Luận văn thạc sỹ quốc tế TECHNICAL EFFICIENCY FULL DISSERTATION

In two bank types, state-owned banks have tendency toobtain better technical efficiency scores than joint-stock ones.. The empirical findingsshow the positive and significant correlation

in Vietnamese commercial banks The findings show that Vietnamese commercialbanks has improving tendency in technical efficiency in the whole period, in spite

of some fluctuations In two bank types, state-owned banks have tendency toobtain better technical efficiency scores than joint-stock ones In addition, thetotal productivity for all banks is calculated by Malmquist index, which indicatesthat Vietnamese banks have been weak at technological advance and technicalefficiency For the purpose of exploring factors impacting level of technicalefficiency in banks, Tobit regression model is conducted The empirical findingsshow the positive and significant correlation between technical efficiency andsome factors of total asset size and equity ratio Meanwhile, technical efficiencyhas negative and significant relationship with loan ratio, non-performing debt,and government ownership

I would like to express my gratitude to my supervisor who is always beside meduring the period of doing this dissertation He shows me how to do an academicbusiness research with the application of theories that I learnt previously

I would like to thank my parents, my wife and my kids who always encourage

me to complete this master course

I would like to show appreciation to my manager and colleagues to assist mesome tasks when I had to struggle with assignments and dissertation in thiscourse

Finally, I honestly recognize the enthusiastic support of all lectures andadministrators in both Leeds Beckett University and Academy of Finance duringthe program

Contents

ACKNOWLEGEMENTS ii

LISTS OF FIGURES AND TABLES v

CHAPTER 1 INTRODUCTION 1

1.1 Research background 1

1.2 Research aims and objectives 3

1.3 Research questions 3

1.4 Research boundary 3

1.5 Significance of the research 3

1.6 Dissertation structure 4

CHAPTER 2 LITERATURE REVIEW 5

2.1 Bank efficiency 5

2.1.1 Definition of efficiency 5

2.1.2 Types of efficiency 5

2.1.3 Bank’s technical efficiency 6

2.2 Measurement of bank efficiency 8

2.3 Model of factors affecting bank efficiency 10

2.3.1 Factors affecting bank efficiency 11

2.3.1.4 Non-performing loans 13

2.3.2 Proposal research model 16

CHAPTER 3 METHODOLOGY 18

3.1 Data Envelopment Analysis (DEA) 18

3.1.2 Malmquist productivity index 22

3.1.3 Specification of inputs and outputs 23

3.2 Analysis of factors affecting banks’ technical efficiency 25

3.2.1 Tobit regression model 25

3.2.2 Variables 27

3.2 Sampling and data sources 27

3.2.1 Sampling plan 27

3.2.1 Data sources 28

3.3 Data analysis 29

3.4 Summary 29

CHAPTER 4 RESEARCH RESULTS 30

4.1 General information of Vietnam’s banking sector 30

4.2 DEA model 33

4.2.1 Descriptive statistics 33

4.2.2 DEA’s results 34

4.3 Malmquist productivity index 39

4.4 Tobit regression 40

4.4.1 Descriptive statistics 40

4.4.2 Tobit regression model 41

CHAPTER 5 DISCUSSION, CONCLUSION AND RECOMMENDATIONS 43

5.1 Discussion 43

5.2 Conclusions 45

5.3 Policy implications 46

5.4 Limitations 49

5.5 Suggestions for future research 49

APPENDICES 1

APPENDIX 1: LEARNING STATEMENT 1

REFFERENCES 7

LISTS OF FIGURES AND TABLES

Table 4.1 Bank types and quantity in Vietnamese banking sector

Table 4.2 Description of inputs and outputs used DEA model

Table 4.3 Average technical efficiency of two groups of ownership in Vietnamesecommercial banks from 2011-2015

Table 4.5 Variations in Total Factor Productivity of All Banks and Its specificelements per year

Table 4.5 Variations of total factor productivity and its specific elements inVietnamese banks in the period of 2011-2015

Table 4.6 Descriptive statistics of variables in Tobit model

Table 4.7 Result of Tobit regression of factors

CHAPTER 1 INTRODUCTION 1.1 Research background

Commercial banks play a vital part in the connection of economic constituents inthe economy and contribution to the development of financial sector Thefinancial system is considered as the backbone of the economy, so the bankingsector is represented for the main pillar of business The growth of nationaleconomy is largely reliant on the operations of banking sector

In Vietnam, banking sector changes strongly current years It grows rapidly interms of both asset and number Together with the operations of traditionalbanking, commercial banks provide a lot of new services and products to bringthe satisfaction to customers Otherwise, these commercial banks also enhancecontinuously facilities, infrastructure as well as employ advanced technologies intheir operations, which supports to enhance the technical effectiveness in eachbank as well as the commercial banking sector

Nevertheless, Vietnamese banking sector in general and commercial banks inparticular have their ineffectiveness in operations during the time of uncertainty.Remarkably, in the period of crisis and economic downturn since 2008,Vietnamese banking industry had experienced numerous difficulties and fell intothe overloaded situation During the time of 2011-2015, the increasing rate ofcredit in banks was just around 20%, the historic lowest level of Vietnamesebanking system Besides, the proportion of non-performing deb had anincreasing trend, reaching 51% in 2013 The absorption of capital has been verypoor in Vietnamese economy So, the major features of Vietnamese bankingindustry were weak growth of loan and deposit, high ratio of non-performing loanand revenues primarily dependent on credit activities Although the nationalgovernment and commercial banks issued several new policies to deal with thesematters, they seemed to be ineffective in encountering these challenges.Therefore, it is important that the authorities and bank’s top executives havelong-term perspectives and propose long-run strategies to manage bankoperations

To look for the reasons for this, it is necessary to analyze and understand theway of performance of commercial bank, especially the elements creating

dissimilarities in performance among commercial banks such as technicaleffectiveness It can say that technical effectiveness plays an important role inbringing the successfulness to commercial banks in the fierce competition ofbanking sector Consequently, if commercial banks want to improve theirperformance as well as the position in the banking sector, they should focus ontechnical effectiveness There are many authors studying this issue And thispaper is desired to provide a helpful viewpoint in relation to how commercialbanks’ technical effectiveness is and what elements influencing the banking

technical effectiveness are Therefore, the topic “evaluation of technical efficiency in Vietnamese commercial banks” is chosen

