Working Paper: NBER ID: w26924
Authors: Casey C. Maue; Marshall Burke; Kyle J. Emerick
Abstract: A vast firm productivity literature finds that otherwise similar firms differ widely in their productivity and that these differences persist through time, with important implications for the broader macroeconomy. These stylized facts derive largely from studies of manufacturing firms in wealthy countries, and thus have unknown relevance for the world's most common firm type, the smallholder farm. We use detailed micro data from over 12,000 smallholder farms and nearly 100,000 agricultural plots across four countries in Africa to study the size, source, and persistence of productivity dispersion among smallholder farmers. Applying standard regression-based approaches to measuring productivity residuals, we find much larger dispersion but less persistence than benchmark estimates from manufacturing. We then show, using a novel framework that combines physical output measurement, estimates from satellites, and machine learning, that about half of this discrepancy can be accounted for by measurement error in output. After correcting for measurement error, productivity differences across firms and over time in our smallholder agricultural setting closely match benchmark estimates for non-agricultural firms. These results question some common implications of observed dispersion, such as the importance of misallocation of factors of production.
Keywords: Productivity; Smallholder Farms; Measurement Error; Agricultural Economics
JEL Codes: O12; Q12
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| measurement error (C20) | productivity dispersion (O49) |
| smallholder farms (Q12) | productivity dispersion (O49) |
| smallholder farms (Q12) | productivity persistence (O49) |
| measurement error (C20) | understanding of productivity dynamics (O49) |
| correcting for measurement error (C20) | productivity differences (O49) |