Firm-level perception of climate policy uncertainty and energy efficiency: Empirical tests based on text analysis and machine learning

This paper examines how firms' perception of climate policy uncertainty (CPU) affects energy efficiency. Using textual analysis and Word2Vec machine learning techniques, we construct a firm-level CPU index based on the annual reports of Chinese listed firms from 2011 to 2020. The empirical results show that higher CPU significantly reduces firms' energy efficiency. Instrumental variable estimations based on the turnover of local environmental officials, together with a series of robustness tests, confirm the validity of the findings. Mechanism analyses further indicate that CPU weakens firms' energy efficiency by suppressing green innovation and reducing environmental governance quality. We also find that external media attention, investor scrutiny, and corporate risk management can mitigate these negative effects. This study contributes to the literature by introducing a novel machine learning–based measure of firm-level climate policy uncertainty and providing new evidence on how policy uncertainty influences firms' green transition. The findings further highlight the importance of stable and transparent climate policies for improving firms' long-term energy efficiency and supporting low-carbon development.