Back in April, I reported on the best work yet regarding the famous Reinhart and Rogoff (R&R) paper that purported to prove that government debt over 90% was catastrophic for economic growth. The work was done by the brilliant young economist Arindrajit Dube. He took the R&R data for high government debt periods, and found that economic growth was slow before the high debt periods, not after. In other words, slow growth causes high debt, not the other way around.
The reason all of this is big news and the reason that many of us are so angry is that R&R have been engaged in a kind of intellectual Whac-A-Mole game. Their paper was very careful not to claim causation: they never said that high debt caused slow growth. This in itself is a problem, I think. If there is no causation, who cares? Especially given the paucity of high debt events. In addition, when politicians all over the world used their research as though it proved causation, R&R said nothing. But R&R did have the intellectual cover (which they have used time and again) that their research, while uninteresting, was not incorrect—at least until the spreadsheet error was discovered. But then Ken Rogoff went all around talking up the causation that they had absolutely no proof of. When anyone points this out, R&R respond that they never said there was causation. Hit Rogoff here, R&R pop up over there. It is very frustrating.
Matt Yglesias puts it well:
Now, two researchers at the University of Michigan, Miles Kimball and Yichuan Wang has redone Arindrajit Dube’s work a bit more carefully. While Dube did find a little causation, Kimball and Wang present a very provocative conclusion, After Crunching Reinhart and Rogoff’s Data, We’ve Concluded That High Debt Does Not Slow Growth. Of course, they do hedge a bit. Miles Kimball and Brad DeLong have been going around on this issue and they both seem to think that there must be some effect of debt on growth, but Kimball’s point is that the data simply don’t show it. I suspect that is because there really is an effect but it is so small that it is wiped out by the noise in the data.
Of course none of this will matter. The Whac-A-Mole game will continue. I doubt that R&R will even respond to this. Everyone knows, after all, that they never said there was causation. Except when they did. But it wasn’t in the paper. So they are not only innocent, they are right. There really should be a way to impeach academics. (Note: I wouldn’t impeach Reinhart who hasn’t behaved too bad. But Ken Rogoff: no professorship for him!)
I was bothered that Kimball and Wang did not mention Dube’s work. I know that they haven’t published a real paper as of yet. But the introduction should go something like this: R&R did research that implied forward causality; Dube did research showing the causation was the other way; here is a more rigorous analysis.