The generalized panel data stochastic frontier model: A review and nonparametric estimation

Recently, the four component generalized stochastic frontier model has become increasingly common in practical applications. However, it remains tethered to potentially restrictive distributional assumptions on all four random components. In this paper, we show that when certain exogenous variables uniquely influence technology, time-varying inefficiency, or persistent inefficiency, all components of the model can be identified nonparametrically. In essence we require separability between the frontier, the conditional mean of time-varying inefficiency, and the conditional mean of persistent inefficiency. Given that our identification hinges on differencing, we recommend using splines or sieves to estimate each of the components of the model. We provide a short application to demonstrate the workings of the method.