Could the Debate Be Over? : Errors in Farmer-Reported Production and Their Implications for the Inverse Scale-Productivity Relationship in Uganda

Gourlay, Sydney ; Kilic, Talip ; Lobell, David

World Bank, Washington, DC

关键词: AGRICULTURAL PRODUCTIVITY;    MAIZE;    YIELD MEASUREMENT;    PLOT AREA MEASUREMENT;    INVERSE SCALE-PRODUCTIVITY RELATIONSHIP;   

DOI  :  10.1596/1813-9450-8192
RP-ID  :  WPS8192

学科分类：社会科学、人文和艺术（综合）

来源: World Bank Open Knowledge Repository

【 摘 要 】

Based on a two-round household panelsurvey conducted in Eastern Uganda, this study shows thatthe analysis of the inverse scale-productivity relationshipis highly sensitive to how plot-level maize production,hence yield (production divided by GPS-based plot area), ismeasured. Although farmer-reported production-basedplot-level maize yield regressions consistently lend supportto the inverse scale-productivity relationship, thecomparable regressions estimated with maize yields based onsub-plot crop cutting, full-plot crop cutting, and remotesensing point toward constant returns to scale, at the meanas well as throughout the distributions of objectivemeasures of maize yield. In deriving the much-debatedcoefficient for GPS-based plot area, the maize yieldregressions control for objective measures of soilfertility, maize genetic heterogeneity, and edge effects atthe plot level; a rich set of plot, household, and plotmanager attributes; as well as time-invariant household- andparcel-level unobserved heterogeneity in selectspecifications that exploit the panel nature of the data.The core finding is driven by persistent overestimation offarmer-reported maize production and yield vis-à-vis theircrop cutting–based counterparts, particularly in the lowerhalf of the plot area distribution. Although the resultscontribute to a larger, and renewed, body of literaturequestioning the inverse scale-productivity relationshipbased on omitted explanatory variables or alternativeformulations of the agricultural productivity measure, thepaper is among the first documenting how the inverserelationship could be a statistical artifact, driven byerrors in farmer-reported survey data on crop production.