SADIFU (Shamba Darasi Imara na Ustadi): Farmer Field Business Schools for Strengthening Smallholder Farmers' Resilience

The SADIFU project evaluates the impact of Farmer Field Business Schools (FFBS), implemented alone or in combination with a gender-transformative Social Analysis and Action (SAA) approach, among small-scale farming communities in rural Tanzania. The study focuses on smallholder farmers who are members of existing Village Savings and Loan Associations (VSLAs), and therefore operate under relatively less binding liquidity constraints compared to non-members. The evaluation is designed to compare the effects of FFBS alone with those of FFBS combined with SAA, thereby identifying the additional contribution of the gender-transformative component to agricultural, economic, and intra-household outcomes.

The intervention has two active treatment arms.

In the first arm (FFBS), farmers participate in group-based Farmer Field Business Schools built on existing Savings and Loan Associations (SLAs). The objective of FFBS is to enhance farmers’ knowledge, improve adoption of CRA practices, increase agricultural productivity and market participation, and ultimately strengthen household income, food security, and resilience against weather shocks.
FFBS is a learning-by-doing model that provides training in climate-resilient agricultural (CRA) practices (e.g. minimum tillage, crop rotation, intercropping), post-harvest handling, nutrition, financial literacy, and farm business management. The program also strengthens access to savings and credit within SLAs, promotes improved record-keeping, and facilitates linkages to input suppliers, buyers, and market actors. The FFBS intervention operates through several mechanisms. Training and experiential learning increase farmers’ knowledge and skills, which are expected to translate into greater adoption of CRA practices. Improved access to savings and credit relaxes liquidity constraints, and, when combined with increased knowledge and skills, enables farmers to invest in agricultural inputs and technologies. Market linkages and business training improve commercialization and price realization. Together, these channels are expected to increase crop productivity, farm income, and food security.

The second treatment arm (FFBS+SAA) includes all FFBS components and adds a gender-transformative Social Analysis and Action (SAA) approach. The addition of SAA is expected to influence outcomes through changes in intra-household dynamics and social norms.
SAA consists of structured community dialogues, household-level engagement, and village-level activities aimed at challenging restrictive gender norms and strengthening women’s agency in agricultural and household decision-making. The approach promotes reflection on gender roles, joint planning within households, and community-level norm change, supported by trained facilitators and gender champions. By promoting critical reflection and dialogue on gender roles, SAA aims to increase women’s decision-making power, control over resources, and confidence. These changes are expected to improve the allocation of labor and inputs, enhance cooperation within households, and facilitate the adoption of agricultural and economic practices promoted through FFBS. By addressing gender-based constraints that limit women’s access to resources, information, and decision-making, the SAA component is expected to amplify the impacts of FFBS on agricultural productivity, income generation, and household welfare.

The target population consists of small-scale farming households in villages with active Savings and Loan Associations in a region highly exposed to weather shocks, like droughts. The intervention particularly targets female SLA members and female-headed households, while also engaging men and broader community members to promote inclusive and sustained change. The intervention is implemented through community-based facilitators, including paraprofessionals and community trainers, each responsible for supporting multiple FFBS groups within their villages.

The interventions rely on several key assumptions. First, farmers are willing to participate in group-based learning and adopt new practices. Second, VSLAs function effectively and provide reliable access to finance. Third, market actors engage with farmers and provide viable market opportunities. Fourth, both men and women participate in SAA activities and are receptive to changes in gender norms. These assumptions are critical for the intervention to achieve the intended outcomes and will be explored empirically through complementary qualitative methods and targeted survey modules.

Overall, the intervention aims to improve smallholder farmers' resilience to weather shocks, agricultural performance, food security, and women's empowerment in vulnerable rural communities.

