Types and Natures of Parametric Methods of Efficiency Analysis The parametric technique can be further divided into three dis-
tinct approaches, which all require a particular functional form to
be specified for the cost or profit frontier. They are the stochastic
frontier approach (SFA), the Thick Frontier Approach (TFA), and
the Distribution Free Approach (DFA) [6,14]. In accordance with
parametric approaches, the efficiency frontier is constructed on the
basis of econometric modelling, usually in the form of Cobb-Doug-
las (log-linear) production function [11]. Therefore, the production
function is defined by the set of explanatory variables (inputs, out
-
puts and other possible explanatory variables) and the two com
-
ponents of this regression´s composite error term (the random
error) and the inefficiency term. The stochastic frontier approach
treats deviations from production function as comprising both ran-
dom error (noise) and inefficiency [10,14]. Therefore, SFA assumes
two-sided distribution (usually normal with zero mean) of the er
-
ror term and one-sided distribution of the non-negative inefficien
-
cy term [14].
DFA, used in panel data, relaxes composite error term of dis-
tributional assumptions. The core inefficiency is distinguished
from random error by the assumption of core inefficiency being
persistent over time, while random errors tend to average out over
time (Equation 4). TFA also does not impose distributional restric
-
tions on the composite error term but assumes that inefficiency
term is different in the highest (thick frontier) and lowest efficien
-
cy quartile of the observed decision making units and the random
error is present within these quartiles [14]. All these approaches
commonly suffer from potential specification errors, because the
specified cost or profit function is at best an approximation to the
true (but unknown) counterpart [6,14].