A Skew Normal Stochastic Frontier Model for Panel Data

Abstract

A Stochastic Frontier defines a random upper bound of a response variable, called output,  as a function of variables called inputs plus an idiosyncratic random error. The observable output is given by the Stochastic Frontier minus a non negative  random variable called subject random inefficiency component.  In this paper a three random component Stochastic Frontier for Panel Data is introduced. Subject heterogeneity is modeled by a random component and   the presence of both a time invariant  and a time dependent random inefficiency component is allowed. In this way subject heterogeneity is not wrongly modeled as inefficiency  and it is possible to disentangle  a time persistent component from the total inefficiency. The role of the closed skew normal distribution in this kind of modeling is stressed.