Abstract
Background and Objective: Grounded in Diener’s cognitive life satisfaction framework, this study develops an integrated MCDM–machine learning pipeline to examine how environmental, social, and economic sustainability predict subjective well-being.
Methods: Panel data (47 countries, 2016–2025, N=470) were used. Sustainability indices followed the triple-bottom-line framework. Winsorization and log transformations were applied. Of five MCDM techniques evaluated, ELECTRE was selected as the primary method based on its superior discriminatory ranking performance and robustness in sustainability benchmarking contexts. An ExtraTrees regressor (R²=0.80) predicted the Cantril Ladder; SHAP decomposed importance. LOCO-CV tested generalizability (MAE=0.38).
Results: Social sustainability dominated (62.1% importance) over economic (21.4%) and environmental (16.5%) dimensions. ELECTRE ranked institutional quality highest in structural importance, with social and environmental indices closely following. SHAP showed non-linear interactions (Δ = −0.02). Lower-middle-income countries converged (slope +0.0065, p=0.02), high-income countries declined marginally—a narrowing gap (n=8 exploratory). Composite index weighting was robust (r=0.98).
Discussion and Limitations: Using only cognitive SWB limits full tripartite model operationalization; affective measures were unavailable. The 10-year panel constrains causal inference. LOCO-CV MAE (0.38, SD=0.19) indicates country heterogeneity rather than misspecification. The non-linear SHAP interactions could not be fully decomposed, warranting future interaction analysis. The n=8 sub-group convergence finding is exploratory. Future phases will incorporate low-income countries and country clustering.
Conclusion and Policy Implications: Social well-being (life expectancy, education, social support, digital inclusion) is the strongest life satisfaction lever, necessitating social cohesion. The ELECTRE-SHAP pipeline offers a replicable tracking template. Future extensions include Fuzzy Cognitive Mapping for causal simulation and affective SWB indicators to operationalize Diener’s full model.
Keywords
Sustainable development; subjective well-being; MCDM; SHAP; social sustainability