Wholesome Lunch to the Whole Classroom: Short- and Longer-Term Effects on Early Teenagers' Weight

Previous studies on the effect of school lunch programs on child obesity have been hampered by effect heterogeneity, self-selection, and stigma-induced under-reporting, having produced mixed findings. Their potential long-lasting effect has also been debated. We study the body-weight effect of a Japanese school lunch program, which provides nutritional lunch to all students at participating municipal junior highs. The lack of means testing and individual participation choice offers easily interpretable causal estimates. By exploiting almost all school lunch coverage for elementary school children nationwide, we construct a difference-in-differences (DID) framework to alleviate potential bias due to unobserved differences across municipalities. Using the 1975?1994 National Nutrition Survey, a nationally representative household survey with measured height and weight, we find a regressive benefit of school lunch: while no statistically significant effect is found for the full sample, we find a significant obesity-reducing effect for the subsamples of children with low socioeconomic backgrounds. The obesity-reducing effect remains at least a few years after graduation, implying effect through not only nutritional contents but also guiding healthy eating behavior. We find little evidence that school lunch reduces underweight. Propensity score weighting, quantile DID analysis, and various falsification tests confirm the robustness of our estimates.