Determinant of non performance loans the case of vietnamese banking sector

ABSTRACT The main purpose of this study is to examine the determinants of non-performing loans NPLs in the case of Vietnamese banking sector by analyzing the unbalanced panel data of 30

HOCHIMINH CITY

VIETNAM

THE HAGUE THE NETHERLANDS

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

DETERMINANTS OF NONPERFORMING LOANS THE CASE OF VIETNAMESE BANKING SECTOR

A thesis submitted in partial fulfillment of the requirements for degree of

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

By

TRUONG NGOC THANH

Academic Supervisor

DR NGUYEN THI THUY LINH

HO CHI MINH CITY, DECEMBER 2016

ABSTRACT

The main purpose of this study is to examine the determinants of non-performing loans (NPLs) in the case of Vietnamese banking sector by analyzing the unbalanced panel data of 30 Vietnamese banks over the period of 2008 – 2012 Both of macroeconomic and bank-specific determinants are employed when modeling the regression of NPLs’ determinants Macroeconomic factors including Gross Domestic Product (GDP) growth rate, unemployment rate, real lending interest rate and sovereign debt are exogenous variables that effect on NPLs Besides that, the study examine the bank-specific determinants by analyzing relevant hypothesis such as ‘bad management’, ‘pro-cyclical credit policy’,

‘skimping’, ‘diversification’, ‘too big to fail’, ‘moral hazard’ hypothesis According these hypotheses, return on equity, inefficiency rate, proportion of non-interest income and leverage ratio are the endogenous variables which effect to NPLs In addition, credit growth rate is added into model to examine its effect on NPLs Moreover, the effects of government intervention and foreign investment

on NPLs are also examined in this study by investigating the difference in NPLs of state-owned banks and fully foreign-owned banks The fixed effect of unbalance panel data is employed to test these hypotheses

Regarding bank-specific factors, the inefficiency rate and credit growth rate statistically affect on NPLs However, return on equity, non-interest income rate, leverage ratio do not statistically significant effect on NPLs According to regression result, it shows the negative and significant relationship between the inefficiency rate and NPLs that is consistent with ‘skimping’ hypothesis Moreover, the relationship between credit growth and NPLs is significant and negative

As the regression result, all of macroeconomic determinants including GDP growth rate, unemployment rate, real lending interest rate and sovereign debt statistically significant affect on NPLs The regression shows the positive and significant relationship between the sovereign debt and NPLs which is consistent with hypothesis The increase in sovereign debt will reduce payment ability that increases the future NPLs However, the regression shows the positive relationship between GDP growth rate and NPLs and negative relationships between the unemployment rate, lending interest rate and NPLs that is not consistent with hypothesis

Regarding the government intervention, the regression shows that return on equity and leverage ratio are affected in state-owned bank that lead to higher NPLs However, the effect of foreign investment

in fully foreign-owned banks on NPLs is not supported in this study

There are some policy implications based on the regression results Firstly, the sovereign debt should

be strictly control in order to enhance the payment ability of debtors Secondly, the underwriting and monitoring loans process should be controlled to reduce NPLs expansion at bank level Finally, the operations of state-owned banks should be controlled to reduce NPLs expansion in state-owned banks

TABLE OF CONTENT

CHAPTER 1: INTRODUCTION 1

1.1 Overview of Vietnamese banking sector and non-performing loans 1

1.2 Research problem 2

1.3 Research objectives and research question 4

CHAPTER 2: LITERATURE REVIEW 6

2.1 Non-performing loans definition 6

2.2 Bank-specific determinants of non-performing loans 7

2.3 Macroeconomic determinants of non-performing loans 12

2.4 Government intervention and foreign investment in banking system 16

CHAPTER 3: METHODOLOGY AND DATA 19

3.1 Methodology 19

3.2 Data 21

3.3 Estimation approach 23

CHAPTER 4: ANALYSIS RESULTS 25

4.1 Descriptive statistics 25

4.2 Economic results 27

4.3 Result discussion 30

CHAPTER 5: CONCLUSION 35

5.1 Main findings and policy implication 35

5.2 Limitation of the study 36

REFERENCES 38

APPENDIX 41

LIST OF TABLE

Table 1: Definition of variables used in modeling NPLs determinants 17

Table 2: Specific calculation of variables 22

Table 3: Methodology test 24

Table 4: Descriptive statistics 25

Table 5: The correlation matrix 26

Table 6: Summarize NPLs 27

Table 7: The regression result 28

Table 8: Regression result of dummy variables 29

Table 9: Empirical evidence for tested hypothesis 34

CHAPTER 1: INTRODUCTION

1.1 Overview of Vietnamese banking sector and non-performing loans

There are three types of ownership in Vietnamese banking sector including state-owned commercial banks, joint stock commercial banks, foreign banks (Kalra, 2012) State-owned commercial banks play an important responsibility in international financial by lending to main sectors in Vietnamese economy In particular, loans of trade and industry sectors central is granted by Bank for Industry and Trade (ViettinBank) while foreign payments is in-charged by Bank for Foreign Trade (VietcomBank)

In additional, loans of agriculture and fishing are supported by Bank for Agricultural Development (AgriBank) Concerning the bank market share, state-owned commercial bank account for large bank market share in 2010 (Kalra, 2012) Besides that, the growth of joint stock commercial banks also contributes in the banking sectors throughout their financial services

In Vietnam, banking sector is under the control of government throughout the State bank operations Besides the financial responsibility, some duties of state-owned bank are expected In particular, loans

of main sectors in the economy are financed by state-owned commercial banks In addition, money supply and demand are controlled by state bank by opening the market operation, reserve system, bank rate policy Moreover, all regulation as well as guideline of banking operations must be complied with state bank’s regulation

The Vietnamese banking system is significantly impacted by the economic depression over the period

of 2008 – 2012 which leads to NPLs expansion The main cause of bank problem is the deterioration

of loan portfolio As the same situation with international banking system, Vietnam experienced with

a period of the housing bubble and rapid growth in the stock market Allowing easy access to loans and rapid credit growth, Vietnamese banking sector had to face with the credit exposure when economy went down According to report of State Vietnamese Bank, the loan portfolio significant increased from 2005 to 2007 Specially, the credit growth rate was 52.42% in 2007 that doubly increases comparing with this in 2006 In addition, high unemployment rate in period of economic downturn strongly impact to the payment debt ability Moreover, the weakness of Vietnamese banking sector is one cause that expand the problem loans Excessive loans, loose credit policy assessment, less mortgage loans, lose control in loan monitoring are the problems of Vietnamese banking sectors

As the consequence, the NPLs rate was 3.4% in 2012 which doubly increases comparing with this in 2009.

