Abstract
Data envelopment analysis (DEA) is a very effective management tool in assessing the performance of a set of decision making units (DMUs). In the efficiency evaluation using classic single stage DEA models, the internal processes of the DMUs are often neglected. In most real-world problems, it may be more realistic to evaluate the efficiency evaluation in two-stage production systems. In some cases: however, the decision-maker must need to identify the most efficient single unit. Numerous methods have been introduced to find the most efficient unit in single stage systems whereas no methods have been proposed for this aim in two stage production systems. Therefore, a new model based on mixed-integer programming was proposed to determine the most efficient DMU in two-stage systems and sub-stages in this study. The most important innovation of the suggested approach is that the most efficient DMUs of both stages can be found separately using only one model. Numerical examples for real world problems and a simulation study were provided for the validity of the proposed model.
Similar content being viewed by others
References
Amin, G. R. (2009). Comments on finding the most efficient DMUs in DEA: An improved integrated model. Computers & Industrial Engineering, 56(4), 1701–1702.
Amin, G. R., & Toloo, M. (2007). Finding the most efficient DMUs in DEA: An improved integrated model. Computers & Industrial Engineering, 52(1), 71–77.
An, Q. X., Chen, H. X., Wu, J., & Liang, L. (2015). Measuring slacks-based efficiency for commercial banks in China by using a two-stage DEA model with undesirable output. Annals of Operations Research, 235, 13–35.
Andersen, P., & Petersen, N. C. (1993). A procedure for ranking efficient units in data envelopment analysis. Management Science, 39(10), 1261–1264.
Azizi, R., & Matin, R. K. (2016). Ranking two-stage production units in data envelopment analysis. Asia-Pacific Journal of Operational Research, 33(01), 1650002.
Charnes, A., & Cooper, W. W. (1984a). Preface to topics in data envelopment analysis. Annals of Operations Research, 2(1), 59–94.
Charnes, A., & Cooper, W. W. (1984b). The non-Archimedean CCR ratio for efficiency analysis: A rejoinder to Boyd and Färe. European Journal of Operational Research, 15(3), 333–334.
Charnes, A., Cooper, W. W., & Rhodes, E. (1978). Measuring the efficiency of decision making units. European Journal of Operational Research, 2(6), 429–444.
Chen, Y., & Zhu, J. (2004). Measuring information technology’s indirect impact on firm performance. Information Technology and Management, 5(1), 9–22.
Chen, C., Zhu, J., Yu, J. Y., & Noori, H. (2012). A new methodology for evaluating sustainable product design performance with two-stage network data envelopment analysis. European Journal of Operational Research, 221(2), 348–359.
Chilingerian, J. A., & Sherman, H. D. (2011). Health-care applications: From hospitals to physicians, from productive efficiency to quality frontiers. Handbook on data envelopment analysis (pp. 445–493). US: Springer.
Ertay, T., & Ruan, D. (2005). Data envelopment analysis based decision model for optimal operator allocation in CMS. European Journal of Operational Research, 164(3), 800–810.
Foroughi, A. A. (2011). A new mixed integer linear model for selecting the best decision making units in data envelopment analysis. Computers & Industrial Engineering, 60(4), 550–554.
Friedman, L., & Sinuany-Stern, Z. (1997). Scaling units via the canonical correlation analysis in the DEA context. European Journal of Operational Research, 100(3), 629–637.
Friedman, L., & Sinuany-Stern, Z. (1998). Combining ranking scales and selecting variables in the DEA context: The case of industrial branches. Computers & Operations Research, 25(9), 781–791.
Galagedera, D. U., Roshdi, I., Fukuyama, H., & Zhu, J. (2018). A new network DEA model for mutual fund performance appraisal: An application to US equity mutual funds. Omega, 77, 168–179.
Golany, B. (1998). An interactive MOLP procedure for the extension of DEA to effectiveness analysis. Journal of the Operational Research Society, 39(8), 725–734.
Gong, Y., Zhu, J., Chen, Y., & Cook, W. D. (2018). DEA as a tool for auditing: Application to Chinese manufacturing industry with parallel network structures. Annals of Operations Research, 263(1–2), 247–269.
