Measuring Business Model Disclosure Quality in Integrated Reports Using NLP Techniques

Corporate disclosure has significantly evolved in recent years, providing stakeholders with a wealth of information in diverse formats. This information is unstructured data, which cannot be easily retrieved and analysed. Consequently, there is a need to transform information in corporate reports into structured data to facilitate in-depth analysis by users of corporate reports. As a starting point to analysing unstructured data in corporate disclosure, we selected the business model aspect of integrated reports. This paper investigates the quality of business model disclosure in 370 integrated reports of JSE-listed companies and compares the evolution of business model disclosure quality over a period of five years. In doing so, we designed a multi-dimensional disclosure index to measure business model disclosure quality based on dimensions such as quantity, dispersion, coverage, and depth. The content analysis was automated by incorporating the multi-dimensional disclosure index into a newly designed text analysis software tool, called Business Model Analysis Tool (BMAT). BMAT uses algorithms based on natural language processing (NLP) techniques. The results from BMAT depict an increase in dispersion, coverage, and depth of disclosure of all components of business models, indicating an enhancement in the quality of business model disclosure. An overall adherence to the International Integrated Reporting Framework was noted.