Understanding the considerable influence of banking system on the development

of national economy, numerous researches have been conducted on the theme ofbank efficiency Recently, several scholars in Vietnam have done studies related

to this topic For instance, Phu (2002) applies the functions of production andcost to evaluate the efficiency of commercial banks But the drawbacks of thisresearch are just to recognize costs and the function of estimating cost in order

to build the model So, the bank efficiency is not mentioned specifically Anh(2004) assess factors causing the inefficiency in Vietnamese rural anddevelopment banks; however the author just establishes the framework Thanh(2010) uses the model of data envelopment analysis to assess the inputefficiency in Vietnamese commercial banks The findings indicate that inputefficiency of banks is rather good; however it is possible to improve the efficiency

of inputs to higher level Truc and Danh (2012) explore that small banks sufferedbad consequences from global economic crisis rather than large and mediumones by using financial ratios to evaluate the technical efficiency in Vietnamesebanks Hung (2008) applies the quantitative approach to investigate theperformance of Vietnamese banking system Nevertheless, a drawback of thisstudy was the dataset before 2008 as many weaknesses of banking sectors wereshown after 2008 Additionally, the study concentrates on too much theexplanation of theories and the conclusions have not specified determinants ofbank efficiency So, it is essential to conduct researches about the bankefficiency in this situation

So, the thesis will bring a deep understanding about the technical efficiency ofbanks and factors impacting the efficiency of banks in the context of Vietnam

Evaluating technical efficiency of banks and exploring its determinants isimportant and useful for policy makers, banks’ managers and investors to makedecisions The thesis is also groundwork to fulfill the policy frames moreappropriately in the course of management of Vietnamese commercial banks

1.2 Research aims and objectives

The purposes of this paper are to discuss the evaluation of Vietnamesecommercial banks’ technical effectiveness in the period of 2011 -2015 and todetermine components influencing technical effectiveness

To achieve these purposes, the paper needs to solve the below objectives:

 Measuring Vietnamese commercial banks’ technical effectiveness in theperiod of 2011 -2015

 Comparing and analyzing the technical effectiveness’s results amongbanks

 Exploring components influencing technical effectiveness of Vietnamesecommercial banks

1.3 Research questions

With the purpose of achieving the research aim and objectives, the author willfind the answers for questions below:

 What levels of technical efficiency are in commercial banks of Vietnam?

 What factors affect the technical efficiency of Vietnamese commercialbanks?

1.4 Research boundary

The focus of the dissertation is to evaluate technical efficiency of banks andreveal factors affecting technical efficiency in banks with technical efficiency asthe key pointer to measure bank performance The sample set includes 10commercial banks operating in Vietnam Data are collected in the time between2011-2015

1.5 Significance of the research

This paper will enrich empirical findings about assessing technical efficiency ofVietnamese banks and understanding the impacting factors by showing practicalproofs in sample set Furthermore, the results of the dissertation will be valuable

source for reference in terms of officials and corporate managers who will gaininsight understandings about the technical efficiency in Vietnamese banks Fromthat, they can conduct suitable strategies and frameworks concerning to bankingcontrolling or evaluating benchmarks

Chapter 2: Literature Review

The theories relating to bank efficiency, measuring methods of efficiency anddeterminants of bank efficiency are reviewed critically by underpinning diversity

of researchers’ views in previous researches Chapter 2 is seen as the importantpillar for the dissertation

Chapter 3: Methodology

Chapter 3 will explain the research methods applied in this paper Theprocedures of selecting methodology and gathering data are portrayed Then, thehypotheses will be developed to examine in next chapter

Chapter 4: Empirical findings

The empirical findings will be shown in Chapter 4 The efficiency of banks will berevealed by comparison, analysis and interpretation in order to identify the mostefficient banks and the most influential factors on efficiency of banks

Chapter 5: Discussion of findings and conclusion

Chapter 5 will discuss differences or similarities between previous studies andobtained findings in this research Key notions and results are summarized andsynthesized From that, the author will make a conclusion about the wholeresearch: research aims and objectives, literature reviews, methodology andachieved results

CHAPTER 2 LITERATURE REVIEW 2.1 Bank efficiency

 Productive efficiency

Productive efficiency is the capability of a company to generate full output at thelowest expenses with the optimal usage of resources (Higson, 2011) Productiveefficiency is obtained in case that a company manufactures goods and services atthe lowest costs When a company reaches the productive efficiency, it isimpossible to manufacture more commodities without utilizing more resources orlowering the quality Thus, the company should find a balance in the exploitation

of resources, manufacturing capacity and commodities’ quality

 Technical efficiency

Technical efficiency is the effectiveness with a specific set of inputs exploited togenerate an output A firm obtains the technical efficiency if it manufactures themaximum volume of output from the lowest number of inputs, including humanresources, capital and technology (Kumbhakar and Lovell, 2003) A firm isregarded as technical inefficiency if it uses too many resources than required.Technical efficiency occurs when it is impossible to increase the output withoutimproving more inputs (Koopmans, 1951) The level of technical efficiency israted from 0 to 1

2.1.3 Bank’s technical efficiency

According to Bhattacharyya et al (1997), technical efficiency in banking sector isunderstood as the capacity to transfer input resources into diverse bankingproducts and services A commercial bank acquires technical efficiency when itsperformance is on the frontier line Under the assumption of constant returns-to-scale (CRS), technical efficiency is evaluated to recognize the inefficiencyresulting from factors of input, output or the size of banks

As stated by the theory of system, commercial banks are efficient when: (1) theycan change inputs into outputs or generate profits, or decrease expenditures toimprove the competitive advantage with other banks; (2) they can ensure theprospect of safe operations

The relationship between the commercial banks’ performance and the growth ofdomestic economic is strongly constructive as commercial banks play the role asfinancial intermediaries which move free cash from citizens to corporationsneeding capital for business and investment Therefore, the change oncommercial banks’ performance will largely influence on the whole economicactivities

