Tuesday, April 16, 2013

Reinhart and Rogoff - potentially very important news

Summary here if you haven't heard about it.

The work comes out of PERI - these are Post-Keynesian types really flexing their muscles. Austrians are interesting, but I do think if you want a serious heterodox contender you have to look to the Post-Keynesians rather than the Austrians.

Anyway - my advice is patience on this. It will be interesting to see Reinhart and Rogoff's response. A major issue seems to simply be a coding error, but there might be different reasons for making some of the other decisions R&R made. It's worth hashing that out.

It is troubling that some of the successful years left out seem to be in the heyday of demand management.

UPDATE: An initial response is here (HT - Ryan Murphy). Scanning I don't see a reference to the coding error. A lot of it makes the very reasonable point that a lot of the general story is preserved, and it's found by other authors. This part bothered me a little:

"Note that because the historical public debt overhang episodes last an average of over 20 years, the cumulative effects of small growth differences are potentially quite large. It is utterly misleading to speak of a 1% growth differential that lasts 10-25 years as small."

True enough, but the whole point is that the gaps close considerably when you correct the coding error, from a difference of 3.3% to this difference of 1%! So yes, a 1% difference is still a big deal when we're talking about growth, but if you're going to knock the PERI authors on that count than you have to also acknowledge how huge it was for them to close that gap by two thirds!!!

It's not surprising that R&R would lash out like that. This has caused a huge uproar, after all. But this criticism doesn't seem quite fair.

The other thing this response notes is that they were never making a causal claim. Indeed the original book was very clear about the crisis-to-debt causal arrow (something I and anyone who's ever read the book can easily verify). This is an important point. As Jonathan Catalan notes in the comment section here and Krugman notes in his post on this, the real problem is the politicos that took the 90% threshold and ran with it.


  1. Correcting R&R's data analysis is important, but I feel that a lot of the popular discussion on this new paper is led by reading too much into R&R's debt overhang paper. They're pretty clear that context matters, and so a a debt overhang beyond 90% might not necessarily be a bad thing -- if, for example, we're in the midst of a recession.

    1. Krugman made this point too - that the real problem is how people have used R&R. But a change like this in even the unconditional result is interesting.

  2. All I can say is that I am shocked. And, given the prominence of R&R, surprised.

    Lies, Damned lies, and Statistics. {shakes head}

  3. "A major issue seems to simply be a coding error, but there might be different reasons for making some of the other decisions R&R made. It's worth hashing that out."

    The real question is, why did they not say what they had done and why, in their methodology section?

    1. Well, the original paper, as stated was just a presentation at an AEA meeting. The JEP paper was actually peer reviewed and went through these issues. What I think is more interesting is what it means for other work, like Kumar and Woo's "Public Debt and Growth" that is at the imf. http://www.imf.org/external/pubs/ft/wp/2010/wp10174.pdf

  4. I'd love to see estimates of how much time and money spreadsheets save, and estimates of how much time and money they waste through errors like this. :(

  5. Hockey Stick! Hockey Stick! Hockey Stick!

    BTW read Noah and learn

  6. Reinhart & Rogoff point out that Hernden and Ash have quite similar results. "they find 0-30 debt/GDP, 4.2% growth; 30-60, 3.1 %; 60-90, 3.2%,; 90-120, 2.4% and over 120, 1.6%".

    I don't know why people are using debt-to-GDP. Since they're talking about overhang it seems more sensible to me to use GDP-to-debt-maintenance-cost.

  7. Dan -- I know I'm entering the conversation late. I've read news buzz around these articles, but not the articles themselves. From what is being discussed in news/blogs/etc., their conclusions seem very simplistic -- a great deal changes over time (culture, technology, etc.) that might also affect growth, etc. I assume you've read the articles and can fill me in -- did they control for/block on any variables?


All anonymous comments will be deleted. Consistent pseudonyms are fine.