Sentiment Analysis of Banks' Annual Reports and Bank Features: LASSO Approach

摘要

The current study investigates the effect of bank features on disclosure tones that have been classified into “positive” “negative” and “net” through 630 annual reports for a sample of 63 listed conventional banks from eight emerging markets namely, Egypt”, “Jordan”, “Qatar”, “Oman”, “Saudi Arabia”, “Kuwait”, “Bahrain” and “United Arab Emirates” covering the period of ten years (2008-2018). To achieve this objective, LASSO regression has been employed as a unique technique to design three models (positive- negative- net). Surprisingly, the results of these models are similar. Our results revealed that bank performance and beta of the bank are statistically significant and have a positive association with tones (positive, negative and net) in the three models. In contrast, the variables, bank size, leverage and market to book ratio are not statistically significant in all models. Such results have suggested that bank managers with good financial performance and stock volatility tend to send signals to many stakeholders in order to distinguish themselves from others. Other banks with a good financial performance may behave differently. They may use negative tone in their reports as a strategy that can help them to avoid blame in the future if they suffer losses.