Twitter and stock returns

Abstract

In this thesis, we investigate whether the sentiment of tweets mentioning stock tickers can be used to predict stock performance. In particular we test for leading and lagged relationships between the percentage of positive and/or negative tweets and the returns of the S&P 500 index. We obtain a longitudinal data set of all tweets mentioning stock tickers over a four-month period amounting to 2,599,277 tweets distributed over 84 trading days. We use daily measures for positive and negative sentiment to generate our explanatory variables. Our results indicate that an increase in the percentage of positive tweets predicts increased stock performance the following day whereas an increase in the percentage of emotional tweets predicts a reduction in stock returns after two and three days. An increase in the percentage of negative tweets may predict a reduction in stock returns.