Spatial-temporal evolution of traffic carbon emission efficiency across 30 provinces in China: a super-efficiency SBM-DEA optimization approach

Abstract

The rapid development of transportation in China has brought heavy environmental burdens including high energy consumption and carbon emission, which calls for a rigorous assessment of traffic carbon emission efficiency to guide the dual-carbon strategic objectives of the country. Although the academic attention is growing, the existing studies have paid little attention to the province-level traffic carbon emission efficiency by simultaneously incorporating slack variables and undesirable outputs in a unified non-radial framework. This paper fills this gap by using a Super-Efficiency Slacks-Based Measure Data Envelopment Analysis (SBM-DEA) model under the assumption of constant returns to scale to evaluate the traffic carbon emission efficiency of 30 Chinese provinces from 2010 to 2023. The evaluation framework combines four input indicators and three output indicators including capital, labor, energy structure and transport activity with undesirable carbon emission outputs. Longitudinal trend analysis is used to explore temporal evolution patterns, and global and local Moran’s I indices are used to explore spatial autocorrelation. The results show that the average efficiency at the national level is 0.214 for the whole year. The efficiency varies significantly across the regions, with the western region having the highest efficiency of 0.349 followed by the northeastern region with 0.158, and the central region with 0.155. The eastern region has the lowest efficiency of 0.117. Efficiency shows a U-shape trend, falling from the peak of 0.360 in 2010 to the trough of 0.156 in 2021 and picking up to 0.184 in 2023. The spatial analysis reveals a strong positive autocorrelation with a mean-efficiency global Moran’s I of 0.368 (p<0.001). The annual global Moran’s I is positive and statistically significant during 2010–2021, but it drops to statistical insignificance during 2022–2023, indicating a gradual dispersion of the clustering pattern. High-high clusters are mainly located in the western provinces and low-low clusters in the eastern coastal areas. The results provide actionable insights for differentiated regional policies and cross-regional green technology collaboration.