"background": "Field research stations are a cornerstone of agricultural innovation in sub-Saharan Africa, yet robust methodological frameworks for evaluating their causal impact on clinical outcomes, such as crop yield and farmer income, are underdeveloped. Existing evaluations often lack rigorous counterfactual analysis. ", "purpose and objectives": "This study aims to develop and apply a quasi-experimental difference-in-differences (DiD) model to quantify the causal effect of a national network of research stations on key agricultural outcomes in Senegal. ", "methodology": "We constructed a longitudinal dataset from household surveys and station deployment records. The causal impact was estimated using a two-way fixed effects DiD model: Y{it = \ + \ (Treatedi \ Postt) + \ + \ +, where Y₈ₓ is the outcome for community i in period t. Inference is based on cluster-robust standard errors at the community level. ", "findings": "Adoption of station-promoted maize varieties increased by 32 percentage points (95% CI: 24, 40) in treatment communities relative to controls. This was associated with a statistically significant average yield increase of 17% (p<0. 01). The effect on household income was positive but not statistically significant at conventional levels. ", "conclusion": "The research station system had a substantial, causal effect on technology adoption and productivity for staple crops. The DiD model provides a validated methodological framework for impact assessment in similar contexts. ", "recommendations": "Policymakers should allocate resources to strengthen station-community linkages to enhance adoption. Future impact evaluations should employ quasi-experimental designs with carefully constructed counterfactuals to isolate causal effects. ", "key words": "impact evaluation, agricultural extension, quasi-experimental design, technology adoption, sub-Saharan Africa", "contribution statement": "This paper provides a novel application of the DiD model to evaluate agricultural research systems, demonstrating its utility for generating
Diop et al. (Sat,) studied this question.