This study of the Supplemental Nutrition Assistance Program (SNAP) enrollment in Georgia counties during the Great Recession (December 2007- June 2009) and its aftermath through 2013 builds upon previous research and fills in several important gaps in current knowledge. Few published studies have examined the effects of SNAP expansion via the American Recovery and Reinvestment Act (ARRA) to this point, making this project a critical step toward increased understanding of how economic and policy changes affected enrollment rates. Although previous research has identified demographic differences in SNAP enrollment rates at the national level, the analyses in this study are focused at the local level using unique county-level data from Georgia. The study data stratify county-level SNAP enrollment rates by race, facilitating an analysis that identifies geographic patterns specific to these subgroups.
In addition to examining changes in SNAP enrollment, the study investigates how these changes in enrollment and benefit levels impacted the geographic distribution of SNAP retailers across the State and county-level food insecurity. By analyzing local-scale factors that drive changes in SNAP enrollment and outcomes that may be affected by SNAP participation, the research provides a textured understanding of how the policies surrounding SNAP recruitment impact community well-being at the local level.
The study uses several methodological approaches in these analyses. First, a pooled cross-sectional data set is used, including data on SNAP enrollment by race and control variables derived from Census and other data sources, to perform fixed- and mixed-effects models examining county-level changes in SNAP enrollment by race and the effects of SNAP enrollment on food insecurity in Georgia. Second, the data on SNAP retailers from USDA and InfoUSA are used to perform first-difference models predicting change in density of SNAP retailers by public use microdata areas (PUMAs) within Georgia. The study also maps these changes using GIS techniques, including local indicators of spatial autocorrelation (LISA) to examine spatial clustering.
The analyses of changes in SNAP enrollment reveal substantial differences in effects for increased benefit levels on single-person than larger households, net of theoretically relevant controls and year fixed effects. Across all racial and ethnic groups, the coefficients for single- person households are two to three times larger than for three-person households. To illustrate the impact, the maximum benefit allotment for three-person households increased by a total of $118 between 2007 and 2013. In practical terms, the study models predict that this increase in benefits over the full time period raises total SNAP enrollment by 7.67 percent, White enrollment by 7.91 percent, Black enrollment by 9.79 percent, and Hispanic/Latino enrollment by 9.20 percent. Likewise, the maximum benefit allotment for 1-person households increased by a total of $45 between 2007 and 2013. The models predict that this increase produces a 7.65 percent increase in total SNAP enrollment, a 7.97-percent increase in White enrollment, a 9.77 percent increase in Black SNAP enrollment, and a 9.14-percent increase in Hispanic/Latino enrollment.
Further, the study finds that rising benefit levels contributed positively to uptake in SNAP participation at higher rates in metropolitan counties for both Black and Hispanic/Latino enrollment but not for total or White enrollment. For Black and Hispanic/Latino SNAP enrollment, results show that the impact of increased benefit levels was two to three times higher in metropolitan than in non-metropolitan counties. Together, these results support the hypothesis that increasing benefit levels serve to offset transaction and stigma costs of program participation, particularly for single-person households as well as for minority populations in urban areas. No known prior research has investigated the intersection of these factors in SNAP enrollment trends. The findings suggest that increased benefit levels promote greater enrollment in SNAP among populations that may be more difficult to enroll even when eligible because of stigma or other transaction costs, most notably Hispanics/Latinos who tend to be under-enrolled.
With regard to SNAP’s effects on store density, the results of multivariate ordinary least squares regression using first differences show that outside of the five central Atlanta counties, changing SNAP enrollment was a significant predictor of changes in store counts, with the exception of midsized stores. This effect was strongest for convenience stores, where a 1,000-person increase in SNAP enrollment predicted a one-store increase over the time period. In the Atlanta core, enrollment had an effect of similar size, the only significant effect of enrollment in these areas. The first and third quartiles of enrollment change were 1.3 and 11.3 respectively (with change measured in thousands), so for the middle 50 percent of PUMAs, increased enrollment accounted for between 1 and 11 new convenience stores.
For large stores, SNAP enrollment showed an effect outside the Atlanta core, with an increase of roughly 14,000 SNAP clients predicting the opening of a new supermarket. Four of the 46 non-core PUMAs had this level of enrollment change. All but one are in the eastern and southern Atlanta suburbs, with the exception was in the northwest corner of the State. Conversely, four PUMAs in southern Georgia had a decrease of 10,000 or more SNAP clients, which was predictive of supermarkets exiting SNAP. Increased SNAP enrollment predicted the growth of specialty retailers in non-core areas, and the magnitude of this effect was similar to that of large supermarkets. There was no significant effect of enrollment on midsized grocers in either core or non-core PUMAs.
The results of this study show how increased SNAP participation resulted in the generation of new stores or the enrollment of existing ones. This is especially notable given the significant economic contraction happening during this period. For most SNAP clients, there were many more places at which to use SNAP benefits after the recession than there were beforehand. The specifics of this effect may not be entirely positive, particularly through increasing the availability of small stores and their often meager selection of healthy foods. It may be that requiring these small stores to offer a greater selection of healthy foods in order to be certified as a SNAP vendor would mitigate these negative effects. Conversely, continued support for programs allowing SNAP to be used for urban agriculture or farmers’ markets could provide a financial boost to these alternative food sites. SNAP’s role as a stimulus is confirmed by these data, but future work should consider how policy could help direct how and where SNAP can improve the health and sustainability of community food systems.
With regard to food insecurity, as the maximum SNAP benefit increases, a significant reduction was observed in both the total food insecurity rate and the child food insecurity rate. The magnitude of these coefficients is small, but when the policy changes that occurred during this time frame are taken into consideration, the magnitude grows in significance. One component of the American Reinvestment and Recovery Act (ARRA) of 2009 was to increase the benefit amount that families receive when they participate in SNAP; average benefit levels increased as a result by approximately 15 percent. Using this information and the observed variation in the maximum benefit variable, the study reveals an approximate $63 increase in maximum benefits for three-person households over the study period. When this is multiplied by the SNAP benefit beta coefficient in the model (approximately .017), the study finds that this aggregate increase in benefits is associated with a decrease in the food insecurity rate of 1.07 percent. This actually represents a 5.4-percent decrease in the mean food insecurity rate during the observation period. Once this additional analysis is conducted, which takes into consideration the actual dollar change in benefits during this time period, the clinical significance of these findings becomes more meaningful.
The magnitude of the SNAP benefit-level coefficient for child food insecurity was much larger than the total food insecurity model (.043 percent versus .017 percent, respectively). When the same analysis is calculated for child food insecurity, the study finds that the approximate $63 benefit increase for three-person household is associated with a decrease in the food security rate of 2.7 percent. This translates into an 8.7-percent decrease in the child food insecurity rate over the study period, a significant reduction in the average county-level child food insecurity rate. While these findings support the assertion that increases in benefit levels may help reduce the number of households experiencing food insecurity, the sunset of the ARRA benefit level increases in October 2014 returned benefit levels to pre-recession amounts. The implications of these findings suggest that higher benefit levels may support households in achieving dietary quantity and quality standards that are necessary to avoid food insecurity.