Determinants of Non-performing Loans: Empirical Study on Ethiopian Commercial

Banks

Zelalem Tsige

A Thesis Submitted to

The Department of Accounting and Finance

Presented in Partial Fulfillment of the Requirements for the

Degree of Master of Science (Accounting and Finance)

Addis Ababa University

Addis Ababa, Ethiopia

May 2013
 Addis Ababa University

School of Graduate Studies

This is to certify that the thesis prepared by Zelalem Tsige, entitled: Determinants of

Non-performing loans: Empirical Study on Ethiopian commercial Banks and submitted in

partial fulfillment of the requirements for the degree of Degree of Master of Science

(Accounting and Finance) complies with the regulations of the University and meets the

accepted standards with respect to originality and quality.

Signed by the Examining Committee:

Examiner_____________________ ( ) Signature_____________ Date___________

Examiner _____________________ ( ) Signature__________ Date _____________

Advisor Wollela Abehodie (PhD) Signature__________ Date ________________

__________________________________________________________________

Chair of Department or Graduate Program Coordinator

ii
ABSTRACT

Determinants of Non-performing loans: Empirical Study on Ethiopian commercial Banks

Zelalem Tsige

Addis Ababa University, 2013

This study examines the bank-specific and macro-economic determinants of Non-

performing loans (NPLs) of commercial banks in Ethiopia. The study adopts a mixed

methods research approach by combining documentary analysis (structured review of

documents) and in-depth interviews. More specifically, the study reviews the financial

records of eight commercial banks in Ethiopia and relevant data on macroeconomic

factors considered for the period from the year 2000 to 2011. The findings of the study

show that, loan growth, financial performance, operational efficiency, effective exchange

rate, inflation rate and gross domestic product have negative and statistically significant

relationship with banks’ NPLs. On the other hand, variables like bank size and state

ownership have a positive and statistically significant relationship with banks’ NPLs.

However, the relationship for average lending rate and income diversification were

found to be statistically insignificant. The study suggests that focusing and

reengineering the banks alongside the key drivers of NPLs could reduce the

probability of loan default in Ethiopian commercial banks.

iii
Acknowledgements

It gives me a great pleasure to extend my sincere gratitude for the help I have received

from different people and institutions. I would have never been able to complete this

thesis without their kind support.

First and foremost, I would like to thank the Almighty God who gave me power and

patience in every endeavor of my life. Next, I would like to extend my deeper

indebtedness to my advisor, Wollela Abehodie Yesegat (PhD) for her invaluable

comments, encouragements and guidance at various stage of the study.

I am thankful to Addis Ababa University for the financial support provided to me during

my thesis work through the Department of Accounting and Finance. My thanks also go to

my employer Arba Minch University for the sponsorship and financial supports.

My special gratitude also extended to the management and staff members of the

Ethiopian Commercial Banks, the National Bank of Ethiopia and the Ministry of Finance

& Economic Development for their cooperation in providing me all the necessary data

required for the study.

Finally, I would also like to express my sincere thanks to my dad, Priest Tsige Alemu,

and my mum, Merechesh Mekonen, who thought me how to make the most of my life. I

am also grateful to my brother, Deacon Addisu Tsige, and my aunt, Wagaye Mekonen

and her daughter Hana Assefa for their encouragement and consistent support. Besides,

my thanks also go to my friends and colleagues who helped me in any form of assistance.

iv
Table of Contents

List of Figures…………………………………………………………………………....vii

List of Tables…………………………………………………………………………....viii

List of Acronyms……………………………………………………………………...….ix

Chapter one: Introduction ............................................................................................1

1.1. Statement of the problem ...................................................................................3

1.2. Broad objective, research questions and hypotheses ...........................................7

1.3. Research Methodology ......................................................................................9

1.4. Significance of the study.................................................................................. 10

1.5. Scope of the study ........................................................................................... 10

1.6. Limitation of the study .................................................................................... 11

1.7. Structure of the study....................................................................................... 12

Chapter two: Literature review ................................................................................. 13

2.1. Theoretical review of Non-performing loans ....................................................... 13

2.1.1. Definition of Non-performing loans .............................................................. 14

2.1.2. Determinants of Non-performing loans ......................................................... 17

2.2. Empirical Studies on the determinants of non-performing loans ....................... 27

2.2.1. Single country studies ............................................................................... 27

2.2.2. Cross country studies ................................................................................ 33

2.2.3. Studies on Ethiopia ................................................................................... 37

v
 2.3. Conclusions and knowledge gap ...................................................................... 39

Chapter three: Research design .................................................................................. 42

3.1. Variable specification, Hypotheses and research qustions .................................... 42

3.2. Research approaches ....................................................................................... 50

3.3. Research Methods adopted .............................................................................. 51

3.3.1. Research method: Quantitative aspect....................................................... 52

3.3.2. Research method: Qualitative aspect ......................................................... 58

Chapter four: Results and Analysis ............................................................................ 60

4.1. Research Hypotheses and Questions .................................................................... 60

4.2. Results ................................................................................................................ 62

4.2.1. Structured review of documents .................................................................... 62

4.2.2. In-depth interview results .............................................................................. 76

4.3. Discussions of Results ........................................................................................ 78

Chapter six: Conclusions and Recommendations ...................................................... 96

6.1. Conclusions ......................................................................................................... 96

6.2 Recommendations .............................................................................................. 100

References ......................................................................................................................102

Appendices ....................................................................................................................111

vi
List of Figures

Figure 3.1: Rejection and non-rejection regions for Durbin-Watson test ……….………55

Figure 4.1: Nonperforming loans of Ethiopian commercial banks………………...…….64

Figure 4.2: Normality test for residuals …………………………………………………70

vii
List of Tables

Table 2.1: Minimum provision requirement…………………………………….……….17

Table 3.1: Description of the variables and their expected relationship ………………...49

Table 3.2: Link between research question/hypotheses, variables and the different

data sources………………………...…….……………………….………58

Table 4.1: Summary of descriptive statistics for dependent and independent variables……63

Table 4.2: Heteroskedasticity Test: White ……………………………………………....67

Table 4.3: Autocorrelation Test: Durbin-Watson ……………………………………….67

Table 4.4: Breusch-Godfrey Serial Correlation LM Test………………….…………….68

Table 4.5: Correlation matrixes of independent variables ………………………………69

Table 4.6: Correlation matrixes of dependent and independent variables ……………....71

Table 4.7: First Step (General) Regression Results………...……………………………73

Table 4.8: Second Step Regression Results……………………………………………...75

viii
List of Acronyms

AIB Awash International Bank

BoA Bank of Abyssinia

BLUE Best Linear Unbiased Estimators

CBB Construction and Business Bank

CBE Commercial Bank of Ethiopia

CIR Cost to Income Ratio

CLRM Classical Linear Regression Model

ECBs Ethiopian commercial banks

EFEX effective exchange rate

GDP Gross Domestic Product

HP Hypotheses

IDV Income Diversification

INFL Inflation

LG loan growth

MoFED Ministry of Finance and Economic Development

NBE National Bank of Ethiopia

NIB Nib International Bank

ix
NPLs Non-performing loans

OLS Ordinary Least Square

OWS ownership structure

RIR real interest rate

ROA Return on Asset

RQ Research Question

UB United Bank

WB Wegagen Bank

x
Chapter one: Introduction

Credit creation is the main income generating activity of banks. The loan portfolio is

typically the largest asset and the predominate source of revenue of banks (Achou and

Tenguh 2008). Even if that is the case, all the risks that banks face, credit risk is

considered as the most harmful as nonperforming loans (NPLs) would impair the bank

profitability and its long-term operation significantly (Ahmed 2006). As noted in

Adebola et al. (2011), increasing amount of NPLs in the credit portfolio of banks is

antagonistic to banks in achieving their objectives and a substantial volume of NPLs

indicates the existence of financial fragility and a cause of worry for banks management

and regulatory authorities.

In the context of the above discussion, many scholars like, Caprio and Klingebiel (2002),

Fofack (2005), Pasha and Khemraj (2009) and Azeem et al. (2012) agreed that, the

existence of banking and financial crises has frequently been allied with a massive

accumulation of NPLs. As noted in Caprio and Klingebiel (2002), during the 1997s East

Asian banking and financial crises, many banks were increased by more than threefold in

their volume of NPLs in the period leading up to the crisis. For instance, in Indonesia,

NPLs of banks represented about 75% of their total loan which led to the collapse of over

sixty banks. Furthermore, Fofack (2005) stated that, most banks in sub-Sahara African

countries were also preceded by a rapid accumulation of NPLs during the 1990s crisis.

For instance, in Benin, all banks were collapsed and their NPLs were above 80% of their

total loans. More recently, there are abundant evidences that the current global financial

1
crisis, which was originated in the US, was also attributed to the rapid default of sub-

prime loans/mortgages (Pasha and Khemraj 2009).

Regardless of the implications of NPLs on smooth functioning of banks, for anticipating

banking and financial crises, the leading causes of these loans remain unknown for most

countries especially in Sub-Saharan Africa (Fofack 2005). This is due to the fact that,

most studies often used NPLs as an explanatory variable to other banking outcomes such

as banks performance and failures. Consequently, there appears to be very limited

number of studies investigated NPLs as an explained variable (Boudriga 2009). Those

limited studies were also conducted in developed countries and banks in developing

economies have so far received inadequate attention in the literature (Swamy 2012).

Coming back to the case of Ethiopia, like other African countries and the rest of the

world experienced, Ethiopian commercial banks (ECBs) were also suffered from serious

financial fragility manifested by high proportion of NPLs (Mehari 2012). Recently, the

NPLs of ECBs have shown a substantial improvement and lowered to an average of 5 %

(National Bank of Ethiopia 2011). However, NPLs of ECBs are still high as compared to

the developing economy banks like, Namibia, Mozambique and Uganda 1. Hence, ECBs

are still expected to reduce their NPLs as low as possible in order to achieve their optimal

profit and ultimately improve the soundness of the financial system.

Despite the above fact, as best of the researcher knowledge, there is only the work of

Negera (2012) that investigated determinants of NPLs in Ethiopian banks. However, the

study was not comprehensive enough as result of different gaps. For instance, the study
1
NPLs of Namibia, Mozambique and Uganda are an average of 1.9%, 2.3% and 2.5% of their total loans
respectively (world bank 2012)

2
did not incorporate macroeconomic variables which have found as a significant

determinates of NPLs in many others studies like, Azeem et al. (2012) and Louzis et al.

(2010). Moreover, the study used only descriptive statistics and correlation matrix for the

entire analysis. However, none of those methods are able to explain movements in a

variable by reference to movements in one or more other variables (Brooks 2008). Hence,

all the above discussed knowledge gaps clearly indicate that, the determinants of NPLs in

ECBs have received inadequate attention in the literature and eventually the need for

further investigation.

In the context of the above discussions, the purpose of this study is to investigate the

determinants of NPLs in ECBs. The remaining discussions in the chapter are arranged in

seven sections. The first section presents the problem statement. Section 1.2 discusses the

broad objective, research question and research hypotheses of the study. Section 1.3

presents the research methodology. Scope of the study, limitations of the study and

significance of the study are discussed in section 1.4, 1.5and 1.76respectively. Finally,

the structure of the study is discussed in section 1.7.

1.1. Statement of the problem

Despite the heavy regulation and ongoing efforts made to control the banking industry in

general and the lending activities in particular, the NPLs problems are still a worldwide

headache and a major concern for both international and local regulators (Boudriga

2009). Given the harmful effects of NPLs on countries economy, in recent years, an

issue of preventing NPLs is highly emphasized on the agenda of banking supervisory

institution and policymakers throughout the world (Socuvkova 2012). As far as NPLs

3
problems left unsolved, it can greatly jeopardized the smooth functioning of banks

through erosion of banks asset and reduction of income through accumulation of losses

and increased provisions to compensate for these losses (Kunt and Detragiache1998). In

the worst scenario, a high level of NPLs in a banking system poses a systemic risk,

inviting a panic run on deposits and sharply limiting financial intermediation, and

subsequently investment and economic growth (Ahmed 2006). According to Fofack

(2005), the fiscal costs of NPLs are also significant. Resolution of these loans is usually

undertaking through the creation of government owned Asset Management Companies

whose main function is to take over NPLs of distressed financial institution.

Consequently, the pressure on government revenues will be aggravated. Usually it

accounted for 10% up to 20% of the GDP to cleaning up banking problems caused by

NPLs (Galindo & Tamayo 2000). This interest is justified by the 1990‟s Sub-Saharan

countries financial crisis that resolution of NPLs accounted for over 17 % and 25% of

GDP in Senegal and Cote d‟Ivoire respectively (Caprio and Klingebiel 2002).

In Ethiopia, the banking environment has undergone many regulatory and financial

reforms like other African countries with the aim of improving profitability, efficiency

and productivity (Lelissa 2007). Despite these changes, currently, the banking industry in

Ethiopia is characterized by operational inefficient, little and insufficient competition and

perhaps can be distinguished by its market concentration towards the big government

owned commercial bank, poor credit risk management practices and eventually less

contribution to the GDP as compared to the developed world financial institutions (Abera

2012, Tefera 2011 and Tilahun 2010). In this regard, there are ample empirical evidences

that operational inefficiency and poor credit risk management are usually associated with

4
sizeable volume of NPLs (Berger and Humphrey 1992, Wheelock and Wilson 1994).

For instance, in 2003 the two biggest state owned commercial banks, Commercial Bank

of Ethiopia (CBE) and Construction and Business Bank (CBB) reported NPLs of 53%

and 40% of their total loans respectively (NBE 2012). These are far from the average

30% NPLs of Sub-Saharan Africa countries that recorded during the 1990‟s crisis

(Fofack 2005). In the same year the two private banks Awash International Bank (AIB)

and Bank of Abyssinia (BOA) also reported NPLs of 25% and 28.5% respectively (NBE

2012). Eventually, the government of Ethiopia being worried about the potential systemic

crisis associated with credit risk and imposed restriction on the proportion of NPLs not to

exceed 5% of their total loan outstanding (NBE 2008).

Following the 2008 NBE declaration2, NPLs of ECBs have shown a significant

improvement and lowered to an average of 5 % (NBE 2011). However, there is a

significant variation on the reduction of NPLs from banks to bank. In some bank the

change is abrupt and surprise while in the others the change is steady and constant. For

instance, in 2006 NPLs of CBE were about 22% of their total loan outstanding.

Surprisingly however, in 2011 it lowered to an average of 0.86%. On the other hand,

within the same time range NPLs of NIB lowered from 8.4% to 5.04 % (NBE 2012). This

is not significant as compared to the performance of CBE.

Despite the above discussion, the recent work of Mehari (2012) argued that, the exciting

reduction of NPLs in ECBs is not resulted from improved credit risk controlling,

measuring and monitoring system. Rather, it is merely from writing off and restructuring

2
The declaration that required all commercial banks not hold a NPLs that exceed 5% of their total loan
outstanding.

5
of loans. As far as both writing off and restructuring of NPLs are a post active

measurement (after the occurrence of NPLs), the issue of preventing NPLs in ECBs is

still in question. Moreover, there are still banks that are not fulfilling the 5% maximum

allowable limit of NPLs. For instance, in 2011 NPLs of CBB was 7%. In addition, NPLs

of Cooperative Bank of Oromia (CBO) and Nib International Bank (NIB), in 2010 was

about14.58% and 7.37% of their total loans respectively (NBE 2012). These are far from

the 5% maximum acceptable limit.

In context of the above discussion, many scholars like, Geda (2006), Tilahun (2010),

Tefera( 2011), Mekasha (2011) and Mehari (2012) agreed that, despite the existence of

some improvement in the general risk management practice, ECBs are still weak on their

credit risk management. In connection to this, the 2009/10 survey of NBE revealed that,

significant proportion of commercial banks in Ethiopia lack: appropriate risk

management strategies and programs, timeframe to review risk management documents,

adequate risk management budget, up-to-date and relevant data for informed decisions,

risk communication, risk management integration with human resource management and

policies, risk management audit review, and initiatives to identify risks. The survey

results conclude that, the existing risk management guidelines should be reviewed and

commercial banks should be required to produce risk management programs acceptable

to the NBE by giving due attention to credit, operational and liquidity risks. Among risks

that ECBs face, credit risk has been dominant risks over the last two years, and will

continue to be key risk over the next five years (NBE 2009/10). Accordingly, all the

above problems (especially the lack of risk management audit review and initiatives to

identify risks) indicate that, ECBs provide less consideration to their credit risk

6
management and ultimately the need for reexamine their credit risk management so as to

reduce their NPLs.

Therefore, all the above discussed problems in the ECBs in relation to credit risk

management in general and NPLs in particular along with the above deviant observation

of NPLs at CBB, NIB and CBO (together with the knowledge gap to be established in the

literature review) need extensive research to understand the cause of NPLs.

1.2. Broad objective, research questions and hypotheses

In the context of the problems highlighted above, the general objective of this study is to

investigate the determining factors of NPLs in Ethiopian commercial banking sector.

Having this general objective, the following sections discussed the research questions and

hypothesis that have been used for investigation purpose.

Research questions (RQ)

In line with the broad objective highlighted above, the following two specific research

questions were formulated:

RQ1. What are the determinants of banks’ non-performing loans in Ethiopia

commercial banking sector?

RQ2. What are the likely causes for the existence of variation on NPLs

performance among Ethiopian commercial banks?

7
Hypotheses of the Study (HP)

The hypotheses of this study were formulated by referring the existing theories and past

empirical studies that have been conducted on the determinants of bank‟s NPLs. The

hypotheses of this particular study are intended to catch the determinants of NPLs

quantitatively through structured review of documents. In line with the broad objective of

the study the following ten hypotheses were formulated.

HP1: There is a significant positive relationship between loan growth of a bank and

bank’s NPLs.

HP2: There is a significant positive/ negative relationship between operational efficiency

of a bank and bank’s NPLs.

HP3: There is a significant negative relationship between financial performance of a

bank and bank’s NPLs

HP4: There is a significant positive/negative relationship between Income diversification

of a bank and bank’s NPLs.

HP5: There is a significant positive relationship between state ownership of a bank and

bank’s NPLs.

HP6: There is a significant negative relationship between size of a bank and bank’s

NPLs.

HP7: There is a significant negative relationship between real GDP growth and bank’s

NPLs.

8
HP8: There is a significant positive relationship between real interest rate and bank’s

NPLs.

HP9: There is a significant positive/negative relationship between inflation and bank’s

NPLs.

HP10: There is a significant positive/negative relationship between exchange rate and

bank’s NPLs.

1.3. Research Methodology

In order to achieve the broad objective of the study, a mixed methods research approach

was adopted. However, by considering the nature of the study, quantitative research

approach was dominantly used. To have a better insight and to gain a richer

understanding about the research problem, the quantitative method was supplemented

with the qualitative one. To collect the necessary data the study mainly used survey of

documents. The survey was administered in a structured document review mode, and it

was intended to elicit relevant data that was used to assess the factors that had impact on

the NPLs of ECBs. In addition to the survey method, in-depth interviews were also

conducted with selected ECBs credit officers.

The target populations of this study were all commercial banks registered by NBE and

under operation in the country at least for the last twelve years. Accordingly, the study

covered two government owned commercial banks namely, Commercial Bank of

Ethiopia (CBE) and Construction and Business Bank (CBB). Besides, the study

included the six leading private commercial banks in the country in terms of both branch

9
network and market share namely, Awash International Bank (AIB), Bank of Abyssinia

(BoA), Dashen Bank (DB), Nib International Bank (NIB), United Bank (UB) and

Wegagen Bank (WB).

1.4. Significance of the study

Despite the devastating effects of NPLs on nations‟ economy in general and banks

operation in particular, the issue of NPLs in most developing economies like Ethiopia

have so far received inadequate attention in the literature. As result, this study will extend

the existing literature by providing evidence on the determinants of NPLs in Ethiopian

commercial banking context by utilizing both macroeconomic and bank specific

variables. Apart from contributing to the literature, this study may also have important

practical implications for commercial banks mangers and bank regulators authorities in

dealing with NPLs management. Moreover, it may also help other researchers as a source

of reference and as a stepping stone for those who want to make further study on the

issue of NPLs in the Ethiopian banking context afterwards. Finally, it may provide a

possible opportunity to all stake holders to gain deep knowledge about the leading cause

of NPLs in Ethiopian commercial banking sector.

1.5. Scope of the study

For the sack of robustness of the results, the scope of the study was restricted to all

commercial banks that are registered by the NBE and that have at least twelve years data

(i.e., 2000-2011). As a result, the study included the two large government owned

commercial banks namely, CBE and CBB and six leading private commercial banks in

10
the country in terms of both branch network and market share namely, AIB, DB, BoA,

WB, UB and NIB. Hence, commercial banks that are established newly in the country

and that do not have a minimum of twelve years data were left in this study.

1.6. Limitation of the study

It obvious that, incorporating all independent variables in a single econometrics model is

impossible. And, that is why a disturbance term is usually incorporated in econometrics

model (Brooks 2008). Hence, the study limited to six bank specific variables (banks‟

loan growth, income diversification, operational efficieincy, financial performance, and

ownership structure and bank size) and four macroeconomic variables (inflation, real

interest rate, exchange rate and GDP growth). Different regulatory factors such as

capital adequacy ratio, loan loss provision and time factors like the May 2011 NBE

directive that require all commercial banks to purchase National Bank of Ethiopia bonds

(contribution for the Great Renaissance Dam) worth 27% of their loan disbursements,

and macroeconomic variables like unemployment rate were left in this study.

Since the quantitative part of this study is mainly analyzed by the ordinary listing square

(OLS) method, all the limitations associated with this method might hinder the outcome

of the study. To mitigate the limitation associated with OLS, diagnostic tests that insure

the validity of the data and econometrics model were conducted. However, the

autocorrelation test measured by the default Durbin-Watson (DW) test statistics did not

able to detect whether the autocorrelation problem existed or not. To the contrary, results

obtained from the Breusch-Godfrey (BG) test statistics shown the existence of

11
autocorrelation problem in this particular study. Hence, the inconsistence among the two

tests might impair the outcome of the study.

As a result of the confidentiality policy of banks, the study limited to the officially

disclosed financial data of banks and the personal perception of credit officers‟ of

selected Ethiopian commercial banks towards the issue. Hence, intentionally or

unintentionally the respondent might be biased. Finally, resource and time constraints

were also some of the factors that hindered the outcome of the research.

1.7. Structure of the study

This study is organized into five chapters. Introduction of the study with respect to

Problem statement and objective of the study are presented in chapter one. Chapter two

presents literature review of the study. In this chapter both the theoretical and empirical

reviews pertaining to the determinants of bank‟s NPLs are discussed. Research design

and methodology are presented under chapter three. This is followed by the results and

analysis of different data source in chapter four. Finally, chapter five presents the

conclusions and recommendations.

