Harnessing Aspect - Based Sentiment Analysis: How A re Tweets Associated with Forecast Accuracy?

For firms in the consumer sector of the economy, tweets about service quality reflect consumer satisfaction, which determines firms' future earnings. Our study responds to anecdotal evidence indicating that analysts have adopted opinion mining to scrutinize Twitter data in order to detect shifts in consumer behavior and make earnings forecasts. If this anecdotal evidence is accurate, certain tweet characteristics may be associated with the accuracy of earnings forecasts for these firms. Our study draws on the literature on consumer satisfaction and firm earnings to identify possible tweet characteristics and hypothesize their associative relationships with analyst forecast accuracy. We use the airline industry as the study context and extract tweets related to airline service quality from publicly available Twitter data and analyst forecast data from the Institutional Brokers' Estimate System Academic. We apply content analysis, followed by aspect-based sentiment analysis, to the downloaded tweets. Using regressions, we find that the breadth of coverage and the number of posters are associated positively with forecast accuracy. The valence of tweets differentiates their effects on forecast accuracy: negative tweets enhance forecast accuracy to a greater extent than do positive tweets. We do not detect any association between tweet subjectivity and analyst forecast accuracy. There is a marginal negative association between tweet dispersion and forecast accuracy. We conclude by discussing theoretical and practical contributions.