(Final?) Notes on the Reinhart/Rogoff saga

Robert Samuelson and Anders Aslund have a go at defending the R&R results, and the authors themselves further respond. A couple of notes on all of this.

As pointed out elsewhere, the claim that the UMASS paper actually supports R&R’s alleged core finding of a significant relationship between high debt and slow growth is flat wrong.1 Yes, the midpoint average growth rate of high-debt countries (over 90 percent of GDP) is lower than the average growth of lower-debt countries, but the differences are not statistically significant. Really, people should give up on this one.

More importantly, the real argument all along has been two-way causality: data showing that there is lower growth at high debt levels does not show that high debt causes low growth. A finding of statistical association between high debt and slow growth would surprise precisely nobody, but there is a better case to be made that slow growth leads to high debt rather than vice-versa. Arin Dube demonstrated this very well in his note on Reinhart and Rogoff’s entire data-set, and for the U.S. John Irons and I did the same with Granger causality tests on the U.S. data in July 2010, nearly three years ago. Paul Krugman kindly referred to our results in his blog saying “John Irons and Josh Bivens have the best takedown yet of the Reinhart-Rogoff paper (pdf) claiming that debt over 90 percent of GDP leads to drastically slower growth.” So the causality problem has been well known for some time. By the way, we compiled the data we used ourselves because emails to Reinhart and Rogoff requesting their data went unreturned. Perhaps if they had shared their data at that time their actual weighting procedure would have become clear much sooner and even their spreadsheet error could have been corrected. Kudos to the UMASS authors for being more persistent than us and for the work they did.

Lastly, Reinhart and Rogoff argue against the thesis that slow-growth causes debt rather than vice-versa based simply on the fact that high-debt episodes are often long-lived:

“The fact that high-debt episodes last so long suggests that they are not, as some liberal economists contend, simply a matter of downturns in the business cycle.”

Paul Krugman has more on this, but I think the simple point is that “downturns in the business cycle” seems to be an intentionally narrow claim being put forward by Reinhart and Rogoff. It’s true that outright recessions in modern times tend to not last years and years. Yet slow growth caused by a persistent lagging of aggregate demand behind potential output can indeed go on for years and years. The U.S. economy, for example, is a full five years and counting from the beginning of the Great Recession, and a full 39 months into the official recovery, yet there remains an output gap of just under 6 percent of GDP.

So, one could argue (if one was feeling tendentious) that budget deficits in 2010, 2011 and 2012 were not driven at all by the economy being in outright recession in those years (hence not driven by “downturns in the business cycle”), since the economy wasn’t in an official recession. Yet these deficits were certainly driven by a shortfall of aggregate demand.

A clue that this is indeed what is going on in lots of the Reinhart and Rogoff data is their other persistent finding that the slower growth associated with (again, associated with, not caused by) high-debt  in their data is not durably associated with high interest rates. The classic argument for how adding to public debt harms an economy runs through higher interest rates, and without these it seems much more likely that the causality runs the other way.

Finally, I’d just like to add that another problem with the larger R&R narrative that both their book and their paper have constructed. This narrative  presents  financial crises as consigning economies to slow growth through some iron law of nature—and their 90 percent debt threshold is clearly intended to imply that efforts to use fiscal stimulus to combat this slow growth in the wake of the U.S. housing bubble’s burst will be thwarted. But this narrative is wrong—most of their sample includes country experiences before modern Keynesian macroeconomics was discovered. Since then, we have known how to deal with aggregate demand-led recessions, and the only real danger posed by these sorts of recessions is that policymakers would be too ignorant to respond correctly. Don’t get me wrong—that’s a real danger, and it’s happening today! But policy failure, not “debt” or “financial crisis” should be clearly identified as the stumbling block to full recovery.


1. I say “alleged core finding” because the real core finding was the claim that a 90 percent debt/GDP ratio was a clear tipping point over which growth slowed. This is obviously the finding that stuck in the public debate, and the authors themselves just helped to push this impression.