The non-parametric approach DEA is a widely nonparametric and powerful data analytic tool,
which is commonly applied in the research and practitioner com-
munities to determine the relative efficiencies of the DMU [25].
Non-parametric methodology has been applied, using the concept
of a reference group of efficient decision-making units that produce
a similar output (peer group) [3]. According to Toma P et al. [3] the
most efficient DMUs are those situated on the frontier, the others
can reach this allocation if: they reduce the inputs, while maintain-
ing a constant output; they increase the outputs, while maintaining
the inputs constant; they perform a combination of the two previ-
ous solutions [3]. The distance between each DMU and its related
best point is a measure of the inefficiency, that is, how much it is
possible to expand the output given the input or how much it is
possible to reduce the input given the output.
In DEA, we use the concept of “reference set”, which is useful to
identify the best production unit with which to compare all the oth-
er observations. The DEA method is applied by adopting two dif-
ferent approaches that are both based on the concept of technical
efficiency, defined as the ability of the DMU (the decision-making
unit of production), given the existing technology, to produce the
highest level of outputs from a given combination of inputs (output
model- oriented), or alternatively, to use the least possible amount
of inputs to obtain a given output (model input-oriented). There
-
fore, this nonparametric methodology provides guidance on how
the inefficient production units could become efficient, using the
concept of reference group of efficient decision making units that
produce a similar output (peer group) [3].
The basic efficiency measure utilized in DEA is the ratio of out
-
put to input, but this measure is only applicable to cases of a sin-
gle input and output. This analysis requires two-dimensional data
where all DMUs are plotted on a two-axis graph and an efficient
frontier constructed. All units that lie on the efficient frontier are
defined as efficient units (i.e., they have a 100% efficiency score),
while all those that do not lie on the frontier are defined as ineffi
-
cient units. Their locations from the efficient frontier are then used
to calculate their efficiency scores. The frontier thus ‘‘envelops” the
whole data [18]. In the case of multiple input and/or multiple out
-