Thursday, 18 April 2013

Reinhart, Rogoff and Persistence on Austerity

Big news online yesterday were that researchers were finally able to replicate Reinhart and Rogoff''s (henceforth R&R) results on debt-to-GDP ratios and growth. As it appears, the authors did some serious mistakes including leaving several observations out of their sample (for no apparent reason) and some significant coding errors in Excel. Although many (including the authors) could argue that mistakes happen, it is not that often that these mistakes dictate fiscal policy in the world. The Guardian, for example, questions how many people have become unemployed due to this mistakes, while Azizonomics presents several occasions where such researches have been quoted by politicians, in defense of their actions.

The grander question would of course be why did politicians and economists place so much attention to one specific paper in the first place, especially when results were never so extreme. Let's us assume for a second that their results hold (presented below) and study them carefully.

Source: The Wall Street Journal
What notably hits most readers is that the average growth rate of real GDP when the debt-to-GDP ratio exceeds 90% becomes negative. What most did not consider was their methodology: R&R give equal weight to all countries after they had averaged their growth rates for any bracket. Thus, using this methodology means that real GDP growth has almost a 50-50 probability of being positive or negative. Then, this means that although there is no real GDP growth on average, it is very likely that we will have some nominal growth (which will in fact assist in repaying debt, if we are talking about the Eurozone).

Edward Hugh makes a another significant point: it all depends on the country and how developed it is. This would mean that it is easier for a developing country to allow her debt burden to increase, although it would be quite rough for a developed one (whether countries are allowed to print money or control monetary policy is of course another issue).

This prompted yours truly to make a little research of his own. In the following charts the reader may observe the average real GDP growth of a small sample of developed countries for 10-point debt-to-GDP brackets up to 200% based on annual observations (where brackets are absent no data were found in those-data range is dependent on the country).
United States
United Kingdom
Spain (quarterly data)
As not to bore the reader with too many charts, the following presents the average of the averages, based on the R&R methodology. 

The data indicate that not only does a debt-to-GDP ratio of over 90% not decrease growth, but even as it increases over 100%, real GDP growth appears to persist, even on extreme levels. Although the dataset employed above uses data from 1949-2011, WWII could have something to do with the initial periods where debt was extremely high and growth persisted. Yet, this argument seizes to have any rational base for debt burdens of up to 120% (or higher depending on the country) or periods much later than WWII (as the specific case of Belgium and others indicates).

In addition, Arindrajit Dube argues that the above mentioned paper exhibits problems with causality. It not that debt causes growth, it is growth (or no growth to be fair) which causes the debt burden to increase (after a debt-to-GDP ratio of 30%). His "exercises" appear to be sound both economically as well as mathematically, yet although mathematics and statistics indicate a specific pattern, the truth is not so simple.

As with most things in life, it lies somewhere in between. In distinct cases (in which statistical aggregation is not helpful) both could be significant in predicting their future paths. In Dube's analysis, it appears that a decrease in the growth rate will cause an increase to the debt-to-GDP ratio. The question could be what caused this shock? If this is exogenous (e.g. a recession caused by bank failures) then Dube's analysis could be true. Yet, if a state wishes for some inane reason to increase its debt burden (say to increase social benefits for voters to like their government) it may have any effects on GDP depending on the way this is measured:

An increase in benefits has no direct impact on output as they do not constitute any part of it, yet it affects both consumption and investment. Thus, even as the debt burden will be increased, the growth rate of GDP will also be increased. Nevertheless, if a government insists on such policies, it may shift causality is on the side of R&R once more: debt can cause growth (or contraction). The mechanics behind this are simple. The government cannot sustain the increase in debt forever thus it will have to use some kind of consolidation for it to be sustainable, and through the mechanics better described here, it will negatively affect growth rates.

The simple conclusion is that people should be extremely cautious of non-replicated results. More importantly, when these results affect policy people should be even more cautious on how they use them, as they affect the lives of millions. As Richard Feynman, after having been led astray on the neutron-proton coupling constant by reports of "beta-decay experts" stated, "since then I never pay attention to anything by "experts". I calculate everything myself."

P.S. For those interested in the spreadsheets just send me an email

No comments:

Post a Comment