A Cardan's discriminant approach to predicting currency crashes

This paper models large swings in exchange rates by introducing nonlinearity via the generalized normal (GEN) distribution [Lye, J.N., Martin, V.L., 1993. Robust estimation, nonnormalities and generalized exponential distributions. Journal of the American Statistical Association 88, 261-267]. As the distribution allows for bimodality, a switch between modes may give rise to currency crashes. A statistic known as Cardan's discriminant, based on the shape parameters of the GEN, is used to detect bimodality. The Cardan's discriminant is found to reliably predict currency crashes for eight emerging countries and generate relatively low false signals for stable currencies.