ABSTRACT
Contents 9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .174 9.2 The Hybrid Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .176 9.3 Adding Intraday Data to the Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . .183 9.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .193 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .193 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .194
Abstract: Despite significant advances in volatility forecasting over the last 20 years, since the first introduction of ARCH and GARCH models, comparatively little attention has been paid to issues such as the sensitivity of out-of-sample volatility forecasts to changes in structural components of the models. We propose and estimate a hybrid GARCH model where the parameters are Kalman filtered, and the model is assessed and compared with standard GARCH methods, using several measures of efficacy. The framework is then extended to cope with intraday data, and we investigate whether working with intraday summaries brings any informative advantage in comparison to working with daily data only. As this chapter does not concern itself with intraday volatility forecasting, but focuses on daily volatility forecasting,
daily summaries are developed that allow for intraday microstructure. These summaries are then calibrated and used for the daily forecasting problem. Kalman-filtering methods are then applied to these, and the results are then shown to compare favorably with the methods of the earlier section.
