The household return to home-grown school feeding: why income understates the welfare gain in Malawi

Home-grown school feeding programs (HGSFPs) are expanding across Africa on the assumption that benefits propagate through income: participation raises farm revenue, and richer households invest more in food and schooling. We test this directly using survey data from 662 smallholders in Kasungu District, Malawi (332 non-participants, 330 HGSFP participants, all supplying through the district’s sole cooperative procurement channel). Identification combines a two-step endogenous switching regression with regime-specific inverse Mills ratios as the primary estimator, with three propensity-score-matching algorithms, inverse probability weighting, and entropy balancing as robustness; the estimators converge. Participation roughly doubles agricultural revenue (ATT 0.72–0.88 log points, 106–141%), uniformly across household-head gender and age [ F (3, 651) = 0.48, p = 0.693]. It also improves food security (Household Food Insecurity Access Scale falls 1.79 points) and sharply raises voluntary school contributions (MWK 5,331 per child). Baron–Kenny decomposition shows income mediates 100% of the revenue effect by construction, 23% of the food-security effect, and under 1% of the school-contribution effect. We test several candidate direct channels: participants receive significantly more extension advice and derive a higher farm income share, consistent with a productivity account; we find no evidence for home consumption, credit access, or trust in schools as carriers. HGSFPs, on this evidence, are bundled financial-and-social-capital interventions whose returns extend well beyond the income channel conventional cost–benefit logic captures—and income-only evaluations understate the social return most where the stakes for children are highest.