Understanding ESG Scores Through Network Analysis: A Study Using Graph Neural Networks

Graph neural networks (GNNs) are powerful tools for analyzing graph-structured data and have numerous applications. In this study, GNNs are used to explore the connections between a set of Italian companies, focusing on how sharing decision-makers influence their Environmental, Social, and Governance (ESG) scores. The research compares various GNN models, including GraphSAGE, Graph Attention Networks, and Graph Neural Additive Networks, with traditional classification techniques such as kNN, Random Forest, Decision Trees, and Naive Bayes. The results demonstrate that GNNs provide a higher node classification accuracy and suggest that more central nodes tend to have higher ESG scores, indicating that companies with greater network connectivity may provide higher ESG performance. This suggests that sharing decision-makers could be a strategic tool to enhance efforts in sustainability and social responsibility.