Zusammenfassung:
The growing emphasis on sustainability within the business community has led to a progressive integration of Environmental, Social, and Governance values (ESG) into the investment process. This study aims to identify those key fund characteristics that best predict financial performance, with a special focus on the individual impact of ESG Pillar scores. Using the Extreme Gradient Boosting (XGBoost) algorithm, a machine learning technique known for enhancing predictive accuracy, we analyze cross-sectional data on Euro-denominated equity mutual funds with a global scope over a five-year period (2020-2024). In addition, this paper evaluates the advantages of the XGBoost algorithm in predicting mutual fund returns by comparing its performance against two benchmark models: OLS regression and a deep learning architecture. Our findings reveal that ESG Pillar Social score is the second most important predictive factor and that it is positively associated with fund performance, whereas ESG Pillar Environmental score ranks fifth in predictive power and it shows a negative relationship with performance. These insights offer practical value for values-driven investors, financial advisors, and fund managers by supporting investment decisions that align financial performance with sustainability considerations. This study addresses a critical gap in the literature by analyzing the individual effects of each ESG pillar, rather than relying on aggregated ESG scores as commonly done in prior research. The use of cross-sectional data provides a detailed representation of these relationships over a five-year span. Our approach expands the literature by using advanced machine learning to show links between sustainability and fund returns.
