VoxEU Column Global crisis

External liabilities and crisis risk

Debt seems to be a lightning rod for crises. This column presents new research showing that the ratio of net foreign liabilities to GDP, and in particular its net external debt component, is indeed a significant crisis predictor for both advanced economies and emerging markets. Large current-account deficits and real exchange rate appreciation – the standard predictors – still matter, but we should be thinking more about net external debt.

Much has been written about the causes of the global financial crisis of 2008 – the role of the US subprime crisis as a triggering event, the generalised period of easy credit and financial excesses fuelling growing economic and financial vulnerabilities, the failures to properly regulate large systemic financial institutions. Nevertheless, our understanding of the intensity with which the crisis has affected different countries remains modest (Rose and Spiegel 2009, 2011). A look back at the advanced and emerging economies that before the crisis had large debtor positions vis-à-vis the rest of the world – more precisely, large net foreign liabilities – shows that a large number of those were forced to seek external support and/or restructure their debt (Figure 1, red bars).1

Figure 1. Net foreign liabilities of selected countries prior to the 2008-09 crisis

In a new paper (Catão and Milesi-Ferretti 2013) we seek to understand whether the size and composition of a country’s net external liabilities do indeed help predict an external crisis. We also ask whether – based on the history of external crises up to 2006 – a simple model including these variables would have correctly pointed out the countries most severely engulfed in the global financial turmoil of the Great Recession.

There is a long and distinguished literature on currency crises (Frankel and Rose 1996, Eichengreen et al. 1996, Kaminsky and Reinhart 1999, Berg and Pattillo 1999, Gourinchas and Obstfeld 2010, Frankel and Saravelos 2012). While this literature has identified significant factors (such as real exchange rate appreciation, large current-account deficits, and, for emerging and developing economies, large external debt), it finds it difficult to predict crises ‘ex ante’ and even more ‘out of sample’ (that is, predicting a crisis that we know has occurred using data only for the period preceding the crisis).

We revisit this question by including size and composition of external liabilities among the variables potentially explaining a crisis. Our analysis also encompasses the recent wave of crises in the wake of the Great Recession, and focuses on major external crises – specifically, default/re-scheduling of sovereign obligations and/or sizeable multilateral financing, measured as IMF loans of at least twice the size of the respective country quota. It includes advanced economies and emerging markets, excluding instead low-income countries.

Our first pass at this panel dataset suggests that higher net external liabilities do raise crisis risk. The first panel in Figure 2 plots the average value of net foreign assets as a share of GDP for countries affected by a crisis, within a window starting five years before the crisis and ending five years after. The panel singles out the average behaviour around crises for both the period 1970-2006 (blue line) and the most recent set of crises (red line). It clearly shows that before crises net foreign assets are negative, deteriorating, and large (50% of GDP for ‘older’ crises and 70% of GDP for recent crises).

But are all forms of external liabilities the same? Does it matter whether a country’s assets are held by the private sector or by the central bank in the form of reserves? And how about other factors – the level of economic development, large current-account deficits, an appreciated currency, and unstable external environment – that the literature has identified as important factors in affecting the likelihood of an external crises? Finally, once we take into account level and composition of net foreign liabilities as well as a handful of other macro variables, can we do a better job at predicting crises out of sample? Let’s tackle these questions one at a time.

Figure 2. Cross-country means of external sector variables around external crises

In-sample evidence on external crisis determinants

We disaggregate net foreign assets into their debt and equity components. Specifically, we calculate the difference between external debt assets (such as foreign debt securities or foreign deposits held by domestic residents, including the central bank) and debt liabilities (domestic debt securities held by nonresidents, foreign loans etc). We do the same for the net equity position, comprising portfolio equity (foreign shares held by residents minus domestic shares held by foreigners) and foreign direct investment (foreign firms controlled by domestic residents minus domestic firms controlled by foreign residents). First and foremost, we find that the net external debt component is the one that matters: crises are preceded by periods of high and increasing net debt liabilities (second panel of Figure 2). This intuitive result had not been established for both advanced and emerging economies in previous work. To be sure, it has been documented that countries with large external debt positions tend to be more vulnerable, but the focus has been on gross external debt (see, e.g., Reinhart and Rogoff 2010). However, this result was obtained in samples that crucially excluded advanced economies – since the latter tended to have much fewer crises and much larger gross external debt, reflecting the extent of their financial integration. When we enter external debt assets and external debt liabilities separately in probit regressions on external crisis determinants, their estimated coefficients have a similar magnitude and opposite sign, indicating that net debt is what fundamentally matters.

In contrast, we find no systematic link between a country’s net equity position and crisis risk (third panel of Figure 2) – the most notable stylised fact being the larger level of net equity liabilities in recent crises, reflecting the boom in foreign direct investment and portfolio flows during the past decade. The fourth panel suggests that large current-account deficits are another salient feature of pre-crisis periods, and all the more so for the most recent crises.

