School meal quality and academic performance

All California public schools over a five-year panel (school-by-grade-by-year observations); contract subsample restricted to districts with a vendor contract in at least one year

California public elementary, middle and high school students, ages ~5-18. Well-fed by global standards; ~half economically disadvantaged and eligible for free/reduced-price lunch.

Difference-in-differences exploiting frequent turnover of school-meal vendor contracts. Vendors were classified healthy vs standard by Healthy Eating Index scores computed by nutritionists at the Nutrition Policy Institute (healthy = above median HEI). Outcome is the California STAR standardized test, independently administered. School-by-grade and year fixed effects. Placebo treatment coded one year before the actual contract yields a coefficient <0.01 SD and insignificant, so pre-trends are flat; falsification tests on observable school characteristics are also null. This isolates meal QUALITY holding meal provision roughly constant, which is closer to a nutrition channel than any program-level study.

Contracting with a healthy meal vendor raises test scores 0.03-0.04 SD relative to in-house provision (preferred estimate 0.036 SD, significant at the 0.1% level; contract subsample 0.036-0.040). Standard (non-healthy) vendors show a positive but insignificant effect for the whole population, though a significant positive effect for disadvantaged students - consistent with a pure calorie/palatability channel. Effect for disadvantaged students 0.045 vs 0.032 for advantaged, difference not statistically significant. No detectable effect on obesity rates. Cost is at most ~$80 per test-taker per year, implying <=$222 per 0.1 SD - competitive with the most cost-effective known interventions and far cheaper than STAR class-size reduction.

Low. Vendor contract turnover is a district procurement decision, not a family choice, and the placebo/pre-trend tests rule out schools with improving students selecting into healthy vendors. What the design cannot rule out is that a district switching to a healthy vendor is simultaneously improving on other unobserved margins.