Menu Training and Vendor Linkages in school feeding programs to support Smallholder Farmers in Nigeria

The project aims to pilot linking school feeding programs with smallholder farmers who can supply food to school feeding programs. We aim to achieve this through two types of interventions: (i) training of school vendors to improve and tailor school menus; and (ii) creating linkages between school vendors and smallholder farmers. Thus, the project targets both vendors/cooks involved in school feeding programs and potential smallholder farmers who can be linked with school feeding programs. We plan to conduct this pilot study in Osun State, located in the South-West region of Nigeria. Osun is recognized as one of the more successful states in implementing the Nigerian Homegrown School Feeding Program (NHGSFP) and is currently the only state where the program is active. Through this initiative, we aim to generate evidence that can improve program efficiency by analyzing menu plans and procurement procedures. Using this evidence, we intend to collaborate with the Government of Nigeria (GoN) and the World Food Programme (WFP) to extend successful interventions to other states across the country. The primary reference groups for this study are the cooks/vendors and smallholder farmers involved in sales to cooks supported by the school feeding program. We target all cooks involved in school feeding services and smallholder farmers in the targeted communities who are able to supply the food to the cooks, including produce oil-palm, sweet potatoes, yam, cocoyam and vegetables.

2.1. Study objectives
This study aims to evaluate alternative mechanisms and interventions to support integration of local value chains into school feeding programs to benefit smallholder farmers in Nigeria. The impact evaluation includes a cluster randomized controlled trial aimed at evaluating the cost-efficiency and impact of adding training activities to an ongoing school feeding program to improve smallholder sales and revenues. The primary research questions we aim to address in this project include:
a) Can school feeding menu training improve the quality of school menus (including subjective evaluation of hygiene standards) and cooks’ welfare and empowerment outcomes (?
b) Can school feeding menu training and linking school feeding programs to smallholder farmers improve access and quality of school feeding menus
c) Can school feeding menu training and linking school feeding programs to smallholder farmers improve farmers’ sales and revenue?

2024-06-21

2024-07-26

a) The proportion of smallholder farmers selling their agricultural products to schools as well as to traditional markets.
b) Farmers' sales volume of locally grown agricultural outputs to both the market and schools (cooks).
c) Farmers' sales revenue from locally produced agricultural outputs to both the market and cooks.
d) Cost per unit of school meal to schools and cooks

a) The proportion of producers selling agricultural products to both traditional markets and directly to schools/cooks’ measures smallholder farmers’ participation and contribution to school feeding programs. This will be measured by eliciting participation in sales to nearby schools/cooks as well as more broadly to markets. We aim to measure this on a quarterly basis and hence by eliciting participation in the sale of agricultural produce.
b) Farmers' sales volume of locally grown agricultural outputs to both the market and cooks/vendors denotes the quantity of agricultural products sold within a specified period. This measurement aids in evaluating the demand and supply for locally sourced goods in both marketplaces and among cooks/vendors, offering farmers valuable information to adjust their production and marketing strategies accordingly.
c) Farmers' sales revenue from their outputs to both the market and cooks represents the total income generated from selling agricultural products within specified market segments. This financial metric reflects farmers’ intensive margin of participation in supplying fresh produce to schools and domestic markets.
d) Price per unit of agricultural produce sold to schools and markets. This measures access to fair and competitive prices to smallholders and cooks. Farmers and cooks can effectively reduce both the transaction costs as well as the time spent searching for products and markets. This approach enhances operational efficiency and profitability while ensuring the quality of their menus remains intact.

a) Menu quality and quantity assessment.
b) Awareness of food safety and accessibility to fresh produce.
c) Subjective evaluation of hygiene standards.
d) Empowerment and well-being women cooks.

