Macro economic determinants of credit risks in the asean banking system

1.5 Hypothesis of the study: This paper will examine the impacts of five macroeconomic factors to the NPLs rate, thus the five hypotheses are as follows: H1: Gross Domestic Product GDP

UNIVERSITY OF ECONOMICS ERASMUS UNVERSITY ROTTERDAM

HO CHI MINH CITY INSTITUTE OF SOCIAL STUDIES

VIETNAM THE NETHERLANDS

VIETNAM – THE NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

MACROECONOMIC DETERMINANTS OF

CREDIT RISK IN THE ASEAN BANKING

SYSTEM

BY

NGUYEN CHI THANH

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

HO CHI MINH CITY, DECEMBER 2016

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL STUDIES

HO CHI MINH CITY THE HAGUE

VIETNAM THE NETHERLANDS

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

MACRO ECONOMIC DETERMINANTS OF CREDIT RISK IN THE ASEAN BANKING

SYSTEM

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

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

By

NGUYEN CHI THANH

Academic Supervisor:

DR NGUYEN VU HONG THAI

HO CHI MINH CITY, DECEMBER 2016

DECLARATION

I declare that the wholly and mainly contents and the work presented in this thesis (Macro Economic Determinants of Credit risk in the ASEAN Banking System) are conducted by myself The work is based on my academic knowledge as well as my review of others’ works and resources, which is always given and mentioned in the reference lists This thesis has not been previously submitted for any degree or presented

to any academic board and has not been published to any sources I am hereby responsible for this thesis, the work and the results of my own original research

NGUYEN CHI THANH

ACKNOWLEDGEMENT

Here I would like to show my sincere expression of gratitude to thank my supervisor,

Dr Nguyen Vu Hong Thai for his dedicated guideline, understanding and supports during the making of this thesis His precious academic knowledge and ideas has motivated me for completing this thesis

Besides, I would like to express my appreciation to the lecturers and staff of the Vietnam – Netherlands Program at University of Economics Ho Chi Minh city for their willingness and priceless time to assist and give me opportunity for this thesis completion

Next, I would like to thank all of my classmates for their encouragement and their hard work, which become a good example for me to do the thesis I wish all of us will graduate at the same date

Lastly, I would like to express my love to my families for their unlimited supports which has led to the completion of this course research project

ASEAN: Association of Southeast Asian Nations

DGMM: the difference generalized method of the moments estimator

FE & RE: Fixed-effect and Random-effect estimator

GDP: Gross domestic product

NPLs: Non-performing loans

OECD: Organization for Economic Cooperation and Development

OLS: Ordinary Least Square

SGMM: the system generalized method of the moments estimator

The impact of credit risk, which is caused by the increase in the non-performing loans (NPLs), on the performance and stability of banking system as well as economic activities have recently raised many interests from researchers and policy makers Motivated by the close connection between the NPLs and macroeconomic environments as proposed by many researchers, this paper will empirically examine the determinants of non-performing loans in commercial banking systems of the five ASEAN countries in the period of 2002 to 2015 The research uses a sample of 162 banks in these countries with 11 variables of macroeconomic and bank-specific factors and applies the System Generalized Method of Moments estimator (SGMM) for dynamic panel models

The empirical results in this paper indicate that the movement of NPLs in the commercial banks of the five studied countries is associated with both macroeconomic variables and bank-specific factors For the macroeconomic condition, an increase in unemployment rate and the appreciation of domestic currency are found to significantly increase the NPLs In addition, bank with higher returns on asset and leverage ratio and low ratio of equity to total assets will have lower rate of NPLs Moreover, with the application of additional statistical analyses, the results indicate that the findings of the main model of this paper are consistent and robust

DECLARATION i

ACKNOWLEDGEMENT ii

ABBREVIATION iii

CONTENTS v

APPENDIX 1

LIST OF TABLES 2

CHAPTER 1: OVERVIEW OF RESEARCH 3

1 Introduction: 3

1.1 Backgrounds: 3

1.2 Problem statements: 4

1.3 Research objectives: 5

1.4 Research questions: 6

1.5 Hypothesis of the study: 6

1.6 The importance of research: 6

1.7 Structure of Research: 8

CHAPTER 2: LITERATURE REVIEWS 9

2.1 Theoretical reviews: 9

2.2 Empirical reviews: 13

2.3 Conclusion: 22

2.4 Research Hypothesis: 23

CHAPTER 3: DATA AND METHODOLOGY 27

3.1 Data collection: 27

3.2 Econometric methodology – The NPLs measurement: 28

3.3 The variables definition and measurement: 32

3.3.1 The dependent variable – the Non-performing loans: 32

3.3.2 Macroeconomic variables: 32

3.3.3 Microeconomic variables – bank-specific determinants: 34

3.4 Econometric strategy – The system GMM estimator: 38

CHAPTER 4: RESULTS AND DISCUSSIONs 40

4.1 Summary statistics: 40

4.2 Unit root tests: 41

4.3 Empirical results: 41

CHAPTER 5: OTHER ANALYSIS AND ROBUSTNESS CHECK 51

CHAPTER 6: CONCLUSION, POLICY IMPLICATIONS & LIMITATIONS OF THE REASEARCH 56

6.1 Main findings: 56

6.2 Policy implications: 57

6.3 Limitations: 58

6.4 Future research recommendation: 58

REFERENCES 59

APPENDIX 66

APPENDIX

Appendix 1: Number of banks in each country

Appendix 2: xtabond2 model selection criteria

Appendix 3: Correlation of variables

Appendix 4: Additional analyses and Robustness checks

Appendix 5: Additional analyses and Robustness checks

AP

LIST OF TABLES

Table 1: Description of variables

Table 2: Summary statistics

Table 3: Unit root tests for NPLs estimations variables

Table 4: Results with SGMM and fixed-effect estimations

CHAPTER 1: OVERVIEW OF RESEARCH

1 Introduction:

Banks are the financial intermediaries who play an important role in the development

of a country In the financial sector, a commercial bank is a funding channel, which can allocate the cash flows in the economy through their financial services as well as traditional services (taking deposits and make business loans) Whenever a loan is approved, banks gain profits from the borrowers by loan interest rate and services fees However, banks would expose to credit risk from this service because borrowers could suddenly lost their abilities to pay the loan in time, namely the non-performing loans (NPLs) The main reason for that comes from the movement of the macroeconomic environment, which directly impacts to the revenues and business activities of bank borrowers

Therefore, this paper will conduct an examination about how the economics determinants affect the bank credit risk In this chapter, the backgrounds, problem statements, research objectives, research questions, significance of the research and the layouts will be discuss around this issue

