新质生产力视角的碳效率影响因素研究——基于LightGBM和SHAP方法

Research on Influencing Factors of Carbon Efficiency from the Perspective of New Quality Productive Forces: Based on LightGBM and SHAP Methods

摘要:

在全球气候治理紧迫性加剧与中国双碳目标深入推进的背景下，厘清新质生产力对碳效率的关联特征对城市绿色转型至关重要。本文构建涵盖数据要素、新劳动者、新劳动资料、技术研发、新生产组织与新劳动对象六个维度的新质生产力测度体系，采用非径向方向性距离函数测算地级市碳效率，并分解为效率变动与技术变动。进一步结合LightGBM模型与SHAP方法，系统识别各要素与碳效率的关联特征。结果表明：数据要素、新劳动者与新劳动资料是碳效率预测中贡献度最高的关键特征要素，而技术研发、新生产组织与新劳动对象则表现出潜在影响。在非线性关系上，各要素普遍呈现阈值效应，数据要素与新劳动资料呈现U型影响结构，新劳动者与新劳动对象的作用效果表现出显著的成熟度依赖与规模门槛特征，新生产组织则呈现阶段性正向影响。机制分解显示，效率变动主要由数据要素与新生产组织优化资源配置影响，而技术变动则更多受新劳动资料与新劳动对象的前沿技术创新支撑。本研究旨在识别新质生产力的绿色效应模式，为城市低碳精准施策提供参考。

Abstract:

Against the backdrop of intensifying global urgency in climate governance and the deepening advancement of China's "Dual Carbon" goals, clarifying the influencing factors of new quality productive forces on carbon efficiency is crucial for urban green transformation. This paper constructs a measurement system for new quality productive forces encompassing six dimensions: data factors, new-type workers, new-type means of labor, technology R&D, new production organization, and new-type objects of labor. We employ the Non-radial Directional Distance Function (NDDF) to measure carbon efficiency at the prefecture-level city level and decompose it into efficiency change and technological change. Furthermore, by integrating the LightGBM model with SHAP (SHapley Additive exPlanations) methods, we systematically identify the association characteristics between each element and carbon efficiency. The results indicate that data factors, new-type workers, and new-type means of labor serve as core driving forces for improving carbon efficiency, whereas technology R&D, new production organization, and new-type objects of labor exhibit potential influences. Regarding nonlinear relationships, most elements demonstrate threshold effects. Specifically, data factors and new-type means of labor display a U-shaped influence structure. The effects of new-type workers and new-type objects of labor show significant maturity dependence and scale threshold characteristics, while new production organization presents a phased positive impact. Mechanism decomposition reveals that efficiency change is primarily driven by the optimization of resource allocation by data factors and new production organization, whereas technological change is more supported by frontier technological innovations in new-type means of labor and new-type objects of labor. This study aims to identify the green effect patterns of new quality productive forces, providing a reference for precise urban low-carbon policy-making.