The Great Recession was the most important macroeconomic event in a generation. Policymakers are scrambling to guard against any repeat of these cataclysmic events.
One strand of this effort is aimed at setting up an early warning system. The idea is to provide early warnings of macroeconomic or financial vulnerabilities and impending danger. This work seeks to create statistical models that link crisis causes to their effects; it is based on econometric models of the incidence of macroeconomic and financial crises.
Can an early warning system work?
These models, however, can work no better than our understanding of the causes of crises. But how much do we actually know about the causes of crises, even major ones like the Great Recession? Is such a warning system led by the IMF and others likely to succeed?
Many believe that an early warning system can potentially provide helpful information to policymakers.
- Frankel and Saravelos (2010) argue that a variety of early warning indicators, especially international reserves, help understand why certain countries were hit harder during the Great Recession.
- Lane and Milesi-Ferretti (2010) and Berkmen et al. (2009) point to cumulative credit growth and fixed exchange rates affect crisis severity.
- Giannone et al. (2010) provide evidence that countries with more liberal credit market regulations suffered more serious crises.
- Both Blanchard et al. (2010) and Lane and Milesi-Ferretti (2010) find that countries whose trading partners declined further during the crisis fared worse themselves.
- Blanchard et al. (2010) demonstrate that the fraction of GDP held in short-term external debt is closely negatively linked to growth for developing countries.
It might seem like the problem faced by those constructing early warning systems lies in winnowing down the vast set of plausible causes of the crisis, rather than in finding ones that work.
New evidence: Warning on the warning system
Our recent work leads us to be sceptical. We restrict ourselves to examining the cross-country incidence of the Great Recession – that is, we ask why countries like Iceland and Latvia have been hit so hard while China and Brazil have not. We do this since it has historically proven difficult to link crisis causes to their timing; it is easier to link the cross-country incidence of crises to their causes. Of course a successful early warning system must predict both the incidence and the timing of future crises, but one might as well to try to solve the easier problem first.
It is easier to understand the cross-country incidence of the Great Recession, but it is not easy. A review of the developing literature on the causes of the crisis brings up many positive results. This researcher or that is able to understand some manifestation of the crisis using one or more potential causes.
But our statistical analysis shows that these results are often sensitive to a variety of apparently innocuous assumptions. Our robustness checks examine the collective set of all the variables that have been successfully linked by other researchers to the cross-country incidence of the crisis. We show that many positive results in the literature are untrustworthy, since they tend to be sensitive to things like the precise measures of the intensity of the crisis and the exact sample of countries examined.
The six most successful variables in understanding the Great Recession are charted in Figure 1. This scatters the 2008-09 growth in real GDP across countries (on the vertical-axis) against six potential causes of the crisis that seem to work consistently well. They are:
- the 2006 current account (expressed as a percentage of GDP);
- the 2006 measure of credit market regulation estimated by the Heritage Foundation;
- 2000-06 growth in bank credit (again as a percentage of GDP);
- the country’s trade with the US as a fraction of its 2006 trade;
- the 2003-06 growth in stock market capitalisation; and
- the log of 2006 real GDP per capita.
Even though these variables have been carefully chosen, none are tightly correlated with the cross-country incidence of the Great Recession.
Figure 1. Scatter-plots of 2008-09 growth against early warning indicators
A few kernels of wheat remain after the chaff is removed. The degree of credit market regulation appears to have a strong consistent negative effect on crisis incidence. More short-term external debt (a variable only available for poor countries) and a worse current account are also correlated with more intense Great Recession experiences across countries.
You never go down to the same river twice
Even if we limit ourselves to the few consistent successes, the results turn out to be of little value going forward. We take the model which best fits the Great Recession and ask how well it would have worked earlier.
We examine two global downturns: the 1991 downturn associated with the Gulf War (for Americans) and German Unification (for Europeans), and the 2001 “dot.com” recession. Unfortunately, we find that even the best estimates gleaned from the Great Recession are not stable predictors of the incidence of global recessions at different periods of time.
Countries with large current-account deficits seemed to suffer more in the Great Recession. The single most robust result of the Great Recession analysis is the negative and significant effect of the credit market regulation variable; looser credit market regulation seemed to hurt recently. But for the two previous global recessions, this effect is often positive and significantly so in a number of the samples; it is never significantly negative. Such sample sensitivity reinforces our scepticism concerning the value of such exercises.
In our earlier research on early warning systems (Rose and Spiegel 2010b and 2010c), we found that it was difficult to reliably link macroeconomic or financial indicators from 2006 or earlier to a variety of financial and real manifestations of the 2008 crisis. Our more recent research corroborates our scepticism. Despite a broad search, we have been unable to find consistent strong linkages between pre-existing variables that are plausible causes of the Great Recession and the actual intensity of the recession.
It is natural for economists to generalise from experiences of a few particularly salient countries to make generalisations, though it is often inappropriate. Our poor results are simply telling us that the pre-conditions for the crisis in the US (or Iceland, or Latvia …) often do not describe other countries particularly well. Credit growth was high before 2008 in Australia, Canada, and South Africa, yet these countries seemed to have weathered the crisis well. Real housing prices actually fell in Japan, Germany and Portugal, yet these countries were hard hit. Since it is difficult to understand the cross-country incidence of the great recession even in retrospect, we are dubious about the potential for a comparable early warning forecasting model going forward.
Berkmen, Pelin, Gaston Gelos, Robert Rennhack and James P Walsh (2009), “The Global Financial Crisis: Explaining Cross-Country Differences in the Output Impact”, IMF Working Paper WP/09/280.
Blanchard, Olivier, Hamid Faruqee and Mitali Das (2010), “The Initial Impact of the Crisis on Emerging Market Countries” unpublished.
Frankel, Jeffrey A and George Saravelos (2010), “Are Leading Indicators of Financial Crises Useful for Assessing Country Vulnerability?”, NBER Working Paper 16047.
Giannone, Domenico, Michele Lenza and Lucrezia Reichlin (2010), “Market Freedom and the Global Recession” unpublished.
Lane, Philip R, and Gian Maria Milesi-Ferretti (2010), “The Cross-Country Incidence of the Global Crisis”, unpublished.
Rose, Andrew K and Mark M Spiegel (2010a), “Cross-Country Causes and Consequences of the 2008 Crisis: An Update”, CEPR Discussion Paper 7901.
Rose, Andrew K and Mark M Spiegel (2010b), “Cross-Country Causes and Consequences of the 2008 Crisis: Early Warning”, Global Journal of Economics,forthcoming.
Rose, Andrew K and Mark M Spiegel (2010c), “Cross-Country Causes and Consequences of the 2008 Crisis: International Linkages and American Exposure”, Pacific Economic Review, forthcoming.