The Supplemental Nutrition Assistance Program (SNAP) is the largest Federal food and nutrition program and the second largest Federal means-tested transfer program. Participation in the program and real average monthly payments grew rapidly in the past decade. We investigate the impacts of changes in SNAP benefit payments on earnings in nonmetro and metro counties during the 2000 to 2010 period using three econometric methods: i) ordinary least squares fixed effects (OLS-FE) regressions, ii) instrumental variables fixed effects (IV-FE) regressions, and iii) spatial discontinuity OLS and IV first difference (SD-OLS-FD and SD-IV-FD) regressions. The estimated impacts vary considerably across the estimation methods, and all methods suffer significant weaknesses. The OLS-FE regressions appear to be affected by unobserved heterogeneity and endogeneity biases. The instrumental variables used in the IV regressions (state-level SNAP policy variables) are found not to be exogenous in most cases, and are weak in the SD-IV-FD regressions. The most plausible results are found using the SD-OLS-FD model, but large standard errors in this model limit the statistical power to draw confident conclusions. The most interesting finding is of a more positive (or less negative) association of SNAP payments with earnings during the Great Recession. This difference supports our hypothesis that the positive impacts of SNAP payments are likely to be larger during a recession. Several avenues for further research are suggested.