Many reactions were implemented by State bank of Vietnam to solve the bank’s NPLs The number

of policies was implemented including increasing capital adequacy ratio to 9%, increasing restriction for lending credit, establishing Vietnam asset management company (VAMC), buying NPLs of weak banks, restructuring weak banks, issuing new loan classification, etc In addition, minimum of charter capital of banking sector was increased Interest rate ceilings were re-imposed to control operation of banking sector as well stable the economy However, the NPLs rate was not significantly improved According the World Bank’s report, the NPLs declined to 3.107% by the end of 2013 because of transferring bad loans to the VAMC However, the NPLs in 2013 also emphasizes that this rate could

be 9% if all restructured loans were included (Mellor, Minh, & Thuc, 2014) In the other sides, according to rating agency Moody’s estimation, NPL could be higher and exceed 15% in the case of implement international standard assessment

The concern of NPLs was raised in Vietnamese banking sectors in recent years In addition, the root cause of NPLs of bank’s sector was examined to find out best measure for NPLs solving Therefore, the main purpose of this research is to examine the determinants of NPLs in the case of Vietnamese banking sector in order to find out the appropriate policy implication for solving banking NPLs

1.2 Research problem

Reviewing empirical studies, there are many approaches to examine the determinants of NPLs On the one hand, macroeconomic factors could be employed to evaluate their effect on NPLs Berge and Boye (2007) conclude that real interest rate and unemployment are highly sensitive with the problem loans They find out that one of primary contribution in real interest rate and unemployment rate improvement is the problem loans’ declining (Berge & Boye, 2007) Besides that, according to study

of Reinhart and Rogoff (2011), they made conclusion that NPLs could be considered as the one root cause of banking crisis According International Monetary Fund working paper, basing on the NPLs

in Central, Eastern and South Eastern Europe, the research indicates that strong feedback of macroeconomic condition including GDP growth, unemployment and inflation on NPLs (Klein, 2013) The econometric result suggests GDP growth is one of the macro explanatory of NPLs Besides that, the significant linkage between macroeconomic condition and NPLs is also supported by the

investigation of determinant of NPLs of 85 banks in three countries including Italy, Greece and Spain (Messai & Jouini, 2013) However, this approach does not consider the effect of banking specific variables that illustrate the characteristic of each bank, which generates different effect on the risk exposure at the bank level

On the other hand, some empirical studies attempt to find out the linkage between bank-specific variables and NPLs including bank capitalization, bank profitability, bank regulation, etc This approach is more powerful in explanation of difference of banking NPLs For instance, using the aggregate banking data from 59 countries, internal factor including the capital adequacy ratio, prudent provisioning policy, private or foreign ownership, strengthening the legal system have significant impact on banks’ NPLs (Boudriga, Taktak, & Jellouli, 2009) Moreover, the insolvency of financial institution is also the result of high NPLs (Farhan, Sattar, Chaudhry, & Khalil, 2012) In addition, other study attempts to find out impact of ownership status or market power on NPLs It generally accepted that NPLs associated with the inefficiency, failures of the banks in the financial crisis period (Ahmad & Bashir, 2013)

Other approach to examine NPLs’ determinant is analyzing the effect of both macroeconomic and bank-specific factors on NPLs In particular, the macroeconomic and microeconomic factors are combined to examine the NPLs of commercial and saving bank in Spain It concludes that all macroeconomic and microeconomic factors have specific effect on NPLs (Salas & Saurina, 2002) Using the data of Greek banking system, the empirical study combines both macroeconomic and bank-specific factors to assess NPLs’ determinant This study finds out that bank-specific factors have a different impact on NPLs of different loan categories including mortgage, business and consumer loan portfolios (Louzis, Vouldis, & Metaxas, 2011)

Government intervention and foreign investment are also considered as the endogenous variables that affect to NPLs Some arguments show that government intervention play important role to manage economic in which market failure are balanced (Garcıa-Marco & Robles-Fernandez, 2008) Other arguments supported for private-sector monitoring hypothesis Regarding foreign investment, it is general accepted that bank will get advantages from experience of management as well as capital from foreign investment However, its effect varies in different studies

In summary, the financial problem raise more concern in the NPLs in recent years The determinants

of NPLs are examined in many empirical studies However, the determinants of NPLs in the case of

Vietnamese banking sector are not examined Therefore, this study will examine the NPLs’ determinants in the case of Vietnamese banking sector

1.3 Research objectives and research question

1.3.1 Research objectives

As discussion above, the main purpose of this study is to examine the determinants of NPLs The unbalanced panel data of 30 Vietnamese banks over the period of 2008-2012 is used in this study Both macroeconomic and bank-specific factors are employed in order to model the NPLs’ determinant In particular, this study will examine the effect of exogenous variables including GDP growth, unemployment rate, lending interest rate and sovereign debt on NPLs The endogenous variables including return on equity, inefficiency rate, non-interest rate, leverage ratio and credit growth are also examined In addition, the effect of government intervention and foreign investment

on NPLs is investigated by assessing the difference of NPLs in state-owned bank and fully owned bank In finally, the policy implication for NPLs solving is suggested after examining the regression results

foreign-1.3.2 Research questions

According to the research objectives, this study will attempt to answer following research questions The first question is which factors will affect on the NPLs The second question is how they affect on NPLs The third question is what the cause of these effect And the final question is which policy applicant could be raise from analyzing the effect of these factors

