Economic Impact Assessment of Agricultural Extension Programmes:a Review of Methods, Evidence, and Challenges

Abstract

Rigorous assessment of the economic impacts of agricultural extension programmes has been a cornerstone challenge for development economists and policy evaluators since the mid-twentieth century. Despite decades of accumulated research, significant methodological heterogeneity, evidence gaps, and contested findings continue to limit the confidence with which policymakers can make investment decisions.This paper provides a comprehensive critical review of the methods, evidence, and challenges in assessing the economic impacts of agricultural extension programmes. We synthesize the dominant evaluation paradigms, critically assess their strengths and limitations, document the range of impact estimates in the published literature, and identify the primary challenges confronting the field. A systematic review of 135 peer-reviewed studies, meta-analyses, and programme evaluations published between 1980 and 2024 was conducted, drawing on Web of Science, EconLit, Scopus, IFPRI, and World Bank publication databases. Studies were categorised by evaluation methodology, region, commodity system, and impact outcome metric. Benefit-cost ratio estimates are highly sensitive to evaluation methodology: observational studies report median BCRs of 3.2 versus 2.1 for RCTs, a 52% differential attributable primarily to positive selection bias. Seven identifiable sources of bias — including publication bias (+35%), selection bias (+42%), and scaling discount (−38%) — create systematic distortions in the evidence base. RCTs, while providing the most credible causal estimates, suffer from external validity limitations and scaling discounts of 30–50%. No single method dominates across all evaluation criteria; method choice should be guided by context, data availability, and the specific policy question. The field of extension impact assessment faces a dual challenge: improving the internal validity of causal estimates while simultaneously strengthening external validity and scalability relevance. Mixed-methods approaches combining experimental estimation of programme effects with structural models of behavioural mechanisms offer the most promising path forward. Standardised reporting protocols and pre-registration of evaluation designs are urgently needed to address publication bias.