CMStatistics 2018: Start Registration
View Submission - CFE
Title: Stochastic frontier model choice with unobserved heterogeneity: A Monte Carlo study Authors:  Antonio Carvalho - Heriot-Watt University (United Kingdom) [presenting]
Jan Ditzen - Heriot-Watt University (United Kingdom)
Abstract: Stochastic frontier models often assume the existence of a one-sided error term with an economic interpretation regarding technical or cost efficiency measurement. The true random and true fixed effects stochastic frontier models are popular solutions for dealing with unobserved heterogeneity in this context. Given the nature and assumptions of the models, it is natural to consider the Hausman test to make decisions on which model to use, as is done in random effects vs fixed effects models in panel data econometrics. However, in these specific models, an unbiased estimate of the efficiency term is potentially more important than an efficient estimate. We argue that using the Hausman test to make a decision on which model to use can potentially lead to an incorrect choice. A Monte Carlo simulation is set up to assess if the correlation between true and estimated efficiencies and the preservation of original efficiency rankings can be used as selection criteria, particularly if they are consistently favorable to the less restrictive fixed effects model, independently of the data generating process.