An application of spatially autoregressive models to the study of US county mortality rates

County mortality rates in the US tend to be associated with social and economic resources of counties and the unequal distribution of these resources across space. The processes that generate these social and economic inequalities are often tied to geographical location. In this paper, we present an application of spatially autoregressive models of US county mortality rates that control for the social and economic conditions that often influence mortality rates and the effects of spatial structure of counties in the US. We suggest that arguments are missing from the social science and demographic literatures to offer possible explanations for the spatial patterning of county mortality rates and ecological correlates of these rates at the county level. We find that, after controlling for spatial structure in the data, several key social variables become insignificant in the analysis. We suggest that spatial statistical models are valuable tools in the social and behavioural sciences but that the use of these methods needs to be well grounded in considerations about the spatial process inherent to the outcome studied, and the applications of these methods should not be used solely for post hoc statistical correction.