Social sentiment and impact in US equity market: an automated approach

Abstract

AbstractIn this study, a database of approximately 50 million tweets was used for the estimation of the positive and negative sentiment factors for 2557 companies operating in US stock market. For each company, the sentiment factors were calculated through the mean equations on GARCH models of different orders. Our findings show that, for 503 companies the negative factor effect has a greater impact than the positive factor effect. The period analyzed was from October 2022 to January 2023, using hourly observations. Results provide evidence to support that there is an asymmetric effect from the factors traveling to the stock market and it takes at least an hour the signal to travel. The investors and regulatory agents can find useful the results given that news has been demonstrated a source of influence in the market. Therefore, news impact can be modeled into portfolio theory using GARCH which is easy to implement and to interpret. Given the exposure of prices and volatility to news, it can be considered that these findings provide evidence to support efficient market hypothesis. Modeling returns and volatility for the assets through GARCH family is a widely known tool. Including the news sentiment on social media is dually a novelty: the empirical demonstration of the effects of social comments on the stock performance and volatility, in addition to the use of a large data set of social network comments in an hourly frequency.