The impact evaluation involves a cluster randomized controlled trial (c-RCT) design that will be implemented in about 300 schools supported by the school feeding program in Osun State, Nigeria. For this purpose, we closely work with school feeding program implementers during detailed planning stages. Schools and associated farmers will be randomly assigned to one of three treatment groups:
a) Standard of care (control) group: Schools with the standard school feeding for the duration of the trial.
b) Training (intervention) group (T1): the cooks in these schools will receive additional training focusing on menu planning, which will be implemented alongside the standard school feeding program.
c) Training + farmer linkages (training plus intervention) group (T2): the cooks in this group of schools receive similar training on school menu planning while farmers associated with these schools are linked with the cooks in neighbouring schools.

The random assignment into the three groups above will be conducted after the baseline line data collection. Smallholder farmers with some potential to serve and supply to nearby school feeding programs were identified in consultation with cooks, nutrition experts, and agriculture officers.

Cluster Randomization:
A c-RCT is a field experiment where clusters of schools are randomly assigned to intervention groups instead of individual cooks. Therefore, the unit of randomization (school) differs from the unit of analysis (cooks or farmers).

We randomize at the level of school using a complete listing from the school currently implementing school feeding program using a random number generator in Excel. This happens after the baseline data collection and after all inclusion and exclusion criteria are implemented.

School

Yes

100 schools

Planned number of observations: Cooks: 700 Farmers: 1500

T1 : 100 schools
T2 : 100 schools

We first identify the number of clusters (schools in our case) and farmers needed for detecting a minimum impact on the primary and secondary outcomes stated above. The number of farmers and clusters, as well as the inherent similarity of farmers within a cluster matter in sample size calculations for cluster-randomized studies. This discordance between the unit of randomization and the unit of analysis creates special methodological challenges at every stage of the trial. These challenges arise essentially because individuals in the same cluster tend to respond more similarly than individuals from different clusters, violating the assumption of statistical independence required for the application of standard statistical methods. The outcome measure is now characterized by two separate sources of variation, one within clusters and the other between clusters. Thus, it requires intra-cluster correlations (ICC) estimation. Thus, we are doing power calculations for cluster randomized trials, and therefore it is important to know what the likely intra-cluster correlation will be for the study. To determine our sample size, we conduct power calculations prior to collecting the data. The accepted approach to estimating intra-cluster correlations for sample size calculations is to use values calculated from existing comparable datasets. We focus on the primary outcome indicators identified above, particularly those related to smallholder farmers’ market participation rates. These outcomes are related to the extensive and intensive margin of participation in sales to schools and other markets. We are considering an indicator variable for sales to schools as well as total sales volume/value to cooks. We then used these data to estimate the sample size that would provide an 80 percent chance (the power of the test) of correctly rejecting the null hypothesis that no change occurred, with a 0.05 level of significance. Across these sales’ volume/value outcomes, the mean intra-cluster correlation is 0.08. We compiled evidence on slightly similar programs from the literature to identify a reasonable Minimum Detectable Effect (MDE). For example, Gelli et al document that a slightly similar program in Ghana generated a 7.5 percent increase. Based on this we consider a MDE of 7.5% in 2025. We use recent data coming from the Living Standards Measurement Survey (LSMS) to compute mean and standard deviations associated with the farm outcomes listed above. The share of farmers selling to markets appears to be 40% and the average sale associated the important agricultural products (oil-palm, sweet potatoes, yam, cocoyam and vegetables) relevant for schools feeding programs. Using these values, we find that we need at least 300 schools to identify the MDE identified above for the outcomes listed above. Power calculations based on available clusters in targeted districts and resource availability suggested 300 clusters (schools) per intervention arm and 1500 smallholder farmer households per cluster. The primary outcomes of the trial include the sales and revenues. For these outcomes, assuming an inter cluster correlation coefficient (ICC) of 0.08, a sample size of 300 school leads to a minimum detectable effect size (MDES) of 7.5%. Within each chosen school, we compile a list of all cooks. We either include all cooks from that school or randomly select 700 of cooks to participate in the study, ensuring representation from both control and treatment groups. This selection process extends to cooks from various communities for both baseline and end line surveys.