A lot of researchers carried out studies concerning to examination of bankefficiency In the research of Casu and Molyneux (2003), they investigate theimpact of the efficiency score of European banking system on banks in eachcountry The authors use Tobit regression to evaluate the impact of eachcountry’s specific factors concerning to bank efficiency to measure determinants

of banks in Europe The findings show that the efficiency levels of bankingsystem are different between European countries due to specific country-factorslinked with banking advanced technology

Isik and Hassan (2002) assess the input and output efficiency in banks byapplying non-parametric and parametric approaches in Turkish banks in theperiod of 1988-2006 The obtained results indicate that dissimilar characteristics

of banks considerably impact the efficiency and different banking characteristicsimpact the variations in bank efficiency; and the efficiencies in cost and profithas a little decrease across periods

In the research of Chen, Skully and Brown (2005), bank efficiency before andafter Chinese government conducted the deregulation scheme in 1993-2000 areinvestigated to reveal the difference of Chinese banks Results show banks whichare owned by the state and smaller sized are often more efficient than those ataverage size Moreover, Chinese banks often have better technical efficiencythan allocate efficiency The banks have better cost efficiency after the scheme ofderegulation in 1994

In Canada, technical efficiency of banks in the period of 1983-1987 is explored

by two scholars named Nathan and Neave (1992) Researchers uncover that thecost efficiency in large-sized banks were not as good as small-sized banks

The examination of Berger, Hanweck and Humphrey (1987) on 413 subsidiaries

of national banks and 241 commercial banks owned by the government in 1983

is carried out The inputs of efficiency consist of capital and labor, and theoutputs of efficiency are call deposit, term deposit, mortgage credit volume,installment credit volume The economic efficiency of 241 commercial bankshaving state ownership was averagely at 0.96, while this figure was at 0.98 insubsidiaries of national banks So, subsidiaries of national banks perform moreproductively than commercial banks

Fukuyama (1993) uses the DEA method to assess the efficiency of 143commercial banks in Japan The inputs of bank efficiency include: labor, tangible

assets and intangible assets; and the outputs of bank efficiency comprise:incomes from lending and financial services The study result indicates that theefficiency of most banks enhanced with bank size, and banks having total asset

of more than 8 billion JPY got the best productivity

In the context of Vietnam, the efficiency of banks have been paid attention andevaluated by scholars, such as: Huong (2002) shows some suggestions toenhance the efficiency on investment activities in commercial banks, Dan (2004)applies the statistical method to investigate the technical efficiency in commercialbanks, Binh (2005) measures the competitiveness level of domestic banks afterjoining the global and regional integration agreements

In the measurement of technical efficiency, the functions of production and costare applied by Phu (2002) to evaluate the efficiency of commercial banks.Nevertheless, the drawback of this paper is to fairly recognize costs and use costfunction to figure out the boundary of the framework Therefore, the inefficientaspect at banks is not apparently revealed On the other hand, Anh (2004)measures unproductive factors in a state-owned commercial bank – Agribank,but merely the function is built Thanh (2010) judges how commercial banks inVietnam use efficiently input resources by applying DEA method The findingindicates that input resources are used efficiently by commercial banks, but atmoderate level and highly potential to improve the efficient usage of inputresources in the future By employing traditional methods of analyzing financialratios, Truc and Danh (2012) evaluate technical efficiency in commercial banksand discover that small banks are more easily vulnerable by the consequences ofglobal downturns than other medium banks

In general, studies about bank efficiency are undertaken in developed nations,rather than developing economies Researches concerning to technical efficiencymeasurement are not executed deeply in the case of Vietnam Therefore, it isnecessary to make a specific analysis paper about bank efficiency in Vietnam

2.2 Measurement of bank efficiency

In order to evaluate the efficiency of banks or corporations, four methods areused popularly, namely: stochastic frontier approach (SFA), data envelopment

analysis (DEA), distribution-free approach (DFA) and thick frontier approach(TFA) SFA and DEA are applied most by scholars to assess bank efficiency.

SFA method was developed by Aigner, Lovell and Schimidt (1977) and Meeusenand van den Broeck (1977) with the aim of calculating the corporate efficiency inproduction The method is built on the ground of the econometric theory andunder the assumption that input and output has a stochastic relationship SFAmethod is applied to evaluate efficiency in various industries, such as: Greene(2004) and Gerdtham et al (1999) assess efficiency of healthcare centers, Diazand Sanchez (2008) examine the efficiency of manufacturing enterprises, Wang(2003) investigate the efficiency of financial institutions

Charnes, Cooper and Rhodes (1978) developed DEA approach deriving from thetheory of Farrell (1957) about the application of non-parametric approach andmathematical practice to measure the level of efficiency DEA concerns to theusage of linear encoding method to construct the frontier of non-parametricpiecewise with the purpose of calculating efficient point on the frontline DEAassesses the efficiency of decision-making units (DMU) which are corporations,banks or institutions… The result of DEA shows whether DMUs can achieve theefficiency or not Researchers implement this approach to measure the efficiency

in many sectors, such as retailing sales (Thomas et al., 1998; Keh and Chu,2003); farming sector (Mao and Koo, 1997)

Besides, there are a lot of researches using DEA and SFA approaches to evaluatebank efficiency Ferrier and Lovell (1990), the assessment of around 600 banks

in the United States are done in terms of cost structure in 1984 The outcomes ofefficiency generated from DEA approach are found to be higher than those fromSFA approach Thus, DEA is more flexible to run data

Two these approaches of DEA and SFA are also used by Sheldon (1994) toestimate cost efficiency of Switzerland banks in the period of 1987 -1991 It isfound that the cost efficiency of banks calculated by SFA is at 56%, whereas DEAmeasures the cost efficiency of banks at 3.9% The considerable gap betweentwo approaches in efficiency measurement proves the doubt about theappropriate estimate of cost function Nevertheless, the research of Resti (1997)which applies two similar approaches to measure efficiency of Italian banksshows the little divergence between two outcomes

In the United States, the performance of banks is assessed in the study of Bauerand Berger (1984) by using SFA method The operation of US banking system isdetected to be inefficient, but not in the aspect of scale or scope This efficiency

is explained by the burden for banks to cut operational expenses, unitecompulsorily with other healthier banks, or exit the market SFA approach isimplemented in the different study of Altunbas et al (1994) to analyze Germanand Italian banks

