Skip to main content
No Access

Clustering the changing nature of currency crises in emerging markets: an exploration with self-organising maps

Published Online:pp 24-46

Currency crises are a recurring phenomenon. To increase the understanding of their changing nature, this paper analyses the evolution of currency crises and assesses them in the generation framework of theoretical models. The self-organising map (SOM), a neural network-based clustering and visualisation tool, is used for clustering pre-crisis periods for emerging market economies. The clustering results are used for finding differences in crises between decades, whereafter the decade clusters are compared with the theoretical framework. We conclude that for emerging market economies, this paper shows that the crises in the 1970s and 1980s are of a different nature and that the crises in the late 1990s are, in comparison with the preceding decades, determined by weaker warning signals. This illustrates that predicting the Asian crises with a priori models was a difficult task. Further, the empirical results indicate that the generation framework is not as clear-cut as theory points out.


currency crisis, early warning, financial instability, SOMs, self-organising maps, clustering, changing nature, emerging markets, neural networks, visualisation, emerging economies