2024-04-01

2027-01-31

Following the causal pathway of the theory of change, the primary outcome measures of this trial are:

Agricultural services and productive resources
1. Use of agricultural advisory services: Binary indicator equal to one if the household received any agricultural advisory service (e.g. weather, marketing, or agricultural practices information) during the past 12 months, and zero otherwise.
2. Productive use of financial resources: Binary indicator equal to one if the household used a loan or savings from the savings and loan association (SLA) for productive purposes (i.e., farm inputs, livestock, non-farm business, labor, farmworkers, or protection against climate-related shocks) in the past year, and zero otherwise.
3. Value of financial resources used for productive purposes: Total amount (USD) of loans and savings from the SLA used for productive purposes in the past year.

Climate-resilient agricultural practices
4. Adoption of climate-resilient agricultural practices: Binary indicator equal to one if the household adopted minimum tillage, intercropping, or crop rotation during the last agricultural season, and zero otherwise.

Agricultural production
5. Any planted area lost to a weather shock: Binary indicator equal to one if the household reported any planted area lost to a weather shock during the last agricultural season, and zero otherwise.
6. Area of planted land lost to a weather shock: Total area of planted land lost to a weather shock during the last agricultural season (hectares).
7. Crop yield: Yield (kg/acre) of maize and sunflower during the last agricultural season, measured separately for each crop.
8. Value of crop production: Total value of crop production (USD) during the last agricultural season.

Gender norms and empowerment
9. Gender norms and attitudes index: Measures support for gender-equitable norms by recording the proportion of gender equality-friendly statements with which respondents agree, reported separately for women and men.
10. Attitudes towards intimate partner violence index: Measures the proportion of scenarios in which respondents consider a husband justified in hitting or beating his wife, reported separately for women and men.
11. Women's empowerment: Women's empowerment score ranging from 0 to 1, measured using the Five Domains of Empowerment of the abbreviated Women's Empowerment in Agriculture Index (A-WEAI), comprising input into productive decisions, ownership and decision-making over assets, control over income, group membership, and workload.

Household food security and nutrition
12. Food security: Binary indicator equal to one if the household is classified as food secure based on the Household Food Insecurity Access Scale (HFIAS) over the past four weeks, and zero otherwise.
13. Food insecurity: Binary indicator equal to one if the household is classified as food insecure based on the Household Food Insecurity Access Scale (HFIAS) over the past four weeks, and zero otherwise.
14. Household dietary diversity: Household Dietary Diversity Score (HDDS), measured as the number of food groups consumed by the household during the previous day.

These outcomes are measured at baseline and endline using structured household surveys administered to both male and female respondents within sampled households.

This study employs a stratified cluster randomized controlled trial (RCT) to evaluate the impacts of Farmer Field Business Schools (FFBS), implemented either alone or in combination with a gender-transformative Social Analysis and Action (SAA) approach, on small-scale farming households in rural Tanzania. Villages are the unit of randomization because the interventions are delivered at the community level.

The study is conducted in the districts of Iringa, Kilolo, and Wanging'ombe. A total of 139 villages met the program eligibility criteria, including the presence of an active savings group (VICOBA or VSLA) meeting minimum membership requirements, sufficient participation by women and crop farmers, and no prior exposure to FFBS or gender-transformative interventions. From these, 120 villages were selected and stratified into six strata defined by the combination of district (Iringa, Kilolo, and Wanging'ombe) and agroecological zone (hot/mild and cold/very cold) before being randomly assigned to one of three study arms: (1) FFBS only (40 villages); (2) FFBS combined with SAA (40 villages); and (3) a control group receiving no project intervention during the evaluation period (40 villages).

Within each study village, fourteen households were randomly selected from savings group membership lists, yielding a baseline sample of 1,680 households (560 per study arm). Baseline data were collected in March 2024, and endline data will be collected from the same households between October and November 2026. The study is designed to estimate both the impact of FFBS relative to the control group and the additional impact of integrating SAA with FFBS.

Not available

Randomization was conducted in-office using a computer-generated procedure in Stata at the village level. Villages were randomly assigned using the randtreat command under a stratified randomization design.

