Land Measurement, Yield Bias, and the Distribution of Productivity Growth Among Mozambican Smallholders

Farmers in sub-Saharan Africa often overstate plot size, causing yields calculated from self-reported land area to be systematically understated. A widely used correction predicts GPS-measured area from self-reported area in a subsample of measured plots and applies the estimated relationship to all farms, yet its validity has received little empirical scrutiny. Using three waves of Mozambique’s national agricultural survey (2017, 2020, and 2023) and evidence from maize and cowpea, we evaluate the assumptions underlying this correction, assess whether GPS-measured plots provide an unbiased benchmark, and examine how corrected yields alter conclusions about productivity growth. Self-reported area is a consistently strong predictor of GPS-measured area across survey waves and crops, and corrected and uncorrected yield estimates are nearly identical, supporting the validity of the adjustment. However, the GPS-measured subsample in 2023 disproportionately represents smaller, more productive farms, indicating that GPS-based yields should not automatically be treated as an unbiased benchmark. The correction eliminates average yield bias of 15 to 22 percent and shows that maize productivity gains between 2020 and 2023 were concentrated among the smallest farms, with substantial differences across provinces that national averages conceal. These findings strengthen confidence in GPS-based yield adjustment while highlighting the importance of accounting for benchmark selection bias.