12
Chapter two: Literature review

In the preceding chapter, background information of the study with respect to the

research problem and objective of the study were discussed. The purpose of this chapter

is to discuss both theoretical and empirical issues pertaining to the determinants of bank‟s

NPLs. The review has three sections. The first section 2.1 presented theoretical review of

NPLs. Section 2.2 presents a review of empirical studies that have been conducted so far

on determinants of bank‟s NPLs. Finally, conclusions on the literature review and

knowledge gaps are presented in section 2.3.

2.1. Theoretical review of Non-performing loans

According to Issa (2009), from a pragmatic point of view, the rationale behind the

existence of banks is the provision of different types of loans, which in turn are

considered as the main source of the banking profits. Therefore, commercial banks

attempts to invest as much of the available funds as possible, in the form of loans

and credit facilities so as to maximize their profit. This in turn results in the majority

of commercial banking assets being in the form of loans and credit facilities

(Achou and Tenguh 2008). Despite the loan portfolio is typically the largest asset and the

predominate source of revenue of banks, the function of granting credit is not free of risks

(Casu et al. 2006). In practice, loans are considered as the types of investment

which have the highest levels of risks with regards to the difficulty of the funds'

recovery. Commercial banks are exposed to numerous difficulties regarding the

protection and recovery of funds granted in the form of loans and credit facilities.

According to Casu et al (2006), the main difficulty that the commercial banks are

13
exposed to the failure of borrowers to repay their obligations on time. As noted in

Heffernan (2005) the failure of the commercial banks' clients to repay their

obligations caused the emergence of NPLs, and is considered the most serious

financial problems facing commercial banks. Hence, the following sections discussed

the meaning, classification and determinants of NPLs in detail.

2.1.1. Definition of Non-performing loans

The term „‟bad loans‟‟ as described by Basu (2003), is used interchangeably with NPLs

and impaired loans as identified in Fofack (2005). Berger and De Young, (1997) also

considers these types of loans as “problem loans”. Thus these descriptions are used

interchangeably throughout the study.

Theoretically, there is no global standard to define NPLs which could be applied to all

economies of the world (Hou 2006 and Bloem and Gorter 2002). Variations exist in

terms of the classification system, the scope, and contents. Such problem potentially adds

to disorder and uncertainty in the NPL issues (Hou 2006). Thus, the definition of NPLs

varies from one banking system to another according to banking laws and regulations

(Issa 2009). In practical terms, Quantitative and qualitative criteria are used

individually or collectively by credit institutions to identify the situation of the

loan. A quantitative criterion uses numbers of days or months to determine the

weakness of borrowers to repay their debt, while a qualitative criterion uses all the

information about the future of loans and borrowers (Bloem and Gorter 2002).

In referring to the period of NPLs, Rose (2002 p.118) defined NPLs as "a loan is

placed in the NPLs category when any scheduled loan repayment is past due for

14
more than 90 days". In addition, Bloem and Freeman (2005 p.8) give the definition

of NPLs as "a loan is NPLs when payments of interest and/or principal are past

due by 90 days or more, or interest payments equal to 90 days or more have been

capitalized, refinanced, or delayed by agreement". Others consider NPLs as a

borrower stopping to repay the installments in a period of over six months. For

instance, Cho (2002 p.10) define NPLs as "a loan was considered NPLs only when it

was past due six months or more while provisioning requirements".

In light of the above discussion, a study for the International Monetary Fund (IMF),

Cortawarria et al. (2000) define NPLs according to region where they originate from.

For instance, in countries like France, Spain, Portugal, Switzerland and Norway, loans

became NPLs when principal and interest uncollected for more than 90 days. Others

countries like Greece and Italy used more than 90 days. In countries like U.K and

Germany there is no explicit criteria to be used in determining loans as good or bad.

As per NBE(2012 p.3), NPLs are defined as “loans or advances whose credit quality has

deteriorated such that full collection of principal and/or interest in accordance with the

contractual repayment terms of the loan or advances in question”. It further provides that:

“Short term loans are NPLs when principal and/or interest is due and uncollected for

90(ninety) consecutive days or more beyond the scheduled payment day or maturity.

Medium and long term loans are NPLs when principal and/or interest is due and

uncollected for 12(twelve) consecutive months or more beyond the scheduled payment

day or maturity”.

15
According to NBE (2012) directive, Ethiopian commercial banks are required to classify

their loans as pass, special mention, substandard, doubtful and loss in accordance with

Bank for International Settlements (BIS)3 standards as presented below:

Pass: loans in this category are fully protected by the current financial and paying

capacity of the borrower and not subject to any criticism.

Special mention: Short term loans past due for 30 days or more, but less than 90 days

and medium and long term loans past due for 6 month or more, but less than 12 months.

Substandard: Short term loan past due for 90 days or more, but less than 280 days and

medium and long term loans past due for12 months or more, but less than 18 months

Doubtful: Short term loan past due for 280 days or more, but less than 360 days and

medium and long term loans past due for18 months or more, but less than 3 years.

Loss: Short term loan past due for 360 days or more, and Medium and long term loans

past due for3 years or more.

According to NBE (2012) directive, all Ethiopian commercial banks required holding

provisions for each loans mentioned above so as to absorb the potential losses in their

loans portfolio. The minimum provision requirements are mentioned below:

3
Bank for International Settlements (BIS) is an international banking supervisory institution that was
established by the central bank governors of the Group of Ten countries in 1975. It consists of senior
representatives of bank supervisory authorities and central banks from Belgium, Canada, France, Germany,
Italy, Japan, Luxembourg, the Netherlands, Spain, Sweden, Switzerland, the United Kingdom, and the
United States (Coen 2010).

16
Table2.1 Minimum provision requirement

Classification of loans Minimum percentage

pass 1%

Special mention 3%

Substandard 20%

Doubtful 65%

loss 100%

Source: NBE (2012)

Among the above classified loans, the last three loans such as, substandard, doubtful and

loss are considered as NPLs (NBE 2012). In fact, such type of classification is in

accordance with BIS standards except some variation on the percentage of provision

requirement. For instance, as per BIS standard doubtful loans are required 50% provision

whereas under the NBE directive banks are required to hold 65% provision.

2.1.2. Determinants of Non-performing loans

Unfortunately, there is no particular theoretical framework that emphasizes on the

determinants of NPLs (Issa 2009). However, with the major contribution of Akerlof

(1971), the asymmetric information as concepts have been used to analyze the

individual behaviour in the market in relation to having knowledge in transactions

or exchanges. These concepts can be extended to NPLs, since, NPLs are the result

of a particular behavioural pattern emerging from moral hazard on the side of

17
borrower and adverse selection on the side of lenders (Issa 2009). Therefore, the

concepts of asymmetric information can be examined to give further meaning and

to understand behavioural aspects of NPLs.

According to Arestis and Sawyer (2001), the first important theoretical concept in

relation to NPLs, as the articulation of asymmetric information, is the adverse

selection issue. Adverse selection problem occurs before the transaction takes place,

in the event that the lender's inability to distinguish between a high-risk borrower and

a low-risk borrower is compromised. In this regard, Hafer (2005) noted that,

increasing the interest rate and required additional collateral lead the low risky

clients to go elsewhere in order to obtain loans, while the high risky clients will

accept the conditions at hand. In other word, those who want to take on big risks are

likely to be the most eager to take out a loan, even at a high rate of interest, because they

are less concerned with paying the loan back.

As noted in Breuer (2006) conflict of interest between bank managers and shareholders

may aggravate the adverse selection. Bank managers have short term decision horizons

because their reputations are strongly influenced by public perceptions of their

performance, as evidenced by short term earnings. Managers‟ reputations suffer if they

fail to expand credit when the economy is expanding and bank earnings are improving.

This herd behavior will result in some loans going to customers with higher default risk.

In addition, the macroeconomic condition may also aggravate the adverse selection.

During the expansion phase of the economy banks characterized by a relatively

low number of NPLs, as both consumers and firms face a sufficient stream of

income and revenues to service their debts. However as the booming period continues,

18
credit is extended to lower quality debtors and subsequently, when the recession

phase sets in, NPLs increase (Fisher 1933, Minsky 1986, Kiyotaki and Moore

1997, Geanakoplos, 2009).

According to Arestis and Sawyer (2001), the second theoretical concept derived

from asymmetric information is `moral hazard', which can be applied to NPLs. A

borrower may have incentives to misallocate funds for personal use and to undertake

investment in unprofitable projects that serve only to increase their personal power or

stature. Thus, a lender is subjected to the hazard that the borrower has incentives to

engage in activities that are undesirable from the lender‟s point of view: that is, activities

that make it less likely that the loan will be paid back. In addition, banks credit managers

may intentionally provide loans to lower quality debtors.

Ultimately, it could be concluded that asymmetric information often leads to the

emergence of the economic and financial problems especially NPLs in the credit

market. Therefore, it can be said that the economic and commercial banks will

not operate as efficiently as they should, in the absence of sufficient information,

related to both clients and the general environment. Hence, by having theory of

asymmetric information as standing point the following sections discussed the determents

of NPLs that are intended to cover under this study. It should be noted that all the

determinants of NPLs that are discussed in the following sections do not necessarily have

a direct relationships with theory of asymmetric information.

19
2.1.1.1. Bank specific factors

Bank-specific variables refer to those factors which characterized individual banks.

Those factors can be influnecd by managerial decisions and usually associated with the

specific policy choices of a particular bank with regard to its efforts to maximize

efficiency and improve its risk management. Hence, bank specific variables that are

usually theorized as determinats of NPLs are include , loan growth, financial

perfrormance, bank size, ownership structure, the quality of the loan portfolio and

operatinal efficicecy of bank‟s. Hence, the following part of this particular section clearly

presents the bank-specific variables that are used in this study.

Loan growth: The credit policy of the bank plays an essential role in determining the

subsequent levels of NPLs. Loan growth has a direct (positive) association with the

volume of NPLs reported by commercial banks (Sinkey and Greenwalt (1991), Keeton

(1999), Salas and Saurina (2002), Jimenez and Saurina (2006) and Metaxas et al (2010)).

To maximize the short run benefits, managers seek to rapidly expand credit activities and

may hence take inadequate credit exposures. In this regard, Keeton (1999) suggests that

rapid growth of loans can be triggered by return maximization strategies. Particularly,

during periods of economic growth, the financial institutions engage in market share

conquest campaigns discarding the necessary assessment of credit quality of borrowers

(Fernandez De Lis et al., 2000). The search for rapid growth of loans is achieved by

either reducing interest rate charged to borrowers or by lending to lower credit quality

borrowers. This will lead, through adverse selection reasoning in which banks lend to

lower credit quality borrowers and ultimately increase the probability of NPLs.

20
Operational efficiency: Theoretically, the relationship between operational efficiency of

a bank and bank‟s NPLs might appear to be largely unrelated, because operations

personnel typically do not participate in screening and monitoring loan customers, and

because loan officers typically do not participate in overseeing operations costs.

However, issues of NPLs and cost of efficiency are in fact related in several important

ways. Hence, the impact of operational efficiency on NPLs can be positive or negative.

In one hand, a number of researchers have found that failing banks tend to be located far

from the best-practice frontier (Berger and Humphrey 1992, Barr and Siems 1994,

DeYoung and Whalen 1994, Wheelock and Wilson 1994). A number of other studies

have found a positive relationship between low cost of efficiency and NPLs even among

banks that do not fail (Hughes and Moon 1995, Resti 1995). According to Berger and

DeYoung (1997), the positive association among low cost efficiency and NPLs is

hypothesized as „bad management‟ hypothesis. The basic argument here is that, low cost

efficiency considers as signal of poor managerial performance (bad management), which

can greatly affects loan granting behavior of a bank. As „bad‟ managers, they may a)

have poor skills in credit scoring and therefore choose a relatively high proportion of

loans with low or negative net present values, b) be less than fully competent in

appraising the value of collateral pledged against the loans, and c) have difficulty

monitoring and controlling the borrowers after loans are issued to assure that covenants

are obeyed. Hence, under the „bad management‟ hypothesis, there is a positive

association between low cost efficiency and NPLs. On the other hand, low cost efficiency

may have a negative impact on NPLs. According to Berger and DeYoung (1997) this

relationship is hypothesized as „Skimping‟ hypothesis. According to this view, there

21
exists a trade-off between allocating resources for underwriting and monitoring loans and

measured cost efficiency. In other words, banks which devote less effort to ensure higher

loan quality will seem to be more cost-efficient; however, there will be a growing number

of NPLs in the long run. Hence, banks which have adequate budget to screening loan

customers, appraising collateral, and monitoring and controlling borrowers after loans are

issued seem to be cost inefficient in the short run. However, the volume of NPLs tends to

be lower as compared to banks which do not have adequate budget to ensure higher loan

quality. Thus, under the skimping hypothesis, the association between measured cost

efficiency and NPLs is negative.

Financial performance: the financial performance of a bank is usually related to the

risk taking behaviour of managers (Hu et al. (2004), Jimenez and Saurina (2006), Jellouli

et al (2009), Metaxas et al (2010) and Vogiazas and Nikolaidou (2011)). As noted in Hu

et al. (2004), profitable banks are less engaged in risky activities as they have less

pressure to create revenues. Profitable banks have an opportunity to choose a loan

applicant who has strong financial performance and lower risk. Hence, as the profitability

of banks increases, the probability that managers engaged in risky investment will reduce

and ultimately the probability that loans become a nonperforming loans will also reduce

with the same manner. To the contrary, unprofitable (inefficient) banks might engage in

risky lending activity in particularly when managers have short term incentives. As long

as banks engaged in risky activity the likelihood that loans become default is high and

ultimately resulted with sizeable volume of NPLs.

Income diversification: The traditional argument based on Diamond (1984) suggests the

wisdom of not putting all eggs in one basket. Recently however, there is no general

22
consensus on the benefit of income diversification. On one side, there are supporter of the

concept of portfolio theory which states that banks can reduce firm-specific risk by

diversifying their portfolios as it makes possible the compensation for losses in some

products by gains in others (Winton 1999, Templeton and Severiens 1992 and Gallo et al.

1996). Hence, the potential losses on the loan activity might be overcome by looking for

non interest sources of revenues (financial revenues and capital gains). On the other side,

scholars like, Maksimovic and Philips (2002), DeYoung and Roland (2004) and Stiroh

(2006) argued that, diversification of revenue does not guarantee low level of NPLs.

Because, too many operating items make the banks lose their focus on specialized field

and reduce their monitoring effectiveness that may increase the probability of failure.

Hence, banks should focus on a single line of business so as to take greatest advantage of

management‟s expertise which ultimately reduced the probability of NPLs.

Ownership structure: According to the Coase Theorem, cited in Hu (2006), the

assignment of property rights (ownership) will not affect economic efficiency as long as

the transaction cost is zero. However, the real world is imperfect and the transaction cost

can be sufficiently high. In an imperfect world with high transaction costs, ownership

does matter to economic efficiency and making different ownership types is associated

with different transaction costs (Cooter and Ulen 2000). In this regard, most existing

literature suggested that state-owned banks are usually associated with high NPLs than

privately owned banks. Salas and Saurina (2002) argue that to enhance the economic

development of the country, state-owned banks have more incentives to fund riskier

projects and to allocate more favorable credits for small and medium firms. Private

institutions clearly have an incentive to solve adverse selection and moral hazard

23
problems and lend to borrowers who have productive investment opportunities.

Governments have less incentive to do so because they are not driven by the profit

motive. The absence of a profit motive also means that state-owned banks are less likely

to manage risk properly and be efficient (Hu 2004). This inadequate risk taking behavior

(compared to the return profile) will lead to a higher level of NPLs.

Bank size: The existing literature provides evidence that suggests a negative association

between size of a bank and bank‟s NPLs (Salas and Saurina (2002), Hu et al. (2004),

Cole et al. (2004), Micco et al. (2004), Garcia and Robles (2007) and Swamy (2012)). As

noted in Hu et al. (2004), large banks have more resources and are more experimented for

efficient information gathering, processing and analyzing to tackle moral hazard and

adverse selection and ultimately better deal with bad borrowers. Small banks, on the

contrary, may be exposed to the adverse selection problem because of the lack of

sufficient competencies and experience to effectively assess the credit quality of

borrowers. In addition, Cole et al. (2004) suggested that, smaller banks adopt small

business loan underwriting practices. Hence, the extents that the failure rates of small

businesses are higher than those of larger and established firms.

2.1.1.2. Macroeconomic factors

The existing literature provides evidence that suggests a strong association between NPLs

and macroeconomic factors. Several macroeconomic factors which the literature proposes

as important determinants of NPLs are: real GDP growth, inflation rate, effective

exchange rate, real interest rate, unemployment rate, broad money supply (M2) and GDP

per capital (Salas and Suarina 2002, Fofack 2005 and Jimenez and Saurina 2005). This

24
study only considers the growth in real GDP, annual inflation rate, real interest rate and

the real effective exchange rate.

Real GDP growth: there is an inverse relationship between GDP growth and the level of

NPLs reported by commercial banks (Salas and Suarina(2002), Jajan and Dhal (2003),

Fofack (2005) , Hou (2006) , Jimenez and Saurina (2005), Pasha and Khemraj (2009),

Louzis et al. (2010) and Azeem et al. (2012)). The explanation provided by the literature

for this relationship is that, Changes in business cycle impact the credit worthiness of

borrowers in terms of repayment capacity. Hence, strong positive growth in real GDP

usually translates into more income which improves the debt servicing capacity of

borrower which in turn contributes to lower NPLs. Conversely, when there is a slowdown

in the economy (low or negative GDP growth), the economic activities in general are

decreasing and the volume of cash held for either businesses or households is

decreasing. These conditions contribute in deteriorating the ability of borrowers to

repay the loans, which lead to increase the likelihood of delays their financial

obligations and thus banks‟ exposure to credit risk increase. In this regard, Hou (2006)

noted that, each NPL in the financial sector is viewed as an obverse mirror image of an

ailing unprofitable enterprise.

Real Interest rate: Asymmetric information and the resulting adverse selection problem

can lead to “credit rationing,” in which some borrowers are denied loans even when they

are willing to pay a higher interest rate (Stiglitz and Weiss 1981). This occurs because as

interest rates rise, prudent borrowers are more likely to decide that it would be unwise to

borrow, whereas borrowers with the riskiest investment projects are often those who are

willing to pay the highest interest rates. In this general setting, a higher interest rate leads

25
to even greater adverse selection; that is, the higher interest rate increases the likelihood

that the lender is lending to a bad credit risk and ultimately increases NPLs (Sinkey and

Greenwalt (1991), Jimenez and Saurina (2006), Pasha and Khemraj (2009), Ahmad et al

(2009) and Metaxas et al (2010)).

Inflation: Inflation affects borrowers‟ debt servicing capacity through different channels

and its impact on NPL can be positive or negative (Fofack 2005, Pasha and Khemraj

(2009) and Nkusu 2011). The explanation provided by the literature for this relationship

is that, higher inflation can make debt servicing easier by reducing the real value of

outstanding loans particularly when the loan rates are fixed (banks do not adjust rates in

accordance to the inflation change to maintain their real returns). However, it can also

weaken some borrowers‟ ability to service debt by reducing real income. Moreover, when

loan rates are variable(adjusted in accordance to the inflation change), inflation is likely

to reduce borrowers‟ loan servicing capacity as lenders adjust rates to maintain their real

returns or simply to pass on increases in policy rates resulting from monetary policy

actions to combat inflation. Against this background, the relationship between NPL and

inflation can be positive or negative.

Real effective Exchange rate: like inflation a change in effective exchange rate can also

affects borrowers‟ debt servicing capacity through different channels and its impact on

NPL can be positive or negative (Nkusu 2011). As noted in Pasha and Khemraj (2009),

depreciation of the exchange rate can have mixed implications on borrowers‟ debt

servicing capacity. On the one hand, it can improve the competitiveness of export-

oriented firms. As long as the value of domestic currency depreciated (lower), export-

oriented firms can dominate the international market at lower price (since their

26
production cost is covered in domestic currency which has lower value than foreign

currency and their revenue is collected in foreign currency which has higher value as

compared to the domestic currency. Hence, depreciation of exchange rate can improve

the debt-servicing capacity of export-oriented borrowers. On the other hand, it can

adversely affect the debt-servicing capacity of borrowers who borrow in foreign currency

(import-oriented firms).

2.2. Empirical Studies on the determinants of non-performing loans

There are many studies that have been conducted so far in determinants of bank‟s NPLs.

Their results have shown that, banks NPLs are determined by either internal or external

factors. Hence, the following section presents the empirical evidence on the determinants

of bank‟s NPLs with a particular focus on those studies that have been conducted more

recently, as far as they are the best indicators of the current situation. Hence, section

2.2.1 presents studies that have been conducted in single country. This is followed by a

review of cross countries studies in section 2.2.2. Finally, Section 2.2.3 presents related

studies that have been conducted in the Ethiopian context.

2.2.1. Single country studies

One of the earliest studies on the determinants of NPLs is the work of Keeton and Morris

(1987), who investigated the fundamental drivers of loan losses for a sample of nearly

2,500 US commercial banks for the period 1979–1985. Using simple linear regressions,

they found that local economic conditions along with the poor performance of certain

27
sectors explain the variation in loan losses recorded by the banks. The study also reported

that commercial banks with greater risk appetite tend to record higher losses.

Several studies which followed the publication of Keeton and Morris (1987) have since

proposed similar and other explanations for problem loans in the US. For instance,

Sinkey and Greenwalt (1991) investigated the loan loss experience of large commercial

banks in the US from 1984 to 1987 by using a simple log linear regression model. They

found that both internal and external factors explain the loan loss rate of US banks. These

authors found a significant positive relationship between the loan loss rate and internal

factors such as high interest rates, excessive lending, and volatile funds. In addition, they

reported that depressed regional economic conditions also explain the loss rate of the

commercial banks. In addition, Keeton (1999) analysed the impact of credit growth and

loan delinquencies in the US banks from 1982 to 1996 with a vector auto regression

model. The result has shown that, there is a strong relationship between credit growth and

impaired assets. Specifically, rapid credit growth was associated with lower credit

standards and contributed to higher loan losses in US banks.

Studies that have been conducted in other financial systems also provided similar results

to those in the US. For instance, Bercoff et al (2002) examined the fragility of the

Argentinean Banking system over the periods of 1993-1996 by using survival analysis.

They argue that NPLs are affected by both bank specific factors and macroeconomic

factors. In addition, Salas and Saurina (2002) investigated the determinants of problem

loans of Spanish commercial and saving banks using a dynamic model and a panel

dataset covering the period 1985-1997. The study revealed that real growth in GDP, rapid

28
credit expansion, bank size, capital ratio and market power explain the variation of NPLs

in Spanish commercial banking sector.

