Abstract
In this paper, we propose a general approach to find the closest targets for a given unit according to a previously specified criterion of similarity. The idea behind this approach is that closer targets determine less demanding levels of operation for the inputs and outputs of the inefficient units to perform efficiently. Similarity can be interpreted as closeness between the inputs and outputs of the assessed unit and the proposed targets, and this closeness can be measured by using either different distance functions or different efficiency measures. Depending on how closeness is measured, we develop several mathematical programming problems that can be easily solved and guarantee to reach the closest projection point on the Pareto-efficient frontier. Thus, our approach leads to the closest targets by means of a single-stage procedure, which is easier to handle than those based on algorithms aimed at identifying all the facets of the efficient frontier.
Similar content being viewed by others
Notes
We would like to explicitly make the distinction between setting targets and benchmarking in the sense that targets are the coordinates of a projection point, which is not necessarily an observed DMU, whereas benchmarks are real observed DMUs, which are not the concern in this paper.
Note that, as pointed out by Briec and Leleu (2003), many of the distance functions commonly used in the context of the efficiency evaluation do not measure a distance in a mathematical sense.
Silva et al. (2003) explicitly states “To find closest targets one needs to use multi-stages procedures so that we can maximise....” the value of the used efficiency measure, “.... while at the same time assuring projection on the efficient frontier”.
It should be noted that the possibility of setting targets from points obtained by slightly modifying projections on the weakly efficient frontier may be risky since these points might be outside the production possibility set, and so, they would violate the axioms that generate the technology.
The extreme efficient units are the DMUs spanning the efficient facets of the frontier that cannot be expressed as a linear combination of the other DMUs. For a formal definition, see Charnes et al. (1991).
Olesen and Petersen (1996) use a similar set of constraints in a model intended to assess the efficiency with reference to FDEFs.
Note also that, since (mADD) determines the minimum L1-distance from any point (X0, Y0) in the PPS to the Pareto-efficient frontier, this model goes one step further the work by Briec in which the author finds the minimum L1-distance to the weak efficient frontier. This also applies with the approach in Charnes et al. (1996).
The comments in the previous footnote on the approach by Briec with respect to the L1-distance also apply in the case of the L∞-distance.
It can be shown that the change of variables suggested in Charnes and Cooper (1962) to make linear a continuous linear fractional program also leads to the optimal solution of the model proposed here.
References
Ali AI, Seiford LM (1993) The mathematical programming approach to efficiency analysis. In: Fried HO, Lovell CAK, Schmidt SS (eds) The measurement of productive efficiency: techniques and applications. NY Oxford University Press, New York
Banker RD, Charnes A, Cooper WW (1984) Some models for estimating technical and scale inefficiencies in data envelopment analysis. Manage Sci 30(9):1078–1092
Bogetoft P, Hougaard JL (1999) Efficiency evaluations based on potential non-proportional improvements. J Product Anal 12:233–247
Briec W (1999) Hölder distance function and measurement of technical efficiency. J Product Anal 11(2):111–131
Briec W, Leleu H (2003) Dual representations of non-parametric technologies and measurement of technical efficiency. J Product Anal 20:71–96
Briec W, Lemaire B (1999) Technical efficiency and distance to a reverse convex set. Eur J Operat Res 114:178–187
