Predicting Currency Crisis: An Analysis on Early Warning System from Different Perspective
​
Nor Azuana Ramli
​
​
Abstract. Numerous market analysts and researchers have attempted to foresee currency crisis by using diverse methodologies although some of them believe the currency crisis cannot be predicted. Varieties of methodologies from different fields have been introduced to model currency crisis but the problem highlighted here is the prediction results is not very accurate. In this paper, an early warning system model based on Chi-squared Automatic Interaction Detector (CHAID) algorithm is used to see if it can gives better accuracy in predicting such crisis. For comparison, logistic regression is chosen to be applied as it is one of the popular methods in application of early warning system for currency crisis. Variables for this study are taken from the International Country Risk Guide instead of using macroeconomic indicators. By using data sets on 8 countries from first quarter 1984 until fourth quarter 2011, the results from this analysis showed CHAID algorithm present better accuracy in predicting currency crisis than logistic. It also proved that it is not impossible to predict any crisis as long as the right methodology and indicators are used in an analysis. The CHAID algorithm can be used in developing early warning system for banking, economic or debt crisis since it gives good prediction results.