Last modified: 2015-09-17
Abstract
This work-in-progress research is based on the financial information that may be present in Twitter data streams. While this area of research contains many possibilities, the focus of this paper is the potential information in tweets leading up to firms’ quarterly earnings announcements. Quarterly earnings announcements can have a major impact on stock prices especially if the earnings are a “surprise.â€Â Earnings surprises will be the main dependent variable of this study, and are defined as the percentage difference between a firm’s reported earnings and analyst expectations prior to the announcement.
Twitter’s “$†tag will be used to screen for tweets referencing stock tickers. Additionally Twitter provides a geo-tag that will allow for screening of tweets within a given radius of a firm’s headquarters. This increases the probability of the tweet containing a real information leak. The tweets will then be grouped into three moods based on sentiment analysis. This analysis is conducted by assigning values to key words within the tweets which are then totaled resulting in a designation for each tweet of positive, negative, or neutral. The final step of the analysis will be to test if there is any predictive power in tweets leading up to an earnings announcement.