Tail behaviour of the euro

Abstract

This paper empirically analyses risk in the euro relative to other currencies. Comparisons are made between a subperiod encompassing the final transitional stage to full monetary union with a subperiod prior to this. Stability in the face of speculative attack is examined using Extreme Value Theory to obtain estimates of tail exchange rate changes. The findings are encouraging. The euro's common risk measures do not deviate substantially from other currencies. Also, the euro is stable in the face of speculative pressure. For example, the findings consistently show the euro being less risky than the yen, and having similar inherent risk to the Deutsche mark, the currency that it is essentially replacing.

Acknowledgements

The author acknowledges the finanical support from University College Dublin President's awards. The author would like to thank Stephen Figlewski, Donal McKillop, Kate Phylaktis and participants of New York's University Salomon Center's conference on the ‘Euro: Valuation, Hedging and Capital Market Issuses’, the European Financial Management Association Annual Meetings and the Statistical and Social Inquiry Society of Ireland for their helpful comments.

Notes

More recently the Asian currency crises sent warning signals as to the stability of nominal exchange rates (for a review see Corsetti et al., 1998).

These extrapolated estimates are described as out-of-sample estimates in the Extreme Value Theory literature.

Jensen's inequality says that for the exchange rate, s, expressed in foreign currency per unit of the local currency, E(1/s) ≠ 1/E(s). For example, whilst the DM/US rate was the reciprocal of the US/DM rate, their expected values were not reciprocals (Diebold, 1988).

The Deutsche mark is incorporated as a precursor to the euro.

This may be due to investors perceiving that the euro is a continuation of the German currency.

The BHHH algorithm is used for convergence after the simplex method provides the initial starting values. The starting values for the conditional variance are given by the unconditional variance measures. The model was fitted assuming the conditional t-distribution to allow for fat-tails (Baillie and Bollerslev, 1990). AIC and BIC criteria were used for model selection.

This interpretation is given in the context of comparing different exchange rate regimes, fixed and floating, using tail estimates (Koedijk et al., 1992).

The Extreme Value estimates cover fully the final transitional period, 1999–2001, to a full monetary union.