Experimental Design
The trial is part of phase two of a school feeding programme that was being implemented by the World Food Programme in four districts in southern Malawi. The current phase covers four education districts in northern Malawi: Rumphi, Mzimba South, Mzimba North and Nkhata bay. The eligibility of a primary school for inclusion in the programme is determined by the implementing partner (i.e. WFP) based on the feasibility of establishing a functional school meals programme in the school. The minimum criteria applied for eligibility are listed below:
- Exclude schools providing school meals to children in the 2023/24 academic year
- Have a functional school management committee (SMC)
- Have a functional parent teachers’ association (PTA)
- Have a functional procurement committee
- Have a functional finance committee
- Have a functional water source
- Have a tree grown woodlot
- Distance to nearest farmer organisation (FO) is less than 20km
It was decided by government and the funding organization to cover 200 primary schools and 78 ECD centers in the current phase of the school feeding programme. The study includes all 200 primary schools and 78 ECD centers to maximize statistical power. The four districts were assigned specific numbers of schools as part of the programming decision. Accordingly, the distribution of treatment schools is:
- Mzimba North: 50
- Mzimba South: 65
- Rumphi: 50
- Nkhata bay: 35
The 78 treatment ECD centers are distributed between the four districts in proportion to the number of schools. Therefore, the distribution of treatment ECD centers is:
- Mzimba North: 19
- Mzimba South: 26
- Rumphi: 19
- Nkhata bay: 14
Operationally, each treatment ECD center need to be paired with a treatment primary school because the procurement of food items for ECD centers will be handled by primary schools. Therefore, the randomization processes proceeded in two stages. In the first stage, we randomly selected 200 treatment and 200 control primary schools, out of a total of 526 eligible schools, as per the predetermined allocation of schools within each district stratum. Treatment is randomly assigned at a school level using a matched quadruplet randomisation strategy to ensure balance. Schools are matched into quadruplets within districts based on:
- Pupil-to-Qualified Teacher Ratio (PQTR)
- Attendance rate (number of students present / number of students enrolled) on the day of implementing partner’s (WFP) screening visit
- 2019/20 Primary School Leaving Certificate Exam (PSLCE) Pass Rate
- Presence of a study school within a 3km radius to account for spillovers
Using Mahalanobis distance matching, we group four similar schools into a quadruplet. Within each quadruplet, two schools were randomly assigned to treatment and two to control to ensure within-district balance. Thus an equal number of treatment and control schools were selected per district.
In the second stage, all ECD centers within a 2km radius from a treatment or control primary school were assessed for eligibility based on a series of indicators to determine their suitability for functional school meal programme. Subsequently, 78 treatment schools were randomly selected from the list of treatment schools with at least one eligible ECD center within a 2km radius. The associated ECD centers constitute the treatment group for ECD centers. If there is more than one eligible ECD center within a 2km radius from a selected school, the center with the highest eligibility score is included. Similarly, 78 control schools were randomly selected from the list of control schools with at least one eligible ECD center within a 2km radius. The associated ECD centers form the control group for ECD centers.
All children attending treatment ECDs and primary schools receive school meals while no student in the control schools participates in any formal school feeding scheme. For each school, the list of communities/villages where students most likely reside as reported by the school head teachers is collected. Typically, schools receive students from 2-6 surrounding villages. The village that supplied the highest proportion of students to the school is designated as the primary village. If that village had at least 200 households, no other village is required for sampling. If the primary village had less than 200 households, the second largest village that was a similar distance away from the school as the primary village was identified and included in the sampling. More villages were selected using a similar selection process if required to reach a pool of approximately 300 households collectively for each school.
Households are selected using random-walk protocol in each sampled village. The field team begins by identifying a central location as a starting point. Each enumerator is assigned a cluster of about 50 households from which to sample 3-5 households. Based on the number of households in the clusters, the supervisor defines a common sampling interval for every enumerator to follow in each village. If the household is eligible (i.e., has at least one child enrolled in the target primary school between grades 1 and 7), the enumerator proceeds to list the members of the household using the provided tablet. In the case where there is more than one eligible child per household, the tablet randomly selects the child for learning assessment (one per household). Where possible, a replacement child is also identified, in case the child originally sampled was not available at the time of the survey. The target total sample to be selected was 12 households with eligible children enrolled at the target primary school.
For ECDs, the enumerators identified a cluster of households that contained the ECD within the village. The enumerators then sampled 5 households within that cluster with an eligible ECD-age student (i.e. between 36 and 71 months of age), that may or may not be enrolled at the target ECD. An additional 5 children were randomly selected from the ECD center’s roster, and then In addition, 5 children were randomly selected from the roster at the ECD center and then followed to their homes. The target for each ECD was 10 households with eligible aged children.