FinPred: An explainable machine learning tool to accurately predict bidirectional CSR–CFP relationship

• This study predicts the relationship between corporate social performance (CSP) and corporate financial performance (CFP). • We introduce a new explainable machine learning model to accurately predict the CSPCFP relationship. • We leverage 70 different features and comprehensively analyze their impact. • Our model offers high predictive accuracy and valuable insights into CSP-CFP dynamics. • FinPred as a standalone predictor is publicly available at: https://github.com/MLBC-lab/financial_predictor. The relationship between corporate social performance (CSP) and corporate financial performance (CFP) remains complex, with methodological challenges complicating causal analysis. We introduce FinPred, an explainable, integrative machine-learning pipeline that unifies 70 relevant features at the firm, governance, and industry level to examine both how CSP predicts next-period CFP and how CFP predicts next-period CSP, using separate explainable models. Using Random Forest regression, FinPred achieves normalized mean square error (NMSE) < 0.2 for most targets, delivering high predictive accuracy and transparent insight into CSP–CFP dynamics. The results help reconcile inconsistencies in the literature by showing that: (i) CSR and CSI are strongly path-dependent; (ii) CSP’s financial impact is dimension-specific, with community, diversity, and environmental practices exerting the clearest influence on market valuation; and (iii) market value added (MVA) creates managerial incentives that shape CSP decisions. Practically, FinPred’s explainability flags sharp declines in MVA as early indicators of irresponsibility risk, enabling boards, investors, and regulators to intervene proactively. We also release code and a reusable pipeline to support replication and extension.