Association for Public Policy Analysis & Management

The Community Eligibility Provision (CEP) of The Healthy, Hunger-Free Act of 2010 allows schools or districts to provide free school meals to all students regardless of income, if at least 40 percent of students participate in other specific means-tested programs (i.e. SNAP, TANF, etc.). Advocates claim Universal Free Meals (UFM) will reduce stigma, food insecurity, hunger, and administrative burden, and improve nutrition and student readiness to learn. Critics, on the other hand, worry UFM may adversely affect student weight outcomes (i.e., increase obesity) by inducing students to eat too much (two breakfasts, perhaps) and/or less nutritious foods. There is, unfortunately, little empirical evidence on the academic and health impacts of UFM adopted under CEP. Schwartz & Rothbart (2017) estimate the impact of UFM adopted by a set of New York City (NYC) middle schools, finding positive academic effects but little impact on weight. The effects of district-level policies may well differ from school-level policies and, more generally, students outside NYC may respond differently. In this paper, we use data on New York State (NYS) schools and districts to explore these, examining the effects of UFM adopted under the CEP on test scores and obesity.

We combine data on 698 school districts for 2009-2017 from the NYS Departments of Education and Health, NYS Comptroller, and the American Community Survey. Data include student demographics, school meal participation rates, enrollment, financial resources, test scores and new data on obesity rates (overall and among primary and secondary students separately). We also use school-level data on demographics, meals participation rates, enrollments, and test scores to explore heterogeneity in estimated effects.

We use two key empirical strategies. First, we use a difference-in-differences approach, exploiting the precise timing of CEP adoption. While CEP adopting districts are likely different from others, it is also likely that the precise timing of adoption is conditionally random among NYS districts. Thus, we compare early and late adopting districts in our difference-in-differences design. Second, we employ a synthetic control design, constructing a weighted combination of non-CEP districts that are observationally similar to CEP districts. We explore heterogeneous treatment effects by school and district characteristics to assess differences in response to CEP. We derive credibly causal estimates of the academic and health consequences of UFM, informing the debate on the extent to which and for whom UFM delivers benefits (increased test scores and reductions in obesity) or costs (both direct costs and increased obesity). The findings from this study will aid policymakers targeting expansions or contractions of UFM policies.