Abstract. Keywords: Data Envelopment Analysis (DEA), decision making unit (DMU), efficiency, Korea Securities Dealers Automated Quotation (KOSDAQ)

Transcription

1 , pp The Efficiency Comparative Evaluation of IT Service Companies using the Data Envelopment Analysis Approach Focus on KOSDAQ(KOrea Securities Dealers Automated Quotations) Companies - HyunWoo Goh Dept. of Industrial Management System Engineering, Seokyeong University 16-1 Jungneung-Dong, Sungbuk-Ku, Seoul , KOREA hwgoh@skuniv.ac.kr Abstract IT paradigm is shifting rapidly from technology-oriented to service-oriented and, therefore, the interest on the efficiency of the IT service industry is increasing. In this study, we measure and compare the efficiency of 32 KOSDAQ(Korea Securities Dealers Automated Quotation) IT service companies. We applied the data envelopment analysis(dea) to measure the relative efficiency, which is a linear programming model for measuring the relative efficiency of decision making units(dmus) with multiple inputs and outputs. Assets and capital were selected as input factor, and sales and operating profit as output factor. The technical efficiency was analyzed by CCR-O, and the pure technical efficiency and the scale efficiency by BBC-O. The rank value analysis was performed applying the super efficiency method to distinguish between the DMUs with the same efficiency level. On the basis of analysis results, the degree of improvement was suggested for inefficient companies by benchmarking efficient companies. We expect that the results of efficiency evaluation and the benchmarking method to improve efficiency obtained in this study will provide fruitful implications for decision making of IT service companies. Keywords: Data Envelopment Analysis (DEA), decision making unit (DMU), efficiency, Korea Securities Dealers Automated Quotation (KOSDAQ) 1. Introduction IT industry is sensitive to the economic situation, and its effect of creating demand and employment is great. It is also a subject of public sector investment. When large-scale investment is made for IT industry, its backward linkage effect is very large. Therefore, we need to constantly strive to efficiently distribute the limited resources and to provide a new growth engine. The domestic IT industry is led by manufacturing sectors such as semiconductors, display panels, LED, smart phones, and is moving around the world as a leader. But IT service sector constitutes a mere 1.1% share of the world market, causing an imbalance in the IT industry. IT services include all new services that are related to information and communications, as well as the traditional IT services such as consulting, systems implementation, systems integration, systems operation, infrastructure and its operations, outsourcing, and so on. According to KSIC(Korea Standard Information Classification), IT services are classified as follows: KSI 6201: Computer programming services, KSI 6202: computer system integration, consult management services, KSI 6209: other information technology and computer management services, KSI 6311: Data processing, hosting and related, KSI 6312: portals and other internet information. ISSN: IJSEIA Copyright c 2015 SERSC

2 Meanwhile, KITSA(Korea Information Technology Association) subdivides IT services further than KS does, as follows: consulting, system integration with software development, system integration with package integration (i.e., ERP, SCM, CRM, SEM etc.), network integration, hardware support and integration, DB construction, IT outsourcing, Processing service, network management, data center management, IT education and training. As IT manufacturing sector requires a large investment, the efficiency of the investment is very important and many studies have been conducted to evaluate them. But the assessment is negligent due to the investment of the relatively small scale in IT service industry. However, it is necessary to evaluate the efficiency of IT service industry in the situation where the paradigm is shifting rapidly from IT as a technology to IT as a service. This study compares the efficiency of IT service companies, and suggests a method to improve the efficiency of inefficient enterprises through benchmarking. For this purpose, we use the multi-criteria decision method of DEA, which are widely used to evaluate the efficiency of a company. 2. Theoretical Background 2.1. DEA Model [16] The Data Envelopment Analysis (DEA) is a scientific decision technique, which is one of the multi-criteria decision making method, to measure the relative efficiency of decision making units (DMUs) by the weight limitation of multiple inputs and out puts. DEA developed by Charnes et al., (1978), is a linear programming based technique. The DEA ration form is designed to measure the relative efficiency or productivity of a specific DMUk. The DEA formulation is given as follows. Suppose that there is a set of n DMUs to be analyzed, each of which uses m common inputs and s common outputs. Let k (k= 1,, n) denote the DMUs whose relative efficiency or productivity is to be maximized. max h k = s r=1 u rky rk m i=1 v ik X ik (1) s. t. s r=1 u rk Y rk m i=1 v ik X ik 1, for j = 1,, n u rk > 0, for r = 1,, s v ik > 0, for i = 1,, m Where u rk is the variable weight given to the r th output of the k th DMU, v ik is the variable weight given to the i th output of the k th DMU, u rk and v ik are decision variables determining the relative efficiency of DMU k, Y rj is the r th output of the j th DMU, and X ij is the i th input of the j th DMU. This also assumes that all Y rj and X ij is positive. h k is the efficiency score, and is less than and equal to 1. When the efficiency score of h k is 1, DMU k is regarded as an efficient frontier. CCR(Charne, Cooper, Rhodes) output-oriented model is expressed in the following model, which is composed of vector p and q as variables concerning to the input and output respectively. There are two types of CCR and BCC(Banker, Charnes, Cooper) models. One type is the input oriented model, in which inputs are maximized, and the other is the output 206 Copyright c 2015 SERSC