The rest of study will be arranged as follows Chapter 2 briefly presents the theories and empirical studies regarding NPLs’ determinant In this part, specific influence of each factor on NPLs will be analyzed basing analyzing the result of previous studies Chapter 3 will provide methodology analysis

of previous empirical literature This part will give overview of all methodologies were applied in previous study and suitable mythology will be selected to analyze NPLs’ determinants in Vietnamese banking sector Detailed data and data sources are also presented in this part Next chapter will present the analysis results The descriptive statistic as well as economic results is provided in this part This

part also provides regression explanation and comparison with expectation of literature review The conclusion as well as policy implication will be presented in final chapter Chapter 5 also provides research limitation as well as guideline for future studies

CHAPTER 2: LITERATURE REVIEW

2.1.Non-performing loans definition

Non-performing loans are loans either in default or close to being in default It means that the borrower cannot pay the loan back in full In generally, three kinds of debts could be defined as NPLs Firstly, debts whose interest and principal are past due by 90 days or more compared with stipulated time governed in credit contract Secondly, at least 90 days of interest payments have been capitalized, refinanced or delayed by agreement Finally, payments are less than 90 days overdue, but there are other good reasons to doubt that payments will not be made in full In particular, the loan is considered

as NPLs if they belong to following exposures (Basel III, 2011) Firstly, all exposures are classified

as the defaulted or impaired loans in which loans experience with the deterioration of their creditworthiness Secondly, other exposures have more than 90 days past due Thirdly, one exposure could be considered as the NPL if there is evidence that the customer could not fully pay principal or interest

NPLs definition used in empirical researches is consistent Louzizs at el (2011) used to dataset of Greek banks to analyze the impact of NPLs According that, NPLs refer to loan which are 90 days past due Basing on the study of Louzis at el (2011), Klein (2013) employed the NPLs of Central, Eastern and South-Eastern Europe to investigate their determinants in which NPLs is defined as the loan with 90 days past due

In Vietnamese banking sector, definition of NPLs is nearly the same with international cases The NPLs is the loan is classified as Group 3 to Group 5 (Decision 493/2005/QD_NHNN, 2005) As stipulated in loan classification regulation, following debts are classified into group 3 to group 5 Firstly, NPLs are debts overdue for a period of more than 90 days Secondly, debts are restructured and extended payment term Thirdly, loan issuing to customer who is not allowed or restricted to get loan as regulation is considered to classify as NPLs In addition, loans must be classified to higher group if there is evidence of disadvantage change in environment or business that negatively effect to payment ability of customer

In summary, NPLs in this study is understand as the NPLs rate which is rate of non-performing loans over the gross loans NPLs used in this study will based on the regulation of Decision 493/2005/QD_NHNN and other empirical studies in which non-performing loans is loans with 90

days past due Next sections will continue to examine the effect macroeconomic and bank-specific factors on NPLs by reviewing the discussion in theory and previous empirical studies

2.2 Bank-specific determinants of non-performing loans

Besides major studies investigating the effect of macroeconomic factors on NPLs, fewer empirical research attempt to analyze the effect of bank-specific factors on NPLs While macroeconomic factors are reflected as exogenous for bank performance, bank-specific is the endogenous factors which directly effect on credit exposure Difference in bank regulation, bank capacity as well as profit enhance will generate different change in NPLs In general, moral hazard, operating efficiency, loan diversification, banking leverage, credit policy, etc is one of the bank-specific factors which are usually used to analyze NPLs Following is the discussion regarding effect of specific factors to NPLs

by analyzing relevant hypotheses in previous empirical studies

2.2.1 ‘Bad management’ and ‘Skimping’ hypothesis

According to ‘bad management’ hypothesis, cost efficiency is considered as the endogenous factor effecting to NPLs This hypothesis suggests that NPLs will expand in bank with low cost efficiency (Berger & DeYoung, 1997) Problem loans strongly links with the management which is reflected in internal control system Furthermore, bank with bad management do not have enough skill to manage bank’s operation as well as credit risk According that, poor skill in risk definition, risk assessment will perform in poor skill of defining risk policy, underwriting, collection as well as problem loans solving

In contrast with ‘bad management’ hypothesis, the ‘skimping’ hypothesis suggests that the high cost efficiency is associates with the NPLs’ increase According this hypothesis, resources is allocated for management including underwriting and monitoring loans which are traded off with cost efficiency

In this assumption, it is expected that the bank with high cost efficiency will lead to NPLs expansion

in future by less focusing on loan monitoring including underwriting, debt collection, credit scoring, etc (Berger & DeYoung, 1997)

Employing Ganger-causality technique, Berger and DeYoung (1997) proposes that NPLs associate with cost efficiency Using U.S commercial banks in 1985 and 1994, the data and regression result

confirm the strong linkage between cost efficiency and NPLs According that, the increase in future NPLs is the consequence of low cost efficiency that is consistent with ‘bad management’ hypothesis

In addition, ‘bad management’ hypothesis is also supported by the research in Czech banks in 1994 and 2005 Applying GMM model, this research finds that NPLs is forecasted by cost efficiency’s deterioration (Podpiera & Weill, 2008) However, the study of Spanish banks over the period 1985-

1997 find out cost efficiency in statistically affects on problem loans that do not support for hypothesis estimation (Salas & Saurina, 2002) Basing on aforementioned research, ‘bad management’ hypothesis is considered to analyze bank-specific determinants of NPLs in mortgage, business and consumer loans in Greek banking system (Louzis, Vouldis, & Metaxas, 2011) The inefficiency rate

is use as proxy for this hypothesis The regression gives result that the linkage between cost efficiency and NPLs is statistically significant According that, this study suggests that one of leading indicators

of the increase in NPLs is the low cost efficiency This conclusion is quite consistent with ‘bad management’ hypothesis and the results of large of studies In sum, the ‘bad management’ hypothesis

is more supported by different empirical studies Therefore, this hypothesis will be continued to examine by measure the effect of inefficiency of bank on NPLs

Hypothesis 1: Higher cost inefficiency in bank operations is associated with higher NPLs

