Quantitative Comparison of China’s Multi-Level Carbon Peaking Policies Based on Natural Language Processing

Pragmatic sustainability emphasizes that policies must adapt to the reality of multi-level governance to balance targets and feasibility. To explore how this concept is embodied in China’s carbon peaking policies, this study adopted natural language processing (NLP) and machine learning methods to conduct a systematic quantitative analysis of 316 carbon peaking policy documents spanning from the national to county levels in China. The findings reveal that the policy system presented a distinct logic of pragmatic coordination. The application of legal instruments decreased with descending administrative levels, whereas that of supervision instruments showed the opposite trend; central-level targets were more flexible, while local governments demonstrated higher policy intensity in specific targets and livelihood-related sectors. The regional differences in policy intensity were closely associated with local economic development and energy structure, indicating that future policy optimization should more thoroughly implement the principle of common but differentiated responsibilities in target decomposition and dynamic adjustment. This study not only provides a novel quantitative perspective for investigating pragmatic sustainability in carbon peaking policy texts but also offers critical empirical evidence for synergistically advancing SDG 13 (climate action) with other SDGs.