Convergence in exchange rates: market's view on CE-4 joining EMU

Abstract

We empirically analyse currency fluctuations in four central European states (CE-4) against the USD and Euro, employing daily data over 1 January 1994 to 10 October 2005 and constructing a dynamic correlation coefficient based on the estimates of a bivariate generalized autoregressive conditional heteroscedasticity model. We find evidence of convergence in exchange rate volatilities between CE-4 currencies and the Euro. In other words, from the US market's point of view, currencies of the CE-4 region and the Euro tend to behave quite similarly. This degree of synchronicity is in line with the composition of currency baskets and the share of the Euro as a trade-invoicing currency in the CE-4 economies.

Acknowledgements

The authors are grateful to Balázs Égert, Evžen Kočenda, Claudio Paiva and workshop participants at the IMF Institute for providing valuable comments. Special thanks go to Martin Spilka for his kind language assistance. The opinions expressed in the article are those of the authors and are not necessarily endorsed by the Czech National Bank.

Notes

¹ Estonia, Lithuania and Slovenia are in the ERM II since the end of June 2004. Cyprus, Latvia and Malta joined the ERM II in May 2005. A minimum two years of ERM II membership is formally required to qualify for the Euro adoption. The Czech Republic, Hungary, Poland and Slovakia plan to introduce Euro by 2010–2011.

² Bernard and Galati (2000) apply a 6-month rolling window on daily data of the Dow Jones and the US Dollar in order to estimate a time-varying correlation between the stock and the exchange rate markets.

³ Alternatively, Calvet et al. (2004) use the bivariate Markov-switching multifractal model to estimate co-movements in exchange rate volatility for the Deutsche Mark, British Pound and Japanese Yen relative to the US Dollar.

⁴ Due to a high number of unknown parameters (for example, in a bivariate specification it is equal to eleven: (N × (5N + 1))/2), BEKK-GARCH is often accused of complexity, but it is said to be suitable when the number of time series is less than or equal to 4. Otherwise, multivariate GARCH can be substituted by various extensions of a univariate GARCH. One of them is the dynamic conditional correlation (DCC) model, which represents a non-linear combination of a univariate GARCH. Van Dijk et al. (2006) extend the DCC model by allowing for structural breaks in unconditional correlations among five European currencies. For a good survey of various multivariate GARCH specifications, see Bauwens et al. (2006).

⁵ The same correlation coefficient estimated for major currency pairs is also used by Castrén and Mazzotta (2005).

⁶ ECU is used instead of Euro for the pre-1999 period.

⁷ Note that volatility of the estimated dynamic correlation coefficient is relatively low for the Czech Republic and Slovakia, and higher for Hungary and particularly Poland. Such differences in volatility may be due to exchange rate regimes (e.g. crawling pegs in Hungary and Poland; changes in the crawl are interpreted as volatility), the degree of openness and share of trade with the Euro area (lower for Poland), capital flows, etc.

⁸ CE-4 currencies and the Euro can be viewed as a particular game theory case of interaction between the anchor and satellite currencies. See Engelmann et al. (2004) for a broader discussion of interrelations between currencies.

⁹ Information about composition of the currency basket and Russian international reserves is taken from RIA Novosti and ITAR-TASS news.