Catching Up or Falling Behind? Governance Thresholds and the Formation of ESG Convergence Clubs

Whether nations converge toward a shared sustainability equilibrium or stratify into persistent clubs carries fundamental implications for global climate governance. We apply a five-layer econometric framework to a rating-agency-independent environmental, social, and governance (ESG) panel of 137 countries over 1990–2023 constructed from nine primary macro-level indicators. The Phillips–Sul log-[Formula: see text] test rejects full-panel convergence across all ESG pillars; club detection identifies two convergence clubs (28 high-ESG and 14 mid-ESG countries) alongside a divergent majority of 95 countries whose ESG gap relative to Club 1 widens from 0.125 (2000) to 0.281 (2023). A complementarity-trap model predicts that governance and ESG investment are strategic complements with a coordination threshold separating low- from high-ESG equilibria. Hansen fixed-effects threshold regression with 500-bootstrap validation confirms a critical Rule of Law threshold at [Formula: see text] below which the income elasticity of ESG performance is 0.050, rising to 0.090 above it; a second threshold at [Formula: see text] reveals a third frontier regime in which the elasticity rises further to 0.162, indicating a monotonically increasing institutional dividend. We address endogeneity using a legal-origin instrumental variable (first-stage [Formula: see text]) and Arellano–Bond Generalized Method of Moments (GMM), both consistent in sign with panel estimates. The mean group panel Autoregressive Distributed Lag (ARDL) establishes long-run error correction with the Rule of Law long-run coefficient (4.276) greatly exceeding the income coefficient (0.127). A governance-conditional Environmental Kuznets Curve (EKC) shows the emissions turning point ranging from $13,717 to $68,597 Purchasing Power Parity (PPP) across governance quartiles—a roughly fivefold governance dividend. XGBoost achieves 77.4% cross-validated accuracy using country-average features, rising to 92.0% using five-year rolling windows, with SHapley Additive exPlanations (SHAP) confirming the Rule of Law as the dominant feature in both. Structural clustering of the divergent majority identifies three subgroups requiring differentiated policy responses, and nine historical threshold-crossing countries show an average ESG gain of 0.094. All findings should be interpreted as associational rather than causal, given the inconclusive endogeneity tests; nevertheless, the consistency of the governance–ESG association across multiple estimators and specifications supports its empirical robustness.