Using dea models to evaluate the efficiency of lienvietpostbank in hanoi

Nội dung text: Using dea models to evaluate the efficiency of lienvietpostbank in hanoi

1. USING DEA MODELS TO EVALUATE THE EFFICIENCY OF LIENVIETPOSTBANK IN HANOI Luong Tuan Anh Tran Van Trang Vietnam University of Commerce Abstract The DEA (Data Envelopment Analysis), a popular efficiency benchmarking method for bank’s branches in the world is the model chosen for this research because it does solve the difficulties of evaluating the effectiveness of a business unit using several to many inputs and outputs. It has proven to be a practical method since it does not only provide a relative efficiency score for a set of selected decisions making units (DMUs), it also give clear information on the changes needed to be made on the inputs and outputs in order to improve the productivity. The main objective of this research is to comply with the need of LVB and find a scientific method that would help the bank improve their branches' efficiency. In order to reach this global goal, we will have to answer two research questions: Firstly, what are the suitable criteria for the evaluation of LVB’s branches? And secondly how to access the efficiency of bank branches? Key words:DEA model, efficiency, bank, LienVietPostbank 277
2. 1. INTRODUCTION Banks are very important entities in the economy of every country in the world. Their main activities, includingpayment services, receiving deposit and giving out loans serve to help organizing, facilitating business and commercial transactions, as well as develop markets and economic growth. In addition, it also operates as pillars in directing and managing disperses funds toward production units and regulates the cash flow domestically or between Vietnam and foreign countries. Moreover, they participate actively in the inflation control.Banks are arguably even more important in Vietnam since the State Bank of Vietnam is under thedirect control of the government.Furthermore, four biggest Vietnamese banks, namely Vietcombank, Vietinbank, Agribank and BIDV, have more than 50% of their shares that belong to the state. In this situation, the banks are required to become a key mean to regulate the monetary flow and apply the economic and monetary policy for the Vietnamese government. In fact, the banking sector has contributed a large part in the fast developing period of Vietnamese economy after the crisis in 2009. Unfortunately, the speculative bubble on both the Vietnamese stock exchange and the real estate market exploded in 2011 and many banks were on the verge of bankruptcy since the safety and efficiency control were neglected. In fact, most of the banks took profits of the fast growing economy from 2009 to 2011 in order to expand quickly and recklessly giving out loans without much control aiming for fast but non-durable development. As a solution to avoid any similar crisis and improve the banking sector, Vietnamese banks should focus more on their business performance. In fact, it is suggested by Vu (2013) that the State Bank of Vietnam (SBV) should first try to improve the management and operation of Commercial Banks. Lienvietpostbank (LVB) is one of the rare banks succeeding to avoid grave consequences of the crisis in 2011 while other commercial banks such as Habubank had to merge with SHB to avoid bankruptcy. Meanwhile, the bank is still affected by the stagnation of the Vietnamese economy and the leaders recognized the fact that in this situation, improving the efficiency of the branches is a good method to ensure further development. It must also be noted that Lienvietpostbank is the result of the merger of Lien Viet bank and the Postal Credit Funds in 2010. For this specific reason, its enlarged network may be heterogonous in term of performance; hence they need to evaluate the efficiency of the branches. Moreover, according to Mr. NGUYEN, Vice President of the Board of Directors of LienVietPostBank (LVB): “Our Bank is not in a dire situation but it’s true that the economic regression is affecting us and since the opening of new agencies is limited by the government we cannot focus on an geographical extension strategy so we are actually very interested in perfecting the effectiveness of individual branches meaning to maximize the productivity with the accessible resources.”Hence, it is clear that a research on scientific method to benchmark and finding ways to improve the efficiency of the branches at LVB is important for the bank. The main objective of this research is to comply with the need of LVB and find a scientific method that would help the bank improve their branches' efficiency. In order to reach this global goal, we will have to answer two research questions: Firstly, what are the suitable criteria for the evaluation of LVB’s branches? And secondly how to access the efficiency of bank branches? 278
