On the predictive ability of several common models of volatility: an empirical test on the FOX index

This paper provides empirical evidence supporting the notion that engineering efforts in modelling future volatility are worthwhile. It is found that the more complex Exponentially Weighted (EWMA) and GARCH models yield prevailing weekly and monthly rolled out-of-sample volatility forecasts on the Finnish Options Index. The extension to the current literature is twofold: first, an array of common predictors is comparatively evaluated with a regression-based efficiency test and with several conventional as well as two asymmetric error statistics; secondly, the GARCH is fitted to low frequency return data, eschewing aggregating or scaling daily forecast, and conclude it not to be preferred to the EWMA on weekly volatility frequency, yet superior on the respective monthly.