CFE 2019: Start Registration
View Submission - CMStatistics
B0391
Title: A semi-parametric approach for stochastic frontier models with exogenous variables Authors:  Michael Wiper - Universidad Carlos III de Madrid (Spain) [presenting]
Abstract: The stochastic frontier model assumes that a firm's output y can be modeled as $y = \beta X - u + v$ where $X$ is a vector of inputs, $\beta X$ is an efficiency frontier, $v$ is a random error and $u$ is a non-negative, inefficiency term which may depend on exogenous variables. Up to now, most research has assumed a parametric form (exponential, half normal, ...) for modeling the inefficiency distribution, and when exogenous variables are introduced they have typically been included as linear regressors in the location term of this distribution. We address this problem by introducing a semi-parametric, Bayesian model based on Dirichlet process mixtures. Our approach is illustrated with real data applications.