By employing DEA approach, technical efficiency of commercial banks isindicated to be negative association with bank size in the paper of Field (1990).The efficiency of bank subsidiaries in Canada is also explored by DEA method(Wu et al, 2006) The efficiency of commercial banks in Greek in the period of2000-2004 is evaluated by Pasiouras (2008) He points out that the provisionamount for credit loss has positive relationship with the efficiency level of banksmeasured by DEA approach By using the same analysis method, Brazilianbanks’ efficiency is revealed in the aspect of cost, technical and allocative (Staub

et al., 2010) The scholar gave the empirical result that the cost efficiency ofBrazilian banks is less than those of EU and US banks Meanwhile, cost efficiency

of state-ownership banks is higher than those in overseas and private banks Overall, two approaches of SFA and DEA are applied singly or jointly toinvestigate and evaluate efficiency of companies, banks or specific industry.However, studies relating to efficiency are carried out mainly in advancedeconomies So, there is a large room to conduct research in developingeconomies

2.3 Model of factors affecting bank efficiency

CAMEL framework is often used to identify determinants of bank efficiency,consisting of capital, asset, management, earning and liquidity Besides, factorsimpacting bank efficiency are selected on the ground of previous researches andrequests of directors in bank management Upon considering prior studies aboutbank efficiency, such as: Altunbas et al (1994), Das (2002), Forster and Shaffer(2005), Staub et al (2010), and Pasiouras (2008) Five factors are chosen toevaluate the influence on bank efficiency, comprising: bank size, loan ratio,capital ratio, non-performing loan, type of ownership

2.3.1 Factors affecting bank efficiency

2.3.1.1 Bank size

Bank size is measured by the natural logarithm of total bank assets Logically, it

is predictable that bank size and efficiency have positive relationship The size commercial banks possess the abundant resources of capital, employees,technological infrastructure, thus they have higher probability to increase theoverall efficiency However, when problems occur due to mistakes in bankoperations and administration, these banks would suffer worse consequencesthan small banks Thus, small-size banks are evaluated to obtain betterefficiency in the difficult period

larger-In previous studies, researchers found that bank size have strong relationshipwith the efficiency It is assumed that large banks have competent and skillfulmanagers or have advantage of economic scale in cost as the directorsemphasize dominantly on profits (Evanoff and Israilevich, 1991)

The earlier researches show different results about the correlation between sizeand efficiency In the paper of Berger, Hunter and Timme (1993), the affirmativerelationship between efficiency and size may not apparent because factorsrepresentative are not convincing Larger companies able to manufacture at thefull capacity will reach the efficient point It means that greater banks maygenerate higher level of profits at a specific cost as they have progressivelyimproved in size during a given time Inversely, small banks are difficult to haveenough capacity to obtain the productivity in a short term Besides, it islikelihood that companies with higher level of efficiency have better competitiveadvantage, and consequently, they are increasingly large Reda and Isick (2006)support that bank size have significant association with efficiency

Findings about the linkage between size and efficiency are conflicting in differentresearches Ataullah and Le (2006) and Ajlouni, Hmedat and Hmedat (2011)realize that larger banks tend to be better efficient when they enlarge the size ofasset In the research of Chen, Skully and Brown (2005), banks with large andsmall size have tendency to perform better than medium-size ones However,this finding is opposite with the US banks where the curve of average cost is flat,signifying the efficiency of medium-sized banks

By contrast, Almumani (2013) indicates that the performance of small-size banks

is better than medium and large-size ones in the aspect of technical efficiency.Isik and Hassan (2002) show the findings about the inverse relationship betweensize and efficiency Small and medium-size banks tend to spend large proportion

of budgets to operate activities, but they can get better technical and scaleefficiency than larger-size ones DeYoung and Nolle (1994) and Kaparakis, Millerand Noulas (1994) find that smaller banks can obtain the efficiency in profithigher than large banks and efficiency evaluation reveals no scale biases on thebank size Havrylchyk (2006) and Avkiran (1999) also confirm these findings byshowing the insignificant relationship between bank size and efficiency

From the results of existing literature, the researcher assumes that:

H1: The relationship between bank size and technical efficiency is positive inVietnamese commercial banks

2.3.1.2 Equity capital

Equity capital is total amount of capital contributed by investors/shareholders ofbanks The ratio of equity and total assets is the proportion of equity capitalamount of banks and its asset amount The higher ratio of equity results in theincrease of return on equity In other words, stakeholders can mitigate risk andget better profit when investing in the bank Commercial banks tend to use highratio of leverage which is contributed by depositors and creditors, while the ratio

of equity accounts for only 8% However, the proportion of equity has asignificant role in maintaining the stable activities of banks and is an importantsource for banks to develop and expand the business in long term The equitycapital is useful to minimize the potential risks arising from banking operationsand protective actions in case of insolvency If banks have the amount of lossmuch higher than the capital, they may be in threat of insolvency and default.Therefore, the maintenance of adequate capital is important for commercialbanks

The proportion of equity capital could have positive and negative impact on theefficiency of commercial banks In existing literature, the relationship betweenequity ratio and bank efficiency is found to be different Girardone et al (2004)indicate the opposite relationship between equity and inefficiency in Italianbanks Das (2002) also shows that the Indian banks had higher ratio of equitywill perform better in the period of 1995 – 2001 Besides, Pasiouras et al (2007)

affirm that Greek banks’ technical efficiency is positive with the amount of equitycapital during the time interval of 2000-2004 The positive correlation betweenequity and bank efficiency is supported by different scholars as Mester (1996),Pastor et al (1997), and Carbo et al (1999)

However, the research of Casu and Molyneux (2003) reveals that the correlationbetween equity ratio and efficiency was insignificant in European banks in thetime of 1993 -1997

The majority of studies explore that bank’s efficiency and the equity ratio havepositive relationship Therefore, it is assumed that

H2: The relationship between equity and technical efficiency is positive in Vietnamese commercial banks.

