Intergenerational nutrition benefits of India’s national school feeding program

Child stunting remains a major global public health challenge, with 149 million children short for their age and over half residing in Asia. In India, 38% of children were stunted in 2015–2016. Stunting reflects chronic undernutrition and is linked to deficits in development, human capital, and economic productivity. While global strategies emphasize nutrition-specific interventions during the first 1000 days, there is growing recognition that investments across the life course, addressing underlying determinants, are also important. In India, maternal height and education are among the strongest predictors of child stunting. The Mid-Day Meal (MDM) scheme, launched nationally in 1995, provides a free cooked meal (mandated minimum 450 kcal and at least two food groups) to children in government and government-assisted primary schools (classes I–V; ages 6–10 years). MDM has documented benefits for school attendance, learning, hunger, and resilience to shocks, suggesting potential intergenerational effects via improved female education and health. However, no prior studies had examined whether benefits to program participants carry over to their children’s nutrition. This study asks whether maternal exposure to the MDM scheme during primary school is associated with improved linear growth (HAZ) in their future children, and explores pathways and heterogeneity by socioeconomic status (SES).

The literature documents positive effects of school feeding and transfer programs on school participation and educational outcomes for girls in low- and middle-income countries. Evidence on impacts of transfers on anthropometry is less developed, often due to short evaluation horizons, though some studies (including from high-income contexts) show lasting health benefits from timely transfers. Intergenerational effects have been documented for cash transfers, health insurance, and other social programs on later-life outcomes such as income, education, nutritional status, and mortality. Frameworks for intergenerational transmission of health disparities highlight parental SES, adolescent development, and young adult capacities. In India, women’s height and education are key determinants of child stunting. Despite widespread implementation of school feeding programs globally, evidence on intergenerational nutrition effects is scarce. This study fills that gap by examining whether MDM exposure in girls’ primary school years translates into improved child growth outcomes in the next generation.

Data sources and design: The study combines multiple nationally representative datasets spanning 1993–2016: National Sample Surveys of Consumer Expenditure (NSS-CES; rounds circa 1993/94, 1999/2000, 2004/05, 2011/12), National Family Health Survey (NFHS-4, 2015–2016), and India Human Development Surveys (IHDS). MDM coverage is measured as the proportion of primary school-age children (ages 6–10) receiving free mid-day meals, constructed by state, socioeconomic stratum (SES deciles), and birth cohort. Coverage rose from ~6% (1999) to 32% (2004) and 46% (2011), with substantial inter-state variation. Exposure construction: Maternal exposure to MDM is assigned based on mothers’ birth year (1980–1998 cohorts), state of residence, and SES, mapping the share of girls in their primary school-age window reported to receive mid-day meals in NSS-CES. Because MDM exposure spans multiple school years, coverage is smoothed across years within 5-year intervals using linear interpolation (and tested with log-linear smoothing) to approximate cumulative exposure during ages 6–10. Sensitivity analyses use raw, unsmoothed coverage and alternative single-year coverage measures. Outcomes and covariates: Child height-for-age z-scores (HAZ) are computed from NFHS-4 anthropometry for children under five. Controls include child age, sex, birth order, maternal antenatal care (4+ visits), institutional delivery, urban/rural residence, region, and access to Integrated Child Development Services (ICDS) and Public Distribution System (PDS). Errors are clustered at the district level; models include district or state fixed/random effects as specified. Empirical strategy: 1) Birth cohort fixed effects models relate child HAZ to maternal cohort MDM coverage, controlling for maternal birth-year fixed effects, state, SES, and interactions (state-specific birth-year fixed effects), plus child- and service-access covariates. Heterogeneity by SES is examined via interactions (poor: SES deciles 1–3; middle: 4–6; non-poor: 7–10). 2) Controlled interrupted time series / difference-in-differences (DID) leverage staggered state roll-out (late 1990s to early 2000s) comparing intervention vs. control states before/after program start, estimating changes in HAZ slopes for children born to exposed vs. non-exposed maternal cohorts, with district/state effects and SES subgroup analyses. 3) Regression decomposition estimates the share of national HAZ improvements (2006–2016) attributable to MDM by combining estimated effect sizes with exposure duration. Robustness and sensitivity: Analyses test alternative smoothing (log-linear), use raw coverage, exclude specific states with early coverage, and assess heterogeneity by SES and by caste/religion proxies. Additional analyses examine associations of MDM exposure with plausible pathways among women (education, height, fertility, and health service use).