3. 2. LITTERATURE REVIEW AND RESEARCH MODEL 2.1 Notion of “efficiency” The first notion to access in this work is “efficiency” or more precisely “operational efficiency”. For several sources, “operational efficiency” is also considered as the productivity of a business entity. Coelli, Rao, O’Donnell and Battese (2005) stated that an informal definition for efficiency is productivity. In fact it refers to the ratio between outputs produced by the inputs consumed. However, they are still different as a “technically efficient” firm which is meant to be able to produce the maximum quantity of outputs for a certain number of inputs while its productivity can still improve by changing the inputs to a different level (economy of scale). In fact, if the maximum production for 1 input unit equal 3 output units, a firm is efficient if it has this setup. Nevertheless, if thanks to the economy of scale, by using 5 input units it can produce 25 output units. The firm can still improve its productivity. The Organization for Economic Co-operation and Development (OECD) gives a similar definition for the term: “Efficiency is a measure of how economically resources/inputs (funds, expertise, time, etc.) are converted to the results.” (German Federal Ministry for Economic Cooperation and Development [FMECD], 2010). It is noted in this definition that the inputs can be of multiple types, whether financial, material or even the time. The same can, of course, be said about the outputs since “result” has large meaning. According to the authors of an organization evaluation method called Data Envelopment Analysis (DEA), the “efficiency” of a unit is the actual outputs that it can produce while exploiting some precise inputs (Charnes, Cooper and Rodez (1978). Since this is a common definition largely accepted for entities of businesses, companies, organizations, the definition of the term “efficiency” in this article would also be: = 𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐 Papers on the efficiency of 𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆banks or banks' branches also use this generic definition as the factors chosen as inputs and outputs vary a lot𝒊𝒊𝒊𝒊𝒊𝒊𝒊𝒊𝒊𝒊 in these researches (Shahroudi&Assimi (2010), Hassan &Ader (2009), Lin, Lee & Chiu (2009), Wu, Yang & Liang (2006), Yavas& Fisher (2005)) hence in this paper the above definition is chosen for the notion of “efficiency”. 2.2 Firm’s efficiency evaluation method In order to evaluate the efficiency of firms, we have accessed four methods presented byCoelli, Rao, O’Donnell and Battese (2005): the Least-squares Economics Production models (LS); the Total Factor of Productivity (TFP), the Data Envelopment Analysis (DEA) and Stochastic Frontiers (SF). They differ from each other in several ways as it is presented in the table 1 below. 279
4. Table 1 Summary of the properties of the four methods Attributes LS TFP DEA SF Parametric method yes no no yes Account for noise yes no no yes Can be used to measures Technical efficiency no no yes yes Allocative efficiency yes no yes yes Technical change yes no yes yes Scale effects yes no yes yes TFP change yes yes yes yes Data used Cross sectional yes yes yes yes Time series yes yes no no Panel yes yes yes yes Basic method requires data on: Input quantities yes yes yes yes Outputs quantities yes yes yes yes Inputs prices no yes no no Outputs prices no yes no no Source: Coelli , Rao, O’Donnell, Battese, (2005, p 312) The first 2 methods can’t be used to measure the technical efficiency since they assume that all the firms researched are already efficient. The goal of this article is to measure the efficiency of bank branches and since such condition is unlikely realistic, we will not consider them any further. The DEA and SF methods are usually used for inefficiency research. Practically, they have the same characteristics as the first 2 methods presented in table 1. The advantage of the SF over DEA is that it does accounting for noises, which means it can recognize random errors on the sample. Nevertheless, this comes at the disadvantage of being a parametric method. A parametric method must assume that the data come from a type of probability distribution, hence we have chosen the DEA to avoid any mistakes in case the conditions set on the data are not verified. Taking into account the fact that DEA is the prominent method used for efficiency benchmarking in the banking field, the use of the third method here (DEA) would be better in this case. The creators of the DEA model (Cooper & Rhodes, 1978), in their originated text, described DEA as a “mathematical programming model applied to observational data [that] provides a new way of obtaining empirical estimates of relations - such as the production functions and/or efficient production possibility surfaces – that are cornerstones of modern economics”. Formally, DEA is a method that is meant to evaluate the efficiency of Decision Making Units (DMUs) based on a set of outputs produced by using a set of inputs. This method delivers not strict but relative efficiency of the DMUs after processing quantitative data. Since its beginning more than 30 years ago, DEA has indeed become a very popular 280
