핵심 정량지표 기반 ESG 평가등급 예측 및 ESG 평가기관 간 비교 연구

This study presents an initial attempt to enhance the objectivity and efficiency of ESG evaluation by developing a predictive model based on a limited set of quantitative indicators. Using ESG rating data for listed companies from the Korea ESG Standards Institute (KCGS) for 2022 to 2024, the analysis employed four quantitative indicators—greenhouse gas emissions, energy consumption, renewable energy ratio, and industrial accident rate—to examine the predictive feasibility of a minimum-indicator-based framework. Composite scores constructed from these indicators were weighted using both equal and asymmetric schemes and then used as input variables for machine learning models, including Linear Regression, Random Forest, and XGBoost. The experimental results demonstrated that the XGBoost model achieved the most robust predictive performance, indicating its potential suitability for ESG-related applications. In addition, a comparative analysis of ESG ratings from KCGS, Sustainalytics, and Refinitiv revealed both common patterns and divergent characteristics, particularly in the assessment of higher- and lower-grade firms. These findings suggest that meaningful explanatory power can be achieved without relying on a highly complex indicator system, with important implications for simplifying ESG evaluation frameworks. Furthermore, the study highlights that the results may serve as a practical foundation for developing automated ESG consulting programs that could support more efficient, cost-effective sustainability assessments in the future.