Methods: I build OLS regression models with fixed effects using pooled data from the 2003 and 2013 Current Population Survey-Food Security Supplement and from the RUCC. The RUCC are a continuum produced by the USDA every ten years using the decennial Census that assigns each US county a code, ranging from 1 (completely urban) to 9 (completely rural). The independent variable, metro-area designations, is constructed using the RUCC and takes on four values: large, medium, small, and non-metro areas. The outcome of interest, food insecurity, is represented by a binary variable, with a value of 1 assigned if a household experienced very low or low food insecurity in the last year and 0 if not. I also include a vector of control variables previously shown to be associated with food insecurity.
Results: The findings of this study suggest a non-linear relationship between food insecurity and metro-area designations. All metro areas experienced higher odds of food insecurity compared to non-metro areas, but households residing in the largest metro areas had lower increased odds of food insecurity (4.5%) than medium and small metro areas (13.9% and 17.3%, respectively). A non-linear relationship persists between metro-area designations and food insecurity persists when disaggregating by year, although the pattern of this relationship changes from 2003 to 2013.
Conclusions/Implications: Differences in the odds of food insecurity across the metro-area designations suggest that using a binary rural-urban definition of location is likely insufficient in understanding the nuanced role of location on food insecurity. The fact that the relationship between metro-area designations is non-linear implies that there are additional factors beyond the population of a metro-area designation that contribute to differences in food insecurity status. Further research is needed to identify what these metro area-specific characteristics might be in order to optimize policies and programs designed to reduce food insecurity going forward.