Machine learning and econometric analysis of the heterogeneous welfare effects of agricultural cooperative membership in Mpumalanga, South Africa

This study applies machine learning and econometric analysis to examine the heterogeneous welfare effects of agricultural cooperative membership among smallholder maize farmers in Mpumalanga, South Africa. Using two complementary approaches, we analyzed data collected from 210 farmers, comprising both cooperative members and non-members. Machine learning techniques, including the Extra Trees Regressor (ETR), were employed to identify key welfare determinants, followed by the Inverse Probability Weighted Regression Adjustment (IPWRA) to estimate treatment effects. Results indicate that cooperative membership significantly enhances farmers' welfare, with farm size, access to infrastructure, and years of experience as critical contributors. Cooperative members benefit from improved income and reduced input costs, highlighting the role of cooperatives in promoting rural development and poverty reduction. The findings advocate for increased policy support to strengthen cooperative frameworks and encourage broader participation among smallholder farmers.