The impact of ESG scores on financial performance with a comparative analysis of traditional, machine learning and fuzzy logic approaches in BIST 100 companies

This doctoral dissertation investigates the relationship between Environmental, Social, and Governance (ESG) scores and the financial performance of publicly listed companies in Turkey, focusing on firms included in the BIST 100 index between 2010 and 2024. The study aims to address the methodological challenges and reliability issues surrounding ESG scoring by comparing traditional provider-based, machine learning-based, and fuzzy logic-based ESG evaluation methods. Drawing on an integrated research design, the thesis first applies panel data regression analysis to examine the effects of ESG scores and their sub-dimensions on key financial indicators, including Return on Assets (ROA), Return on Equity (ROE), and Tobin's Q. Subsequently, various supervised and unsupervised machine learning models—such as Random Forest, XGBoost, and Artificial Neural Networks—are employed to forecast financial performance and benchmark the predictive power of alternative ESG scoring frameworks. In addition, the study utilizes multi-criteria decision-making (MCDM) techniques, including entropy weighting and the TOPSIS method, to rank companies in terms of both financial and sustainability performance. The findings reveal that machine learning-based ESG scores exhibit higher predictive accuracy for financial performance compared to traditional and provider-based models. Moreover, the results indicate that the integration of ESG sub-dimensions, especially governance and environmental factors, enhances the explanatory power of financial forecasting models. Sectoral analysis demonstrates significant heterogeneity in the ESG–financial performance nexus, with industry-specific dynamics influencing the strength and direction of the relationship. The thesis contributes to the literature by offering a comprehensive methodological comparison of ESG scoring approaches, highlighting the added value of data-driven and explainable machine learning models in sustainable finance. The integrated analytical framework developed in this study provides practical guidance for investors, policymakers, and corporate managers seeking robust and transparent tools for sustainable value creation and decision-making.