Agricultural interventions in developing-region smallholder systems: A harmonized outcome–metric meta-analysis of productivity, livelihood, and welfare outcomes

Agricultural interventions are central to smallholder productivity, livelihoods, and food-security policy in developing regions, but their evidence base is fragmented across outcomes, effect metrics, designs, and contexts. This study provides a quality-controlled, harmonized outcome–metric meta-analysis of agricultural intervention evidence in developing-region smallholder systems. A cleaned impact-evaluation dataset was audited for effect estimates, sampling variances, standard errors, exclusions, and within-study dependence. The final analytical base comprised 26 retained studies and 135 model-ready effects. Because the evidence included mean differences, regression coefficients, marginal effects, and only two log response ratios, effects were not combined into a single universal estimate. Outcomes were grouped into harmonized families, while effect metrics were kept separate. Ten outcome–metric families met the minimum threshold for quantitative synthesis. The best-populated families were Economic_Livelihood_MD and Productivity_MD, which produced positive average family-level estimates of 1452.27 and 263.70, respectively. However, both were accompanied by very high heterogeneity, wide prediction intervals crossing zero, and sensitivity to dependence-oriented checks. Food-security-related evidence was present but measurement-fragmented and statistically inconclusive. The findings support cautious family-level evidence of productivity and livelihood gains, but not universal effectiveness, intervention ranking, unified percentage effects, or direct environmental sustainability claims.