A couple of
days ago Daron Acemoglu and James Robinson published a critique of Piketty’s

*Capital in the 21*. It is published here. Because of the renown of the authors, perhaps more than because of its intrinsic quality, it is a review worth reading. I read it today and my brief reaction to the three main critiques by Acemoglu and Robinson is as follows.^{st}century
1) Piketty totally neglects institutions. This is hard to
understand since Piketty's explanation for a large part of changes in
inequality in the US, France and elsewhere are precisely institutional: higher and
then lower income and inheritance tax rates, abolition of slavery. It is so
much so that in my class and presentations I have dubbed Piketty's approach "a political theory
of income distribution". So I really fail to see any validity in this
critique.

Actually, that part of the critique is fundamentally
dishonest. It proceeds as follows. First, Acemoglu and Robinson establish the
equation Piketty=Marx. Then then criticize Marx for ignoring institutions, more
or less correctly (but clearly that has nothing to do with Piketty). Then, since
they have already decided that Piketty is really Marx, they barely give one or
two examples of Piketty’s lack of concern with institutions and discuss, as many
authors have done before, Piketty’s “fundamental laws”. They seem to imply that
having “fundamental laws” (never mind that these laws are an identity and a
dynamic equilibrium condition) somehow means that one does not care about
institutions. Very bizarre.

2) Lots of inequality increase is due to higher inequality of labor incomes. This is true especially for the United States and no one disputes it; neither does Piketty. He actually mentions it repeatedly especially when he discusses the increasing share of labor incomes in the top 1%. But he is also aware of potentially very high inequality which may exist when both capital and labor incomes are heavily concentrated and often among the same people. In other words, just saying that inequality is driven by rising concentration of labor incomes does not make such inequality “benign” and somehow “fair”.

3) Panel regressions. This is a novelty. Acemoglu and Robinson test whether r-g is correlated with increase in inequality (measured by the top 1% share). They find that the sign of the coefficient is in most cases negative, that is different from the one predicted by Piketty. This is a good approach and I am sure it will be repeated a number of times. I expect an entire cottage industry of similar cross-country regressions trying to prove or disprove Piketty’s contention. Acemoglu and Robinson’s first set of regressions assumes an equal _net_ rate of return for all countries. Thus the right-hand side variable is not r-g but simply “g”. This approach is surely wrong. Piketty's rate of return is defined after taxes which are of course not the same in all countries. It is not at all obvious that all countries face even the same gross rate of interest, much less that the net or gross rate of return to capital is the same. Thus only the second set of Acemoglu-Robinson regressions with national estimates of rates of return makes sense. But there, the results are inconclusive. Moreover, there are no usual controls at all except for the country and year dummies. What would be the results if one introduced policies and outcomes with respect to labor, taxation, government expenditures, level of income and the panoply of the usual controls that we use in inequality regressions? And aren't time dummies precisely related to Piketty's argument about (say) globalization and tax competition holding r high? But they are not discussed at all.

2) Lots of inequality increase is due to higher inequality of labor incomes. This is true especially for the United States and no one disputes it; neither does Piketty. He actually mentions it repeatedly especially when he discusses the increasing share of labor incomes in the top 1%. But he is also aware of potentially very high inequality which may exist when both capital and labor incomes are heavily concentrated and often among the same people. In other words, just saying that inequality is driven by rising concentration of labor incomes does not make such inequality “benign” and somehow “fair”.

3) Panel regressions. This is a novelty. Acemoglu and Robinson test whether r-g is correlated with increase in inequality (measured by the top 1% share). They find that the sign of the coefficient is in most cases negative, that is different from the one predicted by Piketty. This is a good approach and I am sure it will be repeated a number of times. I expect an entire cottage industry of similar cross-country regressions trying to prove or disprove Piketty’s contention. Acemoglu and Robinson’s first set of regressions assumes an equal _net_ rate of return for all countries. Thus the right-hand side variable is not r-g but simply “g”. This approach is surely wrong. Piketty's rate of return is defined after taxes which are of course not the same in all countries. It is not at all obvious that all countries face even the same gross rate of interest, much less that the net or gross rate of return to capital is the same. Thus only the second set of Acemoglu-Robinson regressions with national estimates of rates of return makes sense. But there, the results are inconclusive. Moreover, there are no usual controls at all except for the country and year dummies. What would be the results if one introduced policies and outcomes with respect to labor, taxation, government expenditures, level of income and the panoply of the usual controls that we use in inequality regressions? And aren't time dummies precisely related to Piketty's argument about (say) globalization and tax competition holding r high? But they are not discussed at all.

Finally, I agree with their point that the focus on top
shares is limited and that lots of important changes take place along the
entire income distribution. But as well known, the focus on top shares is driven
by the very nature of fiscal data used by Piketty and his coauthors.

I do not discuss Acemoglu-Robinson analysis of South Africa vs. Sweden increase
in inequality because I really fail to see a great virtue in it. As I
unfortunately have to confess, I often find reading Acemoglu-Robinson
descriptions of political changes quite superficial: they read like Wikipedia entries
with regressions. I had the same feeling here too.