Along with the expansion of the economy as well as financial liberalization process in developing countries, the financial sector have been grown with surprising rate Besides, the improvements of technology and management procedures help banks making decisions to grow in financial markets However, the occurrences of two big economic recessions in 1997 and 2007 have significantly affected the banking systems

in developing countries It associated with the deteriorated quality of bank assets due to

a massive increase in the NPLs, which has a close connection to the economic cycle When borrowers are unable to fulfill their obligations to the loans, it would become credit risk of banks, which is one of the significant risks among many kinds of risks that most of the commercial banks are exposed Credit risk is distinguished by two components which are systematic and unsystematic credit risk (Castro, 2013) and in fact, it is very hard to set an efficient credit risk management policy and procedure for the banking system This is because of the unpredictable natures of economic

environment that have the impacts to banking-specific factors as well as risks in banking industry Therefore, this impact has raised many serious concerns to researchers and policy makers to understand the relation between credit risk and the business cycle in order to ensure the stability of a banking system

The beginning of recent crisis exploded since the collapse of the Lehman Brothers, the fourth-largest U.S investment bank It is because of the subprime mortgage crisis, many loan defaults makes the bank illiquidity to prevent from the crisis Moreover, the depositors do a massive withdraw their money out of the bank as they lost their confidence in the banks As a result, the bank do not have enough money to do business and indirectly cause the Washington Mutual bankruptcy Since the Lehman Brother do business around the world, it also leads banks in many countries face the credit risk Making loan is the traditional function provided by the bank but it also causes the credit risk, which come from the borrowers who are inability to pay back the loans as they promised Following to Castro (2013), the increase of bad loans in banks’ balance sheet leads to the problem of liquidity and insolvency, which is the signal for banking crisis

In the case of illiquidity and insolvency, banks will lose their abilities to pay to their debtors and fail to meet their obligations As a shock have happened, banks will be considered as loss and could be forced to shut down From there, both banks and their debtors will be struggled by loss and it will effect to economy Therefore, it is crucial

to raise awareness to the credit risk in order to determine the cause of risks and prevent banks from illiquidity and insolvency problems

Consequently, if banks need to control the credit risk efficiently, they must understand the factors that cause the credit risk However, as suggestion of Garr (2013), the nature

of macroeconomic environment is unforetold and also associates with various microeconomic factors, which makes banks’ credit risk management become a very complicated and tough objective in order to manage the credit risk Lack of knowledge and experience in credit risk management can leads banks to more serious risks Besides, Ratnovski (2013) points a view that credit risk management may become a burden rather than a solution for banks because it could drain a certain amount of

resources and time of banks For more specific, the managers also have to put many effort in knowledge and experiences to deal with it and it could raise the administrative cost while a low return on highly liquid assets cannot be compensated the cost A credit risk program requires time to take effect and resources (such as capital and labors) to

be employed and managed for a long time in order to prevent banks from a sudden attack of credit risk Therefore, if the credit risk policy and procedure are not based on the real situation of the factors that impact to credit risk, they will be loss because their money and time for the costly program are wasted, but also they will suffers a significant raise of the credit risk problems

As a result, it has led to many interests of researchers and policy makers in finding the factors that can lead to the bank credit risk, so that they can understand these factors and build an effective credit risk management to limit the probability of credit risk

The paper will examine the influence of macroeconomic environment factors to the non-performing loans ratio (NPLs) in the five countries of ASEAN (Indonesia, Malaysia, Philippine, Thailand and Vietnam) covering a 13-year period of time from

2002 to 2015, which are in the same development rate in the area However, due to the lack of NPLs data at countries level, the NPLs ratio of individual commercial bank will

be examined and in order to prevent from bias and to ensure the model consistent, other bank-specific factors will be adopted in this paper, there are 162 commercial banks’ information collected The data for macro determinants is collected from the World Bank data while bank-specific ones is from the Bank Scope-Fitch’s International Bank Database Finally, the objectives of this paper are as follows:

- To examine the impacts of macroeconomic determinants to the NPLs ratio of the commercial banks in the five countries of ASEAN

- To study the nature of the commercial banks’ specific factors toward the NPLs

in the five countries of ASEAN

- To find an appropriated method to measure the relationship between macroeconomic factors and the NPLs ratio

- To ensure the consistent of the chosen method through the application of robustness check and additional analytical tests

- Give recommendation to policy makers

- How do banks’ management in these countries affect their NPLs?

1.5 Hypothesis of the study:

This paper will examine the impacts of five macroeconomic factors to the NPLs rate, thus the five hypotheses are as follows:

H1: Gross Domestic Product (GDP) has a significant negative relationship with

bank credit risk in the five studied ASEAN countries

H2: Interest rate has a significant positive effect on bank credit risk in the five studied ASEAN countries

H3: Inflation rate has a significant impact on bank credit risk in the five studied ASEAN countries

H4: Exchange rate appreciation has a significant relationship with bank credit risk in the five studied ASEAN countries

H5: Unemployment rate has a significantly positive impact on bank credit risk

in the five studied ASEAN countries

Numerous existing papers are conducted to examine the credit risk determinants within

a country or a category of countries (such as in Europe, OECD or developed countries)

or a limit of determinant category In this study, the potential determinants of bank credit risk, which are applied in the model, are 11 factors (including five main

macroeconomic determinants and six additional bank-specific factors) This is also the first paper that examines the impacts of these variables on the NPLs of commercial banks in five ASEAN countries (Indonesia, Malaysia, Philippine, Thailand and Vietnam) from 2002 to 2015 In addition, due to the nature of the data sample in this paper and the limit of related research papers, the research methodological design will follow an extensive approach through the dynamic panel data econometric techniques that serve as a robust cross-validation of the results as well as several additional analysis and robustness tests

The results of the research will assist a better understanding into the key factors of credit risk in the commercial banks of studied countries In addition, the paper will propose useful information in explaining what cause the bank credit risk and in evaluating the performance of the banks toward the NPLs According to Demirguc-Kunt and Detragiache (1998), banking system of a country with high inflation rate, unemployment and interest rate seem to have higher bank credit risk and banking crisis would be easily occur Therefore, this study will give more understanding in the connection of the economic developments and the credit risk as well as the information

on how the banks’ operation and the economic condition within these countries is For more specific, the investor and depositor will know how and when the bank performances are in the stable and sound condition through knowing nature of the economic and bank specific factors With this knowledge, their banking activities are much easier to make exact decisions to use their fund and prevent from bad investments

In addition, the result will provide to bank managers an efficient loan and credit risk management policy with the information of which economic and bank specific determinants of the bank influence credit risk Therefore, with information such as increase in the inflation rate, interest rate or domestic currency appreciation, banks could issue an appropriated approach to monitor, evaluate and control for bank risk exposures with a more precise way Consequently, an efficient credit risk management policy will help bank management more effective in capital allocation, banking performance, operating cost and profitability

1.7 Structure of Research:

This research paper is organized in six chapters Chapter 1 is the introduction and overview the general idea of the study context Chapter 2 gives the literature reviews of the previous studies in both theoretical and empirical frameworks for the effect of the macroeconomic factors on the bank credit risk and it also describes the proposed hypotheses development for the study Chapter 3 consists of the data and research methodology which includes the research methodology, data collection methods, the model description and variable description

