Using stochastic frontier analysis to assess the performance of public service providers in the presence of demand uncertainty

Abstract Public service providers generally make input decisions before the demand for their services is known. To account for demand uncertainty, we divide the decision-making process into two distinct stages: a resource planning stage in which managers choose inputs to meet future demand, and a production stage in which the chosen inputs are used to maximize revenue and/or meet demand. We use stochastic frontier analysis (SFA) models to estimate how well managers perform in each stage. We apply the methodology to hospital and health service managers in the Australian state of Queensland. We use Bayesian methodology to impose inequality constraints on the parameters of our models. We compare Bayesian predictions of efficiency with results obtained using data envelopment analysis (DEA) models. We find that Bayesian predictions of variable cost efficiency exhibit less variation than estimates obtained using the DEA approach; this is likely due to the fact that DEA models do not allow for noise. Our Bayesian predictions of technical efficiency are much lower than those obtained using the DEA approach; this is likely due to the fact that the DEA estimator we used is upwardly biased in finite samples.