Evaluating large-scale government investments in fertilizer adoption: The Ethiopian experience

Abstract

We evaluate the impact of a large Government of Ethiopia intervention to raise fertilizer supply by establishing five fertilizer blending facilities supplying fertilizers tailored to local soil nutrient profiles. We rely on the phased geographic rollout of blending facility establishment to identify the causal effect on fertilizer use, application rates, crop yields, gross crop revenue, and household consumption. Combining effects of multiple treatment periods, each estimated using a doubly robust difference-in-difference model, we find that the blending facilities increased the probability that farmers adopt the new blended fertilizers by 22 percentage points and increased application rates by 17 kg/ha (baseline adoption was zero). The facilities mostly induced farmers who previously used DAP to switch to NPS, and we find large decreases in DAP adoption (by 22 percentage points, 47% of the control group base mean) and application rates (16 kg/ha, 52% of the control group base mean) yet no impact on overall fertilizer adoption or application rates. Though the new blended fertilizers were expected to perform better, there is no evidence they improved crop yields, crop gross revenue, or household consumption. The effect of the intervention was more pronounced (with larger increases in NPS use and larger decreases in DAP use) for farms located near demonstration plots, which the Government used to train farmers about the agronomic response to the new fertilizers. We confirm results using three large-scale longitudinal datasets and show that they are robust to choices of specification, treatment definition, and inference assumptions.