Association for Public Policy Analysis & Management

The National School Lunch Program (NSLP), authorized by Congress in 1946 to “safeguard the health and well-being of the Nation's children”, is now the second largest nutrition assistance program in the U.S., subsidizing 30 million meals for school children each school day. In traditional public schools, lunch is provided free to students with household income up to 135% of the federal poverty line, and at a reduced price to students with household income up to 185%.  A growing number of schools (and districts) adopt an alternative, introduced in 1980, Universal Free Meals (UFM), under which school lunch is free for all students, regardless of income. UFM advocates argue that it will reduce the stigma that limits participation and address hunger and food insecurity for needy students, while reducing administrative burden and improving student outcomes. Critics are skeptical about benefits, charging that school lunch may increase obesity and have other adverse health effects. 

Despite its longevity, rigorous research into the impact of school lunch on student outcomes is relatively thin, due, in part, to the scarcity of data linking participation in school lunch to student performance, demographics, etc. and also to the difficulty of disentangling the causal effect of the program from the effects of poverty. Two notable exceptions draw on national survey data -- Hinrich (2010) finds positive long run effects on education (but not health) and Schanzenbach (2009) finds deleterious effects on obesity. In this paper, we exploit differences in the timing of UFM adoption by NYC public schools to estimate its impact on student participation in school lunch and academic and health outcomes. More specifically, we use detailed student-level data to estimate the impact of UFM on school lunch participation, academic performance, obesity, BMI and attendance using a difference-in-differences design with school (or student) fixed effects and a range of student and school control variables.

Unlike previous work that relies on survey responses about participation in school lunch, we use daily data on student lunch transactions for 300,000 students in 450 schools. These allow us to explore the heterogeneity of the effect across subgroups defined by gender, race/ethnicity, grade, individual eligibility status and, critically, prior participation in school lunch. To our knowledge, this is the first study to use such individual, daily, longitudinal data on lunch participation.  Finally, we explore the responsiveness of student outcomes to lunch participation, using UFM as an instrumental variable to ameliorate potential endogeneity.

Preliminary results indicate that UFM increases school lunch participation– with a larger effect for students in high school and middle school than elementary school, and larger for “full pay” students than those who had been individually eligible. UFM increases academic performance for middle school students by as much as 0.1 standard deviations with no significant effect for elementary school students. We see little overall effect on BMI and obesity, but find some evidence that SLP reduces weight and obesity in high schools -- due, perhaps, to the better nutritional value of school lunch compared to alternatives including fast food.