The Signals Approach to Early Warning of Currency Crisis: Does It Work and Is There a Better Way?

Abstract

In recent years, a number of researchers have claimed success in systematically predicting which countries are more likely to suffer currency crises, most notably Kaminsky, Lizondo, and Reinhart (1998) and its extension in Kaminsky ((1998). In this note, we assess the success of this approach, emphasizing out-of-sample testing. First, we try to answer the question: If we had been using the KLR model in late 1996, how well armed would we have been to predict the Asia crisis? Second, we analyze a more general probit-based model for predicting currency crises. In the process, we test several basic assumptions underlying the signals approach.