Many studies attempts to analyze effect of diversification on NPLs, however, the proxy for this variable distinguish in each research First of all, size of bank is used as the proxy for diversification According this approach, more diversification opportunities are driven by bigger size of bank which allows reducing NPLs (Salas & Saurina, 2002) The research in India is also supported this argument

which bigger size will reduce NPLs (Ranjan & Dhal, 2003) Secondly, the entropy index regarding share of different revenue is set up to analyze the diversification hypothesis However, the result shows statistically insignificant effect of diversification on NPLs (Hu, Li, & Chiu, 2004) Income growth is used to consider verifying impact of benefit from diversification on NPLs Because of high correlation between income growth and net interest income, diversification’s benefit does not generate any effect

on NPLs reduction (Stiroh, 2004) In the other hand, proportion of non-interest income is counted when modeling NPLs determinant (Louzis, Vouldis, & Metaxas, 2011) On that ground, this ratio will reflect the dependence of income in interest rate income This study suggests that NPLs negatively related to banks size as well as the proportion of non-interest income over total income As the aforementioned discussion, this study will employ the proportion of non-interest income over the total income to examine the effect of diversification on NPLs

Hypothesis 2: Higher non-interest income ratio is associated with lower NPLs

2.2.3 ‘Moral hazard’ hypothesis and ‘Too big to fail’ hypothesis

In ‘moral hazard’ hypothesis, the moral management is considered as variables which effecting to NPLs According that, it is assumed that the increase in NPLs is the consequence of bank’s low-capitalization (Berger & DeYoung, 1997) The moral hazard incentive will vary in different bank that

is influenced by bank’s manager This hypothesis assumes that the riskiness of loan portfolio expanded

by bank’s manager because of thin capital Furthermore, the hypothesis suggests that there is excessive risk taking in low-capitalization bank As the result, credit risk’s increase will associate with problem loans expansion that increases future NPLs

Salas and Saurina (2002) strongly confirm this hypothesis The lagged of solvency ratio is used to analyze the moral hazard hypothesis As the result, lagged solvency of bank is significant negative with NPLs that is consistent with moral hazard hypothesis In addition, moral hazard hypothesis is tested in the case of Greek banks However, the increase of solvency ratio - a proxy for moral hazard hypothesis insignificantly effects on NPLs declining (Louzis, Vouldis, & Metaxas, 2011)

‘Too big to fail’ hypothesis is based on the moral hazard problem of key banks in the economy According that, bank is supported by other institutions in case of incident, there is less effort for defend and recover the risk (Stern & Feldman, 2004) This hypothesis assumes that moral hazard problem will maintain key banks having many customers or playing a key role in the banking system of one

country The liquidity and solvency of key banks will strong link with other bank in the economy Because of their nature, the failure of this bank could create domino effects which continuously effecting to other banks and their creditors The failure in whole of banking system is beginning if there is no prevention of institutions Therefore, government usually plays a role to support and maintain the operation of key banks As the consequence, the moral hazard problem happens in which risk prevention is not strongly prevented The banks have a probability to accept excessive risk and issue loan for lower quality customer that increases future problem loans In sum, according too big

to fail hypothesis, the increase in NPLs could be driven from the moral hazard problem in the large bank

This hypothesis is not clearly supported by empirical study According the study regarding U.S banks, the research suggests that riskier portfolio in large bank is motivated by U.S government in 1980s which supports for too big to fail hypothesis (Boyd & Gertler, 1994) In the other hand, the study analyze to US bank performance over the period 1983-2003 by investigate size classes do not give evidence for this hypothesis (Ennis & Malek, 2005) This hypothesis is also applied to analyze NPLs’ determinant in Greek banks This hypothesis is strong supported at all loan categories including mortgage, business loans (Louzis, Vouldis, & Metaxas, 2011) However, in the case of consumer

loans, this hypothesis is not supported

In sum, both ‘moral hazard’ and ‘too big to fail’ hypothesis examine the effect of moral management

on NPLs According these hypotheses, the lower solvency ratio associates with higher leverage ratio that lead to an increase the future NPLs

Hypothesis 3: Higher leverage ratio is associated with higher NPLs

2.2.4 ‘Bad management II’ and ‘Pro-cyclical credit policy’ hypothesis

‘Bad management II’ hypothesis suggests that bank performance is considered as the proxy for management skill in lending activities Lower quality of skill management in lending activities associates with low performance that will deteriorate future loans Therefore, past performance or earnings negatively link with NPLs (Louzis, Vouldis, & Metaxas, 2011) In addition, bank with high profitability, which is not under pressure increasing loan portfolio and profitability, will be more careful when assessing new loan and risk exposure reduction will belong Using return on equity as proxy for bank profitability, Godlewski (2004) reports the negative relationship between return on

equity and NPLs This indicates that higher bank profitability will be lower NPLs (Godlewski, 2004) Furthermore, ‘bad management II’ hypothesis is applied to consider NPL’s determinant in Greek banking system The return on equity is used as the proxy for this hypothesis The regression result suggests that there is negative relationship between NPLs and earnings in the case of mortgage loans (Louzis, Vouldis, & Metaxas, 2011)

In contrast, ‘pro-cyclical credit policy’ hypothesis suggests distinguish assumption with ‘bad management II’ hypothesis According that, positive relationship between current bank performance and future NPLs is expected that supports for liberal credit policy According to the model of Rajan (2004), the credit policy is affected by earning expectation as well as management’s concerns regarding short-term reputation As the consequence, to increase bank’s profitability, the current earnings and future problem loans are distorted and inflated Loan loss provision is also used to adjust current earnings Therefore, future NPLs positively links with past earnings This assumption is consistent with the result when analyzing risk taking behavior and ownership in the Spanish banks This empirical study argues that higher bank profitability will associated with higher NPLs (Garcıa-Marco & Robles-Fernandez, 2008) In the other side, relationship between bank profitability and NPLs is not supported when lagged return on asset is used as proxy of bank profitability The researchers argue that return on asset is appropriate when applied in firm level instead of country level (Boudriga, Taktak, & Jellouli, 2009)

As aforementioned empirical studies, most of studies supported for ‘bad management II’ in which bank with high profitability will be more carefully in granting credit that lead to NPLs reduction

