Social Media Analysts' Skills: Insights from Text-implied Beliefs

Abstract

This paper analyzes the extent to which nonprofessional social media investment analysts (SMAs) form correct beliefs about stock returns. Contrary to the wisdom-of-the-crowd hypothesis, over half of the SMAs form beliefs that correctly align with future returns. However, there is substantial heterogeneity in the SMAs' ability: some 13% high-skilled SMAs produce a one-week three-factor alpha of 61 bps, while the remaining 87% generate only 6 bps. Firm and industry specializations are the distinctive traits of the high-skilled SMAs. Although SMAs extrapolate and herd, their expectations are not systematically wrong. Consistent with theory, the extrapolation and herding intensities depend on SMAs' skills.

Keywords: Nonprofessional analysts, Expectation formation, Social networks, Social finance, Machine learning, Textual analysis

JEL Classification: G11, G12, G14