The geographical polarization of per capita output: an analysis of the Italian case based on provincial data

Economic convergence at the sub-national level could have spatially explicit manifestations reflecting convergence clubs and other forms of geographical clustering that are not captured by an overall inequality measure. By decomposing the Theil index of per capita income inequality into between and within groups of neighbouring locations, the degree of spatial polarization is measured as the share of inequality accounted for by the between group component. The choice of the partition can fundamentally change any inequality measure decomposition, both qualitatively and quantitatively. In order to overcome this issue I propose a kernel approach to the spatial polarization measure based on the Theil index computed on the spatial moving averages. This allows to detect a polarization curve measuring the degree of geographic concentration as a function of a spatial scale parameter, denoting the geographic dimension of the groups. The analysis of per capita income data of Italian provinces in XX century shows the existence of a multiple polarization, that is per capita income is polarized at different spatial scales. The forces sustaining the polarization are stronger at medium and high spatial scales. The analysis also shows the presence of a long-run geographic club convergence with converging clubs and a strengthening of the polarization, mainly referable to regional differences; the reinforcement of North � South dualism is more limited.