Forecasting corporate financial performance using sentiment in annual reports for stakeholders' decision-making

This paper is aimed at examining the role of annual reports' sentiment in forecasting financial performance. The sentiment (tone, opinion) is assessed using several categorization schemes in order to explore various aspects of language used in the annual reports of U.S. companies. Further, we employ machine learning methods and neural networks to predict financial performance expressed in terms of the Z-score bankruptcy model. Eleven categories of sentiment (ranging from negative and positive to active and common) are used as the inputs of the prediction models. Support vector machines provide the highest forecasting accuracy. This evidence suggests that there exist non-linear relationships between the sentiment and financial performance. The results indicate that the sentiment information is an important forecasting determinant of financial performance and, thus, can be used to support decision-making process of corporate stakeholders.