Hatami-Marbini, A., & Saati, S. (2020). Measuring performance with common weights: Network DEA. Neural Computing and Applications, 32, 3599–3617.
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.
Kao, C., & Hwang, S. N. (2014). Multi-period efficiency and Malmquist productivity index in two-stage production systems. European Journal of Operational Research, 232(3), 512–521.
Kao, C., & Liu, S. T. (2019). Cross efficiency measurement and decomposition in two basic network systems. Omega, 83, 70–79.
Karsak, E. E., & Ahiska, S. S. (2005). Practical common weight multi-criteria decision-making approach with an improved discriminating power for technology selection. International Journal of Production Research, 43(8), 1537–1554.
Lam, K. F. (2015). In the determination of the most efficient decision making unit in data envelopment analysis. Computers & Industrial Engineering, 79, 76–84.
Li, Y., Chen, Y., Liang, L., & Xie, J. (2012). DEA models for extended two-stage network structures. Omega, 40(5), 611–618.
Li, Y., Lei, X., Dai, Q., & Liang, L. (2015). Performance evaluation of participating nations at the 2012 London Summer Olympics by a two-stage data envelopment analysis. European Journal of Operational Research, 243(3), 964–973.
Li, X. B., & Reeves, G. R. (1999). A multiple criteria approach to data envelopment analysis. European Journal of Operational Research, 115(3), 507–517.
Liu, J. S., & Lu, W. M. (2010). DEA and ranking with the network-based approach: A case of R&D performance. Omega, 38(6), 453–464.
Liu, J. S., Lu, W. M., Yang, C., & Chuang, M. (2009). A network-based approach for increasing discrimination in data envelopment analysis. Journal of the Operational Research Society, 60(11), 1502–1510.
Moreno, P., & Lozano, S. (2014). A network DEA assessment of team efficiency in the NBA. Annals of Operations Research, 214(1), 99–124.
Örkcü, H. H., Özsoy, V. S., Örkcü, M., & Bal, H. (2019). A neutral cross efficiency approach for basic two stage production systems. Expert Systems with Applications, 125, 333–344.
Premachandra, I. M., Zhu, J., Watson, J., & Galagedera, D. U. (2012). Best-performing US mutual fund families from 1993 to 2008: Evidence from a novel two-stage DEA model for efficiency decomposition. Journal of Banking & Finance, 36(12), 3302–3317.
Seiford, L. M., & Zhu, J. (1999). Profitability and marketability of the top 55 US commercial banks. Management Science, 45(9), 1270–1288.
Sexton, T. R., & Lewis, H. F. (2003). Two-stage DEA: An application to major league baseball. Journal of Productivity Analysis, 19(2), 227–249.
Sexton, T. R., Silkman, R. H., & Hogan, A. J. (1986). Data envelopment analysis: Critique and extensions. New Directions for Evaluation, 32, 73–105.
Sinuany-Stern, Z., Mehrez, A., & Barboy, A. (1994). Academic departments efficiency via DEA. Computers & Operations Research, 21(5), 543–556.
Sueyoshi, T. (1999). DEA non-parametric ranking test and index measurement: Slack-adjusted DEA and an application to Japanese agriculture cooperatives. Omega, 27(3), 315–326.
Toloo, M. (2012). On finding the most BCC-efficient DMU: A new integrated MIP–DEA model. Applied Mathematical Modelling, 36(11), 5515–5520.
Toloo, M. (2014a). An epsilon-free approach for finding the most efficient unit in DEA. Applied Mathematical Modelling, 38(13), 3182–3192.
Toloo, M. (2014b). The role of non-Archimedean epsilon in finding the most efficient unit: With an application of professional tennis players. Applied Mathematical Modelling, 38(21–22), 5334–5346.
Toloo, M. (2015). Alternative minimax model for finding the most efficient unit in data envelopment analysis. Computers & Industrial Engineering, 81, 186–194.
Toloo, M. (2016). A cost efficiency approach for strategic vendor selection problem under certain input prices assumption. Measurement, 85, 175–183.
Toloo, M., & Ertay, T. (2014). The most cost efficient automotive vendor with price uncertainty: A new DEA approach. Measurement, 52, 135–144.
