Stochastic Frontier Analysis with Generalized Errors: inference, model comparison and averaging

03/16/2020 ∙ by Kamil Makieła, et al. ∙ 0

Our main contribution lies in formulation of a generalized, parametric model for stochastic frontier analysis (SFA) that nests virtually all forms used so far in the field and includes certain special cases that have not been considered so far. We use the general model framework for the purpose of formal testing and comparison of alternative specifications, which provides a way to deal with model uncertainty - the main issue of SFA that has not been resolved so far. SFA dates back to Aigner et al. (1977) and Meeusen and van den Broeck (1977), and relies upon the idea of compound error specification with at least two error terms, one representing the observation error and the other interpreted as some form of inefficiency. The models considered here are based on the generalized t distribution for the observation error and the generalized beta distribution of the second kind for the inefficiency-related term. Hence, it is possible to relax a number of various potentially restrictive assumptions embedded in models used so far. We also develop methods of Bayesian inference that are less restrictive (though more demanding in terms of computation time) compared to the ones used so far and demonstrate inference on the latent variables (i.e., object-specific inefficiency terms, which are important for, e.g., policy-oriented analyses).

READ FULL TEXT
POST COMMENT

Comments

There are no comments yet.

Authors

page 28

page 29

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.