The information content of sentiment indices for forecasting Value at Risk and Expected Shortfall in equity markets

The aim of this paper is to investigate the impact of public sentiment on tail risk forecasting. In this framework, we extend the Realized Exponential GARCH model to directly incorporate information from realized volatility measures and exogenous variables. Several indices related to social media and journal articles regarding the economy and stock market volatility are considered as potential drivers of volatility dynamics. An application to the prediction of daily Value at Risk and Expected Shortfall for the Standard & Poor's 500 index provides evidence that combining the information content of realized volatility and sentiment measures can lead to significant accuracy gains in forecasting tail risk.