By using the panel regression model, the work of Jimenez and Saurina (2006) analyzed

credit cycles, credit risks, and prudential regulation of Spanish commercial banks. The

study utilized the ratio of NPLs as a dependent variable, and on the other side,

independent variables are represented by the growth rate of GDP, the real interest rate,

total collaterized loans, size of the banks, and two Herfindhal indexes (one for

region and one for industry). The study found that the macroeconomic control variables,

i.e. growth rate of GDP and real interest rates are both significant. The acceleration of

GDP as well as decline in real interest rates brings about decline in problem loans. The

size of the bank does not have a significant association to the problem loan ratio. Finally,

the rate of loan growth found to be highly significant at the 1 percent level and has

positive association with problem loans.

Hu et al (2006) examined the relationship of ownership structure, size of banks and

income diversification with NPLs of commercial banks in Taiwan with a panel dataset

covering the period 1996-1999. The study shows that banks with higher government

ownership recorded lower NPLs. Hu et al (2006) also show that bank size is negatively

related to NPLs while diversification has not found a significant association with banks

NPLs in Taiwan commercial banking sector.

Pasha and Khemraj (2009) empirically examined the determinants of NPLs in the

Guyanese banking sector using a panel dataset. The result revealed that, both

microeconomic and macroeconomic variables have significant association with NPLs.

29
Among macroeconomic variables, exchange rate and GDP growth found out a significant

relationship with NPLs. Specifically, the real effective exchange rate has a significant

positive impact on NPLs. This indicates that whenever there is an appreciation in the

local currency the NPLs portfolios of commercial banks are likely to be higher. On the

other hand, the study found that, GDP growth is inversely related to NPLs, suggesting

that an improvement in the real economy translates into lower NPLs. Among micro-

specific variables, Interest rate and credit growth have a significant positive association

with NPLs. This indicated that, banks which charge relatively higher interest rates and

lend excessively are likely to incur higher levels of NPLs. However, contrary to previous

studies, the study does not support the view that large banks are more effective in

screening loan customers when compared to their smaller counterparts.

Ahmad et al (2009) examined factors affecting banks credit risk in Jordan banks from

1995-2005 using dynamic panel regression analysis. Their findings reveal that at the

macroeconomic level, GDP, inflation and market interest rate have significant impact on

credit risk. While at microeconomic level, previous NPLs, loan growth, loan

concentration and bank size are significant determinants. Specially, conditions associated

with good economic periods contribute in decreasing the banks‟ credit risk exposure.

Louzis et al (2010) examined the determinants of NPLs in the Greek financial sector by

using dynamic panel regression method. The study utilized a panel data set comprising

the 9 largest Greek banks (which account for approximately 90% of Greece‟s banking

sector) for the period 2003-2009 includes both a period of growth as well as the

downturn. The study concluded that macroeconomic variables, specifically the real GDP

growth rate, the unemployment rate and the lending rates have a strong effect on the level

30
of NPLs. Furthermore, bank specific variables such as financial performance and

efficiency have also a strong relationship with the level of NPLs.

Ramachandran et al (2011) empirically investigated determinants of the credit risk in the

Indian commercial banking sector by using an econometric model. The study utilized a

panel data at bank level for 22 public sector banks and 15 private sector banks. The study

revealed that both macroeconomic and bank specific factors play crucial role in

determining the credit risk of the commercial banking sector. Specifically, the lagged

nonperforming assets had a strong and statistically significant positive influence on the

current non-performing assets. There is also a significant inverse relationship between the

GDP and the credit risk for both public and private sector banks.

Jusoff et al (2011) investigated the relationship between NPLs, interest rate and inflation

rate in Malaysian banking sector for the period January 2006- December 2009. The

study utilized a vector Error correction model (VECM) to determine whether interest rate

and inflation rate may affect the NPLS based on 48 monthly data. The long run

relationship shows that interest rate has a significant relationship towards NPL. On the

other hand, there is no significant relationship between inflation rate and NPL. In the

short run relationship, the finding shows that both inflation and interest rate cannot

influence the NPLs in Malaysian commercial banking sector.

Vogiazas and Nikolaidou (2011) investigated the credit risk determinants of the

Bulgarian banking sector by means of time series modeling approach. The study was

motivated by the hypothesis that macroeconomic variables such as, interest rates,

financial markets‟ and bank-specific variables have a role to play on the NPLs in the

31
Bulgarian banking system. Using monthly data series from January 2001 to December

2010 that covers both the booming period and the recent global financial turmoil.

Findings indicated that, the macroeconomic and financial markets‟ variables,

specifically the unemployment rate, the construction index, the industrial

production index and the real effective exchange rate jointly with the credit

growth and the global financial crisis influence the NPLs of Bulgarian banks.

Farhan et al (2012) investigated the economic factors causing NPLs in the Pakistani

banking sector using a primary data collected via a structured questionnaire from 201

bankers who are involved in the lending decisions or analyze the credit risk or

handling NPLs portfolio. Correlation and regression analysis was carried out to analyze

the impact of selected independent variables(Interest Rate, Energy Crisis,

Unemployment, Inflation, GDP Growth, and Exchange Rate) on the NPLs of

Pakistani banking sector. According to the results Pakistani bankers perceive that

Interest Rate, Energy Crisis, Unemployment, Inflation, and Exchange Rate has a

significant positive relationship with the non-performing loans of Pakistani banking

sector while GDP growth has significant negative relationship with the non-performing

loans of Pakistani banking sector. This study also discusses how good loans are turning

into bad loans due to disaster in energy sector of Pakistan and how these energy crisis are

badly affecting the banking sector of Pakistan

Recently, Swamy (2012) examined the determinants of default risk of Indian commercial

banks using panel data set over the period of 1997-2009. The study used quantitative

method with econometrics module. A multiple regression result revealed some interesting

32
inferences contrary to the perception of few opinion makers. Lending rates have been

found to be not so significant in affecting the NPLs, which is contrary to the general

perception. Bank size has a significant and negative association with level of NPLs,

indicating that large banks may have better risk management procedures and technology

which definitely allows them to finish up with lower levels of NPLs compared to smaller

banks. Further, this study suggests that private banks and foreign banks have advantages

in terms of their efficiencies in better credit management that contains NPLs. That

indicates that bank privatization can lead to better management of NPLs.

Very recently, Azeem et al. (2012) also investigated the determinants of NPLs of US

commercial banks from 1985-2010 by using correlation and regression tests. Their

result have shown that a significant association between NPLs and real interest rate and

GDP. Particularly, real interest rate has a significant positive relationship with NPLs of

US commercial banking sector while GDP growth has significant negative relationship.

2.2.2. Cross country studies

Fofack (2005) investigated the leading causes of NPLs during the economic and banking

crises that affected a large number of countries in Sub-Saharan Africa in the 1990s. The

study used correlation, causality analysis and pseudo-panel models based on data drawn

from 16 African countries. The result revealed that both macro and microeconomic

factors have significance association with NPLs of banks in Sub-Saharan African

countries. The econometric analysis have shown a strong causality between NPLs and,

economic growth, real exchange rate appreciation, the real interest rate, net interest

margins and interbank loans. Specifically, the dramatic increase in the size of NPLs is

33
largely driven by macroeconomic volatility and reflects the vulnerability of undiversified

African economies, which remain heavily exposed to external shocks. Macroeconomic

stability and economic growth are associated with a declining level of NPLs; whereas

adverse macroeconomic shocks coupled with higher cost of capital and lower interest

margins are associated with a rising scope of NPLs. Interesting enough, inflation does

not appear to be particularly significant in explaining the dynamics on NPLs.

Masood and Aktan (2009) investigate the factors that contributed to the increase in non-

performing loans (NPLs) in large state-owned commercial banks over the period 1999–

2001 and 1996–1998 in Turkey and Pakistan respectively. This paper uses the survey-

based methodology to investigate the determinants of NPLs. Data are mainly obtained

from credit managers of Turkey and Pakistan commercial banks through questionnaire.

The study found that external government intervention; loans to insiders and poor credit

risk assessment have significant influence in the size of NPL in Turkey. The result also

show that communication facilities provided to credit managers, credit managers' years of

service in the banks and years of experience in the bank of credit managers, were the

significant determinants of non-performing loans in Pakistan.

Jellouli et al (2009) empirically analyzed the determinants of problem loans and the

potential impact of both business and institutional environment on credit risk exposure of

banks in the Middle East and North Africa (MENA) countries region based on a sample

composed of 46 commercial banks from 12 countries over the period 2002-2006. The

study used a random-effects panel regression model that controls for cluster effects at the

country level. Their results have shown that, foreign participation from developed

34
countries reduces the NPL level; highly capitalized banks experience high levels of

NPLs, high credit growth is associated with a reduced level of NPLs. Concerning

business environment factors, it appears that only the relevance of information published

by credit bureaus favorably impacts the credit exposure of banks.

Boudiga (2009) empirically analyzed the cross-countries determinants of NPLs and the

potential impact of regulatory factors on credit risk exposure. The study used aggregate

banking, financial, economic and legal environment data for a panel of 59 countries over

the period 2002-2006. The study used the ratio of NPLs as dependent variable. On the

other hand, the independent variables includes capital to risk-weighted assets minus the

required minimum capital, one year lagged loan loss reserves to total loans ratio, one

year lagged return on assets ratio, percentage of state-owned banks, percentage of

foreign ownership, percentage of assets held by the five largest banks and one

year lagged real GDP growth. The results imply that higher capital adequacy ratio

and prudent provisioning policy are correlated with a decrease in NPLs. In

addition, private ownership, foreign participation and bank concentration have a positive

association with NPLs. Interestingly, the regulatory devices are either

counterproductive or do not significantly improve the credit risk exposure for

countries characteristic with weak institutions, corruption environment and little

democracy. Further, the presence of state does not enhance the NPLs problem either.

Espinoza and Prasad (2010) attempted to ascertain the determinants of NPLs in the Gulf

Cooperative Council (GCC) banking sector. The study used data drawn from 80 banks in

the region over the period of 1995–2008. The result of multiple regressions models

supported the view that both macroeconomic factors and bank-specific characteristics

35
determine the level of NPLs. In particular, the study found strongly significant and

inverse relationship between real GDP growth and NPLs. The study also showed that

global financial market conditions have an effect on NPLs of banks. Among bank

specific variables factors, efficiency and past expansion of the balance sheet were found

to be significant. High credit growth in the past could generate higher NPLs in the future.

Finally, larger banks would also have lower NPLs in GCC banking sector.

Nkusu (2011) empirically investigated the link between NPLs and macroeconomic

performance on a sample of 26 advanced countries that spans the period from 1998 to

2009 using two complementary approaches. First, the study investigated the

macroeconomic determinants of NPL in panel regressions and confirms that adverse

macroeconomic developments are associated with rising NPLs. Second, the feedback

between NPLs and its macroeconomic determinants is investigated in a panel vector

autoregressive (PVAR) model. The findings of the study suggested that, deterioration in

the macroeconomic environment such as adverse shock to GDP growth, higher inflation,

unemployment or falling asset prices is associated with debt service problems, reflected

into rising NPL. Conversely, a favorable macroeconomic environment is associated with

subdued NPL.

Castro (2012) analyzed the link between the macroeconomic developments and the

banking credit risk in a particular group of countries those recently affected by

unfavorable economic and financial conditions such as Greece, Ireland, Portugal,

Spain and Italy (GIPSI). Employing dynamic panel data approaches to those five

countries over the period 1997-2011. The result conclude that the banking credit

risk is significantly affected by micro and macroeconomic variables. Specifically,

36
GDP growth has a significant negative relationship with banks NPLs. On the other hand,

NPLs of banks have a significant positive association with rises of unemployment

rate, interest rate, credit growth increase and appreciation of real exchange rate.

De Bock and Demyanets (2012) examined how credit growth and asset quality in 25

emerging markets (EM) relate to different domestic and external factors by using

dynamic panel regressions analysis over the period 1996-2010. The panel regressions

result has shown that a slowdown in economic growth, a weaker exchange rate or terms

of trade, and rapid credit growth are independently associated with higher NPL levels.

2.2.3. Studies on Ethiopia

In the context of Ethiopia, there appears to be very limited studies on the determinants of

bank‟s NPLs. To the knowledge of the researcher there is only the work of Negera (2012)

that investigated the determinants of NPLs in the context of ECBs. However, there are

also other studies that investigated the other aspect of NPLs in ECBs. These studies are

the work of Tilahun (2010) and Ayalew (2009). Thus, this particular section provides a

detailed review of those related studies conducted in the context of Ethiopia.

Negera (2012) assessed the determinants of NPLs in Ethiopian commercial banking

sector using a survey data collected from both private and state owned commercial Banks

using a self-administered questionnaire. In addition to the survey data, the study used

interview with senior bank officials. A descriptive statistics and correlation matrix were

used so as to analyze the data. The findings of the study shows that poor credit

assessment, failed loan monitoring, underdeveloped credit culture, lenient credit

terms and conditions, aggressive lending, compromised integrity, weak institutional

37
capacity, unfair competition among banks, willful default by borrowers and their

knowledge limitation, fund diversion for unintended purpose, over/under financing by

banks ascribe to the causes of loan default. However, the study outcome failed to

support the existence of relationship between banks size, interest rate they charge

and ownership type of banks and occurrences of nonperforming loans.

On the other hand, the study of Tilahun (2010), identified the underlying NPLs

management difficulty of ten privately owned commercial bank in Ethiopia given in

to consideration that the managements are different from bank to banks based on their

perception towards the NPLs. The study applies mixed research approach. The data

obtained from questionnaires and unstructured interviews presented and analyzed through

descriptive statistics. The result revealed that, Credit Policy and supervision by the

management has less contribution to the NPLs. Most of the NPLs are caused by factors

after the loan released like moral hazard of the borrower, ineffective monitoring, and

operational loss of the borrower. Ineffective monitoring system has contributed to the

NPLs because of less amount and type of information assessed, understaffing and

lack of communication among branches of the bank. Dealing with NPLs on a timely

manner is also another problem identified by the paper. The directive from NBE

have a positive impact by creating safe ground the bank can do business and by giving

absolute right to the banks to take their collateral at any time the bank need to do so

without notifying courts.

By using a mixed research approach, Ayalew (2009), investigated legal problems in

realizing nonperforming loans of Ethiopian commercial Banking sector. The finding

of the study indicated that legal gaps that exist in procedural laws and institutional

38
problems affect the resolution process. Furthermore, the study argued that, issuing

appropriate laws covering financial securities, establishing a comprehensive

institutional framework including Asset Management Companies (AMCs) with clear

accountability and transparency are found to be very important.

2.3. Conclusions and knowledge gap

The literature review that are discussed so far showed that, banks NPLs are determined

by macroeconomic and bank specific factors. The empirical evidence shows that,

favorable macroeconomic conditions, such as sustained economic growth, low

unemployment and interest rates, tend to be associated with a better quality of bank loans.

The studies in general depicted the association between real GDP growth, inflation,

real interest rate, unemployment rate and effective exchange rate . On the other hand,

bank specific factors like, bank size, financial performance, operational efficiency, rapid

loan growth, ownership type, income diversification, risk assessment and monitoring are

found to be having significance on the occurrence of NPL.

However, Most of the literature that are discussed so far appeared to have focused on

studies that were conducted in the banking sector of developed economies (such as united

state of American, Spanish, Greek and Italian) and some emerging economies (such as

Indian, Chinese, Malaysian, and Indonesia). Consequently, the Banking sectors in most

developing economies like Ethiopia have so far received inadequate attention in the

literature. Moreover, NPLs of different countries does not necessarily share identical

immediate causes since those studies were based on the data from diverse countries.

Apart from the data originated from, those literatures by themselves provided

39
contradictory conclusions because of different models and methodologies they used.

Hence, their results may not be applicable to Ethiopian banking sector.

In the context of Ethiopia, the related study conducted by Negera (2012), assessed the

determinants of NPLs in Ethiopian commercial banks by using bank-specific variables.

Accordingly, this study clearly failed to identify macroeconomic determinants of NPLs

which have found as a significant determinates of NPLs in many others studies like,

Azeem et al. (2012) and Louzis et al. (2010). Furthermore, the study used only

descriptive statistics and correlation matrix for the entire analysis. However, none of

those methods are able to explain casual relationship between variables (i.e., movements

in a variable (dependent) by reference to movements in one or more other variables

(independent)). For instance, a correlation between two variables measures only the

degree of linear association between them (Brooks 2008).

In addition, the work of Tilahun (2010) was not mainly intended to investigate the

determinants of NPLs, the major aim of the study was to identify the underlying NPLs

management difficulty of privately owned commercial bank of Ethiopia without

including the two biggest state owned commercial banks that have higher market share in

the industry. Moreover, the study used only a descriptive statistics for the entire analysis

without considering a lot of limitations associated with it. Similarly, the work Ayalew

(2009) mainly emphasized on the legal problems in realizing NPLs of Ethiopian banking

sector. Thus, the cause of NPLs was not the concern of the study since legal issues

comes after the occurrence default loans.

40
Furthermore, the data that has been used in the analysis of all the above studies (i.e.,

studies that are conducted in Ethiopian commercial banking context) are mainly obtained

from banks officials‟ through questionnaire and interviews. Hence, intentionally or

unintentionally the respondent might be biased.

In general, the lack of sufficient research on the determinants of NPLs in the context of

Ethiopia banking sector and the existence of knowledge gap in the area initiate this study.

Hence, the purpose of this study is to investigate the determinants of NPLs in Ethiopian

commercial banking sector by utilizing an econometrics model so as to estimate both the

macroeconomic and bank specific determinants of NPLs.

41
Chapter three: Research design

The preceding chapter presented the review of litreature on the determinats of NPLs and

identified the existing knowledge gap. The purpose of this chapter is to discusses the

research design. The chapter is organized in to the following three sections. The first

section 3.1 presents variabe specification of the study with respect to the research

hypotheses. This followed by research approachs in section 3.2. Finally , Section 3.3

discusses the research method used in the study.

3.1. Variable specification, Hypotheses and research qustions

As already shown in the first chapter, the broad objective of this research is to investigate

the determinats of NPLs in the context of ECBs. In line with the broad objectiv ten

hypotheses and two specific research questions were formulated. Hence, subsection 3.1.1

presents the dependent variable (NPLs). Then the independent variables that are selected

and categorized into bank-specific and macroeconomic determinants of banks‟ NPLs are

presented in subsection 3.1.2.

3.1.1. Dependent variable

As mentioned in the literature review part of this study, there is no global standard to

define NPLs at the practical level. Variations exist in terms of the classification system,

the scope, and contents. Neversless, as far as this study intends to ivestigate the

determinats of NPLs in ECBs, the measurement of NPLs is in accordance with the NBE

directive. As per the NBE(2012)directive, NPLs are are clacified as Substandar,

42
Doubtfull and Loss. Hence, the ratio of NPLs to total loans was used as aproxy for this

study.

3.1.2. Independent variables

Pervious researches on the determinants of banks‟ NPLs have shown that, independent

variables that can explain the variation on NPLs are classified into bank-specific and

macroeconomic variables (Azeem et al. 2012, Delgado and Vallcorba 2007, Louzis et al.

2010 and Aktan 2009). The bank-specific variables are internal factors and controllable

for banks‟ managers while the macroeconomic variables are uncontrollable and hence

external. Therefore, the following subsections presented the bank-specific and

macroeconomic variables used in the econometrics model of this study.

3.1.1.1. Bank specific variables

The existing literature provides evidence that suggests a strong association between NPLs

and several bank specific variables. The bank specific variables that are usually theorized

as determinats of NPLs are include, loan growth, financial perfrormance, bank size,

ownership structure, the quality of the loan portfolio and operatinal efficicecy. Hence, the

following part of this subsection presents the bank-specific variables used in this study.

Loan growth: as mentioned in the literature, to maximize the short run benefits,

managers seek to rapidly expand credit activities. The search for rapid growth of loans is

achieved by either reducing interest rate charged to borrowers or by lending to lower

credit quality borrowers (Fernandez De Lis et al., 2000). This will lead, through adverse

selection reasoning (lending to lower credit quality borrowers) and ultimately increase

43
the probability of NPLs. Empirically; various Studies found strong positive relationship

between rapid credit growth and NPLs (Sinkey and Greenwalt 1991, Weinberg 1995,

Keeton 1999 Salas and Saurina (2002), Jimenez and Saurina (2006) and Metaxas et al

(2010)). Hence, a positive relationship between loan growth and NPLs is expected in this

study. The variable used to capture credit growth was constructed by finding the annual

percentage change in the loan portfolio for each Ethiopian commercial bank.

HP1: There is a significant positive relationship between loan growth of a bank and

bank’s NPLs.

Operational efficiency: as shown in the literature, the relationship between operational

efficiency of a bank and bank‟s NPLs can be positive or negative. A number of

researchers have found a positive (Berger and Humphrey 1992, Wheelock and Wilson

1994) and a negative (Hughes and Moon 1995, Resti1995) relationship between cost of

efficiency and NPLs. Hence, the sign of the coefficient estimate among operational

efficiency and NPLs is remain indeterminate in this study. Cost to income ratio of

Ethiopian commercial banks used as a proxy measurement to capture the operational

efficiency of banks.

HP2. There is a significant positive/negative relationship between operational efficiency

of a bank and bank’s NPLs

Financial performance: As noted by Hu et al. (2004), profitable banks are less engaged

in risky activities as they have less pressure to create revenues. At the opposite,

inefficient institutions might engage in risky lending in particularly when managers have

short term incentives. In this regard many scholars found a negative association between

44
financial performance of a bank and bank‟s NPLs (Jimenez and Saurina (2006), Jellouli

et al (2009), Metaxas et al (2010) and Vogiazas and Nikolaidou (2011)). Hence, a

negative relationship is expected in this study. In this study the financial performance of a

bank was measured by the ratio of Return on asset (ROA).

HP3: There is a significant negative relationship between financial performance of a

bank and bank’s NPLs

Income diversification: as mentioned in the literature review part of this study, there is

no general consensus on the benefit of income diversification. In this regard, vast

empirical literature found both a significant positive (Lepetit et al. (2007), Stiroh (2002,

2006) and negative (Winton 1999, Templeton and Severiens 1992) association between

income diversification and NPLs. Hence, the sign of the coefficient between income

diversification and NPLs is indeterminate in this study. The variable used to capture the

income diversification of banks was measured by the ratio of non interest income to total

asset of banks.

HP4: There is a significant positive/negative relationship between income diversification

of a bank and bank’s NPLs.

Ownership structure: most existing literature suggested that state-owned banks are

usually associated with high level of NPLs than privately owned banks. Empirically,

Salas and Saurina (2002), Hu et al. (2004), Micco et al. (2004), Barth et al. (2004) ,

Garcia and Robles (2007) and Swamy (2012) suggested a positive association between

state ownership of banks and the volume of NPLs . Hence, in this study a positive

relationship between NPLs and state owned banks is expected. The variable used to

45
capture the ownership structure of banks was measured by dummy variables (1=state

owned banks and 0 =private banks).

