The Supplemental Nutrition Assistance Program (SNAP) serves more than 42,000 people to reduce food insecurity and improve nutritional well-being by providing $126 average monthly benefits per person (2017). Although the effects of SNAP participation on food insecurity are well documented, the degree to which SNAP reduces food insecurity of households with a member(s) with a disability—households that are more likely to participate and less likely to be food secure—is less recognized. In November 2013, SNAP benefits decreased by around 5 percent, on average, due to the expiration of the American Recovery and Reinvestment Act. If this decrease impacts SNAP’s effectiveness in reducing food insecurity, the effects may differ between households with and without a disabled member. However, considerably less research has been conducted on this issue.
To fill these research gaps, this study uses the public- and restricted-access National Health Interview Survey (NHIS) 2011–2015 and examines how much the effects of SNAP participation on alleviating food insecurity differ by the presence of a member(s) with a disability in a household and who in the household has disabilities––household head, spouse/partner, and/or children. The study also estimates how much the downsized benefits reduce SNAP’s effectiveness for households with and without a disabled member. An analytical sample for this study comprises 64,209 households below 200 percent of the Federal poverty line, which is broken down into 29 percent SNAP participant households and 25 percent food-insecure households.
To obtain more efficient estimates, a copula-based endogenous switching regression model is employed in which distributions of unobservables are featured by the Ali-Mikhail-Haq (AMH)–Plackett copula function that replaces joint normal distribution and is selected as the best fitting combination of copulas among 25 alternatives. To aid model identification, Sate-specific SNAP policy variables––broad-based categorical eligibility, a vehicle restriction for SNAP asset test, and the combined application process for SNAP––function as instruments. Based on the parameter estimates, the program effects are estimated on the framework of the unconditional and conditional average treatment effects on the treated.
For overall households, SNAP participation reduces the probability of food insecurity by 13.2 percentage points. For those with and without a disabled member, SNAP participation reduces the probability of food insecurity 15.3 and 11.4 percentage points, respectively. For households headed by a person with and without a disability, SNAP participation reduces the probability of food insecurity by 17 and 11.9 percentage points, respectively. On the other hand, SNAP has a greater effect on reducing the probability of food insecurity for households headed by a person with a spouse/partner without a disability than those with disabilities or households headed by a person without a spouse/partner by 13.9 and 11.5 percentage points, respectively. There is little difference in the effects of SNAP by children’s disabilities. Additionally, the results suggest that the downsized SNAP benefits reduce the program’s effectiveness by 3.7 percentage points for households with a member(s) with a disability and 3.9 percentage points for households without a disabled member, with very few differences between those two cohorts.
Overall, this study finds that SNAP participation is more effective in mitigating food insecurity for households with members with disabilities than those without a disabled member and overall households, and the effectiveness of SNAP substantially varies by who in the household has disabilities. This study adds to the literature that the downsized SNAP benefit is related to decreased program effectiveness. In particular, the program effects are rarely sensitive to the use of State-specific SNAP policy variables, but vary by a selection of copula joint distribution functions, meaning that distributional assumptions on unobservables are decisive on the estimated program effects, rather than the use of the instruments that satisfy exclusion restrictions.