Mixed-methods research in the age of analytics, an exemplar leveraging sentiments from news articles to predict firm performance

Investors and companies have always aspired to make informed investment decisions by using diverse information sources. With the explosion of information sources on the web and emergence of predictive analytics, many investors moved beyond traditional financial measures, as key predictors of firm performance, to textual content from analysts’ reports. Empirical research suggests that these information sources complement each other by providing a clear picture of firm performance, but remains silent on the role of additional textual content that continues to emerge and reach more potential investors on the web. We build on this line of research to examine the effect of textual content from business journals in conjunction with summary measures on cumulative abnormal returns. We use sentiment analysis with machine learning and econometrics methods to examine content extracted from textual articles about S&P 500 index companies that are published in the Wall Street Journal (years 2013–2016). Textual analysis of business journals in conjunction with quantitative measures revealed direct and interaction effects on abnormal returns over time. We also tested for robustness by replicating the analysis with different variable operationalization and observe consistent patterns. Relative to positive sentiments, negative sentiments have more profound effects on cumulative abnormal returns. The effect of positive sentiments becomes weaker when past quantitative measures are high. As information sources continue to emerge on the web, this work makes key contributions to the practice of sentiment analysis in financial markets.