HP5: There is a significant positive relationship between state ownership of banks and

bank’s NPLs.

Bank size: The empirical evidence relating to the impact of bank size on NPLs

suggested an inverse relationship (Rajan and Dhal, 2003; Salas and Saurina, 2002; Hu et

al, 2006). According to these studies, the inverse relationship means that large banks have

better risk management strategies and technology which definitely allows them for

efficient information gathering, processing and analyzing which finish up with lower

levels of NPLs as compared to smaller banks. Hence, a negative relationship between the

size of a bank and bank‟s NPLs is expected in this study. The size of a bank was

measured by the Natural log of total assets of each bank.

HP6: There is a significant negative relationship between size of a bank and bank’s

NPLs.

3.1.1.2. Macroeconomic variables

Apart from bank specific variables, there is abundant empirical evidence that suggests

that several macroeconomic factors are important determinants of NPLs. Several

macroeconomic factors which the literature proposes as important determinants of NPLs

are: annual growth in GDP, the annual inflation rate, real effective exchange rate

(REER), annual unemployment rate, broad money supply (M2) and GDP per capital

(Salas and Suarina, 2002; Rajan & Dhal, 2003; Fofack, 2005; and Jimenez and Saurina,

46
2005). This study only considers the growth in real GDP, inflation, real interest rate and

effective exchange rate.

Real GDP growth: the empirical evidence suggested a negative relationship between the

growth in real GDP and NPLs (Salas and Suarina, 2002; Rajan & Dhal, 2003; Fofack,

2005; and Jimenez and Saurina, 2005). The explanation provided by the literature for this

relationship is that strong positive growth in real GDP usually translates into more

income which improves the debt servicing capacity of borrower which in turn contributes

to lower NPLs. Hence, a negative relationship between GDP and NPLs is expected in this

study. The variable used to capture real GDP growth was constructed by finding the

annual percentage change in the real GDP.

HP7: There is a significant negative relationship between real GDP growth and bank’s

NPLs.

Real interest rates: empirically, several studies report that high real interest rate is

positively related to NPLs (Sinkey and Greenwalt (1991), Fofack (2005) and Jimenez and

Saurina(2005)). The basic argument here is that, as interest rates rise, prudent borrowers

are more likely to decide that it would be unwise to borrow, whereas borrowers with the

riskiest investment projects are often those who are willing to pay the highest interest

rates. Hence, a positive relationship real interest rate and bank‟s NPLs is expected in this

study. In this study, the average lending rate of Ethiopian banks used as proxy

measurement.

HP8: There is a significant positive relationship between real interest rate and bank’s

NPLs.

47
Inflation: as mentioned in the literature, inflation affects borrowers‟ debt servicing

capacity through different channels and its impact on NPL can be positive or negative.

Empirically, Fofack (2005) found a positive relationship between inflation and NPLs in a

number of Sub-Saharan African countries with flexible exchange rate regimes. On the

other hand, Smadi (2010) found a negative association between inflation and NPLs in

Jordanian commercial banking sector. Hence, the relationship is indifferent in this study.

In this study, annual inflation rate was used as a proxy measurement.

HP9: There is a significant positive/negative relationship between inflation and bank’s

NPLs.

Real effective Exchange rate: like inflation a change in effective exchange rate can also

affects borrowers‟ debt servicing capacity through different channels and its impact on

NPL can be positive or negative. Hence, the sign of the relationship between exchange

rate and NPL is indeterminate in this study. The variable was measured by the annual

effective Exchange rate of Ethiopian birr in terms of dollar.

HP10: There is a significant positive/negative relationship between effective exchange

rate and bank’s NPLs.

In addition to the above ten hypothesis, the following two specific research questions

were also formulated.

RQ1. What are the determinants of banks’ non-performing loans in Ethiopia

commercial banking sector?

RQ2. What are the likely causes for the existence of variation on NPLs

performance among Ethiopian commercial banks?

48
The following table 3.1 presents the summary of hypothesized expected sign for the

relationship between variables (dependent and independent) along with the proxy

measurement for each variable.

Table 3.1 Descripton of the variables and their expected relationship

Variables Measure Notation Expected

sign

Nonperforming loans Non-performing loans/gross loans ratio NPL

Loan growth Percentage change in loan growth LG +

Operational efficiency Cost/income ratio CIR +/-

Income diversification Non-interest income/total assets ratio IDV +/-

Financial performance Return on asset ROA -

Size Natural log of total assets SIZE -

Ownership structure Dummy, “1” denote stated owned banks OWS +

and “0” for private banks

Exchange rate Annual effective Exchange rate of EFEX +/-

Ethiopian birr in terms of dollar

Real interest rate The average lending rate of banks RIR +

Economic growth Real GDP growth (in %) GDP -

Inflation The annual inflation rate INFL +/-

49
3.2. Research approaches

As noted in Creswell (2003, p.13) in terms of investigative study there are three common

approaches to business and social research namely, quantitative, qualitative and mixed

methods approach. Quantitative research is a means for testing objective theories

by examining the relationship among variables (Creswell 2009, p.4). In quantitative

research approach there are two strategies of inquiries namely, survey design and

experimental design. The chief advantage of this approach is making generalizations for a

broader population based on the findings from the sample. However, as noted in Dunn

(1999) quantitative research approach has its own disadvantages. For instance, the sample

selected may not represent the total population and the researchers know much about

the collective or average experience of research participants, but not their

individual experiences (Dunn 1999).

On the other hand, Qualitative research is an approach in which the investigator

often makes knowledge claims based primarily on the multiple meanings of individual

experiences, socially and historically constructed meanings, participation in issues

with an intent of developing a theory or pattern (Creswell 2003, p. 18). As noted in

Sarantakos (2005, p. 45 cited in Yesegat 2009, p. 73) qualitative research approach

uses strategies of inquiry such as narratives, ethnographies, grounded theory

studies, or case studies. Qualitative research design has advantages of flexibility and

emergent without being constrained by standardized procedures (Liamputtong and

Ezzy 2005, p. 204, cited in Yesegat 2009, p. 74). However, qualitative research

design has also its own weaknesses.

50
As noted in Dunn (1999) the demerits of this approach includes; absence of quick

response, difficulty, inefficiently, and lack of generalization among others. Finally,

mixed research is an approach to inquiry that combines or associates both qualitative and

quantitative forms (Creswell, 2009).

3.3. Research Methods adopted

Based on the above discussions of the three research approaches and by considering the

nature of the study a mixed methods research approach was used. The decisive argument

here is that the use of both quantitative and qualitative approaches in combination

provides a better understanding of research problems than either approach achieves alone.

As noted in Greene et al. (1989, p. 259 cited in Yesegat 2009, pp.75-76) adopting a

mixed methods approach has a number of benefits. The first benefit is triangulation

pertaining to a situation where researchers seek convergence, corroboration,

correspondence of results from quantitative and qualitative methods to increase validity

of constructs and inquiry results. Secondly, by mixing methods complementarily,

researchers seek elaboration, enhancement, illustration, clarification of the results from

one method with the results from the other method. Thirdly, by mixing methods with

developmental intent, researchers seek to use the results from one method to help develop

or inform the other method. Finally, to increase the scope of inquiry mixed method with

expansion intent seeks to extend the breadth and range of inquiry by using different

methods for different inquiry components.

Mixed method research involves both collecting and analyzing quantitative and

qualitative data either sequentially or concurrently. Accordingly, the quantitative method

51
was dominantly used to investigate determinants of NPLs of ECBs. Following to this, the

qualitative method was used to support the quantitative findings and to gain

additional insight into the factors that may affect NPLs of ECBs. Hence, the

following sections present consecutively the quantitative and qualitative aspects of the

research method.

3.3.1. Research method: Quantitative aspect

According to Leedy & Ormord (2005), among quantitative strategies of inquiry, a survey

research is the most common and economical way of conducting business research. As a

result, in order to generalize the findings to the whole banks operated in the

country, in the current study the researcher adopted a survey design that was

administered through structured review of documents.

3.3.1.1. Survey design: structured review of documents

As noted in Creswell (2003), the purpose of survey is to generalize description of trends,

attitudes, or opinions from a sample to a population so that inferences can be made about

some characteristic, attitude, or behavior of this population. Moreover, Fowler (1986)

noted that, it is also rational to use survey designs because of its benefits such as the

economy of the design and the rapid turnaround in data collection and identifying

attributes of a large population from a small group of individuals. Hence, by having the

above discussion, it is logical to apply survey method for this study. The subsequent

discussions present the structured review of documents in respect of sampling design and

data collection.

52
Sample design

The target population for this study was all commercial banks that were registered by

NBE and operational in the country. Currently, the country has two public-owned and

eighteen private commercial banks which are operating throughout the country (NBE

2012). However, as a result of lack of 12 years data (that was required for the analysis

purpose) in most of the newly established private banks the number of sampled banks

were reduced to eight. The researcher believes that the sample size is sufficient to make

sound conclusion about the population as far as it covers around 40% of the total

population. Moreover, the inclusion of CBE in the sample which takes the lions share in

the country‟s banking activity makes the sample more representative and reasonable.

Data collection

In order to achieve the stated objective, a panel data were collected through structured

document review. As noted in Brooks (2008) using a panel data has the following

advantages: First, and perhaps most importantly, researchers can address a broader range

of issues and tackle more complex problems with panel data than would be possible with

pure time-series or pure cross-sectional data alone. Second, by combining cross-sectional

and time series data, researchers can increase the number of degrees of freedom and the

power of the test by employing information on the dynamic behaviour of a large number

of entities at the same time. Finally, the additional variation introduced by combining the

data in this way can also help to mitigate problems of multicollinearity that may arise if

time series and cross sectional are modelled individually. Hence, audited financial

statements of eight banks (CBE, CBB, AIB, DB, WB, BoA, UB and NIB) for 12

53
consecutive years (i.e., from 2000-2011) were collected. The secondary data that were

collected through structured document reviews were mainly from the records held by

NBE and the banks themselves. Moreover, in order to analyze the relationship that exists

between NPLs and macroeconomic variables, macroeconomic data were also collected

for the same years. Those macroeconomic data were mainly gathered from the records

held by NBE and MoFED through structured document review

Data analysis techniques

So as to achieve the stated objective, the collected panel data was analyzed using

descriptive statistics, correlation matrix and multiple linear regression analysis. The

descriptive statistics (Mean values and standard deviations) was used to analyze the

general trends of the data from 2000 to 2011 based on the sample of 8 banks, and the

correlation matrix was also used to examine the linear relationship between the dependent

variable and independent variables. Finally, a multiple linear regression model was used

to determine the relative importance of each independent variable in explaining the

variation of NPLs in ECBs. Accordingly, a two step multiple linear regression equations

were run. In the first step (general) regression equation, all the proposed independent

variables (i.e., LG, CIR, ROA, IDV, SIZE, STATEDUMMY, INFL, GDP, EFEX and

RIR) were regressed with respect to the dependent variable (NPLs). To this end, only the

significant variables that were found from the first step regression equation were

regressed once again.

54
The multiple linear regressions model was conducted by the ordinary listing square

(OLS) method using EVIEWS 4 6 econometric software package. The rational for

choosing OLS is that, if the Classical Linear Regression Model (CLRM) assumptions 5

hold true, then the estimators determined by OLS will have a number of desirable

properties, and are known as Best Linear Unbiased Estimators(Brooks 2008). In addition,

as noted in Petra (2007) OLS outperforms the other estimators when the following

holds; the cross section is small and the time dimension is short. Therefore, as far

as both the above facts hold true in this study it is rational to use OLS. Thus, the

following section discussed the CLRM assumptions.

Test for Heteroscedasticity: as noted in Brooks(2008), the variance of the errors must

be constant (homoscedasticity). If the errors do not have a constant variance, they are said

to be heteroscedastic. If this problem ignored, the standard errors could be wrong and

hence any inferences made could be misleading. Hence, To test for the presence of

heteroscedasticity, the popular white test was employed in this study(Brooks 2008).

Test for Autocorrelation: This is an assumption that the errors are linearly independent

of one another (uncorrelated with one another). If the errors are correlated with one

another, it would be stated that they are auto correlated (Brooks 2008). To test for the

existence of autocorrelation, the popular Durbin-Watson test was employed. As noted in

Brooks (2008) the rejection / non-rejection rule would be given by selecting the

appropriate region from the following figure:

4
EViews is software package, providing the tools most frequently used in practical econometrics
5
The conditions for OLS optimality

55
Figure 3.1 Rejection and non-rejection regions for Durbin-Watson Test

Figure 3.1 shows as Durbin-Watson has 2 critical values: an upper critical value (dU) and

a lower critical value (dL).

Test for normality: As noted in Brooks (2008) a normal distribution is not skewed and

is defined to have a coefficient of kurtosis of 3. One of the most commonly applied tests

for normality; the Bera-Jarque formalizes these ideas by testing whether the coefficient of

skewness and the coefficient of excess kurtosis are zero and three respectively. Brooks

(2008) also stated that, if the residuals are normally distributed, the histogram should be

bell-shaped and the Bera-Jarque statistic would not be significant at 5% significant level.

Test for Multicollinearity: An implicit assumption that is made when using the OLS

estimation method is that the explanatory variables are not correlated with one another.

To test the presence of multicollinearity problem in regression model the study used a

correlation matrix of independent variables. The problem of multicollinearity usually

arises when certain explanatory variables are highly correlated. As noted by Hair et al.

(2006) correlation coefficient below 0.9 may not cause serious multicollinearity problem.

In contrary to this, Kennedy (2008) argued that as any correlation coefficient above 0.7

could cause a serious multicollinearity problem leading to inefficient estimation and less

reliable results.

56
Model estimation

So as to investigate the bank-specific and macroeconomic determinants of bank NPLs,

the following general multivariate regression equations were adopted:

Yi,t = α + βXi,t + εi,t………………………………………………….………….(1)

Where Yi,t is the NPLs ratio of bank i at time t, with i=1… N, t=1… T, α is a constant

term, Xi,t is the explanatory variables(bank specific and macroeconomic variables) of

bank i at time t and εi,t the disturbance term. As noted in Brooks (2008) the rational for

the inclusion of disturbance term are: first, even in the general case where there is more

than one explanatory variable, some determinants of Y i,t will always in practice be

omitted from the model. Second, there may be errors in the way that Yi,t is measured

which cannot be modelled. Finally, there are bound to be random outside influences on

Yi,t that again cannot be modelled. For example, a computer failure, human behaviour.

Hence, the disturbance term intends to mitigate the above problems.

Based on the general model provided above and on the base of selected variables the

empirical model used in the study was as follows:

NPLi,t= βo + β1 LGi,t + β2ROAit + β3CIRit + β4IDVit +β5OWSit + β6 SIZEit + β87GDPit

+ β8 INFLit + β9 EEXR i,t + β10 RIRi,t +Ɛ i,t ……………………………….. (2)

Where: LGi,t=loan growth of bank i at time t, CIRit = cost to income ratio of (operational

efficiency ) bank i at time t, ROAit = return on asset of bank i at time t, IDVit = Income

diversification for bank i at time t , SIZEit = The natural logarithm of total asset for bank i

at time t, OWSit = ownership structure of bank , GDPt = Real GDP growth at time t ,

57
INFLt = Inflation rate at time t, EEXR i,t = effective foreign exchange rate at time t ,

RIRi,t = real interest rate (average interest rate )of banks at time t, Ɛ it = the error term.

3.3.2. Research method: qualitative aspect

To supplement the gap that might not be captured by the quantitative survey ,

unstructured interviews with five senior Ethiopian bank officials (credit vice presidents

and senior credit committee members) were conducted. The interviewees were from both

private and state owned banks namely CBE, CBB, AIB, DB and WB who had over 8

years credit experience in addition to their several years of banking experience. The

interviews were conducted to know about both the internal and external factors

affecting the non-performing loans of banks in Ethiopia. The interview questions were

focused on the identification of factors that were not incorporated in the quantitative part

of this study. To this end, the result obtained from in-depth interviews were analyzed

using triangulation with the findings of the structured record reviews. As a result, the

response of the interviewees‟ for the interview questions were used for supporting the

result obtained from analysis of structured record reviews.

Finally, links between research question/hypotheses and variables on the one hand

and different data sources on the other hand are presented in table 3.2 below.

Table 3.2 Link between research question/hypotheses, variables and the

different data sources

Research question and hypotheses Variables Data sources

RQ1. What are the determinants of banks’ Factors affecting
NPLs in Ethiopia commercial Banks‟ Non- In-depth unstructured face-
banking sector? performing loans to-face interviews with

58
RQ2. What are the likely causes for the senior Ethiopian bank
existence of variation on NPLs officials
performance among banks?
HP1: There is a significant positive Dependent
relationship between loan growth of a Bank-specific data from
variable:
bank and bank’s NPLs. Income statement and
HP2: There is a significant positive/negative Non-performing Balance sheet held by
relationship between operational NBE and the banks and
loans
efficiency of a bank and bank’s NPLs. macroeconomic data from
Independent the records held by NBE
HP3: There is a significant negative
and MOFED
relationship between financial variables:
performance of a bank and bank’s
-Loan growth
NPLs
HP4: There is a significant positive/negative
relationship between Income
-Operational
diversification of a bank and bank’s
NPLs. efficiency
HP5: There is a significant positive -Income
relationship between state ownership
of a bank and bank’s NPLs. diversification
HP6: There is a significant negative -Financial
relationship between size of banks
and bank’s NPLs. performance
HP7: There is a significant negative -Size
relationship between real GDP
growth and bank’s NPLs. -Ownership
HP8: There is a significant positive structure
relationship between real interest
rate and bank’s NPLs. -Exchange rate
HP9: There is a significant positive/negative -Real interest rate
relationship between inflation and
bank’s NPLs. -Economic
HP10: There is a significant growth
positive/negative relationship
between exchange rate and bank’s -Inflation
NPLs.

59
Chapter four: Results and Analysis

The preceding chapter presented the research design along with the justification for

adopting a mixed methods research approach in this particular study. The purpose of this

chapter is to present results and analysis of data obtained from different methods

involved in this study. Accordingly, the remaining part of the chapter is organized into

three sections. The first section 4.1 presents research hypotheses and questions as

presented in the previous chapter. This is followed by the results of both documentary

analyses (structured review of documents) and in-depth interview in section 4.2. Finally,

Section 4.3 presents detail analysis of results.

4.1. Research Hypotheses and Questions

As mentioned in chapter one the broad objective of this study is to investigate the

determinants of NPLs in ECBs. So as to achieve this broad objective the study developed

the following ten hypotheses and two specific research questions:

HP1: There is a significant positive relationship between loan growth of a bank and

bank’s NPLs.

HP2: There is a significant positive/negative relationship between operational efficiency

of a bank and bank’s NPLs.

HP3: There is a significant negative relationship between financial performance of a

bank and bank’s NPLs

60
HP4: There is a significant positive/negative relationship between Income diversification

of a bank and bank’s NPLs.

HP5: There is a significant positive relationship between state ownership of a bank and

bank’s NPLs.

HP6: There is a significant negative relationship between size of banks and bank’s NPLs.

HP7: There is a significant negative relationship between real GDP growth and bank’s

NPLs.

HP8: There is a significant positive relationship between real interest rate and bank’s

NPLs.

HP9: There is a significant positive/negative relationship between inflation and bank’s

NPLs.

HP10: There is a significant positive/negative relationship between exchange rate and

bank’s NPLs.

In addition to the above ten hypotheses the following two specific research questions

were also formulated:

RQ1. What are the determinants of banks’ NPLs in Ethiopia commercial

banking sector?

RQ2. What are the likely causes for the existence of variation on NPLs

performance among banks?

61
4.2. Results

The purpose of this section is to present the results of data obtained from different data

sources. Accordingly, the results of the documentary analysis (structured reviews of

documents) and in depth interviews are presented in the following subsections.

4.2.1. Structured review of documents

As mentioned earlier, the major purpose of this study is to investigate the determinants of

NPLs in ECBs. To achieve this objective, quantitative research approach along with

survey design (structured review of documents) was dominantly used. To this end, the

necessary data gathered from the documents held by NBE, MoFED and the banks

themselves were analyzed using descriptive statistics, correlation matrix and multiple

linear regression analysis. Accordingly, the following discussion presents the results

of the documentary analysis as follows. Section 4.2.1.1 presents results of the

descriptive statistics followed by tests for the classical linear regression model

assumptions in section 4.2.1.2. Section 4.2.1.3 presents the correlation analysis among

the dependent and independent variables. Finally, the outcomes of the panel data

regression analysis are presented in section 4.2.1.4.

4.2.1.1. Descriptive statistics

The summery of descriptive statistics that was intended to give general descriptions about

the data (both dependant and independent variables) is presented in Table4.1. The total

number of observation for each variable was 96 (i.e., data for 8 banks for the period from

the year 2000 to 2012). Accordingly, mean, median, standard deviation, minimum and

62
maximum values of each variable were used so as to show the overall trend of the data

over the period under consideration.

Table 4.1: Summary of descriptive statistics for dependent and independent variables

Variables Observations Mean Median Max Min Std. Dev.

NPLs 96 0.133 0.087 0.535 0.0086 0.118
CIR 96 0.673 0.523 3 0.190 0.433
LG 96 0.180 0.175 0.719 -0.140 0.153
IDV 96 0.030 0.030 0.062 0.009 0.010
SIZE 96 21.797 21.678 25.461 18.778 1.364
ROA 96 0.032 0.036 0.056 0.003 0.013
INFL 96 0.109 0.089 0.364 -0.106 0.117
EFEX 96 0.101 0.088 0.165 0.083 0.024
GDP 96 0.086 0.108 0.126 -0.021 0.044
RIR 96 0.113 0.111 0.127 0.105 0.008
Note: Non-performing loans (NPLs), Cost to income ratio (CIR), Loan growth (LG), Income diversification
(IDV), Size (SIZE), Return on asset (ROA), Inflation (INFL), Effective exchange rate (EFEX), Growth
domestic product (GDP) and Real interest rate (RIR)
Source: Financial statements of banks, MoFED reports and own computation

As can be seen from table 4.1, for the total sample, the mean of NPLs was 13.3% with a

minimum of 0.86 % and a maximum of 53.5%. This indicates that, from the total loans

that ECBs disbursed, an average of 13.3% were being default or uncollected over the

sample period. The lowest NPLs ratio that ECBs experienced over the sample period

was 0.86 %. On the other extreme, the highest NPLs ratio of ECBs was 53.5% which was

in excess of the average 30% NPLs recorded in sub-Saharan African countries during the

1990‟s financial crisis (Fofack 2005). The disparity between the minimum 0.86 % and

the maximum 53.5% of NPLs indicate the margin that NPLs ratio of ECBs ranged over

the sample period. The standard deviation (0.118) of NPLs also shows the existence of

high variation among ECBs in terms their loan recovering capacity as compared to other

63
variables like ROA, IDV and RIR. As shown in figure 4.1, the NPLs ratio of Ethiopian

private banks was relatively lower as compared to state owned banks and shown small

variation among banks except some deviant observation in Awash International Bank

(AIB) and Bank of Abyssinia (BoA). AIB and BoA were the only private banks that

experienced NPLs ratio over 30% of their total loans. On the other hand, NPLs ratio of

state owned banks shows high disparity and tends to be higher as compared to private

banks. As the raw data indicates, both the maximum 53.5% and the minimum 0.86 % of

NPLs were recorded in Commercial bank of Ethiopia (CBE). This clearly indicates the

existence of variation on NPLs of ECBs in terms of their ownership structure.

