KOF working papers // KOF Swiss Economic Institute, ETH Zurich 288
We examine the statistical power of fundamental and behavioural factors with regards to stock returns of the Dow Jones Industrials Index. With a novel sentiment dataset from over 3.6 million Reuters news articles, we find signifcant correlations between Reuters sentiment and stock returns. We show with vector autoregression and error correction models that sentiment can explain and predict changes in stock returns better than macroeconomic factors. Considering positive and negative sections of Reuters sentiment, we find that negative sentiment performs better in simple trading strategies to predict stock returns than positive sentiment, while the sentiment effect remains over months.
Reuters sentiment stock returns out-of-sample forecasts vector error correction model