Trading on Twitter: Using Social Media Sentiment to Predict Stock Returns

Abstract

Decision making is often based on the rational assessment of information, but recent research shows that emotional sentiment also plays an important role, especially for investment decision making. Emotional sentiment about a firm's stock that spreads rapidly through social media is more likely to be incorporated quickly into stock prices (e.g., on the same trading day it was expressed), while sentiment that spreads slowly takes longer to be incorporated into stock prices and thus is more likely to predict stock prices on future days. We analyzed the cumulative sentiment (positive and negative) in 2.5 million Twitter postings about individual S&P 500 firms and compared this to the stock returns of those firms. Our results show that the sentiment in tweets about a specific firm from users with less than 171 followers (the median in our sample) had a significant impact on the stock's returns on the next trading day, the next 10 days, and the next 20 days. Interestingly, sentiment in tweets from users with fewer than 171 followers that were not retweeted had the greatest impact on future stock returns. A trading strategy based on these findings produced meaningful economic gains on the order of an 11–15% annual return.

Internet big data and capital markets: a literature review

“…The basic data used for empirical study are quantity-type indicators, including the number of microblogs published, shares, and comments made (Cheng and Lin 2013;Mao et al 2012;Ruiz et al 2012;Sul et al 2016;Xu and Chen 2016). The main content-type indicators include the frequency with which specific keywords appear or sentiment indexes (Bartov et al 2015;Bollen et al 2011aBollen et al , 2011bCheng and Lin 2013;Sprenger et al 2014a;Xu and Chen 2016).…”

“…Sprenger et al (2014b) collected Twitter data related to listed companies, used the Naive Bayes algorithm for text classification of market sentiments, and constructed a daily overall indicator to reflect those sentiments. Sul et al (2016) collated information in tweets related to listed companies, examined the text contents, and determined market sentiments by analyzing the lexical categories (parts of speech) using the Harvard-IV dictionary. The authors similarly established a daily overall indicator of market sentiments.…”

“…The second method is based on the lexical classifications in dictionaries. The relevant keywords are extracted and used for the classification of collated text (Bartov et al 2015;Da et al 2015;Sul et al 2016). In the field of financial research, the main distinction between these two methods lies in the sources of keywords: the source of the first method is text information from the samples, whereas the sources for the second method are actual dictionaries or inventories of lexical categories developed on the basis of dictionaries.…”

“…To address this issue, Loughran and McDonald (2011) developed an inventory of lexical categories that is more appropriate for financial research. However, Sul et al (2016) highlighted that the inventory by Loughran and McDonald (2011) is more suitable for the analysis of formal financial documents (e.g., 10 K filings). Since microblog class text employs colloquial and popular terminologies, the Harvard-IV dictionary is still deemed more appropriate for their analysis.…”

“…In their study on Twitter, Sprenger et al (2014a) used hashtags to collect data from text information. Sul et al (2016) used the $ symbol and a stock's abbreviation as the search criterion.…”