CONCENTRATION AND COMPETITION IN ETHIOPIAN BANKING

INDUSTRY(A PANEL DATA ANALYSIS)

By

GEBEYEHU, Hailemariam Nega

THESIS

Submitted to

KDI School of Public Policy and Management

In Partial Fulfillment of the Requirements

For the Degree of

MASTER OF DEVELOPMENT POLICY

2018
CONCENTRATION AND COMPETITION IN ETHIOPIAN BANKING
INDUSTRY(A PANEL DATA ANALYSIS)

By

GEBEYEHU, Hailemariam Nega

THESIS

Submitted to

KDI School of Public Policy and Management

In Partial Fulfillment of the Requirements

For the Degree of

MASTER OF DEVELOPMENT POLICY

2018
Professor Shun WANG
CONCENTRATION AND COMPETITION IN ETHIOPIAN BANKING
INDUSTRY(A PANEL DATA ANALYSIS)

By

GEBEYEHU, Hailemariam Nega

THESIS

Submitted to

KDI School of Public Policy and Management

In Partial Fulfillment of the Requirements

For the Degree of

MASTER OF DEVELOPMENT POLICY

Committee in charge:

Professor Shun WANG, Supervisor

Professor Seulki CHOI

Professor Yooncheong CHO

Approval as of December, 2018

i
 TABLE OF CONTENTS
LIST OF TABLES ....................................................................................................................................... iv
LIST OF FIGURES ..................................................................................................................................... iv
ABSTRACT.................................................................................................................................................. v
CHAPTER ONE ........................................................................................................................................... 1
1. INTRODUCTION ............................................................................................................................ 1
1.1. BACKGROUND OF THE STUDY ................................................................................................. 1
1.2. STATEMENT OF THE PROBLEM ................................................................................................ 3
1.3. OBJECTIVE OF THE STUDY ........................................................................................................ 6
1.3.1. GENERAL OBJECTIVE ................................................................................................................ 6
1.3.2. SPECIFIC OBJECTIVES ............................................................................................................... 6
1.4. SIGNIFICANCE OF THE STUDY .................................................................................................. 6
1.5. ASSUMPTIONS ............................................................................................................................... 6
1.6. SCOPE OF THE STUDY ................................................................................................................. 7
1.7. LIMITATION OF THE STUDY ...................................................................................................... 7
1.8. ORGANIZATION OF THE STUDY ............................................................................................... 7
CHAPTER TWO .......................................................................................................................................... 8
2. SURVEY OF LITERATURE ................................................................................................................... 8
2.1. THEORETICAL SURVEY ............................................................................................................... 8
2.2. EMPIRICAL LITERATURE .......................................................................................................... 12
CHAPTER THREE .................................................................................................................................... 15
3. RESEARCH METHODOLOGY ............................................................................................................ 15
3.1. DATA TYPE AND SOURCE ......................................................................................................... 15
3.2. METHOD OF DATA ANALYSIS.................................................................................................. 15
CHAPTER FOUR ....................................................................................................................................... 24
4. DATA ANALYSIS AND PRESENTATION ........................................................................................ 24
4.1. DESCRIPTIVE ANALYSIS ........................................................................................................... 24
4.1.1. OVERVIEW OF ETHIOPIAN BANKING INDUSTRY ............................................................ 24
4.1.2. CONCENTRATION IN ETHIOPIAN BANKING INDUSTRY ................................................. 26
4.1.3. COMMERCIAL BANKS INCOME COMPOSITION IN ETHIOPIA ....................................... 26
4.2. ECONOMETRIC ANALYSIS ........................................................................................................ 29
4.2.1. UNIT ROOT TEST....................................................................................................................... 29
4.2.2. RESULT AND DISCUSSION ..................................................................................................... 29

ii
 4.2.2.1. Competition and Contestability Test (Model A)........................................................................ 29
4.2.2.2. Interest Vs Non- Interest Income (Substitute or complement products)? (Model B) ................ 31
4.2.2.3. Mis-specification of Empirical P-R Model test ......................................................................... 31
4.2.2.4. Competition among Private Banks (Model D)........................................................................... 32
4.2.2.5. How Non- Performing loan affects Income of Banks? (Model E) ............................................ 32
4.2.2.6. The Effect of NBE policy on Banks Income Statement ............................................................ 35
CHAPTER FIVE ........................................................................................................................................ 36
5. CONCLUSION AND RECOMMENDATION ...................................................................................... 36
5.1. CONCLUSION ................................................................................................................................ 36
5.2. RECOMMENDATION ................................................................................................................... 37
REFERENCES ........................................................................................................................................... 38
APPENDIXES ............................................................................................................................................ 42