3 oriented model in which the outputs are maximized. As the focus is on maximizing multiple outputs, this paper employs the output-oriented CCR and BCC model. s.t. qy 0 = 1 px + qy 0 p 0, q 0 min px 0 (2) x 0 and y 0 are the input and output vector of DMU 0. In formular (2), X and Y variables mean that the matrix of inputs and outputs respectively. Let an optimal solution of the output-oriented model is obtained from p = v θ, q = u θ (3) It is clear that (p, q ) is feasible for LP 0. The optimal solution comes from the equation (4). p x 0 = v θ = η (4) x 0 = x 0 t (5) y 0 = η x 0 + t + (6) Where t and t + are the slack variables of input and outputs related to DMU 0 There are various extension models of the CCR approach, among which the BCC model is representative. The BCC approach calculates the efficient frontier group spanned by the convex hull of the existing DMUs. In hence, the condition, which is eλ = 1, is added in the CCR model considering the variable return to scale characterization, which accounts for increased return to scale (IRS), decreased return to scale(drs), and constant return to scale(crs). The BCC output oriented model is expressed as from formular (7) Procedure of DEA [18] s.t. uy 0 = 1 vx uy v 0 e 0 v 0 u 0 v 0 free in sign min z = vx 0 (7) In this section, we explain out method for using DEA to assist in evaluation of companies. The method is composed of five steps, which are briefly explained below. The overall flow of the decision procedure is provided in Figure 1. Copyright c 2015 SERSC 207

4 Collect detailed company STEP 1 Data collection information STEP 2 Factor selection Choose input/output factors STEP 3 Model selection Select proper DEA model as approach Run the DEA model with Choose DEA software STEP 4 specific ranking method Choose DEA a ranking method Result Analysis STEP 5 Analyze the result to assist in decision making Figure 1. The Procedure for Evaluation of Efficiency using DEA Data Collection For each component of the different company alternatives, all pertaining information on potential factors (e.g., costs, revenues etc.,) should be collected and recorded. The quality of the collected data will in part determine the quality of the eventual decision Factor Selection The main select the factors which could directly affect the decision. The set of factors should be limited in size, and accordingly, only the main factors which significantly affect the decision should be included in the set. Too many indices will cause the result of losing discriminatory power [19] Model Selection According to the property of the factors and the decision purpose, we can select the most appropriate DEA model as our approach. Model types might change based on the calculation of the projection(ccr, ADD) or problem/factor characteristic(such as inputoriented or output oriented models) may dictate the selection of the model Run the DEA Model There are several software packages for DEA calculations such as Frontier Analyst, DEAFrontier, DEA-solver, EMS, etc. In this study, DEA-solver is used Result Analysis When a same efficiency score is present in the results, a ranking of the results with a specific DEA ranking method such as cross-efficiency method or super efficiency method might be necessary. This ranking will break the same. During the analysis of the results, special attention should be directed to the meaning of the model parameters to obtain the most appropriate result Preceding Research Several studies using DEA have been proposed on the analysis of efficiency of domestic IT industry. Oh[19] analyzed the efficiency of 48 telecommunication-related companies. He selected employee, building area and equity as input factors and operating profit, goal attainment, and management members as output factors. 208 Copyright c 2015 SERSC

