A novel UMIDAS-SVQR model with mixed frequency investor sentiment for predicting stock market volatility

Abstract

Recently, exploring the impact of investor sentiment on stock market volatility becomes popular yet challenging. Two issues on mixed frequency data and nonlinear relationship modelling have to be addressed simultaneously. To this end, we combine unrestricted mixed data sampling (UMIDAS) and support vector quantile regression (SVQR) to propose a novel UMIDAS-SVQR model under the framework of quantile regression. The UMIDAS-SVQR model can be estimated by solving a quadratic programming problem. Thus, we implement the nonlinear quantile regression on mixed frequency data by introducing a kernel function. We then apply the proposed UMIDAS-SVQR model to predict weekly volatility of SHSE and HS300 in mainland China, using mixed frequency investor sentiment as predictors. The empirical results show that the UMIDAS-SVQR model is promising and superior to several competing models in terms of accuracy and robustness. Additionally, we find that the models with investor sentiment are usually superior to those without considering this across different markets and quantile intervals.