2.3.1.3 Loan to total assets

The ratio of loan to total assets is the proportion of total lending volume on totalbank’s asset This ratio reveals the risk level of liquidity in banks as it shows theallocation of low liquid asset in total asset When banks provide larger volume ofloans, the operational cost is reduced and thus increasing progressively theamount of loans to customers (Isik and Hassan, 2003)

Sufian (2009) explores that Malaysian banks having higher proportion of loanstend to obtain higher level of efficiency in 1997 Supporting this result, Sufianand Habibullah (2010) and Sufian and Noor (2009) testify that banks with moreefficiency often have larger volume of loan in the cases of Thailand and Islamic.However, Barr et al (1999) indicate that higher proportion of loan on total assetswill result in the lower efficiency of banks because of higher potential risks From the findings above, it is supposed that

H3: The relationship between the ratio of loan and bank’s technical efficiency ispositive

2.3.1.4 Non-performing loans

Non-performing loans are loans whose interest and principle has been paidoverdue for a long time Non-performing loans arise from controllable anduncontrollable factors (Berger and Mester, 1997) Controllable factors includequality management, credit appraisal and policies, loan structure and credentials.Uncontrollable factors consist of harsh business environment, unexpected

modification in regulations and laws, problems from borrower’s business andcatastrophic events Due to the impact of external factors, loss from loans isimpossible to discard totally The understanding of this will be helpful to acceptthe risk of credit, which results in lower profit Therefore, bank managers shouldtake into account the loss arising from loans because bank employees aredifficult to predict exactly if the borrower can pay interest and principle on duedate or not

Several researches about the efficiency have examined the level of bankefficiency in the relation with the quality of asset In earlier papers, assessment

of non-performing loans have straightly been combined to manage asset quality

in the functions of cost and profits, thus indicators of efficiency are achieved.Other scholars applied different methods, and include the variable of non-performing loans as in the regression model (Reda & Isik, 2006; Mester, 1994)

In the research of Das and Ghosh (2006), regardless of selecting input andoutput, the non-performing loan and inefficiency have positive relationship.Berger and DeYong (1997) state the hypothesis of “bad management” that thegrowth of non-performing loan reduces the efficiency of financial institutionsbecause of increasing cost in administration, supervision and sales-off

Some academicians as Wheelock & Wilson (1995) and Barr & Siems (1994)explore that weak banks tend to be positioned far from the optimal practicefrontier Thus, despite of high proportion of non-performing loans, these banksstill have tendency to show the efficiency in cost Studies of Karim, Chan &Hassan (2010) and Kwan & Eisenbeis (1995) provide supportive evidence thatefficiency and non-performing loans have increasing correlation in banks Fanand Shaffer (2004) also affirm that the association between non-performingloans and efficiency in US large banks was negative, but no statisticallysignificant

The author proposes the hypothesis 4:

H4: The relationship between non-performing loans and technical efficiency isnegative in Vietnamese commercial banks

2.3.1.5 Ownership structure

The importance of ownership structure has been concentrated on the stage ofreforming banking activities, especially in developing nations The big inquiry is

how to maximize the structure of ownership and management in banks toenhance the mutual benefits of stakeholders and directors (Spong et al., 1995).The ownership of government is supported by the belief that the actions ofgovernment aims to ensure the best advantages for the general society; so, thestate ownership is useful to reduce common costs and improve efficiency Thecontribution of the government in the operation of banking system is hoped toencourage more capital to develop the domestic economy and bring other socialbenefits for the society instead of focusing on economic benefits (Megginson,2005)

However, the expectations about advantages of state-owned banks are failureand their banking activities have worse efficiency than private banks It isbelieved that the purpose of the state in banking control is primarily to increaseits advantages rather than social improvement in spite of risk aspects andunproductive asset distribution

The correlation between ownership structure and efficiency in banks has beenstudied by many scholars, especially in economies that the state has a key part

in the operation of banking system Several empirical results indicate that theassociation between state ownership and efficiency in banks is converse Creditfor political and social aims often fall into the inefficiency Although thegovernment monitors and directs commercial banks towards some specificbenefits, the financial detriments counterweight the returns Conversely, theprivate ownership is thought to be improve the performance of banks byhandling the troubles between directors and stakeholders and encourageinvestors to investigate the execution of management Therefore, the elimination

of state ownership in the banking sector by privatizing may be most expected.Das (2002) indicates that Indian banks can enhance their performance if thecapital is contributed more by the state Sathye (2003) also uncovers thatprivate commercial banks tend to be weaker in efficiency than public andoverseas commercial banks in the context of India Supporting this finding, Unal,Aktas and Acikalin (2007) prove that state-owned banks have the sameefficiency with private banks and even perform better at several facets

Sufian and Noor (2009) hold the opposite view that the efficiency score of privatebanks is higher than those of public banks in the situation of Italian economy.Hussein (2003) reinforces the above outcome when doing a research on 17

banks in Sudan, Islamic in the period of 1990-2000 with the application ofstochastic cost frontier model He finds that in three groups of state-owned,private and overseas banks, state-owned banks are the worst at cost efficiency

In the situation of Vietnam, commercial banks have been progressivelyprivatized; however the state still holds a large percentage of shares in severalkey banks The state-owned banks are defined as the holdings above 50% by thestate Joint-stock banks have an insignificant part of state capital or none Due tothe abundant experience in operating banking services and higher volume ofcapital, state-owned banks are supposed to be more efficient that joint-stockbanks Thus, the hypothesis 5 is

H5: The relationship between state-owned banks and technical efficiency ispositive

2.3.2 Proposal research model

This dissertation will test 5 hypotheses:

H1: The relationship between size and technical efficiency in Vietnamese banks ispositive

H2: The relationship between equity capital and technical efficiency inVietnamese banks is positive

H3: The relationship between loan ratio and technical efficiency in Vietnamesebanks is positive

Bank technical efficiency

Non-performing loans

(-)

State ownership

(+)

H4: The relationship between non-performing loans and technical efficiency inVietnamese banks is negative

H5: The relationship between state-owned banks and technical efficiency inVietnamese banks is positive

CHAPTER 3 METHODOLOGY 3.1 Data Envelopment Analysis (DEA)

DEA was built by Charnes, Cooper and Rhodes (1978) for the sake of measuringcorporate efficiency in the public sector, especially on non-profit corporationswhere the amount of account profit is at insignificant value, different outputs aregenerated with different inputs, and it is complicated to recognize efficientcorrelation of input and output The formation of this approach arises whenprices might be unavailable or unreliable and hypotheses of cost minimizationand profit maximization might be unsuitable (Bauer et al., 1998)