iii
LIST OF TABLES
Table 1: Five Banks Asset and Deposit concentration in Ethiopia ............................................................... 4
Table 2: Interpretation of the Panzar-Rosse H-Statistic.............................................................................. 10
Table 3: Ratio of selected financial performance indicators....................................................................... 25
Table 4: Top Three banks Concentration ratio ........................................................................................... 26
Table 5: Share of Non-interest income in Ethiopian Banking industry ...................................................... 28
Table 6: Summary of H - Statistics from Model A ..................................................................................... 30
Table 7: Coefficient of Share of Non-interest Income (SNII) in Model C ................................................. 31
Table 8: Summary result of H- statistics that used to test mis-specification of P-R model ........................ 32
Table 9: Summary of competition result among private banks and comparison with Model A................. 32
Table 10: The effect of Credit Cap and NBE Bill purchase on Banks Income statement .......................... 35

LIST OF FIGURES
Figure 1: Five Banks ACR & DCR in Ethiopia: Trend ................................................................................ 5
Figure 2: Trend in Share of Non-interest income ....................................................................................... 28
Figure 3: The effect of Non-performing Loan on Interest, Non- Interest and Total Income ...................... 34

iv
 ABSTRACT
The paper aims to investigate the degree of concentration and competition (market contestability) in the
Ethiopian commercial banking industry using the Panzar-Rosse (P-R) Model. It also aims to empirically
test for misspecification of the P-R model. The impact of different control variables on income types and
thereby on competition is also examined. Finally, the paper attempts to formulate a testable hypothesis
whether different income types are substitutes or compliments.

All commercial banks operating in Ethiopia are considered for the analysis. Panel Random Effect
Bootstrap Estimation is employed using quarterly data from 2002/03 (QIII) to 2014/15 (QIV). A total of
16 econometric panel models are estimated. In calculating prices of labor, physical capital and funds the
most widely recommended proxies are used in the course of estimation.

The results indicated that though the industry was highly contestable in early periods of the analysis, it
still remains highly concentrated. The competition appears to be higher among private banks, though it
endogenously comes from the existence of Commercial Bank of Ethiopia.

Following Bikker et al (2006), we tested for misspecification of the empirical P-R Model and proved that
the use of dependent variables scaled by asset sizes is biased towards competitive market conditions.
Credit ceilings are found to have negative and significant influence on income statements of banks and
then on competition. We could not find significant evidence about the impact of NBE bills on income and
competition.

Finally, we have formulated a testable hypothesis in identifying whether the relationship between income
components is a substitute or compliment type. As far as the information we have, this is the first
formulation in the analysis of income statement of banks from microeconomic theory perspective. Our
results indicated that interest and non-interest components of income are non- compliments in the
Ethiopian commercial banking industry.

Based on our findings we would like to forward recommendations to policy makers and researchers. For
policy makers, as our findings showed a non-favorable outcome of credit ceilings on income and
competition, we recommend as much as possible such policies must be taken as last option and for a very
short period. For researchers, we recommend the use of unscaled dependent variables in identifying
market structure of the banking industry. Moreover, we would like to encourage researchers to test our
newly formulated hypothesis in the banking industry so that we can get feedback.

Keywords: competition, Panzar-Rosse, Commercial Banking, Panel GLS, Ethiopia

v
 CHAPTER ONE

1. INTRODUCTION

1.1. BACKGROUND OF THE STUDY

It has been argued that banks play an important role in the process of economic growth and
development. Traditionally, banks transfer money from surplus groups (savers) to deficit
groups (borrowers). In this process, savers can earn an interest/income for their deposit and
borrowers get money to start/expand their business or buy anything they want. The banking
sector’s role as a source of finance for productive investment and an accelerator of economic
growth has been stressed by different economists. For instance, Pagano (1993) suggested
three ways in which the banking sector might affect economic growth. First, it can increase
the productivity of investments. Second, an efficient financial sector reduces transaction cost
and thus increase the share of saving channeled into productive investments. Third, financial
sector development can promote savings. Moreover, Schumpeter (1934) argues that in a
well-functioning financial system entrepreneurs will have a better opportunity to fund their
new ideas and hence encourage technological innovation. To reap these benefits from the
banking system the structure, competitiveness and efficiency of the system is vital.

