06 April 2011

Does Government Debt Reduce Growth?

Kenneth Rogoff, a very respected economist, repeated in the FT his claim that there is a threshold for government debt above which growth begins to slow.

According to my recent research with Carmen Reinhart, debt-to-income ratios are already at, or near, postwar highs across advanced economies. Many are close to the roughly 90 per cent debt-to-income threshold which, historically, begins to be associated with lower growth.
This claim has been questioned by others who point out that the result does not come from the well researched book This Time Is Different, but from another short paper. Paul Krugman in his blog has made some good points challenging the research. (link added) In particular Krugman questions the direction of causality. Does low growth increase debt, not just becase of automatic stabilisers but also because the numerator in the debt ratio is lower?

I am also concerned at how the idea of a threshold is arrived at. The paper simply slices the data into four sets where the debt ratio is below 30%, 30%-60%, 60%-90% and above 90% and then compares median and mean growth rates. So the 90% "threshold" is manufactured by the methodology. It does not emerge from the data.

I don't have access to the dataset they used and so I decided to find an easy to assemble dataset which could be used to test for a threshold around 90% debt to GDP ratio. The main criterion would be a set of data including a number of countries and periods when government debt was high.

Looking into Eurostat I was able to find the data for the 15 countries which were already members of the EU in 1996 and looked at the numbers for general government gross debt and growth rates in each year between 1996 and 2007 inclusive. I thought it reasonable to cut off the data in 2007 which is both before the disruptions of the Great Recession, and also because the data is less likely to be revised in future.

This gave me 180 data points which I put on a scatter chart. Here is the result, click on the chart for a closer look:


I see no sign of a threshold at or near the 90% debt ratio. You might see a slight correlation between high debt and low growth, but there is not much. Trying a linear regression gives an R2  of 0.09 which is not significant and so I haven't added a trendline.

I don't know how to show the median lines using Excell, but I did calculate that the median growth rate in the dataset is 3.05%. Of the 34 data points which lie above the 90% debt ratio there are 15 above and 19 below this median. So this data does not support Reinhart and Rogoff's claim.

Of course other data sets should be used, say EU 27 plus Switzerland and the non EU Nordic countries, and over a longer period. I look forward to seeing the results of such research.