This paper focuses on nominal exchange rates, specifically the US dollar rate vis-à-vis the Euro and the Japanese Yen at a daily frequency. We model both absolute values of returns and squared returns using long-memory techniques, being particularly interested in volatility modelling and forecasting given their importance for FOREX dealers. Compared with previous studies using a standard fractional integration framework such as Granger and Ding (1996), we estimate a more general model which allows for dependence not only at the zero but also at other frequencies. The results show differences in the behaviour of the two series: a long-memory cyclical model and a standard I(d) model seem to be the most appropriate for the US dollar rate vis-à-vis the Euro and the Japanese Yen respectively.
Fractional integration long memory exchange rates volatility