Can unobserved land quality explain the inverse productivity relationship?

Abstract

An inverse relationship between both farm productivity and labor intensity, and farm size, is a common empirical finding in developing country agriculture. The traditional explanation has been imperfect labor markets. Recently, it has been suggested instead that the inverse relationship is a statistical artifact resulting from omitted land quality. Using a farm-level data set from Java, I investigate whether omitted variable bias can explain the inverse relationship. I show that the inverse relationship and accompanying wage responses are inconsistent with a model of neoclassical farm behavior that ignores omitted variable bias. Instrumental variables techniques yield parameter estimates in which the inverse relationship is eliminated and the estimated wage elasticities are more in line with economic theory. Further econometric investigation hints that a model of omitted land quality may be a possible source of the inverse relationship. These results emphasize the importance of considering the sources of cross-sectional variation in estimating microeconomic relationships.