Like in other sectors, competition in the banking sector has different strands. Proponents of
competition argue that the degree of competition in the financial sector can matter for the
efficiency of production of financial services and for the quality of financial products and the
degree of innovation in the sector (Claessens etal, 2004:563).Moreover, in the absence of
competition, banks with highest market power will earn more rents by charging higher
interest rates on business loans and this might lead to financial instability as higher interest
rates may increase the riskiness of loan portfolios (Allen etal, 2008:5).On the other hand,
excessive bank competition has proven to be a bad omen to financial stability as witnessed in
the 1997 East Asian financial crisis. In support of this, Keeley (1990) documented that
increased competition in the 1980s eroded monopoly rents and led to an increase in bank
failures in the United States. Xavier(2001) argued that regulators have traditionally tried to

1
 restrict competition in banking sector with the aim of avoiding excessive risk-taking. Though
there are different arguments regarding bank competition, it is believed that normal
competition is desirable. Normal in the sense that the competition will not have a negative
effect on the workings of an economy.
Knowing the fact that competition is necessary, regulatory bodies are very concerned about
the degree of concentration in the banking sector. For instance, if there is high concentration
in the banking industry, depositors would have less return/interest income/for their deposits
and borrowers would be charged high lending interest rates. Hence, this will affect the
economy negatively by making the cost of funds higher for investors resulting in higher
output prices for buyers. Past empirical studies on the relationship between bank
concentration and monopoly power did not reach with similar argument. In Traditional
Industrial Organization theory, by looking at the market structure (concentration index) of
the industry we can easily infer about the competitiveness nature (conduct) of the industry. In
this case, the market with high concentration is characterized by absence of competition and
vice versa. But, this paradigm misses the competitive outcome that might come with a
concentrated market. In the New Industrial Organization (NEIO) theory, there may be a
competitive outcome even in a highly concentrated market if the market is contestable. A
market is contestable; if there is a relatively free entry and exit in the industry, low sunk cost
etc. (see Baumol etal, 1982 for more).Therefore, policy toward bank consolidation cannot
rely solely on structural measures (Shaffer, 1994).

Different scholars try to measure the degree of banking competition by using alternative
methods of measurement. Generally, there are two types of methods that help to measure
competition in banking industry: structural and non – structural methods (the detail is
presented in the literature part).

2
 1.2. STATEMENT OF THE PROBLEM

In Ethiopia, the financial sector constitute of the banking, microfinance and insurance companies
of which the banking sector plays the dominant role. According to the National Bank of Ethiopia
(2015) report, banking sector takes 92.8 percent of asset, 76.8 percent of capital and 57.7 percent
of total branch network from the financial sector. Currently, there are 18 commercial banks (2
public and 16 private) and 1 specialized bank (Development Bank of Ethiopia) operating in the
country under one regulatory body, the National Bank of Ethiopia (NBE, 2014). Development
Bank of Ethiopia (DBE) is the specialized bank because unlike other commercial banks, DBE
doesn’t mobilize deposits and its source of funding is long term government bonds. Hence, the
bank does not compete with commercial banks in mobilizing deposits. In addition, the bank only
provides long term credit to priority projects and is not profit oriented.

In the last decade, the National Bank of Ethiopia undertook different policy measures that have a
direct/indirect impact on bank competitiveness. These include, credit cap (2009), revision of
minimum paid up capital (2011) and National Bank of Ethiopia (NBE) Bill purchase (2011).The
credit cap policy which was introduced in 2009 limited the outstanding credit of commercial
banks. The policy was only applied to private banks as public bank lending was towards priority
sectors and policy makers believed that limiting their credit would have a higher negative impact
on the economy. The policy came into effect in order to contain the rampant inflation at the time.
The credit cap policy was lifted in 2011 but followed by another policy called NBE Bill purchase.
The NBE bill purchase policy function in such a way that when private banks disburse a loan to
the economy, they should buy the NBE-bill with the 27 percent of their total planned
disbursement. The NBE bill purchase policy believed to have an effect on competition in the
industry as commercial banks would compete more for collecting deposit in order to fulfill their
NBE-bill obligation. In the same period, the minimum paid-up capital requirement to establish a
bank was changed from Ethiopian Birr (ETB) 70 million (USD 4.3 million) to 500 million (USD
25 million). NBE introduced this in order to strengthen the stability of domestic commercial
banks. All the above polices might have an effect on the level of competition in Ethiopian
banking industry but have not been studied so far.

