Multivariate Transvariation Analysis and Currency Crises

The aim of this paper is to understand currency crises by selecting the macroeconomic fundamentals with a high predictive power, defining the latter as the variable ability of discriminating between two groups of countries: the sound and the distressed. We consider a sample of over one hundred countries which experienced a currency crash, following Frankel and Rose (1996) and we apply the statistical methodology, transvariation analysis, which measures the amount of overlap between the distributions of the sound and the distressed countries. The result of this methodology is a ranking of the groups of variables who tend to better distinguish the two groups. In order to test the informative power of the selected indicators, we calculate the leave one out predictive error using a transvariation based linear discriminant function. Our results seem to outperform Frankel and Rose (1996) probit analysis.