5. method for efficiency benchmarking in the banking sector across the world. In fact, researches on banks have been done using DEA in Canada, Belgium, China, Thailand, USA hence it would be highly probable that such a method is adapted to this case of study. 2.3 Research model for Lienvietpostbank Based on the definition of efficiency and the literatures using the DEA method in the banking sector, we have built the following research model for the LVB: Inputs (Resources) LVB’s Branch (DMUs) Outputs Outputs (Services) (Revenues) Figure 1: Efficiency research model based on in-and-outputs of Bank Branch Source : ModifiedafterGemmel et al. (2002), Paradi et al. (2011), Yavas et al. (2005) In this model, the branches of LVB will use a set of inputs (both financial and non- financial) to produce two types of outputs: revenues and services. Based on all of the mentioned literatures of the DEA model, the inputs are clarified to elements subjected to minimize and outputs to those we would try to maximize in order to reach efficiency at a DMU in this model. Subjected to the DEA process by Charnes, Cooper, Rhodes (1978) the efficiency of a DMU (LVB’s branch) is the ratio between the sum of weighted outputs and the sum of weighted inputs. As such, the efficiency score of r-th DMU is given by the following formula: 281
6. =1 × = 𝑚𝑚 × ∑𝑗𝑗=1 𝑣𝑣𝑗𝑗 𝑦𝑦𝑗𝑗𝑗𝑗 𝜃𝜃𝑗𝑗 𝑙𝑙 - is the efficiency score of the∑ DMU𝑖𝑖 𝑢𝑢𝑖𝑖 𝑥𝑥𝑖𝑖𝑖𝑖 - is the weight of the i-th input 𝜃𝜃 - is the value of the i-th input used by branch i 𝑢𝑢𝑖𝑖 - is the weight of the i-th output 𝑥𝑥𝑖𝑖𝑖𝑖 - is the value of the i-th output used by branch j 𝑣𝑣𝑗𝑗 - is the number of inputs 𝑦𝑦𝑗𝑗𝑗𝑗 - is the number of outputs 𝑙𝑙 The maximum𝑚𝑚 value for the efficiency score in this case is 1 and all of the weights must be higher than 0 to avoid any input or output to be neglected. 2.4 Possible inputs and outputs criteria The selection of inputs and outputs will be done based on the list taken from several researches made on the efficiency of bank branches. The list is presented in the following table: Table 2: Researches using DEA in banking industry with their inputs and outputs Authors Inputs Outputs Shahroudi et al. (2010)) Staff’s cost Resources Non staff’s cost Expenses Remained owes Commission Income Profit Staub et al. (2009) Financial credit Deposits Interest expense Loans Capital Investments Staffs Hassan et al. (2009) Fixed assets Total loans Total investments Other incomes fromassets Lin et al. (2009) Number of staff Loan operating amount Interest expense Earnings Deposit operatingamount Operating revenue Current depositoperating Interest revenue amount Mokhtar et al. (2008) Total deposits Total earning assets Total overhead expenses Bdour et al. (2008) Staffs Total deposits Total assets Net direct credits Total operating expenses Operating income Kumar and Gulati Physical assets Net interest income (2008) Labor Non- interest income 282
7. Loanable funds Mostafa (2007) Assets Net profits Capital ROA ROE Sufian (2007) Total deposits Total loans Fixed assets Other incomes Ramanathan (2006) Fixed assets Loans Deposits Other incomes Short-term deposits ROA Personnel expenses Wu (2006) Labor Deposits General expenses Incomes Loans Yavas and Fisher (2005) Number of employees : FTE Retail deposits Lobby hours per week Small business deposits Number of ATMs Safety deposit boxes Average waiting time in line for service Source :Modified after Yavas, & Fisher. (2005), Shahroudi et al. (2010) The criteria that will be selected from this list must satisfy 5 conditions in order to assure the reliability of this research: 1. Consistent with previous works (Yavas et al., 2005). In fact, selecting inputs and outputs which are consistent with the previous research will help narrow down the possible indicators. 2. Vary from a branch to another (Yavas et al., 2005). This condition would permit me to investigate different level of efficiency between branches and succeed in finding interesting results. 3. Kept to a low number (Gemmel et al., 2011) so the data processing is feasible in for myself and also to avoid the overlapping of indicators such as the number of employees and the wage paid. 4.Approved by the bank managers (Gemmel et al., 2011, Yavas et al., 2005, Shahroudi et al., 2010) to make sure that the indicators selected are ones that fit in with the policy of the banks. This work will then be realistic and beneficial for LienVietPostBank. 5.Data from all branches available (Yavas et al., 2005) since one missing data in a single DMU would make the data processing impossible since DEA method calculate the relative efficiency of a branch in comparison with the whole group. 283