5 Ham[10] performed an analysis of the management efficiency level of the Korean telecommunications companies through a comparison with some of the world s major telecommunications companies. Kim[12] studied the efficiency of the wired and wireless communications companies like SKT, KTF. He selected investment and marketing cost as input factors and revenue as output factor. Kim[14] analyzed the efficiency of top 17 companies of SI sector in KOSDA. He selected total assets, employee, and training cost as input factors and operating profit as output factor. Kim and Kang[13] studied the efficiency of domestic telecommunications companies using DEA technique. They presented projection value so that inefficient companies benchmark efficient companies. They selected assets, capital, and employee as input factors and net income, operating income, ordinary income, and sales as output factors. 3. Analysis of Efficiency In this section, we apply the five step method presented above to evaluate the efficiency of company. There are 41 IT services companies listed on the domestic stock market: 8 in KOSPI and 33 in KOSDAQ. We selected 32 KOSDAQ-listed domestic companies as DMUs. The input and output factors selected from previous efficiency studies for IT industry, considering the characteristics of KOSDAQ. As a result, we selected employee, asset, capital and R&D cost as input factors and sales, operating profit and net profit as output factors Data Collection and Factor Selection There are 41 IT services companies listed on the domestic stock market: 8 in KOSPI and 33 in KOSDAQ. Thirty two domestic companies except a foreign company in KOSDAQ are chosen as the target of analysis. There must be a constant correlation between the factors in order to have validity between the input and output factors of DEA model. In this study, we chose two input and output factors respectively among addressed in previous studies, which are highly correlated. Their correlations are shown in the Table 1. Regression analysis was performed to see the influences between input factors and output factors. Asset and capital as input factors have high correlation coefficient in the correlation analysis. So capital was excluded according to the multicollinearity principle. Regression analysis was conducted for the impact of asset on sales and operating profit. Their regression analysis is shown in the Tables 2 and 3. Table 1. Correlation between Input and Output Factors Asset Capital Sales Operating profit Asset (**) 0.73(**) 0.6(**) Capital (**) 0.78(**) Sales (**) Operating profit 1 **Correlation is significant at the 0.01 level (2-tailed) In Table 2, the impact of asset on sales is significant. And in Table 3, the impact of asset on operating profit is significant too. After all assets appeared to affect sales and operating profit. Then the choice of factors that seem reasonable Copyright c 2015 SERSC 209

6 Table 2. Regression Analysis between Asset and Sales Standardized Coefficients Std. Error t Sig. 95% Confidence Interval for B Lower Upper Bound Bound (Constant) (I)Asset E Dependent Variable: (O)Sales Table 3. Regression Analysis between Asset and Operation Profit Standardized Coefficients Std. Error t Sig. 95% Confidence Interval for B Lower Upper Bound Bound (Constant) (I)Asset Dependent Variable: (O)Operation profit Meanwhile, it is very important to select the appropriate number of DMU and factors in the DEA model. It is because, if the number of DMU and selection of factor are not appropriate, the result of efficiency analysis can be quite wrong. That is, when the number of input and output factors is excessively large compared to the number of DMU, the ability to distinguish between efficient DMU and inefficient DMU becomes poor. Then, the reliability of the results becomes low. There have been studies to determine the number of input and output factors, in order to avoid excessive measure of efficiency in the application of DEA model (see Table 4), where m is the number of input elements, s is the number of output elements, and n represents the number of the DMU to be analyzed. Table 4. Relationship between the Number of Appropriate DMU and the Number of Input/Output Factors Researchers Banker, Charnes, Cooper(1984) Boussofiane, Dyson, Thanassoulis(1991) Fitzsimmons(1994) the number of the appropriate DMU(n) and the number of input(m) / output factor(s) n > 3(m + s) n > 3(m s) n > 2(m + s) Taking account of three criteria in the previous studies presented above, the number of input / output factors and the number of DMU selected in this study are considered reasonable. The data collected finally is shown in Table 5. Table 5. Final Data Collection (unit: one hundred million won) DMU Asset Capital Sales Operation profit D D , D D D D6 1, ,851 6 D , Copyright c 2015 SERSC

