Duke Economics Working Paper #05-02
Can we improve forecasts of currency crises by using a large number of predictors? Which economic variables are most important to predict currency crises? Are the most important factors that trigger currency crises internal? This paper evaluates the performance of traditional Leading Indicators and a new Diffusion Index (DI) method as Early Warning Systems to monitor the risk and forecast the likelihood of the recent currency crises in East-Asia. We find that, relative to traditional Leading Indicators (such as money multipliers, terms of trade, etc.), the DI performs quite well in real-time, with a higher probability of providing correct signals. We also find that the economic variables that help in forecasting future crises are output growth, the Balance of Payments, interest rates and money growth, and that their relative importance changes dramatically around the time of the crises. Finally, by using a structural Factor Augmented VAR, we show that a U.S. interest rate shock played an important role in triggering the East-Asia currency crises. This suggests that the adoption of a fixed exchange rate regime might be risky for developing countries.
JEL Classification Codes: F30, F37
Keywords: currency crises, forecasting, leading indicators, diffusion index, exchange rates
†January 2005. We would like to thank Lutz Kilian, Oscar Jorda, Alessandro Tarozzi, and seminar participants at the 2004 Conference on Forecasting at Duke, U.C. Riverside, and the 2005 Econometric Society Winter Meetings for comments. Pedro Duarte and Viktor Todorov provided skilled research assistance.
Corresponding author: Barbara Rossi, Department of Economics, Duke University, Durham NC27708 USA.
E-mail: brossi@econ.duke.edu. Tel.: (919) 660 1801. Fax: (919) 684 8974.
PDF version (1.16MB)