DEA approach is viewed as a standard for efficiency evaluation (Canhoto &Dermine, 2003) This approach was firstly applied in banking sector by Shermanand Gold (1985) Currently, DEA is one of the most popular methods to measureefficiency of banks

 Definition and Assumptions

DEA frontier is built as a part of linear combination linking the group of bestobservation in sample set, so producing a convex Production Possibility Set(PPS) Therefore, scores of DEA efficiency for a particular decision-making unit(DMU) are not expressed by an absolute figure, however relative to other DMUs

in the particular group of data (Casu and Molyneux, 2003) Cooper et al (2006)indicate that the method’s name is called as the technique it covers observationsfor the purpose of recognizing a frontier which is employed to assessobservations signifying the performance of measured units by recognizing thesources and volumes of inefficiency in every input and output for single DMU andDMUs are positioned on the frontier of efficiency

DEA model was firstly constructed on the foundation of constant return to scale(CRS) which is often seen as CCR framework After that, Banker, Charnes andCooper (1984) broadened this model to variable returns to scale (VRS) Itpermits the recognition of whether a DMU is functioning at growing, constant, ordeclining return to scale and generally acknowledged as BCC model The model isalso benefit for identifying separately technical and scale inefficiency Technicalefficiency is the score of efficiency in CRS framework, while scale efficiencyentails running data in both two models of CRS and VRS and equals to the ratio

of CRS and VRS scores

The score of scale efficiency = CRS score/ VRS score

It is not required pre-designated weights for many inputs and outputs, but itrecognizes a group of most productive DMUs, where the scores of DMU’sefficiency are measured The score of efficiency reveals a chain of weights asregulated from the data, and the highest efficiency point is made by one forevery output and one for every input

A suggesting group for an inefficient DMU is set up the efficient DMUs, which aninefficient DMU is assessed contrary The suggesting group is the non-parametricform of the frontier of efficiency according to the parametric methods Besides,this suggesting group permits management to position and comprehend theessence of inefficiencies by assessing the inefficient DMU with its equivalentefficient complement (Sherman and Gold, 1985)

The slack based measure (SBM) of efficient DMUs is recommend by Tone (2001)for the sake of integrating coincident input surplus and output deficits, andtherefore explained as the outcome of input and output inefficiencies Moreover,the process solves with the regular zero apportioned weights in the traditionalassessments by distributing weights to entirely inputs and outputs of DMUs withthe exemption of non-positive data Likewise, SBM tries to locate the maximumpractical returns in contrary to the CCR concentrating on recovering themaximum proportion of real output on real input

Andersen and Petersen (1993) enlarge the expansion of DEA to deeper examineand rate the efficient DMUs from prior frameworks Their perspective assessesDMU with a linear grouping of efficient DMYs in the trial, so DMU is discarded.This directed to Tone (2002) combining the SBM and super-efficiency model(SEM) built initially by Andersen and Petersen (1993) in order to get the slack-based evaluation of super-efficiency The SEM is viewed as an evaluation ofstability, in case that the input data is exposed to inaccuracy or fluctuationacross periods, and offers ways of assessing the aspect of which thesemodifications could happen without infringing that DMU’s position as being anefficient element (Cook and Seiford, 2009)

Casu and Molyneux (2003) recommend that Tobit regression model is applied toexamine the impact of numerous factors relating to specific nation andenvironment on the efficiency of banks and condensed data can be comprised,despite of several questions for verifying the validity of this method (Simar and

Wilson, 2007) Besides, the bootstrapping approach is implemented to correctthe integral dependency of DEA applied in regression model Grosskopf (1996)states that dependency is validated as the efficiency score of DEA is relative,rather than absolute index of efficiency; so breaking the individual assumption inthe model of regression and proposing that typical process is unenforceable(Grosskopf, 1996) The assumptions of DEA on the ground of unit uniformity areevaluated as followed: (1) units are (1) units are presumed to involving in thesame operations, generating equivalent output set and popular technologies areimplemented; (2) the same series of resources is accessible to units; (3) unitsare functioning in the same environment (Dyson et al., 2001); (4) the sufficientsample size ought to be bigger than the outcome of input and output numbers(Cooper et al., 2006) or the quantity of sample needs to be triple than those ofinputs and outputs at least

Strengths of DEA

The advantages of DEA is that it is not essential for explicit or apparently stateddescription of a purposeful form and enacts less organisation on the form ofefficiency frontier and data is permitted to express themselves (Wheelock andWilson, 2006) It solves different inputs and outputs specified in various units ofmeasurement, and concentrates on most efficient frontier, instead of tendencyfor centralized population (Chen et al., 2005)

Siems and Barr (1998) propose that DEA model is helpful for policy-makersbecause of balancing off-site supervising instrument Moreover, they presentfeatures of DEA model to comprise: a firm cost-effective and mathematicalunderlining; substitute practical and combined or theoretical best-practice units;the capacity to take into consideration trade-offs and replacements between thestandard metrics, and the suggestion for enhancement on different corporateaspects

In comparison of three approaches: DEA, SFA and TFA, Bauer et al (1998)indicate several proofs to show that the efficiency is fluctuated across period andthe efficiency is almost the same in the stability in two approaches of parametricand non-parametric A significant distinction between methods is that DEAoverall is more stable than other approaches Likewise, Sickles (2005) examinesefficiency measurement of parametric and non-parametric, consisting of DEA,and shows that DEA is a better measurement tool of technical efficiencies varying

across time for corporations, and also implements well in place of accurate andpredicted efficiency of companies, especially, once there is a growing trend in thenumber of cross-sections and time series

Limitations of DEA

However, the most disagreement of DEA is concerning on the lack of ability tocombine random error For banks, their costs which are lower than themeasurement level would be categorized as most efficient, and other adverseimpact out of the management of banks would be recognized as inefficiency(Mester, 1996) The case becomes severer when there occurs random error onthe frontier of efficiency or a leading recommending set, it impacts the deliberateefficiency of all companies that are contrasted (Bauer et al, 1998)

