CMStatistics 2018: Start Registration
View Submission - CFE
Title: Realized GARCH model adding robust measures of skewness and kurtosis Authors:  Cesar German Santamaria - Universidad Nacional de San Martin (Argentina) [presenting]
Abstract: Some researchers have started to incorporate higher moments into their volatility models, e.g., the GARCHSK model, that considers the conventional measures of the sample skewness and kurtosis based on daily data. With the availability of high-frequency data, the estimation of volatility has moved from traditional daily models to realized models. This shift aimed to provide more accurate short-term risk models. One of them is the RGARCH model. Following this trend, we propose the Realized Generalized Autoregressive Conditional Heteroskedasticity Robust Skewness and Robust Kurtosis (RGARCHRSRK) model, which incorporate not only the realized measure of volatility, but also robust measures of skewness and kurtosis, where the standardized residuals follow a modified Gram-Charlier expansion. We found empirically that the proposed model is statistically significant and empower the estimation of parametric Value at Risk (VaR) in comparison with the other RGARCH, GARCHSK and GARCH base models.