Misspecification preferred: The sensitivity of inefficiency rankings

Ruggiero compared the two popular parametric frontier methods for cross-sectional data – the stochastic frontier and the corrected OLS – in a simulation study. He demonstrated that the inefficiency ranking accuracy of the established stochastic frontier is uniformly inferior to that of the misspecified Corrected OLS (COLS) (which lacks an error term). The reason for his result remains unclear, however. In this paper, a more extensive simulation study is therefore conducted to find out whether the superiority of COLS is simply due to small sample sizes or to poor performance of the inefficiency level estimator.