Chapter 4 will present and interpret the results of the econometric analysis with respect

to the research’s theoretical and empirical analyses, which are linked to the hypotheses

of the research paper It will show the relationship of the economic factors and the NPLs ratio of banks Furthermore, chapter 5 conducts additional analysis and robustness test

in order to examine the consistent of the estimator and finally chapter 6 will suggest some policy implications, the limitations and the final conclusion of this thesis

CHAPTER 2: LITERATURE REVIEWS 2.1 Theoretical reviews:

Credit risk is defined as the risk from borrowers who have lost their ability to pay loans back to lenders partially or totally In recent years, many banks in the world experienced substantial losses and reduction of capital provision due to rapid deterioration in assets’ quality This not only increased banks’ exposure to economic crisis but also restricted bank lending ability with both direct and indirect consequences to the financial stability and economic activities Therefore, the need for the credit risk analysis is crucial because it is not only to ensure a stable banking system for a prosperous economic growth but also can raise the awareness to the regulatory authorities to prevent a possible crisis in the future Castro (2013) identifies factors affecting systematic and unsystematic credit risk separately The factors influencing the systematic credit risk are: macroeconomic factors, changes in economic policies and political changes or changes in the goals of leading political parties While unsystematic credit risk is affected by specific factors: (i) individual-specific factors namely individual personality, financial solvency, capital and credit insurance; (ii) company-specific factors namely management, financial position and reporting, sources of funds, their ability to pay the loan and specific factors of the industry sector

2.1.1 Business Cycle and Risk:

The relationship between the economy and financial system has been argued in a number of theories Within the framework of business cycles, the connection between macroeconomic factors and loan quality is emphasized by linking to the movement of business cycle with financial vulnerability and banking performance Specifically, Messai and Jouini (2013) offers a theoretical models from Williamson (1987), which emphasizes the nature of credit risk and proposes the impact of business cycle to the financial sector of a country In addition, Messai and Jouini (2013) also summarized theoretical review for this relation, the phases of the business cycle relating to banking performance have been studied in order to express the relationship between the macroeconomic environment (such as the yearly GDP growth, the real interest rate, the annual inflation rate, the exchange rate and the unemployment rate) and the quality of

loans During the economic expansion phase, there are only a relatively small proportion of bad loans, borrowers are confident to have adequate income or more cash held to repay for their loans in time of deadlines Therefore, lenders may not pay much attentions to the credit standards and allow more risk (Koch and McDonald, 2003) orthe increased ability of creditors to repay loans leads to reducing of credit risk for lenders (Salas and Saurina, 2002) However, when economic conditions worsen, the studies of Jiménez and Saurina (2006) for Spanish banks and Bohachova (2008) for members of Organization For Economic Cooperation And Development (OECD) reach the conclusion that banks are vulnerable to adverse selection in their financial decisions and moral hazard behavior of their creditors so that this causes an increase in risk of loans

2.1.2 Interest Rate and Risk:

It is also argued that higher interest rate, mostly induced by monetary policy, associates highly with debt burden due to higher interest payment, which leads to high rate of NPLs For instances, following the theory of asymmetric information, borrowers are able to face adverse selection problem as interest rate surges, it is call “bad risk” (Bohachova, 2008), the result of loan applicants is probably adverse with the borrowers’ selection In order to pay for their loans, instead of using the loans on safe projects with low returns, creditors tend to have strong motive to riskier projects with much more higher income In addition, when interest rate increases, banks will earn more returns from new loans and floating interest loans while borrowers have to stand with higher payments and then the probability of increase in credit risk would occur on banks’ balance sheets (Demirguc-Kunt and Detragiache, 1998) However, from the view of the bank side, banks diversify their financial roles in the market, they conduct asset transformation and they lend to a large number of borrowers as well as borrow from a large number of depositors (Williamson, 1987) Moreover, in some countries with interest rate liberalization, because of rises in the costs of funds and the culture of high-risk behaviors; higher rates are charged to high-risk borrowers in order to mitigate risks, hence banks overall risk exposure increases more (Fofack, 2005)

When the economy went down, the return on bank assets deteriorates more than the rate that must be paid on depositors and banks would reduce profits or face losses As bank’s assets are composed of long-term fixed interest rate loans, thus banks cannot handle for the return on assets quickly enough As a result, banks would raise short-term lending interest rate in order to deal with their liability payments (Mishkin, 1996)1 In addition, when borrowers are likely to be exposed to debt burden, banks also face to a large risky loan portfolio, thus a higher net interest margin is required to compensate the higher risk of default (Ahmad and Ariff, 2007), which leads to a systematic banking sectors problems

2.1.3 Inflation and Risk:

Another factor that should be considered is the inflation, which is caused by the restrained money supply growth and the disposed nominal depreciation of the domestic currency; inflation influences to both banks’ decisions and borrowers’ behaviors to loans For more specific, inflation is unpredictable and an increase in inflation makes the prices of goods and services go up, thus the volatility of firms’ profits will rises as well as their debt obligations (Peyavali, 2015) An increased rate of inflation also have

a negative effect on real rates of return on bank assets as well as incomes of existing borrowers thereby making the quality of previously extended loans worse and resulting

to credit rationing (Bohachova, 2008) In addition, if variable loan rates are applied, high inflation leads borrowers to adverse selections because banks will prefer to adjust the lending rates to keep their real returns stable or the government conducts monetary policy to fight against inflation (Nkusu, 2011) On the other hand, disinflation also affects loan quality because in a previously high-inflation economy, there are high real interest rate, which makes the earnings of borrowers declined and encourages risks similar to a rise in nominal interest rate (Mishkin, 1996)

2.1.4 Exchange Rate and Risk:

Exchange rate, which indicates the value of domestic currency in terms of another, is also one of macroeconomic sources of economic instability as well as bank risk

1 Most of the United States banking panics follow an increase in short-term lending interest rates

exposure Because of no currency matching between the income of borrowers and their loan debts, for loans nominated in foreign currency, depreciation of domestic currency increases debts and debtors’ incapacity to pay the loans and then banks would face to loan defaults (Curak et al., 2013) When domestic currency depreciates, the rate of impaired loans would increase, especially for loans nominated in foreign currency Credit risk for bank loans is likely to increase to importers and decrease to exporters, thus bank’s overall risk exposure will be determined by its net vulnerability to exporting

or importing borrowers As the foreign currency appreciates, it costs more to purchase foreign goods and services, thereby more units of domestic currency are required to secure the same quantity of imported goods and services than before Accordingly, the demand of financial support for bank credit will increase to cover the raising costs and

it would reduce the firm’s profitability, then firm will encounter the problem to serve interest and principal of loans (Poudel, 2013) On the other side, Bochahova (2008) also expresses two theoretical interactions of exchange rate movement on banks’ credit risk For more specific, banks’ volatility could increase due to the domestic currency depreciation when banks liabilities denominated in foreign currencies are higher than their foreign exchange assets In addition, a great rate of domestic currency depreciation could lead to disintermediation as depositors decide to withdraw their funds from banks

to invest directly to other “hard currency assets” with higher returns, thus banks will face capital shortage and bank credit risks will increase

