Soft information in public firms’ quarterly earnings calls: Isolating the effects of value-relevant information and linguistic tone in financial text

Thumbnail Image
Journal Title
Journal ISSN
Volume Title
School of Business | Master's thesis
Degree programme
82 + 4
In this thesis, I study the effects that the soft information in public firms’ quarterly earnings call transcripts have on stock returns. More specifically, I measure both the value-relevant information (sentiment) communicated in the earnings calls as well as the linguistic tone of the statements, by studying firms’ cumulative abnormal returns following quarterly earnings calls, both on portfolio-level and with a regression analysis on earnings call -level. My findings can be divided into methodological contributions and the uncovered financial phenomena. Firstly, I provide strong evidence in support of utilizing large language model -based solutions in extracting sentiment information from financial texts, as opposed to the dictionary-based methods that have been predominant in the financial academic literature so far. Furthermore, I show that using a tool like this together with a dictionary-based method that is designed to measure the linguistic tone of financial text makes it possible to isolate the two effects. Leveraging these benefits of my approach, I recognize three distinct stock return phenomena related to either earnings call sentiment or tone. Firstly, I find a strong positive relation between the sentiment of an earnings call Q&A session and the cumulative abnormal returns for that firm. Moreover, unlike some of the previous studies, I find that this stock price impact is nowadays practically immediate, with this information being incorporated into stock prices in the initial reaction period. Secondly, I find evidence that the sentiment information in the calls’ presentation section leads to a relatively similar immediate stock price reaction, but then to a subsequent reversal, that makes the long-term informational value in this section non-significant. Finally, I find that when controlling for the value-relevant information, a higher linguistic tone of the call’s presentation section predicts lower abnormal returns, in line with the theory of strategic communication, which states that managers use of a more positive tone constitutes bad news due to their incentives to downplay negative news. I find this negative stock price reaction to be immediate only for firms that reported a negative earnings surprise, whereas firms that reported positive earnings surprise this effect takes longer to be incorporated into share prices.
Thesis advisor
Knüpfer, Samuli
earnings call, sentiment analysis, linguistic tone, abnormal returns
Other note