The Financial Times takedown of Thomas Piketty’s inequality data demonstrates a range of errors from the trivial all the way up to errors which seem inexplicable apart from their importance in bolstering his thesis. As Business Insider pointed out late Friday, FT isn’t just saying Piketty made mistakes, they are also suggesting he may have “exaggerated” his results.
In his long blog post on the subject, FT’s Chris Giles’ identifies 7 different classes of errors in Piketty’s data. Some of these are as simple as reading the wrong figure from a chart of source data. This is certainly an error but it’s correction is unlikely to do any harm to Piketty’s thesis or his reputation. Simple mistakes happen.
Then there are some non-trivial errors, like Piketty’s decision to use a simple average to combine data from three different countries. As FT points out, it’s difficult to see why a population-weighted average wouldn’t be a better choice in this case. As a result, the validity of Piketty’s Europe data is in doubt. But again this could be explained as simple carelessness and, by itself, it may not alter his results.
The problems seem most acute for Britain, where Prof. Piketty shows
rising concentrations of wealth among the richest since 1980, when his
source data does not. This appears to be the result of swapping between
data sources, not following the source notes, misinterpreting the more
recent data and exaggerating increases in wealth inequality.
In particular, Giles goes in to great detail about Piketty’s data selection for the Britain graph:
Prof. Piketty ends his series taking at face value the level of the HMRC
data, despite HMRC saying clearly (see section 1-g) the data is not
suited for that purpose, nor is it consistent with the old Inland Revenue Series which Prof. Piketty uses for earlier years. This latter point is also clearly stated in the notes to the source data.
Here Giles is referring to a note in Piketty’s source data which reads “[The data] is not a suitable data source for estimating total wealth
in the UK, or wealth inequality across the whole of the wealth
population; the Wealth and Asset survey is more suitable for those
purposes.” Did Piketty miss this warning? Or did he simply forget to explain why he chose to ignore it?
What makes this a difficult problem for Piketty to recover from is the fact that, apart from his jump to an inappropriate data set, Piketty’s thesis fails. As Giles’ concludes, Piketty’s graph of the wealth of the top 10% in
Britain “does not match any of the underlying data” after the 1970s.
If you correct Piketty’s Britain data, the increase in inequality among the top 10% goes away. And if you use a proper weighted average to combine this to create a trend for Europe as a whole, the rise in inequality is either very faint or non-existent.
So, to sum up, there is no outright accusation at this point that Piketty tweaked his data. However, we do know that he selected a data set which he should have know (because there was a clear written warning) was inappropriate to make his case. If Giles is correct, then without his use of this inappropriate data, Piketty’s thesis for Britian and for Europe fails.
What explanation can Piketty offer? If he admits he didn’t know about the warning that came with his Britain data then he’s admitting his central thesis is in doubt. If, on the other hand, he did know about the data warning then he needs to justify his choice to ignore it. The first choice undercuts his thesis. The second seems like special pleading.
There is a third and fourth possibility. Maybe Giles made a mistake and Piketty can abandon the dodgy data without undercutting his thesis for Britain or Europe. Or perhaps he can offer an ironclad explanation for why –despite the written warning–this data is appropriate to use. Until Piketty responds in detail to Giles’ critique we won’t know for sure. All we know for certain now is that Piketty’s initial response is insufficient.