Using the variable specifications and transformations provided by FL we replicated, exactly, their WLS regression results using S-Plus (S + SpatialStats) software and the open source software, R. (The original FL analysis had been carried out using SAS.) Focusing specifically on the final and most fully elaborated FL model, we examined the regression residuals for spatial autocorrelation. A Moran’s I statistic of 0.326 ( p < 0.001) strongly suggests that standard regression estimates cannot be trusted, but, by itself, does not determine how we should proceed. We commenced the reanalysis by specifying and estimating a simple regression model using SpaceStat in order to obtain the useful diagnostic statistics that are part of the SpaceStat package. The model is faithful to the FL model except that it is fit using OLS rather than WLS—due to limitations of SpaceStat. The OLS diagnostics demonstrate the likely presence of a spatial error process in the child poverty data, although we will present the results of both a spatial error model and a spatial lag model.