Figure 4.1: Nonperforming loans of ECBs

NPL
.6

.5

.4

.3

.2

.1

.0
CBB - 00
CBB - 04
CBB - 08
CBE - 00
CBE - 04
CBE - 08
BOA - 00
BOA - 04
BOA - 08

DB - 00
DB - 04
DB - 08

UB - 00
UB - 04
UB - 08
AIB - 00
AIB - 04
AIB - 08

NIB - 00
NIB - 04
NIB - 08

WB - 00
WB - 04
WB - 08

Source: Financial statements of banks

Regarding the independent variables of the model there are some interesting statistics that

have to be mentioned. For instance, the size of banks which was measured by natural log

of total asset revealed the highest standard deviation (1.364), which means, it was the

64
most deviated variable from its mean compared to other variables. This indicates the

existence of high variation among ECBs in terms of their size. Another interesting

observation was the cost-to-income ratio of banks which indicated by the range

between 300% and 19.1%. The standard deviation of 0.433 also indicates relatively

high operational efficiency disparity among ECBs. In addition, the average loan growth

of ECBs was 18%, with a minimum of -14% and a maximum of 71.9%. A negative sign

of loan growth indicates the existence of different conditions that decreased the loans

disbursement practice of Ethiopian banks over the sample period. The standard deviation

of 0.153 also shows high disparity among ECBs in terms of their loan disbursement

practice next to size and operational efficiency. On the other hand, the mean of ROA was

3.2% with a minimum of 0.3% and a maximum of 5.7%. That means during the period

under consideration sampled ECBs earned an average of 3.2 cents of profit before tax for

a single birr invested in their assets. The standard deviation for ROA was 0.013 which

indicates that the profitability variation between the selected banks was very small.

Among bank-specific variables, the smallest standard deviation was reported in income

diversification of banks which was 0.010. This indicates the existence of less variation

among ECBs in terms of diversifying their source of revenue.

Among macroeconomic variables employed in this study inflation had a higher standard

deviation which was 0.117. This implies that inflation rate in Ethiopia during the study

period remains somewhat unstable. On the other hand, the average lending rate of ECBs

was 11.3%; the standard deviation (0.008) was the lowest of all the variables used in this

study. This indicates that the average lending rate of ECBs was highly stable over the

sample period. In addition, the average real GDP growth in Ethiopia for the last twelve

65
years was 0.086 (8.6%), with a standard deviation of 0.045 implies the economic growth

in Ethiopia during the sample period remains stable as compared to the inflation rate.

Moreover, the Standard deviation of exchange rate (0.024) indicates the existence of less

volatility of dollar in terms of Ethiopian birr over the period under consideration. Thus, it

can be concluded that, the macroeconomic variables were relatively stable over the

sample periods as compared to bank specific variables with the exception of some

instability on inflation rate.

4.2.1.2. Test results for the classical linear regression model assumptions

As mentioned in the methodology part of this study, as far as the assumptions of classical

linear regression model hold true, the coefficient estimators of both α (constant term) and

β (independent variables) that are determined by OLS will have a number of desirable

properties, and usually known as Best Linear Unbiased Estimators (BLUE). Hence, the

following sections discuss results of the diagnostic tests (i.e., heteroscedasticity,

autocorrelation, multicollinearity and normality) that ensure whether the data fits the

basic assumptions of classical linear regression model or not.

Test for Heteroscedasticity

To test for the presence of heteroscedasticity, the popular white test was employed

(Brooks 2008). As shown in table 4.2, all versions of the white test statistic( F-statistic ,

Chi-Square and Scaled explained SS) gave the same conclusion that there was no

evidence for the presence of heteroscedasticity in this particular study since the p-

values for all versions of the test statistic were in excess of 0.05. Accordingly, the null

66
hypothes that the variance of the errors is constant (homoscedasticity) should not be

rejected.

Table 4.2 Heteroscedasticity Test: White

F-statistic 1.265153 Prob. F(61,34) 0.2312

Obs*R-squared 66.64071 Prob. Chi-Square(61) 0.2892
Scaled explained SS 56.20589 Prob. Chi-Square(61) 0.6499

Source: Financial statements of banks, MoFED reports and own computation

Test for Autocorrelation

As mentioned in the previous chapter to empirically analyze the determinants of NPLs

in ECBs, 96 observations and ten regressors along with an intercept term were used

in the model. Accordingly, the relevant critical values for 96 observations and 10

regressors in Durbin-Watson test statistic table have shown an upper critical value (dU)

of 1.765 and a lower critical value (dL) of 1.335 which is an intermediate region where

the null hypothesis of no autocorrelation can neither be rejected nor not rejected. Thus, as

shown in table 4.3, the Durbin-Watson test statistic of this study (1.40) was clearly

between the upper limit (1.765) and the lower limit (1.335) and thus the null hypothesis

of no autocorrelation neither rejected nor not rejected.

Table 4.3 Autocorrelation Test: Durbin Watson

Variables DW test static result

All bank-specific & macro-economic 1.40
Source: Financial statements of banks, MoFED reports and own computation

67
To mitigate the above problem, the Breusch-Godfrey test statistic was used. As shown in

table 4.4, all versions of the Breusch-Godfrey test statistic (F-statistic and Chi-Square)

show the existence of autocorrelation problem in this particular study. However, the

problem was not strong since the p-values (especially F-statistic) were very much closer

to the highest significant level in this test (5%). Hence, the autocorrelation problem of

this study can be tolerable. In this regard Brooks (2008) noted that, the coefficient

estimates derived using OLS in the existence of autocorrelation problem may not be Best

Linear Unbiased Estimators (BLUE), but they are still unbiased.

Table 4.4 Breusch-Godfrey Serial Correlation LM Test:

F-statistic 3.121483 Prob. F(2,83) 0.0493

Obs*R-squared 6.715652 Prob. Chi-Square(2) 0.0348

Source: Financial statements of banks, MoFED reports and own computation

Test for Multicollinearity

According to Kennedy (2008) multicollinearity problem exists when the correlation

coefficient among the variables are greater than 0.70. As shown in table 4.5, in this study

there is no correlation coefficient that exceeds or even closer to 0.70. Moreover, as

noted in Brooks (2008) if the independent variables are not correlated with one another,

adding or removing a variable from a regression equation would not cause a change on

the values of the significant level and the coefficients estimates of other variables. In this

study a two step multiple regression analysis were made (i.e., in the first step all the

proposed variables were regressed and then, only the significant variables that were

found in the first regression analysis were regressed once again). Hence, as shown in

68
table 4.8 dropping the insignificant variables (i.e., RIR and IDV) from the regression

equation of this study did not cause any change on the values of other variables (both in

terms of the significance level and the coefficient estimates). Therefore, it can be

concluded that there is no multicollinearity problem in this particular study.

Table 4.5 Correlation matrixes of independent variables

CIR LG ID SIZE ROA GDP INFL EFEX RIR

CIR 1
LG -0.132 1
ID -0.302 0.167 1
SIZE -0.558 -0.373 0.083 1
ROA -0.624 0.200 0.666 0.299 1
GDP -0.369 0.011 0.365 0.364 0.536 1
INFL -0.326 -0.156 0.260 0.391 0.367 0.383 1
EFEX -0.255 -0.226 0.506 0.491 0.454 0.303 0.331 1
RIR 0.092 -0.095 0.175 0.076 0.162 0.074 0.157 0.465 1

Source: Financial statements of banks, MoFED reports and own computation

Test for normality

In this study, the normality of the data was checked with the popular Bera-Jarque test

statistic (Brooks 2008). According to Bera-Jarque test statistic, normally distributed data

is not skewed and has a coefficient kurtosis of 3. As shown in figure 4.2, the coefficient

kurtosis(3.15) of the data in this particular study was much closer to 3, and the Bera-

Jarque statistic had a P-value of 0.853 implying that there was no evidence for the

presence of abnormality in the data. Thus, the null hypothesis that the data is normally

distributed should not be rejected since the p-value was considerably in excess of 0.05.

69
Figure 4.2 Normality test for residuals: Bera-Jarque

16
Series: Residuals
14 Sample 1 96
Observations 96
12
Mean -1.05e-16
10 Median -0.001798
Maximum 0.129683
8 Minimum -0.097811
Std. Dev. 0.044426
6
Skewness 0.118542
Kurtosis 3.151676
4

Jarque-Bera 0.316859
2
Probability 0.853483
0
-0.10 -0.05 -0.00 0.05 0.10

Source: Financial statements of banks, MoFED reports and own computation

4.2.1.3. Correlation analysis

As mentioned in the third chapter, the purpose of correlation matrix in this particular

study was to show the linear association between the dependent and independent

variables. Accordingly, the ownership structure of Ethiopian banks (measured by a

dummy variable i.e., 1=state owned banks and 0= private banks) was excluded from the

correlation matrix since a dummy variable is a qualitative data that do not have any linear

association with any of other variables. As can be seen in table 4.6, return on asset

(ROA), income diversification(IDV) and loan growth(LG) of a bank were the most

negatively correlated bank-specific variables with the movement of bank‟s NPLs with a

correlation coefficient of -0.593, -0.480 and -0.473 respectively.

70
Those correlation results clearly indicate the existence of strong inverse linear association

among the above mentioned variables and NPLs of Ethiopian banks. In other words, as

bank‟s ROA, LG and IDV increases, NPLs of Ethiopian banks moves to the opposite

direction. In addition, size of a bank also negatively correlated with the movement of

NPLs. However, the magnitude of the correlation coefficient (-0.02) was very small as

compared to the above variables. To the contrary, the movement of cost to income ratio

was positively correlated with NPLs. On the other hand, all the macro economic

variables, real GDP growth, inflation, effective exchange rate and real interest rate show

a negative linear association with NPLs. This implies as the above macroeconomic

variables increase, NPLs of Ethiopian commercial banks moves towards the opposite

direction. The magnitude of the correlation coefficient for real GDP growth (-0.45) and

effective exchange rate (-0.40) had shown a strong inverse linear association with the

movement of NPLs as compared to other macro economic variables.

Table 4.6 Correlation matrix of dependent and independent variables

NPL LG CIR ID SIZE ROA GDP INFL EFEX RIR

NPL 1
LG -0.51 1
CIR 0.22 -0.13 1
ID -0.52 0.16 -0.30 1
SIZE -0.16 -0.37 -0.55 0.08 1
ROA -0.64 0.09 -0.62 0.66 0.33 1
GDP -0.45 0.01 -0.36 0.36 0.36 0.53 1
INFL -0.39 -0.16 -0.32 0.26 0.39 0.36 0.38 1
EFEX -0.41 -0.24 -0.25 0.50 0.49 0.45 0.30 0.33 1
RIR -0.23 -0.05 0.09 0.17 0.07 0.16 0.07 0.15 0.46 1

71
Source: Financial statements of banks, MoFED reports and own computation

It should be noted that all the above correlation coefficient results merely show the linear

association among the dependent (NPLs) and independent variables. In other words, the

relationships are not casual (i.e., the change in one variable is not resulted from the

movement of the other variable).

4.1.1.1. Regression results

As mentioned in the previous chapter, in this study a two step multiple linear regression

equations were run. In the first step (general) regression equation, all the proposed

independent variables (i.e., LG, CIR, ROA, IDV, SIZE, STATEDUMMY, INFL, GDP,

EFEX and RIR) were regressed with respect to the dependent variable (NPLs). To this

end, only the significant variables that were found from the first step regression equation

were regressed once again. Table 4.7 shows the first step regression results. The R2 and

the adjusted- R2 statistics of the model were 83.72% and 81.81% respectively. These

results are intended to show how well does the model containing the explanatory

variables that can explain variations in the dependent variable and usually known as

goodness of fit statistics (Brooks 2008).

Thus, the adjusted- R2 of this study indicates that, 81.81% of the variation on the

dependent variable (NPLs of ECBs) was explained by the changes in the independent

variables. In other words, the change in annual inflation rate, real interest rate, effective

exchange rate, real GDP growth, loan growth, income diversification, size, performance,

ownership structure, and operational efficiency collectively explain 81.81% of the

72
variation in NPLs ratio of ECBs. In contrary, the remaining 18.19 % of changes on the

NPLs of ECBs were explained by other factors which were not included in the

econometrics model of this study. Thus it can be concluded that, all the independent

variables used in this study collectively, were good explanatory variables of NPLs in

ECBs. Thus, the null hypothesis of F-statistic (the overall test of significance) that the R2

is equal to zero was rejected at 1% significance level (p-value =0.0), which enhanced the

reliability and validity of the model.

Table 4.7 First Step (General) Regression Results

Variable Coefficient Std. Error t-Statistic Prob.

C 0.255358 0.124459 2.051750 0.0433

CIR -0.027556 0.015833 -1.740337 0.0854
LG -0.293450 0.036765 -7.981801 0.0000
IDV 0.646428 0.680856 0.949434 0.3451
SIZE 0.009856 0.004643 2.122890 0.0367
ROA -2.829388 0.686064 -4.124087 0.0001
STATEDUMMY 0.076057 0.013080 5.814793 0.0000
INFL -0.235798 0.047757 -4.937420 0.0000
GDP -0.451706 0.135930 -3.323086 0.0013
EFEX -1.531906 0.297262 -5.153390 0.0000
RIR 0.063976 0.690610 0.092636 0.9264

R-squared 0.837277 Durbin-Watson stat 1.4011

Adjusted R-squared 0.818133
F-statistic 43.73601
Prob(F-statistic) 0.000000

Source: Financial statements of banks, MoFED reports and own computation

73
As shown in table 4.7, the coefficient estimate of real GDP growth, effective exchange

rate, inflation rate, loan growth and financial performance of a bank (ROA) were

negative and statistically significant at 1% significance level. The coefficient estimates of

the aforementioned five independent variables were -0.451, -1.531, -0.235, -0.293 & -

2.829 respectively. The negative sign of the coefficient estimate with 1% significant level

indicate the existence of strong inverse relationship between NPLs and the above

mentioned independent variables. Thus, it can be concluded that, an increase on those

variables lead to a decrease in NPLs of Ethiopian commercial banks. In addition, the

coefficient estimate of operational efficiency was negative and statistically significant at

10% significance level. However, the magnitude of the coefficient estimate (-0.027) was

very low as compared to the above variables. On the other hand, the coefficient estimate

of ownership structure (STATEDUMMY) was positive and statistically significant at 1%

significant level. This clearly indicates that, state owned banks tend to have high level of

NPLs as compared to privately owned banks. In addition, size of a bank had also a

positive and statistically significant (at 5% significance level) association with NPLs.

However, the magnitude of the coefficient estimate (0.009) was the lowest of all the

variables used in the model. Hence, the explanatory power of bank size on the variation

of NPLs in ECBs was very small as compared to other variables. Furthermore, income

diversification (IDV) and average lending rate (RIR) had also positive coefficient

estimate of 0.646 and 0.063 respectively. However, the results of t - statistics for both

IDV and RIR were insignificant since the p-values were considerably in excess of 10%.

Hence, based on the above results it can be conclude that, both bank-specific (loan

growth, financial performance, ownership structure, operational efficiency and size of a

74
bank) and macroeconomic (real GDP growth, inflation and effective exchange rate)

variables were the determinants of NPLs in Ethiopian commercial banks.

As mentioned earlier, only the significant variables(LG, CIR, ROA, SIZE,

STATEDUMMY, INFL, GDP and EFEX ) that were found in the first step regression

analysis were regressed once again in order to ensure the reliability and the consistency

of the first step regression results (both in terms of the coefficient estimates and the level

of significance). Table 4.8 shows the second step multiple regression results in which the

insignificant variables (IDV and RIR) were dropout. Comparing the results of the two

regression analysis, major differences were not found. As shown in table 4.8, the R2

(83.55%) and the adjusted- R2 (82.04%) statistics in the second step regression were

much closer to the R2 (83.72%) and the adjusted- R2 (81.81%) results obtained in the first

step regression. Similarly, the results of Durbin-Watson statistics in both the first and

second step regression were almost equal. In addition, significant variables that were

found in the first step regression were remained significant (with the same significance

level) in the second step regression. Moreover, the sign and the magnitude of coefficient

estimates in both the first and second step regression were almost similar.

Table 4.8 Second Step Regression Results

Variable Coefficient Std. Error t-Statistic Prob.

C 0.275838 0.098027 2.813900 0.0061

LG -0.293679 0.036521 -8.041285 0.0000
ROA -2.523769 0.598648 -4.215785 0.0001
CIR -0.025478 0.015142 -1.682544 0.0960
SIZE 0.009189 0.004528 2.029289 0.0455
STATEDUMMY 0.073909 0.012794 5.776680 0.0000

75
 GDP -0.445923 0.134830 -3.307285 0.0014
INFL -0.234712 0.047214 -4.971184 0.0000
EFEX -1.434699 0.248540 -5.772513 0.0000

R-squared 0.835550 Durbin-Watson stat 1.365317

Adjusted R-squared 0.820429
F-statistic 55.25469
Prob(F-statistic) 0.000000

Source: Financial statements of banks, MoFED reports and own computation

Based on the above discussions, it can be concluded that the results obtained from the

first (general) regression analysis were consistence with the result of the second

regression analysis, which enhanced the reliability and validity of the data used in the

model.

4.2.2. In-depth interview results

As mentioned in the previous chapter the purpose of qualitative research approach in this

particular study was to supplement the data obtained from structured review of

documents and to investigate other factors that could not obtained from documents.

Accordingly, unstructured interview with five senior Ethiopian bank officials (credit vice

presidents and senior credit committee members) was conducted. The interviewees were

from both private and state owned banks namely CBE, CBB, AIB, DB and WB who had

over 8 years credit experience in addition to their several years of banking experience.

All the interviewees were interviewed independently at different times. To this end, the

interview questions were focused on the identification of factors affecting the NPLs

of ECBs in general by giving due attention on factors that were not incorporated in the

76
quantitative part of this study. In addition, the interview questions tried to identify

how those factors can influence NPLs of ECBs, the major determining factors among

the influential factors and their general opinion regarding the matter. As per the interview

results, the general reasons which lead to the emergence of NPLs in ECBs can be

grouped into the following three major categories:

The first category includes the internal factors associated with the specific policy choices

of a particular bank and usually termed as bank-specific determinants. Those factors

include variables that can be influenced by managerial decisions of a bank such as loan

growth, size, financial performance and operational efficiency of bank. As per the

interviews, loan growth, performance and operational efficiency of banks have a negative

association with NPLs of ECBs. On the other hand, state ownership of a bank has

positive relationships with the volume of NPLs. Besides, inadequate credit risk

management from identifying, measuring and monitoring of credit were also considered

as the major internal factors that can affect NPLs of ECBs. Moreover, collateral based

lending system, absence of adequate man power, lack of comprehensive studies on the

credit applicants, lack of follow-up on the borrower's activities (failure to follow up

the collateral provided by the borrowers) were also considered as the major internal

determinants of NPLs in ECBs.

The second category that was considered as the most important reason to the

emergence of NPLs in ECBs was factors related to the client (borrower). As per the

interviews, those factors include, providing false information , using the loan for other

purposes that are undesirable from the banks' point of view (fund diversion), bad

intentions of the client with respect to non-payment of the loan at the maturity date

77
(willful default). In addition, operational losses of borrower were also cause a loan

default to Ethiopian banks especially in the event of absence of adequate collateral.

The third category was factors related to the external conditions. As per the interviews,

those factors include variables such as macro economic variables (real GDP growth,

inflation and effective exchange rate), the international conditions, government regulation

and the surrounding natural environment. All the above external factors can adversely or

positively affect the loan quality of ECBs. Under the macroeconomic conditions

inflation, effective exchange rate and real GDP growth had a negative association with

the volume of NPLs in ECBs. However, the importance of average lending rate in

determining the variation of NPLs in ECBs had not supported by the interviewees. On the

other hand, the absence of credit rating agency in the country, the national bank

regulation, the surrounding natural environment and international competition in the

global market can also affect the NPLs of Ethiopian banks in different aspects.

4.3. Discussions of Results

The preceding sections presented the result of the documentary analysis and in-depth

interviews. The purpose of this section is to discuss the results obtained from different

data sources. The hypotheses and specific research questions presented in section 4.1 deal

with the casual relationship between NPLs and such factors as loan growth, financial

performance, operational efficiency (cost to income ratio), income diversification,

ownership structure, bank size, growth domestic product, inflation, effective exchange

rate and real interest rate. The subsequent discussions hence try to present the analysis of

results in respect of those factors in an orderly manner.

78
Loan growth

As mentioned in the literature review part of this study, a rapid loan growth in the

banking sector may aggravate the problem of asymmetric information (adverse selection)

and ultimately increases the probability of NPLs. In this regard, most of the empirical

studies suggested a positive relationship between loan growth of a bank and bank‟s

NPLs. Despite this fact, the coefficients estimates of loan growth in this particular study

revealed a negative association with NPLs reported by ECBs. As can be seen from table

4.7, a variable used to capture loan growth of a bank (LG) had a coefficient estimate of -

0.2934 and statistically significant at 1% significance level (P-value=0.0). This indicates

the existence of strong inverse relationship between loan growth of a bank and bank‟s

NPLs in ECBs. That means an increase in loan disbursement practice of Ethiopian

banks, lead to a decrease in their volume of NPLs. In this general setting, it can be

concluded that, loan growth of a bank was key determinate of NPLs in ECBs. The

finding was consistent with previous studies of Pasha and Khemraj (2009), Jellouli et al

(2009), and Vogiazas and Nikolaidou (2011).

Similarly, the result obtained from interview also clearly supports the regression output.

As per the interview conducted with credit vice president and senior credit committee

members of selected ECBs, loan growth was one of the major factors that can affect

Ethiopian banks NPLs negatively. According to the interviewees, the existing loan

disbursements practice of ECBs was not aggressive (rapid) as compare to the demand of

loans in the market. This is due to the fact that, the current ever increasing economic

growth in Ethiopia results a high demand of loans in many sectors of the economy.