When we put all these variables together to estimate the probability of a crisis, our results suggest that net external debt is indeed an important explanatory variable for crisis risk. Current-account deficits, an appreciated currency, lower per capita income, and an unstable external environment (reflected in high spreads and financial market volatility), also contribute to raising the risk of a crisis. And finally, a country’s reserves do provide some insulation against the risk of a crisis, even holding constant the net external position of the country. This makes sense – after all, it is easier for the central bank to mobilise reserves in a difficult situation than to rely, for example, on deposits overseas by private individuals or firms. Overall, our simple model does a very good job in explaining crises according to a variety of metrics. Adding additional variables to our model does not really improve performance – even for variables like domestic credit growth that several studies (such as Jordá et al. 2011, and Shularick and Taylor 2012) found as an important determinant of crises. We do find that higher credit growth increases the risk of a crisis when taken as its sole determinant, but its effects are captured through other variables, such as the current-account balance and the net external debt position, that are already included in our model (a finding consistent with Bruno and Shin 2013, and Lane and McQuade 2013).

Could we have predicted recent crises?

As mentioned above, early warning models of currency crises are widely perceived as failing out of sample: too many false alarms and bad misses, such as the 1997/98 Asian crises. How does our model fare in this respect?

Using the parameters estimated over 1970-2006, Figure 3 reports the estimated probabilities of crisis for each year after 2006 for both countries that actually experienced a crisis as well as countries whose probability of crisis according to the model was ‘high’ (18% or above). The model correctly predicts Greece and Portugal, and also singles out Spain as a high-risk case. It also correctly predicts the crises in the Dominican Republic, Jamaica, Latvia, Romania (at an 18% threshold), and Serbia. While it gives a ‘false’ alarm for Lithuania (predicting a crisis that did not occur according to our definition), the model correctly picks up the major recession that ensued. Ecuador and Hungary are instead clear misses (in the form of a very low point estimates for crisis probability one year ahead), whereas Turkey is a rather special case because the disbursement of the pre-approved final tranche of IMF lending in 2008 brought its IMF exposure over 200% of quota which our coding classifies as crisis, even though country risk was clearly dropping.

Figure 3. Estimates of crisis probabilities for 2008-11 (out of sample)

Conclusions

We find these results encouraging, both because of the explanatory power of our model and because the variables we selected to explain external crises were perfectly plausible candidates before the occurrence of the Great Recession. Indeed, our model does not contain measures of banking exposures and other financial vulnerabilities that have been highlighted as key factors in explaining the most recent wave of crises. Yet we need to be modest – we do have the benefit of hindsight and our model misses crises in countries with clear vulnerabilities. Predicting crises remains difficult work.

Editor’s note: Disclaimer: The views expressed here are those of the authors and do not necessarily represent those of the institutions with which they are affiliated.

References

Berg, A and C Pattillo (1999), “Are Currency Crises Predictable? A Test”, IMF Staff Papers 46, 107-138.

Bruno, V and H Shin (2013), “Global Factors in Capital Flows and Credit Growth”, VoxEU.org, 7 July.

Catão, L and G Milesi-Ferretti (2013), “External Liabilities and Crises”, IMF WP 13/113.

Eichengreen B, A Rose, and C Wyplosz (1996), “Contagious Currency Crises”, NBER WP 5681.

Frankel, J and A Rose (1996), “Currency Crashes in Emerging Markets: An Empirical Treatment”, Journal of International Economics 41, 351–66.

Gourinchas, P-O and M Obstfeld (2012), "Stories of the Twentieth Century for the Twenty-First", American Economic Journal: Macroeconomics 4(1), 226-65.

Jordá, O, A Taylor and M Schularick (2011), “Financial Crises, Credit Booms, and External Imbalances: 140 Years of Lessons”, IMF Economic Review 59, 340-378.

Kaminsky G, S Lizondo, and C Reinhart (1998), “Leading Indicators of Currency Crises”, IMF Staff Papers 45, 1-48.

Lane, P and P McQuade (2013), “Domestic Credit Growth and International Capital Flows”, Scandinavian Journal of Economics, forthcoming.

Reinhart, C and K Rogoff (2010), “Growth in a Time of Debt”, American Economic Review, Papers and Proceedings 100, 573-578.

Rose, A and M Spiegel (2009), “Could an Early Warning System Have Predicted the Crisis?”, VoxEU.org, 3 August.

Rose, A and M Spiegel (2011), “Cross-Country Causes and Consequences of the Crisis: An Update”, European Economic Review 55, 309-324.

Shularick, M and A Taylor (2012), “Credit Booms Gone Bust: Monetary Policy, Leverage Cycles, and Financial Crises, 1870-2008”, The American Economic Review 102(2), 1029-61.


1 A country’s NFL position consists of the difference between the claims of a country’s residents on the rest of the world and the claims of the rest of the world on the country’s residents. External assets and liabilities can take the form of debt contracts (for example, a bond issued by the domestic government and bought by a foreign investor) or equity contracts (for example, the purchase by a domestic resident of a share issued by a foreign company or the foreign direct investment of a domestic multinational in an affiliate located overseas).

2,100 Reads