Earlier today Thomas Piketty released a 10 page supplemental appendix to his book Capital and the 21st Century, responding to the raging controversy around his data. The appendix was styled as a response to the Financial Times investigation, although it is relevant to issues that have also been raised elsewhere including on this blog.
My initial take is that Piketty’s response contains much verbiage and a bit of fist-shaking, but surprisingly little additional information that would rectify some of the criticisms that have been made. This probably won’t stop those who are already sympathetic to Piketty’s political message from taking his retort as gospel, but I’d like to point out at least one reason why, perhaps, they should start asking some tough questions.
In one of his harshest passages for the Financial Times, Piketty says the following:
“What is troubling about the FT methodological choices is that they use the estimates based upon estate tax statistics for the older decades (until the 1980s), and then they shift to the survey based estimates for the more recent period. This is problematic because we know that in every country wealth surveys tend to underestimate top wealth shares as compared to estimates based upon administrative fiscal data.”
On this point, I have no reservations in saying that Piketty is absolutely correct. He does stray a bit in the next sentence by suggesting that this sourcing mixup will necessarily bias the data in favor of declining inequality, but as a methodological critique it is entirely to raise questions about this sort of apples-to-oranges merging of disparate data sources using widely divergent estimation techniques for wealth.
But Piketty also has a problem: he too is guilty of the exact same type of data-mixing that he now ascribes to the Financial Times, only more so. In fact, his retort to Chris Giles could just as easily describe the veritable Frankenstein of a graph that Piketty himself assembled to portray rising wealth inequality in the United States.
I’ve focused many of my criticisms on this graph (Figure 10.5) for exactly the same faults that Piketty now imputes to the Financial Times, only his are worse. A more technical examination of what Piketty did in his spreadsheets may be found in this post, but briefly summarized, Figure 10.5 is assembled from three different authors and alternates back and forth between estate taxes and surveys no less than four times:
- 1910s-1950s: Estate Tax Data (Kopczuk-Saez)
- 1960s: Survey Data (Wolff)
- 1970s: Estate Tax Data (Kopczuk-Saez), partially calibrated to Survey Data (Wolff)
- 1980s: Survey Data (Wolff), partially calibrated to other Survey Data (Kennickell)
- 1990s-2010s: Survey Data (Kennickell)
It bears mention that Piketty tends to rely upon two survey-based data sets over an estate tax data set in the latter part of his graph. While he justifies doing so on the relatively high reputation of the US survey data vis-a-vis its European counterparts, this is a judgment call in itself that actually goes contrary to what he chastises the Financial Times for doing. His tradeoffs also have the effect of exchanging a more complete annual set of figures in the estate tax estimates for sparsely populated survey estimates that contain only one or two data points per decade. At bare minimum, it seems that Piketty’s retort to the Financial Times is attempting to impose methodological standards that his own data manipulation also exhibits, and in a more severe way.