Volatility forecasts: the role of asymmetric and long-memory dynamics and regional evidence

This article seeks to examine the forecasting performance of nine competing models for daily volatility for stock market returns of 33 economies. Whilst volatility is an important variable in many financial applications including those relating to areas of risk management there exits little consensus with regard to the most appropriate model. The results of this article seek to bring some closure to the debate. Our results suggest that in 70% of our cases the GARCH-class of model provide the best forecasts and in particular models that account for either asymmetry or long-memory dynamics. Outwith the GARCH-class, the moving average model provides reasonable forecasts.