Consequently, the existing unparallel high demand of loans in the market (as compared to

79
the limited supply of loans) provides an opportunity to ECBs so as to choose loan

applicant who has strong financial performance and lower risk.

In addition, the traditional lending system of Ethiopian banks in which every branch

managers approved a loan had changed and currently loan applications can only be

approved at district or at head office level. Hence, the existing centralized loan

application evaluation system enables banks to use their management expertise and

reduce risk associated with lending. Furthermore, ECBs began to establish a close

relationship with the borrowers through customer relation managers (CRMs) whose

major function is reducing information asymmetries between the borrowers and their

banks. Each CRMs of Ethiopian banks are responsible for the activity of borrowers

under their supervision. Hence, the segregated duty provided to each CRMs increase

their influence on borrowers management through frequent dealings over time and

enable them to get important non public information. Such information advantages allow

banks to improve decision making if borrowers face distress. Moreover, a close

relationship between a lender and a borrower enables a bank to be better in offering

future loan and reduce risk and uncertainty associated with lending.

Financial Performance

The coefficient estimate of banks financial performance (measured by the ratio of ROA)

revealed negative and statistically significant association with NPLs. The magnitude of

the coefficient estimate (-2.83) for ROA was the largest of all the variables used in the

model. This indicates that, ROA had a great impact in explaining the variation of NPLs

in ECBs. Moreover, the coefficient estimate was statistically significant at 1% significant

80
level (p-value of 0.0). This implies that, an increase in the ratio of bank‟s ROA, leads to a

decrease in NPLs of ECBs. The finding was in consistent with prior expectation and

theory that indicated profitable banks are less engaged in risky activities as they have less

pressure to create revenues and ultimately resulted with lower volume of NPLs (Kwan

and Eisenbeis (1995), Berger and DeYoung (1997), Barth et al., (2002)). In this general

setting, it can be concluded that, as the financial performance (profitability) of Ethiopian

banks increase, the likelihood that banks engaged in risky activities would reduced and

ultimately the probability that loans became NPLs will reduce with the same manner.

Therefore, the findings suggested that, financial performance (profitability) of banks was

a vital determinant of NPLs in ECBs. This finding was also consistent with the previous

study of Hu et al. (2004), Godlewski (2004), Jellouli et al (2009) and Metaxas et al

(2010). Similarly, the result obtained from interview supports the regression output. This

is due to the fact that, the financial performance of Ethiopian banks determines the risk

taking behaviour of managers.

As per the interviewees, Ethiopian commercial banks are profitable enough and hence

managers have less pressure to create revenues. This condition provides Ethiopian

commercial banks an opportunity to compromise risks so as to choose loan applicant who

has lower risk. In this regard, all Ethiopian commercial banks are reluctant to provide

loan to the transport sector (which can generate high interest income) due to high risk

associated with the sector. Moreover, this result was also consistent with the existing

reality in the Ethiopian banking industry where the NPLs shows a parallel decrease as the

financial performance (profitability) of banks grow up.

81
Operational efficiency

As mentioned in the literature the impact of operational efficiency on NPLs of banks can

be positive or negative. According to Berger and DeYoung (1997), the positive

association among low cost efficiency and NPLs is hypothesized as „bad management‟

hypothesis. The basic argument here is that, low cost efficiency considers as signal of

poor managerial performance (bad management), poor skills in credit scoring, monitoring

and controlling the borrowers after loans are issued. On the other hand, low cost

efficiency may have a negative impact on NPLs. According to Berger and DeYoung

(1997) this relationship is hypothesized as „Skimping‟ hypothesis. According to this

view, there exists a trade-off between allocating resources for underwriting and

monitoring loans and measured cost efficiency. In other words, banks which devote less

effort to ensure higher loan quality will seem to be more cost-efficient; however, there

will be a growing number of NPLs in the long run.

Despite the above fact, the coefficient estimate of operational efficiency (measured by

cost to income ratio of banks) in this particular study revealed a negative relationship

with the volume of NPLs reported by ECBs. The inverse relationship implies that, as the

volume of cost to income ratio of Ethiopian banks increases, their size of NPLs will

decline. However, the magnitude of the coefficient estimate of -0.0275 was not strong as

compared to other variables like ROA and IDV implying the operational efficiency of

Ethiopian banks had a little impact in explaining the variation of NPLs in the industry. In

addition, coefficient estimate was statistically significant at 10% significant level (p-

value=0.0854) which was the highest significant level. Hence, the findings suggested

that, operational efficiency of banks was one of the modest determinants of NPLs in

82
ECBs (both in terms of significance level and the magnitude of the coefficient estimate).

This was in favor of „skimping‟ hypothesis that banks which have adequate budget to

screening loan customers, appraising collateral, and monitoring and controlling

borrowers after loans disbursement resulted with lower volume of NPLs as compared to

banks which seem to be cost efficient but, do not have adequate budget to ensure higher

loan quality. Similarly, the result gathered from interview clearly supports the regression

output. As per the interview results, cost of efficiency was considered as a major

determinant that affects NPLs of Ethiopian banks negatively. The inverse relationship

among cost of efficiency and NPLs of ECBs was justified by the ever increasing cost

incurred by Ethiopian banks so as to achieve improved credit risk management through

adequate loans screening, monitoring and controlling of borrowers. Hence, it can be

conclude that, an increase in operating cost of Ethiopian commercial banks can enhance

the loan quality of banks and ultimately reduced the probability of NPLs.

Income diversification

Despite the absence of general consensus on the impact income diversification on bank‟s

NPLs, in this study the ratio of non-interest income to total assets which was used to

measure the level of income diversification of ECBs revealed a positive coefficient of

estimate. This indicates as the source of revenue increases (diversified), the size of NPLs

would also increase. In other words, the size of NPLs tends to be higher in banks that

have diversified source of revenue (income other than interest) than banks that are

specialized on the traditional source of revenue (interest income). The intensity of the

coefficient estimate (0.64) was the third largest of all the variables used in the model next

83
to performance (ROA) and effective exchange rate (EFEX). However, the coefficient

estimate of income diversification was statistically insignificant since the p-value

(0.3451) was considerably in excess of the acceptable 10% significance level. That means

the importance of income diversification in explaining the variation of NPLs in the

Ethiopian banking industry was not significant. Hence, income diversification was not

considered as the determinant of NPLs in ECBs. This finding was consistence with

previous studies of Hu et al. (2004) and Jellouli et al (2009). Regardless of the findings

of the regression analysis, the results obtained from interview had no consensuses

on the importance of income diversification in explain NPLs of ECBs. Most of the

interviewees suggested that, income diversification of a bank was one of the major

determinants that can affect NPLs of ECBs negatively since the potential losses on the

loan activity of banks might be overcome or compensate by looking for non interest

sources of revenues. On the other hand, some interviewees did not support the

significance of income diversification in explaining the variation of NPLs in ECBs.

Therefore, the conclusion about the impact of IDV on the NPLs of ECBs remains

ambiguous and further research is required.

Ownership structure

The coefficient estimate of ownership structure of banks which was measured by a

dummy variable (1=state owned banks and 0=private banks) revealed a positive and

statistically significant at 1% significance level (p-value of 0.00). This result was

agreement with the previous expectation and theory that suggested state owned banks

tends to have high volume NPLs. A positive association between state ownership and

84
NPLs in Ethiopian banking sector indicates that, the level of NPLs tends to be higher in

state owned banks of Ethiopia than privately owned banks. However, the magnitude of

the coefficient estimate (0.076) was very small as compared to other variables like

financial performance and effective exchange rate. The finding was consistence with the

previous studies of Salas and Saurina (2002), Hu et al. (2004), Micco et al. (2004), Barth

et al. (2004) , Garcia and Robles (2007) and Swamy (2012). Moreover, the result

generated from the interview also supports the regression output.

As per the interview result, ownership structure of Ethiopian banks was one of the

determinants of NPLs and the positive association between state ownership and NPLs in

ECBs could explain in the following two facts. First, in developing countries like

Ethiopia the mandate of economic development usually given to state-owned banks, thus

so as to achieve the development program of a country state-owned banks have more

incentives to fund riskier projects and to allocate more favorable credits for small and

medium firms. Hence, the high risk taking behavior of state owned banks may hinder the

quality of their loan portfolios. As per the interviewees, commercial bank of Ethiopia (the

largest state owned Ethiopian bank) was the only bank that providing a fund for

government real estate programs (condominiums). Second, the lack of a profit motive in

state-owned banks (as compared to private banks) may hinder the efficient credit risk

management of banks and ultimately lead to higher level of NPLs.

As can be seen in figure 4.1, the variation of NPLs among private banks was very small

and in the same way the variation of NPLs among the two state owned banks also small.

However, the variation of NPLs between private and state owned banks was significantly

high. When we see merely the average NPLs ratio of ECBs over the sample period, NPLs

85
of state owned banks tends to be higher. Currently however, NPLs ratio of state owned

bank (especially CBE) considerably small as compared to the privet banks. For instance,

NPLs of CBE lowered from 22% in2006 to an average of 0.86% in 20011. Having the

above discussions the interviewees were asked their own opinion on the likely causes of

NPLs disparity among state especially CBE and privet banks. In this regard,

interviewees from CBE suggested that, the existing abrupt change in their volume of

NPLs was resulted from the implementation of business process reengineering (BPR) that

improved the overall system of the bank in general and their lending activity in particular.

Accordingly, the traditional lending system in which every branch managers approved a

loan had changed and currently loans applications could only approved at district or at

head office level after the passage of different channels. Hence, as per the interview, the

existence of improved loan screening, measuring, monitoring and controlling system in

CBE enable to reduced their NPLs as low of the industry.

On the other hand, most of the interviewees from privet banks credit officer were

strongly questioning about the existence of Political reasons that made them incapable to

compete with state owned banks especially CBE. As per the interviewees, state owned

banks get strong support from government. For instance CBE has a priority to write letter

of credit in the case of foreign trade, to involve in government Treasury bill market, bond

market and to pay salary for federal organizations. The stiff competition from state

owned banks forces private banks to involve in risky businesses like loans and advances.

As result, the reduction of NPLs in Ethiopian private banks was not significant as

compared to the performance of CBE.

86
Bank Size

As mentioned in the literature, many scholars suggested an inverse relationship between

size of a bank and bank‟s NPLs. large banks have better risk management strategies and

technology which definitely allows them for efficient information gathering, processing

and analyzing which finish up with lower levels of NPLs as compared to smaller banks.

Despite this fact , the coefficient estimate of bank size (measured by the natural log of

total assets) in this particular study found to be positive and statistically significant at

5% significance level (P-value of 0.0345). This indicates that, as the size of Ethiopian

banks increases, their volume of NPLs would also increase. In other words, in ECBs, the

level of NPLs was relatively higher in banks that have large size. However, the

magnitude of the coefficient estimate (0.009) for bank size was the smallest of all the

variables used in the model implying that the responsiveness of bank size on the variation

of NPLs was very low as compared to other variables. Hence, the finding suggested that,

bank size was the modest determinate that can affect NPLs of ECBs positively. This was

inconsistence with prior expectation and theory that suggesting larger banks have more

resources for efficient information gathering, processing and analyzing to tackle moral

hazard and adverse selection and ultimately deal with lower volume of NPLs. In addition,

the finding was not in agreement with those reported by the majority of previous studies

such as Salas and Saurina (2002), Rajan and Dhal (2003), Hu et al, (2006) , Jellouli et al

(2009) and Espinoza and Prasad (2010). This is due to the fact that, larger banks in

Ethiopia are state owned banks that have less incentive to profit as compared to private

banks(small banks). Hence, in order to achieve the development program of country

Ethiopian state-owned (large) banks may have more incentives to fund riskier projects

87
that ultimately increase the volume of NPLs. Correspondingly, the result generated from

the interview also supported the output of the regression analysis fully. That is Ethiopian

banks NPLs reduce as the size of the banks decreases. Moreover, this result was also

consistent with the result of the descriptive analysis that smaller banks are enjoying lower

NPLs than larger banks of the country.

Gross domestic product

The coefficient estimate of real GDP growth revealed a negative and statistically

significant association with NPLs of Ethiopian commercial banks at 1% significance

level (p-value of =0.00). The magnitude of the coefficient estimate (-0.45) indicates the

existence of strong inverse relationship between real GDP growth and NPLs of

Ethiopian banks which was in accordance with the theory and prior expectation. This

implies that, an increase in the real GDP growth, certainly lead to a decrease in the

likelihood that loans became default. This is due to the fact that, strong positive growth in

real GDP usually translates into more income which improves the debt servicing capacity

of borrower (households and businesses). Whenever there was a positive GDP growth,

the economic activities in general were increasing and the volume of cash held for either

businesses or households was increasing. These conditions contributed to decrease the

likelihood that borrowers delay their financial obligations. Hence, the findings suggested

that, real GDP growth was one of the vital determinants of NPLs in Ethiopian

commercial banks. This result was in consistent with the findings of Salas and

Suarina(2002), Jajan and Dhal (2003), Fofack (2005) , Hou (2006) Jimenez and Saurina

(2005), Pasha and Khemraj (2009), Louzis et al. (2010) and Azeem et al. (2012). In

addition, the result obtained from the interview also highly supported the output of the

88
regression analysis since strong performance in the real economy results in lower NPLs.

Moreover, this result also consistent with the existing reality in the Ethiopian banking

industry where the volume of NPLs shows a parallel decrease as the economy grows up.

Inflation

As mentioned in the literature review part, inflation affects borrowers‟ debt servicing

capacity through different channels and its impact on NPL can be positive or negative.

Higher inflation rate can make borrowers debt servicing easier by reducing their real

value of outstanding loans. However, it can also weaken some borrowers‟ ability to

service debt by reducing their real income. Nevertheless, in this study the coefficient

estimate of inflation was negative and statistically significant at 1% significant level (P-

value of 0.000). The negative coefficient estimate of inflation (-0.235) indicates a strong

inverse association with NPLs. That means an increase in inflation rate; lead a decrease

in NPLs. This result was inconsistent with the findings of Fofack (2005), Pasha and

Khemraj (2009), Louzis et al. (2010) and Azeem et al. (2012). As the existing theories

suggested this relationships appeared in the banking system where the lending rate is not

adjusted to the inflation change.

Correspondingly, the result obtained from interview revealed the existence of similar fact

in Ethiopia banking system. As per the interview conducted with the credit vice

president and senior credit committee members of the selected banks, inflation was

one of the major factors that can affect NPLs of ECBs negatively. The inverse

relationship between inflation and NPLs in ECBs is due to the fact that, in Ethiopia the

maximum lending rate is determined by National bank of Ethiopia and ECBs are unable

89
to adjust their lending rate in accordance with different factors such as time value of

money, inflation and business risk. For instance, the average annual inflation rate in

Ethiopia over the period of consideration was 11% with a maximum of 36.4%. Despite

this fact, the average lending rate of ECBs never exceeds 12.75% over the sample period.

This clearly indicates the lending rate in Ethiopia was far below from the market interest

rate since ECBs were not allowed to adjust their lending rate to compensate the existing

high inflation rate. In this general setting, it could be conclude that, the existing higher

inflation rate in Ethiopia was in favour of borrowers since it can make debt servicing

easier by reducing the real value of outstanding loans.

Real effective Exchange rate

As shown in the literature review part of this study, the impact of exchange rate on

bank‟s NPLs have mixed implications. Hence, the sign of the relationship between

exchange rate and NPL can be positive or negative. A depreciation of the exchange rate

can have mixed implications on NPLs of banks. On the one hand, it can improve the

competitiveness of export-oriented firms and ultimately increase their ability to service

debt (Fofack, 2005). On the other, it can negatively affect the debt-servicing capacity of

borrowers who borrow in foreign currency (import-oriented firms). Despite this fact, the

coefficient estimate of effective exchange rate (EFEX) in this particular study revealed

negative association with NPLs of ECBs. This result was inconsistent with the findings of

Jajan and Dhal (2003), Fofack (2005), Pasha and Khemraj (2009), De Bock and

Demyanets (2012) and Castro (2012). The magnitude of the coefficient estimate (-1.53)

result of EFEX was the second largest amount next to performance of a bank (ROA).

This indicates that, EFEX had a great impact in explaining the variation of NPLs in

90
Ethiopia commercial banks. Moreover, the coefficient estimate of EFEX was statistically

significant at 1% significant level (p-value = 0.0). This implies that, an increase in EFEX

(i.e., a depreciation of Ethiopian birr in terms of dollar); lead to a decrease in NPLs of

ECBs. More specifically, as the value of Ethiopian birr depreciated in terms of dollar, it

can increase the competitiveness export-oriented Ethiopian firms in the international

market. This due to the fact that, the operating cost of export-oriented Ethiopian firms

was very less as compared to the international firms since the value domestic currency

was very small in terms of foreign currency (dollar). Consequently, the debt-servicing

capacity of export-oriented Ethiopian firms would improve. This result was in accordance

with the import substitution policy of Ethiopian government that encouraging export-

oriented firms. In this regard, ECBs are providing loans with lower interest rate to export-

oriented firms so as to encourage the export sector. For instance, commercial bank of

Ethiopia providing a loan with 7.5 % interest rate for export-oriented firms which is

considerably lower than the stated 9.5% interest rate for any other sectors. Hence, the

lower interest rate for export-oriented firms can also make their debt servicing easier.

On the other hand, as the theory suggested, depreciation of domestic currency in terms of

foreign currency obviously deteriorate the performance of import-oriented firms. In

Ethiopia the trading system was often manifested by deficit (high import). In this

condition, a negative association between exchange rate (appreciation of dollar) and

NPLs was surprised. As per the interviewee‟s opinion, the import market in Ethiopia

was dominated by limited numbers of importers that have strong relationship with banks.

Whenever the value of dollar appreciated, small and medium importers go out of market

and those limited large importers dominate the market monopoly. This allows them to

91
manipulate the price as they wish and maximize their profit. Hence, the profit originated

from the monopolistic advantage of importers enables their debt servicing easier.

Real interest rate

The coefficient estimate of real interest rate (measured by the average lending rate of

ECBs) was positive which was in accordance with prior expectation and theory (Sinkey

and Greenwalt (1991),Fofack (2005), Jimenez and Saurina(2005)).This implies that an

increase in the average lending rate of ECBs, ultimately increase the size of their NPLs.

That means as interest rates rise, prudent borrowers are more likely to decide that it

would be unwise to borrow, whereas borrowers with the riskiest investment projects are

often those who are willing to pay the highest interest rates. Hence, higher interest rate

leads to greater adverse selection that increases the likelihood that the lender is lending to

a bad credit risk which ultimately increases the volume of banks NPLs. However, the

magnitude of the coefficient estimate (0.064) was the smallest of all variables used in the

model next to size and operational efficiency of banks. This indicates that, the average

lending rate of ECBs had a very limited impact in explaining the variation of NPLs in

Ethiopia banks.

In addition, the coefficient estimate of real interest rate was statistically insignificant

since the p-value (0.93) was considerably in excess of the acceptable (10%) significance

level. Hence, the findings suggested that, there was no significant association among real

interest rate and bank‟s NPLs in ECBs as far as the parameter for this variable was

insignificant as illustrated by the large p-values of 0.93. Similarly, the result generated

from the interview supported the output of the regression analysis fully. As per the

92
interview conducted with credit vice presidents and senior credit committee members of

ECBs, real interest rate was not considered as the determinate of NPLs since banks are

not allowed to make any adjustment on their lending rate so as to compensate losses from

the existing high inflationary rate. Consequently, the volatility of the average lending rate

of ECBs being very small, in fact this interest was justified in the summary of

descriptive statistics that real interest rate had the smallest standard deviation of all the

variables used in the model. This implies the existence of very small variation on the

average lending rate of ECBs over the sample period. Therefore, as long as the variation

in the average lending rate of ECBs very small, the significance in explaining the

variation of NPLs would also be very small or insignificant. The finding was

inconsistence with the result of Swamy (2012), De Bock and Demyanets (2012), Azeem

et al. (2012) and Castro (2012).

In addition to the discussions so far, this study also identified several other factors that

have their own impact on NPLs in the context of ECBs. As the interview results shown in

section 4.2, inadequate credit risk management from identifying, measuring and

monitoring of credit, and collateral based lending system were considered as the major

internal factors that can affect NPLs of ECBs. In relation to credit risk management

practice of ECBs, deficiencies in credit information was considered as the reason for

NPLs since sole reliance on the incorrect and inadequate information provided by

the clients especially who had not previously dealt with the bank significantly affect

the loan quality of banks.

As per the interviewees, Ethiopian commercial banks are considering collateral as the

prime factor for assessing loan application in all conditions and most important factors

93
such as the repayment capacity of the client, the feasibility of the project and the

management experience of the company are coming after the issue of the collateral.

Hence, weaknesses of estimating the value of the guarantees provided by the

clients can cause a high level of NPLs in Ethiopian banks. This is due to the fact that,

employee of the bank may intentionally (moral hazard) or unintentionally (due to lack of

experience) overvalued the collateral provided by the loan applicants. In this condition

borrowers usually preferred to default at the date of maturity since the market value of

the collateral is lowered than the value of loan outstanding. Furthermore, absence of

adequate man power, lack of comprehensive studies on the credit applicants (the

financial analyses of the customer), lack of follow-up on the borrower's activities or

failure to follow up the collateral provided by the borrowers were also suggested by

the interviewees as the major internal determinants of NPLs in ECBs. As per the

interviewees, borrowers fund diversion for unintended purpose was a major determinate

of NPLs in ECBs. For instance, borrowers who took a loan for a new investment purpose

or for expanding his/her existing business may use the loan for other unproductive

activates like buying luxury car. In addition, borrowers who obtained short term loans

from banks through overdraft (in any other form of short term loan application) may use

the fund for capital expenditures that do not generate immediate income for servicing

their debt in accordance to the sated agreement. In addition, intentional or willful default

of borrowers at the date of maturity also the major factors of NPLs in Ethiopia banks. As

per the interview, the rationale behind the intentional default of borrowers was usually

the market value of the collateral. Borrower usually makes their own cost and benefit

analysis between the market value of the collateral and the value of loan outstanding.

