Nonparametric Estimation and Inference of Production Risk

Abstract

This paper proposes a nonparametric approach for estimation of stochastic production functions with categorical and continuous variables, and then develops procedures that allow for inference on production risk. The estimation is based on the kernel method and the inference is based on a bootstrapping approach. We establish the asymptotic properties of our proposed estimator. Monte Carlo simulation results suggest that our proposed nonparametric procedure is more robust and outperforms other existing parametric and nonparametric methods. In addition, we empirically illustrate the proposed nonparametric approach using long‐run corn production data from university field trials in Wisconsin that examines the performance of genetically modified corn varieties. Specifically, the proposed nonparametric procedure is used to empirically examine the production risk effects of categorical genetically modified variety variables and a continuous planting density variable.