Volatility filters for FX portfolios trading: the impact of alternative volatility models

Abstract

Dunis and Miao (2005) find that the addition of volatility filters with RiskMetrics forecasts can improve the performance of moving average convergence and divergence (MACD) models that replicate typical currency fund managers introduced as in Lequeux and Acar (1998). The motivation of this paper is to test whether alternative volatility models forecasts can further improve the models’ performance with such filters. The two alternative volatility forecast models used in this paper are GARCH model as in Bollerslev (1986) and stochastic volatility model with Markov switching (MS) based on Hamilton (1994) and Roche and Rockinger (2003). Our results show that volatility filters using alternative volatility models fail to enhance the performance of the simpler filters using RiskMetrics forecasts in terms of annualized return and Sharpe ratio.

Notes

¹ See, among others, Akgiray (1989), Bollerslev et al. (1992), Pagan and Schwert (1990), West and Cho (1995) and Chong et al. (1999).

² We use the notation of the International Organization for Standardization (IOS) for all the exchange rates considered.

³ Since the EUR/USD exchange rate only exists from 4 January 1999, we follow the approach of Dunis and Williams (2002) to apply a synthetic EUR/USD series from 2 January 1995 to 31 December 1998 combining the spot USD/DEM and the fixed EUR/DEM exchange rate. The synthetic EUR/GBP, EUR/JPY and EUR/CHF are created following the same approach.

⁴ The AFX dynamic currency index is available at www.cibef.com.