A longitudinal analysis of CSR disclosure for BIST companies: A text mining approach

Official URL: https://risc01.sabanciuniv.edu/record=b2756176_(Table of contents)

With environmental problems being more salient in daily life, corporations increasingly started to report activities they undertake that do not solely entail financial interest but are mostly related to their impact on the society, otherwise known as their Corporate Social Responsibility (CSR) activities. In the business and management literature, CSR reporting has particularly become a major research interest and a great source for understanding CSR behavior. Despite the wide interest in analyzing CSR reporting in the last decades, the range of methods for analysis remain narrow, mainly dominated by the widely used content analysis method. In this thesis, we followed a novel text mining approach to examine the annual reports of BIST companies from 2007 to 2020. For this purpose, we firstly prepared an ESG dictionary to extract keywords from the annual reports and assigned aggregate environment, social and governance scores to each report. Descriptive results for all data showed that governance related information has the highest salience among all ESG categories while environment salience has an upward trend. As a secondary task, we employed two different clustering algorithms, k-medoids and hierarchical (agglomerative) clustering, to group all reports based on their ESG salience. Our analysis revealed 3 distinct groups of reports and showed that the share of the group with high environment scores have increased significantly in 2020.