Are emerging market currency crises predictable? A test

This paper analyzes the predictability of emerging market currency crises by comparing the often used probit model to a new method, namely a multi-layer perceptron artificial neural network (ANN) model. According to the results, both models were able to signal currency crises reasonably well in-sample, but the forecasting power of these models out-ofsample was found to be rather poor. Only in the case of Russian (1998) crisis were both models able to signal the crisis well in advance. The results reinforced the view that developing a stable model that can predict or even explain currency crises is a challenging task.