Targeting, bias, and expected impact of complex innovations on developing‐country agriculture: evidence from Malawi
研究马拉维一项参与式行动研究项目，发现新技术测试中系统性地偏向资源较好的农户，早期虽显示玉米产量和收获价值提升，但存在不可观测因素的选择偏差，提醒关注项目设计和农户分类以改进靶向。
Abstract Agronomic analyses of new technologies are often conducted under carefully controlled research station programs or trials managed by self‐selected farmers. Oftentimes, the technologies are then scaled up with minimal evaluation under real‐world conditions. Yet, the interim step between agronomic trials and large‐scale promotion is crucial to generate evidence on the social and economic impact of technologies that is both internally valid and generalizable. The article focuses on a participatory action research program in Malawi designed to test and identify scalable technology options to intensify the smallholder sector and contribute to poverty reduction and food and nutrition security. We examine the socioeconomic characteristics of farmers testing technologies and find evidence of systematic targeting of better‐endowed farmers. After controlling for observable differences using matching and a doubly robust estimator, we find evidence of early positive effects on maize yield and harvest value, although placebo tests suggest possible selection on unobservables. We note that attention should be given to program design and household characterization to better define and improve targeting criteria, technology selection, and external validity.