The Piketty results are well-known, but new research casts some doubt on them. But the debate is not over: some of the new data make his hypotheses stronger. (Hat Tip: Bradley Hansen.)
The paper everyone should read is by Richard Sutch (UC-Riverside). He concludes that Piketty’s estimates for wealth inequality between 1810 and 1910 are unreliable. The catch? Sutch’s data show bigger swings: a higher rise during the Gilded Era, a steeper fall during the Great Depression, and a bigger recent increase.
I am not sure why this would change Piketty’s overall conclusions.
There is also a paper by Carlos Góes (IMF) that essentially says that increases in r−g do not, empirically, lead to increases in income inequality. It is a good paper, but I doubt the conclusions. VAR models estimate the correlation between multiple variables and their lagged values. They are theory-free, by definition! Which means that when Góes finds that increases in r−g do not lead to increases in the returns to capital, there are two reasonable responses:
- Doubt the data. See his definition on page 6; it certainly is not clear that r is correctly measured. (He uses a better measure in the robustness checks.)
- Ask why the relationship does not seem to work. If it is because market forces push r back down, then Piketty has a problem. If it is because institutions react to redistribute wealth and income when r rises, then he does not. For an example of a great paper that explores the latter, see here.
After all, it is basically an identity that the proposition that a rate of return on capital higher than the growth rate of income will lead to a higher income share for capital, ceteris paribus. The empirical debate does not really revolve around the truth of that identity; rather, it revolves around (1) whether r > g is true; (2) if #1 is true, what other things might cancel out the relationship between r−g and inequality.
You can find a great example of the a Piketty critique along these lines here, by Daron Acemoglu and James Robinson. They do not stop with the long-term data showing that the identity does not seem to hold; they dive into the examples of Sweden and South Africa to show why and how it does not hold.
Conclusion: this debate ain’t over yet!