Adoption and dis-adoption of push-pull technology in East Africa: Evidence from survival analysis and machine learning

Publication Type

Journal Article

Journal Name

Crop Protection

Publication Date

4-1-2026

Abstract

This study examines the timing to adoption and dis-adoption, as well as the key drivers influencing these transitions, for Push-Pull Technology (PPT) in East Africa using household-level panel data from extensive surveys conducted in Ethiopia, Kenya, Rwanda, Uganda, and Tanzania. Using discrete-time proportional hazard models, we estimate the hazard functions for PPT adoption and dis-adoption and complement this with machine learning–based survival analysis through Extreme Gradient Boosting (XGBoost) to evaluate and enhance predictive performance. Both approaches consistently identify peer networks measured by the number of known PPT adopters, frequent extension contact, and farmer group membership as the most influential drivers of adoption and long-term retention. Farmers' positive perceptions of PPT's pest-control effectiveness (against stemborer, Striga, and fall armyworm) significantly accelerate uptake and reduce exit risk. Access to diverse information sources also speeds adoption, while participation in PPT training programs and repeated training sessions are essential for sustaining use over time. These findings suggest that complementary to ongoing efforts to improve input access, enhancing farmers' perceptions through targeted education, visible adopter clusters, and evidence-based demonstrations, together with embedded extension services, repeated interactive training, and country-tailored strategies, offer a roadmap for accelerating uptake and securing the sustained use of PPT across East Africa.

Keywords

Agricultural innovation, Discrete-time proportional hazard model, Duration analysis, Extreme gradient boosting, Social interaction, Technology adoption

Link to Full Text

Share

COinS