Hypothesis 4: Profitability negatively related with lower NPLs

2.2.5 Credit growth

In this hypothesis, the credit growth in banking sector is analyzed to find out their effect on NPLs Credit growth is the increasing rate of banking credit loan which reflects the speed of credit growth According that, it assumed that rapid growth in credit loan will effect on quality of risk control Because of large credit loan assessment, the quality of underwriting as well as credit loans is not ensured, which enhance risk exposure and NPLs in the future

Many recent empirical studies give clear evidence that is consistent with this hypothesis Using data

of Argentine banking sector, the study suggests there is strong linkage between credit growth and

impaired loans The study make conclusion that impaired loan and credit growth are relevant (Bercoff, Giovanni, & Grimard, 2002) In addition, rapid past credit or branch expansion is associated with NPLs Using the dataset of Spanish problem loans of both commercial and saving banks, the research concludes that rapid credit growth associated with problem loans (Salas & Saurina, 2002) Furthermore, Jimenez and Saurina (2006) suggest the positive linkage between credit growth and impaired loans The research concludes that lagged there credit growth positively effects to the loan losses It explained that low quality of customer and mortgage loans declining increase in the period

of economy downturn According that, the credit is more risky which affecting loan losses (Jimenez

& Saurina, 2006) This conclusion is also affirmed by other researches Khemraj and Pasha (2009) used the dataset of Guyanese banking system to analyze the relationship excessive lending and NPLs Using the panel data with fixed effect model, the regression result is consistent with those of Jimenez and Saurina (2006) According that, excessive lend could generate the likelihood of higher NPLs (Jimenez & Pasha, 2009)

In the contrast, high credit growth could be considered as high profitability in a certain period With high bank profitability, bank is not under pressure to spread out rapidly their market that affects on credit quality Ahmad and Bashir (2013) supported the negative relationship between credit growth and NPLs They argued that large bank would diversify the loan portfolio and reduce risk by increasing their market However, this dimension is rarely supported

Hypothesis 3: Credit growth is positively associated with higher NPLs

2.3 Macroeconomic determinants of non-performing loans

Reviewing the empirical studies regarding NPLs’ determinant, the major of study assess NPLs at the aggregate level by investigating macroeconomic environment GDP growth rate, unemployment and lending interest rate are general investigated when modeling macroeconomic determinants of NPLs

2.3.1 Economic growth

Many previous studies confirm the linkage between NPLs and business cycle GDP growth gets the negative effect to the NPLs rate (Salas & Saurina, 2002) GDP growth rate and other macroeconomic factors such as family indebtedness, rapid past credit are taken into account to explain the credit risk

in Spanish bank over the period 1985-1997 Two categories of banks are taken into account to analyze the determinant of problem loans including commercial and saving banks The result shows that NPLs

is negatively affected by GDP growth rate This is explained that the ability to serve the debt including problem debt is improved by macroeconomic development (Salas & Saurina, 2002) As the same expectation of previous studies, the NPLs of Italian banks are largely affected by the business cycle over the period 1985-2002 (Quagliariello, 2007) This study suggests that bad debt as well as loan loss

is tentative to be low in period of rapid growth As the result, the research confirms that the revolution have significant impact on new bad debts (Quagliariello, 2007)

Furthermore, the effect of business cycle on credit default is affirmed by analyzing the relationship between production cycle and credit default in Turkish financial system (Ciftera, Yilmazerb, & Cifter1, 2009) At the different time scale over the period 2001-2007, this study finds that production cycle generated impact on NPLs and the effect vary in different levels Reviewing the data of 26 advanced countries over the period 1998-2009, the result show the strong linkage between NPLs and macroeconomic exposure (Nkusu, 2011) This study finds that the macroeconomic performance is vulnerable by sharp increase in NPLs Nkusu (2011) also points out the key indicator of macroeconomic performance is GDP growth which effected to NPLs

Based on previous study regarding the determinants of NPLs, the study of NPLs in Greece affirmed the macroeconomic impact on NPLs The result shows that GDP growth rate mainly explains the NPLs of all loan categories in Greek bank (Louzis, Vouldis, & Metaxas, 2011) Using dynamic panel data of Greek banks database, this study attempts to combine both macroeconomic and bank-specific factors to analyze NPLs of mortgage, business as well as consumer loan portfolio The research suggests that all GDP growth rate statistically impacts on NPLs However, the quantity effect to different category of loans is not consistent The mortgage and business loans are less sensitive to the change of macroeconomic factors compared with consumer loans As regression result, the increase

of GDP growth rate is associated with NPLs declining It could be explained that the slowdown of economic generate negative effect on NPLs

In addition, basing on the NPLs in Central, Eastern and South Eastern Europe, the research indicates that strong feedback of macroeconomic condition including GDP growth, unemployment and inflation

on NPLs (Klein, 2013) The econometric result suggests GDP growth is one of the macro explanatory

of NPLs Besides that, the significant linkage between macroeconomic condition and NPLs of bank

is also supported by the investigation of determinant of NPLs of 85 banks in three countries including Italy, Greece and Spain (Messai & Jouini, 2013)

Hypothesis 6: Economic growth negatively related with lower NPLs

2.3.2 Unemployment

Unemployment is other primary contribution in the increase of NPLs Many previous studies confirm the linkage between NPLs and unemployment rate Current income as well as unemployment rate effect could be considered as the probability of credit default (Rinaldi & Sanchis-Arellano, 2006) Using dataset of seven euro areas over the period 1989-2004, this study finds that unpredictability of future income is the consequence of current income as well as unemployment, which effect on the likelihood of credit default In addition, Berge and Boye (2007) conclude that unemployment is highly sensitive with the problem loans The NPLs in household as well as enterprise sector in the Norges bank is respectively investigated They find out that one of primary contribution on unemployment rate improvement is the problem loans’ declining (Berge & Boye, 2007)

