Sensitivity analysis of efficient units in the presence of non-discretionary inputs

Discretionary models of data envelopment analysis (DEA) assume that all inputs and outputs can be varied at the discretion of management or other users. In any realistic situation, however, there may exist “exogenously fixed” or non-discretionary factors that are beyond the control of a DMU’s management, which also need to be considered. This paper discusses and reviews the use of super-efficiency approach in data envelopment analysis (DEA) sensitivity analyses when some inputs are exogenously fixed. Super-efficiency data envelopment analysis (DEA) model is obtained when a decision making unit (DMU) under evaluation is excluded from the reference set. In this paper by means of modified Banker and Morey’s (BM hereafter) model [R.D. Banker, R. Morey, Efficiency analysis for exogenously fixed inputs and outputs, Operations Research 34 (1986) 513–521], in which the test DMU is excluded from the reference set, we are able to determine what perturbations of discretionary data can be tolerated before frontier DMUs become nonfrontier.     

DEA cross-efficiency evaluation under variable returns to scale

Cross-efficiency evaluation in data envelopment analysis (DEA) has been developed under the assumption of constant returns to scale (CRS), and no valid attempts have been made to apply the cross-efficiency concept to the variable returns to scale (VRS) condition. This is due to the fact that negative VRS cross-efficiency arises for some decision-making units (DMUs). Since there exist many instances that require the use of the VRS DEA model, it is imperative to develop cross-efficiency measures under VRS. We show that negative VRS cross-efficiency is related to free production of outputs. We offer a geometric interpretation of the relationship between the CRS and VRS DEA models. We show that each DMU, via solving the VRS model, seeks an optimal bundle of weights with which its CRS-efficiency score, measured under a translated Cartesian coordinate system, is maximized. We propose that VRS cross-efficiency evaluation should be done via a series of CRS models under translated Cartesian coordinate systems. The current study offers a valid cross-efficiency approach under the assumption of VRS—one of the most common assumptions in performance evaluation done by DEA.     

关于要素不可控条件下的DEA模型的特性分析

在 DEA模型的实际应用中 ,往往要求某些要素的投入是不能改变的 ,即这些要素对于决策者是不可控的 .1 986年 ,Banker和 Morey为了克服此类问题 ,提出了具有固定不变投入的 DEA模型 ,被称为要素不可控条件下的 DEA模型 .详细研究了要素不可控条件下的 DEA模型 ,并且建立了要素不可控条件下的DEA(弱 )有效性与 (弱 ) Pareto解的等价关系     

On relations between DEA-risk models and stochastic dominance efficiency tests

In this paper, several concepts of portfolio efficiency testing are compared, based either on data envelopment analysis (DEA) or the second-order stochastic dominance (SSD) relation: constant return to scale DEA models, variable return to scale (VRS) DEA models, diversification-consistent DEA models, pairwise SSD efficiency tests, convex SSD efficiency tests and full SSD portfolio efficiency tests. Especially, the equivalence between VRS DEA model with binary weights and the SSD pairwise efficiency test is proved. DEA models equivalent to convex SSD efficiency tests and full SSD portfolio efficiency tests are also formulated. In the empirical application, the efficiency testing of 48 US representative industry portfolios using all considered DEA models and SSD tests is presented. The obtained efficiency sets are compared. A special attention is paid to the case of small number of the inputs and outputs. It is empirically shown that DEA models equivalent either to the convex SSD test or to the SSD portfolio efficiency test work well even with quite small number of inputs and outputs. However, the reduced VRS DEA model with binary weights is not able to identify all the pairwise SSD efficient portfolios.     

