Spatial Patterns and Geographic Determinants of Welfare and Poverty in Tunisia

Previous poverty analysis in Tunisia concluded that the poor population is concentrated in interior areas, especially in the northwest and center west. Thus more information on the spatial dimension of welfare and poverty may be of interest for any poverty alleviation programs as poverty may be associated to geographic locations. However, the analysis of the spatial dimension cannot be limited to the addition of some variables to our econometric model. We have to consider the neighborhood effects and the heterogeneity of households’ behaviors in more disaggregated geographic units using specific tools of spatial and geographical analysis. First, we conduct an exploratory spatial data analysis (ESDA) based on a geographical information system (GIS), to visualize the “local” spatial structure of poverty. Second — to deal with spatial autocorrelations and unobserved spatial heterogeneity of the households’ behaviors — we use a spatial autoregressive model (SAR) and a geographical weighted regression model (GWR) respectively. Spatial and non-spatial models are compared according to their prediction performances. SAR and GWR spatial models are found superior to the traditional non-spatial regression model, and give a better approximation of the Tunisian poverty map.