7 3.2. Efficiency Analysis D D D10 5,253 2,153 4, D D D D14 2, , D D , D ,079-8 D18 4, D , D D21 1, , D D D D25 1, , D D D28 2, , D D D31 1, , D Max Min Average SD As the focus is on maximizing multiple outputs, this paper chooses the output-oriented CCR and BCC model [6] CCR-O Model DEA presents the DMU evaluated effectively, called 'reference set', to the inefficient DMU for benchmarking efficient DMU. Inefficient DMUs can set the target to be an efficient DMU through this 'reference set'. This is because, as the 'reference set' is a model which inefficient DMU company can refer to, both referred DMU and referring DMU consist of groups of homogeneity in the input and output structures. And the DMUs with higher weight λ value in the 'reference set' appear to have higher similarity of the input and output structures. As shown in Table 6., which is the analysis of the CCR-O model, there are 4 DMUs, D7, D15, D17, and D19, whose efficiency score is effectively analyzed as 1. The remaining 28 DMUs are found to be less efficient than these. Table 6. The Efficiency Score and Rank of CCR-O Model DMU Score Rank Reference(Lambda) # of reference D D D D D D D D D D D D D D D D D D7 1 1 D D D D Copyright c 2015 SERSC 211

8 D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D Reference count for D7 is 15 times, D15 13 times, D17 14 times, and D19 22 times. Reference count of D19 is relatively very high. Reference count is the number used for different evaluation of DMU analysis, which means that the higher the value is, the more often it is used for the other evaluation of DMU analysis. When we take a look at the relationship between an inefficient DMU and the reference group as for an example, D1 must refer to D17 in order to improve efficiency and the level of efficiency of D1 is found to be 11.22% of D17. This is similar for the other inefficient DMUs. DEA presents a target value to be an efficient DMU to inefficient DMU. This allows improving the efficiency of inefficient DMU more easily to be efficient DMU by telling the target value of what, how, and how much to be improved. It can be calculated using the efficiency ranking, the reference set, and weight λ in Table 6. DEA calculates a target value compared to the efficient DMU to improve the inefficient DMU. The target value is referred to as projection values shown in the Table 7. Table 7. Projection of CCR-O Model Asset Capital Sales Operation profit DMU Score Data Proj. Data Proj. Data Proj. Data Proj. D D D D D D D D D D D D Copyright c 2015 SERSC

9 D D D D D D D D D D D D D D D D D D D D BCC-O As analyzed by CCR-O model, D7, D15, D17, and D19 are found to be efficient. The remaining 28 DMU is found to be inefficient. Then, we applied the BCC-O model to analyze whether the causes of 28 inefficient DMUs were due to scale factor or pure technical factors. As shown in Table 8., the result of applying BCC-O model, DMUs whose efficiency score is 1 are D2, D6, D7, D8, D10, D13, D15, D17, D19, D28 and D DMUs are found to be efficient, and the remaining 21 DMUs inefficient. Table 8. The Afficiency Score and Rank of BCC-O Model DMU Score Rank Reference(Lambda) # of reference D D D D2 1 1 D2 1 3 D D D D D D D D D D D D D6 1 1 D D7 1 1 D7 1 5 D8 1 1 D8 1 4 D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D Copyright c 2015 SERSC 213

10 D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D D The average efficiency score is , higher than , the average of the CCR-O model. Reference count of D2 is 3 times, D6 10 times, D7 5 times, D8 4 times, D10 5 times, D13 8 times, D15 8 times, D17 14 times, D19 11 times, D28 2 times, and D29 3 times. D17, D19, ad D6 have been referenced relatively more. When we look at the relationship between the inefficient DMUs and the reference group, D1 needs to refer to D6 in order to improve efficiency. The efficiency of D1 was found to be 16.41% of the efficiency of D17. Other inefficient DMUs are same. In order to improve the inefficient DMU, projection obtained by comparing with the effective DMU is shown in Table 9. DMU Score Data Table 9. Projection of BCC-O Model Asset Capital Sales Operation profit Proj. Data Proj. D D D D D D D D D D D D D D D D D D D D D D D D D D D D Data Proj. Data Proj. 214 Copyright c 2015 SERSC

