The Good, The Bad, and The Trending: Using Social Media Data To Test Theories of Momentum
40 Pages Posted: 15 Mar 2019 Last revised: 6 Jun 2019
Date Written: June 3, 2019
I use a unique data set from investor centric social media site StockTwits to test the Hong and Stein (1999) two trader theory of momentum against Daniel, Hirshleifer, and Subrahmanyam (1998) self attribution theory. Tests of individual securities and portfolios produce results which are largely consistent with Hong and Stein (1999). For individual securities I find that trailing measures of sentiment have predictive power over future stock returns. Portfolios generated using StockTwits data are found to have strong explanatory power over the daily momentum factor from Carhart (1997).
Keywords: Microblogging, Sentiment Analysis, StockTwits, Stock Returns, Momentum
JEL Classification: G11, G12
Suggested Citation: Suggested Citation