8. 3. METHODOLOGY This research is in fact divided into 3 steps: the first one is the selection of inputs and outputs, the second one is the collection of data at LVB branches and the last one is to process using the DEA model in order to benchmark the LVB’s branches. The selection of inputs and outputs will be done using interviews with the bank’s managers. This is a crucial step that must be conducted carefullysince we can select from a large variety of factors and mistakes would lead to irrelevant results, failing the research in the same process. In fact, if we just choose the factors without taking into account the opinion of LVB’s leaders we may fail to get the suitable inputs and outputs from the bank since these factors (the resources and the results) also depend on the policy and decisions of the chiefs. Additionally, the interview will permit to make sure that the data of selected factors are available at all the branches of the research. After reflecting on the importance of the interviews to be conducted, we have chosen the sample of this research: 20 branches of LVB present in Hanoi, they are chosen to facilitate the communication since we will interview the directors of those branches. 3 other interviews with top level managers of LVB will be conducted for deeper and more global opinion on the selection of inputs and outputs. These interviews will be conducted with a Vice President of the Board of Directors and 2 Deputy CEO responsible for the Human resources Division and the Product Division. Once the inputs and outputs are selected, data from the DMUs will be collected in the internal reports of 2014, the latest year of activity. The data collected will be secondary data so we must proceed carefully as these data were not collected for the purpose of this research. The documents that we will collect the data are the balance sheet (for capital, total assets ), income statement (income, profit, operation revenue, interest revenue ), statistics report of the HR division (numbers of staffs, labor ), activities reports (number of transactions, number of clients, number of accounts etc.) Finally, the collected data will be processed using DEA software designed by Doctor Zhu. This final step will provide the efficiency score for each of the 20 branches and the results that will be used to discuss the DEA method. 4. RESEARCH FINDINGS 4.1 Inputs and outputs selected for LVB’s branches The results of the interviews of LVB’s managers have permitted to choose 6 factors, satisfying the conditions set in point 2.4. They are presented in the following table: Table 3. Units and sources of the inputs and outputs No Inputs Outputs Unit Sources . 1 Staffs Number of employees HR Division 2 Total cost Vietnamese Dong (VND) Product Division 3 ACPD Number of clients Product Division 4 Interest income VND Product Division 284
9. 5 Non-interest income VND Product Division 6 Profits before taxes VND Source : Synthesized from the interviews with managers of LVB As such, the final research model of this paper is as in the next figure: OUTPUTS (Revenues) OUTPUTS (Services) Interest income (VND) - - Average number of clients - Non-interest income (VND) (number) - Profits before taxes (VND) LVB’s Branch (DMUs) INPUTS (Resources) - Number of staffs (number) - Total cost (VND) ModifiedafterGemmel et al. (2002), Paradi et al. (2011), Yavas et al. (2005) Figure 2: Efficiency research model based on in-and-outputs for LVB’s branches In this model, the branches of LVB will use a set of inputs (both financial and non- financial) to produce two types of outputs: revenues and services. Based in all of the mentioned literatures of the DEA model, the inputs are clarified to elements subjected to minimize and outputs to those we would try to maximize in order to reach efficiency at a DMU in this model. Subjected to the DEA process by Charnes, Cooper, Rhodes (1978) the efficiency of a DMU (LVB’s branch) is the ratio between the sum of weighted outputs and the sum of weighted inputs. 285
10. The maximum value for the efficiency score in this case is 1 and all of the weight must be higher than 0 to avoid any input or output to be neglected. The DEA method uses a mathematical method: linear programming to determine the efficiency score. The linear programming is, in fact, set to find a common set of weights for all the selected branches with the optimization which is to minimize the inputs. 4.2 Efficiency evaluation of LVB’s branches Processing the data of 20 branches provided the results presented in the table 4: Table 4: Efficiency score of LVB’s branches LVB’s Branch Efficiency score DEA ranking A 0.93245 8 B 1.00000 1 C 0.50188 15 D 1.00000 1 E 1.00000 1 F 1.00000 1 G 0.37279 20 H 0.73319 11 I 0.94623 7 J 1.00000 1 K 0.77979 10 L 0.61869 14 M 0.42921 19 N 0.97252 6 O 0.78384 9 P 0.47521 16 Q 0.43667 18 R 0.66826 12 S 0.66464 13 T 0.46054 17 These results show that there are 5 branches judged to be efficient among the ones tested. For the rest, the efficiency score ranged widely and the lowest is 0.37279 which means the branch G is only 37% efficient. 286
