Financial development, energy consumption, and environmental degradation: A threshold and causal machine learning approach with explainable AI

This study examines how financial development shapes environmental degradation under uneven globalization, where cross-border capital flows, trade openness, energy-market interdependence, and heterogeneous institutional capacities jointly influence low-carbon transition. Although prior research has examined the finance–environment nexus, limited evidence explains how financial development affects emissions through energy-consumption channels across nonlinear thresholds and heterogeneous country contexts. Using a panel dataset of 82 emerging and advanced economies from 1990 to 2021, this study investigates the relationships among financial development, energy consumption, and CO₂ emissions by integrating Panel Threshold Regression, mediation analysis, causal machine learning, and explainable artificial intelligence. The results show that the environmental effect of financial development is regime dependent. At lower levels of financial development, financial deepening is associated with lower CO₂ emissions, whereas beyond estimated thresholds, further financial expansion is linked to higher carbon intensity. Energy consumption is identified as a key transmission channel through which financial development affects environmental degradation. Causal machine learning further reveals substantial heterogeneity, with stronger nonlinear effects in emerging economies and weaker associations in advanced economies. SHAP-based explainable AI confirms the relative importance of energy use and different dimensions of financial development in predicting emissions. These findings suggest that financial development alone does not automatically promote environmental sustainability under globalization; its effects depend on financial maturity, energy structure, institutional context, and countries' capacity to redirect globally integrated finance toward low-carbon transition.