3
In addition, the banking industry in Ethiopia can be characterized by the existence of high
concentration even after the sector was opened to private sector in 1991. Concentration in
financial sector can be measured by looking at n-banks Asset Concentration Ratio (ACR) and n-
banks Deposit Concentration Ratio (DCR). If we look at the share of assets held (5 banks ACR)
and deposit mobilized (5 banks DCR)by the five Ethiopian largest banks, it is 84.4and 83.8
percent, respectively, at the end of June 2014(ibid).Though concentration levels have come down
recently, they still exceeds the world average(see Table 1) 1 . By looking at the level of
concentration in the sector, some argue that the banks in Ethiopia are operating in the absence of
competition. However, even in the presence of market concentration, there might exist
competitive outcomes (see Jackson, 1992). In the case of Ethiopia, there is no empirical study
that relates concentration and competitiveness. In other words, there is no study about the
contestability of the banking sector. Hence, the level and trend of bank concentration and
competition in Ethiopia must be analyzed comprehensively. In this regard, over time
concentration in the industry will be related with its competitiveness in order to infer about
contestability of the banking industry.

Table 1: Five Banks Asset and Deposit concentration in Ethiopia

Particulars Asset Deposit World (ACR)

2007 90.8 90.5 85.6
2008 89.4 88.4 85.2
2009 89.2 86.3 84.0
2010 86.0 84.2 83.5
2011 86.9 85.0 81.1
2012 88.0 85.3 NA
2013 85.7 84.6 NA
2014 84.4 83.8 NA
Source: Own computation based on NBE & GFDD data base

1
It would be very nice if the comparison in bank concentration is done between Ethiopia and other developing
countries. But, the researcher cannot find a data for those countries after the year 2011, the same for world average.

4
Figure 1: Five Banks ACR & DCR in Ethiopia: Trend

Fig 1: 5 Banks ACR & DCR in Ethiopia: Trend

92.0
90.0
88.0
In percent

86.0
84.0
Asset
82.0
Deposit
80.0
78.0 World (ACR)
76.0
2007 2008 2009 2010 2011 2012 2013 2014
Year

Furthermore, in the Growth and Transformation Plan (GTP) II period (2015/16-2019/20), the
Ethiopian government believes that mega projects like establishment of sugar corporations, rail
way construction, and hydroelectric dams etc. will speed up economic growth and development
of the country. But, these projects will not come into effect if there is inadequate funding. Hence,
banks in Ethiopia are expected to efficiently mobilize savings /deposits from the economy in
order to avail enough credit for projects. This clearly shows us the importance of the sector. In
this process, the level of competition in the banking sector matters and hence knowing the level
of banking sector competitiveness is of paramount importance for future policy measures.

Finally, although the nature of concentration and competition in the banking industry is deeply
studied in developed economies, a limited number of studies have been done in developing
countries. Moreover, to the best of researcher knowledge, there is only one published research
work in the case of Ethiopia. But, the study used a time period from 2000 until 2007 and also
concluded the industry is incontestable without conducting any tests. Moreover, the study only
estimates the static competition level and did not consider the dynamics. Hence, the current study
uncovers the static and dynamic nature of concentration and competition in the Ethiopian
banking industry by including the recent time period. In the same fashion competition in private
banks is analyzed. Moreover, the effect of different policy and control variables on banking
competition is ascertained.

5
 1.3. OBJECTIVE OF THE STUDY

1.3.1. GENERAL OBJECTIVE

Drawing from the aforementioned problems, the research mainly examines the static and
dynamic change in concentration and competition (market structure) in the Ethiopian banking
industry.

1.3.2. SPECIFIC OBJECTIVES

More specifically the research identifies the extent of market contestability in the Ethiopian
banking industry. Moreover, the level of competition among private banks, by excluding public
banks from the sample, and impact of policy and other control variables on competition in the
banking industry is investigated in this study. The study also tests the Bikker etal (2006)
hypothesis which says using scaled dependent variable is biased towards competitive market
structure. Finally, the study formulates a testable hypothesis regarding the relationship between
traditional and non-traditional banking activity

1.4. SIGNIFICANCE OF THE STUDY

The study help for many purposes concerning about competitiveness in Ethiopian commercial
banking industry. More specifically, it provides helpful information for; the regulatory body
(National Bank of Ethiopia), investors, both who are already engaged in and planning to join the
sector in Ethiopia, bank managers. Moreover, by filling the information gap that prevailed in the
country regarding commercial banks competition level it can be used by other researchers who
intend to conduct research on the same or related areas.

