In earnings calls, corporate executives provide pertinent and timely information beyond the headline numbers. Indeed, textual data from narrative disclosures, including in earnings calls, have been widely used in the finance and accounting literature to predict company performance and stock returns.1

The information provided in narrative disclosures is generally considered an effective measure of “company sentiment,” but why this measure is useful is not well understood. One widely cited paper explains that textual data, such as the data found in earnings calls, are useful because they provide “hard-to-quantify aspects of firm fundamentals” unavailable in other types of company disclosure.2

In their paper “Corporate Disclosure: Facts or Opinions?Vitaly Meursault of the Philadelphia Fed and Shimon Kogan of IDC Herzliya in Israel and the University of Pennsylvania’s Wharton School examine the varied content of earnings calls and how corporate messaging affects firms’ financial performance. As the authors point out, executives customarily convey a mix of objective information (facts) and subjective information (opinions) in earnings calls. Facts essentially reflect “the state of the firm observable by the manager,” they explain, while opinions reflect “the state of the mind of the manager.”

Based on computational linguistic literature, the authors develop a new methodology to classify statements as objective or subjective and apply this approach to disentangle the content of earnings calls. From this, they create an annotated data set containing information on over 4,000 firms and over 85,000 quarterly observations for the period 2008 to 2019.3 Next, they train a machine-learning model to capture the text attributes of subjective and objective sentences in their data set, allowing them to automate, on a large scale, the process of deciphering the content of earnings calls.

Specifically, they use their model to differentiate sentences as predominately objective, for instance, “for the period ending June 30, the occupancy rate was 80.3 percent”; subjective, and therefore representing an opinion that two equally informed executives could disagree on, such as, “we had a great quarter”; or irrelevant, for example, “we now turn to slide 10.”4

Meursault and Kogan find that both objective and subjective information is useful to investors and can affect firm fundamentals and stock returns. They report that executives use both facts and opinions in earnings calls across all firm types, industries, and years analyzed. Facts and opinions, they show, account for about equal weight in earnings calls overall. Exploring further, they discover that while the distribution of subjective and objective text is about the same over time, it varies substantially depending on the complexity of the firm. Large and growth companies use more subjective communication than small and value companies, and the greater use of subjective language among large and growth companies occurs across industries.

CEOs use more subjective language in earnings calls than chief financial officers (CFOs), and both CEOs and CFOs use more subjective language in the Q&A section at the end of the call than during opening remarks. When observing particular executives engaged in earnings calls over time, the authors observe that the communication style of CEOs, with respect to the use of subjective language, depends more on the individual CEO than on the company as a whole. However, the opposite is true for CFOs, as CFOs typically follow customary company practices to guide their use of subjective language in earnings calls. The authors suggest that this finding is consistent with the idea that CEOs have more latitude than CFOs in how they communicate with investors.

A higher prevalence of subjective information in earnings calls is associated with higher postcall investor disagreement — as measured by either abnormal stock-trading activity or social-media-based measures of disagreement.5 Further, the authors show that a higher concentration of opinions in earnings calls is associated with higher returns attributable to anomalies identified in previous research.6

The authors also determine that facts, rather than opinions, as expressed by announcing firms have a stronger statistical association with firms’ performance in the reporting quarter. (They measure firm performance based on corporate earnings.) In the subsequent quarter, however, both facts and opinions are equally important in explaining firms’ future performance.

In summary, Meursault and Kogan suggest that investors interpret and process facts and opinions contained in earnings calls differently, with important implications for the firm. In particular, the way in which company executives present information has a significant impact on the amount of investor disagreement, the extent of anomalous returns, and firm fundamentals.

  1. The views expressed here are solely those of the authors and do not necessarily reflect the views of the Federal Reserve Bank of Philadelphia or the Federal Reserve System.
  2. See, for example, Werner Antweiler and Murray Z. Frank, “Is All That Talk Just Noise? The Information Content of Internet Stock Message Boards,” Journal of Finance, 59:3 (2004), pp. 1259–1294; and Zheng Tracy Ke, Bryan T. Kelly, and Dacheng Xiu, “Predicting Returns with Text Data,” working paper (2020).
  3. Paul C. Tetlock, Maytal Saar-Tsechansky, and Sofus Macskassy, “More Than Words: Quantifying Language to Measure Firms’ Fundamentals, Journal of Finance, 63:3 (2008), pp. 1437–1467.
  4. The authors’ data on earnings calls are from the Capital IQ Transcripts database, which is part of the Wharton Research Data Services (WRDS) platform.
  5. Specifically, the authors manually tag 3,673 earnings call sentences and then train their machine-learning model to identify relevant text attributes. If a sentence contains both facts and opinions, they classify the sentence according to what is determined to be the main point of the sentence.
  6. The authors use the StockTwits platform to obtain three social-media-based measures of investor disagreement as found in J. Anthony Cookson and Marina Niessner, “Why Don’t We Agree? Evidence From a Social Network of Investors,” Journal of Finance, 75:1 (2020), pp. 173–228.
  7. As Meursault and Kogan explain, previous research shows that anomalous returns are greater following the release of earnings and that, after the disclosure of earnings information, investors’ beliefs are pushed “closer to the truth,” sending stock prices of undervalued firms higher and overvalued firms lower. See Joseph Engelberg, R. David McLean, and Jeffrey Pontiff, “Anomalies and News,” Journal of Finance, 73:5 (2018), pp. 1971–2001.