Slow Recoveries Cause Financial Crises

No, you didn’t read that title wrong. In this post I hope to outline the argument against the view, popularized by Reinhart, Reinhart, and Rogoff, that recoveries following financial crises are naturally slower. This is a conjecture that is in the same realm as the “balance sheet recession” hypothesis. But first, a little context.

Yesterday, I wrote a post defending Miles Kimball against an attack by Paul Krugman. The gist is that Krugman read up until Kimball mentioned the Reinhart, Reinhart, and Rogoff (R,R, and R) paper, and disparaged him. I informed Miles about my post on Twitter today, and had a bit of correspondence on my views. I had mentioned that I’ve never bought into the view that financial crises cause slow recoveries. I had a prior post that I thought had made it on Modeled Behavior, but I couldn’t find it, so I assume that it ended up getting posted on “www.cheapseatsecon.com”, which the internet ate, and I don’t have the archives unfortunately. In any case, I told Miles that I would write a new post.

Touching off with their paper, The Aftermath of Financial Crises, continuing with their book This Time is Different, and focusing their attention on public debt in their 2012 paper, Reinhart (x2) and Rogoff have been documenting the correlation between financial crises and subsequent rates and levels of GDP growth. This line of research is certainly extremely useful, however I think that R, R, and R go wrong when they ascribe to financial crises causality for slow growth.

Immediately after I heard this conjecture, I was skeptical. However, it has since taken the economics profession by storm…so I’m defending a minority here, but then again, most of the profession thinks that monetary policy has been highly accomodative over the last six years (same link), so consensus can be wrong.

The first problem that you will come across in this line of thinking is that, if you dig through The Aftermath of Financial Crises, you will find that far from being always followed by slow recoveries (as is common to hear in the press — uncorrected by R, R, and R), their findings actually show that recoveries following financial crises are on average slower than recoveries from other types of recessions. You can see this in the image below from their paper.

Using Reinhart and Rogoff’s definition of ‘systemic crises’, we can see that there is wide variability in the various recoveries they studied. This is evident not only in employment levels, but also in price movements.

Another problem that would plague any study like this is identification. There are far too many confounding factors when doing a cross-country analysis of recessions to make a blanket statement like “financial crises cause slow recoveries”, even with a very large data set. Unfortunately, the US has had but one (or two) episodes which Reinhart and Rogoff classify as “systemic crises”. That one data point is the Great Depression, and the possible second is the Great Recession. In both cases, Reinhart and Rogoff get causality wrong. This is likely because they date the financial crisis in the Great Depression from 1929, but there was no financial crisis in 1929. Similarly, many people will date the financial crisis during the Great Recession to 2007, or maybe even before…but in fact, the crisis didn’t start until late 2008/early 2009.

Finally, the experience of the US does not lend itself to the conclusion that financial crises cause slow recoveries. A more trenchant analysis would lead you to the conclusion that financial crises are caused by a monetary policy unable or unwilling to accommodate an increase in the demand for money (or money-like assets). Even in the unlikely case that monetary policy is truly impotent at the zero lower bound, it is still the case that it is monetary policy that is providing the drag on recovery, not the financial crisis.

So, while there are certainly interesting questions to be investigated stemming from R, R, and R’s survey of historical data (i.e. what causes the variability in recovery rates from financial crises?), the conclusion that financial crises cause slow recoveries is not one that jumps out of the analysis…unfortunately it is the one that has been latched onto by the profession and the press.

Update: Here is another paper that analyzes, and cuts against the idea that financial crises cause slow recoveries. I think that their conclusion is problematic as it doesn’t seem to take into account the unwillingness of the Fed to adopt a monetary policy stance consistent with return to the pre-crisis trend level of NGDP growth.