Measuring national sustainability: ESG scores from corporate data

Environmental, Social, and Governance (ESG) metrics have become central to sustainability assessment, yet the link between national conditions and composite ESG performance remains largely unexplored. We develop a bottom-up national ESG rating by aggregating the distribution of listed firms’ ESG scores for twelve developed economies between 2013 and 2022. Several aggregation schemes—mean, median, Sen’s inequality-adjusted index, and a dispersion-adjusted mean—are benchmarked, and the resulting rankings prove highly consistent, supporting the median as the headline measure. National ratings are then compared with World Bank indicators of environmental efficiency, social welfare, and governance quality through panel fixed-effects regressions and four machine-learning models (Random Forest, Gradient Boosting, Support Vector Regression, and CatBoost), assessed via cross-validation and explainability tools. CatBoost achieves the highest predictive accuracy and balanced use of predictors. Energy intensity and under-five mortality consistently act as dominant negative drivers, while gender representation and demographic maturity contribute positively. A pillar-level (E, S, G) panel-VAR analysis reveals strong within-pillar persistence and asymmetric cross-effects led by the social dimension. Overall, the framework provides a transparent bridge from firm-level data to national ESG performance, delivering robust and interpretable evidence for policy evaluation and sustainable investment screening.