Does low carbon city pilot promote urban carbon unlocking?-- A heterogeneity analysis based on machine learning

Carbon lock-in has become a major obstacle to green development in all countries, and carbon unlocking (CUL) is of great significance in accelerating the process of carbon neutrality worldwide. Based on the panel data of 292 prefecture-level cities in China from 2006 to 2020, this paper constructs a PSM-DID model based on machine learning to empirically test the impact of LCCP policy on urban CUL. The regression result shows that the low carbon city pilot (LCCP) policy significantly increases the level of CUL by 1.25 % in the pilot cities, which is still valid after range of robustness tests, and the mechanism analysis shows that green innovation and green habits are important ways for the policy to affect CUL. Based on the generalized random forest (GRF) algorithm, we further find that the policy can increase the CUL level by 0.6 %, and the average treatment effect (ATE) of the sample is concentrated between 0.004 and 0.006, which exhibits obvious heterogeneity. This paper chooses population quality, openness level and industrial structure as heterogeneous factors, which are finally found to have a U-shaped or inverted U-shaped relationship with ATE. This study provides useful insights for China and other countries to break the carbon lock-in and realize green development.