1.5. ASSUMPTIONS
For conducting this research manageably assumptions needs to be considered. The research
assumed; Banks are multi-product firms, the objective of a bank is profit maximization or cost
minimization and there exists free entry into and exist from the banking industry

6
 1.6. SCOPE OF THE STUDY
The overarching aim of the research work is to find out the concentration and competition of
commercial banks operating in Ethiopia. To achieve this objective 19 commercial banks
operating in the country are included in the study. And the time span cover from FY 2002/03 to
2014/15 and the frequency of data is on quarterly basis.

1.7. LIMITATION OF THE STUDY

The study faces lack of time in relation with data and financial constraints. Given the limitation
the researcher exerts the maximum effort to meet the research objectives.

1.8. ORGANIZATION OF THE STUDY

The research paper is organized in five chapters. The first chapter includes the introduction parts
where the background, statement of the problem, objective, significance, assumptions and scope
of the study is covered. In the second chapter both theoretical and empirical literatures reviewed,
in deep, which is related to the problem under study. Following the literature review section, the
methodology and data discussion appears in chapter three. Under the methodology: data type,
source, model specification and estimation procedure are discussed in detail. Chapter four
presents the finding of the study analyzed using both descriptive and econometric analysis. In the
last chapter conclusion and policy implications are presented based on the finding of the study.

7
 CHAPTER TWO

2. SURVEY OF LITERATURE

2.1. THEORETICAL SURVEY

This part presents the theoretical review of literatures regarding definition and alternative
measurements of banking sector competition and concentration. Moreover, it reviews the
different theories about the nexus between bank competition/concentration and bank fragility
/stability.

Definition

Competition can be defined as the state of rivalry among suppliers of a product. Depend on the
level of competition any industry/market is classified as perfectly competitive, monopolistically
competitive, oligopoly or monopoly.

2.1.1. MEASUREMENT OF COMPETITION IN BANKING

Banking sector competition can be measured based on two different methods: structural
(Traditional IO, market structure is exogenous) and non-structural (NEIO, market structure is
endogenous). In the first case, structure-conduct-performance (SCP) and Efficient Structure
Hypothesis (ESH) are the dominant theories in measuring the level of bank competition. In the
SCP approach, the degree of competition is easily inferred from the level of concentration in the
industry, in which competition is minimal in highly concentrated market and vice versa.

Due to lack of empirical result that fully support traditional IO theoretical framework (eg:
contestable market), the New Industrial Organization (NEIO) theory forward different method
for measuring the level of competition in an industry by making market structure an endogenous
variable which will be determined by a model. In the NEIO theory market structure cannot be
determined by simply looking at the number of firms or level of concentration in the industry.
The Panzar-Rosse H-statistics (Panzar and Rosse, 1987) and the Bresnahan Index (Bresnahan,
1982) are the two prominent methods that are used to empirically test bank competition based on
the NEIO theory. Both methods assume that banks are profit maximizer and market structure of

8
an industry is endogenously determined from profit maximization condition. The PR method is
highly applied for measuring competition in banking sector.

The Panzar _ Rosse Test (H – Statistic)

Panzar and Rosse [1977, 1987] developed a test for competitive market conditions based on a
reduced-form revenue equation of the firms. The test was based on empirical observation of the
impact of variations in factor input prices on firm-level revenues.

This test was derived from a general banking market model 2, which determines the equilibrium
output and the equilibrium number of banks, by maximizing profits at both the bank level and
the industry level. This implies, first, that bank i maximizes its profits, where marginal revenue
equals marginal cost [Bikker, 2004).

𝑅𝑖′ (𝑦𝑖 , 𝑛, 𝑧) = 𝐶𝑖′ (𝑦𝑖 , 𝐼𝐼𝑖 , 𝑡𝑖 ) ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ (1)

𝑅𝑖∗ (𝑦 ∗ , 𝑛∗ , 𝑧) = 𝐶𝑖∗ (𝑦 ∗ 𝐼𝐼𝑖 , 𝑡𝑖 ) ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ (2)

Market power is measured by the extent to which a change in factor input prices (∂IP i ) is
reflected in equilibrium revenues (∂R*) earned by bank i. In order to identify the nature of the
market structure (monopoly or oligopoly, monopoly competition or perfect competition) the
Panzar and Rosse model (P-R) provides a measure called the H- statistic. They show that sum of
elasticity of the total revenues, with respect to changes in banks input prices, allows inference
about the banks competitive conduct.