Prior to randomization, villages were stratified into six strata defined by the combination of district (Iringa, Kilolo, and Wanging'ombe) and agroecological zone (hot/mild and cold/very cold). Within each stratum, villages were randomly assigned to one of three study arms: (1) FFBS only, (2) FFBS combined with SAA (FFBS+SAA), or (3) a control group. The target allocation was equal across arms, with 40 villages assigned to each arm overall.

A randomly generated seed was used for treatment assignment to ensure reproducibility and eliminate researcher discretion in the randomization process. Because the number of villages in some strata was not divisible by three, the randomization algorithm allocated the remaining villages to preserve the planned overall balance across treatment arms while maintaining stratification.

Randomization was completed before baseline data collection to allow timely programme implementation. The randomization code and final treatment assignment are archived and will be made publicly available with the replication materials.

Randomization takes place at the village level. Villages are the unit of treatment assignment because the intervention (FFBS and FFBS+SAA) is delivered at the community level through village-based Savings and Loan Associations (SLAs) and collective training activities.

Eligible villages were randomly assigned to one of three study arms: (1) FFBS only, (2) FFBS combined with SAA (FFBS+SAA), or (3) a control group. Households within villages were not individually randomized.

Yes

The planned number of clusters is 120 villages, which constitute the unit of randomization. The villages are located across the districts of Iringa, Kilolo, and Wanging'ombe and were stratified into six district-by-agroecological-zone strata prior to random assignment. Villages were allocated equally across the three study arms, with 40 villages assigned to each arm: (1) FFBS only, (2) FFBS combined with SAA (FFBS+SAA), and (3) a control group

Within each village, 14 households are sampled for survey data collection, resulting in a total planned sample of 1,680 households. The number of participants is approximately 38,400 direct beneficiaries.

The trial includes 120 villages, which are allocated equally across three study arms (40 villages per arm):

Control: 40 villages receiving no FFBS or SAA intervention during the project period.
FFBS: 40 villages receiving the Farmer Field Business Schools (FFBS) intervention.
FFBS+SAA: 40 villages receiving FFBS combined with the Social Analysis and Action (SAA) gender-transformative approach.

Within each village, 14 households are randomly sampled for survey data collection, yielding a planned sample of 1,680 households (560 households per study arm). For modules on gender norms and women's empowerment, both one adult woman and one adult man are interviewed in each sampled household, resulting in up to 3,360 individual respondents for those outcomes.

We conduct power calculations using baseline data and the planned cluster-randomized design. The calculations assume 40 villages per study arm, 14 households per village, a two-sided significance level of 5%, and 80% statistical power. We base the calculations on pairwise comparisons between two study arms and estimate minimum detectable effects (MDEs) separately for each outcome using its baseline mean, standard deviation, and intracluster correlation coefficient (ICC). We also account for expected precision gains from stratified randomization and the planned ANCOVA specification. To capture the gains from stratification, we estimate the share of between-village variation explained by the randomization strata, which ranges from 3% to 43% across outcomes. For the ANCOVA adjustment, we consider baseline-endline correlations of 0.4 and 0.7, representing conservative and optimistic scenarios, respectively. The accompanying table reports the baseline mean, standard deviation, ICC, and MDE for each outcome. We express MDEs in the original outcome units, as a percentage of the baseline mean, and in standard deviation units. Under the unadjusted design, MDEs range from 0.22 to 0.39 standard deviations. Accounting for stratification and ANCOVA improves precision, reducing this range to 0.19-0.35 standard deviations under the conservative scenario and 0.15-0.27 under the optimistic scenario. For most outcomes, the adjusted MDEs lie between 0.15 and 0.25 standard deviations. For example, the MDE for maize yield decreases from 23% of the baseline mean under the unadjusted design to 19.8% under the conservative scenario and 15% under the optimistic scenario. Overall, these results suggest that the study is sufficiently powered to detect moderate and policy-relevant changes in the main outcomes included in the program's theory of change, although smaller effects may remain below the minimum detectable threshold.