Maize Productivity and Input Subsidies in Malawi: A State-Contingent Stochastic Production Frontier Approach

We make cross-sectional comparisons of productivity in a risky agricultural setting. To make meaningful comparisons, we find it necessary to define a new productivity index that satisfies important axioms from index number theory (e.g., transitivity). The index can be computed without any information on output or input prices. However, it cannot be computed without an estimate of a state-contingent production frontier. We use maximum likelihood methods to estimate a state-contingent stochastic production frontier that explicitly allows for variations in input quality. We find that differences in productivity are mainly due to differences in environment and scale-mix efficiency. In turn, we conjecture that differences in scale-mix efficiency are partly driven by variations in access to input subsidies. The maximum likelihood estimator appears to do a poor job of disentangling the effects of technical inefficiency and statistical noise.