Furthermore, reviewing the data of 26 advanced countries over the period 1998-2009, the result show the strong linkage between NPLs and macroeconomic factors (Nkusu, 2011) This study points out the key indicator of macroeconomic performance including unemployment effect to NPLs The linkage of unemployment rate and NPLs is confirmed when analyzing panel data of Greek banks The result shows that unemployment rate mainly explains the NPLs of all loan categories in Greek bank (Louzis, Vouldis, & Metaxas, 2011) According to regression result, the NPLs are also positively affected by unemployment rate It could be explained that NPLs is reduced when unemployment rate declined and customer had enough capacity to pay the overdue debts Basing on the NPLs in Central, Eastern and South Eastern Europe, Klein (2013) indicates strong feedback of macroeconomic condition including unemployment and inflation on NPLs The econometric result suggests that unemployment rate is one of the macro explanatory of NPLs

Hypothesis 7: The higher unemployment rate is associated with higher NPLs

2.3.3 Lending interest

Besides other macroeconomic factors, lending interest rate is other macroeconomic factors affecting

to NPLs Many previous studies confirm the linkage between NPLs and lending interest rate This study concludes that riskier financial position is set up for household having debt increase (Rinaldi & Sanchis-Arellano, 2006) In addition, Berge and Boye (2007) conclude that real interest rate is highly sensitive with the problem loans The NPLs in household as well as enterprise sector in the Norges bank is respectively investigated They find out that one of primary contribution in real interest rate is the problem loans’ declining (Berge & Boye, 2007)

Based on previous study regarding the determinants of NPLs, the study of NPLs in Greece affirmed the effect of lending interest rate on NPLs of all loan categories in Greek bank (Louzis, Vouldis, & Metaxas, 2011) According that, the NPLs are positively affected by real lending rate It could be explained that higher lending interest rate is usually charged for riskier loans that have more ability to debt default

Hypothesis 8: Higher lending interest rate is associated with higher NPLs

2.3.4 Sovereign debt

Sovereign debt plays an important role in investigating NPLs’ determinant, especially after recent financial crisis There are two effects of sovereign debt on banking system Firstly, because of the public finance failure, market evaluation ‘ceiling’ is set up As the consequence, the bank’s liquidity

is affected in which lending is decreasing and debtors could not be able to finance their debt This lead

to credit default in the banking system In addition, fiscal measures are applied in the case of high sovereign debt As the consequence, the social expenditure and wage for government are cut Affecting by these measures, the debtors are shocked and could not be able to serve their debts which increase future NPLs Therefore, it is expected that sovereign debt will increase future NPLs

Many studies give evidence for the linkage between sovereign debt and financial crisis According to the regression result, this empirical study concludes that financial crisis leads to sovereign debt (Reinhart & Rogoff, 2011) Based on previous study, Louzis et al (2011) confirm the effect of government debt on NPLs by investigating all loan categories in Greek banks This study uses ratio

of central government debt over the nominal GDP as the proxy for sovereign debt According to regression, the results show that sovereign debt statistically effect to NPLs of all loan categories including mortgage, business and consumer loan portfolio

Hypothesis 9: Sovereign debt positively related with higher NPLs

2.4 Government intervention and foreign investment in banking system

2.4.1 Government intervention

Government intervention plays important role in bank’s performance by controlling regulation as well

as credit policy The effect of government intervention in bank varies in different studies In the one hand, government intervention could be offset the market failure, particularly in Pigouvian where the monopoly as well as the information asymmetries exist As the consequence, government intervention play as important role to manage economic in which market failures are balanced and social welfare

is strengthened In addition, other empirical researches suggest that state-owned bank with better regulation will reduce more NPLs compared with private-sector monitoring By examining the relationship between risk and ownership structure, Garcia-Marco and Robles-Fernandez (2008) document that commercial banks are more risk exposure compared with state-owned banks

In the other hand, some argument suggests that the private-sector monitoring in which government intervention is excluded will be more effective In this case, the private-sector is allowed to control bank regulation which no more effected by government intervention With profitability enhance, private-sector is more incentive to control their profitability, cost efficiency which reduce NPLs Moreover, others suggest that state-owned banks are incentive to finance riskier project in order to finance for public good and country’s economic development As the consequence, that lead to higher NPLs Using dataset regarding bank regulation and supervisory of 107 countries, Barth et al (2004) argues that empower the private-sector to monitoring of banks are associated with better banking soundness including outcomes, higher development as well as small NPLs This argument is strongly confirmed by regression result According that, the private-sector is more effective when controlling NPLs compared with government intervention (Barth, Jr., & Levine, 2004) The role of private-sector monitoring is also confirmed in the report of Boudriga et al (2009) This paper indicates that due to weaker credit recovery capacities compared with private monitoring, state-owned banks tend to increase NPLs (Boudriga, Taktak, & Jellouli, 2009) As the same result, the investigating ownership reform in China confirms the role of private monitoring in generating better performance and NPLs reduction (Lin & Zhang, 2009) According aforementioned researches, this study will investigate the effect of government intervention in NPLs, particular in state-owned banks

Hypothesis 10: State-ownership is positively associated with higher NPLs

2.4.2 Foreign investment

Foreign investment generates negative effect in banks’ performing, particularly in NPLs The bank efficiency seems to be higher in bank with higher foreign investment ratio compared with domestic ownership Bank with foreign investment is usually get advantages from experience of management

as well as capital from foreign investment With their advantages, the strictly internal control system

is usually set in foreign investment bank that reduces credit exposure In addition, because of capacity independence, foreign investment bank is rarely effected by profitability enhance in which control points are loosen to archive sales target In particular, developing countries get experience of human capital improvement including skill and technology from foreign investment company (Lensink & Hermes, 2004) All of that will increase banks’ soundness that positively encourages reducing bad loans

This hypothesis is supported by many empirical studies Barth et al (2004) investigate the effect of foreign bank entry limitation on bank performance by counting the number of bank entries to economy

in investigating period According to regression results, the study finds that the bank fragility positively links with the limitation in foreign investment entry Besides that, poor bank development

is associated foreign investment entry limitation In conclusion, the effect of foreign investment varies

in different examinations Therefore, foreign investment is applied in this study to analyze their effect

on bank performance, particularly in NPLs Besides that, the study also attempts to analyze the distinction of NPLs’ determinant in bank with direct foreign investment and the others

