From Words to Value: The Relationship between CEO Letters and Financial Performance

Sammanfattning

Seminar Date: 2024-05-29 Course: BUSN79 - Degree Project in Accounting and Finance Authors: William Gustafsson and Jacob Linder Examiner: Anna Glenngård Five key words: CEO letters, sentiment analysis, agency theory, firm performance, Tobin's Q. Purpose: Discern with the help of a new machine learning model whether the sentiment expressed in CEO letters can offer meaningful insights into the firm's performance. Methodology: The new algorithm FinBERT was used to collect sentiment scores, which were used as the main independent variables in this study. OLS Regressions with fixed effects were used. The dependent variables were ROAt+1, Tobin's Qt+1, ΔROAt+1, and ΔTobin's Qt+1. Theoretical perspectives: The study draws from agency theory, information asymmetry, signaling theory, and prospect theory. Empirical foundation: The study focuses on ten years of data from the FTSE 100, 2013 to 2022. The final sample consists of around 700 observations in the different regressions. Financial firms and companies whose letters were exceedingly long were excluded. Conclusions: Negative sentiment in the CEO letter appears to negatively affect next year's ROA and Tobin's Q. For ΔROAt+1 and ΔTobin's Qt+1, the Negative score has a positive relationship.