Attachments
By Amina Mendez Acosta and Biniam Bedasso
As school meal programmes continue to expand globally, so too is the body of evidence from impact evaluations. Given the multiplicity of objectives and the diversity of contexts in which school meals programmes are implemented, there is still a need for more rigorous evidence on the impacts of school feeding and the underlying causal mechanisms. Reviews over the past decade have shown mixed effects of school meals on nutritional and educational outcomes. We have conducted a rapid review of the most recent evidence, and have included studies which came out in the last five years, in addition to those that were published in the decade before that. As a result, we have added ten new evaluations, some of which focus on the large national programmes in India and China, that have not been included in any systematic review to date.
So, what does the latest evidence from the last 15 years, a period of rapid expansion in school meals programmes across the developing world, tell us?
What did the existing evidence tell us?
Earlier reviews demonstrate that school meals have a positive impact on some educational outcomes. A study by Snilstveit et al. in 2015 reviewed 16 programmes in developing countries and found that school meals improved enrollment, attendance, and learning by small but meaningful amounts. Wang et al. in 2021 updated this by reviewing 19 studies published through 2019, highlighting notable gains in height, weight, and attendance but no significant improvement in mathematics scores. Other reviews, like Wall et al. in 2022 focusing on Africa and Cohen et al. in 2021 looking at countries worldwide, also found school meals benefit education. Our review adds studies conducted more recently to this evidence base, and includes an analysis of the associated cost data.
The scope of our review
We drew on existing systematic reviews to compile the evidence on school meals programmes located in low- and middle-income countries and published after 2010. We also looked for eligible studies from papers that cite those reviews in addition to a more general search of school feeding studies in institutional websites. In total, we found 27 studies from 17 countries in Latin America, sub-Saharan Africa, and Asia. China and India both have five studies each on their respective national school feeding programmes while the rest of the countries have either one or two studies. Figure 1 in the PDF shows a map of the countries included in our review. Twenty-three of the 27 programmes covered by the studies provided breakfast or lunch in school while the rest offered snack and/or take-home ration.
The majority of the evaluations in our review (21 of them!) use either a randomized controlled trial, a difference-in-differences or a regression discontinuity approach to estimate the effect of school feeding on outcomes. Half a dozen of the studies use somewhat less rigorous methods such as instrumental variables and matching techniques. We list the studies in Table A4 of the online appendix, available here.
Key findings
1. School feeding continues to generate a sizable impact on school access and learning outcomes
We compiled data from all school feeding studies that report impacts on enrollment and attendance, finding an average effect size of 0.13 and 0.12 standard deviations respectively. Notably, the studies show even larger effects on learning, with increases of up to 0.15 standard deviations. This suggests that having access to school meals could improve test scores by as high as 15 percent of the usual variation in test scores across the study samples. Studies using more rigorous analytical methods report slightly lower but still significant effects on learning outcomes, around 0.08 standard deviations. The effect sizes on attendance are the most robust to changes in evaluation methodology.
To put these figures in context, a review of over 200 education interventions found a median effect size of 0.07 standard deviations for improving access outcomes. For interventions targeting learning, the average reported effect size is 0.1 standard deviations.
It's also important to note that school feeding programmes often also have goals relating to nutrition and social protection, in addition to education. As a result, their overall social impact likely exceeds the effect size measured solely in terms of education outcomes—something that is not typically true for traditional education interventions.
2. The effect sizes are comparable to other interventions aimed at improving education outcomes
How do these effect sizes compare to other interventions aimed at achieving similar outcomes, or to earlier school feeding programmes? Figure 3 presents effect sizes of various interventions based on the comprehensive review by Snisltveit et al. at the International Initiative for Impact Evaluation.
The impact of school feeding interventions reported in the past 15 years is at par with the effect sizes of earlier studies published prior to 2015. In fact, more recent evidence shows even greater effects on learning and attendance (it’s important to note that there are some differences between the earlier systematic review and our rapid review, particularly regarding the inclusion of quasi-experimental evaluations without baseline data). On the other hand, the sizable decline in the effect on enrollment may be related to many middle-income countries approaching universal primary enrollment already.
While some interventions—such as community-based monitoring, cash transfers, and multilevel interventions—achieve larger impacts in specific areas like enrollment or attendance, school feeding programmes show significant and consistent effects across all three dimensions: enrollment, attendance, and learning, both in earlier studies and the most recent evaluations.
