Saturday, May 24, 2014

Piketty Debunked? Not So Fast

Obviously, it couldn't last. Piketty's book was so universally lauded by such a chorus of great minds (Solow, Wolf, Krugman, even Summers with some reservations), and an even greater chorus of lesser minds (the journalists seduced by his 'rockstar' persona and Gallic charms), that is was only a matter of time before someone pushed back. And just as the book's original reception was probably overdone - it's a great book, but there are plenty of possible criticisms and counter-arguments - so the counter-offensive is similarly overcooked.

Chris Giles in the FT takes issue with Piketty's data on rising wealth inequality, raising some interesting questions about how he fits together different data sources to create a two-century long series, and how this series compares to other available data. Not surprisingly, different sources give different numbers - indeed vastly different numbers. In particular, Giles was struck by the marked difference between the wealth figures used by Piketty for the UK, and those published recently by the UK Office for National Statistics, which suggest much lower levels of inequality. Intrigued, he dug further using Piketty's data, which is posted online (although I couldn't get it this afternoon - maybe he's checking the files?).

Giles comes up with this chart:

By adding the raw data from the UK Inland Revenue figures, used by Piketty and colleagues to extend the series, and the ONS data, which they did not use, Giles is able to throw doubt on the uptick in wealth inequality Piketty claims for the period since the 1970s. This, alongside other errors and questionable decisions, means that the books 'central findings... no longer seem to hold'.

That's quite a big claim. Is it true? Well, here I'll just look at the data for the UK. There is a problem with the chart posted just above - it takes Piketty et al's series from 1810, and then, for the final few decades of the series, throws in other data sources which show a lower level of inequality. The picture that leaps out of the page is that instead of a U-shaped trend over the 20th century, claimed by Piketty, the FT's alternative data suggests a flatlining distribution since the 1970s, thus invalidating the claim of rising inequality.

But this is misleading, because by throwing in new data that gives a lower figure in the same chart, the visual impact is of a different trend that is not really supported by the data. The IRS numbers that Giles throws in raw were used by Tony Atkinson and Piketty to construct the longer series, with adjustments to attempt to make them consistent with different sources for earlier periods. The fair test of whether Piketty's trend exists or not is to compare the IRS numbers with data for the earlier period. In fact those numbers track the trend of the Piketty series fairly closely, but with lower absolute values. As for the ONS numbers, they give us only 3 data points over a 6 year period, which happens to include the biggest shock to asset prices since the Great Depression and the Second World War. You can argue that the ONS's lower estimate of wealth inequality is better than Piketty's, but you can't say anything about the trend.

However, the IRS numbers do still suggest rising wealth inequality. If we place these numbers in their own chart, with no tricky juxtaposition against the longer data series, this is what we get:

A clear, though unspectacular upward trend in wealth inequality as measured by the Inland Revenue data, on which the Atkinson series used by Piketty is based. It's also worth pointing out that the HMRC numbers, although not much help with the long-term picture, do attribute 70% of wealth in the UK to the top 10%. As far as I can see, the main difference between Piketty and Giles's alternative data is that the latter suggests that inequality started to rise in the late 1980s rather than the 1970s.

I'll leave the rest to the specialists in constructing historical series of income and wealth distributions, a complicated field in which Piketty has a long track record. If he's made mistakes in the data work used in the book, boo to him. But if he hasn't, it would probably be the first 700 page academic book not to make any.

The problem is that, having been anointed the 'rockstar' economist, Piketty is set up to fail, and the normal business of argument and questioning which any successful research project spawns will inevitably become politicised and dramatised. The FT is, for once, guilty of a kind of tabloid simplification that is quite out of character: the headline 'Piketty findings undercut by errors', 'Rockstar economist's wrong sums on rising inequality' reminds me of the debate around climate change and the 'hockey stick' graph. And not in a good way.

Doubtless one percenters will be happy to find that they are not doing as well compared to the rest of us as we thought. Economic liberals that have sweated to produce justifications for rising inequality in terms of productivity and innovation will maybe revise down their estimations of the contribution of the one per cent. OK, I'm being sarcastic. The direction of policy and the phenomena we observe all around us, never mind the data, suggest that wealth and income are increasingly concentrated. But of course documenting long-term trends with multiple kinds of data is difficult, as the climate scientists could attest. Maybe Bill Gates, Steve Jobs, or indeed Yaya Toure and Tony Blair, are exceptional and unrepresentative anecdotes from a world characterized by fairly evenly distributed wealth. But I really doubt it.