Does sentiment affect stock returns? A meta-analysis across survey-based measures

Abstract

We are the first to meta-analyze the literature on the relationship between sentiment and stock returns, a topic that reacts to the history of systemic events causing asset bubbles in financial markets. We focus on three questions — whether the literature is biased; what is the “true effect” beyond this bias; and what are the key determinants of the variance among the estimates in the literature. To answer those questions we collect 1311 point estimates from 30 primary studies and use state-of-art meta-analytical approaches. Both linear and non-linear tests for publication bias suggest that the “true effect” of an improvement in sentiment is non-negligible and negative. In the majority of specifications, researchers tend to report this effect as being much stronger than it actually is. Next, using Bayesian model averaging we show that the effect of sentiment on future returns is significantly stronger for individual investors than for large institutions, and in US stock markets compared with European ones. The effect also depends on several data and model characteristics.