The Persistence in Income Disparities: Evidence from Italian NUTS-2 and NUTS-3 Regions using Dynamic Spatial Panel Models

This study investigates stochastic income convergence across Italian regions at two spatial scales—NUTS-2 (regions) and NUTS-3 (provinces)—in a context of an affluent North and the less-developed South. Using dynamic panel data models, both non-spatial and spatial, for 2001–2021, the analysis explores the roles of persistence, unobserved regional heterogeneity, and spatial dependence in income gap dynamics. Models without fixed effects show near-unit-root behavior, indicating high persistence or divergence, whereas including fixed effects substantially reduces persistence, providing evidence of conditional stochastic convergence at both spatial scales. Spatial dependence is limited at the NUTS-2 level, with coefficients generally small and insignificant, but at the NUTS-3 level, spatial interactions are economically meaningful and significant, highlighting localized spillovers in provincial income disparities. Ignoring these interactions at the finer scale can overstate persistence. To correct for Nickell bias in dynamic fixed-effects models, System GMM estimators are employed. Spatial System GMM results show strong but stable persistence, with coefficients around 0.86–0.87. The implied half-life of income shocks is roughly five years at the regional level and four to four-and-a-half years at the provincial level, indicating faster adjustments locally. Overall, the study finds weak but robust conditional stochastic convergence, emphasizing the importance of accounting for both spatial dependence and unobserved heterogeneity in understanding regional income dynamics in Italy.Unlike previous studies, our results quantify the persistence and half-life of income shocks, highlighting the difference between NUTS-2 and NUTS-3