An exploratory indicator of disclosure consistency in ESG environmental reports: Application to Chinese corporations

The rapid expansion of Environmental, Social, and Governance (ESG) reporting has increased the availability of corporate environmental information while raising concerns about the consistency and interpretability of disclosed content. A key challenge is to assess the alignment between the narrative tone of ESG reports and the reporting density of specific environmental strategies. We develop an exploratory indicator of disclosure alignment that integrates these two dimensions. Using Natural Language Processing (NLP), we quantify (i) sentiment, as a proxy for narrative tone, and (ii) substance, defined as the normalized reporting density of specific environmental strategies identified through keyword-frequency analysis. The resulting Sentiment–Substance Indicator provides a comparative text-based framework for characterizing narrative–substance alignment within ESG reports and is not intended to infer greenwashing or actual environmental performance. The analysis focuses on 2024 ESG and Sustainability Reports from a purposive sample of 20 large, high-visibility Chinese corporations across five sectors: Finance, Energy, Food and Beverage, Technology, and Automotive. The sample provides a benchmark for examining the feasibility and interpretability of the indicator and is not intended to support statistically representative sector-level inference. Within the selected sample, corporations in the Technology group show relatively higher narrative–substance alignment, combining near-neutral sentiment with comparatively high reporting density of environmental strategies. Corporations in the Finance group show the contrasting pattern, with higher sentiment and lower reporting density of environmental strategies, resulting in greater narrative–substance divergence. Energy, Automotive, and Food and Beverage occupy intermediate positions in terms of relative alignment. NBS-related activities are reported across all five groups, with the highest aggregated investment, tree-planting, and carbon-offset values among the selected Energy corporations. The proposed indicator provides a transparent framework for examining disclosure alignment across contrasting corporate contexts. While larger and more representative datasets are required to validate its performance and broader applicability, the study illustrates the potential of NLP-based indicators to complement existing approaches to ESG disclosure analysis.