"background": "Agricultural field research stations are critical for generating improved crop varieties and agronomic practices. However, methodological weaknesses in their experimental designs often undermine the validity and generalisability of reported yield gains, limiting their impact on smallholder farming systems. ", "purpose and objectives": "This review critically evaluates the methodological frameworks employed by field research stations in Tanzania. Its primary objective is to propose and detail a robust quasi-experimental framework specifically designed for causal yield assessment in these settings. ", "methodology": "We conducted a systematic review of experimental designs used in station-based agronomic research. The proposed framework centres on a generalised difference-in-differences model, Y{it = \0 + \1 (Treatmenti \ Postt) + \ + \ +, where \ and \ are station and time fixed effects. Inference relies on cluster-robust standard errors to account for spatial autocorrelation. ", "findings": "The analysis identifies a prevalent over-reliance on simple, unreplicated demonstration plots, which fails to control for confounding spatial and temporal heterogeneity. A synthesis of comparative studies suggests that yield improvements attributed to new technologies may be overestimated by 15-25% when using these conventional, less rigorous designs. The proposed quasi-experimental design directly addresses these sources of bias. ", "conclusion": "Current methodological practices at many stations are insufficient for establishing credible causal claims about yield impacts. Adopting a formal quasi-experimental framework is both feasible and necessary to enhance the scientific rigour and policy relevance of station-derived evidence. ", "recommendations": "Research stations should institutionalise the use of staggered treatment rollout and maintain consistent control plots to enable robust counterfactual analysis. Donors and policymakers should mandate and fund the capacity building required for this methodological shift. ", "key words": "quasi-experimental design, agricultural research, impact evaluation, yield gap, causal inference, fixed effects", "contribution statement
Mwakyembe et al. (Sat,) studied this question.