A practical guide to volatility forecasting through calm and storm

We present a volatility forecasting comparative study within the autoregressive conditional heteroskedasticity (ARCH) class of models. Our goal is to identify successful predictive models over multiple horizons and to investigate how predictive ability is influenced by choices for estimation window length, innovation distribution, and frequency of parameter reestimation. Test assets include a range of domestic and international equity indices and exchange rates. We find that model rankings are insensitive to the forecast horizon and suggestions for best practices emerge. While our main sample spans from 1990 to 2008, we take advantage of the near-record surge in volatility during the last half of 2008 to ask whether forecasting models or best practices break down during periods of turmoil. Surprisingly, we find that volatility during the 2008 crisis was well approximated by predictions made one day ahead, and should have been within risk managers’1% confidence intervals up to one month ahead.