Title: Consistent estimation for true fixed effects stochastic frontier models
Authors: Luca Grassetti - University of Udine (Italy) [presenting]
Ruggero Bellio - University of Udine (Italy)
Abstract: Inference is considered for true fixed effects stochastic frontier models, which are employed in panel data settings to separate time-invariant heterogeneity from efficiency. Such models have a noteworthy theoretical appeal, yet the estimation of their structural parameters is hindered by the incidental parameter problem, which may be severe for settings with a large number of short panels. We propose an estimation approach that exploits the equivariance property of maximum likelihood estimation to obtain a simple marginal maximum likelihood estimator of the structural parameters. This results in root-$n$ consistent estimation, and it can be applied to a wide array of distributional specifications without the need for simulation techniques. Another important feature of the proposal is its computational simplicity, and we illustrate the use of the TMB R package for automatic differentiation to obtain a scalable and efficient implementation.