Erkie Asmare 1 and Andualem Begashaw

Summary and Conclusion
In summary, under parametric methods of efficiency analysis 
there are three approaches, namely; Stochastic Frontier Approach 
(SFA), Distribution-Free Approach (DFA), and thick frontier ap
-
proach, (TFA). On the other hand, nonparametric methods of effi
-
ciency analysis have two main approaches namely; Data Envelop-
ment Analysis (DEA), Free Disposal Hull (FDH) methods. Therefore, 
these two approaches have been used to obtain estimates of farm 
efficiencies. Both parametric and nonparametric methods of effi
-
ciency analysis have their own advantage and disadvantage. For 
instance, nonparametric techniques make no accommodation for 
noise. Parametric method on the other hand, requires specification 
of the technology, which may be restrictive in most cases. DEA does 
not model noise, but is able to impose axiomatic properties and 
estimate the frontier non-parametrically, while SFA cannot impose 
axiomatic properties, but has the benefit of modeling inefficiency 
and noise [1]. 
Based on the robustness of different techniques in ranking pro-
ductive units, DEA can improve the accuracy of parametric tech-
niques. The flexibility of DEA permits the introduction of relevant 
issues such as non-discretionary variables, categorical variables, or 
constrained multipliers [4]. In terms of consistency, the parametric 
approaches were found to be more stable than the non-parametric 
ones. This is perhaps because the DEA tends to confound random 
disturbances with inefficiency due to its non-stochastic nature. On 
the other hand, this nonparametric methodology provides guidance 
on how the inefficient production units could become efficient, us
-
ing the concept of reference group of efficient decision-making 
units that produce a similar output.
Since both parametric and non-parametric techniques have 
their own merits and demerits, the selection of a suitable estima-
tion method has been quite controversial and still unclear. There-
fore, a parametric and nonparametric method of efficiency analysis 
is not direct competitors but rather complements: in the trade off 
between DEA and SFA something is sacrificed for something to bar
-
gain. Hence, joint use of parametric and non-parametric measure-
ment techniques of efficiency is a novel issue in the recent empirical 
literature. 
Recommendations
The problems of DEA models can be solved either by collecting 
or by measuring accurately all relevant variables. Another and 
more feasible alternative for including statistical noise in the model 
is to use a parametric approach for the estimation of the production 
function. In order to overcome the limitation of the construction of 
confidence intervals, using a bootstrap DEA method allows us to 
validate the results and enables us to obtain confidence intervals 
and adjusted efficiency scores.

---