Do Better Surveys Change Policy Conclusions? Evidence from Burkina Faso’s 50x2030 Initiative?
Conference
10th International Conference on Agricultural Statistics
Format: CPS Paper - ICAS 2026
Keywords: hypothesis test, instrumental variable, public-policy, survey-methodology
Abstract
Improving agricultural survey methodology is a central objective of the 50x2030 Initiative, yet little evidence exists on whether better data collection changes the economic estimates that inform policy. We exploit a methodological transition in Burkina Faso’s Enquête Permanente Agricole (EPA)—comparing 534,000 plot observations from the traditional survey (2008–2021) with 51,000 from the redesigned 50x2030 instrument (2023)—to test whether survey methodology shapes estimated productivity elasticities. Ordinary least squares (OLS) production functions yield dramatically different policy signals across periods: the mineral fertilizer coefficient drops from 0.068 (p 2,000). The divergence in OLS estimates is consistent with changes in the underlying bias structure rather than shifts in agronomic productivity. These findings demonstrate that improving survey data is necessary but not sufficient for better policy: without upgrading analytical methods, richer data can actually worsen OLS-based policy recommendations.