This paper focuses on the analysis on income inequality in Italy at the municipal level of the areas defined by the National Strategy for Inner Areas. We discuss an analysis of the economic and spatial dynamics of the phenomenon through the construction of the Gini's coefficient and the estimation of the regression model for the evaluation of the determinants of inequality. We highlight the influence of the spatial dimension on income inequality in Italy. Inequality appears to be greater in densely populated urban centers with a strong incidence of tertiary activities and young population. Conversely, in the inner areas, the distribution of income is more balanced due probably to the weakness of the social and economic structure that determines low levels of income and job opportunities mainly in the agricultural sector. Sustainability 2020, 12, 1622 2 of 18 Sustainability 2020, 12, x FOR PEER REVIEW 11 of 18 355 Figure 2. Map of significance of the LM index for the values of the Gini coefficient. Source: own 356 elaboration.
357Consequently, the analyses are based on a classic linear regression model. Table 4 shows its main 358 results. The histogram of the residuals is showed Figure 3.
359A first element of reflection is that the model explains 60% of the observed variability, 360 confirming the strong territorial characterization of inequality in Italy. A second consideration is that, 361 except L, all the selected variables are quite significant. In detail, the t ratio between coefficients and 362 standard errors reports the relative variation of Gini coefficent, associated with a unitary variation of 363 the explanatory variables. The last columns on the right show the p values and the relative 364 significance under the hypothesis of robust standard errors with respect to heteroskedasticity (i.e., in 365 the presence of a nonconstant variance between the residuals or discrepancies between observed and 366 estimated values of the dependent variable, a possible cause of distorted estimations of the model 367 parameters).
369Lisa significance map High-High (928) High-Low (39) Low-High (27) Low-Low (809)