آیا نرخ ارز دو گانه و بحران بدهی خاص هستند؟

In the literature on currency and banking crises it has become the standard procedure to distinguish pure currency crises, pure banking crises and combined (“twin”) currency and banking crises. We show theoretically and empirically that a similar differentiation should be chosen with regard to currency and debt crises. Twin currency and debt crises differ from both pure currency and pure debt crises in their determinants, course of events, and economic consequences. We find that each type of crises has a unique set of macroeconomic causes. We also identify internal contagion and selection bias effects, which may lead to biased empirical estimates if twin crises are not treated as a specific type of crises. Such a separation allows in significantly improving the efficiency of early warning systems especially for debt and twin crises.

In the literature on currency and banking crises it has become the standard procedure to treat pure currency crises, pure banking crises and combined (“twin”) currency and banking crises as separate classes of crises. For example, in empirical studies the output effects and frequencies are usually analyzed separately for each of these crises.1 In contrast, in the literature on debt and currency crises such a distinction is less common- even though empirically twin debt and currency crises are at least as frequent as the more prominent twin currency and banking crises.2 Therefore, we address the following question: Should twin debt and currency crises also be regarded as a specific type of crisis, which should be analyzed separately from pure currency and pure debt crises (see Table 1)? Can such a classification improve the analysis of fiscal and exchange rate crises? Most importantly, does it on the empirical side improve the quality of early warning indicators and the prediction of financial crises? If debt and currency crises are interrelated due to common causes and/or direct contagion effects from one crisis to the other, an explicit consideration of these interrelations may indeed enhance the results of empirical analyses and the forecasts of early warning systems.In the literature on currency and banking crises it has become the standard procedure to treat pure currency crises, pure banking crises and combined (“twin”) currency and banking crises as separate classes of crises. For example, in empirical studies the output effects and frequencies are usually analyzed separately for each of these crises.1 In contrast, in the literature on debt and currency crises such a distinction is less common- even though empirically twin debt and currency crises are at least as frequent as the more prominent twin currency and banking crises.2 Therefore, we address the following question: Should twin debt and currency crises also be regarded as a specific type of crisis, which should be analyzed separately from pure currency and pure debt crises (see Table 1)? Can such a classification improve the analysis of fiscal and exchange rate crises? Most importantly, does it on the empirical side improve the quality of early warning indicators and the prediction of financial crises? If debt and currency crises are interrelated due to common causes and/or direct contagion effects from one crisis to the other, an explicit consideration of these interrelations may indeed enhance the results of empirical analyses and the forecasts of early warning systems.

Our theoretical and empirical evidence strongly suggests to regard twin debt and currency crises as a specific type of crisis, which should be analyzed separately from pure currency and pure debt crises. Such a classification helps to better understand the interrelations between fiscal and exchange rate crises. Each of the three types of crises – pure currency, pure debt and twin debt and currency crises – is characterized by a unique set of causes and consequences. Theory suggests a number of possible interrelations between debt and currency crises, which include common causes, contagion effects from one crisis to the other, and complementary budget financing aspects. Recapitulating the results of our GLM estimations this new approach to treat twin crises as a distinct type of crisis also significantly improves the predictive power of early warning systems. Our model predicts 36% of all currency crises, 75% of all debt crises, and 50% of all twin crises correctly, while we only get 1.6% false alarms for the prediction of currency crises, 12% false alarms in the case of debt crises, and 0.3% false alarms for currency crises. With these results our approach performs significantly better than traditional early warning systems on currency and debt crises which typically do not differentiate between pure currency crises (or pure debt crises, respectively) and twin crises. As major reasons for the inferior performance of the traditional approach, we identify internal contagion and selection bias effects, which may lead to biased results if twin crises are not treated separately. While for example in traditional empirical analyses short-term debt is often found to be a significant predictor for debt crises, we show that in fact it has no significant influence on the risk of pure debt crises. It rather increases the probability of currency crises, which then can trigger debt crises via contagion effects. Short-term debt thus seems to be an important explanatory variable in traditional debt crisis samples only because of its relevance for twin crises, which is carried forward to the entire sample. The example of the foreign exchange reserves shows how selection bias effects affect the predictive value of empirical analyses. Reserves have a significant negative influence on the probability of pure debt crises, but a positive influence on twin crises, i.e., pure debt crises are associated with comparatively low reserve ratios, while twin crises are connected with comparatively high reserve ratios. A possible explanation for this result is that high reserve ratios are less important under floating exchange rates than in fixed exchange rate systems. As countries with fixed exchange rates are more likely to become subject to speculative attacks and currency crises in terms of our crisis definition, debt crises are more likely to be associated by simultaneous currency crises in countries with high reserve ratios (i.e., fixed exchange rate regimes) than in countries with low reserve ratios (i.e., flexible exchange rates). We also find that growth seems to raise the probability of pure currency crises, while it does not significantly influence the crisis risk in the combined currency and twin crises sample. A possible explanation for this finding is that countries with weak monetary but strong real fundamentals are likely to be able to confine crises to the monetary sector. Countries with weak monetary fundamentals are in general likely to encounter currency crises, but they are likely to experience a simultaneous debt crisis as well only if growth is low. Thus with higher growth rates the probability of pure currency crises declines simply as the probability of twin crises increases. The differentiation between currency, debt, and twin currency and debt crises improves the performance of early warning systems based on macroeconomic indicators and helps to better identify key variables explaining the economic intuition behind these crises. Thus the definition of crisis events appears to be fundamental for this kind of analysis.