Accounting for heavy tails in stochastic frontier models.

This paper aims at introducing a new class of stochastic frontier models that can take account for fat tails composed errore. Quite surprisingly, all the stochastic frontier models proposed in literature cannot handle situations where the empirical distribution of the composed error has heavy tails. These situations are instead very common in applications. In particular, we will propose to model the composed error with the skew-t distribution. This is equivalent to assume a Student-t distribution for the measurement error and a half-t distribution for the inefficiency. In this way, we extend quite naturally, the stochastic frontier model where a normal distribution is assumed for the symmetric error and a half-normal distribution is assumed for the inefficiency term.