Decomposing agricultural productivity growth using a random-parameters stochastic production frontier

This study makes two key contributions to the agricultural productivity literature. First, it demonstrates, using US agricultural state-level data, how a random-parameters stochastic frontier model can be used to account for environmental heterogeneity across decision-making units. Second, it uses the estimated parameters of the model to compute and decompose a productivity index that satisfies several key axioms from index theory. Because the decomposition explicitly accounts for both observed and unobserved environmental effects, we are able to obtain a more realistic and flexible assessment of productivity growth. We find substantial differences between productivity results generated using a model with random slope parameters and those generated using a more conventional model with constant slope parameters.