Association for Public Policy Analysis & Management

In addition to being among the oldest and largest child development programs in the United States, the National School Lunch Program (NSLP) is a crucial data source in American public education. The program provides free lunch to students whose household income is less than 130% of poverty and reduced-price lunch to students who household income is between 130% and 185% of poverty, based on student or parent reports of total household income.

Since free or reduced price lunch participation is often the sole available indicator of student socioeconomic status available in K-12 administrative data, the NSLP also plays a central role in the allocation of school finances as well as social science relating to schools. School finance policies – including federal Title I funds as well as state and local weighted student funding formulae – use NSLP measures to target supplemental funds to poor students. Likewise, educational researchers routinely categorize students as poor or non-poor based on their participation in the NSLP and represent school poverty based on the rate at which students qualify for free or reduced price lunch.

In this paper, we investigate the validity of free or reduced price lunch program participation as a proxy for students’ household income. We merge K-12 administrative data from two sources – a mid-sized California school district and the state of Oregon – with household income data derived from household members’ annual IRS 1040 filings. Pilot work using the California data indicate that we can match more than 90 percent of students in our administrative data with IRS household income records. We are currently matching the Oregon data. These merged data make it possible to precisely measure students’ household incomes and contrast these measures with school-reports of students’ free or reduced price lunch participation.

Preliminary analyses suggest that NLSP participation data align poorly with student income. Considerable variation in household income-to-poverty ratios among students who qualify for free lunch, students who qualify for reduced-price lunch and student who qualify for no free lunch. Approximately half of the California students for which we have both IRS-reported household income data and NSLP program participation data appear to be misclassified in the NLSP data either because they receive free or reduced-priced lunches for which they do not appear to be income-eligible or because they do not receive free or reduced-price lunches despite having a low enough household income to qualify. However, nearly 80% of students that receive free or reduced price for which they do not appear eligible where program eligible at some point in their educational career, suggesting that students may be grandfathered into the NSLP. In this district at least, the level of program eligibility misclassification varies across racial/ethnic groups and school, with income-eligible white students having notably high levels of program non-participation.

Further analyses using the Oregon data will explore cross-school variation in measurement error associated with NSLP data; its implications for educational research; and alternative measures, including Michelmore & Dynarski’s (2016) measure of persistent free lunch eligibility, to address these issues.