Exchange rate instability: determinants and predictability

The paper is concerned with exchange rate instability, by which we mean large changes in exchange rates. The paper has two objectives. First, we search for plausible determinants of currency crashes. To do this we examine annual panel data for a large sample of developing countries. The work is non-structural, taking the form of probit regressions which link currency crashes to a variety of candidate causes. We examine a comprehensive set of both foreign and domestic explanatory variables. The list includes: foreign conditions; the vulnerability of the country to a crash; the level of external indebtedness; the composition of this debt; measures of domestic government policy; and measures of the state of well-being of the economy. The second objective of the paper is to examine the predictability of exchange rate crashes. For this, we use quarterly time series data from eight countries in the European Monetary System to calibrate two structural models of speculative currency attacks. The first model relies on a traditional monetary approach to exchange rate determination in a target zone, and the second uses a Markov-switching model with time-varying transition probabilities. Both models are calibrated using nontraditional methods to determine model parameters, as we are primarily interested in our models' ability to predict the exchange rate regime that will prevail in the next quarter. In consonance with the older literature on empirical exchange rate models, we do a reasonable job in-sample of predicting currency crises, but a very poor job of forecasting currency crises one quarter ahead.