Signal or Noise - The predictive value of social media features on market returns

Abstract: This study aims to investigate the relationship between social media data and the main market returns; stock price return, trading volume and volatility for the Swedish stock market. Using a data set of stock-specific social media mentions, classified by state of the art machine-learning algorithms, for the period of 2014-2016, we construct proxies for investor sentiment, attention and agreement. We then perform a time series analysis on daily financial and social media data, testing against an adjusted version of the OMXS30 index and a sample of ten individual firms. Our results show a significant positive relationship between lagged changes in investor sentiment and stock price returns, implying that data from social media contains valuable information for investors. However we find a negative relationship between the volumes of social media mentions, a proxy for attention, and trading volume. Our study contributes to the existing literature on the relationship between social media data and financial markets, as well as extends the relevance of previous research to the Swedish stock market.