Separating managerial inefficiency and external conditions in data envelopment analysis

In this paper, we highlight the importance of appropriately dealing with non-controllable inputs in technical efficiency evaluations by using DEA. In order to do this, the two most important options that exclusively use DEA methodology for the incorporation of these variables – the one-stage model by Banker and Morey [Operations Research 34(4) (1986a) 513] and the three-stage method developed by Fried and Lovell [Searching the Zeds, Working paper presented at II Georgia Productivity Workshop, 1996] – are compared both methodologically and empirically. At the same time, we propose a modification to the latter model which allows us to improve its results and interpretation. The education sector has been selected for the empirical application, the reason being that it has the desirable feature that, in the productive process, the students' socio-economic and family status (a non-controllable input) has a direct influence on the school results.The results obtained show the superiority of the multi-stage approach. It is argued that the model developed by Banker and Morey does not deal appropriately with inefficient units, as producer's behaviour in this model does not reflect the objective situation faced by such DMUs.     

Measuring super-efficiency in DEA in the presence of infeasibility

Super-efficiency data envelopment analysis (DEA) model is obtained when a decision making unit (DMU) under evaluation is excluded from the reference set. Because of the possible infeasibility of super-efficiency DEA model, the use of super-efficiency DEA model has been restricted to the situations where constant returns to scale (CRS) are assumed. It is shown that one of the input-oriented and output-oriented super-efficiency DEA models must be feasible for a any efficient DMU under evaluation if the variable returns to scale (VRS) frontier consists of increasing, constant, and decreasing returns to scale DMUs. We use both input- and output-oriented super-efficiency models to fully characterize the super-efficiency. When super-efficiency is used as an efficiency stability measure, infeasibility means the highest super-efficiency (stability). If super-efficiency is interpreted as input saving or output surplus achieved by a specific efficient DMU, infeasibility does not necessary mean the highest super-efficiency.     

Super-efficiency infeasibility and zero data in DEA

Lee et al. (2011) and Chen and Liang (2011) develop a data envelopment analysis (DEA) model to address the infeasibility issue in super-efficiency models. In this paper, we point out that their model is feasible when input data are positive but can be infeasible when some of input is zero. Their model is modified so that the new super-efficiency DEA model is always feasible when data are non-negative. Note that zero data can make the super-efficiency model under constant returns to scale (CRS) infeasible. Our discussion is based upon variable returns to scale (VRS) and can be applied to CRS super-efficiency models.     

Combining the assumptions of variable and constant returns to scale in the efficiency evaluation of secondary schools

Our paper reports on the use of data envelopment analysis (DEA) for the assessment of performance of secondary schools in Malaysia during the implementation of the policy of teaching and learning mathematics and science subjects in the English language (PPSMI). The novelty of our application is that it makes use of the hybrid returns-to-scale (HRS) DEA model. This combines the assumption of constant returns to scale with respect to quantity inputs and outputs (teaching provision and students) and variable returns to scale (VRS) with respect to quality factors (attainment levels on entry and exit) and socio-economic status of student families. We argue that the HRS model is a better-informed model than the conventional VRS model in the described application. Because the HRS technology is larger than the VRS technology, the new model provides a tangibly better discrimination on efficiency than could be obtained by the VRS model. To assess the productivity change of secondary schools over the years surrounding the introduction of the PPSMI policy, we adapt the Malmquist productivity index and its decomposition to the case of HRS model.     

互联网金融背景下我国上市商业银行的效率实证研究

本文通过数据包络分析(DEA)中的CRS模型和VRS模型,对我国16家上市商业银行于2007至2014年在互联网金融背景下的效率进行了测度。根据16家商业银行的综合效率、纯技术效率、规模效率的变动情况,来分析互联网金融对商业银行造成影响和冲击。结果表明,商业银行的整体效率波动与互联网金融环境的变化基本一致,并且,商业银行的规模效率变化与综合技术效率变化保持一致,股份制商业银行的规模效率一直高于国有商业银行。     

A linear programming framework for free disposal hull technologies and cost functions: Primal and dual models