11. 5. CONCLUSIONS AND SUGGESTIONS 5.1 Discussion on the results The result from the data processing was obtained as planed and we have obtained valid efficiency scores for each of the 20 selected branches. These scoresprovided by the DEA method have shown us that LVB’s branches are indeed heterogeneous in terms of effectiveness. Fiveout of the branches were judged to be efficient for the selected set of inputs and outputs while some others (5 branches) have efficiency score lower than 0.5 (less than 50%). This fact confirmed the heterogeneity of the branches at LVB and that the bank can still improve the performance of their offices by a large margin. We should also note that since the result of this method are only relative efficiency scores, the five branches which have the grade of 1 are only efficient in comparison with 15 other units in this study and for the selected inputs and outputs only. In fact, any change in sample or inputs/ outputs will change the result of the research. Consequently, DEA method should be the most suitable for specific research with clear objectives. 5.2 Recommendations This paper has proven that the DEA model is a suitable model for the evaluation of bank branches, especially for branches in the same banks since it offers a relative efficiency score. Furthermore, the flexibility in terms of inputs and outputs selection should permit the bank to freely conduct research for different objectives in order to evaluate their offices. We would like to recommend LVB or any other bank to implement this method to evaluate and manage the performance of their branches. 5.3 Limitations and further research First and foremost, the importance of inputs and outputs selection which has to be done by interviews has limited greatly the number of units in the sample. Since researches using this method have also been done with large sample (more than 1000 branches) abroad, we also suggest further study with larger sample in Vietnam. Another limitation of this research comes from a characteristic of DEA method, relative efficiency does, indeed, show the branches which are already efficient. Based on this model, it would mean that it will not possible to improve the efficiency of these units. We must also note that by pushing further in the DEA model, it will also provide slack value for each factor used in the research. These slack values are the amount of change on inputs and outputs that would make an inefficient branch efficient. For this reason, it is highly recommended that further research should be carried out using the same model but by pushing the result to the point where we can obtain slack values. 287
12. References - Charnes A., & Cooper W.W. & Rhodes E. (1978) “Measuring the efficiency of decision making units”, Eur. J. Opl. Res 2, 429-444. - Coelli , Timothy J., &Rao, DS. Prasada, & O’Donnell, Christopher J., &Battese, George E An Introduction to Efficiency and Productivity Analysis Second Edition. New York. Springer Science and Business Media Inc. - Gemmel P., Bourgonjon F., & Leuven V., “Efficiency benchmarking of bank branches for staffing purposes using DEA”, 2002 - Hassan T., Mohamad S., and Bader M., “Efficiency of conventional versus Islamic banks: evidence from the middle east”, International Journal of Islamic and middle eastern finance and management, 2 (1), 46-65, 2009. - Lin T., Lee C., and Chiu T., “Application of DEA in analyzing a bank’s operating performance”, Expert systems with applications, 36, pp. 8883-8891, 2009. - Paradi Joseph C., Yang Z., & Zhu H., “Two-stage evaluation of bank branch efficiency using data”, Omega, Volume 39, Issue 1, January 2011, Pages 99–109 - Shahroudi K., Assimi S., “Performance Evaluation of Banks Using DEA (Case study: GuilanSaderat Bank Branches)”, Iranian Journal of Optimization 2(2010) p.375-387 - VU Van Thuc, “Restructuration of the Vietnamese banking system”, Development & Integration scientific journalvol 10, may-june 2013, 17-31. - Wu D., Yang Z., and Liang L., “Using DEA-neural network approach to evaluate branch efficiency of a large Canadian bank”, Expert systems with applications, 31, 108-115, 2006. - Yavas, Burhan F. and Fisher, Dorothy M., “Performance Evaluation of Commercial Bank Branches Using Data Envelopment Analysis”, Journal of Business & Management. Apr2005, Vol. 11 Issue 2, p89-102. 14p. 288