Realized GARCH, Volatility Risk Premium and CBOE VIX

Hao and Zhang (2013) demonstrated that GARCH family option pricing models under the LRNVR fail to explain the variance risk premium suggested by CBOE VIX index. In this paper we show that a variant of Realized EGARCH model proposed in Hansen et al.(2012) and futhrer developed in Hansen and Huang(2013), under an exponentially affine pricing kernel, generates reasonable level of variance risk premium. We find empirical evidence that Realized EGARCH model provides significantly better forecasting performance for CBOE VIX index, both in-sample and out-of-sample, than GARCH and EGARCH model under LRNVR, Heston-Nandi GARCH model under both the LRNVR and a variance dependent pricing kernel of Christoffersen et al.(2013).