What do the Piketty data problems really mean?

Great post from Tyler Cowen.

"In some ways the new FT criticisms may not matter much, although I think not in a way which is reassuring for Piketty.  There were already several major problems with Piketty’s analysis and also empirics, including what Alex has called the asset price problem.  He wrote:

According to four French economists, Piketty’s measure of the capital stock is greatly influenced by the Europe-US housing bubble that preceded the financial crisis.

Adjusting for that factor seems to make the main results go away, and that is a purely empirical problem which has not been answered, at least not yet.

Another pre-existing empirical problem is that 19th century data seem to indicate that a “Piketty world,” even if we take it on its own terms, far from being a disaster, would likely be accompanied by rising real wages and declining consumption inequality, albeit rising wealth inequality.

That hasn’t been answered either, although a few people have suggested (without serious back-up) that if wealth inequality is going up that has to lead to political problems, or problems of some kind or another, and thus it can’t be something we can approve of or accept with equanimity, because inequality is really really bad, and therefore Piketty is somehow right anyway.  That’s a weak response to begin with and furthermore it doesn’t fit the available data.

Empirically, inheritances aren’t nearly as important as Piketty seems to suggest.
On Twitter Clive Crook wrote of the:

…distance between treacherous data and super-bold conclusions an issue at the outset. This underlines the point.

Now, when you cut through the small stuff, the new empirical problem seems to be that UK revisions, combined with a population-weighted series for Europe, contradicts Piketty’s claim of rising wealth inequality for Europe.   I would call that a serious problem.  I am not impressed by the “downplaying” responses which focus on coding errors, Swedish data points, and the other small stuff.  Let’s face up to the real (new) problem, namely that robustness suddenly seems much weaker.  You can’t argue that population-weighting is “the right way to do it,” but it is an entirely plausible way to estimate the wealth inequality trend.  If Piketty’s results don’t survive population weighting (and what are apparently the superior UK numbers), that suggests the overall rise in European wealth inequality is not very robust to how the pie is carved up and also that it is not backed by dominant, “rule the roost” sorts of forces.

It should be noted that Piketty’s response to the new criticisms was quite weak.  Maybe he’s not to be blamed for what was surely a rapid and caught-off-guard response, and perhaps there is more to come, but it doesn’t reassure me either.  He also should have run it by a PR person first (for instance, don’t start your response with a sentence ending in an exclamation point.)

That said, don’t focus on Piketty.  When evaluating debates of this kind, never ever confuse a) is he right? with b) “how much should we raise/lower the relative status of the author as a result of the new exchange”?  So responses like “he made all his data freely available,” or “he admits all along how complicated this all is,” address b) but not the more important a).  And if you are seeing people focus on b) rather than a), they have a problem themselves.  On empirical grounds it does seem we have another reason for thinking Piketty’s central claim isn’t quite right, at least not for the reasons he sets out, and perhaps not quite right altogether.

Addendum: Ryan Avent has a good survey of some key issues and responses."

In some ways the new FT criticisms may not matter much, although I think not in a way which is reassuring for Piketty.  There were already several major problems with Piketty’s analysis and also empirics, including what Alex has called the asset price problem.  He wrote:

According to four French economists, Piketty’s measure of the capital stock is greatly influenced by the Europe-US housing bubble that preceded the financial crisis.

Adjusting for that factor seems to make the main results go away, and that is a purely empirical problem which has not been answered, at least not yet.
Another pre-existing empirical problem is that 19th century data seem to indicate that a “Piketty world,” even if we take it on its own terms, far from being a disaster, would likely be accompanied by rising real wages and declining consumption inequality, albeit rising wealth inequality.
That hasn’t been answered either, although a few people have suggested (without serious back-up) that if wealth inequality is going up that has to lead to political problems, or problems of some kind or another, and thus it can’t be something we can approve of or accept with equanimity, because inequality is really really bad, and therefore Piketty is somehow right anyway.  That’s a weak response to begin with and furthermore it doesn’t fit the available data.
Empirically, inheritances aren’t nearly as important as Piketty seems to suggest.
On Twitter Clive Crook wrote of the:

…distance between treacherous data and super-bold conclusions an issue at the outset. This underlines the point.

Now, when you cut through the small stuff, the new empirical problem seems to be that UK revisions, combined with a population-weighted series for Europe, contradicts Piketty’s claim of rising wealth inequality for Europe.   I would call that a serious problem.  I am not impressed by the “downplaying” responses which focus on coding errors, Swedish data points, and the other small stuff.  Let’s face up to the real (new) problem, namely that robustness suddenly seems much weaker.  You can’t argue that population-weighting is “the right way to do it,” but it is an entirely plausible way to estimate the wealth inequality trend.  If Piketty’s results don’t survive population weighting (and what are apparently the superior UK numbers), that suggests the overall rise in European wealth inequality is not very robust to how the pie is carved up and also that it is not backed by dominant, “rule the roost” sorts of forces.
It should be noted that Piketty’s response to the new criticisms was quite weak.  Maybe he’s not to be blamed for what was surely a rapid and caught-off-guard response, and perhaps there is more to come, but it doesn’t reassure me either.  He also should have run it by a PR person first (for instance, don’t start your response with a sentence ending in an exclamation point.)
That said, don’t focus on Piketty.  When evaluating debates of this kind, never ever confuse a) is he right? with b) “how much should we raise/lower the relative status of the author as a result of the new exchange”?  So responses like “he made all his data freely available,” or “he admits all along how complicated this all is,” address b) but not the more important a).  And if you are seeing people focus on b) rather than a), they have a problem themselves.  On empirical grounds it does seem we have another reason for thinking Piketty’s central claim isn’t quite right, at least not for the reasons he sets out, and perhaps not quite right altogether.
Addendum: Ryan Avent has a good survey of some key issues and responses.
- See more at: http://marginalrevolution.com/marginalrevolution/2014/05/what-do-the-piketty-data-problems-really-mean.html#sthash.vHgd7K36.dpuf