Brockett PL, Rousseau JJ, Wang Y, Zhow L (1997) Implementation of DEA models using GAMS. Research Report 765, University of Texas, Austin
Charnes A, Cooper WW (1962) Programming with linear fractional functionals. Naval Res Logist Quart 9:181–186
Charnes A, Cooper WW, Thrall RM (1991) A structure for classifying and characterizing efficiency and inefficiency in data envelopment analysis. J Product Anal 2:197–237
Charnes A, Cooper WW, Golany B, Seiford L, Stutz J (1985) Foundations of data envelopment analysis for Pareto-Koopmans efficient empirical productions functions. J Economet 30:91–107
Charnes A, Cooper WW, Golany B, Seiford L, Stutz J (1987) A dimensionless efficiency measure for departures from Pareto optimality. Research Report CCS 480, Center for Cybernetic Studies, The University of Texas, Austin
Charnes A, Rousseau JJ, Semple JH (1996) Sensitivity and stability of efficiency classifications in data envelopment analysis. J Product Anal 7:5–18
Cherchye L, Van Puyenbroeck T (2001) A comment on multi-stage DEA methodology. Operat Res Lett 28(2):93–98
Coelli T (1998) A multi-stage methodology for the solution of orientated DEA models. Operat Res Lett 23(3–5):143–149
Cooper WW, Park KS, Pastor JT (1999) RAM: a range adjusted measure of inefficiency for use with additive models, and relations to other models and measures in DEA. J Product Anal 11(1):5–42
Cooper WW, Seiford LM, Tone K (2000) Data envelopment analysis: a comprehensive text with models, applications, references and DEA-solver software. Kluwer Academic Publishers, Boston
Färe R, Lovell CAK (1978) Measuring the technical efficiency of production. J Econ Theor 19:150–162
Frei FX, Harker PT (1999) Projections onto efficient frontiers: theoretical and computational extensions to DEA. J Product Anal 11:275–300
González E, Álvarez A (2001) From efficiency measurement to efficiency improvement: the choice of a relevant benchmark. Eur J Operat Res 133:512–520
Lozano S, Villa G (2005) Determining a sequence of targets in DEA. J Operat Res Soc 56:1439–1447
Olesen O, Petersen NC (1996) Indicators of ill-conditioned data sets and model misspecification in data envelopment analysis: an extended facet approach. Manage Sci 42:205–219
Olesen O, Petersen NC (2003) Identification and use of efficient faces and facets in DEA. J Product Anal 20:323–360
Pastor JT, Ruiz JL, Sirvent I (1999) An enhanced DEA Russell graph efficiency measure. Eur J Operat Res 115(3):596–607
Ray SC (2004) Data envelopment analysis. Theory and techniques for economics and operations research. Cambridge University Press
Silva MCA, Castro P, Thanassoulis E (2003) Finding closest targets in non-oriented DEA models: the case of convex and non-convex technologies. J Product Anal 19:251–269
Takeda A, Nishino H (2001) On measuring the inefficiency with the inner-product norm in data envelopment analysis. Eur J Operat Res 133:377–393
Thanassoulis E (2001) Introduction to the theory and application of data envelopment analysis: a foundation text with integrated software. Kluwer Academic Publishers, Boston
Tone K (2001) A Slacks-based measure of efficiency in data envelopment analysis. Eur J Operat Res 130(3):498–509
Acknowledgements
We are deeply grateful to Prof. Timo Kuosmanen both for his comments and for his interest in the paper. We are also grateful to two anonymous referees for their constructive comments. Finally, we would like to thank the Ministerio de Educación y Ciencia (MTM2004-07473) for the financial support.
Author information
Authors and Affiliations
Corresponding author
Appendix: proof of the theorem
Appendix: proof of the theorem
First, we prove the sufficient condition.
Let \({({\tilde {\hbox{X}}, \tilde {\hbox{Y}}})\in \hbox{D}_0.}\) Since this point is efficient, the optimal value of the following pair of dual LP problems will be zero:
Since \({({\tilde {\hbox{X}}, \tilde {\hbox{Y}}})\ne 0_{{\rm m+s}}},\) we can find a set of non-negative scalars \({\tilde {\lambda }_{\rm j} ,\;\hbox{j}\in \hbox{E},}\) with at least one of them being strictly positive, such that \({\left({\tilde {\hbox{X}},\tilde {\hbox{Y}}} \right)= ({\sum_{{\rm j}\in {\rm E}} {\tilde {\lambda }_{\rm j} \hbox{X}_{\rm j} }, \sum_{{\rm j}\in {\rm E}} {\tilde {\lambda}_{\rm j} {\rm Y}_{\rm j} } } )}.\) Moreover, for complementary slackness, for each \({\tilde {\lambda }_{\rm j} > 0,}\) the corresponding dual constraint is binding, that is, \({\exists \;\tilde {\nu}_{\rm i} \ge 1,\;\hbox{i}=1,\ldots,\hbox{m},}\) and \({\tilde {\mu}_{\rm r} \ge 1,\;\hbox{r}=1,\ldots,\hbox{s},}\) satisfying \({-\sum_{{\rm i}=1}^{\rm m} {\tilde {\nu}_{\rm i} \hbox{x}_{\rm ij} +\sum_{{\rm r}=1}^{\rm s} {\tilde {\mu}_{\rm r} \hbox{y}_{{\rm rj}} =0}}}.\)
Then, let us define \({\tilde {\hbox{d}}_{\rm j} =\tilde {\hbox{b}}_{\rm j} =0,}\) for each \({\tilde {\lambda}_{\rm j} > 0,}\) and \({\tilde {\hbox{b}}_{\rm j} =1}\) and \({\tilde {\hbox{d}}_{\rm j} =\sum_{{\rm i}=1}^{\rm m} {\tilde {\nu}_{\rm i} \hbox{x}_{\rm ij} -\sum_{{\rm r}=1}^{\rm s} {\tilde {\mu}_{\rm r} \hbox{y}_{\rm rj} \ge 0}},}\) for each \({\tilde {\lambda}_{\rm j} =0.}\) Finally, since \({({\tilde {\hbox{X}}, \tilde {\hbox{Y}}})}\) dominates DMU0, we can also define \({\tilde {\hbox{s}}_{{\rm i}0}^{-} =\hbox{x}_{{\rm i}0} -\tilde {\hbox{x}}_{\rm i} \ge 0,\;\hbox{i}=1, \ldots,{\rm m},}\) and \({\tilde {\hbox{s}}_{{\rm r}0}^{+} =\tilde {\hbox{y}}_{\rm r} -\hbox{y}_{{\rm r}0} \ge 0,\;\hbox{r}=1,\ldots,\hbox{s}.}\)
Therefore, \({\tilde {\lambda }_{\rm j},\;\tilde {\hbox{d}}_{\rm j} ,\;\tilde {\hbox{b}}_{\rm j},\;\hbox{j}\in \hbox{E},\;\tilde {\nu}_{\rm i},\;\tilde {\hbox{s}}_{{\rm i}0}^{-} ,\;\hbox{i}=1,\ldots,\hbox{m},}\) and \({\tilde {\mu}_{\rm r},\tilde {\hbox{s}}_{{\rm r}0}^{+},\;\hbox{r}=1,\ldots,\hbox{s},}\) satisfy all the desired conditions.
Conversely, let \({\tilde {\lambda}_{\rm j},\;\tilde {\hbox{d}}_{\rm j} \ge 0,\;\tilde {\hbox{b}}_{\rm j} \in \{{0,1}\},\;\hbox{j}\in \hbox{E},\; \tilde {\nu}_{\rm i} \ge 1,\;\tilde {\hbox{s}}_{{\rm i}0}^{-} \ge 0,\;\hbox{i}=1,\ldots,\hbox{m},}\) and \({\tilde {\mu}_{\rm r} \ge 1, \; \tilde {\hbox{s}}_{{\rm r}0}^{+} \ge 0,\;\hbox{r}=1,\ldots,\hbox{s},}\) be scalars satisfying (t.3)–(t.7) and define \({\left({\tilde {\hbox{X}},\tilde {\hbox{Y}}} \right)=({\sum_{{\rm j}\in {\rm E}} {\tilde {\lambda }_{\rm j} \hbox{X}_{\rm j} }, \sum_{{\rm j}\in {\rm E}} {\tilde {\lambda }_j Y_j } } )}.\) Then we will show that \({\left({\tilde {\hbox{X}},\tilde {\hbox{Y}}} \right)\in \hbox{D}_0 .}\)
Obviously, \({({\tilde {\hbox{X}}, \tilde {\hbox{Y}}})}\) dominates (X0, Y0) as a consequence of the constraints (t.3) and (t.4). Therefore, we only need to prove that \({\left({\tilde {\hbox{X}}, \tilde {\hbox{Y}}} \right)}\) belongs to the efficient frontier. To do it we will show that this point is a linear combination of extreme efficient DMUs lying on the same efficient facet of the frontier and, consequently, that point will also be an efficient point.
For each \({\tilde {\lambda}_{\rm j} > 0,}\) we have \({\tilde {\hbox{d}}_{\rm j} =\tilde {\hbox{b}}_{\rm j} =0}\) by virtue of (t.6) and (t.7) (note that there always exists such a positive intensity as a result of equalities (t.3)–(t.4) and the fact that at least one input and one output of DMU0 is non-zero). Hence, all the extreme efficient units with a non-zero coefficient in the linear combinations (t.1) and (t.2) satisfy the equation \({-\sum_{{\rm i}=1}^{\rm m} {\tilde {\nu}_{\rm i} \hbox{x}_{\rm i} +\sum_{{\rm r}=1}^{\rm s} {\tilde {\mu}_{\rm r} \hbox{y}_{\rm r} =0}}}.\) Since, in addition, these multipliers hold (t.5) and they all are strictly positive (in particular, greater than or equal to the unity), we have that such equation defines a facet of the efficient frontier.
Rights and permissions
About this article
Cite this article
Aparicio, J., Ruiz, J.L. & Sirvent, I. Closest targets and minimum distance to the Pareto-efficient frontier in DEA. J Prod Anal 28, 209–218 (2007). https://doi.org/10.1007/s11123-007-0039-5
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11123-007-0039-5