Social impacts reflected in CSR reports: Method of extraction and link to firms innovation capacity

Abstract

Assessing and comprehending the social impact of firms at global and local level is a pressing concern for both researchers and policy-makers. To address this concern, our paper contributes to the stream of literature that studies the content of Corporate Social Responsibility (CSR) reports (which are also referred to as non-financial statements, sustainability reports or parts of annual reports) using text mining methods. We present a novel approach called Standard-based Impact Classification method (SBIC method), which employs natural language processing (NLP) and supervised machine learning techniques to identify the types of social impacts reflected in CSR reports. We deploy a Random Forest model which we train on reports adhering to Global Reporting Initiative (GRI) framework, enabling the identification of social impact in the majority of CSR reports that do not conform to this standard. Our proposed SBIC method serves as a valuable tool for comparing the social impacts generated by firms, industries, or countries. We showcase an application of our approach by examining the relationship between a company’s social impact and its innovation capacity. Our findings support the existing literature consensus that CSR activities generally exhibit a positive correlation with a firm’s ability to innovate. Furthermore, we reveal that specific types of social impacts have a more pronounced influence on innovation capacity.