Toloo, M., & Salahi, M. (2018). A powerful discriminative approach for selecting the most efficient unit in DEA. Computers & Industrial Engineering, 115, 269–277.
Torgersen, A. M., Førsund, F. R., & Kittelsen, S. A. (1996). Slack-adjusted efficiency measures and ranking of efficient units. Journal of Productivity Analysis, 7(4), 379–398.
Wang, Y. M., & Chin, K. S. (2010). Some alternative DEA models for two-stage process. Expert Systems with Applications, 37, 8799–8808.
Wang, C. H., Gopal, R. D., & Zionts, S. (1997). Use of data envelopment analysis in assessing information technology impact on firm performance. Annals of Operations Research, 73, 191–213.
Wang, Y. M., & Jiang, P. (2012). Alternative mixed integer linear programming models for identifying the most efficient decision making unit in data envelopment analysis. Computers & Industrial Engineering, 62(2), 546–553.
Wanke, P., & Barros, C. (2014). Two-stage DEA: An application to major Brazilian banks. Expert Systems with Applications, 41(5), 2337–2344.
Xie, J., Zhu, X., & Liang, L. (2020). A multiplicative method for estimating the potential gains from two-stage production system mergers. Annals of Operations Research, 288, 475–493.
Zhu, J. (2000). Multi-factor performance measure model with an application to Fortune 500 companies. European Journal of Operational Research, 123(1), 105–124.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
1.1 Appendix A. Proofs of theorems
Theorem 1
The MILP model (4) is feasible.
Proof
To ease of notation, we let \( I_{r}^{u} = \frac{1}{{\left( {s + q + m} \right)\hbox{max} \left( {y_{rj} } \right)}};\; r = 1, \ldots ,s \), \( I_{p}^{w} = \frac{1}{{\left( {s + q + m} \right)\hbox{max} \left( {z_{pj} } \right)}}; \;p = 1, \ldots ,q \), and \( I_{i}^{v} = \frac{1}{{\left( {s + q + m} \right)\hbox{max} \left( {x_{ij} } \right)}}; \;i = 1, \ldots ,m \).
For \( I_{r}^{u} \) and \( I_{p}^{w} \), in the third set of constraints of model (4), there are two situations.
One is \( \sum\nolimits_{r = 1}^{s} {I_{r}^{u} y_{rj} } - \sum\nolimits_{p = 1}^{q} {I_{p}^{w} z_{pj} } \le I_{2j} \) for \( \forall j \in \left\{ {1, \ldots ,n} \right\} \) and the other is \( \sum\nolimits_{r = 1}^{s} {I_{r}^{u} y_{rj} } - \sum\nolimits_{p = 1}^{q} {I_{p}^{w} z_{pj} } \ge I_{2j} \) for \( \exists j \in \left\{ {1, \ldots ,n} \right\} \).
Let’s start with \( \sum\nolimits_{r = 1}^{s} {I_{r}^{u} y_{rj} } - \sum\nolimits_{p = 1}^{q} {I_{p}^{w} z_{pj} } \le I_{2j} \).
