There are data errors, interpretations and prejudice

Posted on

According to the FT Thomas Piketty's book 'Capital' has been 'undermined by errors'. Chris Giles, who works for the paper, says he has found spreadsheeting mistakes and some inconsistent use of data in Piketty's work. The claim Giles then makes as a consequnce is that when he reworks the data he can find no increase in wealth in recent years, contrary to Piketty's claim.

In response Piketty, who has clearly not been given enough time to provide a meaningful explanation by the FT, has dismissed the claim, saying that, first of all he has underestimated wealth differences because offshore has not been included and secondly that his research is part of a great body of work that shows increasing wealth concentration. The implication is very obvious: he is saying that even if there is any error that does not change the reality of what is happening.

I have not checked the data. I do know something about using data to prove hypotheses. I know a great deal about the reaction of those who do not like the conclusions to work that challenges the prevailing world view from what is considered a left wing perspective, so let me offer a view, and not technical analysis.

First, I would be astonished if the odd transcription error or even odd occasional formula mistake did not exist in Piketty's massive work. He is human after all. But Giles has found no apparent systematic fault of the type in Reinhart and Rogoff. The claimed spreadsheet errors really do look like occassional mistakes, at worst.

Second, and much more importantly, some of Giles' claims are simply about choices. Whether to weight data or not is always contentious. Do and you're damned (as I know) and don't and you're damned. This is not an error or getting anything wrong. That is the sort of decision social scientists face, and it's why social science is misnamed because there are no absolutely right or wrong answers.

Third, Giles makes the absurd assumption that Piketty's data sources are reliable. They're not. As I have noted very recently, current HMRC and ONS data on wealth in the UK differ by approximately £2 trillion and both will underestimate offshore welath, dramatically. So sometimes a researcher has to use their judgement on data and even correct for it (I had to decide what to do with straightforward data inconsistencies in data reporting from Companies House in my recent work).

And then there's Giles wish to say Piketty is wrong which is very clear above all else is what he writes. Whatever else Giles is he's not objective. He is pursuing a vendetta. And in the process he is ignoring, as Piketty has already pointed out, all to evidence to the contrary of Giles' opinion. After all, we thought inequality was rising before Piketty, and with good reason. And the evidence is all around us. The IMF and World Bank were of that view before 'Capital'. There is no reason to change that view now. If it's wrong and we are to take no action on it then it is going to take a lot more than a transcription error on a single line of 19th century data to suggest we need to change tack now.

In fact, without saying Giles has not found some issues that Piketty may wish to respond to in more detail ( which no doubt, he will), all that Giles has so far proved is that he has a rather childish determination to undermine an argument he finds offensive to his wealthy patrons, and that he is takng considerable delight in his pedantry whilst at the same time ignoring the broader evidence and displaying very little knowledge of the conundrums that dealing with data present.

I won't change my mind on the reality and dangers of inequality even if Piketty was proven wholly wrong (which appears to be very, very unlikely). But, more than that, I expect this attack to backfire. It appears insubstantial. It wreaks of desperation. And it will open a wider debate that will reveal the compelling evidence for action.

Giles may feel smug for a moment. But that, I suspect, is his whole problem.