Abstract: With the increasing complexity of global supply chain networks, quantifying the supply chain risks faced by enterprises and effectively managing them has important practical significance. This paper uses the Term Frequency-Inverse Document Frequency algorithm in machine learning to mine the textual information in the annual reports of listed companies. It constructs supply chain risk indicators at the enterprise level and examines the impact of corporate ESG performance on supply chain risks from the perspective of fulfilling social responsibility. The findings show that good ESG performance of enterprises can promote the establishment of long-term cooperative relationships with upstream and downstream enterprises, and can enhance the discourse power of enterprises in the supply chain. Therefore, good ESG performance of enterprises can reduce their supply chain risks. The heterogeneity analysis reveals that if companies face high competitive pressure, or if they are heavily polluting enterprises, or if they are located downstream of the industrial chain, their ESG performance has a more significant effect on reducing supply chain risks. Therefore, enterprises should actively practice ESG concepts and reduce supply chain risks.

Key words: ESG performance, supply chain risks, spillover effect