If \( \sum\nolimits_{p = 1}^{q} {I_{p}^{w} z_{pj} } - \sum\nolimits_{i = 1}^{m} {I_{i}^{v} x_{ij} } \le I_{1j} \), the model has a possible solution. Conversely, if \( \sum\nolimits_{p = 1}^{q} {I_{p}^{w} z_{pj} } - \sum\nolimits_{i = 1}^{m} {I_{i}^{v} x_{ij} } \ge I_{1j} \) \( \exists j \in \left\{ {1, \ldots ,n} \right\} \), we consider the set of deviation variable \( d_{1j} > 0 \left( {j = 1, \ldots ,n} \right) \) meet the conditions \( \sum\nolimits_{p = 1}^{q} {I_{p}^{w} z_{pj} } - \sum\nolimits_{i = 1}^{m} {I_{i}^{v} x_{ij} - d_{1j} } \le I_{1j} \). Let \( \delta_{1} = \mathop {\hbox{max} }\limits_{j} \left\{ {{{d_{1j} } \mathord{\left/ {\vphantom {{d_{1j} } {\sum\nolimits_{{{\text{i}} = 1}}^{m} {x_{ij} } }}} \right. \kern-0pt} {\sum\nolimits_{{{\text{i}} = 1}}^{m} {x_{ij} } }}} \right\} \) and \( v_{i}^{a} = I_{i}^{v} + \delta_{1} \) for \( \forall {\text{i}} \in \left\{ {1, \ldots ,m} \right\} \). Then we can write \( \delta_{1} > 0,v_{i}^{a} > I_{i}^{v} \) for \( \forall i \in \left\{ {1, \ldots ,m} \right\} \), and using the \( {{\delta_{1} \ge d_{1j} } \mathord{\left/ {\vphantom {{\delta_{1} \ge d_{1j} } {\sum\nolimits_{p = 1}^{q} {x_{ij} } }}} \right. \kern-0pt} {\sum\nolimits_{p = 1}^{q} {x_{ij} } }} \), we obtain
for \( \forall j \in \left\{ {1, \ldots ,n} \right\} \). Thus, \( I_{p}^{w} \left( {p = 1, \ldots ,q} \right),v_{i}^{a} \left( {i = 1, \ldots ,m} \right) \), \( I_{r}^{u} \;\left( {r = 1, \ldots ,s} \right) \) are in feasible region.
Consider the \( \sum\nolimits_{r = 1}^{s} {I_{r}^{u} y_{rj} } - \sum\nolimits_{p = 1}^{q} {I_{p}^{w} z_{pj} } \ge I_{2j} \) \( \exists j \in \left\{ {1, \ldots ,n} \right\} \). Let a set of deviation variable \( d_{2j} > 0 \left( {j = 1, \ldots ,n} \right) \) meet the conditions \( \sum\nolimits_{r = 1}^{s} {I_{r}^{u} y_{rj} } - \sum\nolimits_{p = 1}^{q} {I_{p}^{w} z_{pj} } - d_{2j} \le I_{2j} \). Let \( \delta_{2} = \mathop {\hbox{max} }\limits_{j} \left\{ {{{d_{2j} } \mathord{\left/ {\vphantom {{d_{2j} } {\sum\nolimits_{p = 1}^{q} {z_{pj} } }}} \right. \kern-0pt} {\sum\nolimits_{p = 1}^{q} {z_{pj} } }}} \right\} \) and \( w_{p}^{a} = I_{p}^{w} + \delta_{2} \) for \( \forall p \in \left\{ {1, \ldots ,q} \right\} \) for this case. Then we can write \( \delta_{2} > 0, w_{p}^{a} > I_{p}^{w} \) for \( \forall p \in \left\{ {1, \ldots ,q} \right\} \), and using \( {{\delta_{2} \ge d_{2j} } \mathord{\left/ {\vphantom {{\delta_{2} \ge d_{2j} } {\sum\nolimits_{p = 1}^{q} {z_{pj} } }}} \right. \kern-0pt} {\sum\nolimits_{p = 1}^{q} {z_{pj} } }} \), i.e., \( \delta_{2} \sum\nolimits_{p = 1}^{q} {z_{pj} } \ge d_{2j} \) we have
for \( \forall j \in \left\{ {1, \ldots ,n} \right\} \).
If \( \sum\nolimits_{p = 1}^{q} {w_{p}^{a} z_{pj} } - \sum\nolimits_{i = 1}^{m} {I_{i}^{v} x_{ij} } \le I_{1j} \) \( \forall j \in \left\{ {1, \ldots ,n} \right\} \), the model has a feasible region.
