Predicting Equity Index Volatility with Text Information

Abstract

We attempt to forecast the volatility of returns of several equity indices using text data. We consider a variety models and methods for incorporating text information with standard volatility forecasting models. We nd that certain representations of the text data can provide some increased accuracy in volatility forecasting. Finally we backtest a trading strategy based on the forecasts of the models for one of the datasets of interest and nd that text information provides better performance for some time periods.