CMStatistics 2021: Start Registration
View Submission - CFE
Title: Predictive power of the variance premium Authors:  Yuze Liu - Fernuniversität in Hagen (Germany) [presenting]
Abstract: The predictive power of the variance premium (VP) on the equity premium is tested in different aspects for the S\&P 500 index. The VP is defined as the difference between implied volatility and the physical expectation of conditional variance. We consider various versions of the VP: VIX, Structural VIX and Corridor VIX are employed as implied volatility measurement. 5-min realized volatility (RV) forecasts in an extended HAR framework including jumps and daily quadratic variation and the recently MF2-GARCH model featuring leverage effect and the lower frequency of conditional variance are considered as the physical measures. First, the forecasting performance of different conditional variance measures is compared to select the best physical measurement. Second, an extensive comparison of the predictive power is performed by using different VP measures. The performance is evaluated by in-sample and out-of-sample predictive power and predictive regression tests. Third, we discuss the difference in the predictive power between VP, implied volatility and conditional volatility. The results are discussed in light of the usefulness of newly proposed VP measures in comparison to implied or conditional volatility, respectively.