Have Exchange Rates Become More Closely Tied? Evidence from a New Multivariate GARCH Model

We analyze the time-dependence of exchange rate correlations using a new multivariate GARCH model. This model consists of two parts. First, we transform the exchange rate changes into their principal components and specify univariate GARCH models for all components. Second, we use the inverse of the principal components construction to transform the conditional component moments back into those of the exchange rate changes themselves. The model is easy to estimate, as it requires only univariate GARCH estimations. Nevertheless, it outperforms the popular constant conditional correlations and factor GARCH models. We …nd that the major U.S. dollar exchange rates have become more loosely instead of closely tied since the eighties. Key words: correlations, multivariate models, GARCH, factor models, exchange rates. JEL classi…cation: C32, C52, F31. ¤Correspondence to: Franc Klaassen, Department of Econometrics, Tilburg University, P.O. Box 90153, 5000 LE Tilburg, the Netherlands; tel: +31-13-4668229; fax: +31-13-4663280; E-mail: [email protected]. I would like to thank Frank de Jong, Harry Huizinga, Theo Nijman, Bertrand Melenberg, Siem Jan Koopman and Bas Donkers for their helpful comments.