Improving (E)GARCH forecasts with robust realized range measures: Evidence from international markets

This paper empirically shows that (E)GARCH volatility forecasts may be improved by inserting an appropriate exogenous variable in the volatility equation. Several realized measures were tested as regressors and the robust to microstructure effects and/or jumps realized range-based measures provided the best results. The out-of-sample forecasts were computed in two steps. Firstly, the (E)GARCH-X model was estimated. Secondly, an ARFIMA forecast of the realized range was plugged into the volatility prediction formula. The methodology was illustrated in a comprehensive study involving fifteen market indices from developed stock markets. It was also shown that the inclusion of realized range-based measures as regressors reduces persistence and renders the past squared returns with no remaining explanatory power. We use two evaluation criteria to compare the forecasting performance of the (E)GARCH-X model and the Realized GARCH model. © 2017, Springer Science+Business Media New York.