Forecasting Stock Market Volatility With Investor Sentiment And Macroeconomic Factors

It is an important project for domestic financial sector to prevent and defuse risks.Preventing excessive volatility of the stock market and making appropriate policies are significant.In the long run,the stock market and the real economy are closely linked,and the stock market volatility is affected by the macroeconomic operation.But in the short run,the subjective sentiment factor of investors also causes volatility of the stock market.The integration of big data method and traditional econometrics provides a new idea for the measurement of subjective sentiment;the rise of mixed-frequency data modeling provides a new method for studying data with different frequencies.This paper is designed to integrate multi-source information,introduce investor sentiment and macro-economy from the perspective of mixing volatility component model,and make more accurate measurement and prediction of stock market volatility.Because the frequency of macroeconomic data is lower than that of sentiment and stock market volatility data,traditional research methods require the same frequency of data,resulting in information loss when the frequency is reduced.In the framework of mixed-frequency,this paper directly models the data of different frequencies,decomposes the total volatility of the stock market into long-term and short-term components,introduces different low-frequency macroeconomic indicators to explain the long-term components which change slowly,and compares the prediction results of several models.At the same time,the short-term components of the mixed-frequency model are not discussed in depth in the existing research.This paper introduces the daily investor sentiment to explain the short-term components of the rapid change,which is an extension of the existing model.In the measurement of investor sentiment,text analysis method is used to extract text sentiment from Oriental Fortune Forum,and combined with other indicators,state space model is used to extract common factor to construct composite investor sentiment indicator.This construction method is also relatively novel.Compared with the traditional principal component analysis method,it can reduce the noise components that have nothing to do with sentiment in each variable.Most of the GARCH-MIDAS models constructed in this paper have better prediction result than the same-frequency benchmark model.The macroeconomic and investor sentiment introduced have significant positive impact on the stock market volatility in the long term and short term respectively,and have prediction function on the stock market volatility.Macroeconomic prosperity index has the best predictive result.Compared with single indexes,the composite investor sentiment index constructed has better prediction effect.