94
Hence, whenever the market value of the collateral less than the value of the outstanding

loan borrowers preferred to default since there is no rule that allowed banks to claim any

other personal property of borrowers more than the agreed collateral. As per the

interviewees, the national bank regulation can also affect the NPLs of Ethiopian banks in

different aspects. For instance, the 2009 directive which required all Ethiopian banks not

to hold NPLs in excess of 5% of their total loans outstanding pressurized banks not

engaged in risky investment. In contrary, the absence of rules that allowed banks to

claim beyond the pledged collateral could encourage willful defaulter in the condition

where the market value of the collateral lower than the value of outstanding loans. In

addition, due to the absence of credit rating agency in the country, the loan repayment

capacity of borrowers was exclusively measured by bank‟s credit officers. Hence, the

exclusively burden of Ethiopian banks in screening the loan application may hinder the

quality of their loan portfolios. Besides, the surrounding natural environment also affects

the loan quality of Ethiopian banks. Particularly, loans granted to the agriculture sector

greatly affected by the climate change. Whenever a bad climate condition like flood and

famine happened in the country the income of the agriculture sector greatly affected and

ultimately their debt servicing capacity reduced significantly. Furthermore, international

competition in the global market also affects the loan repayment capacity of export-

oriented Ethiopian firms. As per the interviewees, there was situation in which Ethiopia

coffee exporters face difficulty in servicing their debt up on the maturity due to the high

competition in international market from Brazilian coffee exporters.

95
Chapter six: Conclusions and Recommendations

The previous chapter presented the analysis of the findings obtained from different data

sources.The purpose of this chapter is to discuss the conclusions and recommendations.

Accordingly, the chapter is organized in two sections, the first section, 6.1 presents the

conclusions of the study. And, the second section, 6.2 presents the recommendations that

provided based on the findings of the study.

6.1. Conclusions

The broad objective of this research was to investigate bank specific and macroeconomic

determinants of NPLs in ECBs. To achieve this broad objective, the study used mixed

methods research approach. More specifically, quantitative research approach along with

survey design (structured review of documents) was dominantly used. In addition, to

have a better insight and to gain a richer understanding about the research problem, the

quantitative method was supplemented with the qualitative method (unstructured

interviews). To this end, the collected data from a sample size of eight Ethiopian

commercial banks over the period of 2000 to 2011were analyzed using descriptive

statistics, correlation matrix and multiple linear regression analysis. The analyses were

made in line with the stated hypotheses and specific research questions formulated in the

study. In doing so, previous studies on determinants of bank‟s NPLs have been

reviewed and as per the literature NPLs of banks‟ usually expressed as a function

of internal and external determinants. The internal determinants refer to those factors

which characterized individual banks and usually associated with the specific policy

choices of a particular bank such as loan growth, performance, the quality of the loan

96
portfolio and operational efficiency. On the other hand, the external determinants are

variables that are not related to bank management but reflect the economic , legal and

the surrounding natural environment that can affect the loan quality of banks. The

macroeconomic factors which the literature proposes as important determinants of NPLs

are: annual growth in GDP, the annual inflation rate, real interest rate, real effective

exchange rate (REER), annual unemployment rate, broad money supply (M2) and GDP

per capital. Accordingly, in this study, six bank specific variables (i.e., loan growth,

ownership structure, size, income diversification, financial performance, operational

efficiency of banks) and four macroeconomic variables (i.e., real GDP growth, annual

inflation rate, real interest rate and effective exchange rate) were included.

Consequently, the empirical findings of this particular study suggested the following

conclusions:

First, among bank specific variables, loan growth (LG), financial performance measured

in terms of return on asset (ROA) and ownership structures of banks were found to be a

major determinant of NPLs in ECBs with 1% significance level. Particularly, the ROA

of banks had a negative relationship with the volume of NPLs reported by ECBs

suggesting that as the profitability of banks increases, the likelihood that managers

engaged in risky lending activity decreases and ultimately reduce NPLs. The findings

also suggested a negative relationship among loan growth of a bank and NPLs of ECBs

which was inconsistent with the prior expectation. On the other hand, the ownership

structure of Ethiopian banks measured by a dummy variable (1=state owned banks and

0= privately owned banks) had a positive association with NPLs of Ethiopian commercial

banks. This implies state ownership increase the probability of NPLs and thus the volume

97
of NPLs tends be higher in state owned Ethiopian banks than privately owned banks. Size

of a bank had also a positive association with NPLs of ECBs. This indicates the larger

banks of Ethiopian have relatively higher volume of NPLs than smaller banks. The

operational efficiency of banks had a negative association with NPLs of ECBs which was

in favor of „skimping‟ hypothesis that banks which allocate adequate budget to screening

loans, appraising collateral, and monitoring and controlling borrowers after loans

disbursement resulted with lower volume of NPLs as compared to banks which seem to

cost efficient but, do not have adequate budget to ensure higher loan quality. However,

the magnitude of the coefficient estimate and the level of significance was not strong as

compared to other variables. Hence, operational efficiency of a bank found to be the

modest determinants of NPLs in ECBs.

Second, with respect the macroeconomic variables, real GDP growth, annual inflation

rate and effective exchange rate were found to be statistically significant determinants of

NPLs in ECBs. In particular, inflation rate had a negative association with the levels of

NPLs reported by Ethiopian banks suggesting that the absence of adjustment on the

lending rate of ECBs (to compensate the inflation rate) enhanced the debt servicing

capacity of borrowers by reducing the real value of the outstanding loans. The findings

also suggested a significant inverse relationship among real GDP growth and NPLs of

ECBs which indicates whenever there was a strong positive economic growth, the

volume of cash held for either businesses or households was increasing. These conditions

contributed to decrease the likelihood that borrowers delay their financial obligations.

In addition, the effective exchange rate had also a negative association with NPLs of

Ethiopian commercial banks. This implies the depreciation of Ethiopian birr in terms of

98
dollar reduced the volume of NPLs reported by Ethiopian banks through increasing the

competitiveness of export-oriented Ethiopian firms in international market and it can also

increase the monopolistic power of large importers of the country by getting out of

market those small and medium importers. Hence, this conditions increase the debt

servicing capacity of borrowers (both import and export-oriented).

Third, as the interview suggested, other internal factors such as absence of adequate man

power, lack of comprehensive studies on the credit applicants, lack of follow-up on

the borrower's activities or failure to follow up the collateral provided by the

borrowers were also the major internal determinants of NPLs in ECBs. In addition,

factors related to the borrowers such as providing false information to the bank, using

the loan for other purposes that are undesirable from the banks' point of view (fund

diversion), willful default and operational losses of borrower were also the determinants

of NPLs in ECBs. Moreover, external variables such as the international conditions

(competition), absence of credit rating Agencies Company in the country, government

regulation, and the surrounding natural environment were also affect the NPLs of ECBs.

Lastly, the result of average lending rate and income diversification of Ethiopian banks

did not support an important association with NPLs reported by Ethiopian banks. In fact,

the result of income diversification and the average lending rate of Ethiopian banks

showed a positive relationship with NPLs. However, the association was statistically

insignificant since the p-values for both IDV and RIR are in excess of 10%. Hence, both

income diversification and average lending rate of Ethiopian banks were not found to be

important determinants of NPLs in Ethiopia commercial banks.

99
6.2 Recommendations

Based on the findings of the study the following possible recommendations were

forwarded:

Loan growth, financial performance, operational efficiency, GDP growth rate, effective

exchange rate and inflation rate were the significant drivers of NPLs in Ethiopian

commercials banks. Hence, focusing and reengineering the institutions alongside

these indicators could reduce the probability of nonperforming loans in Ethiopian

commercial banks. Specifically, commercial banks need to consider the performance of

the real economy when extending loans given the reality that NPLs are likely to be lower

during the periods of economic growth.

As per the interviewees, borrowers related factors (such as fund diversion, willful default

and providing false information) and internal factors (such as lack of comprehensive

studies on the credit applicants and lack of follow-up on the borrower's activities)

were quite important determinants of NPLs in Ethiopian commercial banks. Thus,

Ethiopian commercial Banks that were considered in this study should put in place a

vibrant credit process that would encompass issues of proper customer selection, robust

credit analysis, authentic sanctioning process, proactive monitoring and follow up and

clear recovery strategies for sick loans.

Currently, Ethiopian commercial banks that were sampled in this study were considering

collateral as prime factor for assessing loan application in all conditions and hence,

providing appropriate focus for factors such as repayment capacity of the client, the

100
feasibility of the project and the experience of the management of the company in credit

approval process could improve the quality of their loan portfolios.

Finally, the study sought to investigate the determinants of NPLs in Ethiopian

commercial banks. However, the variables used in the econometrics model did not

include all factors that can affect NPLs of Ethiopian commercial banks. Thus, future

research could incorporate external factors such as regulatory environment (capital

adequacy ratio and loan loss provisions) and corporate governance.

101
 Reference

Abera, A2012, „Factors Affecting Profitability: An Empirical Study on Ethiopian Banking

Industry‟, MSC thesis, Addis Ababa University

Achou, FT & Tenguh, CN 2008, „Bank performance and credit risk management‟, MA Degree

Project in Finance, Skovode Univercity

Adebola, S, Yusoff, S & Dahalan, D2011, „Determinants of nonperforming loans in Islamic

Banking system in Malaysia‟, Arabian Journal of Business and Management Review,

Vol. 1, No.2, PP.20-30

Ahmad, HN & Ariff , M 2007, „Multi-country study of bank credit risk determinants‟,

International Journal of Banking and Finance, Vol. 5, No.1, PP.135-152

Ahmed, SZ 2006, „An investigation of the relationship between Non-performing Loans,

Macroeconomic Factors, and Financial Factors in context of Private Commercial Banks

in Bangladesh‟, Independent University, Bangladesh

Akerlof, G1971, Theory of asymmetric information, 1st edn, Oxford Universities press

Arestis, P & Sawyer, MC2006, „A Handbook of Alternative Monetary Economics‟, Cheltenham,

Edward Elgar Publishing Ltd

Ayalew, T2009, „Legal Problems in Realizing Non-Performing Loans of Banks in Ethiopia‟, MA

thesis, Addis Ababa University

Azeem, M, Kouser, R& Saba, I2012, „Determinants of Non Performing Loans: Case of US

Banking Sector‟, Romanian Economic Journal, Vol. XV, No.44, PP.125-136

Basu, S2003, „Why do Banks Fail?‟, International Review of Applied Economics, Vol. 17,

pp. 231-248

102
Bebczuk, RN2003, „Asymmetric Information in Financial Markets: Introduction and

Applications‟, Cambridge University Press, New York

Berger, NA and DeYoung, R1997, „Problem Loans and Cost Efficiency in Commercial Banks‟,

Journal of Banking and Finance, Vol. XV, No.21, pp.849-870

Berger, NA, & Udell, D2004, „The Institutional Memory hypothesis and the procyclicality of

bank lending behaviour‟, Journal of Financial Intermediation, Vol.13, No.4, PP. 458-495

Bercoff, J, Julian, G & Franque, G2002, „Argentinean Banks, Credit Growth and the Tequila

Crisis: A Duration Analysis‟ (Unpublished).

Bloem, AM & Freeman, R2005, „The Treatment of Nonperforming Loans in Macroeconomic

Statistics‟, IMF Working Paper WP/01/209, International Monetary Fund.

Boudriga, A, Neila, B & Jellouli, S2009, „Does bank supervision impact nonperforming loans:

Cross-country determinants using aggregate data‟, Munich Personal RePEc Archive

(MPRA), No. 18068, PP.1-28.

Brooks, C 2008, Introductory Econometrics for Finance, 2nd edn, Cambridge University Press,

New York

Breuer, JB2006, „Problem bank loans, conflicts of interest and institutions‟, Journal of Financial

Stability, Vol.2, No.3, pp. 226–285.

Caprio, G & Kilngebiel, D2002, „Episodes of Systemic and Borderline Financial Crisis‟, World

Bank (Unpublished)

Castro, V 2012, „Macroeconomic determinants of the credit risk in the banking system: The case

of the GIPSI‟, University of Coimbra and NIPE, Portugal.

Casu, B, Girardone, C & olyneux, P2006, „Introduction to Banking‟, Harlow Financial Times

103
Creswell, JW 2003, Research design: qualitative, quantitative and mixed methods approaches,

2nd edn, Sage Publications, California.

Creswell, JW 2009, Research design: quantitative, qualitative and mixed methods approaches,

3rd edn, Sage Publications, California.

Coen, B2010, „the Basel Committee on Banking Supervision: Its Global Role and Current

Initiatives‟, United Nations General Assembly Working Group on the World Financial

and Economic Crisis and its Impact on Development, New York

Cortawarria, L, Dziobek, CH, Kanya, A & Song, I2000, „ Loan Review, Provisioning, and

Macroeconomic Linkages‟, IMF Working Paper, Washington Dc.

De Bock, R & Demyanets, A2012, „Bank Asset Quality in Emerging Markets: Determinants and

Spillovers‟, IMF Working Paper WP/12/71, International Monetary Fund

Espiloza, R, & PRASAD, A 2010, „Nonperforming Loans in the GCC Banking System and

their Macroeconomic Effects‟, IMF Working Paper WP/10/224, International

Monetary Fund.

Fofack, H2005, „Nonperforming loans in Sub-Saharan Africa: Causal Analysis and

Macroeconomic implication‟ World Bank Policy Research Working Paper 3769,

//econ.worldbank.org.

Fowler, FJ 1986, Survey research methods, 3rd edn, Sage Publications, USA

Garcia, T and Robles, MD2007, „Risk-taking behavior and ownership in the banking industry:

The Spanish evidence‟, Journal of Economics and Business, Vol. 60, No. 4, PP. 332-354.

Geda, A 2006, „Structure and performance of Ethiopia‟s financial sector in the pre & post reform

period: with special focus on banking‟, Research paper no. 2006/112, Addis Ababa

University.

104
Gujarati, DN 2004, Basic Econometrics, 4th edn, McGraw Hill, Boston

Fernandez, S, Martinez, J and Jesus, S2000, „Credit Growth, Problem Loans and Credit Risk

Provisioning in Spain‟, Working Paper No. 0018.

Fisher, I1933, „The Debt Deflation Theory of Great Depressions‟, Econometrical, Vol.1, PP.

337-57

Fuentes, R & Carlos, M2003, Institutional arrangements, credit market development and loan

repayment in Chile, School of Business and Economics, University of Chile

Hartelius, D2010, „Analyzing nonperforming loans in Central and Eastern Europe based on

historical experience in emerging market‟, Global Financial Stability Report,

Washington

Hair JF, Black, WC, Babin, BJ, Anderson, RE & Tatham, RL 2006, Multivariate data analysis,

6th edn, Pearson Education, New Jersey.

Hafer, RW2005, „The Federal Reserve System‟, Greenwood Publishing Group.

Hermes, N and Lensik, R1996, „Financial Development and Economic Growth: Theory and

Experiences from Developing Countries‟, Routledge press, London

Hefferan, SA2005, „Modern Banking‟, Chichester, Wiley.

Hou, Y2006, „the non-performing loan problem: some bank level evidence‟, University of

Birmingham Edgbaston, Birmingham

Hu, J, Li, Y, and Chiu, Y2004, „Ownership and nonperforming loans: Evidence from Taiwan's

banks‟, The Developing Economics, Vol. 42, No.3, PP. 405-420

Issa, AG 2009 „Analysis of non-performing loans in the Libyan state-owned commercial banks:

perception analysis of the reasons and potential methods for treatment‟, PHD thesis,

Durham University

105
Jusoff, K, Ahemed, W, Muhamad, S & Asari, H2011, „analysis of Non-performing loans,

interest rate, inflation rate‟, world applied science journal, Vol.12, PP.41-48

Jellouli, S Taktak, BN Boudriga, A 2009, „Bank specific, business and institutional environment

determinants of nonperforming loans: Evidence from Middle East and North African

countries‟, economic research forum, Cairo, Egypt.

Jimenez, G and Jesus, S2005, „Credit cycles, credit risk, and prudential regulation‟, Working

Paper, No. 0018

Jiminez, G & Saurina, J2006, „Credit cycles, credit risk and prudential regulation‟, International

Journal of Central Banking, Vol.2, No.2, PP. 65-98

Kearns, A2004, „Loan Losses and the Macro economy: A Framework for Stress Testing Credit

Institutions‟ Financial Well-Being‟, Journal of Financial Intermediation, Vol.12, No.3,

PP.111-121.

Keeton, WR1999, „Does faster loan growth lead to higher loan losses?‟ Economic Review, Vol.

XV, No.21, pp. 57-75.

Keeton, WR & Morris, CS1987, „Why Do Banks‟ Loan Losses Differ?‟, Federal Reserve Bank

of Kansas City Economic Review,PP.3-21.

Kiyotaki, N & Moore, J1997, „Credit Cycles‟, Journal of Political Economy, Vol.105, No. 2,PP.

211-248.

Kolb, RW & Rodriguz, R J 1996, Financial Institutions and Markets, Oxford, Blackwell

Publishers, Inc

Kunt, DA & Detragiache,E 1998, „The Determinants of Banking Crises in Developing and

Developed Countries‟, International Monetary Fund, Vol. 45, No. 1, pp. 81-109

106
Leedy, PD & Ormrod, JE2005, Practical research: planning and design, 8th edn, Upper Saddle

River publication, Pearson

Lelissa BT 2007, „The impact of financial liberalization on the ownership, market structure &

performance of the Ethiopian banking industry‟ MBA thesis, Addis Ababa University.

Lepetit, L, Nys, E, Rous, P & Tarazi, A2008, „Bank income structure and risk: An empirical

analysis of European banks‟, Journal of Banking & Finance, Vol. 32, No. 8, PP.1452-

1467

Li, Y2003, „the Asian Financial Crises and Non-performing Loans: Evidence from Commercial

Banks in Taiwan‟, International Journal of Management, Vol. 20, No.1, PP. 69-73.

Louzis, Pd, Vouldis, TA & Metaxas, LV 2010, „Macroeconomic and bank-specific determinants

of non-performing loans in Greece: a comparative study of mortgage, business and

consumer loan portfolios‟, Bank of Greece Working Paper 118, Athens, Greece.

Mehari, M 2012, „credit risk management in Ethiopia banking industry‟, MBA thesis, Addis

Ababa University

Mekasha, G 2011,‟ Credit Risk Management and Its Impact on Performance on Ethiopian

commercial Banks‟, MSc thesis, Addis Ababa University

Micco, A, Panizza, U and Yanez, M2004, „Bank Ownership and Performance‟, Inter-American

Development Bank, working paper 518

Masood,O & Aktan, B2009, „Dterminats of Non- performing loans :comparative analysis‟

АКТУАЛЬНІ ПРОБЛЕМИ ЕКОНОМІКИ,Vol.102, No.12, PP.251-163

Nkusu,W 2011, „Nonperforming Loans and Macrofinancial Vulnerabilities in Advanced

Economies‟ International Monetary Fund, Vol.11, No.161

107
Negera, W 2012, „Determinants of Non Performing Loans The case of Ethiopian Banks‟

research report, University of South Africa

National bank of Ethiopia 2009/10, „survey on risk management practice of Ethiopian

commercial banks‟, National Bank of Ethiopia, Addis Ababa, Ethiopia

National bank of Ethiopia 20010/11, „Annual report‟, National Bank of Ethiopia, Addis Ababa,

Ethiopia

Pasha,S and Khemraj, T 2009,” The determinants of non-performing loans: an econometric case

study of Guyana”, The Guyana Bank(Unpublished)

Petra T 2007, Panel data; fixed effects random effects dynamic panel data models,

Cambridge University Press, New York.

Rajan, R1994, „Why Bank Credit policy Fluctuate: A theory and Some Evidence‟, Quarterly

Journal of Economics, Vol.109, No. 2, PP.399-441.

Rajan, R and Sarat, CD2003, „Non-performing Loans and Terms of Credit of Public Sector

Banks in India: An Empirical Assessment‟, Reserve Bank of India, Vol. 24, No.3, pp. 81-

121.

Ramachandran, A, Thiagarajan,S & Ayyappan, S2011, „Credit Risk Determinants of Public and

Private Sector Banks in India‟, European Journal of Economics, Finance and

Administrative Sciences, Vol.34, PP.1450-2275

Resti, A1995, „Linear Programming and Econometric Methods for Bank Efficiency Evaluation:

An Empirical Comparison based on a Panel of Italian Banks‟, working paper

Robinson, MS2001, „the Microfinance Revolution: Sustainable Finance for the Poor‟,

Washington, the World Bank

Rose, P2002, „Commercial Bank Management‟, McGraw-Hill Companies, Inc.

108
Salas, V and Saurina, J2002, „Credit risk in two institutional regimes: Spanish Commercial and

Savings banks‟, Journal of Financial Services Research, Vol. 22, No 3, PP. 203-224.

Safakli, O2007, „Credit Risk Assessment for the Banking Sector of Northern Cyprus‟, Banks and

Bank Systems, Vol.2, PP. 21 -31.

Shehzad, C, Haan,T & Schltens, B2010, „The impact of bank ownership concentration on

impaired loans and capital adequacy‟, Journal of Banking & Finance, Vol. 34, pp. 399-

408.

Sinkey, JF and Greenawlat, MB1991, „Loan-Loss Experience and Risk-taking Behaviour at

Large Commercial Banks‟, Journal of Financial Services Research, Vol.5, No 1, PP. 43-

59

Socuvkova, A 2012, „Nonperforming Loans in China: Do We (Still) Need to Worry?‟ MA thesis,

Charles University, Prague

Stiroh, KJ2002, „Diversification in banking: Is noninterest income the answer?‟ Federal Reserve

Bank of New York, PP. 1-26.

Stiroh, KJ2006, „A portfolio view of banking with interest and noninterest activities‟ Journal of

Money, Credit and Banking, Vol. 38, No.5, PP.1351-1361.

Stiroh, KJ & Rumble, A2006, „the dark side of diversification: The case of US financial holding

companies‟ Journal of Banking and Finance, Vol.30, PP. 2131-2161.

Strauss, AL & Corbin, JM1998, Basic of Qualitative Research, 2nd edn, Thousand Oaks, Sage

Publications

Swamy, V 2012, „Impact of macroeconomic and endogenous factors on non performing bank

assets‟, The International Journal of Banking and Finance, Vol.9, No.1, PP.27-47

109
Templeton, WK, & Severiens, J1992, „The effect of nonbank diversification on bank holding

companies‟, Quarterly Journal of Business and Economics, Vol.31, No.4, PP. 3-16.

Tefera, T 2011, „Credit risk management and profitability of commercial banks in Ethiopia „,

MSc thesis, Addis Ababa University

Tilahun, D 2010, „privately owned commercial banks in Ethiopia: issues of nonperforming

loans‟, MSc thesis, Addis Ababa University

Vogiazas, D & Nikolaidou, E 2011, „Credit risk determinants in the Bulgarian‟, Vogiazas-

Nikolaidou, PP. 177-189

Wheelock, D &Wilson, W1994, „Explaining Bank Failures: Deposit Insurance, Regulation, and

Efficiency‟, Review of Economics and Statistics, Vol.77, PP. 689-700.

World bank 2012, Bank nonperforming loans to total gross loans (%)

Wydick, B2008, „Games in Economic Development‟, Cambridge University Press, New York

Winton, A1999, „Don't Put All Your Eggs in One Basket: Diversification and Specialization in

Lending‟, University of Minnesota.

