Erkie Asmare 1 and Andualem Begashaw

The Choice Between Parametric and Nonparametric 
Methods
Unfortunately, there is no commonly accepted methodology of 
efficiency analysis currently, but the field is divided between two 
competing approaches: Data Envelopment Analysis (DEA) and Sto
-
chastic Frontier Analysis (SFA) [1]. The choice of estimation meth
-
od has been an issue of debate, with some researchers preferring 
the parametric and others prefer the nonparametric approach [7]. 
In addition, parametric and nonparametric techniques are the two 
approaches that have been used to obtain estimates of efficiencies 
but the choice of which approaches to use is still unclear [19]. Since 
both parametric and non-parametric techniques have their own 
merits, the selection of a suitable estimation method has been quite 
controversial. Furthermore, the choice of methodology appears to 
influence the policy implications derived from the analyses [6].
In contrary, Toma P et al. [3] Reported that the two approaches 
of efficiency measures (parametric and non-parametric) methods 
achieve highly correlated results in most cases. In addition, stud
-
ies on efficiency measurements argue that a researcher can safely 
choose any of the methods since there are no significant differenc
-
es between the estimated results [8,19]. Generally, the parametric 
technique is likely to be more attractive than DEA in cases where 
the data suffer from serious measurement errors, random events 
and difficulty in identifying inputs and outputs. On the other way, 
DEA may be a better choice when random disturbances are less of 
an issue, and price information is not available [6].
Consistency between Parametric and Nonparametric 
Measures
A study by Wang W [6], reveal that the choice among the vari-
ous frontier methods has no important effect on the estimated EE 
scores. This evidence suggests that the two models tend to reach 
moderately consistent rankings with each other once the true 
structure of pooling data is omitted. As nonparametric measures of 
performance (efficiency) are widely used by managers and policy
-
makers, it is informative to correlate the frontier efficiency scores 
with some conventional performance measures commonly used in 
the financial industry [6]. 
The parametric methods SFA and DFA seem to perform rela-
tively consistently with conventional performance measures. In 
some cases, parametric approaches were found to be more stable 
than the nonparametric ones [6]. This is perhaps because the non-
parametric method tends to confound random disturbances with 
inefficiency due to its non-stochastic nature, and the SFA and DFA 
have to specify a particular functional form for estimation, which 
may be miss-specified [6].