Using Variation in State Policies to Estimate the Effects of SNAP on Child and Adult Health

Background and Purpose

The Supplemental Nutrition Assistance Program (SNAP) is a substantial part of the social safety net, but its effects on outcomes other than nutrition and food insecurity are not well understood. In particular, the relationship between SNAP and health has received only limited attention. SNAP receipt may alleviate hunger and stress and thus improve physical or mental health. As a near-cash benefit, SNAP receipt may free up resources that can be used for health promotion. Using variation in state SNAP policies as a natural experiment, this paper estimates whether the receipt of SNAP benefits is associated with child and adult physical and behavioral health.

Methods

This paper uses nationally-representative data from the 2008-2013 waves of the National Health Interview Survey (NHIS) and matched state data from the USDA’s State SNAP Policy Database.  Following state rules, we restrict our sample to SNAP-eligible NHIS respondents on the basis of household income, the presence of older adults or disabled persons in the household, and participation in other public programs.

We assess whether SNAP participation (measured as a simple indicator and months of participation in the past year) is associated with a number of adult and child health outcomes in the NHIS, including self-rated health, whether respondents delayed necessary care, whether respondents reported having a stomach illness or cold in the past two weeks, whether respondents needed but couldn’t afford necessary care, adult and child depressive symptomology, and many others.

To address concerns about non-random selection into SNAP participation, we use instrumental variables analysis. This technique hinges on using “instruments” that are as-good-as-randomly assigned and strongly associated with SNAP participation but not the health outcomes of interest. States vary in their policies regarding the SNAP program, providing an arguably random source of variation in participation that can minimize concerns about selection when studying program impacts. This approach has been used in previous SNAP research.  We created a standardized index of nine state SNAP policies (alpha=0.80), which is strongly associated with SNAP participation, and controlled for a number of additional potential confounders including health insurance coverage, family income and size, the year and month of the survey, and a number of state-level variables.

Results

Among others, results suggest that SNAP participation is associated with significant improvements in general health for children and adults, significantly higher probability of very good or excellent health for both adults (0.24) and children (0.43), significantly lower probability that adults reported a stomach illness in the past two weeks (-0.07), and that adults and children had to go without needed dental care (-0.40 and -0.47, respectively) or glasses (-0.27 and -0.20, respectively). SNAP participation was not significantly associated with child or adult behavioral health.

Conclusions

Our results indicate that SNAP participation has positive impacts for children and adults that extend beyond food security and nutrition. They suggest the importance of redoubling efforts to promote participation in SNAP and the potential value of expanding benefits or eligibility as a means to promote health among low-income Americans.