Association for Public Policy Analysis & Management

The Supplemental Nutrition Assistance Program (SNAP) administered near-cash benefits to over 45 million Americans, valued at almost $74 million dollars, in 2015. The SNAP program is well studied, yet little is known about its effect on the prevalence of obesity-related diseases or associated Medicaid expenditures. This study provides new evidence of the effects of SNAP on clinically-diagnosed obesity and co-morbidities - including coronary heart disease, type 2 diabetes, high blood pressure, and metabolic syndrome - and related Medicaid expenditures. My analysis uses multiple waves of the National Health Interview Survey (NHIS) linked with longitudinal Medicaid eligibility and claims data.  

Prior evaluations of SNAP have focused on estimating its effect on health status and obesity. Many of these studies relied on self-reported outcomes from the same year as reported SNAP participation. These studies are subject to two limitations. First, self-reported measures of health are unreliable. Second, the effects of SNAP-induced increases in expenditures are unlikely to manifest in the very near term. Consequently, studies that evaluate SNAP with short-panels and self-report outcomes fail to identify unbiased, longer-term effects of the program. This is the first study to examine either the longitudinal relationship between SNAP and clinically diagnosed obesity-related disease or its effect on Medicaid expenditures.

To test the hypothesis that SNAP participation results in improved health outcomes and reduced Medicaid expenditures, I use restricted-use data from the National Center for Health Statistics. These data include survey information for respondents to 11 waves of the nationally representative NHIS, each wave of which has been linked to up to 10 years of Medicaid data. The analytic dataset includes household and individual level characteristics, including SNAP participation, self-reported health variables, and claims-level data for diagnoses, procedures, and Medicaid expenditures. In order to account for endogenous participation in SNAP, I use a bivariate probit model using state-level eligibility and enrollment policies as instrumental variables. Changes in each instrument – for example, simplified income reporting and short-recertification periods – have been shown to affect SNAP participation, and are unlikely to directly impact health outcomes.

Findings from this evaluation of SNAP will identify policy opportunities to improve long-term health outcomes among this at low-income population and reduce Medicaid expenditures.