Abstract
Cover crop adoption rates in the US Midwest have remained low despite their potential to improve soil health, enhance soil carbon sequestration, and increase long-term crop productivity. Existing studies using farmer surveys provide cross-sectional evidence of the farm and farmer characteristics influencing adoption but do not explain adoption dynamics over time and across fields for the same farmer. We combine novel satellite-detected grid-level cover crop planting data with secondary data on field, farm, and farmer attributes from 2011-2021 to track the dynamics of annual field-level adoption decisions for a random sample of farmers in the US Midwest. We quantify the effects of adoption by neighboring farms on explaining the spatial clustering pattern of cover crop adoption at the farm, field, and share of land area scale using linear panel models with an instrument variable approach. We find that a one percentage point increase in neighboring-field adoption intensity increases farm-level annual adoption probability by 28%, field-level annual adoption probability by 7%, and the share of farmer’s land area planted with cover crops by 3 percentage points based on our IV estimates. We further explore the impacts of “learn from self” vs. “learn from neighbors” using a dynamic panel data model. We find that farmers’ previous decisions do not significantly influence their current decisions on the share of land planted with cover crops. Using estimates of neighborhood effects, we show the timing and extent of the spread of this practice based on autonomous contagion from neighbors in the US Midwest. Our findings suggest a potential for education and extension efforts to build on this neighborhood effect to accelerate learning and induce further adoption of cover crops.