Association for Public Policy Analysis & Management

The Supplementary Nutrition Assistance Program (SNAP) is a key strand in the federal social safety net and by far the largest U.S. food assistance program, providing recipient households with over $70 billion in benefits a year. During the Great Recession the program caseload rose to historically high levels, but was slower to decline after the recession officially ended, raising concerns that SNAP recipients were not responding appropriately to improving economic conditions by securing employment that would enable them to leave the program.

Part of this seeming disconnect between SNAP caseloads and economic conditions can attributed to a focus on comparing the SNAP caseload to labor market indicators at the national, and even state, level. Indicators at this level of aggregation may obscure important heterogeneity in local, and industry-specific, labor markets that are most relevant to able-bodied SNAP participants’ employment prospects.

This paper examines the relationship between local labor market conditions and local SNAP caseloads. We evaluate two definitions of local labor market areas (LMAs): one based on county boundaries and another based on commuting zones (CZs). Even when local LMAs are delineated in a way that reflects economic integration rather than political and administrative units—-as CZs attempt to do—, there may still be important spatial interdependence across areas that can bias estimates from conventional (i.e., non-spatial) regression methods. To address potential spatial and temporal correlation in the (observed and unobserved) determinants of SNAP caseloads across LMAs, this paper employs spatial panel models, which to our knowledge have not been previously used to study the determinants of movements in the SNAP caseload over time. We isolate the effect of demand-side factors by estimating models using Bartik-style instruments for local labor market conditions, and drill down to industry-level indicators within LMAs in an effort to better capture local labor demand for SNAP recipients (as opposed to local labor demand more generally). We have collected monthly county-level SNAP caseload data from 43 states, spanning a period from approximately 2000 to 2016 (the span of years varies for some states). This breadth of states and long panel permit us to exploit considerable variation in local economic conditions as well as in state-level SNAP policies.

The spatial models we use provide information on the effect of labor shocks in one LMA on the SNAP caseload in the same LMA as well as in neighboring LMAs (and potentially beyond, depending on the structure of the spatial weight matrices). The results from this study provide program administrators and policymakers with a more complete picture of the factors driving SNAP caseloads at the local level and a guide for how labor market shocks in nearby areas may eventually impact SNAP caseloads in their own area. The results from this paper highlight the importance not only of local economic conditions for SNAP caseloads but also the economic interdependence of LMAs.