Search
Browse By Day
Browse By Time
Browse By Person
Browse By Policy Area
Browse By Session Type
Browse By Keyword
Browse Artificial Intelligence Presentations
Program Calendar
Sign In
Search Tips
Cash transfer programs are widely used to reduce poverty in low- and middle-income countries, yet their long-run impacts remain uneven across beneficiaries. This paper examines the extent and structure of heterogeneous treatment effects in Zambia’s Child Grant Programme (CGP), focusing on households not subsequently exposed to the Harmonized Social Cash Transfer (HSCT). The central question is whether identical transfers generate divergent long-run outcomes and, if so, what explains these differences. We use seven-year panel data from a randomized program rollout to analyze long-term welfare impacts. To capture heterogeneity, we combine causal machine-learning methods with a conceptual framework grounded in poverty-trap dynamics. Specifically, we estimate household-level conditional average treatment effects (CATEs) on per-capita consumption using causal forests and complement these estimates with distributional analysis, LOWESS smoothing, and K-means clustering. The results reveal substantial dispersion in treatment effects, ranging from negative impacts to gains exceeding 0.6 standard deviations across the distribution. Households with high returns (“high flyers”) are systematically characterized by stronger baseline conditions, including higher asset holdings, better market access, lower dependency ratios, and more favorable environmental factors such as rainfall. In contrast, low-return households tend to face structural constraints that limit productive investment. LOWESS estimates show a smooth, positive relationship between predicted treatment effects and realized post-treatment outcomes, indicating a continuous gradient in responses. Clustering analysis further identifies a relatively homogeneous group of low-return households alongside multiple high-return clusters with distinct economic profiles. The combined evidence from our empirical strategy is consistent with nonlinear accumulation dynamics, in which initial conditions shape households’ ability to convert transfers into sustained gains. These findings suggest that while cash transfers improve average welfare, their transformative potential is concentrated among a subset of households. The results underscore the importance of incorporating heterogeneity into program evaluation and highlight the need for complementary interventions and more targeted “cash-plus” approaches in structurally constrained environments.