Session Chair: Andrea Lu, The University of Melbourne
Tony Cookson, University of Colorado Boulder
Chen Xue, University of Cincinnati
Should Retail Investors Listen to Social Media Analysts? Evidence from Text-Implied Beliefs
Chukwuma Dim, Frankfurt School of Finance and Management
This paper uses machine learning to infer nonprofessional social media investment analysts' (SMAs) beliefs from opinions expressed about individual stocks. On average, SMA beliefs predict future abnormal returns and earnings surprise. However, there exists substantial heterogeneity in SMAs' ability to form beliefs that yield investment value. Some 13% high-skilled SMAs form beliefs that yield a sizeable one-week three-factor alpha of 61 bps, while the rest of the SMAs generate only 6 bps. Firm and industry specializations are the most distinctive characteristics of high-skill SMAs. When forming beliefs, SMAs extrapolate from past returns and herd on the consensus. Herding is, however, less pronounced in bad times and when the consensus is optimistic, but it is more pronounced when the consensus is correct ex-post. SMAs' behavioural bias does not result in systematically wrong beliefs.
Retail Derivatives and Sentiment: A Sentiment Measure Constructed from Issuances of Retail Structured Equity Products
Brian Henderson, George Washington University
Neil Pearson, University of Illinois Urbana-Champaign and Canadian Derivatives Institute
Li Wang, Case Western Reserve University
We use retail Structured Equity Product (SEP) issuances to construct a new sentiment measure for individual stocks. The SEP sentiment measure predicts negative abnormal returns on the SEPs’ reference stocks based on a variety of benchmarks including behavioral factor models and factors based on idiosyncratic volatility, short interest, and the 52-week high effect. Consistent with our interpretation that SEP issuances reflect investor sentiment, aggregate SEP issuances are highly correlated with the Baker-Wurgler sentiment index. Tobit regressions reveal that proxies for attention and sentiment predict demand for SEPs, providing additional evidence consistent with the hypothesis that SEP issuances reflect sentiment.