Yesegat, AW 2009, „Value added tax in Ethiopia: a study of operating costs & compliance‟, PhD

thesis, University of New South Wales

110
111
Appendices

Appendix –I: Tests for the aoutocorrelation: Breusch-Godfrey

Breusch-Godfrey Serial Correlation LM Test:

F-statistic 1.233575 Prob. F(2,83) 0.2965

Obs*R-squared 2.771198 Prob. Chi-Square(2) 0.2502

Test Equation:
Dependent Variable: RESID
Method: Least Squares
Date: 02/15/13 Time: 01:36
Sample: 1 96
Included observations: 96
Presample missing value lagged residuals set to zero.

Variable Coefficient Std. Error t-Statistic Prob.

C 0.031108 0.172057 0.180799 0.8570

CIR -0.001362 0.019511 -0.069784 0.9445
LG 0.006521 0.042203 0.154524 0.8776
ID -0.143472 0.733991 -0.195468 0.8455
ROA 0.012209 0.706521 0.017280 0.9863
SIZE -0.000790 0.006765 -0.116783 0.9073
STATEDUMMY -0.000949 0.014535 -0.065268 0.9481
GDP -0.005649 0.141251 -0.039992 0.9682
INFL 0.010455 0.050995 0.205014 0.8381
RIR -0.228160 0.730141 -0.312488 0.7555
EFEX 0.156564 0.341711 0.458178 0.6480
RESID(-1) 0.185050 0.117844 1.570303 0.1201
RESID(-2) -0.019007 0.113780 -0.167053 0.8677

R-squared 0.028867 Mean dependent var -2.23E-16

Adjusted R-squared -0.111538 S.D. dependent var 0.046298
S.E. of regression 0.048812 Akaike info criterion -3.076363
Sum squared resid 0.197755 Schwarz criterion -2.729108
Log likelihood 160.6654 Hannan-Quinn criter. -2.935997
F-statistic 0.205596 Durbin-Watson stat 1.903684
Prob(F-statistic) 0.997872

112
Appendix –II: Tests for the Heteroskedasticity Test: White
Heteroskedasticity Test: White

F-statistic 1.158071 Prob. F(61,34) 0.3264

Obs*R-squared 64.80804 Prob. Chi-Square(61) 0.3453
Scaled explained SS 75.79256 Prob. Chi-Square(61) 0.0962

Test Equation:
Dependent Variable: RESID^2
Method: Least Squares
Date: 02/15/13 Time: 01:13
Sample: 1 96
Included observations: 96
Collinear test regressors dropped from specification

Variable Coefficient Std. Error t-Statistic Prob.

C -0.081998 5.249821 -0.015619 0.9876

CIR -0.216974 0.177828 -1.220134 0.2308
CIR^2 -0.000902 0.012268 -0.073555 0.9418
CIR*LG 0.037154 0.050004 0.743011 0.4626
CIR*ID -0.463469 0.420205 -1.102960 0.2778
CIR*ROA -0.841825 0.599648 -1.403865 0.1694
CIR*SIZE 0.012972 0.007740 1.675916 0.1029
CIR*STATEDUM
MY 0.025340 0.011094 2.284094 0.0287
CIR*GDP 0.015130 0.091014 0.166234 0.8690
CIR*INFL -0.088356 0.044010 -2.007611 0.0527
CIR*RIR -0.091677 0.514953 -0.178030 0.8598
CIR*EFEX -0.386919 0.589408 -0.656453 0.5160
LG -0.321334 0.262406 -1.224569 0.2292
LG^2 0.016161 0.040317 0.400859 0.6910
LG*ID 1.640400 0.992069 1.653514 0.1074
LG*ROA 0.512312 1.349198 0.379716 0.7065
LG*SIZE 0.015387 0.009058 1.698639 0.0985
LG*STATEDUMM
Y 0.022076 0.018054 1.222756 0.2298
LG*GDP -0.206667 0.297696 -0.694219 0.4923
LG*INFL -0.009974 0.087654 -0.113782 0.9101
LG*RIR -1.182032 1.015904 -1.163527 0.2527
LG*EFEX 0.201281 0.429190 0.468977 0.6421
ID -0.505114 3.977695 -0.126987 0.8997
ID^2 -3.084123 11.25713 -0.273971 0.7858
ID*ROA -20.80727 16.89299 -1.231710 0.2265
ID*SIZE 0.005023 0.184332 0.027247 0.9784

113
ID*STATEDUMM
Y -0.661671 0.473897 -1.396234 0.1717
ID*GDP 3.678380 3.519159 1.045244 0.3033
ID*INFL 0.745566 1.224251 0.608997 0.5466
ID*RIR 5.659607 12.93265 0.437622 0.6644
ID*EFEX 8.122443 9.448506 0.859654 0.3960
ROA 2.107042 3.631732 0.580175 0.5656
ROA^2 6.357651 7.615921 0.834784 0.4097
ROA*SIZE -0.113633 0.164720 -0.689858 0.4950
ROA*STATEDUM
MY 0.299745 0.332729 0.900868 0.3740
ROA*GDP -2.440514 3.688351 -0.661682 0.5126
ROA*INFL -0.650741 1.270417 -0.512226 0.6118
ROA*RIR 9.376420 14.01040 0.669247 0.5079
ROA*EFEX -0.589308 13.96522 -0.042198 0.9666
SIZE -0.092798 0.081666 -1.136309 0.2638
SIZE^2 0.002348 0.001702 1.379269 0.1768
SIZE*STATEDUM
MY 0.001932 0.004547 0.424924 0.6736
SIZE*GDP -0.003828 0.036719 -0.104250 0.9176
SIZE*INFL -0.000849 0.010114 -0.083949 0.9336
SIZE*RIR -0.071241 0.150453 -0.473511 0.6389
SIZE*EFEX -0.099335 0.094669 -1.049283 0.3014
STATEDUMMY -0.012563 0.107042 -0.117363 0.9073
STATEDUMMY*G
DP 0.033020 0.123671 0.266996 0.7911
STATEDUMMY*I
NFL 0.025655 0.018345 1.398449 0.1710
STATEDUMMY*R
IR -0.188681 0.399440 -0.472363 0.6397
STATEDUMMY*E
FEX -0.168046 0.117000 -1.436296 0.1601
GDP -0.444518 2.977339 -0.149300 0.8822
GDP^2 -1.548889 5.194729 -0.298166 0.7674
GDP*INFL -2.693790 6.594780 -0.408473 0.6855
GDP*RIR -10.09996 33.91999 -0.297758 0.7677
GDP*EFEX 22.52461 25.94786 0.868072 0.3914
INFL 1.027623 3.903842 0.263234 0.7940
INFL^2 0.747729 2.043598 0.365888 0.7167
INFL*RIR -8.037512 40.85276 -0.196743 0.8452
INFL*EFEX -0.127917 8.168093 -0.015661 0.9876
RIR 20.90294 87.25894 0.239551 0.8121
RIR^2 -80.86015 385.0919 -0.209976 0.8349

R-squared 0.675084 Mean dependent var 0.002121

Adjusted R-squared 0.092146 S.D. dependent var 0.003683
S.E. of regression 0.003509 Akaike info criterion -8.213097

114
Sum squared resid 0.000419 Schwarz criterion -6.556955
Log likelihood 456.2286 Hannan-Quinn criter. -7.543657
F-statistic 1.158071 Durbin-Watson stat 2.178226
Prob(F-statistic) 0.326376

Appendix –III: First Step (General) Regression Results

Dependent Variable: NPL
Method: Panel Least Squares
Date: 02/15/13 Time: 05:14
Sample: 2000 2011
Periods included: 12
Cross-sections included: 8
Total panel (balanced) observations: 96

Variable Coefficient Std. Error t-Statistic Prob.

C 0.638789 0.106497 5.998165 0.0000

CIR -0.117419 0.022499 -5.218795 0.0000
LG -0.381990 0.048616 -7.857319 0.0000
ID 1.212022 0.850997 1.424237 0.1580
SIZE -0.012154 0.007644 -1.590032 0.1155
ROA -3.796408 0.827198 -4.589477 0.0000
STATEDUMMY 0.075765 0.016900 4.483234 0.0000
GDP -0.505052 0.165705 -3.047894 0.0031
INFL -0.226959 0.058767 -3.862012 0.0002
EFEX -1.368306 0.386677 -3.538631 0.0007
RIR 0.000931 0.008417 0.110629 0.9122

R-squared 0.789352 Mean dependent var 0.133325

Adjusted R-squared 0.764570 S.D. dependent var 0.118078
S.E. of regression 0.057293 Akaike info criterion -2.773812
Sum squared resid 0.279010 Schwarz criterion -2.479980
Log likelihood 144.1430 Hannan-Quinn criter. -2.655041
F-statistic 31.85167 Durbin-Watson stat 1.422553
Prob(F-statistic) 0.000000

115
 Appendix –IV: second step regression results
Dependent Variable: NPL
Method: Panel Least Squares
Date: 05/24/13 Time: 19:02
Sample: 2000 2011
Periods included: 12
Cross-sections included: 8
Total panel (balanced) observations: 96

Variable Coefficient Std. Error t-Statistic Prob.

C 0.275838 0.098027 2.813900 0.0061

LG -0.293679 0.036521 -8.041285 0.0000
ROA -2.523769 0.598648 -4.215785 0.0001
CIR -0.025478 0.015142 -1.682544 0.0960
SIZE 0.009189 0.004528 2.029289 0.0455
STATEDUMMY 0.073909 0.012794 5.776680 0.0000
GDP -0.445923 0.134830 -3.307285 0.0014
INFL -0.234712 0.047214 -4.971184 0.0000
EFEX -1.434699 0.248540 -5.772513 0.0000

R-squared 0.835550 Mean dependent var 0.133403

Adjusted R-squared 0.820429 S.D. dependent var 0.110131
S.E. of regression 0.046669 Akaike info criterion -3.202418
Sum squared resid 0.189485 Schwarz criterion -2.962010
Log likelihood 162.7161 Hannan-Quinn criter. -3.105241
F-statistic 55.25469 Durbin-Watson stat 1.365317
Prob(F-statistic) 0.000000

Appendix- V: Ratio Data

year bank NPL LG CIR ID size ROA INFL GDP EFEX RIR

2000 AIB 0.30270 0.154362 0.62963 0.018445 20.447512 0.0303 0.054 0.059 0.0831 0.12
-
2001 AIB 0.20500 0.203209 1 0.020948 20.625653 0.019846 0.003 0.074 0.0879 0.1275
-
2002 AIB 0.34020 0.119309 1.166667 0.017986 20.829426 0.017086 0.106 0.016 0.0868 0.1075
-
2003 AIB 0.25130 0.20375 0.785714 0.030692 21.060452 0.012848 0.109 0.021 0.087091 0.105

2004 AIB 0.18390 0.154334 0.740741 0.031638 21.294245 0.019774 0.073 0.117 0.086751 0.105

2005 AIB 0.12020 0.266667 0.796875 0.024708 21.523472 0.024708 0.061 0.126 0.08711 0.105

2006 AIB 0.09560 0.310897 0.621622 0.030806 21.806426 0.037576 0.106 0.115 0.090258 0.105

116
2007 AIB 0.07360 0.254777 0.377451 0.033681 22.066131 0.053264 0.158 0.118 0.08957 0.105

2008 AIB 0.08660 0.082542 0.553922 0.035685 22.296040 0.042324 0.253 0.112 0.095569 0.115

2009 AIB 0.05780 -0.00921 0.545455 0.031294 22.583151 0.031449 0.364 0.099 0.118102 0.1225

2010 AIB 0.05470 0.137635 0.441734 0.048206 22.795809 0.044179 0.028 0.104 0.136806 0.1225

2011 AIB 0.14002 0.210829 0.407255 0.052689 23.037384 0.049921 0.181 0.114 0.165292 0.1188

2000 BOA 0.04210 0.509579 0.76 0.029248 20.391980 0.02925 0.054 0.059 0.0831 0.12
-
2001 BOA 0.03200 0.240175 0.485714 0.018973 20.613451 0.039063 0.003 0.074 0.0879 0.1275
-
2002 BOA 0.37950 -0.02691 0.642857 0.013135 20.856047 0.007005 0.106 0.016 0.0868 0.1075
-
2003 BOA 0.28430 0.173053 0.6875 0.014254 21.010698 0.006002 0.109 0.021 0.087091 0.105

2004 BOA 0.17510 0.159044 0.446154 0.016404 21.183850 0.034069 0.073 0.117 0.086751 0.105

2005 BOA 0.12400 0.220421 0.45122 0.022849 21.444514 0.039864 0.061 0.126 0.08711 0.105

2006 BOA 0.04940 0.37137 0.420635 0.019407 21.764955 0.043049 0.106 0.115 0.090258 0.105

2007 BOA 0.10540 0.148373 0.457746 0.01914 21.945864 0.027974 0.158 0.118 0.08957 0.105

2008 BOA 0.12870 0.181797 0.542628 0.022253 22.174866 0.005131 0.253 0.112 0.095569 0.115

2009 BOA 0.14750 -0.03994 0.737904 0.02354 22.423754 0.027574 0.364 0.099 0.118102 0.1225

2010 BOA 0.06980 0.140831 0.739796 0.032962 22.560636 0.03121 0.028 0.104 0.136806 0.1225

2011 BOA 0.03970 0.049066 0.639013 0.0338 22.708122 0.035449 0.181 0.114 0.165292 0.1188

2000 CBB 0.23560 0.004902 2.222222 0.010267 20.696922 0.00308 0.054 0.059 0.0831 0.12
-
2001 CBB 0.344391 -0.00617 1.133333 0.009298 20.690743 0.007231 0.003 0.074 0.0879 0.1275
-
2002 CBB 0.41550 -0.12483 2 0.012526 20.680358 0.008351 0.106 0.016 0.0868 0.1075
-
2003 CBB 0.40090 -0.0198 0.772727 0.016985 20.663516 0.014862 0.109 0.021 0.087091 0.105

2004 CBB 0.35470 0.035471 1.391304 0.033113 20.778701 0.006623 0.073 0.117 0.086751 0.105

2005 CBB 0.27760 0.144691 0.913043 0.034389 21.328674 0.014192 0.061 0.126 0.08711 0.105

2006 CBB 0.19420 0.273113 0.288462 0.037841 21.309384 0.043962 0.106 0.115 0.090258 0.105

2007 CBB 0.17060 0.112867 0.404412 0.05135 21.359313 0.04288 0.158 0.118 0.08957 0.105

2008 CBB 0.15560 0.041817 0.422764 0.034699 21.595396 0.048077 0.253 0.112 0.095569 0.115

117
2009 CBB 0.11450 0.118245 0.551724 0.035494 21.675696 0.040895 0.364 0.099 0.118102 0.1225

2010 CBB 0.06560 0.100629 0.564885 0.033207 21.874471 0.041429 0.028 0.104 0.136806 0.1225

2011 CBB 0.17720 -0.13131 0.796748 0.036805 21.977456 0.035093 0.181 0.114 0.165292 0.1188

2000 CBE 0.26440 0.123707 0.292011 0.016139 23.710361 0.03127 0.054 0.059 0.0831 0.12
-
2001 CBE 0.33770 0.02918 0.302425 0.016474 23.790807 0.009912 0.003 0.074 0.0879 0.1275
-
2002 CBE 0.419734 -0.09302 0.431217 0.018875 23.820923 0.022894 0.106 0.016 0.0868 0.1075
-
2003 CBE 0.400023 -0.14007 0.315327 0.02595 23.909618 0.029587 0.109 0.021 0.087091 0.105

2004 CBE 0.37680 -0.02739 0.430615 0.021019 24.054577 0.017408 0.073 0.117 0.086751 0.105

2005 CBE 0.27520 0.12882 0.387833 0.02231 24.224882 0.023787 0.061 0.126 0.08711 0.105

2006 CBE 0.22450 -0.02797 0.333929 0.027086 24.302582 0.031242 0.106 0.115 0.090258 0.105

2007 CBE 0.14520 0.047443 0.625641 0.028005 24.495015 0.026924 0.158 0.118 0.08957 0.105

2008 CBE 0.05330 0.437165 0.305139 0.028384 24.643574 0.037052 0.253 0.112 0.095569 0.115

2009 CBE 0.03660 0.170621 0.190722 0.02508 24.807745 0.045715 0.364 0.099 0.118102 0.1225

2010 CBE 0.14842 0.129533 0.33357 0.023603 25.029855 0.03785 0.028 0.104 0.136806 0.1225

2011 CBE 0.00860 0.315735 0.236196 0.025493 25.461786 0.037089 0.181 0.114 0.165292 0.1188

2000 DB 0.15950 0.435272 1.136364 0.021965 20.578240 0.02081 0.054 0.059 0.0831 0.12
-
2001 DB 0.10920 0.253501 1.111111 0.033636 20.818576 0.032727 0.003 0.074 0.0879 0.1275
-
2002 DB 0.14220 0.181193 0.837209 0.025572 21.119354 0.026245 0.106 0.016 0.0868 0.1075
-
2003 DB 0.08890 0.31176 0.733333 0.026118 21.411903 0.018584 0.109 0.021 0.087091 0.105

2004 DB 0.07440 0.250296 0.606383 0.028016 21.707963 0.029137 0.073 0.117 0.086751 0.105

2005 DB 0.06720 0.242832 0.669725 0.021053 21.952906 0.028363 0.061 0.126 0.08711 0.105

2006 DB 0.06210 0.294564 0.507538 0.027497 22.237514 0.040695 0.106 0.115 0.090258 0.105

2007 DB 0.05950 0.20662 0.473684 0.027313 22.521835 0.042708 0.158 0.118 0.08957 0.105

2008 DB 0.05890 0.089913 0.446023 0.031933 22.781101 0.042534 0.253 0.112 0.095569 0.115

2009 DB 0.07390 0.015723 0.570621 0.032981 22.998788 0.036166 0.364 0.099 0.118102 0.1225

2010 DB 0.03000 0.118241 0.526652 0.039019 23.237165 0.037076 0.028 0.104 0.136806 0.1225

118
2011 DB 0.03380 0.187942 0.479134 0.046317 23.408389 0.042974 0.181 0.114 0.165292 0.1188

2000 NIB 0.01900 0 3 0.012658 18.878106 0.00633 0.054 0.059 0.0831 0.12
-
2001 NIB 0.01900 0.719048 0.421053 0.032828 19.796925 0.045455 0.003 0.074 0.0879 0.1275
-
2002 NIB 0.08640 0.351852 0.4 0.029963 20.095906 0.041199 0.106 0.016 0.0868 0.1075
-
2003 NIB 0.12340 0.410909 0.410256 0.032768 20.601098 0.021469 0.109 0.021 0.087091 0.105

2004 NIB 0.08770 0.300254 0.368421 0.029671 20.944007 0.039294 0.073 0.117 0.086751 0.105

2005 NIB 0.11220 0.306267 0.321429 0.030023 21.272543 0.038106 0.061 0.126 0.08711 0.105

2006 NIB 0.08470 0.231864 0.422222 0.02664 21.429823 0.039961 0.106 0.115 0.090258 0.105

2007 NIB 0.05560 0.188222 0.495495 0.023399 21.681466 0.04066 0.158 0.118 0.08957 0.105

2008 NIB 0.06730 0.140492 0.432584 0.029315 22.017993 0.043562 0.253 0.112 0.095569 0.115

2009 NIB 0.11160 0.047748 0.450413 0.035781 22.293339 0.045559 0.364 0.099 0.118102 0.1225

2010 NIB 0.07370 0.128044 0.468553 0.048568 22.510180 0.047731 0.028 0.104 0.136806 0.1225

2011 NIB 0.05040 0.07987 0.495822 0.045557 22.685049 0.048369 0.181 0.114 0.165292 0.1188

2000 UB 0.07950 0.579545 1 0.041958 18.778355 0.03497 0.054 0.059 0.0831 0.12
-
2001 UB 0.07750 0.343284 1 0.042056 19.181487 0.037383 0.003 0.074 0.0879 0.1275
-
2002 UB 0.15950 0.177914 1.25 0.022293 19.564904 0.022293 0.106 0.016 0.0868 0.1075
-
2003 UB 0.09930 0.437931 1.083333 0.025586 19.966113 0.014925 0.109 0.021 0.087091 0.105

2004 UB 0.09900 0.244792 0.842105 0.02819 20.328741 0.014837 0.073 0.117 0.086751 0.105

2005 UB 0.08450 0.352445 0.48 0.041938 20.793724 0.040075 0.061 0.126 0.08711 0.105

2006 UB 0.04180 0.409363 0.469697 0.034396 21.192644 0.037523 0.106 0.115 0.090258 0.105

2007 UB 0.04590 0.287943 0.52 0.032066 21.503966 0.039853 0.158 0.118 0.08957 0.105

2008 UB 0.03980 0.241935 0.55 0.033538 21.901921 0.038769 0.253 0.112 0.095569 0.115

2009 UB 0.04620 0.135688 0.708609 0.02902 22.260563 0.028805 0.364 0.099 0.118102 0.1225

2010 UB 0.03760 0.176741 0.469314 0.044098 22.497540 0.042062 0.028 0.104 0.136806 0.1225

2011 UB 0.03350 0.202319 0.472727 0.037799 22.767728 0.041812 0.181 0.114 0.165292 0.1188

2000 WB 0.19100 0.251908 1.818182 0.036965 20.057734 0.01362 0.054 0.059 0.0831 0.12

119
 -
2001 WB 0.13660 0.238372 0.954545 0.037736 20.183698 0.024014 0.003 0.074 0.0879 0.1275
-
2002 WB 0.12940 0.152709 1.470588 0.03096 20.286310 0.018576 0.106 0.016 0.0868 0.1075
-
2003 WB 0.10860 0.288967 1.173913 0.028121 20.605608 0.016873 0.109 0.021 0.087091 0.105

2004 WB 0.12240 0.226287 0.559322 0.038596 20.854294 0.039474 0.073 0.117 0.086751 0.105

2005 WB 0.08410 0.263473 0.560976 0.043317 21.203220 0.038985 0.061 0.126 0.08711 0.105

2006 WB 0.04850 0.370998 0.504065 0.044267 21.538188 0.041611 0.106 0.115 0.090258 0.105

2007 WB 0.05250 0.260789 0.432432 0.038793 21.970298 0.043966 0.158 0.118 0.08957 0.105

2008 WB 0.08390 0.081807 0.409091 0.046545 22.140332 0.046061 0.253 0.112 0.095569 0.115

2009 WB 0.07700 -0.11127 0.501931 0.046698 22.356030 0.05002 0.364 0.099 0.118102 0.1225

2010 WB 0.03470 0.146322 0.498471 0.055381 22.471073 0.055381 0.028 0.104 0.136806 0.1225

2011 WB 0.03510 0.149828 0.434706 0.062027 22.810303 0.056817 0.181 0.114 0.165292 0.1188

120