Investor Sentiment and Stock Returns

Sammanfattning: We examine whether investor sentiment can predict next-day stock returns and if the predictive characteristic differs when discriminating between positive and negative sentiment. We propose two novel sentiment proxies derived from a unique dataset using machine learning algorithms to approximate sentiment. In a sample of 286 Swedish equities from February 2014 to June 2015, we find that (1) investor sentiment can predict next-day stock returns; (2) the negative sentiment has a larger absolute economic effect than positive sentiment; and (3) an indication of that a simple sentiment-based trading strategy can earn substantial risk-adjusted returns.