Empirical Economics Social Sciences (miscellaneous)
Inference on environmental productivity growth with by-production stochastic frontiers
Abstract
This paper presents a semi-parametric by-production stochastic frontier model that enables likelihood-based inference on environmental productivity growth when production generates both desirable output and pollution. In contrast to nonparametric approaches, productivity measurement is embedded in a fully specified stochastic frontier system, allowing formal inference on productivity growth itself as well as its structural drivers. Separate technologies for desirable output and undesirable by-products allow technology-specific productivity growth and its components, namely technical change, efficiency change, and scale change effects, to be identified and evaluated for each sub-technology. To avoid restrictive functional-form assumptions, the frontiers and inefficiency heteroskedasticity are specified using penalized splines, while dependence between the two technologies is modeled via a copula. Monte Carlo evidence shows accurate recovery of the components underlying productivity growth, and an application to agricultural firms using methane emissions illustrates technology-specific productivity dynamics, their sensitivity to functional-form assumptions, and enables formal statistical inference on environmental productivity growth and its drivers.