If \( \sum\nolimits_{p = 1}^{q} {w_{p}^{a} z_{pj} } - \sum\nolimits_{i = 1}^{m} {I_{i}^{v} x_{ij} } \ge I_{1j} \) \( \exists j \in \left\{ {1, \ldots ,n} \right\} \), we consider a set of deviation variable \( \tilde{d}_{1j} > 0 \left( {j = 1, \ldots ,n} \right) \) meet the conditions \( \sum\nolimits_{p = 1}^{q} {w_{p}^{a} z_{pj} } - \sum\nolimits_{i = 1}^{m} {I_{i}^{v} x_{ij} } - \tilde{d}_{1j} \le I_{1j} \). Let \( \tilde{\delta }_{1} = \mathop {\hbox{max} }\limits_{j} \left\{ {{{\tilde{d}_{1j} } \mathord{\left/ {\vphantom {{\tilde{d}_{1j} } {\sum\nolimits_{{{\text{i}} = 1}}^{m} {x_{ij} } }}} \right. \kern-0pt} {\sum\nolimits_{{{\text{i}} = 1}}^{m} {x_{ij} } }}} \right\} \) and \( v_{i}^{b} = I_{i}^{v} + \tilde{\delta }_{1} \) for \( \forall i \in \left\{ {1, \ldots ,m} \right\} \). Then we can write \( \tilde{\delta }_{1} > 0,v_{i}^{a} > I_{i}^{v} \) for \( \forall i \in \left\{ {1, \ldots ,m} \right\} \), and using the \( \tilde{\delta }_{1} \ge {{\tilde{d}_{1j} } \mathord{\left/ {\vphantom {{\tilde{d}_{1j} } {\sum\nolimits_{p = 1}^{q} {x_{ij} } }}} \right. \kern-0pt} {\sum\nolimits_{p = 1}^{q} {x_{ij} } }} \), we obtain
for \( \forall j \in \left\{ {1, \ldots ,n} \right\} \). So, \( w_{p}^{a} \left( {p = 1, \ldots ,q} \right), \) \( v_{i}^{b} \) \( \left( {i = 1, \ldots ,m} \right) \), \( I_{r}^{u} \) \( \left( {r = 1, \ldots ,s} \right) \) are in feasible region. That is, the MILP model for two stages is always feasible. If lower bound solutions are infeasible then, these can be made feasible by increasing the values of input weight (\( v_{i} \)) or of the intermediate measures weights (\( w_{p} \)).
Summing the constraints \( \sum\nolimits_{p = 1}^{q} {w_{p} z_{pj} } - \sum\nolimits_{i = 1}^{m} {v_{i} x_{ij} } \le I_{1j} \) and \( \sum\nolimits_{r = 1}^{s} {u_{r} y_{rj} } - \sum\nolimits_{p = 1}^{q} {w_{p} z_{pj} } \le I_{2j} \) over \( j \) from \( j = 1 \) to \( n \) seperately, we have
and
Summing up the last two inequalities, we can write
Therefore, the objective function of the model (4) has a lower bound. As the MILP model is a minimization problem, model (4) is a feasible solution. This completes the proof.
Theorem 2
Model (4) and model (5) are equivalent.
Proof
From the \( \sum\nolimits_{p = 1}^{q} {w_{p} z_{pj} } - \sum\nolimits_{i = 1}^{m} {v_{i} x_{ij} + S_{1j} } = I_{1j} \) constraint, we can write
Summing up these constraints over \( j \) from \( j = 1 \) to \( j = n \) and considering \( \sum\nolimits_{j = 1}^{n} {I_{1j} } = 1, \) we can write
Similarly, from the third type of constraints in the model (5), summing up these constraints over j from j = 1 to j = n, and considering \( \sum\nolimits_{j = 1}^{n} {I_{2j} } = 1, \) we can write
By summing up the last two inequalities, we obtain
If the constant 2 is ignored, the objective function of the model (5) is written as
which is equal to the objective function of model (4). This completes the proof.
Theorem 3
The following model (6) and model (4) are equivalents.
Proof
In model (6), let \( d_{1j} = 1 - I_{1j} \) and \( \beta_{1j} = 1 - S_{1j} \) for \( j = 1, \ldots , n \). Summing up these equations over \( j \) from \( j = 1 \) to \( j = n \), this results in
and
Similarly, if \( d_{2j} = 1 - I_{2j} \) and \( \beta_{2j} = 1 - S_{2j} \) are taken for \( j = 1, \ldots , n \), it is obtained as
Hence, \( \sum\nolimits_{j = 1}^{n} {\left( {\beta_{1j} + \beta_{2j} } \right)} \) is obtained as \( \left( {2n - 2} \right) - \sum\nolimits_{i = 1}^{m} {v_{i} } \left( {\sum\nolimits_{j = 1}^{n} {x_{ij} } } \right) - \sum\nolimits_{r = 1}^{s} {u_{r} } \left( {\sum\nolimits_{j = 1}^{n} {y_{rj} } } \right). \)
Since \( 2n - 2 \) is a constant, the fact that \( \sum\nolimits_{j = 1}^{n} {\left( {\beta_{1j} + \beta_{2j} } \right)} \) is equal to minimum of
completes the proof.