11 D D D D Efficiency of Scale and Return to Scale In this study, the scale efficiency(se) approach, which is based on the CCR and BCC scores, is applied to measure the relative efficiency of DMUs. Let the CCR and BCC scores of a DMU be θ CCR and θ BCC respectively, the SE is defined by the formular (8). SE is not greater than the maximum efficiency score one. SE= θ CCR θ (8) BCC Return to Scale(RTS) has shown constant as CRS(Constant Returns to Scale) state if n the CCR score and BCC score values are equal or j=1 λ j = 1. RTS has shown increasing as IRS(Increasing Returns to Scale) state if the CCR score and BCC score not n equal and j=1 λ j < 1. And RTS has shown decreasing as DRS(Decreasing Returns to n Scale) state if the CCR score and BCC score not equal and j=1 λ j > 1. If an inefficient DMU is the IRS state, the marginal revenue increases with the increase in input factor. That is, it is possible to improve the efficiency through an increase in scale. If an inefficient DMU is the DRS state, the marginal revenue decreases with increase in the input factors. In other words, improvement of efficiency can be achieved through a reduction in scale. Table 10. Scale Efficiency and Return to Scale DMU CCR Score BCC Score SE RTS D DRS D DRS D IRS D DRS D IRS D DRS D CRS D IRS D IRS D DRS D IRS D IRS D IRS D DRS D CRS D DRS D CRS D DRS D CRS D IRS D DRS D DRS D IRS D IRS D DRS D IRS D DRS D DRS D IRS Copyright c 2015 SERSC 215

12 D DRS D DRS D DRS Average *SE: Scale efficiency *RTS: Return to cale Super Efficiency As the result of CCR-O model analysis, 4 DMUs (D7, D15, D17, and D19) were shown efficient, and as the result of BCC-O model analysis, 11 DMUs (D2, D6, D7, D8, D10, D13, D15, D17, D19, D28, and D29) were found efficient. To evaluate both 4 DMUs and 11 DMUs in each analysis have the same efficiency, the result of applying the Super efficiency model is shown in the table 4 below. The order of efficiency ranking in the CCR-O model is D7, D17, D15, and D19, and in the BCC-O models D10, D17, D19, D15, D6, D29, D28, D2, D8, D7, and D13. Rank Table 11. Rank of Super Efficiency CCR BCC DMU Score DMU Score 1 D D D D D D D D D D D D D D D Conclusions and Further Researches This study analyzed 32 companies listed on the KOSDAQ by 2013 in order to measure the relative efficiency of the domestic IT service companies using a DEA model. We analyzed the technical efficiency through the CCR-O model to measure how much we could increase the output factor in the state that the input factor was fixed, and separated pure technical efficiency from scale efficiency through the BCC-O model analysis. The results show that four companies were efficient in the CCR-O model, and 11 companies were relatively efficient in the BCC-O model. The result of applying the CCR-O model shows that the average value of technical efficiency of business is 67.05% and, in other words, inefficiency of 32.95% exists. Measured by the BCC-O model, pure technical efficiency (PTE) is 77.35% and inefficiency of 22.65% exists. And SE, technical efficiency value divided by the value of pure technical efficiency is found to be 86.15%. This can be seen as the cause of inefficiency of business may be due to rather PTE than TE. In other words, the technical inefficiency appears to show because companies produce outputs less compared to the current level of inputs. This means that, in order to be efficient, companies need to reduce over-invested inputs and benchmark efficient enterprises to increase insufficient outputs. We performed the rank analysis by applying the super efficiency method to distinguish 216 Copyright c 2015 SERSC