𝑚
𝜕𝜕𝑖∗ 𝐼𝐼𝑖𝑖
𝐻 = �� � … … … … … … … … … … … … … … … … … … … . (3)
𝜕𝜕𝜕𝑖𝑖 𝑅𝑖∗
𝑡=1

Where: where: R i – revenue function of bank i,

C i – cost of bank i,

yi – output of bank i,

n – Number of banks,

2
Cournot oligopoly model with profit maximization by collusive Cournot oligopolies

9
IP i – factor of input prices of bank i,

m – Number of factor input prices,

z i – vector of exogenous variables that shift the revenue function,

ti – vector of exogenous variables that shift the cost function.

The estimated value of the H-statistic ranges between -∞ and 1. Moreover, Panzar and Rosse
(1987) have shown that in market equilibrium, perfect competition is indicated by the H-statistic
equal to unity. Due to the fact that, under perfect competition an increase in input prices and thus
in average costs should lead to a proportional price increase and (at the firm level) to a
proportional rise in revenues. Under monopoly, an increase in input prices will increase marginal
costs, reduce equilibrium output and consequently reduce total revenues and the H-statistic is
negative or equal to zero. If the market structure is characterized by monopolistic competition,
the H-statistics will lie between zero and unity (see Table 2).

Table 2: Interpretation of the Panzar-Rosse H-Statistic

Values of H-Statistic Competitive Environment

H<= 0 Monopoly or perfectly collusive oligopoly
0 < H <1 Monopolistic competition
H= 1 Perfectly Competitive

The nature of the estimation of the H-statistic means that one is especially interested in
understanding how revenues react to variations in the cost figures.

However, Bikker, Spierdijk and Finnie (2007) provided empirical evidence that the level of
competition in the banking industry in the existing empirical P-R literature is systematically
overestimated. The reason for the misspecifications is that most studies use different definitions
of the appropriate variable to represent banks’ revenue (different definitions of the dependent
variable in the P-R model). It should be stressed that Bikker et al provided empirical evidence to
show that the scaled P-R model is mis-specified. The reason for this misspecification is that most
studies use scaled versions of bank income as the dependent variable in the P-R model and work
with revenues divided by total assets. However, scaling changes the nature of the model
fundamentally, since it transforms the revenue equation into a price equation.

10
2.1.2. COMPETITION IN BANKING: GOOD OR BAD?

Measuring competition is not an end for policy purpose. Theories relate the level of competition
in banking industry with financial stability and/ fragility. Though there is a belief about a
positive effect of normal competition on stability of the financial sector and hence on economic
development, still there is no an unambiguous argument on the relationship between bank
competition and the resultant effect on stability and / fragility in the sector. Proponents of
competition relate banking competition with stability in the financial system. The theoretical
model of Boyd and De Nicolo (2005) showed that a bank operating under less competitive
system will take higher risk in their loan portfolio as banks charge higher interest rate for
borrowers (will create moral hazard).Supporting this, Cetorelli (2001) stated that monopoly in
banking would lead to lower interest rate for saver and higher interest rate for borrower and
hence distort entrepreneurial incentives toward the undertaking of excessively risky projects.
This collectively will lead to instability in credit market and systematic failure. Moreover, less
competition in banking will let banks to engage in risky activities (due to too-Big- to fail
incentive) and hence will have a high probability to fail (Brian A, etal, 2014).

Contrary to the above argument, opponents of competition relate banking competition with
financial crisis. Different theoretical models based on Charter-value of banking predicted that
bank competition will lead to instability. In this view, as summarized by Beck (2008), profits
provide buffer against fragility and provide incentives against excessive risk taking in a more
competitive banking system. In support of this, Xavier (2010) put two channels how bank
competition creates instability. The first channel is through worsening the coordination problem
of depositors/investors which will foster bank run. The second channel is by letting banks to take
high risk in their portfolios.

11
2.2. EMPIRICAL LITERATURE

Given different implication of the level in bank concentration and competition, many empirical
studies have been conducted at different time and economy. Hence, by dividing the countries
into developed and developing economies, selected empirical literatures on bank competition and
concentration is presented in this section.

The degree of banking competition, including overtime change and determinants, in developed
economy is comphrensively studied from early periods to present. Shaffer (1993) and Nathan
etal (1989) tested the degree of market power in Canadian banking industry with different
method of measurement. The first one employed the Bresnahan's index (with relatively long time
series data) while the latter used a Panzar-Rosse method. Both find banks in Canada is
characterized by the existence of contestability (competition though it is highly concentrated).

