The growing relevance of sustainable finance has led investors, financial institutions and regulators to pay increasing attention to the information content of ESG variables. The article Returns under the lens: the importance of ESG factors, by Gabriele Ginestroni, Daniele Marazzina and Nico Rosamilia, addresses this issue from a quantitative perspective: can raw ESG metrics help predict the direction of future stock returns?

Unlike much of the existing literature, the study does not focus only on aggregate ESG scores, whose construction often depends on proprietary methodologies adopted by data providers. Instead, it analyses the underlying ESG metrics directly, with the aim of assessing whether these data, once properly cleaned and processed, contain useful signals for financial forecasting.

The empirical analysis considers MSCI ACWI constituents over the period 2016–2022, focusing on the manufacturing, financial and information sectors in the United States and Europe. The forecasting problem is framed as a classification task: predicting the direction of one-year-ahead stock returns using ESG metrics, financial indicators and past returns.

A key contribution of the paper is the development of an ESG-oriented data cleaning pipeline, designed to deal with the high dimensionality, missing values and strong heterogeneity that typically characterize ESG datasets across sectors and regions. Several machine learning models are then compared, including tree-based methods and gradient boosting techniques. The results show that XGBoost achieves the best predictive performance.

The study also finds that raw ESG variables provide a meaningful and complementary contribution with respect to traditional financial indicators. In particular, a SHAP-based feature importance analysis shows that Environmental and Governance factors are generally the most relevant for prediction, while Social metrics become more important in specific sectoral and geographical contexts.

Overall, the results suggest that ESG information should not be considered only from the perspective of sustainability reporting or regulatory compliance. When properly processed, raw ESG metrics may also represent useful quantitative signals for strategic asset allocation, portfolio construction and risk management. The proposed approach is especially suited to annual investment horizons and low-turnover strategies, rather than short-term tactical trading.

The article is available open access in Decisions in Economics and Finance:
https://doi.org/10.1007/s10203-026-00585-6