column: Sound economics or faulty statistics

Written by Meghnad Desai | Updated: Apr 29 2013, 06:09am hrs
Reinhart-Rogoff were wrong in postulating a cliff-effect at 90% debt levels but no one denies excessive debt lowers growth

There has been a tiny earthquake in economics over the last fortnight. Reputations have been dented. Conclusions have been challenged and certainties have collapsed.

Or so it would seem. Reinhart and Rogoff, authors of the justly famous book This Time is Different, published an article in the American Economic Review 2010 Growth in a Time of Debt. They claimed to have summarised a lot of data for many economies, developed and emerging, on public debt-GDP ratio and GDP growth rates.

They summarised their results in a simple table where debts were classified into four bucketsbelow 30%, 30-60%, 60-90%, and above 90%. The average growth rates for the four categories were 5.2, 4.9, 2.5, and -0.2. For developed economies, the pattern was similar but with the final number being -0.1% for debt above 90%; for emerging economies the final figure was a robust 1.3% though still on a declining trend. Reinhart and Rogoff also gave the median growth rates which showed a much more gentle decline as debt grew; 1.6% for developed economies and 2.9% for emerging economies as debt went above 90%.

Most remarkably for a research paper based on a large data set, Reinhart and Rogoff carried out no regression analysis nor did they even compute standard deviations for the average growth rates they had calculated. After 30 years of toiling with data sets, I ceased doing econometric work 20 years ago. But I would have been chastised by my peers for dealing with averages without standard deviations. But there it is and the paper became a classic. The numbers, especially about the negative growth rate for debt-income ratio above 90%, went viral. In the febrile atmosphere of US Presidential election, the Conservatives lapped it up. Paul Ryan, a Presidential candidate for the Republicans, cited the Reinhart and Rogoff result. It became a gospel.

When natural scientists get a result after an experiment, other scientists try to replicate the experiment and check if the result is the same. Only after it has been checked by many other scientists does it get publication.

Even economists, during the controversy about Keynesianism versus monetarism, were always trying out each others data sets to check the results. Here nothing seems to have happened for three years. Now a team of three researchers at the University of Massachusetts Amherst (known for its radical political economy teachers)Herndon, Ash and Pollinhave checked out Reinhart and Rogoffs results Does High Debt Consistently Stifle Economic Growth A Critique of Reinhart and Rogoff. They found naive statistical errors, data omissions by oversight and many other problems. But the most serious difference is about the growth rate for debt-income ratio above 90%. Herndon, Ash and Pollin found the sequence of percentage growth rates for the four classes as 4.2, 3.1, 3.2, and 2.2. Thus, the negative relationship between debt and growth is confirmed but the fall-off is not sharp at the high end. Thus, there is no cliff-effect at 90%. Herndon, Ash and Pollin also extend the buckets to five by splitting high debt-income ratios into 90-120% and above 120%; the downward decline is confirmed, though there is no cliff after 90%.

This has led to a mini earthquake in financial journals. Reinhart and Rogoffs reputation has been dented. Bill Gross of PIMCO has changed his stance about debt/growth relationship on the basis of this reversal. The Keynesians who never believed in austerity are happy that they have been vindicated. And yet I wonder whether so much fuss is justified.

To begin with, the Reinhart and Rogoff results for median growth rates showed much less dramatic decline as debt-income ratio went up. Nor did they test for the cliff-effect statistically. What is more, they did not ask as any economists should: Is the causality from debt to growth or the other way around Should they not have tested for a simultaneous causal structure treating debt and growth as jointly endogenous The same critique applies to Herndon, Ash and Pollin who stop at correcting Reinhart and Rogoff but do not probe any deeper.

The lesson is not so much about famous authors making simple mistakes; that is not new. It does, however, tell us that if economists simplify then they are apt to mislead their non-economist readers. The world has become hungry for information and now everything is online and instantaneously becomes news, is swallowed by the markets and priced in. No one waits for a proper scientific check. Somewhat like the now notorious Laffer Curve, politicians hailed a result which chimed with their gut instincts (prejudices). Upon serious examination, many such laws are found to be fragile. The Phillips Curve was hailed as a great policy tool in the early 1960s but fell apart within a decade.

In the meantime, I maintain that high debt-income ratio is harmful to growth till someone carries out a thorough econometric modelling exercise showing otherwise.

The author is a prominent economist and Labour peer