Effects of sustainable soil fertility management practices on maize income in rural Zambia: a double machine learning approach

We study the impacts of sustainable soil fertility management (SFM) practices – animal manure, minimum tillage, and improved fallows, on maize income per hectare in Zambia. Using data from 1,210 rural households, we use debiased/double machine learning (DML) to estimate average treatment effects (ATE) – the expected impact if all farmers adopted a practice – and effects on the treated (ATT), reflecting the actual benefit realised by current adopters. We also implement a DML-causal mediation analysis to understand how these effects occur, with maize yield as the primary mechanism. Results show that individual SFM practices lead to substantial income gains. For example, improved fallows – a short-term agroforestry practice where farmers plant fast-growing, nitrogen-fixing trees or shrubs on fallow land to restore soil fertility, could raise maize income per hectare by 37% if adopted universally. Minimum tillage shows a similar pattern (ATE: 21% higher, ATT: 15% higher), whereas animal manure offers moderate potential (ATE: 20% higher) but lower realised benefits, suggesting room for improved implementation. Farmers who adopted animal manure and minimum tillage achieved, on average, about 18% higher maize income per hectare. Moreover, adopting at least one of the three SFM practices is associated with 30% higher income on average, and 18% higher income among current adopters. Mediation results reveal that over 80% of these effects operate through yield improvements rather than through other pathways. These findings highlight the value of promoting flexible SFM options. Even partial SFM adoption can improve soil fertility, rural incomes, and strengthen resilience to environmental and economic stress.