2.1.5 Unemployment and Risk:

Another theoretical explanation of the source of banking credit risk is viewed from unemployment as an indicator that highly correlate with the economic cycle For households and individuals, an increase in the unemployment rate during economic recession reduces the incomes, resulting cash flow streams be worse and then the probability of on loan defaults could surge While in corporate sector, a decrease in production due to a drop in the consumption and demand for goods, causes revenues loss and a weak liquidity position regarding debts Therefore, it exacerbates bank credit risk (Castro, 2013)

Specifically, the relation between unemployment and NPLs are proposed by Lawrence (1995), who conducted a theoretical model about life-cycle consumption In this model, the probability of loan default is explicitly explained that due to an increase in unemployment, it will induce lower level of income from borrowers and their debt-servicing capacity, thus the probability of credit risk is higher Furthermore, in order to limit the risk and ensure the capital for banks, higher interest rate loan will be offered

to clients with higher risk rates From the model, Rinaldi and Sanchis-Arellano (2006) have extended their study and suggested that the possibilities of NPLs also relied on the unemployment rate, which reflects the current income and the uncertainty regarding to the future income of borrowers as well as the lending rates applied by bank Besides, this model also implies that the volume of loan taken, the amount of investment and the time preference rate also impact the probability of default

Berge and Boye (2007) propose that during periods of cyclical economic recession, as unemployment rises and corporate earnings are diminishing, both NPLs and banks' losses may surge Higher unemployment rate also make borrowers suffer from debt-servicing costs and other costs while banks have to determine their loan provisions following to the borrowers’ expected future flows of income and expenditure It will deteriorate the borrower’s debt-servicing capacity as movements of these factors diverse from expected developments, thus the credit risk will increase

2.2.1 Gross Domestic Product (GDP):

Gross Domestic Product (GDP) can be defined as the monetary value of all the finished goods and services produced within a particular country's borders in a specific time period Following former researches, this paper will use annual growth rate of real GDP

at constant prices as an indicator for both of economic activities and business cycles, which may have directly impact on the banking system in regard to bank risks Most of literatures find a significant influence and a negative relationship between GDP growth and NPLs Specifically, Shu (2002) executes stress testing for the Hong Kong’s banking sector to calculate the volatilities of loan quality between 1995 and 2002 Borrowers’ ability to loan repayment and the banks’ portfolio position are influenced by changes in

macroeconomic determinants, which are considered as the risk factors in the paper The author concludes that higher economic growth or economic expansion highly associates with higher profitability for corporate sector, reducing the default rates while banks’ exposure to risk reduces and then open more chances to lend rapidly Moreover, applying Merton´s methodology to analyze the relationship between Czech bank credit risk and macroeconomic factors, Jakubik (2007) finds that decreases in real GDP growth deteriorates the banks’ loan portfolio quality due to changes in the corporate earnings, wage growth and high unemployment rate, which leads to higher bank credit risk In the case in Tunisia banking system, Zribi and Boujelbène (2011) examine a panel model of ten commercial banks from 1995 to 2008 and use GDP growth as the macroeconomic variable in order to ascertain the bank credit risk They also indicate the negative overall effect of GDP growth on the bank credit risk

Louzis et al (2012) conduct research with dynamic panel approach on a wider range of loans (consumer loans, business loans, and mortgages) in Greek banking system over the period 2003–2009 They conclude that the borrowers’ capacity of loan repayments depends on the phase of the economic cycle In an economic downturn or lower GDP growth, the NPLs will increase for all loan types while in the economic expansion, borrowers will have sufficient and enough incomes to repay their loan Therefore, it can

be expected that NPLs is correlated negatively with the economic cycle, rising at times when economic activity slowdown and deteriorates the quality all loan types

Besides, for cross-country level, according to a study with dynamic panel data method

of Castro (2013), in the period 1997q1–2011q3, regard to banks of Greece, Ireland, Portugal, Spain, and Italy, the paper demonstrates the significant interaction of GDP development and the recent financial crisis to the movement of the bank credit risk Their results show that GDP growth is negatively related to the NPLs, the higher level

of GDP growth causes a higher level of income for borrowers, leads to greater cash flows This also raises the profitability of the bank and lowers the NPLs and bad debts The same conclusion is founded in the papers of Nkusu (2011) in case of 26 advanced countries from 1998 to 2009; or Messai and Jouini (2013) in case of Italy, Greece and Spain for the period of 2004-2008; or Klein (2013) in case of Central, Eastern and

South-Eastern Europe (CESEE) in the period of 1998–2011; or Chaibi and Ftiti (2015)

in case of comparison between French and German economy

On the other hand, there are several researchers found out no significant relationship between GDP growth and bank credit risk For example, Poudel (2013) indicates no significant relationship between GDP and NPLs in 31 Nepalese commercial banks over the period from 2001-2011 It can be explained that during economy downturn, when making new loans, banks tend to carefully qualify their borrowers based on creditworthiness and credit condition of borrowers Besides, banks are well prepared and will categorize their clients and debtors in order to control the amount of NPLs and credit risk Therefore, the volume of credit would be reduced during low GDP growth phase The same result is also supported by Kalirai & Scheicher (2002) in case of Austrian banking system, Fofack (2005) in Sub-Saharan Africa banks and Aver (2008)

in case Slovenian banking system

2.2.2 Interest Rate:

Interest rate is another significant determinant in order to investigate the correlation between the interest rate and credit risk because it directly affects the debt burden of borrowers Since there are many different kinds of interest rate, this paper will choose the real interest rate due to data availability and it is expected to be positive In addition, different types of interest rate usually have a strong correlation with each other: an increases in the interbank rate addresses an increase in monetary policy interest rate and leads to money market rates surge as well as long-term fixed-income securities yields (Bohachova, 2008)

Fofack (2005) finds positive relationship between real interest rate and credit risk in Sub-Sarahan Africa The paper suggests that higher interest rate leads to an increase in cost of borrowing that borrowers would pay to obtain loan, as well as an increase in cost

of deposits that make the commercial banks’ profit decrease Therefore, the default rate will increase In addition, Jiménez and Saurina (2006) with the help of Generalized Method of Moments (GMM) estimator for dynamic panel models also used the real interest rate to investigate the impact of interest rate on loan loss They found a significant and positive relationship between interest rate and loans losses in Spanish

commercial or savings bank between December 1984 and December 2002 Also with study applying the GMM estimator for banking systems in Southeastern Europe, Curak

et al (2013) point out that the higher real interest rate is, the higher possibility of the NPLs of variable rate loan are The explanation is that as the real interest rate increases,

it creates an additional burden for debtors to serve their payment obligations

From the research conducted by Castro (2013) in GIPSI countries, the paper use term interest rate in its regression as benchmark for analysis and it is the most appropriate measurements because banks normally and mostly do lending long-term loans The study finds a significant positive relationship between interest rate and credit risk Also in the research, for robustness check of the model, the long-term interest rate

long-is replaced in turn by the real interest rate and the interest rate spreads, the results of these variables are in same direction, In addition, long-term interest rate is more important to measure the effect of credit risk when loan interest is either higher or lower and higher interest rate will lead to increase the obligation for corporate borrowers and individuals, thus it induces the banks’ credit risk The similar results are found in the research of Quagliariello (2007) between the long-term interest rate, measured by ten-year Italian Treasury bond, and the proxy for credit risk, the loan loss provision In addition, the findings of Solarin et al (2011), which comply on the basis of Auto regressive distributed lag (ARDL) approach on Islamic banks of Malaysia and interest rate, indicates a significant positive long-run impact on NPLs Also in Malaysia, with a test on commercial banks during 2006 till 2010, Asari et al (2011) apply the vector error correction model to discover the effect of interested rate on NPLs They find a strong long-run relationship between interest rate and NPLs while in short run, interest rate do not influence NPLs

