PREDICTING CURRENCY CRISES WITH A NESTED LOGIT MODEL

Abstract.  This paper is the first to apply a nested logit model to measure the probabilities of speculative attacks and the probabilities of successful defences by the central banks. This model allows us to predict the probability not only of speculative attacks but also of successful defences, given attacks. It also provides a framework for analysing the degree to which different factors affect the likelihood of attacks and defences. We find strong evidence that external illiquidity and financial fragility are reliable predictors of currency crises. The results shed light on the validity of the three generations of currency crisis models.