Predicting Currency Crisis Contagion from East Asia to Russia and Brazil: An Artificial Neural Network Approach

Abstract

Studies dealing with currency crisis prediction are often vulnerable to data mining and perform poorly when tested on out-of-sample data. This paper suggests an artificial neural network approach to predicting speculative attacks. The properties of the multilayer perceptron are used to develop a method for predicting currency crises. It is then tested whether the speculative attacks in Russia in 1998 and Brazil in 1999 were predictable, given the then recent turmoil in East Asian countries. Overall, it appears that the multilayer perceptron does a better job at predicting currency crises than a logit model.