The free disposal hull (FDH) model, introduced by Deprins et al. [The Performance of Public Enterprises Concepts and Measurements, Elsevier, 1984], is based on a representation of the production technology given by observed production plans, imposing strong disposability of inputs and outputs but without the convexity assumption. In its traditional form, the FDH model assumes implicitly variable returns to scale (VRS) and the model was solved by a mixed integer linear program (MILP). The MILP structure is often used to compare the FDH model to data envelopment analysis (DEA) models although an equivalent FDH LP model exists (see Agrell and Tind [Journal of Productivity Analysis 16 (2) (2001) 129]). More recently, specific returns to scale (RTS) assumptions have been introduced in FDH models by Kerstens and Vanden Eeckaut [European Journal of Operational Research 113 (1999) 206], including non-increasing, non-decreasing, or constant returns to scale (NIRS, NDRS, and CRS, respectively). Podinovski [European Journal of Operational Research 152 (2004) 800] showed that the related technical efficiency measures can be computed by mixed integer linear programs. In this paper, the modeling proposed here goes one step further by introducing a complete LP framework to deal with all previous FDH models.     

Sensitivity analysis of DEA models for simultaneous changes in all the data

In data envelopment analysis (DEA) efficient decision making units (DMUs) are of primary importance as they define the efficient frontier. The current paper develops a new sensitivity analysis approach for the basic DEA models, such as, those proposed by Charnes, Cooper and Rhodes (CCR), Banker, Charnes and Cooper (BCC) and additive models, when variations in the data are simultaneously considered for all DMUs. By means of modified DEA models, in which the specific DMU under examination is excluded from the reference set, we are able to determine what perturbations of the data can be tolerated before efficient DMUs become inefficient. Our approach generalises the usual sensitivity analysis approach developed in which perturbations of the data are only applied to the test DMU while all the remaining DMUs remain fixed. In our framework data are allowed to vary simultaneously for all DMUs across different subsets of inputs and outputs. We study the relations of the infeasibility of modified DEA models employed and the robustness of DEA models. It is revealed that the infeasibility means stability. The empirical applications demonstrate that DEA efficiency classifications are robust with respect to possible data errors, particularly in the convex DEA case.     

Bridging the gap between the constant and variable returns-to-scale models: selective proportionality in data envelopment analysis

In data envelopment analysis (DEA), the use of constant returns-to-scale (CRS) models requires the assumption of full proportionality between all inputs and outputs. Often such proportionality cannot be assumed, although there may be a subset of outputs proportional to a subset of inputs. By using the variable returns-to-scale (VRS) model, this information is effectively ignored and the efficiency of units is overestimated. This paper develops a hybrid approach that combines the assumption of CRS with respect to the selected sets of inputs and outputs, while preserving the VRS assumption with respect to the remaining indicators. The resulting hybrid returns-to-scale models exhibit better discrimination than the VRS model. In certain cases, their discrimination surpasses that of the CRS model, an example of which is given.     

Additive efficiency decomposition in two-stage DEA

Kao and Hwang (2008) [Kao, C., Hwang, S.-N., 2008. Efficiency decomposition in two-stage data envelopment analysis: An application to non-life insurance companies in Taiwan. European Journal of Operational Research 185 (1), 418–429] develop a data envelopment analysis (DEA) approach for measuring efficiency of decision processes which can be divided into two stages. The first stage uses inputs to generate outputs which become the inputs to the second stage. The first stage outputs are referred to as intermediate measures. The second stage then uses these intermediate measures to produce outputs. Kao and Huang represent the efficiency of the overall process as the product of the efficiencies of the two stages. A major limitation of this model is its applicability to only constant returns to scale (CRS) situations. The current paper develops an additive efficiency decomposition approach wherein the overall efficiency is expressed as a (weighted) sum of the efficiencies of the individual stages. This approach can be applied under both CRS and variable returns to scale (VRS) assumptions. The case of Taiwanese non-life insurance companies is revisited using this newly developed approach.     

