Rogoff and Reinhart Respond
After yesterday's fracas, the two economists offer a lengthy response to the FT. My takeaways:
1. The coding error is confirmed. That by itself raises the mean growth rate of high-debt countries from -0.1% to 0.2%. So the negative growth result is gone.
2. Rogoff and Reinhart offer what is to me a convincing defense of their decision to omit the immediate postwar data from New Zealand, Australia, and Canada, which is that in 2010, when the paper was published, they had just gotten the postwar data from these countries, and had not yet had time to clean it up. I find this convincing because they have subsequently published using the omitted data.
3. Reinhart and Rogoff make a fairly strong defense of their decision to weight by country, rather than country year:
This brings us to the core conceptual issue, which Herndon, Ash and Pollin argue greatly biases our results. They argue that we use an “unconventional weighting of summary statistics.” In particular, for each bucket, we take average growth rates for each country and then take an average of the result. This seems perfectly natural to us, and hardly unconventional. We do not want to excessively weight Greece, for example, which has debt over 90 per cent for 19 years in the 1946-2009 sample. The post-war Advanced Economy experience would quickly reduce to the experiences of Greece and Japan. Our approach has been followed in many other settings where one does not want to overly weight a small number of countries that may have their own peculiarities. Our approach is quite clear from table 1, which also gives the averages for each individual country.
Our 2012 Journal of Economic Perspectives paper, based a much longer time period (1800-2011 versus the 1946-2009 Herndon et al focus on), gives episode-by-episode data for each country, including growth rates and number of years, so the results are quite transparent. The problems with weighting long episodes much more heavily than short episodes, as Herndon et al. suggest, become much more apparent in the longer time series (an earlier version of which was also used in Table 1 of our original AER paper) As we noted in our initial comment yesterday upon just receiving the paper, our JEP paper anticipates most of the aggregation debate and diffuses it by using a case-study approach. That is where this literature is now expanding.
4. Rogoff and Reinhart not-so-subtly accuse Herndon et al of cherry picking the mean figures and ignoring the median figures, which Reinhart and Rogoff also provided. Pass the popcorn: this should trigger some fun back and forth.
5. While the medians and country-level analysis still suggest a strong relationship between debt and lower growth, the mean results are not robust, which means that the idea of a tipping point, where things suddenly get really gnarly because you've got too much debt, is in serious question. The important catch in the Herndon et al critique is not that there was a coding error; it is the sensitivity of their results to New Zealand.
6. That said, Ireland, Greece, etc do seem to offer clear cases of a tipping point where things suddenly get really gnarly, and by roughly the mechanism that Reinhart and Rogoff suggest in their paper: a debt crisis which triggers a fiscal contraction. Note that if these countries weren't in the eurozone there wouldn't even be a question of whether they should do austerity; or rather, the market would quickly answer that question in the affirmative by refusing to lend them money. So for now I'm going with "there is a tipping point, but it's probably country and situation specific, and it's hard to know where it is."
7. It's possible that overturning Rogoff and Reinhart should make us more nervous, not less; we can no longer claim that we are safe as long as we keep the debt under 90%.
8. Macro is hard. I'm tempted to say "there are no robust results in country-level regression analysis". The specific details of a country's economy and history probably matter a lot, but we don't know which ones. Everything's serially correlated with itself and correlated every which way with everything else. As both the critique and the response indicate, you can get incredibly different results by making small changes in your method--and those results are often driven by a handful of countries, or years.
9. There are still ample theoretical and empirical reasons to worry about debt. To name just a few:
a. "Crowding out": the logic of Clintonomics. If the government borrows too much, the private sector doesn't have money to invest. Pretty certain that this doesn't apply in America or Europe right now, but it certainly has and could in other times and places.
b. Debt crisis: these are ugly, and often accompanied by other ugly, destructive things, like hyperinflation. Of course causality runs both ways; countries in trouble are more likely to get into a debt crisis. But the more debt you have on the books, the higher the risk that a downturn tips you into crisis. Sudden fiscal contractions are much worse for the economy and other living things than gradual winddowns.
c. Debt dynamics driving fiscal contraction: Well short of an all-out crisis, if your interest rates start rising faster than inflation, you start having to either raise taxes or cut spending; usually both. That slows your GDP, at least in the near term.
d. Income redistribution: worth noting that repaying a big debt load usually involves cutting spending or raising taxes on middle class folks who have to cut their own spending, and giving that money to capital owners. Some of those owners are outside your country, so you don't even get derivative benefits.
10. This doesn't overthrow the broader scope of their work, such as the finding that you tend to get slow growth after financial crises. They certainly seem to have called that one right.
11. Coding errors happen. They're now easy to catch, in part because researchers are more transparent than they used to be. They're also going to be more common when you have more researchers building big private data sets. Overall, more researchers building big private data sets that they then share with other people is a good thing. Acting as if everyone who screws up a spreadsheet must be a terminal idiot who should resign in shame does not foster the broader goal of better research.
12. The next time you see a big regression analysis that 100% Proves, Through the Magic of Science, That You Are Right About a Big Important Issue . . . well, my friend, Remember the Coding Error.