دانلود رایگان ترجمه مقاله بررسی كاربردهای تحليل پوششی داده ها (نشریه الزویر 2013)

بررسی كاربردهای تحليل پوششی داده ها

A survey of DEA applications

Abstract

The literature of data envelopment analysis (DEA) encompasses many surveys, yet all either emphasize methodologies or do not make a distinction between methodological and application papers. This study is the first literature survey that focuses on DEA applications, covering DEA papers published in journals indexed by the Web of Science database from 1978 through August 2010. The results show that on the whole around two-thirds (63.6%) of DEA papers embed empirical data, while the remaining one-third are purely-methodological. Purely-methodological articles dominated the first 20 years of DEA development, but the accumulated number of application-embedded papers caught up to purely-methodological papers in 1999. Among the multifaceted applications, the top-five industries addressed are: banking, health care, agriculture and farm, transportation, and education. The applications that have the highest growth momentum recently are energy and environment as well as finance. In addition to the basic statistics, we uncover the development trajectory in each application area through the main path analysis. An observation from these works suggests that the two-step contextual analysis and network DEA are the recent trends across applications and that the two-step contextual analysis is the prevailing approach.

1 Introduction

The value of data envelopment analysis (DEA) lies in its capability to relatively evaluate the individual efficiency or performance of a decision making unit (DMU) within a target group of interest that operates in a certain application domain such as the banking industry, health care industry, agriculture industry, transportation industry, etc. All these industries practically adopt DEA for a variety of reasons, as Golany and Roll [85] pointed out that it can be applied to: identify sources of inefficiency, rank the DMUs, evaluate management, evaluate the effectiveness of programs or policies, create a quantitative basis for reallocating resources, etc. Some 30 years after the publication of the influential paper by Charnes et al. [1], the application domain for DEA has grown to such an extent that almost no one in the DEA research community is able to keep track of its development and in particular on how widely DEA is applied to real world applications.

Most previous general literature surveys for DEA place their emphasis on the methodologies, as the following examples show. Seiford and Thrall [2] reviewed early-stage DEA development. Seiford [3] traced the evolution of DEA for the period 1978 through 1995. Cooper et al. [4] evaluated some DEA models and measures. Cook and Seiford [5] performed a comprehensive survey on 30 years of DEA developments since 1978. Liu et al. [6] conducted a citation-based survey and depicted the main DEA development paths. All these surveys have elaborated on methodological topics such as generic DEA models, network models, multiplier restrictions, considerations on the status of variables, data variation, etc.

As we are aware of, the literature offers no survey in regards to the development of DEA applications. The closest comments in the literature on how DEA is applied are: ‘‘In total, 67%1 of the (DEA) articles presented a real-world application’’ [7] and ‘‘Banking, education (including higher education), health care, and hospital efficiency were found to be the most popular application areas’’ [8]. These comments provide some information, yet a more extensive survey is needed in order to benefit DEA researchers and practitioners. After all, the main purpose for developing the DEA method is to apply it.

In the development of any discipline, assessing what has been done can provide practical information in setting up strategies ahead of the next stage for various types of researchers. For example, basic statistics such as the number of application papers suggest the overall usefulness of the developed methodology. Theoreticians may need to find ways to improve their methods upon seeing that they are not frequently applied. Other statistics such as the major application area can inspire theoreticians to develop methods to specifically meet the needs of these areas. In addition, information on how each individual model was applied in applications indicates the trend in methodology adoption and thus helps users of the methods to catch up with the latest technology. A newcomer to a discipline would certainly be eager to know the set of ‘‘must read’’ papers to determine his/her research direction in that discipline.

The purpose of this study is to provide the above statistics and information in the DEA application area through a rigorous analysis. We pursue the answers to the following questions: What is the proportion of application papers in the DEA literature? Exactly how widely is DEA applied to real world applications? What are the major DEA applications? What is the trend of the methodological approach for each application area? Lastly, what are the development trajectories for each application area? Through these research questions, this study contributes to the DEA literature in three major aspects. First, it differentiates between methodological articles and application articles and provides basic statistics on application articles, in contrast with previous similar studies [7,8] that present statistics on the whole set of DEA literature. Second, it provides information on the mostcited methodological works in each application area. Third, through the main path analysis, it identifies papers that stand out in the important historical development path of each major application area.

In order to answer these questions, we classify a set of DEA papers into methodological and non-methodological works and then further segment the empirical works based on the real world problem that is discussed, tested, and validated within each work. After the classification, the basic statistics are summarized. We then present a list of the most-cited methodological papers for each of the five most popular application areas. In the end, the citation-based main path analysis as described in Liu et al. [6] is applied to these five most popular application areas to uncover their development trajectories.

This paper is organized as follows. In the next section we describe the dataset and the method of analysis. Section 3 discusses the basic statistics for the DEA applications. Section 4 presents the most-cited methodological works for the five major DEA applications. Section 5 introduces the main paths of the five most popular application areas. The last section draws conclusions, including implications and insights from the analysis results.