R&D Disclosure and Carbon Performance: A Machine Learning Analysis of Carbon‐Intensive Firms

ABSTRACT The growing urgency of climate change, alongside global sustainable development initiatives, has brought environmental priorities to the forefront of corporate strategy. This study explores how narrative disclosures related to research and development (R&D) predict carbon performance in European industries with high R&D intensity. Guided by the natural resource‐based view (NRBV), the research examines how qualitative R&D narratives act as strategic tools for communicating innovation‐driven environmental strategies. We introduce a novel methodological approach for analyzing unstructured textual data using advanced machine learning (ML) models, including neural networks (NNs), support vector machines (SVMs), and random forests (RFs). Our results show that firms with extensive and positively framed R&D disclosures are more effective in managing carbon emissions and in progressing toward major sustainability targets such as the Paris Agreement and the EU Green Deal. The findings also reveal that regulation and innovation shape distinct patterns in narrative disclosures across sectors, particularly in technology and pharmaceuticals. Moreover, the tone and thematic focus of these narratives offer strategic insights that go beyond traditional financial indicators, effectively linking innovation with sustainability objectives. This research advances the corporate disclosure literature by deepening our understanding of how sustainability and innovation intersect, while also offering practical guidance for firms and policymakers.