Recent Advances in the Panel Stochastic Frontier Models: Heterogeneity, Endogeneity and Dependence

This paper primarily reviews the panel stochastic frontier (SF) model in which the composite error term ϵ_it has four components, i.e. ϵ_it = τi - ηi + vit - uit, where ηi and uit are the persistent and transient inefficiency components, τi is the random firm effect and vit is the random noise. We discuss a variety of model specifications of the four-component SF panel model and estimation approaches developed mostly during the period 2014 to 2021. While discussing these models, we focus on the following issues: heterogeneity, endogeneity of inputs/output(s), inclusion of determinants of inefficiencies via their variances and dependence among the error components. Additionally, we examined models that introduce allocative inefficiency. Models that use both multi-step and system approaches to model allocative inefficiency are also examined.