Besides, DEA only employs input and output data and may not directly consist ofinput prices, so it causes difficulties to approximate allocative inefficiency(Berger, 1993) Nevertheless, this matter is against the initial opinion of Farrell(1957) relating to the difficulties resulting from price efficiency

Bauer et al (1998) indicate that a possible problem for DEA is the matter of identifiers and near-self-identifiers Every company merely could be incomparison with other companies on the frontier with the similar outputs or viceversa adding more limitations applied on the ground of comparability of qualitycontrols

self-Therefore, no any companies in different aspects could lead to companies beingevaluated as highly efficiency or self-identified of 100% because of shortage ofequivalent companies in the limited variables Basically, this matter usuallyarises in case of small size of observations comparing with the input and outputsize, and other restrictions leading to a problem of coordinating all aspects.Brown (2006) indicates that the drawback of DEA is the deterministic andtroubled in assessing variables errors Likewise, Fried et al (2002) recommendthat the majority of DEA frameworks and practically all operative DEAframeworks are deterministic, therefore incapable to take the stochasticconstituent of a DMU This encourages these researchers to build a DEA on theground of model consisting of a stochastic factor proposed to separate theinfluence of luck from the environmental influence and managerial execution.However, this view is cost-effectively and statistically illogical because thedeterministic characteristics of DEA may be strongest element

3.1.2 Malmquist productivity index

In addition to the DEA model, several indexes are applied to measure theefficiency, comprising: Fisher index, Tornqvist index and Malmquist index Thisresearch decides to use Malmquist index due to some benefits Initially, it is notrequired to assume profit maximization or minimization Secondly, it isunnecessary to get data relating to input and output price

Productivity index is significant to investigate the level of efficiency and itsvariation over different period The index of Malmquist shows the variation oftotal factor productivity (TFP) It is calculated by the ratio of the distance ofoutput and input vectors in two successive periods, proportional to a directedtechnology (Coelli et al., 2005) The productivity index of Malmquist is explained

by Färe et al (1994) as the formula below:

In which:

Mj = Malmquist productivity index

Dj = Distance function

x and y = inputs and outputs from t to t+1

The components constituting Malmquist index are displayed in the formula The

proportion of shows the variation of technical efficiency (EFFCH) in

the period of t and t+1 The proportion of is thearithmetical mean of two index which signifies for the change in technology(TECHCH) in the period of t to t+1 So, the variation of Malmquist productivityindex is imparted as the outcome of two components EFFCH and TECHCH

On the other hand, as asserted by Fare et al (1994), the change in technicalefficiency (EFFCH) could be divided into two specific components, consisting ofpure technical efficiency change (PEFFCH) and scale efficiency change (SECH):

SECH is equivalent to the ratio of technical efficiency change (EFFCH) and puretechnical efficiency change (PEFFCH).

With the calculation above, the productivity index of Malmquist is written asfollowed:

Malmquist Productivity Index = EFFCH x TECHCH = PEFFCH x SECH x TECHCH

The Malmquist index is more than 1, meaning the improvement of productivity.Conversely, the index is lower than 1, showing the decline of productivity.Besides, the rise of each component in Malmquist index will result in thecomponents’ value higher than 1

3.1.3 Specification of inputs and outputs

Identifying variables of input and output of efficiency has been an arguablesubject in researchers due to tis methodology in a long period Operations ofbanks are totally distinct with the manufacturing corporations which producephysical products, while banks provide different intangible products, such asfinancial services Some indicators are used in prior researches to measure theoutputs of bank efficiency, comprising: assets, liabilities and net incomes.Meanwhile, other scholars recommend choosing the outputs of bank as totalcustomer deposits and credit volume because they are main services of banksdirectly provided to clients Alternatively, the amount in deposit accounts is

suggested as bank output as these amount is seen as the major profitable sourcefor banks The bigger amount in deposit accounts, the better demand of clientsabout financial services of banks (Goddard et al., 2001).

In order to define variables of input and output for bank efficiency, two popularmethods are often applied, including: production and intermediation methods The method of production emphasizes on physical inputs, such as: capital andlabour because banks are seen as manufacturing enterprises which haveproducts of deposits and credits Banks operate and manage transactions andfinancial data for customers The amount and classification of transactions andfinancial data of customers are seen as good pointers to assess the bank output.From the theoretical view, it is rational to apply this approach Nevertheless, inactuality, this method is not easy to implement due to the nearly inaccessibility

of customers’ data Practically, the volume of deposit and credit are utilized toassess the output of banks, rather than data about detailed histories abouttransactions and documents In the research of Ferrier and Lovell (1990) andFriedetal (1993), this approach was used

The intermediation method was created by two scholars Sealey and Lindley(1997) The intermediation method assumes that banks which have the role offinancial intermediaries are in charge of transferring free cash flows from savers

to borrowers Consequently, the lending volume is regarded as an output, andhuman resources, deposit volume and capital volume are regarded as inputs Thedifference between production and intermediation approaches is to add depositvolume as input variable and consider including operating and interest expenses.Several papers have been implemented this method to examine the technicalefficiency, for example: Avkiran (1999), Fries and Tacis (2005), Drake and Hall(2003)…

With the accessible data sources, implications from prior studies about theefficiency in banks and actual activities in Vietnamese commercial banks, theauthor chooses the variables of input and input for bank efficiency on the ground

of intermediation method The inputs of bank efficiency include lending volume(consisting of total lending for individuals and groups), investment and bankingservices Meanwhile, the outputs of bank efficiency are identified as interest andnon-interest as these are two main sources of income for banks

Variables of bank efficiency input consist of:

- Total fixed assets are measured by total assets deducting lending amountand investment amount for all industries of a bank up to the period of end year,measured unit in million

- Total deposit is total deposit that a bank receives from customers up tothe period of end year, measured unit in million

- Total labour cost is total expenditure for hiring labour of a commercialbank up to the period of end year, measure unit in million

Variables of bank efficiency output consist of:

- Interest and equivalents are total amount of interest and other equalthings that a bank obtains from lending up to the period of end year

- Non-interest returns and equivalents are total amount of returns that abank gets from other activities up to the period of end year

Two variables of bank efficiency output are opted from the research of Barr(1998), Barr, Killgo, and Siems (1999) and Denizer and (2000)