• Maternal MDM exposure and child HAZ: In birth cohort fixed effects models, children born to mothers from cohorts with 100% MDM coverage had HAZ higher by 0.40 SD compared to children of non-exposed mothers (p < 0.05), controlling for extensive covariates and fixed effects. Inclusion of ICDS and PDS access did not attenuate associations. - Heterogeneity by SES: Effects were stronger among poorer households: poor (SES deciles 1–3) +0.53 SD (p < 0.05), middle (SES 4–6) +0.33 SD (p < 0.05) relative to non-poor. Some robustness checks with additional fixed effects attenuated significance but overall effects remained. - Interrupted time series/DID: States that rolled out MDM earlier exhibited faster gains in HAZ among children of exposed maternal cohorts compared to control states, supporting the cohort model findings. - Contribution to national HAZ gains: Based on average exposure (≈2.6 years) and estimated per-year effects (~0.044 SD/year), MDM explains approximately 0.053–0.128 SD or 13.2–31.2% of the average national HAZ improvement between 2006 and 2016. - Pathways (associations among women born 1980–1998): Table 1 indicates that higher MDM exposure is associated with increased years of education (+3.95 years; SE 0.71; p < 0.001), modest, non-significant increases in height (+0.51 cm; SE 0.62; p = 0.163), later age at first birth (+1.62 years; SE 0.07; p < 0.001), fewer children (−0.80; SE 0.03; p < 0.001), greater antenatal care utilization (+0.22; p < 0.001), and higher institutional delivery (+0.28; p < 0.001). - Program coverage trends: Coverage among girls aged 6–10 increased from ~6% (1999) to 32% (2004) and 46% (2011), with substantial state heterogeneity that underpins identification. - Observational cross-sectional motivation: States with higher historical MDM coverage had lower stunting among under-fives in 2016, motivating causal investigation while recognizing potential confounding in simple correlations.

Findings suggest that exposure of girls to India’s MDM during primary school years is associated with improved linear growth in their future children, indicating intergenerational benefits of a large-scale school feeding program. The results align with frameworks emphasizing life-course and socioeconomic determinants of child nutrition: MDM appears to enhance women’s schooling, shift fertility to later ages with fewer births, and increase health service utilization, all factors known to reduce child stunting risk. The stronger effects among poorer households imply that school feeding can contribute to reducing nutrition inequalities by reaching children more likely to attend government schools and benefit from transfers. Demonstrating that a school feeding program, typically justified for education and short-term nutrition benefits, relates to better nutrition in the next generation contributes to policy debates that often prioritize the first 1000 days exclusively. These results support complementary investments across the “next 7000 days,” recognizing that improvements during primary school can yield long-run intergenerational nutrition gains. The broader implication is that social protection and school-based nutrition interventions can be integral components of national strategies to reduce stunting at scale.

This study provides evidence that India’s Mid-Day Meal scheme is associated with significant intergenerational improvements in child linear growth, particularly among poorer households, and may explain a substantial share of national HAZ gains between 2006 and 2016. Associations likely operate through increased female education, altered fertility patterns, and greater maternal health service utilization. Given the scale of school feeding programs globally, these findings underscore their potential contributions beyond immediate educational and nutrition outcomes to longer-term intergenerational health. Future research should strengthen causal identification (e.g., leveraging administrative roll-out data and quasi-experimental designs), follow cohorts longitudinally from school exposure through childbearing, and assess program quality, dietary content, and complementary interventions (e.g., adolescent nutrition, secondary school feeding) to maximize intergenerational benefits.

• Observational design: The study cannot follow randomized cohorts from school age to childbearing; estimates are associational and subject to residual confounding and potential endogeneity (e.g., MDM coverage correlated with unobserved state-level improvements). - Measurement of exposure: MDM coverage is constructed from repeated cross-sections and smoothed; exposure timing and intensity may be measured with error. Sensitivity analyses with raw and alternative smoothing suggest robustness but cannot fully eliminate measurement concerns. - SES and compositional matching: SES is matched across datasets and over time using expenditure proxies and deciles; misclassification or mobility could introduce bias. Additional caste/religion matching was exploratory. - External influences: Concurrent policies (e.g., Supreme Court mandates, other social programs) and broad secular trends may contribute to observed improvements despite extensive controls and fixed effects. - Generalizability of pathways: Associations with women’s height were not statistically significant; mechanisms are inferred from associations rather than directly tested causally.