A geographical analysis of poverty in Illinois

This study examined the influence of demographic and economic variables on the proportion of poverty for each census tract in Illinois. We developed a model that examined both "person poverty", a set of demographic variables, and "place poverty", a set of economic variables. We used a local spatial statistical and a global statistical method to understand the differences between these methods and to explore the spatial patterns exhibited in the geographically-weighted regression (GWR) method. We sought to understand how the effects of these variables vary spatially across Illinois and whether these variations improve our understanding of poverty. The study developed a best-fit geographically-weighted regression (BFG) model that lacks both local and global multicollinearity. The BFG model does confirm that both demographic and economic variables are necessary to achieve a model with strong explanatory power. The study revealed that there are spatial variations in the influence of economic and demographic variables on poverty across the state. While some of these variations can be explained by rural, urban, and suburban geographies, the variations indicate that these patterns exhibit more complexity. The study concluded that the GWR method greatly improves the model fit and reduces the spatial autocorrelation of the residuals when compared to the ordinary least squared method.