3.2 Analysis of factors affecting banks’ technical efficiency

3.2.1 Tobit regression model

The model of Tobit regression is used widespread by researchers to examinedeterminants of technical efficiency in most sectors As established by JameTobin (1858), the model aims to interpret the link between non-negativedependent variable and independent variables by the approach of maximumlikelihood The model is frequently implemented in the study with censored ortruncated data Therefore, this model is known as Tobin probit or censoredregression

The model assumes that y*i is a latent variable which is dependent linearly on xithrough a parameter β This parameter decides the association betweenindependent variables xi and latent variables y*i Besides, εi is normallydistributed error term which is to get random impacts on this association Theobservable variable yi is identified to be equivalent to the y*i latent variable incase that the latent variable is more than or equal to zero

The Tobin regression model has the equation below:

y*i = β’xi + εi

yi = y*i if y*i = β’xi + εi >0 and

yi= 0 if y*i = β’xi + εi ≤ 0

Wherein

y*i is latent variable

yi is the measurement of bank efficiency at bank i

xi is independent variables

β is vector of parameters

ε is normally distributed error term

With the yi and xi of banks, the equation of Tobit regression above can be written

in the form of log likelihood with a given observation as followed:

In a simple form, the equation of Tobit regression can be expressed as followed:

Where ξit is technical efficiency of bank i at the period of t; Djit is dummy variableand Zjit is controlling variables

In this research, Tobit regression model is applied to examine determinants oftechnical efficiency in Vietnamese banks due to its conveniences The firstadvantage is that Tobit model can forecast censored or truncated datainterpreting a clustered distribution by the implementation of maximumlikelihood to predict the coefficients (Walker and Maddan, 2013) The secondadvantage is that the generated result of Tobit regression is more accurate thanthe normal linear regression model because its estimated coefficients areunbiased for every variable x

3.2.2 Variables

Tobit model is applied to investigate determinants of technical efficiency inVietnamese commercial banks Therefore, technical efficiency is defined as thedependent variable and gets the result from the computation of DEA approach.Other factors: size, equity ratio, loan ratio, non-performing loan ratio andownership structure are defined as independent variables Specifically, thesevariables are explained as below:

 Bank size is measured by the logarithm of total assets of banks up to theperiod of end year

 Equity ratio is the proportion of equity capital to total asset value in a bank

up to the period of end year Equity capital is defined as total invested amount

by shareholders into the bank

 Loan ratio is the proportion of lending volume to total asset value in abank up to the period of end year

 Non-performing loan ratio is the proportion of non-performing loan to totallending volume in a bank up to the period of end year This ratio shows the level

of credit risk in a bank The higher ratio means the bank facing higher risk ofdefault

 Ownership structure of a bank is dummy variable

3.2 Sampling and data sources

3.2.1 Sampling plan

 Target population

Target population is “the collection of cases in which the researcher is ultimatelyinterested, and to which he or she wishes to make generalizations” (Sim andWright, 2000) These chosen cases should be relevant and have features that thestudy is designed to gather Some factors affecting the selection of targetpopulation are research topic, accessibility of research units, and time restraints.This study chooses the target population of 10 commercial banks in Vietnam intotal of 40 local banks Because time limits and data accessibility, the number ofsample units is rather small, consisting of 4 state-owned banks and 6 joint-stockbanks

 Sampling technique

The selection of sampling technique depends on various factors in terms oftheoretical frameworks and research situation It should consider somesignificant issues, including: the study essence, research aims and objectives,time convenience and usable budget In business research, two samplingtechniques are often implemented by researchers, namely probability andnonprobability Regarding to probability sampling, the choice of sampling isconducted by random and the likelihood of being selected is determined byconductor In case that the sampling selection is performed properly, theprobability sampling will get the representative samples for large population.Meanwhile, the nonprobability sample allows researchers to add or reduceelements in the sample set in need of their demand and assessment It signifiesthat not all elements of sample set are chosen in the research Yet, the bestrepresentative samples will be selected due to the skilled sampling adoption.These samples will serve best for researcher’s objectives and no reliance onprobability

The non-probability sample has advantages of easy accessibility, quickness andcost saving Thus, the research chooses this method in this study 10 commercialbanks are chosen to investigate the technical efficiency The sample set includesfour state-owned banks and six joint-stock banks These banks operate in thenational scope

3.2.1 Data sources

Secondary data are existing data source gathered by other people, institutions,excluding users In this dissertation, the research chooses to get secondary datafor further analysis because of some advantages of secondary data The firstbenefit of using secondary data is to save time for collection When collectingprimary data, the research has to design a questionnaire, survey or interviews toget data, then process these obtained data Due to lack of the support of otherpeople during the research time, it should spend more time to amass data thanextract them from existing sources Although the spending for getting secondarydata is compulsory, it is not much than the costs for lunch, dinner or presents forbank directors or participants when the researcher would like to ask forinterviews because a small gift will give respondents stimulation to answercorrectly The more benefit of secondary data is the appropriateness of research

aim in this study This research requires historical data over years With theseadvantages, the secondary data is preferred to use in this dissertation

Data about financial indicators are collected from the official websites of banks.Data are obtained from consolidate balance sheet, income statement, cash flowsreport and financial statements in 5 years Additionally, other data are takenfrom issued articles, papers and researches from State bank of Vietnam andother financial institutions

3.3 Data analysis

The analysis and process of data is conducted by the software of Eview 8.0 andVDEA 2.0 More specifically, technical efficiency of banks is measured on VDEA2.0 The evaluation of effect of factors on technical efficiency is done bygenerating results in Eview 8.0 The descriptive analysis and correlation matrixare also processed in Eview 8.0

3.4 Summary

To measure the level of technical effectiveness in banks, the study employ DEAapproach of Charnes et al (1978) as well as Malmquist index in this approach.Toanalyze the components influencing the technical effectiveness of banks, Tobitregression model is utilized In terms of variables, the author uses ownershipstructure, loan to total assets ratio, equity to total assets ratio, non-performingloan ratio and bank size This study will use the secondary data and data from 10commercial banks’ official websites from 2010 to 2015 and then analyzed themthrough using Eview 8.0 software