Construction and analysis of corporate greenwashing index: a deep learning approach

Abstract

Abstract Environmental information disclosure serves as a critical indicator of corporate social responsibility. While the quantity of disclosures in this area continues to rise, strategic disclosure behaviors, particularly greenwashing, are increasingly evident. This covert form of greenwashing significantly obstructs the overall advancement of ecological civilization; thus, identifying and preventing corporate greenwashing presents a key challenge in contemporary research. This study utilizes the corporate social responsibility reports of A-share companies from 2008 to 2023 as the research sample. A corporate greenwashing index is constructed using the deep learning-based MacBERT model, which demonstrates bidirectional processing capabilities. This model effectively considers contextual factors to mitigate ambiguities and extracts fine-grained information at the sentence level. The constructed corporate greenwashing index can capture the quality of corporate environmental information disclosure through text structural features, enriching the methodologies for researching corporate information disclosure and environmental behaviors and possessing significant practical application value. This research not only provides a solid foundation for subsequent exploration but also enhances the quality of corporate environmental information disclosure, assists enterprises in establishing sustainable business models, and comprehensively promotes the construction of ecological civilization.

Keywords

• Deep Learning
• Greenwashing Index
• Corporate Social Responsibility
• Environmental Information Disclosure