International Journal of Information Technology & Decision Making

This retrospective by an early DEA researcher recalls the people, the setting, the formative philosophies, and early implementation experiences with Data Envelopment Analysis. It is a tribute to the founders of DEA, a historical chronicle, and a perspective on the nature and value of this important operations research methodology.

Data Envelopment Analysis (DEA) is a recently developed methodology that is widely used for estimating relative efficiency scores of Decision Making Units (DMUs) that use several inputs to produce several outputs. Model specification in DEA includes aspects such as the choice of inputs and outputs or the adoption of a returns to scale assumption. As pointed out by many authors, it is obvious that the specification of a model is the key to having reliable efficiency scores. In this paper, we are particularly concerned with the selection of variables in DEA models. To be specific, we investigate the performance of several statistical tests existing in the literature that can be used for the selection of variables. In particular, the behaviour of the well-known tests proposed by Banker² and the nonparametric tests recently developed by Pastor et al.¹³ is analyzed in relation to several factors such as sample size, model size, the specification of returns to scale and the type and level of inefficiency. We have drawn some conclusions that will be of help for practical uses, since the observed behaviour of the tests in the different scenarios determined by the specifications of the mentioned factors may provide some useful insight into the choice of an adequate statistical test in the particular context of a given DEA application.

In contrast to parametric approaches which try to optimize a single regression function through the observed data set, DEA optimizes on each individual observation with a objective of calculating a discrete piece-wise linear frontier. In the so-called preliminary DEA efficiency evaluation, whether or not a DMU is efficient is solely determined by the solution of the linear programming problem constructed for this particular DMU. In our proposed ranking method, we assess the efficiency performance of a particular DMU based not only on its efficiency score when the DMU is standing out in the objective function of the linear problem but also its efficiency ratios under the optimal solutions of other n-1 linear programming problems. This constitutes the overall efficiency performance of a DMU. A ranking procedure is developed to rank the n DMUs in terms of the overall efficiency performance.

It is well known in the Data Envelopment Analysis literature that proper variable selection is necessary for the reliable measurement of efficiency. Omitting production relevant variables and/or including irrelevant variables will lead to biased measurement. It is also known that the sample size needs to be large relative to the number of inputs and outputs to prevent classification of efficiency by default. In some empirical settings the number of potential relevant variables is large. Careful selection of an appropriate set of variables is necessary for reliable efficiency measurement. This paper looks at the issue of input selection and uses simulation analysis to develop statistical procedures to provide guidelines for input selection.

This paper proposes an aggregated ratio analysis model which can be utilized to evaluate relative efficiency of decision making units (DMUs). We show that our proposed ratio model is equivalent to the CCR DEA model. This equivalence property offers a great deal of opportunities for DEA to be interpreted and applied in different ways. Our model also offers an insight into the frontier analysis. Whether a DMU is on the frontier or efficient frontier can be informed by using our aggregated ratio analysis. Several results developed in the paper are coincident with that in the literature.

Data envelopment analysis (DEA) identifies an empirical efficient frontier of a set of peer decision making units (DMUs) with multiple inputs and outputs. The efficient frontier is characterized by the DMUs with an unity efficiency score. The performance of inefficient DMUs is characterized with respect to the identified efficient frontier. If the performance of inefficient DMUs deteriorates or improves (up to the frontier), the efficient DMUs still have an unity efficiency score. However, the performance of DMUs may be influenced by the context — e.g. a product may appear attractive against a background of less attractive alternatives and unattractive when compared to more attractive alternatives. With an application to Tokyo public libraries, the current paper presents and demonstrates a context-dependent DEA which measures the relative attractiveness of libraries on a specific performance level against libraries exhibiting poorer performance. The set of libraries are grouped into different levels of efficient frontiers. Each efficient frontier (on a specific performance level) is then used as evaluation context for the relative attractiveness. The performance of the efficient libraries changes as the inefficient libraries change their performance. The context-dependent DEA can also be used to differentiate the performance of efficient DMUs. The context-dependent DEA provides finer DEA results with respect to the performance of all DMUs.

Statistical properties of DEA methods for efficiency estimation are poorly understood and currently the best way forward must be to use bootstrap techniques. The article seeks to extend bootstrap methods to allow investigation of the properties of estimates of inefficiencies due to the slack in the use of resources as well as technical efficiency. In an empirical application, it is found that inefficiency due to slack is a small component of the overall inefficiency and that the DEA technical efficiency estimates have a small downward bias, with confidence intervals that are wide enough to suggest cautious interpretation.

The paper studies the result of Zhu²¹ and establishes a relationship between the efficiency in data envelopment analysis (DEA) and the pareto optimality under multiple objective linear programming (MOLP). It is shown that the DEA/preference structure models in Zhu²¹ can be derived by traditional MOLP techniques. Incorporation of tradeoffs or value judgments is a direct result of using MOLP techniques. New uses of DEA are developed and described. The approach is applied to a set of Chinese cities.

In this paper, we examine the potential gains in physical outputs or earnings on outputs from an optimal reallocation of inputs. When some decision-making units (DMUs) face higher input prices than other DMUs, the Farrell decomposition of cost efficiency can potentially indicate that a firm with lower overall costs of production is less efficient than a firm that uses fewer physical inputs, but has higher costs. We extend our gain functions accounting for cases where DMUs face different input prices. An empirical illustration of our method is provided using data on Japanese banks operating during 2000–2003.

IMF policies have been widely criticized in the aftermath of the Asian crisis. Key critics questioned the appropriateness and the sequencing of financial liberalization programs which, along with insufficient monitoring and inadequate prudential regulations, left the financial sectors of the affected countries highly leveraged and exposed. This paper examines the impacts of similar reforms on the efficiency of the banking system in Tunisia, a country whose economy has been reshaped by the IMF/World Bank prescribed economic adjustment plans since 1987. Using various DEA models and panel data covering the period 1992–1997, we evaluate the individual effects of each component of the reforms on the banking industry overall.

Meanwhile, we compare the effects on banks because of the different ownership structures over time. We also pay particular attention to specific factors that have kept the financial sector in Tunisia relatively stable in the midst of the global market turmoil caused by the Asian crisis.

We employed data envelopment analysis (DEA) to examine the technical and scale efficiency of the sawmill industry of Washington State. We found that there is regional variation in the rate of technical efficiency and that for most years the industry in aggregate operated at a point of modest scale inefficiency. In addition, we examined the industry's rate of productivity growth and technical change between the early 1970s and late 1990s using the Malmquist input-oriented productivity index. We found that the industry experienced a modest average annual decline in productivity and technical change during the 1970s, but experienced strong productivity growth and technical change during the 1980s and 1990s.

For some investors their own personal investment counsellors address their investment strategy; for others automated means are used. To protect investors, the Canadian Government has enacted the "Know Your Client" Act requiring that all investment dealers and vendors of securities must know their clients and advise them on the appropriate investment strategy. This paper uses Data Envelopment Analysis (DEA) in a novel manner by applying it to a large data set of answers to a number of psychological questions. A Slacks Based Model was used to estimate investor risk tolerance. The model analyses the risk profile of the investor and can be used as a guide to match the risk rating of the investment vehicles for the client. Statistical comparisons were also carried out to show how risk tolerance relates to various demographic variables. Finally, the DEA results were validated through comparisons with the commercial system already in use.