Association for Public Policy Analysis & Management

Obesity is a major public health issue. Nearly 33.8% of American adults are clinically obese and approximately a third of children and adolescents are overweight or obese (Ogden et al., 2012). A child who is obese is at an elevated risk to be obese later in life (Biro and Wien 2010). An existing food assistance program that can be used as a tool to reduce childhood obesity rates is the Supplemental Nutrition Assistance Program (SNAP). The program is the largest domestic food assistance program with approximately 1-in-4 children receiving benefits (Ganong and Liebman 2013). This study examines the relationship between where a child lives and where their guardian's local SNAP benefits can be redeemed. We assess whether access to SNAP-authorized stores affects childhood body mass index (BMI).

We combine detailed administrative panel data from both the United States Department of Agriculture (USDA), New York City's (NYC) Department of Education (FITNESSGRAM), and the USDA SNAP Retailer Database. Our sample contains information on almost 1,000 schools and almost 600,000 students per year. This makes us the first study to use student level address data to calculate distances to SNAP-authorized store locations. We estimate how changes in the local SNAP food environment affect a child's BMI. Our main model specification predicts a NYC student's BMI based on either the minimum distance to a SNAP authorized store location or the number of SNAP authorized store locations within a network buffer while controlling for local demographic conditions, school characteristics, census tract fixed effects, and/or student fixed effects. Our identifying variation comes from changes in where SNAP beneficiaries can redeem their benefits.

We examine possible heterogeneity in responses to changes in access to SNAP-authorized store locations and stratify our sample based on the age, race and ethnicity of the student and household income of the guardian. We also determine whether the effect of a change in access varies for different types of stores. Regression models were run predicting the student's BMI based on the distance to the nearest supermarket and the number of supermarkets within a network buffer. Similar models were run for convenience stores and small grocery stores. Finally we run regression models predicting the student's BMI based on the introduction of a new SNAP-authorized store location near the student's home.

Our results are policy relevant for two reasons. First, the results contribute to the existing debate on whether to restrict where SNAP benefits can be redeemed. Second, the results help to clarify the effect of SNAP on childhood obesity.