The paper of Ali and Daly (2010) employs credit risk logit regression for both banking system in USA and Australia for 14 years from 1995Q1 to 2009Q2 and does not find any significant relationship between nominal interest rate or short-term interest rate (6-month) and credit risk in both USA and Australia

2.2.3 Inflation rate:

Inflation rate decreases the purchasing power of currency, when the general price level

of goods and services is rising in an economy until a certain extent Inflation rates are generally associated with the interest rate of loan and affect the efficiency of banking sector as well as the debt obligation of borrowers Through literature reviews, the impact

of inflation on NPLs can be positive or negative On one hand, as inflation rate increase, the real value of loans in nominal rates or variable rates (as adjusted according to the inflation) deteriorates, debtors can easily repay their loans On the other hand, high inflation reduces real value of the profitability while rises cost of capital and thus weakens the debtors' ability to the loan (Curak et al., 2013)

Several studies have found inflation is positively affecting the banks credit risk The results of Demirguc-Kunt and Detragiache (1998), using a multivariate logit econometric model with a large sample of developed and developing countries in 1980-

94, indicate that high rate of inflation is one of the consequences exacerbating risk problems of banking sector High inflation may associate with the high nominal interest rate and banks would find it difficult to perform a maturity transformation In addition,

as the paper‘s empirical evidence shows, when restrictive monetary policies are implemented to control inflation and keep banking sector stability, they lead to a sharp increase in real interest rate; high real rates tend to increase the likelihood of a banking crisis Utilizing a panel data at bank level for both public and private commercial banks

in India, Thiagarajan et al (2011) conduct an investigation in the relationship between current inflation and one year lag inflation with bank credit risk In the public sector banks, the authors find a positive relationship between current inflation and credit risk but no any relationship with one year inflation lag However, in case of private sector banks, the relationship between inflation and credit risk is not significant Furthermore, the research of Badar & Javid (2013) also suggests the significantly positive relationship between inflation and bank credit risk, which examines the impact of macroeconomic

on NPLs of 36 Pakistani commercial banks during the period 2002 to 2011 The study states that inflation will affect the profitability of commercial banks and increase the bank credit risk Because the contractionary fiscal policy from government is

established to increase the interest rate in order to control inflation when the Consumer price index (CPI) increases Thus, increase in interest rate will lead to an increase in the cost of borrowing as well as reduce the borrower's real income and their ability to repay the loan Consequently, NPLs increase while the profitability of banks decreases, at the end banks credit risk will increase Similarly based on sample of 69 banks in 10 Southeastern European countries in the period from 2003 to 2010, Curak et al (2013) concludes that the inflation variable is positive and significant The authors explains that because of currency instability and variable interest rate loan adjusted for inflation, real value of incomes decreases make debtors difficultly repay the loans

In the opposite direction, there are researchers argue that inflation has a negative relationship with credit risk Using bank-level data for 80 countries in the year 1988-

1995, Demirguc-Kunt & Huizinga (1999) suggest that banks credit risk is negatively associated with the inflation Specifically, the researchers believe that it could be the bank float, which help these banks earn higher income Or in case of high inflation, there are delays in crediting customer accounts while banks cost is lower than banks net interest margin and bank profitability In addition, a study of Poudel (2013) on credit risk determinants in Nepalese banks during the period 2001 to 2011 also finds a negative relationship between inflation and credit risk The paper gives an evidence that when there is a sign of a high inflation, the interest rate will increase in order to strike for the inflation Therefore, the Nepalese banks will reduce the volume of loans and only establish their lending activities in assured sectors and allow higher quality loans Furthermore, they will carefully categorize the quality of borrowers, therefore it decrease the bank credit risk The results are consistent with studies of Shu (2002), Zribi

& Boujelbene (2011), Vogiazas & Nikolaidou (2011) and Washington (2014), which also conclude negative relationship between inflation and bank credit risk

However, some other study by Fofack (2005), Aver (2008), Dash and Kabra (2010), Castro (2012) and Chaibi and Ftiti (2015) do not find any influence of inflation to credit risk, in case of Sub-Saharan African, Slovenian, Indian, GIPSI and French (a market-based economy) banking system respectively

2.2.4 Exchange rate:

Exchange rate is defined as the value of one nation’s currency in relation to the value

of other nations’ currencies and volatility of exchange rate is one of the main sources

of economic instability, which could strongly affect the loan quality nominated in foreign currency as well as bank credit risk In this research, the real effective exchange rate (REER)2, will be added to control for external competitiveness of studied ASEAN countries The local currency is appreciated as increases in REER, which leads to the goods and services produced in that country relatively more expensive and affects adversely the competitiveness of export-oriented firms as well as their abilities to pay debt (Nkusu, 2011) Consequently, the ratio of NPLs increases

Considering the nominal effective exchange rate, Shu (2002) examines the exposure of Hong Kong banks’ lending portfolio (loans denominated in foreign currencies or used abroad) to the Mainland China and the nominal effective exchange rate The author points out the banking sector’s exposure to the Mainland and the appreciation of the nominal effective exchange rate deteriorate banks’ portfolios and asset quality during the Asian financial crisis For more specific, appreciated exchange rate increases the burden on foreign borrowers to repay debts In addition, higher import prices cause higher production cost and then lead business failures These factors could result to more foreign loan defaults

Fofack (2005) conducts a research to investigate the correlation between NPLs and the real exchange rate appreciation in the Sub-Saharan Africa countries during the 1990s The author also makes a comparison between CFA and non-CFA countries34 As is shown in the whole countries study, the real exchange rate appreciation may diminish the economic growth of countries which are highly dependent on exports (such as Cameroon and Côte d’Ivoire, the world-top coffee and cocoa exporters), and induced the banking crisis For countries in the CFA, Fofack (2005) points out that bank credit

risk could increase with the association of the aggregate stock of money and real effective exchange rate appreciation These contributions deteriorate banks’ loan portfolios in these countries because it is possibly that exchange rate appreciation occurs together with the changes in terms of trade Banks make loans to support the export activities in the agricultural sector but the decrease in export rate causes banks encountered a pressure from highly accumulated problems loans However, non-CFA countries are not strongly correlated with the real exchange rate, the probably explanation for it is that these countries have conducted a flexible exchange rates regimes, thus there would appear an automatic adjustment process to limit the risk By means of a comparative analysis, the examination the determinants of NPLs of commercial banks in France and Germany, Chaibi and Ftiti (2015) conclude that increase in real effective exchange rate is a significant positive determinant of NPLs in France, a market-based economy, and contributes to an increase in credit risk However,

an exchange rate appreciation significantly contributes to lower NPLs in Germany, a bank-based economy, because it would have improved the ability of those Germans who borrow foreign currency to service their debts and the NPLs ratio decreases Refer to the study of Castro (2012) in GIPSI countries, which share a single currency – the Euro, from 1997 to 2011 and found negative relationship between real effective exchange rate and credit risk The paper includes the lag of REER, with reference to the

