Data Envelopment Analysis

Research Proposal

Data Envelopment Analysis (DEA) was introduced for the first time by Charnes, Cooper and Rhodes in 1979 and converted to one of the most popular methods of performance measurement in Management Science. For instance, more than 1800 articles in refereed journals about DEA have been published until August 2001. 1100 of these have appeared after 1995 that indicates an increase of more than %150(Gattoufi et al 2004).It has been extensively applied in educational, banking, retailing and health care systems (see reference list).

EDA measures the relative efficiency of comparable entities that is called Decision Making Unites (DMUs). These unites have the same activities to transform the relatively similar inputs to similar outputs. The purpose of DEA is to find efficient frontiers for a set of available DMUs. An assumed DMU is efficient if there is no any DMU that can produce more outputs with the consumption of a certain amount of inputs or produce a certain amount of outputs with less consumption of inputs. The former approach is called output-oriented and the latter is called input-oriented. DEA provides information about efficient and inefficient units, as well efficient scores and reference sets for inefficient units. The results of DEA analysis is used as assessment indexes for DMUs in practical applications .There is a reason for assessment of DMUs in practice .It can be for the allocation of existing or additional resources to units, need to make more profitable operations by improving the performance of inefficient units or giving rewards to the most efficient units. The results of analysis attain a framework for these decisions.

One of the underlying assumptions of DEA model is the lack of preference among inputs and outputs. Under these circumstances, a DMU with producing a remarkable amount of outputs from one kind can be considered efficient. Evens so, It has a poor performance on production of another kind of output. Hence, in the original model of DEA, efficiency scores are not necessarily valid indicators of performance. Although DEA calculations ignore the preferred values, efforts to corporate the preferred information in DEA has been resulted to development of new models. However, there is no sufficient research in this issue and integration of preferred values still can be an attractive research topic.

Research objectives

The idea of value judgments in DEA has been adapted from real applications. Empirical applications justify the incorporation of value judgments for the number of reasons:

1. To modify the inputs-outputs weights. Sometimes these weights trend to zero that is not feasible in a real application. In addition, the obtained