Theorem 4
The proposed minimax model (8) is always feasible and bounded.
Proof
Let \( \left( {u^{0} ,w^{0} ,v^{0} ,I_{1}^{0} ,I_{2}^{0} } \right) \) be a feasible solution to model (7) and we proved that such a solution exists in Theorem 1. Let \( d_{1j}^{0} = 1 - I_{1j}^{0} \), \( \beta_{1j}^{0} = 1 - I_{1j}^{0} + w^{0} z_{j} - v^{0} x_{j} \) for \( j = 1 \) to \( j = n \) and \( d_{max1}^{0} = max\left\{ {d_{1j}^{0} - \beta_{1j}^{0} :j = 1, \ldots , n } \right\} \) for the first stage. Summing up \( d_{1j}^{0} = 1 - I_{1j}^{0} \) over \( j = 1 \) to \( j = n \) results in \( \sum\nolimits_{j = 1}^{n} {d_{1j}^{0} } = n - 1 \). For \( \sum\nolimits_{p = 1}^{q} {w_{p} z_{pj} } - \sum\nolimits_{i = 1}^{m} {v_{i} x_{ij} + d_{1j} - \beta_{1j} } = 0 \) constraints, \( w^{0} z_{j} - v^{0} x_{j} + 1 - I_{1j}^{0} - 1 + I_{1j}^{0} - w^{0} z_{j} + v^{0} x_{j} = 0 \) could be proved. In the first stage, while for the most efficient unit, \( w^{0} z_{j} - v^{0} x_{j} \le 1 \) satisfies, for the other units, \( w^{0} z_{j} - v^{0} x_{j} \le 0 \) satisfies. Therefore, for \( d_{{max_{1} }} - d_{1j} + \beta_{1j} \ge 0 \) constraints, \( d_{max1}^{0} \ge v^{0} x_{j} - w^{0} z_{j} \) could be obtained. Hence, when considering the most efficient unit and other units for the first stage, \( d_{max1}^{0} \ge - 1 \) is obtained. Similarly, if \( d_{2j}^{0} = 1 - I_{2j}^{0} \) and \( \beta_{2j}^{0} = 1 - I_{2j}^{0} + u^{0} y_{j} - w^{0} z_{j} \) for \( j = 1 \) to \( j = n \) and \( d_{max2}^{0} = max\left\{ {d_{2j}^{0} - \beta_{2j}^{0} :j = 1, \ldots , n } \right\} \) are taken for the second stage, the same results are obtained. It is proved that \( \left( {u^{0} ,w^{0} ,v^{0} ,I_{1}^{0} ,I_{2}^{0} } \right) \) is a feasible solution to model (8).
To show that the proposed model has the bounded, \( \left( {u^{0} ,w^{0} ,v^{0} ,d_{1}^{0} ,\beta_{1}^{0} ,d_{2}^{0} ,\beta_{2}^{0} ,d_{max1}^{0} ,d_{max2}^{0} } \right) \) is taken any arbitrary feasible solution to model (8). From the constraints of this model,
and similarly
is obtained. Summing up the last two equations,
means the objective function is bounded from below. Since the proposed model (8) is a minimization problem, the proof is completed.
1.2 Appendix B. Detailed graphs on cases in simulation
See Figs. 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20 and 21.
Rights and permissions
About this article
Cite this article
Özsoy, V.S., Örkcü, M. & Örkcü, H.H. A minimax approach for selecting the overall and stage-level most efficient unit in two stage production processes. Ann Oper Res 300, 137–169 (2021). https://doi.org/10.1007/s10479-020-03827-x
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10479-020-03827-x