27 EU members, in the equation to control for external competitiveness and points out

to the fact that an increase in this variable contributes to an increase in the credit risk Furthermore, in the research in CESEE countries by Klein (2013), exchange rate depreciation against the euro, coming along with higher inflation, contribute to higher NPLs in the period of 1998–2011 The same results are also founded by Gunsel (2012) for North Cyprus, by Vogiazas & Nikolaidou (2011) for Bulgaria

However, Hoggarth et al (2005) apply stress tests of UK banks using a VAR approach including changes in the real effective exchange rate and find a little impact on the NPLs Also Aver (2008) in Slovenia has not found any relationship in foreign exchange fluctuation and credit risk

2.2.5 Unemployment rate:

The unemployment rate is an additional information regarding the impact of economic conditions to bank credit risk As an increase in the unemployment rate should influence negatively to the income of individuals as well as corporate sectors and increase the debt burden Therefore, an increase in the unemployment rate is expected to lead to an increase in the banking credit risk

To forecast banks performances as well as to obtain relatively early warnings of unusual performance based on macroeconomic indicators in the U.S., Gambera (2000) conducts

a study with quarterly data from 1987-1999 to investigate the influence of economic variables on loan losses; the results indicate that incomes of households along with unemployment rates are significant factors as well as good predictors for loan losses in the U.S In addition, Messai and Jouini (2013), for a sample of 85 banks in Italy, Greece and Spain for the period of 2004-2008, also come to the result that NPLs

macro-is positively connected with the unemployment rate The same findings about the positive correlation of unemployment to NPLs are corroborated with the study of Nkusu (2011) in 26 Advanced Economies, Farhan et al (2012) in Pakistan, both Vogiazas and Nikolaidou (2013) and Bucur and Dragomirescu (2014) in Romany and Makri et al (2014) in 14 countries of Eurozone

Furthermore, for the Nordic banking system in the period from 1993 to 2005, Berge and Boye (2007) find a positively significant impact of unemployment on NPLs in both households sector to corporate sector Specifically, for the households sector, the income is one of a significant contributions that effects households’ debt-servicing capacity and the volume of problem loans When unemployment is rising, the current income used to service debt of households may reduce and thus it will lead to a higher volume of problem loans Accordingly, households will solve debt servicing problems

by using financial reserves and/or reducing consumption On the other hand, for enterprise sector, the capacity to service debt also depends on their income, which is mostly effected by business cycles As an indicator of the level of economic activity, the higher unemployment is, the lower domestic demand will be This normally leads

to a decrease in corporate earnings as well as their debt-servicing capacity Therefore,

higher unemployment will raise problem loans The same result is found by Kelly (2012), the default probabilities of the residential mortgage book of Irish financial institutions depend on changes in unemployment rate, which have stronger influence on those compared to house price movements Through an unemployment shock in 2006

to 2007, the income falls that limit individuals to repay burden of the mortgage

More recently in the research of Chaibi and Ftiti (2015), as the comparison between the two different economy France and Germany, the result reveals that the credit risk in French banks are more sensitive to the economic environment (the growth of GDP and unemployment) than those in Germany, although the NPL ratios for both France and Germany rise significantly when the unemployment rate increases This can be explained by the fact that mostly loans of German banks are allotted to private sector with high-skilled workers who are less likely to become unemployed

However, refer to a comparative study of 3 types of loans in Greek banks conducted by Louzis et al (2013), all NPL categories are significantly negative impacted by unemployment Among other types of loans, the business NPLs is the most sensitive because firms seem to cut their labor cost before they face debt servicing problems While indicator of consumer NPLs implies that an increase in unemployment prevents households’ ability from service their debts Furthermore, mortgages are the least sensitive NPL type The explanation for this difference is similar to Chaibi and Ftiti (2015), it could be that that mortgage loans are mostly provided to civil servants and high-skilled workers of the private sector, who have less probabilities to be unemployed

In general, while the evidence on the factors affecting systematic risks is mixed, majority of the studies find that most of macroeconomic distributions have a significant effect on loan quality while others do not due to the difference in types of banks system

or country institutions In this paper, the empirical analysis with a proper dynamic panel data approach will extend to a panel of commercial banks in the five ASEAN members Macroeconomic determinants and bank-specific factors as control variables are investigated and assumed to contribute a various important roles on the bank risk of

default Thus, these factors will take a substantial part in the explanation of the credit risk in this study

Prior studies suggest that the economic environment is fundamental to explain the behavior of the credit risk This paper will include five independent macroeconomic variables to examine the effects of economic environment on the bank credit risk in five ASEAN members The hypotheses are as below:

2.4.1 Gross Domestic Product (GDP):

GDP growth is one of variables, indicating the cyclical position of the economy, that significantly effects bank credit risk It has been evidenced that when all other things being equal, advanced economic countries with higher GDP are less susceptible to banking crises (Demirgüç-Kunt & Detagiache, 1998) A study of Jakubik (2007) indicates that a decrease in GDP rate will increase the probability of credit risk since unemployment increase while company’s profitability and value of company’s assets deteriorates As borrowers gain a sufficient rate of income and revenues to service their debts, there is usually a relatively low rate of NPLs during the expansion phase of the economy (Chaibi & Ftiti, 2015) However, for a long booming period, credit is provided

to lower-quality debtors to and subsequently, when the recession phase occurs, NPLs tend to increase (Castro, 2013)

GDP is a basic indicator of the cyclical position of the economy As a low growth rate

in GDP has a negative effects on corporate earnings, wage growth and a rise in unemployment or prices of assets, which will lead to a deterioration in loan portfolio quality Therefore, GDP and bank credit risk is negative relationship and this argument leads to the first hypothesis:

H1: Gross Domestic Product (GDP) has a significant negative relationship with

bank credit risk in the five studied ASEAN countries

2.4.2 Interest rate:

Interest rate is a strong determinant of credit risk because it influences the debt burden

of borrowers, there have positive relationship between interest rate and credit risk A

rise in interest rate makes bank loan portfolio quality worsen as it increases the corporates and households’ cost of financing and decreases the market value of assets (Jakubik, 2007) as well as debt burden of borrowers of corporate or individual (Chaibi

& Ftiti, 2015) Consequently, borrowers’ ability of pay back loan is weaken that leads

to a higher rate of NPLs (Nkusu, 2011; Louzis et al., 2012; Castro, 2013)

