IS MARKET SENTIMENT PREDICTIVE OF STOCK RETURNS?

Abstract

This paper formally studies market sentiment and its performance as a predictor of stock returns in the Singapore context. Market sentiment is determined by applying sentiment analysis to news articles. Sentiment can be decomposed into its constituent variables concerning mood, modality, and subjectivity, among others. I find that market sentiment is a useful predictor of returns, although the extent depends on the sentiment variables incorporated into the model and the predictive model’s implementation, which is subject to the model-selection criteria. Moreover, I draw three key findings from patterns across my various selected models. First, negative sentiment exerts a greater influence on stock returns than positive sentiment. Second, although positive sentiment typically raises returns and negative sentiment typically lowers re- turns, there exists a subjectivity reversal effect where the directions reverse if the sentiment is subjective in nature. Finally, the predictability of stock returns using market sentiment appears non-constant over time. JEL Codes: C58, C80, D84, G14, G17.