Daily FX volatility forecasts: Can the GARCH(1,1) model be beaten using high-frequency data?

Citation

McMillan D & Speight AEH (2012) Daily FX volatility forecasts: Can the GARCH(1,1) model be beaten using high-frequency data?. Journal of Forecasting, 31 (4), pp. 330-343. https://doi.org/10.1002/for.1222

Abstract
Volatility forecasting remains an active area of research with no current consensus as to the model that provides the most accurate forecasts, though Hansen and Lunde (2005) have argued that in the context of daily exchange rate returns nothing can beat a GARCH(1,1) model. This paper extends that line of research by utilizing intra-day data and obtaining daily volatility forecasts from a range of models based upon the higher-frequency data. The volatility forecasts are appraised using four different measures of 'true' volatility and further evaluated using regression tests of predictive power, forecast encompassing and forecast combination. Our results show that the daily GARCH(1,1) model is largely inferior to all other models, whereas the intra-day unadjusted-data GARCH(1,1) model generally provides superior forecasts compared to all other models. Hence, while it appears that a daily GARCH(1,1) model can be beaten in obtaining accurate daily volatility forecasts, an intra-day GARCH(1,1) model cannot be.

Keywords
volatility forecasts; exchange rates; intra-day data

Journal
Journal of Forecasting: Volume 31, Issue 4

People

Professor David McMillan

Professor in Finance, Accounting & Finance