Under the view of the asymmetric information theories, higher interest rate tends to exacerbate the problem of “adverse selection” – the selection of borrowers with high probability of adverse project outcomes or bad risks in the context of credit relationships (Mishkin, 1996) The volatility of high interest rate prevents potential borrowers from risk free projects, thus the risk composition of the pool of loan applicants will shift

toward bad risks In addition, a rise in interest rate will deviate the ex post incentives

for borrowers inducing them to take on riskier projects (Stiglitz & Weiss 1981) It is clear that in a background of information asymmetries, if all factors are held constant,

a rise in interest rate will increases credit risk on balance sheets of banks Hence, the hypothesis is:

H2: Interest rate has a significant positive effect on bank credit risk in the five studied ASEAN countries

2.4.3 Inflation rate:

Inflation also has an influence to the real cost of the loan, it affects borrowers’ debt servicing capacity as well as the performance of the banking sector The literature generally shows that the contribution of inflation on NPL is ambiguous On one hand, higher inflation can lower the real value of outstanding loans, thus capacity of borrower for loan payment is enhanced as the actual amount need to pay back become smaller

As a results, the relationship between inflation and bank credit risk is negative Many other papers also support for this connection such as Shu (2002) for the study in Hong Kong banks, Zribi & Boujelbene (2011) in Tunisian banking system, Vogiazas & Nikolaidou (2011) in Romania, both Aver (2008) and Castro (2013) in the case of Slovenian banking system and Poudel (2013) in Nepal

However, inflation can also limit debtors’ ability to service debt expressed in both nominal and floating rate loans (Curak et al., 2013) by diminishing real income when

wages are sticky and hence erodes the cash flow to pay back the loan In addition, a rise

in inflation would reduce the value of bank assets and therefore influence credit rationing As a result, banks would adjust their lending rates to maintain their real returns and profitability on their balance sheets This would increase burden for new borrowers as well as existing borrowers facing to variable rate loans Thus, a positive relation between inflation and NPLs is concluded by researchers (Demirguc-Kunt and Detragiache, 1998; Rinaldi & Sanchis-Arellano, 2006; Thiagarajan et al.; 2011; Badar

& Javid; 2013; Curak et al., 2013) In order to clearly find out the relationship between inflation rate and NPLs, it will be tested by the following hypotheses:

H3: Inflation rate has a significant impact on bank credit risk in the five studied ASEAN countries

is that domestic currency depreciation could increase credit risk for banks whose the liabilities to assets ratio denominated in foreign currency is higher Other reason is that

a sufficiently depreciated currency can cause disintermediation, which could make banks’ capital insufficient for their operation as depositors withdraw their money and invest to “hard currency assets” with higher returns Consequently, banks will face capital shortage and increase bank credit risks will increase

Based on where a country is in its path of general economic development, it might need

to strengthen the conditions the quality of economic environment, especially financial market efficiency, for development and competitiveness As a result, banking systems

of developing countries generally have a high portion of loans relating to export-import activities, for example lending to the manufacturing industry, which have a highly sensitive proportion to exchange rate movements The depreciation of local currency cause higher costs of imported inputs as well as higher foreign-currency-denominated debt burden of firms While in the case of appreciation, it erodes the profitability and competiveness of export-oriented firms In both cases, firms would stand the problem

to serve interest and principal of their debts and bank credit risk exposure is depended

on the net rate of import-export

Refer to above discussions, the exchange rate have related to credit risk for commercial banks and the following hypotheses are proposed:

H4: Exchange rate appreciation has a significant relationship with bank credit risk in the five studied ASEAN countries

2.4.5 Unemployment rate:

The unemployment rate, which could show the economic conditions of a country, provides an additional macroeconomic factor inducing credit risk The unemployment rate indicates the ability to serve loan payment as well as levels of debts and the problems that banks would have to deal with in lending money to finance the economy Refer to the prior researches, the unfavorable increase in the unemployment rate should impact negatively to the income of households Thus by cutting consumption or savings, they would be capable of servicing their debt However, it could lead to a decrease in the market demand, the total production level and revenues of firms would decline as well as the production cost (such as wages) would increase so that it limits firms’ capability to repay their loans Therefore, the unemployment rate may increase the risk

of credit defaults and it is hypothesized that unemployment rate has a positive impact

on bank credit risk

H5: Unemployment rate has a significantly positive impact on bank credit risk

in the five studied ASEAN countries

CHAPTER 3: DATA AND METHODOLOGY

In chapter 3, data collection methods, model specification, variable definition and measurement and the econometric strategy (the SGMM estimator) will be discussed in details All the elements in this chapter were constructed based upon the purpose of the research The aim of this paper is to identify the main systematic determinants of NPLs

in commercial banks of the five ASEAN members (Indonesia, Malaysia, Philippine, Thailand and Vietnam)

3.1 Data collection:

In accordance with the previous literature reviews and the theories, the sample includes the NPLs rate of individual commercial banks as dependent variable, the reason for this choice is that because the NPLs at country-level is not available enough to investigate

In addition, the macroeconomic factors (the Real GDP growth, Inflation, the Real interest rate, Exchange rate and unemployment rate) is considered as independent variables and several bank-specific determinants as control variables Macroeconomic determinants is collected from World Bank data while the NPLs rate of banks as well

as bank-specific determinants are collected from the Bank Scope-Fitch’s International Bank Database In order to ensure the time length as well as the data validity in the five ASEAN countries, the sample covers the period from 2002 to 2015

Also regarding to the previous papers, this research will only pay attentions to the commercial banks (including state-owned commercial banks, joint-stock commercial banks, city commercial banks, rural commercial banks and foreign commercial banks)

In addition, the data of these commercial banks must be available for at least three consecutive years The reason for this selection is that commercial bank is the main player in banking industry which provides variety of financial services and investment products to the customers Other kinds of banks (State banks, investment banks and Islamic banks) are not considered because they have different objectives rather than profitability Around the five studied ASEAN members, the final panel sample is balanced and consists of 162 banks and 1263 annual observations The detail number

of banks in each country is reported in Appendix 1

3.2 Econometric methodology – The NPLs measurement:

It is needed to consider a suitable methodological approach in this research, which must

be consistent with the expectations of both viewed theoretical frameworks and empirical results as well as the modelling flexibility within data restricted requirements This paper will try to find a suitable cross-country modelling techniques, which can exploit both the time series and the cross-sectional dimensions of the dynamic panel sample in order to find the most appropriate results

A few researches have utilized cross country displaying strategies, however the variety

of issues, the parts of institutional situations as well as the account framework management among nations can bring an uncertainty to the study of cross country boards Consequently, the accumulation of individual banks' information and also the general financial environment for every nation would answer the issue in both the time arrangement and the cross-sectional measurements of the dynamic board test

Due to the restriction in collecting data, the panel data of this research is balanced Following the recent literatures in panel data (Salas and Saurina, 2002; Louzis et al., 2012; Castro, 2013; Chaibi & Ftiti, 2015), a dynamic approach is introduced in order to account for the time persistence in the NPLs structure with a lagged variable of NPLs

