Thursday, April 18, 2013

Reposting from my Forbes blog: the debate on Debt and GDP


Within the past day or so, economics conversations have been all about Rogoff and Reinhart and their critics, Herndon, Ash and Pollin.  The Rogoff and Reinhart (RR) paper purported to show that countries with more debt grow more slowly than countries with less; Herndon, Ash and Pollen (HAP) show that Rogoff and Reinhart’s data contains mistakes, and there is not much dispute about whether Herndon, Ash and Pollin’s corrections are right–they are.
HAP also do a pretty good job of showing that connections between debt to gdp ratio are not robust–they are sensitive to time period and country.  But they do not ask the question about direction of causality between debt and growth (page 3):
For the purposes of this discussion, we follow RR in assuming that causation runs from public debt to GDP growth. RR concludes, “At the very minimum, this would suggest that traditional debt management issues should be at the forefront of public policy concerns” (RR 2010a p. 578). In other work (see, for example, Reinhart and Rogo (2011)), Reinhart and Rogo acknowledge the potential for reverse causality, i.e., that weak economic growth may increase debt by reducing tax revenue and increasing public expenditures. RR 2010a and 2010b, however, make clear that the implied direction of causation runs from public debt to GDP growth.
But the question of direction matters a lot.  Consider a country whose GDP weakens–both tax revenues fall and social spending (on things like unemployment insurance) rises.  This means that in the absence of a policy change, weak GDP leads to higher debt.
There is a simple way to take a first cut at the question of direction of causation–by using a technique known as Granger Causality.  The set up is to try to explain something (such as GDP growth) by looking at its own lagged values and the lagged values of another variable (such as debt-to-GDP ratio).  I took the  data set in Herndon, Ash and Pollen and ran Granger tests using one lag explaining real GDP growth and debt-to-GDP ratios; I ran separate regressions for each country in the data set. I tested for significance at the 90 percent level of confidence.  I am happy to share my results with anyone who is interested (richarkg@usc.edu).
In the tests where I was exploring whether debt-to-GDP “caused” GDP growth, I found that debt’s impact was negative in five countries (AustriaGermany,ItalyJapan and Portugal); positive in four countries (Australia, Canada, New Zealand and Norway), and zero in 11 countries (Belgium, Denmark, Finland, France, Greece, Ireland, the Netherlands, Spain, Sweden, the UK and the US; although France was close to being statistically negative).
RR emphasize that there is a critical point at which debt becomes toxic, and that is at a debt-to-GDP ratio of more than 90 percent.  Doing Granger tests using this variable (on “on-off switch” for a country being at greater than 90 percent), we find that the impact of greater than 90 percent debt on GDP growth is positive in two cases (Australia and New Zealand), and is not statistically different from zero in eight cases (Belgium, Canada, Greece, Ireland, Japan, the UK and the US).  Ten countries have not had debt-to-GDP ratios above 90 percent.
When we look in the other direction, however, the impact of GDP growth on debt is negative 12 times (Australia, Austria, Belgium, Denmark, Finland, Germany, Greece, Ireland, Italy, Japan, Netherlands, and Sweden) and is not statistically different from zero in the eight other countries (Canada, France, New Zealand, Norway, Portugal, Spain, the UK and the US).  Reverse causality IS a big issue here, and until it is really sorted out, we can’t say what the true, structural relationship between GDP and debt really is.