Persistent food insecurity and low agricultural productivity continue to undermine rural livelihoods in Sub-Saharan Africa, where smallholder farmers face increasing climatic and market risks. Crop diversification has been promoted as a potential strategy to enhance resilience and improve welfare, yet rigorous causal evidence remains limited. This paper examines whether crop diversification enhances land productivity and household food security in Malawi, using nationally representative data and Double/Debiased Machine Learning (DML) to address model specification bias and high-dimensional confounding. Results show that diversification significantly increases land productivity, food consumption, and household dietary diversity, with the strongest impacts observed for dietary quality and food consumption. Mechanism analysis highlights market participation, access to credit, and the use of organic fertilizers as critical pathways through which diversification improves outcomes. Heterogeneity analysis reveals that benefits are most significant among rural, small-scale, and male-headed households. Robustness checks, including propensity score matching and sensitivity analysis, confirm the validity of the findings. Overall, the study demonstrates that crop diversification is not merely a risk-coping strategy but a transformative pathway toward productivity growth and food system resilience. By integrating causal inference with modern machine learning, this study advances empirical understanding and provides policy-relevant insights for promoting sustainable and climate-resilient agriculture in Africa.
