Supply Chain Concentration's Impact on ESG Performance in the Context of Carbon Emission Reduction Policy: A Study Leveraging Machine Learning

This paper explores the impact of supply chain concentration on corporate ESG performance in the context of China's carbon emission reduction policy. This paper employs machine learning models for logistic regression and causal inference. Specifically: 1) A double fixed-effects model with both individual and time fixed effects is used. 2) The model clusters robust standard errors at the company level to better capture the causal relationships between variables. 3) A difference-in-differences (DiD) approach is applied to capture the effect of carbon policies. 4) The model controls for relevant variables at both the company level and policy level. 5) The explanatory variable is lagged by one period to eliminate the reverse causality between the explanatory and explained variables. 6) Mechanism tests are conducted from two aspects: mediating effects and moderating effects. Regarding the research data, this paper selects A-share listed companies from 2009 to 2022 as the initial sample, excluding ST, ST*, financial industry companies, and those with incomplete data, resulting in a total of 37,600 core enterprises with 33,941 samples. The data are sourced from the CSMAR Database, Huazheng ESG Database, and the China Energy Statistical Yearbook. Through empirical analysis, this paper finds that supply chain concentration significantly inhibits corporate ESG performance, with carbon information disclosure playing a partial mediating role in this process. Meanwhile, this study also finds that market pressure can mitigate the negative impact of supply chain concentration on ESG performance. The study addresses potential reverse causality by lagging the explanatory variable by one period. Additionally, robustness tests are conducted by replacing the proxy variables of the core explanatory variables, and the results remain valid.