A dynamic panel data specification is generally adopted as follows:

Where: δ is a constant term; the subscripts i = 1,…,N and t = 1,…,T respectively indicate the cross sectional and time dimension of the panel; 𝑁𝑃𝐿𝑖𝑡 is the dependent variable NPLs; 𝑋𝑖𝑡 is a k×1 vector of explanatory variables (macroeconomic and microeconomic factors) other than 𝑦𝑖𝑡−1; β is a k×1 vector of coefficients; 𝑣𝑖 the unobserved individual (country and bank specific) effects and 𝜀𝑖𝑡 is the error term

By construction, some econometric bias can arise with traditional panel data estimators, the OLS estimator turn to be biased and inconsistent even in the case 𝜀𝑖𝑡 are not serially correlated Moreover, in a dynamic panel data model, the RE estimator is also biased while the FE estimator is consistent with a larger T In addition, the size of time in this research is relatively small (14 years), the correlation between the lagged dependent

variable 𝑁𝑃𝐿𝑖𝑡−1 and the unobserved effects 𝑣𝑖would highly biased and it is not a reasonable choice for this estimator in this study

Indeed, the above biases can be taken out by other econometric approaches such as instrumental variables and taking the first-differencing equation (1) to eliminate the individual effects Prior to former papers (Louzis et al., 2012; Castro, 2013; Chaibi & Ftiti, 2015), the generalized method of the moments (GMM) estimator, proposed by Arellano and Bond (1991) and further developed by Arellano and Bover (1995), Blundell and Bond (1998), is more consistent for dynamic panel models with lags of dependent variable Accordingly, the GMM estimator is appropriated to be employed

in this study in order to deal with the mentioned biases in the traditional estimators Based on the first difference transformation of equation (1), the dependent variable with

a lag of order j+1 can be used to assured the respective moment conditions and the new equation is written as follows:

𝑁𝑃𝐿𝑖𝑡− 𝑁𝑃𝐿𝑖𝑡−1 = 𝛼(𝑁𝑃𝐿𝑖𝑡−1− 𝑁𝑃𝐿𝑖𝑡−2) + 𝛽′(𝑋𝑖𝑡− 𝑋𝑖𝑡−1) + (𝜀𝑖𝑡 − 𝜀𝑖𝑡−1) (2)

After the differencing, the individual level effect are eliminated in equation (3), but another bias has been appeared from the possibly endogenous explanatory variables, the dependent variable 𝛥𝑁𝑃𝐿𝑖𝑡 is correlated with new error term 𝛥𝜀𝑖𝑡 and 𝑁𝑃𝐿𝑖𝑡−1 On the assumption given that error term and explanatory variables (exogenous or predetermined ) are serially uncorrelated and not strictly exogenous, Arellano and Bond (1991) suggest that lags of order two and more of the dependent variable and current and lagged values of explanatory variables should satisfy the following two moment conditions:

𝐸[𝑁𝑃𝐿𝑖𝑡−𝑠(𝛥𝜀𝑖𝑡)] = 0 𝑓𝑜𝑟 𝑡 = 3, … , 𝑇; 𝑠 ≥ 2 (4)

The above described orthogonality restrictions are the basis of the one-step GMM estimation Regarding to the assumed statement of independent and homoscedastic residuals for both cross sectional and over time, this approach creates consistent parameter estimates In fact, in the Arellano–Bond methodology, the appearance of

first-order autocorrelation (AR1) in the error terms does not generates inconsistent estimates In addition, the dependent as well as predetermined or endogenous variables are instrumented with their lagged levels while the difference of the strictly exogenous independent variables are instrumented with themselves Hence, it is required the no second-order autocorrelation in the inconsistent differenced equation (Arellano and Bond, 1991) For the consistent GMM estimates in the moment conditions, the presence

of the instruments is critical

From there, Arellano and Bond (1991) proposed a two-step GMM estimator At the beginning, the error term is assumed independent and homoscedastic across countries and over time then it employs the recovered residuals in order to construct a consistent variance-covariance matrix of the moment conditions The two-step GMM estimator is

a convergent estimators and can be seen to be more efficient than the one-step as well

as ease up the homoscedasticity assumption However, from a statistical perspective, Blundell and Bond (1998) demonstrate that for the regression equation in differences, lagged levels of persistent explanatory variables are weak instruments, which can extend the variance of the coefficients While regarding to a conceptual perspective, during the differencing process, the two-step estimator not only disposes of the individual-level influence 𝑣𝑖, which refrains from the cross-bank relationship between bank-specific variables and NPLs; but also the time-invariant explanatory variables, which is only available in cross-sectional information In particular, some of explanatory variables such as leverage (liabilities) and size (total assets) are nearly time-invariant (Chaibi & Ftiti, 2015) As a result, the first-difference of the variables is uninformative, which has the related parameters being unidentified in the first-differenced system

In addition, in the two-step GMM estimator, the increase in efficiency are not significant even in account of heteroscedastic errors (Judson and Owen, 1999) Besides, due to its comparative reliance to measure residuals from the one-step estimator, the two-step estimator will set a bias in standard errors, which would result to questionable asymptotic statistical illation (Bond, 2002; Bond and Windmeijer, 2002) This problem should be considered especially in the case that the extent of cross-section data is

relatively small (Arellano and Bond, (1991); Blundell and Bond, 1998), which is correctly similar to the case of this research

As noticed by Roodman (2009), the difference and the system GMM can cause a problem of instrument proliferation For more specific, the source of the bias comes from the internal instruments, which are generated from past observations of the instrumented variables In fact, it is a trade-off between the lag distances for creating internal instruments and the depth of the sample to estimate in the two-stage least-squares (2SLS) analysis Roodman (2009) also suggests the bigger sample size would increase the instrument count, which could causes misleading asymptotic results for both the estimators and related specification tests, especially in small sample studies Further, Chaibi & Ftiti (2015) also propose the problems from two perspectives Firstly, from the classical one, numerous instruments or instrumental variable estimators generally can over fit endogenous variables Secondly, from more modern problems, they are specific to feasible efficient GMM (FEGMM), in order to identify moments between the errors and the instruments, the sample moments are employed to estimate

an optimal weighting matrix Thirdly, both types of problems can together cause invalid and valid results at the same time because of weakened specification tests

Therefore, Chaibi & Ftiti (2015) suggest two methods to limit the generated instruments’ numbers in difference and system GMM The first approach applies specific instruments instead of all lags, which still generates separate instruments for each period; thus the instrument count is linear in T while the number per period is covered This is equivalent to projecting regressors onto the full instrument set, but the coefficients are constrained on certain lags in this projection to 0 The second one is to combine instruments by adding together smaller sets As no lags are dropped, more information can be retained It is similar to put the constraint in projecting regressors onto instruments that have certain subsets with the same coefficient To eliminate over fitting instruments and instrument proliferation problems, Roodman (2009) suggests the moment conditions of the standard difference GMM as: