Maoyong Fan, Ball State University
Yanhong Jin, Rutgers University
American Journal of Health Economics 1(4): 432-460, 2015
In the United States, children living below the federal poverty line have much higher obesity rates than the national average – 12.5% vs. 10.7% for preschoolers aged 2-5, 21.5% vs. 17.4% for children aged 6-11, and 23.1% vs. 17.9% for adolescents aged 12-19 (USDHH 2010). The largest federal food/nutrition program, the Supplemental Nutrition Assistance Program (SNAP), provides food assistance for low-income households and more than half of its participants are children and adolescents. Because of the higher prevalence of obesity in SNAP child participants than in their counterparts, the SNAP is suspected to have contributed to the childhood obesity epidemic by encouraging consumers to consume high caloric foods1. The task force on childhood obesity established by President Obama in 2010 recommends invigoration of the three largest federal food/nutrition programs, specifically, the SNAP, the School Breakfast Program, and the National School Lunch Program, in the campaign against obesity because they reach millions of children in low-income households (Barnes, 2011). In our recent article in the American Journal of Health Economics, “The Supplemental Nutrition Assistance Program and Childhood Obesity in the U.S.: Evidence from the National Longitudinal Survey of Youth 1997”, we examine whether the SNAP has contributed to childhood obesity among low-income children.
Studies investigating the SNAP effects on childhood obesity have yielded mixed results and are far from a consensus. The SNAP effect ranges from being statistically negative to statistically positive and appears to be age- and gender-specific. Estimating the SNAP effect on the obesity risk and the BMI (body mass index) is challenging for the following reasons. First, there is no controlled random experiment because the SNAP participants must meet eligibility requirements. Second, duration of participation and benefit amount vary significantly across households. Accurate measurements of both duration and intensity of SNPA participation is critical for an accurate estimation of the SNAP effect on weight measures. Third, the BMI changes over a certain period of time and not overnight, which suggests the importance of examining the long-term SNAP effect.
The National Longitudinal Survey of Youth 1997 (NLSY97) is the best available data set because it provides solutions to the challenges we discussed above. First, the NLSY97 collects the self-reported body weight and height measurements over time, which allows us to examine the changes in the BMI and the obesity risk rather than the level to control for the time-invariant confounding factors such as genetic characteristics as well as the pre-treatment BMI. Second, the NLSY97 collects detailed information regarding SNAP participation and the benefits received. It shows that the amount of the SNAP benefits is positively related to the SNAP participation intensity and duration. If the SNAP has significant effects on the weight measures, we would expect long-term participants to be affected more than short-term participants. Thus, we define our main treatment group as participants who enroll in the SNAP at least half of the period under study. Furthermore, the detailed income information allows us to better define eligible nonparticipants. To avoid comparing the SNAP participants with non-eligible individuals who have more resources and are less likely to have food insecurity problems, only eligible nonparticipants whose household income is below 130% of the federal poverty line are included in the comparison group.
Finally, the NSLY97 collects rich information on the personal characteristics and environmental variables that are correlated with the SNAP participation and weight measures. Therefore, we are able to deal with self-selection by using the NLSY97 and employing different-in-difference propensity score matching (DID-PSM) to estimate the SNAP effects on weight measures of adolescent participants aged 12-20.
We find little evidence that the SNAP positively or negatively affects the BMI or the obesity risk among child participants aged 12-20 in one-, two- and three-year durations. The results are insensitive to various robustness checks including redefining the treatment and comparison groups by excluding those who had previous SNAP participation experiences, using an alternative definition of the treatment group based on the SNAP benefits received, using different specifications of the propensity score equation, and employing different semi-parametric techniques (covariate matching and inverse probability weighting). The robustness analyses regarding unobservables also confirm the main results. We thus conclude that the SNAP does not contribute to childhood obesity in low-income households in the United States.
This study highlights two aspects that future researchers should consider when evaluating the SNAP effect on childhood obesity. First, SNAP participation should be more carefully examined because, according to the NLSY97, not all the participants receive equal treatment and the differences in the SNAP benefits are significant across participating households. Second, examining the changes in the weight measures allows researchers to control for important unobservables (e.g., genetic factors) that might contribute to childhood obesity. Using the level of weight status is likely to generate false positive results.
In summary, we conclude that there is no evidence indicating the SNAP has contributed to the childhood obesity epidemic in the United States. The disparity in the prevalence of childhood obesity between low-income and high-income households is likely to originate beyond federal nutrition programs such as the SNAP. Some proposed SNAP changes such as more frequent benefit distribution are likely to be ineffective in reducing childhood obesity. However, changes in the SNAP to induce its participants to develop and maintain healthier dietary habits such as consuming more fresh fruits and vegetables could potentially reduce childhood obesity in the long run. Rigorous research is warranted to assess such potential. The SNAP reaches millions of children in low-income households, and our results provide a solid basis for the proposed reinvigoration of the SNAP by the task force on childhood obesity established by President Obama in 2010.
1 “Stop subsidizing obesity” by Mark Bittman, New York Times, December 25, 2012. URL: http://opinionator.blogs.nytimes.com/2012/12/25/stop-subsidizing-obesity/?_r=0.
USDHH. 2010. Healthy People 2020. edited by US Department of Health and Human Services. Washington DC: US Government Printing Office.
Barnes, M. (2011) “White House Task Force on Childhood Obesity report to the President: solving the problem of childhood obesity within a generation.” In Task Force on Childhood Obesity.