3. School feeding appears more beneficial for children in low-income countries than middle-income countries
Low-income countries surveyed in 2021 by the Global Child Nutrition Foundation all cite "education" as a key objective for their school meal programmes, compared to only 69 percent of high-income countries. Does this imply school feeding interventions are likely to have a higher impact on education outcomes in low-income countries? Our review shows that is indeed the case—at least as far as school participation is concerned. The results show that school meal programmes have a statistically significant effect on enrollment in low-income countries, compared to the effect in middle-income countries which are not statistically significant. This outcome is intuitive, considering that primary enrollment rates are lower on average in low-income countries than middle-income countries.
The average effect on attendance is statistically significant in both income groups. In addition, the average impact among low-income countries is over 25 percent higher than in middle-income countries. The estimates of the effect on learning outcomes are less precise for both groups when countries are divided by income level. But it should be noted that less than one-fourth of the studies in our review focused on low-income countries, highlighting the need to carry out more evaluations on programmes in those countries.
4. The effects are often bigger for disadvantaged groups within the same country
Ten of the 28 studies report impacts specifically for subgroups of the population and almost all of those documenting effects for the relevant subgroup report bigger impacts for girls, younger children, those from poorer families and those from more food-insecure locations. Five of the six studies that report gender-disaggregated outcomes report stronger impact for girls. For example, a randomised control trial (RCT) in Ghana showed double the impacts on test scores for girls and those living in disadvantaged communities. A World Food Programme school meal intervention implemented in an emergency setting in Mali led to higher enrollment and school completion on average, and reduced child labour particularly for girls. In contrast, child labour increased for boys receiving generalized food assistance (Aurino et al. 2019). Three studies analyzing impacts by household income report stronger positive effects on test scores and enrollment for children from low-income households (China's school feeding programme in two studies) or disadvantaged castes (in a study of India's mid-day meal scheme). The Senegal school feeding programme led to higher test scores especially for children younger than 10, while another evaluation of the India programme improved enrollment particularly at the first grade.
5. There is wide variation in the type of cost information reported in these evaluations and which costs are counted
The best source for cost data on school feeding programmes are of course the budget documents and other programme reports, but these documents are not always publicly available. These impact evaluations, on the other hand, are often more accessible to the general public. So how much does school feeding cost according to these studies? The estimates fall on a wide range from $5 per child per year in India to $110 per child per year in China (which comes in the form of a 4-yuan government subsidy per child per day). Part of the variation comes from what is counted in that cost. In some studies, only the purchase price of the ingredients is included while others include labour costs of cooking and serving (which is sometimes zero, if those activities are done by volunteers). Other interventions provide school meals alongside take-home rations, deworming initiatives, or as part of larger programmes that include funding for school kitchens and garden supplies, making it challenging to evenly allocate programme overhead costs across multiple components. One programme in Senegal only provided maize to schools and students were asked to contribute a certain amount to purchase other ingredients such as fresh vegetables, fish or meat, and other grains.
These issues extend to reporting cost-effectiveness. Only two of the studies report cost-effectiveness of providing school meals: a study in Senegal reports $2 per each additional score in math and French and another study in India reports $80 to $90 per one standard deviation increase in reading or math scores.
Closing the evidence gap: Cost-effectiveness and interaction with other interventions
Despite widespread public interest and significant investments in school meal programmes globally, there remains a critical need for more comprehensive evidence on their cost-effectiveness and how they interact with other education interventions. Ongoing efforts, such as the forthcoming systematic review by the Research Consortium for School Health and Nutrition, represent crucial steps in addressing these gaps.
Beyond evaluating the direct effects of school meals on educational outcomes, it is equally important to rigorously examine their indirect impacts, particularly their role in enhancing broader educational reforms. Implementing organizations and researchers must also prioritize the systematic collection and standardization of cost data to enable robust analysis.
Key policy debates focus on the comparative efficacy and cost-effectiveness of school meal interventions in achieving developmental goals. Providing definitive evidence in these areas will be pivotal for shaping strategies to promote human development and social protection in low- and middle-income countries.
Online appendix available here.
We gratefully acknowledge helpful feedback from Susannah Hares, Justin Sandefur, and Rory Todd.