Not to Pile On, But…Correcting Reinhart and Rogoff

April 16th, 2013 at 2:35 pm

I’m late to this and many excellent posts have already gone up citing a new paper that corrects what appear to be fundamental mistakes in Reinhart and Rogoff’s (R&R) influential finding that high debt levels lead to slower growth.  Specifically, they argued that when a country’s debt-to-GDP level gets above 90%, real GDP growth takes a big hit.

The authors of the new paper—Herndon, Ash, and Pollin—replicate R&R’s original work and make various corrections to a) methods and data choices and b) a ”spreadsheet error,” the latter where R&R appear to have left out some important data that has a big impact on their results.  When they correct R&R’s boo-boos, they get the results shown in the figure below.  It’s the last set of bars where Herndon et al’s corrections matter most.  In R&R’s work, countries in their top debt level category have a slightly negative average growth rate ; the corrected data find average growth rates of 2.2% in that category.

It’s a big difference, though I suspect R&R will say, assuming they acknowlege they messed up, that it still shows slower growth.  But that’s been the problem with their work from the beginning.  As I’ve written many times, riffing off of Bivens and Irons for one, if you mush everything together they way they do, you’re likely to get the causality backwards.  You’ll convince yourself that higher debt leads to slower growth when it’s more often the opposite.  Certainly in the US case, the most progress we’ve made against our debt ratios have been in periods of fast growth (and the biggest increases have been in periods of recession, slow growth, or war).

Get this wrong—which, thanks in part to R&R’s advocacy of their flawed results, has been the norm in countries across the globe—and you’ll be pulled, like a moth to a flame, towards austerity, the medieval leeching of growth from already weakened economies.

So kudos to Herndon et al for taking the time to get this right.  Let’s see what happens next, which could, as Matt Yglesias suggests, be nothing.  It’s not like facts are driving this debate.

 

R_Rcorrect

Source: Herndon, Ash, Pollin, Table 3.

Print Friendly

12 comments in reply to "Not to Pile On, But…Correcting Reinhart and Rogoff"

  1. Sandwichman says:

    Pile away, Jared! Even if facts don’t drive this debate, maybe a little pile driving will.


  2. Pablo says:

    Everyone agrees that faster growth is what is needed:

    Simplify tax code.

    Reduce Corporate tax rates.

    Address entitlements to restore confidence.

    Address healthcare costs.


  3. Perplexed says:

    Yeah, it would be kind of mean to “pile it on” to all of their suffering. So how exactly does this work now? Do they have to compensate all of the victims that have suffered needlessly from these austerity policies due to their carelessness or negligence (or worse)? Or are they covered by the same “creditor class” immunity from criminal and civil prosecutions as bankers and their captive legislators?


  4. Lars Olsson says:

    Not to go all Krugman on you, but I’m afraid – due to the reasons you suggest (that facts are not driving this debate) – the 90% figure has already become ensconced in the minds of the Very Serious People as something We All Know.

    In short, it has likely already become yet another in a seemingly endless parade of discredited zombie ideas which, regardless how often they’re refuted, continue to pop back up to partially drive the debate.


  5. David C says:

    Here’s a very interesting followup:

    http://www.nextnewdeal.net/rortybomb/guest-post-reinhartrogoff-and-growth-time-debt

    This post looks at the causality question, ie whether high debt causes slow growth, or whether slow growth causes high debt. Using the data from the HAP paper, Mr. Dube regresses current year debt/GDP on growth in the three prior years and again against the growth in the subsequent three years.

    In both cases, there is a negative correlation between growth and debt/GDP. However, the correlation vs. the prior three years is *much* stronger. This strongly suggests that slow growth causes high debt rather than the other way around.


  6. David C Mace says:

    Even more telling about this kind of `research` even in your graph is the conspicuous lack of error bars. If economists put those in it would most likely show that even the corrected results are all but menaingless.


  7. Barry says:

    Somethings which were pointed out in the Rortybomb summary:

    1) R&R deleted data points, without stating this; all points deleted were against their hypothesis.

    2) R&R weighted, and weighted in an extreme manner, without stating this; the weighting moved the results towards their desired conclusion.

    3) R&R, for three years, did not respond to requests from their colleagues for openness, when their colleagues repeatedly failed to replicate their findings.

    4) R&R made numerous causal claims, and then lied about making them.

    It’s not a matter of an Excel error; it’s deliberate academic fraud.


Leave a Reply

Your email address will not be published.

Current month ye@r day *