Volatility and Sentiment, The Explanatory Power of Social Media Sentiment on Volatility for the U.S. Equity Market

Abstract: In this paper, we analyse to what extent sentiment explains and predicts volatility. We perform simple linear regressions with sentiment as the independent variable for a sample of different dependent variables. These dependent variables are the VIX, the VVIX, ve business-to-business companies, ve business-to-consumer companies and the S&P 500. In order to approximate the volatility of the individual companies and the S&P 500, conditional volatility is estimated through GARCH(1,1) models. Sentiment is derived from Twitter data based on the search term "S&P 500" as well as the individual company names. We nd that sentiment and the implied volatility carry a inverse contemporaneous relationship at the 99% signi cance level while no predictive power is found. The conditional and historical volatility of other instruments, and portfolios, show no significant explanatory nor predictive relationship with sentiment. The relationship between sentiment and volatility is not found but sentiment is proven to be related to either the perception of future volatility or the demand for structured products which have volatility as their underlying instrument.