Hypothesis 11: Full foreign ownership negatively related with lower NPLs

Basing on aforementioned theories and empirical studies, these researches will consider both macroeconomic and bank-specific factors in modeling NPLs’ determinants in different type of banks All macroeconomic and bank-specific factors used to modeling NPLs’ determinants are summarized

as following table

Table 1: Definition of variables used in modeling NPLs determinants

Variables Definition Expected

sign Hypothesis tested

Inefficiency Operating expense/ operating income + Bad management (+) Non-interest income Noninterest income/ Total income - Diversification (-) Leverage ratio Total liability/ total assets + Too big to fail (+) Return on equity Profit/ Total equity - Bad management II (-) Credit growth (Loant - Loan t-1)/ Loant-1 + Credit growth (+) Economic growth GDP growth rate - Economic growth (-)

Lending rate Real lending interest rate + Lending rate (+) Government debt Central Government debts/ Nominal

GDP

+ Sovereign debt (+)

CHAPTER 3: METHODOLOGY AND DATA

3.1 Methodology

Panel data is usually employed in recent empirical studies regarding to modeling the regression of NPLs’ determinants Because of panel data advantages, combining both time series and cross-section data, panel data increase number of observations and degree of freedom This method will reduce co-linearity among variables and reduce risk in omitting variables Therefore, simple pooled regression may not be well designed to examine NPLs and its determinant (Boudriga, Taktak, & Jellouli, 2009) Lin and Zhang (2009) employed year fixed effect to control bank size, change in market and regulatory conditions over the year In order to analyze the determinants and impact of NPLs in macroeconomic performance in Europe, Nir Klein (2003) also used panel data of countries in Europe from 1998 y

2001 In additional, Ahlem Selma Messai and Fathi Jouini (2013) also used panel data of 85 banks from Italy, Greece and Spain period from 2004 to 2008 to assess macro and micro determinants of NPLs Therefore, unbalanced panel data of 30 Vietnamese banks over the period 2008 – 2012 is applied to model the regression of NPLs’ determinants

According to recent researches regarding NPLs’ determinant, dynamic panel regression is employed

to analyze effect of factors in NPLs Salas and Saurina (2002) applied dynamic approach to investigate problem loans in commercial and saving banks in Spanish In addition, Klein (2013) applied the dynamic panel regression to analyze macroeconomic determinant of NPLs Furthermore, dynamic panel regression is widely applied when investigate macroeconomic and bank-specific determinant of NPLs in Greek banks (Louzis, Vouldis, & Metaxas, 2011) The specification of a dynamic panel data

is presented as following:

y it = αy it-1 + βX it-1 + ε it (Eq 1)

where cross section and time are denoted by the subscript i and t respectively yit denotes for NPLs change of observations Firstly, first lag of dependent variable (yit-1) is counted to explain dependent variable Second, Xit-1 is other independent variables According to empirical studies, it is important

to count lagged variables into regression model It is explained that NPLs is the consequence of change

of independent variables However, this impact is not immediately affected and time is necessary to count in the model Therefore, all lagged variables are applied in order to count accurate effect of independent variables on NPLs Finally, εit presents for error term in equation

Applying dynamic panel regression for testing effect of macroeconomic factors, economic specific of model is presented of follows:

NPLs it = NPLs it-1 + β 1 GDP it-1 + β 2 UN it-1 + β 3 RLR it-1 + ε it (Eq 2)

In Equation 2, i is denoted for cross section which present for each banks in panel dataset t will presents for time series from 2008 to 2012 NPLs it-1 is the first lag NPLs GDP it-1, UNit-1 and RLR it-

1 are the first lag of GDP growth rate, unemployment rate and real lending interest rate respectively For testing the ‘sovereign’ hypothesis, lag of government debt is counted into regression to analyze their effects on NPLs Finally, bank-specific factor is counted into regression to analyze their effects Lagged variables are also used when modeling the regression The model could be present as follows:

NPLs it = NPLs it-1 + β 1 GDP it-1 + β 2 UN it-1 + β 3 RLR it-1 + β 4 Debt it-1 + β 5 ROE it-1 + β 6 INEF it-1 +

β 7 NII it-1 + β 8 LR it-1 + β 9 CRE it-1 + ε it (Eq 3)

in which Xit-1 is the lagged variable of bank-specific factors as discussion in aforementioned empirical

studies which is summarized in Table 1:

 ROEit-1: return on equity of bank i year t-1

 INEFit-1: cost inefficiency of bank i year t-1

 NIIit-1: proportion of non-interest income on total income of bank i year t-1

 LRit-1: total liability over total asset of bank i year t-1

 CREit-1: credit growth of loan portfolio of bank i year t-1

Finally, dummy variables are added into Equation 3 for testing effect of state ownership and foreign

investment There are two dummy variables presented as follows:

 STATE variable: state-owned bank is 1 and others is 0

 FOREIGN variable: bank with 100% foreign capital is 1 and others is 0

To analyze effect of these variables on bank-specific factors, dummy variables are generated by multiple dummy variables including STATE and FOREIGN with other bank-specific factors

In addition, some fundamental tests are employed to examine variables including correlation among variables, multi-co linearity and stationary

Multicolinearity is the problem of model in which there is one or more relationship among the regressors As the consequence of multicolinearity, the model exist the large confidence interval in

which the null hypothesis may not be rejected In addition, the t ratio may be statistically insignificant

To test the co linearity, the variance-inflating factor (VIF) is used to measure the degree of estimators

inflated by colinearity

If the means and variance of time series is constant over the time, this time series is said to be stationary In addition, the covariance between two time periods depends only on the distance between the two time periods If time series is non-stationary, the model result is less practical value for forecasting purpose In the case of non-stationary time series, the model will obtain high value of R2and significant t However, the result is unreliable To test the stationary for times series, the unit root test is applied In the unit root test, null hypothesis is said that there is unit root in model The stationary time series is the alternative hypothesis This study will apply the Dickey-Fuller test to check stationary of model

