Forecasting exchange rate volatility: a multiple horizon comparison using historical, realized and implied volatility measures

Recent studies suggest realized volatility provides forecasts that are as good as option-implied volatilities, with improvement stemming from the use of high-frequency data instead of a long-memory specification. This paper examines whether volatility persistence can be captured by a longer dataset consisting of over 15 years of intra-day data. Volatility forecasts are evaluated using four exchange rates (AUD|USD, EUR|USD, GBP|USD, USD|JPY) over horizons ranging from 1 day to 3 months, using an expanded set of short-range and long-range dependence models. The empirical results provide additional evidence that significant incremental information is found in historical forecasts, beyond the implied volatility information for all forecast horizons. Copyright © 2008 John Wiley & Sons, Ltd.