Estimating the direct and indirect effects of improved seed adoption on yields: Evidence from DNA-fingerprinting, crop cuts, and self-reporting in Ethiopia

Farmers' adoption of improved crop varieties could increase yields in low-income countries. However, the presence of measurement error in household surveys poses a challenge to estimating true returns. Using the 2018/19 Ethiopia Socio-economic Survey, we analyze the impacts of how three sources of measurement error: misperceptions of seed varieties, land area, and quantities harvested affect maize yields and input use. These data include DNA-fingerprinting of seed, GPS plot size information, and crop cuts that we compare to farmers’ self-reported estimates of these measures. Results indicate that the measurement error in self-reported seed variety adoption, especially from farmers who did not know they were using improved maize varieties, attenuates their estimated yield gains by 25 percentage points on average. The enhanced genetics of improved seed varieties accounts for a 41-percentage point yield increase over non-improved varieties, and increased input use accounts for a 30-percentage point gain for improved varieties on average. • We estimate the impact of adoption of improved seed varieties on maize yields. • These estimates are affected by measurement error on both sides of regression. • Direct impact of adoption is captured by the genetic effect of improved varieties. • Indirect impact of adoption is observed through the use of complementary inputs. • Measurement error obfuscates the benefits of adopting improved maize varieties.