A growing body of research has explored the influence of the built environment—human- constructed aspects of the physical environment such as transportation infrastructure, land use and city design, and recreational facilities—on physical activity, nutrition, and rates of obesity. Rural areas face special challenges in the realms of food access and health. Urban-rural health disparities in obesity prevalence and nutrition have been widely acknowledged in the recent literature. However, the lack of exploration of intrarural variation is problematic because knowing that an area is “rural” tells us very little of the context of the community: the needs of a family in a small, well-established farming town may have little in common with that of a family in a rather transient warm-winter haven along a freeway. This project begins to fill the existing gap in studies of the rural built environment and food access through the exploration of the impact of intrarural variation and the built environment on participation in the Supplemental Nutrition Assistance Program (SNAP).
This project had two primary aims. The first was to explore the relationship between social and geographical characteristics of rural communities and SNAP participation in those areas, in order to identify factors that lead some communities to underutilize SNAP. The second aim was to develop a typology that could differentiate among rural areas based on these community characteristics, identifying multiple discrete categories of rural communities that can be used to better understand the varying needs of different rural populations across a State. By identifying sociogeographical indicators of SNAP utilization, and the communities where they cluster, we hope to be able to better develop and target outreach programs to rural communities with greatest need.
This study examines data for the rural population of Arizona. Rural populations were defined as those not living in a Census-designated urban area. Based on this definition, this study examined a population of approximately 1.3 million people that accounts for nearly 20 percent of the population of Arizona. Data were drawn from the 2010 U.S. Decennial Census, the 2008-2012 American Community Survey (ACS) Five-Year Estimates, the Arizona Department of Health Services 2012 Primary Care Area Statistical Profiles, and SNAP participation data from the Arizona Department of Economic Security. A spatial analysis technique, dasymetric mapping, was used to process data into a uniform geography at the census-tract level. The first phase of analysis explored the influence of the theoretically specified community factors on retailer access and SNAP participation by structuring a pattern of regressions using sequential canonical analysis, referred to as a cascade model in cognitive psychology. The study then used cluster analysis of the factors in the model to develop community clusters that served as the basis for the rural community typology.
The study used cascade analysis with four outcome variables: human capital, mean drive time to a SNAP authorized retailer, estimated percent of children enrolled in SNAP, and child enrollment in SNAP relative to low-income status. The analyses show that the variables identified as likely to have an influence on SNAP enrollment did have explanatory power. The cascade models show that there are complicated direct and indirect relationships among many of the variables that predict whether low income children in any particular rural area are likely to be enrolled in SNAP. Human capital, drive time to the nearest SNAP authorized retailer, and estimated child SNAP enrollment were significantly correlated with the proportion of low-income children enrolled in SNAP, as were community characteristics such as median age, work engagement, economic sector, income equality, ethnicity, linguistic isolation, migration, and slow life history. Overall, the results of the cascade model help to illustrate how important it is to consider multivariate models that account for many of the variables affecting relative SNAP enrollment simultaneously, rather than relying on bivariate relationships.
To increase interpretability of these data in order to make these results more easily actionable, the study used k-means clustering to create a typology of rural communities in Arizona based on the input variables and the human capital factor. An eight-cluster set was chosen that was both interpretable and that had face validity with those working in these rural communities in Arizona. The clusters suggest areas where similar issues may be affecting communities, even though they are not necessarily near one another geographically. The clusters were named based on their relative levels of the variables included, and on other characteristics of the area, yielding the following typology: Ag/Mining/Forestry, Border City, Border Periphery, Mixed Migrant, Suburb/Historically Mormon, Retirees, Scenic, and Tribal. The clusters with the highest mean human capital (Retiree, Scenic and Mormon/Suburb), had the lowest proportion of child SNAP enrollment. Four clusters had a greater than 10 percent disparity between enrollment and low income: Tribal, Mixed Migrant, Border Periphery, and the Ag/Mining/Forestry tracts. The clusters with the two highest drive times to SNAP retailers (Tribal and Ag/Mining/Forestry) are in this set, as is the cluster with the lowest human capital (Tribal). Better outreach or other interventions in these particular areas might increase the proportions of low-income children enrolled in SNAP.
The approaches laid out provide a concrete way to better understand and describe the variability in rural areas, and can aid in identifying and better advocating for the differing needs across rural populations.