sources so as to make them more homogenous over time and across countries,” Piketty writes in a response posted to the Financial Times. “I have tried in the context of this book to make the most justified choices and arbitrages about data sources and adjustments.”
Piketty doesn’t specify these adjustments with his data sets and Giles points out other times where certain data do not have sources. Piketty should have done a better job explaining his adjustments and specifying his sources, but few economists in the world have been as open and transparent with their data as Piketty has been with his. It wouldn’t make much sense to distort the data and then release the incriminating evidence to the public. In addition, Scott Winship, an inequality scholar at the Manhattan Institute and frequent critic of Piketty, has used Piketty’s U.S. data extensively and understood all of his adjustments.
“Having looked at the U.S. inequality spreadsheet quite a bit, I definitely knew what he was doing in that spreadsheet,” Winship said.
But the data errors that Giles are found are real nonetheless. Do they materially change Piketty’s results? Giles thinks so: “The central theme of Prof Piketty’s work is that wealth inequalities are heading back up to levels last seen before the first world war. The investigation undercuts this claim, indicating there is little evidence in Prof Piketty’s original sources to bear out the thesis that an increasing share of total wealth is held by the richest few.” For Britain, this seems to be true, but it does not seem to bear itself out for France, Sweden or the United States, assuming more errors do not come to light.
Giles constructs alternate series using other sources of data to compare and improve upon Piketty’s work. For France and Sweden, Giles’s data is almost identical to Piketty’s:
The largest differences are for Britain, where wealth inequality for the top 1 percent and top 10 percent are considerably greater under Piketty’s original data:
Finally, for the United States, Giles’s data show, at most, that wealth inequality stayed constant over the past few decades, while Piketty’s show a slight increase:
It’s important to remember that wealth data is subject to significant error. For instance, economists do not agree how to factor capital gains into wealth data: Do you include capital gains as they accrue or only when they are realized? In addition, further back in time, the data becomes even more unreliable.
“These numbers are just imprecise to begin with. The numbers that Giles has come up with are imprecise,” Winship said. “Piketty’s original numbers were imprecise. Piketty probably, in places, talked about the numbers in a way that deemphasized the imprecision and I think it’s fair to whack him for that. But when I look at these charts, to me, if you imagine margin of errors around any of these data points, it sort of looks like nothing has changed much.”
Even if you believe that Giles’s findings dramatically change Piketty’s results, they have little bearing on his economic theory. Giles makes a passing comparison to economists Carmen Reinhart and Ken Rogoff (R&R), who drove a significant part of Republican austerity agenda, but saw their findings disproven in 2013. Liberals celebrated when Thomas Herndon, a graduate student from UMass Amherst, discovered a spreadsheet error in R&R’s results that invalidated their main finding. But unlike Piketty, Reinhart and Rogoff largely had no economic theory to ground their argument that national debt crises occur when a country’s debt level surpasses 90 percent of GDP. Once their data fell apart, their theory had no legs to stand on. On the other hand, Piketty fits data to this theory, but does not depend on it. Piketty’s theory—right or wrong—is largely unaffected by these results.
Danny Vinik is a staff writer at The New Republic.