Distinguishing between observationally equivalent theories of crises

The objective of this paper is to empirically test across alternative, apparently observationally equivalent theories of currency crises. Theories of crises are often difficult to distinguish from each other based on the behavior of commonly used predictors. Using a comprehensive data set on gross external assets and liabilities for 167 countries created bythe World Bank's Latin America and the Caribbean Region and the Development Research Group, this study is able to make a significant move toward redressing this shortcoming. It focuses on identifying potential crisis predictors, as well as testing the validity of the distinct transmission mechanisms implied by various theories of currency crisis. Evidence is presented in support of insurance-based models, suggesting that proxies for contingent liability accumulation are effective crisis predictors.