Stock market analysts could be missing out on a valuable source of information on company fundamentals: Twitter.
Consumer tweets about company brands and products can carry “predictive power” regarding future sales and earnings results, according to a study by Vicki Wei Tang, associate professor at Georgetown University’s McDonough School of Business.
“Even though Twitter provides nonfinancial information that is informative about upcoming sales and such information is easily accessible… analysts do not fully incorporate in their forecasts the implications for upcoming sales of the collective wisdom of users of Twitter,” Tang wrote.
For the study, Tang enlisted the services of social media firm Likefolio to analyze third-party tweets made about 171 consumer-facing companies from 2013 through the end of 2015. The sample included only tweets about company products and brands, excluding tweets directly about the firms, in order to capture customer satisfaction and not investor sentiment.
Tang categorized tweets by tone—positive, negative, or neutral—as well as by whether they indicated user intent to purchase a product.
She found that not only were these tweets predictive of upcoming sales—with predictive power highest when Twitter was the primary source of public information about a product—they went largely ignored by analysts.
Analyst forecast error was negatively correlated with the number of positive tweets and the number of tweets containing purchase intent, meaning that the larger the number of tweets, the lower the analyst’s forecast relative to actual sales.
“Analysts consistently overweight the financial information in prior years but underweight the nonfinancial information on Twitter in forming their forecasts,” Tang wrote.
By incorporating nonfinancial information published on Twitter into stock market analysis, Tang argued that investors can form more accurate opinions about individual stocks, and therefore make better investments.
“The valence and volume of tweets, once summarized, provide useful information in predicting firm fundamentals,” she concluded.
Related: Investing Is One Big Data Problem