Title: Exploiting news analytics for volatility forecasting
Authors: Simon Bodilsen - Aarhus University, CREATES (Denmark) [presenting]
Abstract: Using a large database of macroeconomic and firm-specific news, we study whether news sentiment can be used to enhance prediction of stock market volatility. We construct two types of news indices at the daily frequency, by properly aggregating the sentiment scores of past macroeconomic- and firm-specific news, respectively. Using reduced-form time series models for realized measures of volatility, we find evidence that the index of domestic macroeconomic news is very useful in order to predict future levels of volatility for both individual stocks and the S\&P 500 Index. In particular, we find large gains in the predictions of volatilities at long horizons by including the macroeconomic index in the time series regressions. On the other hand, the predictive power of firm-specific news are found to modest in the general framework.