Estimating non-convex production sets – imposing convex input sets and output sets in data envelopment analysis

In this paper, we present Data Envelopment Analysis models for estimating non-convex production sets. In contrast to the Banker et al. (Banker, R.D., Charnes, A. and Cooper, W.W., 1984. Management Science 30 (9), 1078–1092) model, these models give statistically consistent estimators for production sets that are non-convex, but do have convex input sets and/or convex output sets, as is typically assumed in micro-economic theory. In addition, these models suffer less from a finite sample error than the Free Disposable Hull (Deprins, D., Simar, L., Tulkens, H., 1984. In: Marchand, M., Pestieu, P., Tulkens, H. (Eds.), The Performance of Public Enterprises. North-Holland, Amsterdam, pp. 243–267) model.     

Super-efficiency measurement under variable return to scale: an approach based on a new directional distance function

A modified super-efficiency model based on the directional distance function (DDF) has recently been developed in order to tackle the infeasibility issue in the two exceptions of the Nerlove–Luenberger (N–L) super-efficiency model under variable return to scale (VRS). However, we find that model does not fully eliminate the infeasibility issue. This paper chooses an appropriate reference bundle in the DDF so that the resulting DDF-based VRS super-efficiency model is always feasible. The proposed new model successfully addresses the infeasibility issue of conventional VRS super-efficiency models and fully eliminates the infeasibility issue in the two exceptions of the VRS N–L super-efficiency model. Additional advantages of the new model include: it is unit-invariant and does not need to predetermine any parameter. Theoretical analyses and numerical examples support the practicality and superiority of our model when compared with other super-efficiency models.     

Mixed partial elasticities in constant returns-to-scale production technologies

Recently, Podinovski and Førsund (2010) developed a linear programming approach to the analysis and calculation of a class of mixed partial elasticity measures in variable returns-to-scale (VRS) production technologies. In the current paper, we extend their approach to the constant returns-to-scale (CRS) technologies and formulate linear programs required for the computation of elasticity measures. Among other results obtained in this paper, we prove a new result, valid in both VRS and CRS technologies, that allows us to identify the reason why the corresponding elasticity measure is undefined at the unit. This removes the need for a preliminary sorting of the units into those units where the elasticity measure applies and those where it does not.     

Scale economies and production function estimation for object-oriented software component and source code documentation size

In this study, we investigate the factors that influence the object-oriented (OO) component size and source code documentation. For multiple inputs and multiple outputs, we use data envelopment analysis to illustrate that non-linear variable returns to scale (VRS) economies exist for OO component size and source code documentation. The existence of non-linear variable returns to scale economies indicates that non-linear regression models will perform better than linear regression models. Using empirical data, we compare the performance of non-linear artificial neural network (ANN) forecasting model and linear regression model. Our results indicate that the ANN model performs well when VRS economies exist between multiple inputs and multiple outputs.     

DEA-based hypothesis tests for comparing two groups of decision making units

In this paper, we develop five statistical tests to compare the efficiencies of different groups of DMUs. We consider a data generating process (DGP) that models the deviation of the output from the best practice frontier as the sum of two components, a one-sided inefficiency term and a two-sided random noise term. We use simulation to evaluate the performance of the five tests against the Banker tests (Banker, 1993) that were designed for DGPs containing a single one-sided error term. It is found that while the Banker tests are very effective when efficiency dominates noise, the tests developed in this paper perform better than the Banker tests when noise levels are significant.     

A comment on “Measuring super-efficiency in DEA in the presence of infeasibility”

In a recent paper by Chen [Chen, Y., 2005. Measuring super-efficiency in DEA in the presence of infeasibility. European Journal of Operational Research 161 (1), 447–468], he deals with the infeasibility of super-efficiency DEA models in variable returns to scale (VRS) technology. He provides a necessary and sufficient condition for simultaneous infeasibility of input- and output-oriented super-efficiency DEA models in VRS case, then he claims that both of these models are infeasible only for a rare situation. In this paper, we present some counterexamples and comments to the contention by Chen.