Is Piketty wrong about British and Swedish wealth?
Embarrassingly, I missed this reply by Tim Worstsall to my post “Understanding Piketty, part 1.” My apologies to Mr. Worstall and my readers; despite his writing it August 14, I just discovered it the other day when I was mindlessly looking at site traffic data from Alexa.
In his post Worstall takes issue with Piketty’s claim (which I endorsed) that if Financial Times author Chris Giles was correct about the level of British wealth concentration (the top 10% controlling 44% of UK wealth), then British wealth inequality in 2010 was lower than that of Sweden and, indeed, lower than even the lowest share ever held by the top 10% of wealth owners in Sweden (about 53%) which, he said, “does not look very plausible.”
Worstall’s point was that, surprisingly enough, if we measure wealth inequality by the Gini index, Sweden in 2000 had greater inequality than did the U.K, 0.742 to 0.697 (higher is more unequal). His ultimate source (according to the Wikipedia article he cites) was work by the creators of the Credit Suisse Global Wealth Report, a research effort which Piketty praises as “innovative” in capital21c, p. 623 n. 8.
Let’s first note that even if that were true, it does not get Giles off the hook. Giles, whose error was to tack survey-based UK wealth data for 1990-2010 on to earlier tax-based wealth data (thereby biasing it severely downward; see also Howard Reed in The Guardian), does not dispute that the proper measure for inequality is the wealth share of the top 1% and top 10% of wealth owners. Giles makes no appeal to alternate measures to save himself. Thus, on their agreed measure of wealth inequality, Giles fails to make a dent in Piketty’s data.
However, Worstall’s point is an interesting one on its own merits to Piketty’s attempted reductio ad absurdum. While in part 1, I pointed out that income inequality measured by the Gini index is lower in Sweden than in the United Kingdom, the further fact that wealth inequality is always higher than income inequality within each country does not mean, as I blithely assumed, that the country with lower income inequality will necessarily have lower wealth inequality as well.
The question then becomes which is the more meaningful measure of wealth inequality. The U.K. has higher top 1% and top 10% shares but, evidently, a lower Gini coefficient. As I noted in part 3, Piketty deliberately avoids using the Gini index. As Piketty’s sometime-collaborator Facundo Alvaredo writes, “The most commonly used measure of inequality, the Gini coefficient, is more sensitive to transfers at the center of the distribution than at the tails.” This is not a problem for the top shares measures; they have a much more intuitive meaning than the dimensionless Gini index. One might well argue that there is more political significance for the top 1% of wealth owners to increase their share from 20% to 30% than there is for owners at the 85th percentile to gain a corresponding amount of wealth from those below them. But to make that argument doesn’t prove it’s true.
Alvaredo also elaborates on a way to adjust the Gini index for variations at the top of the distribution, which he attributes to Atkinson. As Alvaredo shows in his paper, it is possible for the unadjusted Gini index to be falling even as the adjusted Gini index is rising. I took a stab at adjusting the figures given by Worstall by taking the top 1% share of Sweden in 2000 as 20% and the U.K.’s as 30%. That gives adjusted Gini indices of (.742*(1-0.2)) + 0.2 = 0.7936 for Sweden and (0.697*(1-0.3)) + 0.3 = 0.7879 for the UK. These are much closer, but the U.K. is still slightly more equal if I have gotten this right. In any event, while Swedish income inequality remains robustly lower using either top shares or Gini index, wealth inequality for Sweden is only lower using income shares, but still not Gini.
There remains an obvious question for Worstall: What is the trend of U.K. wealth inequality using the Gini index? If it increased, then Piketty’s finding of an increase using wealth shares will be robustly backed up with this measure Worstall is preferring. Gini may give Worstall a desirable result for a comparison, but still unpleasantly show that Britain is more unequal in wealth than in 1980. That’s the actual question Piketty and Giles were disputing. However, I have yet to find a such a 1980 Gini index for wealth; as Piketty notes in capital21c, the drawback of the Credit Suisse research is that it does not go back in time very far. Perhaps a reader knows where a series of Gini indices for wealth (unlike income, which is easy to find) can be found.
So the answer to the question in the title is that we don’t actually know. Likely, however, it doesn’t matter as far as trends in inequality since 1980 are concerned. It’s definitely worth thinking about what it might mean that the two wealth inequality measures diverge for ranking Britain and Sweden in 2000, even if we eventually conclude that one measure is definitely better than the other.
And it’s not like Piketty is unaware of a potential for greater wealth inequality in Sweden. In June, he lectured in Helsinki, arguing against a recent trend in the Nordic countries to abolish estate (inheritance) taxes. Sweden abolished its inheritance tax in 2005. Not only does this shift the tax burden to those with lesser wealth, as he argued in Helsinki, but it follows from the argument of the book that it takes away the possibility of generating the most reliable form of data on wealth inequality itself.
Cross-posted from Middle Class Political Economist.
Can you link to Tim’s comment?