Fostering human capital development through school feeding: Evidence from Madagascar

Schools involved in the home-grown school feeding program procure produce from local smallholder farmers’ organizations and store it; volunteers from local communities take turns in preparing the food every day. Children receive a warm daily meal on the school premises during the school day. Meals are prepared following predefined recipes to ensure they are healthy and nutritious.

2024-09-02

2026-06-30

First Stage Outcomes: child dietary diversity, child food insecurity experience scale
Second Stage Outcomes: child mental health/psychological wellbeing (life satisfaction, stress, depression, agency), child physical health (health status, nutritional status), child school attendance, child school progression (dropout, repetition, new enrollment), learning (reading skills, numeracy and mathematics skills), child cognitive ability (attention span, verbal working memory, fluid intelligence)

We leverage the randomized expansion of the national school feeding program in Madagascar to understand if the program is effective in bringing children to the classroom and fostering better health and education. We randomize the provision of school feeding at the commune level (third highest administrative unit). All eligible schools within treatment communes are selected to receive the intervention, while eligible schools in control communes do not receive the intervention during the study period. We compare the outcomes of children enrolled in schools where a school feeding program is implemented to the outcomes of children enrolled in schools where a school feeding program is not implemented. Out of the 138 communes in the sample, we assign 68 communes with 121 eligible schools to the treatment group and 70 communes with 100 eligible schools to the control group. Out of these, 99 schools in the treatment group and 100 schools in the control group are randomly selected for data collection.

Not available

Randomization done in office by a computer

Commune (third highest administrative unit)

Yes

138 communes

1,990 students and 199 school administrators from 199 schools.

Treatment: 68 communes, 99 schools
Control: 70 communes, 100 schools

Power calculations for the school design are based on five measures: household food security (FIES, mean 4.65 and standard deviation 2.21), weight-for-age z-scores (mean –1.12 and standard deviation 2.16), height-for-age z-scores (mean -1.11 and standard deviation 2.31), reading ability ICW index (mean –0.12 and standard deviation 0.28) and math ability ICW index (mean 0.00 and standard deviation 0.96). Data for household food security comes from a household survey targeted at capturing FIES in Madagascar in 2021 by FAO. Weight-for-age, height-for-age, and reading and math ability come from the MICS6 dataset conducted in Madagascar in 2018 by UNICEF. Assuming 100 schools per treatment arm and 10 children per school, two follow-up observations, a correlation of 0.2 between follow-up measurements, an intra-cluster correlation of 0.2, and power of the test of 80%, yields a minimum detectable effect (MDE) of 0.16 standard deviations for food security and height-for-age, 0.17 standard deviations for weight-for-age and math ability, and 0.18 standard deviations for reading ability. With three follow-up measurements, the study would be powered to detect an effect of 0.14 standard deviations for food security, weight-for-age, and reading ability and 0.15 standard deviations for height-for-age and math ability. In conclusion, the experiment is likely to be powered to detect impacts on food security, reading and math ability, and anthropometrics with at least three follow-up rounds of data collection at the child level.