Complementary Components? Experimental Evidence on Bundling Health and Education Interventions

How to cite

Maggio, Daniel; Angrist, Noam; Jeong, Dahyeon; Uckat, Hannah; Lombardini, Simone; Lubozha, Theresa. 2026. "Complementary Components? Experimental Evidence on Bundling Health and Education Interventions." AEA RCT Registry. AEARCTR-0019337. https://doi.org/10.1257/rct.19337-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Daniel Maggio* , Noam Angrist , Dahyeon Jeong , Hannah Uckat , Simone Lombardini , Theresa Lubozha

* Primary investigator

Registry ID
AEARCTR-0019337
DOI
https://doi.org/10.1257/rct.19337-1.0
First registered
Last updated
Status
On going
Trial dates
2025-11-01 – 2026-12-31
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/19337

Abstract

Health and education, in theory, could have large complementarities, with improvements in one yielding productivity gains in the other. Yet, experimental evidence on these complementarities remains limited. We conduct a factorial experiment, randomizing school meals access -- one of the most popular health interventions among governments -- alongside an effective educational approach, targeting instruction. We randomize 125 primary schools in Zambia, a context in which both of these interventions are scaling nationwide. Our factorial design enables us to identify a) the effect of school meals, b) encouragement to implement targeted instruction, and c) effects of school meals combined with targeted instruction. We identify effects and complementarities on several health and education outcomes: children's attendance and learning outcomes as well as their health and food security. Complementarities among these interventions could be consequential. If a child is receiving an education, but hungry, they might not be able to learn; vice-versa, if a child is well-fed, but the quality of education is low, the returns to school meals could be marginal. We analyze cost-effectiveness and welfare effects of each intervention alone, and when combined.

Intervention

The Home Grown School Meals Program

Our trial is embedded in the scale-up of Zambia's HGSM program in Zambia's Luapula Province. The program, already present in each of Zambia's ten provinces, provides participating schools with maize flour from the National Food Reserve Agency, together with funding to purchase perishable items such as beans on the spot market. The intent of the program is for schools to provide each child one mid-day meal each school day that consists of at least a staple grain and a legume. The program is being scaled nationally with support from the World Food Programme and the World Bank, which creates the opportunity to randomize the scale-up given logistical constraints that prevent immediate saturation. Within Mwense, Chembe, and Samfya districts, we study 63 schools that were early receivers of the program, compared to 62 schools that were randomly chosen to be enrolled in the program beginning in January of 2027.

The Incentive Intervention

Our incentive intervention is designed to provide teachers with a small amount of additional compensation for every TaRL session they hold. The intention is both to improve teachers' motivation and overcome any financial barriers associated with the additional scope of work that TaRL places on teachers. To do so, our intervention asks TaRL teachers to send in photographs of their learner-level register for each class that they send. In exchange, they receive 20 ZMW (approx. 1 USD) worth of mobile airtime, which serves as reimbursement for the airtime used to send the photographs, plus a small reimbursement. Teachers are eligible to receive up to three incentives per week, and payments are made bi-weekly. As such, the maximum payment a teacher can receive in a month is 260 ZMW (Approx. 13 USD).

Teacher Interface. To track and administer the intervention, our study partners with the service provider engageSpark to create an automated chatbot for teachers to engage with and submit photos to. The application is administered through WhatsApp and is primarily teacher-prompted but sends reminders to teachers who have not submitted any photos within a week. One challenge of administering a WhatsApp chatbot in a setting such as Luapula is that many teachers will not have smartphones that engage with the chatbot. To address this challenge, given that we can verify that at least one teacher in every school has a smartphone, our intervention will assign teachers unique identifiers that the chatbot will prompt them for before they can submit their photos. Keeping such a backend roster will allow teachers to submit photos through a colleague's phone, but receive the airtime incentive to their own.

Verification. The key challenge for an intervention that aims to incentivize teachers to hold class in a cost-effective manner is verifying that classes have indeed been held without visiting schools. Our intervention has landed on the learning-level register because it is a uniform document across government schools that tracks students' attendance on a daily basis. However, an intervention that provides incentives based on photographs of a form naturally creates perverse incentives for fraud through falsified records or photos that are unrelated to the level register entirely. Additionally, we expect to collect over 9,000 photos over the course of the intervention, and dedicating staff time to verify every register would reduce the cost-effectiveness of the intervention. As such, rather than verifying all photos, we will randomly select a 20 percent subset of photos to be manually verified by a staff member at our research partner, VVOB. This initial check will make sure that photos are 1) of the learning level register and 2) current and relevant to the class in question. In addition to these high-level checks, 25 percent of the subset (so 5 percent of the full sample) will be additionally verified via phone calls to the head teacher at the school. For any submission that is found to be fraudulent or otherwise unsatisfactory through this verification process, the teacher will be informed that their submission has not been accepted and will be given a chance to resubmit. Those who fail to resubmit a satisfactory level-register photo will not receive the incentive for that submission.

Experimental design

Our study sample includes 125 government primary schools in three districts within Luapula Province slated for the scale up of the HGSM program: Mwense, Chembe, and Samfya. Within each school, we randomly sample ten grade three children to participate in our baseline survey between November and December of 2025. Following our baseline survey, schools were randomly assigned to either be enrolled in the HGSM program immediately or to be enrolled in January 2027 (our control). This randomization was stratified by district and whether schools were located in an urban or rural area. Following the initial randomization, our study undertook a second randomization to assign schools to recieve the teacher incentive program described above. This second randomization is crossed with our initial randomization, and schools are chosen within strata based on their district, urban or rural location, and their HGSM group assignment. In total, this results in four, roughly equal-sized intervention arms:

1. Pure Control - Schools that will be enrolled in the HGSM program in January 2027 and will not receive the incentive intervention.
2. HGSM-Only - Schools enrolled in the HGSM program in early 2026 who will not receive the incentive intervention.
3. Incentive-only - Schools that will be enrolled in the HGSM program in January 2027 and will receive the incentive intervention.
4. HGSM + Incentives - Schools enrolled in the HGSM program in early 2026 who will receive the incentive intervention.

Design details

Randomization method
Randomization done in office by a computer
Randomization unit
Primary schools
Clusters
125 schools
Observations
1,249 children
Arms
31 Schools in pure control, 31 schools in incentive only, 32 schools in HGSM only, and 31 schools in HGSM + incentives
Primary outcomes
TaRL Attendance, school attendance, TaRL Sessions held (At the school level), food security
Secondary outcomes
TaRL Sessions held (At the school level), School Attendance, Children's nutritional status (HAZ, WHZ, BMI) and food security

Cited by

No citation data available for this registration yet.

Keywords