Claessens etal (2004) tries to determine what derives bank competition internationally, using 50
countries banking system and employing the PR method. They found that a banking system with
greater foreign bank entry and fewer entry and activity restrictions will have a competitive
outcome. Furthermore, they did not find a negative relationship between competitiveness
measurement and concentration which implies contestability matters for the existence of
competition rather than structure of the market.

Most studies in the European banking industry found that banks in the region are operating under
monopolistic competitive market structure. For instance, Christos etal (2006) for EU, Coccorese
(2002) for Italian, Hempell (2002) for Germany, Hondroyiannis et.al (1999) for Greek, Yildirim
etal (2002) for central and eastern Europe, Molyneux et al. (1994), DeBandt etal (2000) for group
of European countries.

With regard to the level of banking competition in developing countries different studies were
conducted. A recent research work by Haytem etal (2015) showed that less competition in the
banking sector leads to a more resilient banking sector in developing countries. The researchers
first derived the level of competition in Libyan Banking industry using concentration index and
then relate it with the level of banking stability proxied by non-performing loan. Their empirical
result supported the “Competition-fragility” view of banking literature. The study of Samuel

12
(2013) also found a monopolistic competition structure across African sub regional banking
markets.
[

In the way to determine the market structure of banking industry in Zambia Simpasa (2013)
found that Zambian banks earned their revenue under conditions of monopolistic competition.
The finding is supported by the estimate of the Lerner index which suggests that the degree of
competitiveness may not be as low as previously understood. Similar with this, Rufus etal, (2013)
showed that Nigerian banks are operating under monopolistic competitive market structure. He
employed a PR methodology with dynamic Generalized Method of Moments technique for the
period spanning from 1990 to 2009. The estimated H-statistics found in the range between 0.09
and 0.11 which is near to monopoly type market structure.

Bing etal (2013) reported that competition of banks in china has been increasing after 1980’s
progressive financial reforms. Their result is contrary to previous works that are done in China
and this is due to different methodology employed in estimating the level of bank competition.
They used Profit Elasticity (PE) approach while earlier studies used either PR or Lerner Index.

A cross-country study by Diego etal (2010) found that banking sector competition, using non-
structural measure, in the Middle East and North Africa (MENA) region is lower relative to other
regions and has not been improved in recent years. The regions worse credit information
environment and lower market contestability are considered as a reason for the low level of
banking sector competition.

Anthony etal (2008) analyzed the existence of market power in Ghanaian banking industry by
calculating Lerner index of banks for the period 2001 to 2006 and found that banks in Ghana
possess a market power. Bank size, efficiency of banks with respect to staff costs, the
macroeconomic environment and time are the contributing factors for the evidence of market
power in Ghanaian banks. Another study by Thierry etal (2005) for the period 1998 to 2003,
using PR methodology, found a rather monopolistic competitive market structure for Ghanaian
Banks.

David etal (2005) estimate the level of competition in Uganda banking sector using a Panzar
Rosse methodology for a quarterly panel dataset of 15 banks from March 1999 to June 2004.

13
Their result suggested that the Ugandan banking sector is characterized by monopolistic
competition. Moreover, they reported banking competition in Uganda has increased after
privatization of state owned banks and consolidation is taken place in the sector. Similarly,
Jennifer etal (2014) found the positive effect of financial liberalization on banking sector
competition in sub-Saharan countries and it, indeed, enhances financial stability.

Given the structural change undertaken in Africa banking sector, Florian (2014) examined the
changes in competition in seven West African countries using 3 different competition indicators
(PR, Lerner index and Boone indicator).He found that competition in the region is increasing in
the last decades due to the expansion of African banking groups in the referred period.

THE CASE OF ETHIOPIA

As best of the researcher knowledge, there is only one research done regarding competition in
Ethiopian banking industry. Zerayehu etal (2013) tries to determine the degree of bank
competition, using P-R approach, in Ethiopia by only calculating static H-statistics from the
period 2000 to 2007. They find monopolistic competitive market structure. But they failed to
show how competition is changing overtime, to use un-scaled dependent variable as they used a
scale variable in the left hand side of the model, to use appropriate input prices (which is the
heart of P-R model and to use to-date data in their analysis. Moreover, they conclude the market
is incontestable without conducting a test.

