Machine Learning-based Corporate Social Responsibility Prediction

Corporate Social Responsibility (CSR) has long been discussed since the year 1950s [1]. This long-debated topic of adopting CSR often draws divided conclusions due to the fact that the reward for investing in CSR practice is often delayed and unquantifiable. Most companies practice CSR to some extent, but that invites other form of issue like pressure of dressing up CSR [2]. While many studies have pointed to the qualitative benefits [3], [4] of increased CSR engagement in different areas like improved company reputation, brand awareness, lower costs, resource usage reduction and more employee engagement. We, on the other hand, hope to provide a quantitative perspective to this important area. We constructed a few machine learning models to examine the relationship between ESG (Environmental, Social and Governance) scores and the firm’s ROE (Return of Equity). We developed models like Support Vector Machine (SVM), Random Forest, Naive Bayesian, Multilayer Perceptron (MLP) Neural Networks, and Long Short-Term Memory (LSTM) Neural Networks. Neural Networks based models generally achieved good accuracy with the highest accuracy of 81.08% of predictability based on MLP model. We hope this work can establish grounds for quantifying the financial benefit CSR can bring to a firm, and it can well be anticipated using machine learning model with sufficient indicators being measured.