An eclectic approach to currency crises: drawing lessons from the EMS experience

Abstract

This article examines the regime changes in the European Exchange Rate Mechanism (ERM), by applying the duration model approach to quarterly data of eight currencies participating in the ERM, covering the complete European Monetary System history. We first make use of the nonparametric (univariate) analysis, finding that the probability of maintaining the current regime decreases very rapidly for the short durations to register then smoother variations as time increases. Second, we apply a parametric (multivariate) analysis to investigate the role of other variables in the probability of a regime change. In particular we consider three alternative theoretical frameworks to select potential explanatory variables: first- and second-generation models of currency crisis and an eclectic model that combines the explanatory variables suggested by both models. Our results suggest that the Weibull specification of the eclectic model would be the more appropriate to fit our data set, finding that the real exchange rate, the interest differentials and the central parity deviation would have negatively affected the duration of a given regime, while credibility, the level of international reserves and the price level in the anchor country would have positively influenced such duration. Finally, we do not find evidence of observed heterogeneity associated to currencies with different behaviour in the sample, nor the existence in our sample of unobserved heterogeneity caused either by misspecification or omitted covariates.

Acknowledgements

The authors are very grateful to Mark Taylor and an anonymous referee for useful comments and suggestions on a previous draft of this article. Sosvilla-Rivero also gratefully acknowledges financial support from the Spanish Ministry of Science and Technology (SEJ2005-09094/ECON). The views expressed here are those of the authors and not necessarily those of the institutions with which they are affiliated.

Notes

¹ Additional regression analyses (available at request) were run considering these three currencies and both point estimates and their SD were very similar. This result supports the elimination of these three currencies from the original sample.

² In the LIT case, we also consider as change its temporary exit in the third quarter of 1992 and its re-entrance in the fourth quarter of 1996.

³ This kind of data is frequently encountered in biomedical and other investigations. In these studies, failure times are correlated within cluster (subject or group), violating the independence of failure times assumption required in traditional survival analysis. In our case, the 64 changes are distributed among currencies as follows: 11 for the IRL, 10 for the LIT and the DKR, 9 for the BFR, 8 for the FF and the HFL and 4 for the PTA and the ESC.

⁴ Duration models have been widely used in the economic analysis of poverty (see, e.g., Cao, 1996; Finnie, 2000; Jenkins and Rigg, 2002; Hansen and Wahlberg, 2004). It has also been used in the field of Industrial Organization, to analyse for example the life duration of multinational subsidiaries in the UK manufacturing industry (McCloughan and Stone, 1998), or to analyse investment ages (Licandro et al., 1999). See also Sosvilla-Rivero and Maroto-Illera (2003) for a detailed study of the weekly duration of exchange rates regimes in the EMS and Jirasakuldech et al. (2006) for the use of nonparametric duration dependence tests to explore the possibility of rational expectations bubbles on exchange rates.

⁵ In our case, this variable measures the time that passes between two consecutive regime changes in the ERM.

⁶ All estimations were performed using the statistical package STATA 8.1, which permits an appropriate and easy treatment of time-varying covariates.

⁷ The exact definition of the variables as well as the data sources are detailed in the Appendix.

⁸ Alternatively, we would have applied nonparametric unobserved heterogeneity model proposed by Heckman and Singer (1984). However, this model is quite sensitive to the number of mixture points in the distribution of unobserved heterogeneity (Paserman, 2006). Furthermore, Chamberlain (1985) and Yamaguchi (1986, 1991), among others have pointed that by assuming independence among unobserved and unobserved factors, the omitted variable bias, or selection bias, will generally remain.