Predicting stock market volatility based on textual sentiment: A nonlinear analysis

This paper investigates whether and how investor sentiment affects stock market volatility forecasting from a nonlinear theory perspective. With the use of a novel dataset that contains massive articles about stock market analysis obtained from a Chinese investors' community, we construct four sentiment indices to measure investor sentiment by applying textual analysis techniques. Differing from the developed market, we find that the investor sentiment from an emerging market causes stock volatility by a nonlinear pattern rather than a linear style. Furthermore, we show that the investor sentiment improves stock volatility prediction based on the long short‐term memory model. And the predictability is still significant after considering another sentiment proxy variable. Finally, we demonstrate that this improvement of predictive performance is meaningful from an economic point of view.