School administrators often face challenging trade-offs when deciding how best to spend money to maximize student achievement. When making these decisions, the food students eat while at school is rarely considered an important input to the education production function, but previous studies have produced tantalizing evidence that providing students with low-cost, nutritious meals may lead to better academic outcomes. This study adds to this fast-growing literature by exploiting arguably exogenous variation in the receipt of free school lunches generated by the Community Eligibility Provision (CEP) of the Healthy, Hunger-Free Kids Act of 2010.
The CEP allows every student in a school to receive free school lunches if at least 40 percent of its enrolled students meet the identified student population (ISP) criteria. Students who directly qualify for a free lunch or participate in the Supplemental Nutrition Assistance Program (SNAP), the Temporary Assistance for Needy Families (TANF) program, Medicaid, or Head Start meet the ISP criteria, as do homeless students and those in foster care. The CEP also requires that schools follow a set of specific school-lunch guidelines that accompanied the Healthy, Hunger-Free Kids Act. By expanding the proportion of students who receive free school lunches and, at the same time, improving the nutritional content of these meals, the CEP provides an opportunity to examine the complicated interactions between policy, nutritional intake, and cognition in a real-world setting.
The CEP was piloted in 10 States starting in the 2011-12 academic year and was made available to all States in 2014-15. This study’s research strategy exploits the timing of the program rollout in pilot States, as well as the cutoff used to determine eligibility and the amount of funding provided. Specifically, the analysis compares the academic performance of students attending schools that qualify for CEP in 10 pilot States with the academic performance of students attending similar schools in neighboring States. The analysis also uses a regression discontinuity design to compare the academic performance of students who attended schools that fell on either side of the 40-percent CEP eligibility cutoff or the 63.5-percent cutoff where it becomes financially worthwhile for schools to participate.
The most important contribution of this project was the gathering of five major datasets that can be merged. These datasets include school-level test score data (2003-16); school-level characteristics from the National Center for Educational Statistics (NCES) Common Core of Data (2003-14); school-level ISP and CEP data (2014-16); and district-level, test-score data (2009-13). The ability to conduct the appropriate analyses is currently limited because there is no single year in which all five of these datasets are available. However, the study will be able to provide the analyses outlined in the initial proposal as additional, overlapping years of data become available.
In addition, there are several complicating, unforeseen challenges with analyzing the impact of the CEP. First, districts can group schools and use their combined ISP to determine eligibility and funding. Second, Title I also uses a 40-percent cutoff for eligibility, although this cutoff is based on the fraction of students receiving a free or reduced-price lunch rather than the ISP. Third, once a school qualifies for the CEP, it can continue in the program even if its ISP rate drops below 40 percent. Fourth, States are required to report the ISP rate for eligible schools but not for ineligible schools. Fifth, funding increases as a school moves from 40-percent to 63.5-percent ISP, creating a regression kink rather than a regression discontinuity. These are all aspects of the program that make it more difficult to evaluate it using standard empirical approaches. Efforts are underway to address each of these issues.
In the course of conducting this analysis, four critical needs were identified that would facilitate this and future research. First, data on school-level lunch participation rates (separately by free, reduced, and paid) would allow a better understanding of the factors that contribute to whether students eat a school lunch and how these dynamics play out over time to influence the stigma experienced by low-income students. Second, ISP data for all schools (and not just the eligible schools) would dramatically improve the ability to conduct a regression kink design by being able to map the pattern observed for all schools. Third, information of the fraction of students eligible for a free or reduced price lunch is currently available only through 2014. Releasing the Common Core of Data in a timely fashion would aid in merging the various data sources needed to complete this project. Fourth, a school-name crosswalk would make it easier to merge across different school-level datasets.