Crop yields fail to rise in smallholder farming systems in sub-Saharan Africa

Abstract

Drawing on a harmonized longitudinal dataset covering more than 55,000 smallholder farms in six African countries, we analyze changes in crop productivity from 2008 to 2019. Because smallholder farmers represent a significant fraction of the world's poorest people, agricultural productivity in this context matters for poverty reduction and for the broader achievement of the UN Sustainable Development Goals. Our analysis measures productivity trends for nationally representative samples of smallholder crop farmers, using detailed data on agricultural inputs and outputs which we integrate with detailed data on local weather and environmental conditions. In spite of government commitments and international efforts to strengthen African agriculture, we find no evidence that smallholder crop productivity improved over this 12-y period. Our preferred statistical specification of total factor productivity (TFP) suggests an overall decline in productivity of -3.5% per year. Various other models we test also find declining productivity in the overall sample, and none of them finds productivity growth. However, the different countries in our sample experienced varying trends, with some instances of growth in some regions. The results suggest that major challenges remain for agricultural development in sub-Saharan Africa. They complement previous analyses that relied primarily on aggregate national statistics to measure agricultural productivity, rather than detailed microdata.

Keywords: Africa; agricultural productivity; agricultural productivity growth; agriculture; smallholder agriculture.

Fig. 1.

Estimated coefficients of productivity change across different regression models. This figure plots coefficients and 95% CIs of productivity change estimates from various regression models. Model 1 is a simple regression of yield on a linear time trend and country dummies. Model 2 is a plot-level model, controlling for inputs, weather, country dummies, and other control variables. Model 3 is analogous to model 2 but using data aggregated at the household level. Model 4 is a household fixed effects model. Model 5 is a plot-manager fixed effects model. Model 6 is a cluster fixed effects model. Model 7 is analogous to model 2 but using current, instead of constant, prices. See SI Appendix, Table S3 for point estimates, sample size, and a full list of variables.