A large body of researches have widely examined the impact of adopting improved agricultural practices and technologies on general welfare of smallholder farmers. The results of deep literature review show that varies agricultural technologies have significant impacts on different welfare measures identified in the primary studies. However, the estimated effects of technology adoption differ among studies. The current study presents a meta-analysis of empirical estimates using a sample of 52 studies that investigated the impact of improved agricultural technologies in Africa on three key sets of outcome variables: output or expenditure, food security, and poverty. The study also conducted tests for publication bias to see if researchers tend to report results in similar or different ways for the same outcome variable. The findings the study shed light on the ways of identifying potential factors explaining the differences in the effects of estimated technology adoption. Results of the meta-regression analysis revealed that differences in the reported impact of technologies is explained by factors like data type, model specification, sample size, region of the study, and journal type. It was also observed that no publication bias in the studies reviewed for the effect size measures of output (expenditure) and poverty models, but in the food security model there is some evidence of publication bias. One of the core implications of the current study is that, based on the sensitivity of effect sizes to study attributes (i.e. data type, econometric methods, sample size, region of the study, and journal type), interested researchers and academicians need to pay attention to these attributes to provide more reliable estimates for policy interventions. We believe this study provides information useful to interested decision-makers in designing policy intervention measures that could encourage the adoption of improved agricultural practices and technologies in the African context. Finally, the study also highlighted future research directions.