A random walk for agricultural total factor productivity

Abstract

Growth in agricultural total factor productivity (TFP), which explains most of the long-term growth in U.S. agricultural output, may be slowing. The Economic Research Service (ERS) of the USDA is confident that current levels of below-average growth will eventually regain the long-term trend line. Others disagree, arguing instead that due to declining public expenditures on agricultural research, TFP growth experienced a downward and seemingly permanent structural shift about 30 years ago. In this paper, I argue that neither perspective is accurate since agricultural TFP is best modeled as a random walk with drift and thus not governed by a deterministic trend line. When I use a first difference model to accommodate the unit root, I do not find a structural break in the rate of drift. However, I acknowledge that this finding may not be general because I show that my test for a structural break has low power. To add theoretical relevance, I develop a simple model of stochastic innovation and farm technology adoption, and then use simulation results from my model to explain why a random walk for agricultural TFP is a theoretically sound proposition.