Endline Report: Impact Evaluation of the Food for All Project in Kenya (F4APK)

Overview

This impact evaluation assessed F4APK, a five-year FDOV-funded public-private partnership (2015 to end-2020) managed by Solidaridad with Heifer International, Meru Greens, SoilCares Foundation and the Horticultural Crops Directorate. The project targeted 48,500 smallholder households in horticulture and dairy value chains in five Lower Eastern Kenya counties (Meru, Tharaka Nithi, Embu, Machakos, Makueni), with a total budget of EUR 5,352,986. It covered hardware and organisation investments, training and ICT. The evaluation looked at relevance, additionality, effectiveness, sustainability and CSR, focusing on farmer-level effects. It compared treatment and control producer organisations between 2019 and 2021, and the results include the effects of COVID-19.

Key findings

• Project implementation was successful: many output targets were reached or exceeded, and treatment farmers had better access to collection centres and higher training participation than control farmers.
• Dairy commercialisation was the clearest success: the share of milk sold to dairy cooperatives rose by 40 percent relative to the baseline share of 50 percent, and treatment farmers overtook control farmers.
• Dairy treatment farmers became more specialised: the share of production acreage used for dairy rose from 22 percent at baseline to 35 percent at endline, a significant difference-in-differences of +14 percentage points.
• Productivity did not rise significantly relative to control: dairy treatment farmers gained about three quarters of a litre per cow per day from a baseline of 8.5 litres, which was no different from control, and horticulture target crops showed no positive effects except French beans and bananas in the pre-COVID period.
• Horticulture engagement fell: the share of horticulture treatment farmers growing any F4APK crop dropped from 54 to 46 percent, and the share trained in the last 12 months fell from 59 to 14 percent.
• The share of French bean sales through collection centres in the horticulture treatment group fell from 62 to 44 percent. None of the newly installed collection centres were functional in 2020, which was linked to farmer distrust after non-payment by Meru Greens.
• Self-reported food access and life satisfaction followed similar trends in treatment and control groups (for example, 35 percent of horticulture treatment farmers reported food shortages at baseline and 33 percent at endline), so these changes appear unrelated to the project. Female and youth reach rose from 45 percent in 2017 to around 70 percent in 2019-20.

In figures

Recommendations

• RVO should focus projects on lower-level, more realistic outcomes in the results chain.
• Explicitly consider the strength of producer organisations (cooperatives) and commercial partners as a prerequisite and determinant of success during project preparation.
• In multi-partner private sector development projects, ask for functional redundancy so that key value chain functions are played by more than one partner.
• Recognise and plan for micro-level supply constraints such as water and irrigation costs, with more flexibility in strategy, activities and budget to address emerging constraints.
• Investigate the potential of low-cost phone-based information interventions, ideally studied separately so they can be evaluated.
• Make impact evaluation an integral part of the project cycle: implementation should not start without a complete evaluation design, including a credible counterfactual.
• Ask implementing partners to provide digitised monitoring microdata (for example, PO-level production, productivity and price sheets).

Lessons learned

• The F4APK theory of change was too ambitious and demanding for a five-year project spanning two very different value chains.
• The project was more effective in the dairy value chain, possibly because dairy producer organisations were more experienced and cohesive.
• Past sector growth rates do not guarantee 'picking winners': horticulture production fell despite the sector's expected 15-20 percent annual growth, so a good understanding of local supply constraints is needed.
• Activities such as input access and collective marketing that were largely organised through the project may require continued support after the project ends.

Methodology

Quasi-experimental difference-in-differences design comparing farmers in F4APK treatment producer organisations with farmers in control POs in Meru, Embu and Machakos counties. Baseline: 1,237 farmers in 226 POs (February-April 2019). Endline: 1,115 farmers re-interviewed (90 percent, February-March 2021). The analysis was supplemented by project documents, monitoring data and interviews with project management, implementers, extension staff and farmers.

Limitations

• The evaluation started late in the project cycle, so a fully pre-treatment baseline was not possible; the analysis focused on a later-trained sub-set of beneficiaries.
• F4APK was not randomly drawn from the FDOV portfolio, and randomised assignment was not possible, so a quasi-experimental design was used.
• Control farmers also received support from non-F4APK sources, so impact is measured against other programme support rather than no support.
• The COVID-19 pandemic affected both treatment and control farmers and is hard to disentangle from project effects; pre-COVID estimates rely on long recall periods.
• Numbers for some crops (for example, French beans) are small, and changes may partly reflect sample selection.