A1179
Title: Estimating impacts of large-scale agricultural intervention: The case of Ethiopian agricultural growth program
Authors: Eunsun Hong - Korea University (Korea, South) [presenting]
Jisang Yu - Korea University (Korea, South)
Abstract: This study evaluates the productivity impacts of Ethiopia's agricultural growth program (AGP I) by explicitly addressing measurement error, input endogeneity, and aggregation bias. Evaluating large-scale interventions in smallholder settings is often confounded by severe measurement error and heterogeneity in self-reported data. Using harmonized panel data from the Ethiopia LSMS-ISA, a two-stage structural empirical strategy is implemented. First, Cobb-Douglas production functions are estimated at both plot and household levels using instrumental variables (IV) to recover total factor productivity (TFP) purged of simultaneity and attenuation bias. Second, these IV-based TFP measures are employed in a difference-in-differences framework to identify the causal effect of AGP I. Results show that IV estimates diverge significantly from OLS, and that aggregating from plots to households sharply reduces unexplained productivity dispersion. Crucially, once productivity is measured at the household level using IV-based methods and appropriate aggregation, statistically significant evidence is found that AGP I increased TFP. These findings suggest that productivity detected in conventional analyses may reflect granular mismeasurement rather than the absence of genuine policy impacts. The necessity of integrating structural estimation techniques into impact evaluations to avoid misleading conclusions driven by noisy survey data is therefore underscored.
