| Summary: | 碩士 === 元智大學 === 財務金融學系 === 96 === Contrary to that traditional noisy volatility estimates constructed from GARCH
or stochastic volatility model using daily data, the recent proposed realized volatility calculated via high frequency intraday data has been proven to provide better precision assessing latent volatility. The superior information content along with its model-free and nonparametric nature, has made the realized volatility to make the latent volatility process virtually visible. Taking advantage of the feature, this paper uses vector autoregression model to investigate the simultaneous relationships among return, trading volume and realized volatility. We argue that the past studies addressing the relationships between return/volume (or return/volatility, or volatility/volume) while controlling the volatility (or volume, or return) as an exogenous variable may subject to endogeneity bias (simultaneous bias) and thus produce misleading results. Hence, we re-examine several well-documented hypotheses or relationships among these variables through bivariate and a full of trivariate analysis to verify our concerns. We found that trading volume is positive related to volatility contemporarily, justifying the role trading volume plays in conveying market information. In addition, we found the leverage effect in volatility, as well as in trading volume. The relationship between lagged volatility and current return that relates return-risk trade-off no longer exists. To properly examine the volatility feedback hypothesis explaining the return-volatility asymmetry, controlling the leverage effect, we construct a new forward-looking volatility factor that is unrelated to historical information set by extracting information from the implied volatility index (VIX) traded in CBOE. Our results support the coexistence of both the volatility feedback effect and the leverage effect. In sensitivity analysis, we found our obtained results robust to the jump considerations.
|
|---|
