Bootstrapping prediction intervals on stochastic volatility models

Abstract

The parametric bootstrap method is applied to derive the prediction intervals for stochastic volatility models. The study adopts the parameters estimation developed by So et al. (1997) and proves the validity of the proposed bootstrap procedure for this process. The basic stochastic volatility model specifies the mean equation with standard normal error. It is found, via simulation study, that the same algorithm can be employed to the model with heavy-tailed innovations, which demonstrates the potential of the bootstrap techniques. This methodology is also applied to a real data example to predict the daily observations on the S&P 500 index and the results confirm that our interval predictions are satisfactory.

Acknowledgement

This work was supported by the National Science Council in Taiwan under grant number NSC92-2118-M008-009 and the MOE Program for Promoting Academic Excellent of Universities in Taiwan under grant number 91-H-FA07-1-4.