Regional Inequalities in Productivity: Multiscale Evidence on Structural Trajectories in Mexico

This study analyzes the persistence of regional inequalities in productivity in Mexico during the period 2010–2023 through a multiscalar structural approach. By integrating nonlinear functional models, interpretable machine learning, and spatial econometric validation, this study identifies the economic growth trajectories of federal entities and the structural factors underlying them. Using a double-well potential model, states are classified into convergence, stagnation, or divergence regimes, capturing their long-term dynamic behavior. Subsequently, the CatBoost algorithm with SHapley Additive exPlanations decomposition is applied to estimate the marginal importance of variables such as digital connectivity, government efficiency, and labor productivity. Finally, a spatial fixed-effects model and territorial cluster analysis reinforce the consistency of the findings, highlighting the institutional and spatial dimensions of inequalities. Results confirm that regional growth in Mexico responds to persistent structural configurations, implying the need to design differentiated and territorially sensitive policies.