13 between the DMUs with the same efficiency value. Implications of the present findings are as follows. First, the efficiency score measured by the DEA model is an indicator of the relative efficiency among companies, with which we can establish a scheme for improving the efficiency of companies analyzed as inefficient, and present a method of operating the available resources most efficiently. Second, using the CCR model and BCC model, this study presents whether the cause of inefficiencies is due to the inefficiency of the scale or the pure technical inefficiency (inefficiency in production). As a result, it is suggested that because overall, inefficiency of company is due to the pure technical inefficiency, it needs to be reduced through works such as cost savings. This study has the following limitations. First, because we targeted the KOSDAQ companies, it is necessary to extend our analysis to KOSPI companies to include all kinds of companies. Second, DEA model is the relative efficiency measure among similar subjects of evaluation. So it is difficult to measure the absolute efficiency. Third, input factors and output factors were reasonably selected. However, if we analyze with other factors, the result may be different. Acknowledgement This research was supported by Seokyeong University in References [1] N. Adler and B. Golany, Including Principal Component Weights to Improve Discrimination in Data Envelopment Analysis, Journal of Operation Research Society, vol. 53, no. 9, (2002), pp [2] W. R. Bitman and N. Sharif, A Conceptual Framework for Ranking R&D Projects, IEEE Transactions on Engineering Management, vol. 55, no. 2, (2002), pp [3] R. D. Banker and W. W. Cooper, Some Models for Estimating Technical and Scale Inefficiencies in Data Envelopment Analysis, Management Science, vol. 39, no. 9, (1984), pp [4] A. Boussofiand, R. D. Dyson and E. Thanassoulis, Applied data envelopment analysis, European Journal of Operational Research, vol. 52, no. 1, (1991), pp [5] A. Charnes, W. W. Cooper and E. Rhodes, Measuring the Efficiency Government-Sponsored R&D Projects: A Three Stage Approach, Evaluation and Program Planning, vol. 32, no. 2, (2009), pp [6] A. Charnes, W. W. Cooper and E. Rhodes, Measuring the Efficiency of Decision Making Units, European Journal of Operation Research, vol. 2, no. 6, (1978), pp [7] W. W. Cooper, L. M. Seiford and K. Tone, Data Envelopment Analysis: A Comprehensive Text with Models, Applications, References and DEA-Solver Software, Second Edition, Springer, New York, (2006). [8] J. A. Fitzsimmons and M. J. Fitzsimmons, Service management for Competitive Advantage, McGrow-Hill Inc., (1994). [9] H. W. Goh, Evaluating Efficiency of Financial Performance for KOSDAQ corporations, Proceedings International Workshop Business 2015, SERSC, (2015), pp [10] S. C. Ham, Comparative Analysis of Efficiency for Korea Telecom Companies and Major Global Telecommunications Companies, Information and Communication Policy Research, vol. 13, no. 4, (2006), pp [11] J. S. Hong, C. J. Yang and H. Y. Lee, Comparative Evaluation of Efficiency of the Korean IT sectors: A Data Envelopment Analysis Approach, Journal of the Korea Management Engineers Society, vol. 17, no. 1, (2012) March, pp [12] H. Kim, A study on Korean Telecommunications Operators Expenditure Efficiency: DEA Approach, Master s Thesis, Information and Communications University, (2006). [13] J. K. Kim and D. Y. Kang, Efficiency Analysis of Domestic Apartment Construction Company Using DEA Model, Journal of Korea Contents Association, vol. 8, no. 7, (2008), pp [14] S. H. Kim, An Analysis of management efficiency of Korean IT Service Firms, Master s Thesis, Information and Communications University, (2007). [15] H. Lee, Y. Park and H. Choi, Comparative Evaluation of Performance of National R&D Programs with Heterogeneous objectives: A DEA approach, European Journal of Operational Research, vol. 196, no. 3, (2009), pp Copyright c 2015 SERSC 217

14 [16] S. Lee, G. Mogi, M. Koike, K. S. Hui and J. Kim, DEA scale efficiency approach on measuring the relative efficiency values of energy technologies in the sector of mid-term strategic energy technology development plan, KORMS/KIIE Proceeding, vol. 5, (2009), pp [17] J. S. Lee, Using Data Envelopment Analysis and Decision Trees for Efficiency and Recommendation of B2C Controls, Decision Support Systems, vol. 49, (2010), pp [18] C. Y. Lin, G. E. Okudan, An Exploration on the Use of Data Envelopment Analysis for Product Line Selection, Industrial Engineering & Management Systems, vol. 8, no. 1, (2009) March, pp [19] H. J. Oh, Study on Performance Evaluation of Information and Communication Enterprises by DEA Model, Ph.D Dissertation, Kyonggi University, (2002). [20] J. C. Paradi, S. Smith and C. Schaffnit-Chatterjee, Knowledge Worker Performance Analysis Using DEA: An Application to Engineering esigh Team at Bell Canada, IEEE Transactions on Engineering Management, vol. 49, no. 1, pp , (2002). [21] J. M. Wagner and D. G. Shimshak, Stepwise Selection of Variables in Data Envelopment Analysis: Procedure and Managerial Perspectives, European Journal of Operational Research, vol. 180, (2007), pp Author HyunWoo Goh, he received the B.S., M.S. and Ph.D. degrees from the Dept. of Industrial Engineering, Hanyang University, Seoul, Korea, in 1987, 1989 and 1995, respectively. He was a General Director of Korea Industrial Management System Society from Currently, he is a Professor of Dept. of Industrial Management System Engineering, Seokyeong University, Seoul, Korea. His research areas include Data Envelopment Analysis (DEA), Supply Chain Management (SCM) and system dynamics. 218 Copyright c 2015 SERSC