Economic growth in Italian NUTS 3 provinces

This paper analyzes the conditional \(\beta \) -convergence hypothesis for NUTS 3 Italian provinces. A methodology for the simultaneous treatment of spatial dependence and spatial heterogeneity is developed. Spatial dependence is introduced in the economic model by assuming a spatial Durbin model specification. The absence of data experienced by some economic variables at the NUTS 3 level is addressed through a modified version of the Bayesian interpolation method introduced by Palma and Benedetti (J Geogr Syst 5:199–220, 1998). Spatial heterogeneity is taken into account by identifying convergence clubs. For this purpose, we use the modified simulated annealing algorithm introduced by Postiglione et al. (Comput Econ 42:151–174, 2013). The methodology is compared with the heteroscedastic approach proposed by Kelejian and Prucha (J Econom 157:53–67, 2010).