3.2 Data

As the same situation with international banking system, from 2008 to 2012, Vietnam experienced with a period of the housing bubble and rapid growth in the stock market Vietnamese banking sector had to face with the credit exposure when economy went down The NPLs rapidly increased and under controlled Therefore, this study will use data of Vietnamese banking sector from 2008 to 2012 in order to evaluate NPLs under the period of international financial crisis and economic depression in Vietnam

Macroeconomic factors including GDP growth rate, unemployment rate, real lending interest rate and government debt will be collected from World Bank database These indicators over the periods 2008 – 2012 is used to analyze the effect of macroeconomic factors on NPLs in period of economic downturn According to previous empirical studies, many data sources were employed to extract banking data including Bankscope database, Financial Soundness Indicators (FSI) data drawn from the IMF (2007), the World Bank database, Federal Reserve database However, bank-specific factors are presented in financial statements which are available in Bankscope database The financial statement in Bankscope database is generally prepared based on two standards Firstly, local generally accepted accounting principle (local GAAP) is applied to prepare financial statements Local GAAP regulated by government varies in different countries The other standard is International Financial Reporting Standard (IFRS) Reviewing data of Vietnamese banks, most of financial statements are

prepared based on local GAAP In sum, to analyze the effect bank-specific factors, this study will use information in financial statements of Vietnamese banks that extracted from Bankscope database Unbalanced panel data of 30 Vietnamese banks over the period 2008 – 2012 are used in order to analyze determinant of NPLs In order to test influence of government intervention, four large state-owned commercial banks that count of 13% total sample are included in this study In Vietnamese banking sector, commercial banks have to follow regulation issued by State bank of Vietnam According that, government will intervene in the financial institutions However, in state-owned bank, the government has more power bank operations Besides supplying financial services, state-owned bank plays an important role to adjust the economy Therefore, the effect of government intervention

in state-owned banks will directly effect on bank-specific factors that impacts to NPLs Moreover, there are only 6 fully foreign-owned banks which count of 20% total sample are included in data in order to check effect of foreign investment in Vietnamese banking sector All reaming banks are joint-stock commercial banks There are 20 joint-stock commercial banks that are equivalent 67% of total sample Detail of banks are used is described in Appendix 1 Besides that, detailed calculation of variables is described in Table 2

Table 2: Specific calculation of variables

Variables Specific calculation Unit Explanation

GDP GDP growth rate % The ratio is used to measure the growth

rate of the economy

UN Unemployment rate % This ratio measure the rate of unemployed

individuals in labor force

RLR Real lending interest rate % The amount is charged to borrower for the

use of assets

DEBT Central Government debts/

nominal GDP

% The amount of money that is borrowed by

one country’s government

ROE Profit/ Total equity % This ratio measures the operating

efficiency of one company

INEF Operating expense/ operating

Variables Specific calculation Unit Explanation

 Operating income: earnings before interest and tax

NII Noninterest income/ Total

income

% The ratio measure the percentage of

non-interest income over the total of gross revenue

LR Total liability/ total assets This ratio measure the ability to finance

of variables VIF test is employed to check multicolinearity in regression If the VIF equal 1, there is

no correlated among variables in regression If the VIF run from 1 to 5, the moderately correlated exists in the regression In the contrast, there is highly correlated in regression if VIF larger than 5 In addition, the unit root test of Dickey Fuller is applied to test the stationary in variables If the variables are stationary, we will continue analyze the effect of each factor on NPLs

The specific test is applied to find out appropriate model to analyze NPLs determinant According to theory, there are two general methodologies to test panel data including fixed effect model and random effect model According to the fixed effect model, the true effect size of all samples is similar and consistent Besides, the fixed effect model assumes that only the sampling error is generate the difference of effect size Therefore, due to the better information having from effect size of larger studies, the information of small studies is could be ignored In the other sides, random effect assumes that it does not exist one true effect model for all studies and each study will generate different effect size Therefore, according to random effect, all studies are counted when estimating the mean of distribution effect even small studies with small weight in samples This is completely different with assumption of fixed effect model According to characteristic of data, fixed effect or random effect model will be applied to test panel data In this study, fixed effect model is applied to test effect of

each variable on NPLs Detail of methodology is presented in Table 3

Table 3: Methodology test

Inefficiency + Bad management (+) Test coefficient β 6 in Eq.3

Non-interest income - Diversification (-) Test coefficient β 7 in Eq.3

Leverage ratio + Too big to fail (+) Test coefficient β 8 in Eq.3

Return on equity - Bad management II (-) Test coefficient β 5 in Eq.3

Credit growth + Credit growth (+) Test coefficient β 9 in Eq.3

Economic growth - Economic growth (-) Test coefficient β 1 in Eq.3

Unemployment + Unemployment (+) Test coefficient β 2 in Eq.3

Lending rate + Lending rate (+) Test coefficient β 3 in Eq.3

Government debt + Sovereign debt (+) Test coefficient β 4 in Eq.3

CHAPTER 4: ANALYSIS RESULTS

4.1 Descriptive statistics

The yearly time series for NPLs at bank level are employed in this study NPLs of Vietnamese banking sector over the period 2008 – 2012 is used to analyze the bank-specific determinants In addition, macroeconomic factors are also counted in model of NPLs determinant The statistics of all variables

used when modeling determinant of NPLs is presented in Table 4

Table 4: Descriptive statistics

According to descriptive statistics, NPLs of sample is from 0.02% to 11.4% This gap reflects the significant distinguish in NPLs at bank level The mean of NPLs sample is 2.37% which indicates the average of NPLs is still at low level Unemployment rate as well as real lending interest rate is significant change over the period 2008 – 2012 which reflect the unstable in the economic In 2009, the unemployment rate is 2.6% This ratio sharply decreases in 2012 Real lending interest rate significantly fluctuate in this period which consistent with the fluctuation in NPLs over years In addition, the government debt increases twice time compared with previous period Reviewing the bank-specific determinants, these variables also give significant change in the bank level over the period of economic downturn It could be explained that, under the economic downturn, NPLs as well

as other factors are significant impacted