To sum up, studies regarding the level of competition in banking industry revealed that most
markets both in developed and developing economy operate under Monopolistic competitive
market structure. Moreover, empirics showed that existence of competition in banking industry
in a concentrated market.

Misspecification Test/Methodological

Bikker etal (2006), using sample of more than 18,000 banks in 101 countries over 16 years,
demonstrated that the level of competition in the existing Panzar- Rosse(P-R) empirical
literature is systematically overestimated and that the tests on both monopoly and perfect
competition are distorted. This is due to the use of bank revenues divided by total assets (price of
asset) as dependent variable and/ including scale variable in the P-R model instead of unscaled
bank revenues. If we used scaled dependent variable, the estimated elasticity’s appear to be
price-to-price rather than Revenue-to-Price elasticity (which is the original PR measure of
elasticity).

14
 CHAPTER THREE

3. RESEARCH METHODOLOGY
3.1. DATA TYPE AND SOURCE
The study used secondary type of data. The study used audited commercial banks financial
statement, National Bank of Ethiopia banking database, World Bank database and IMF data base.
In addition, different Books, Internet and journals are used.

3.2. METHOD OF DATA ANALYSIS

In the way to show competition in Ethiopian Banking industry, we used both descriptive and
econometric analysis.

3.2.1. DESCRIPTIVE ANALYSIS

The researcher uses descriptive statistics for the analysis of the collected data. It is used because
of its simplicity & clarity to see what is going on in the industry at a glance and helps to support
the finding we get in econometric analysis. The analysis of the secondary data generated from
different sources is graphically represented with the help of excel

3.2.2. ECONOMETRIC ANALYSIS

In order to answer the important question we posed in the introduction section, the paper
employed the following data and model.

3.2.2.1. DATA

The span of our data source covers the time period 2002/03 to 2014/15. We used quarterly data
taken from all commercial banks operating in the country. (The list of banks is available in
appendix1). We then classify the period into 3 subs – periods. We do this for two reasons: first it
helps us to see the trend of competition over time. Second, it makes our estimation more
powerful as the data become balanced in the whole sample period.

Following previous literatures and theoretical foundations, we included three dependent and
different independent variables.

15
Dependent Variables include Total income, Interest income and ratio of Total income to Total
Asset.

Explanatory Variables basically includes Price of Labor, Price of Capital and Price of Fund. But
in addition to these variables we also used other control variables like ratio of Non-performing
loan to loan, share of non-interest income, share of total asset, Loan to deposit ratio and dummies
(policy variables and ownership of the bank).All variables are in logarithm form except the
dummies.

The main variables in P – R specification is price of inputs which includes price of labor, price
of capital and price of fund that helps to calculate the elasticity of Revenue to input prices. Hence,
the researcher used proxies for those variables that are recommended in lots of literature. In
literatures, while estimating the H- Statistic, they failed to find the exact proxies due to lack of
data. But this paper address the problem of proxies (price of inputs) as the researchers exert a
maximum effort to collect the data’s. The data’s and variables include;

1. Price of Labor (PL): In literature it is peroxided by the ratio of total salary expenses to
total asset. But, it is not the exact price paid for the employees. Bikker etal and others
suggested using ratio of total salary expense to total number of employees as the proxy
for price of labor is best; this paper does that. As we used quarterly data, we didn’t
simply divide the quarterly salary expenses by the total employees which are reported by
banks’ balance sheet. What we do is: as salary expenses is a flow concept, we subtracts
the existing quarter total salary expense from its previous quarter which gives the exact
expenses paid to employees in that quarter. Mathematically,

SQ t = SQ t * - SQ t-1
Where: SQ t is total salary expense paid in that quarter
SQ t * is total salary expense paid up to that quarter
SQ t-1 is total salary expense paid upto the previous quarter

Then, PL = SQt / L
Where; L is total number of employees (labor) in that quarter.

16
 2. Price of Capital (PK): Mostly, because of lack of data, it is peroxided by dividing
Depreciation by fixed asset or by dividing non – interest expenses by fixed asset. But,
dividing general expense by number of branch is the best proxy for price of capital as
suggested by previous literatures. Hence, we used the recommendation and calculate
price of capital as;

PK = General expense / number of branches

3. Price of fund (PF): No problem with this proxy in previous works as well.it is calculated
by dividing total interest expense by total deposit. But in our case we take care when
divide the variables because we used quarterly data. So, we follow the same step that
used in calculating quarterly salary expenses. Finally, we calculate price of fund as;

PF = interest expense / total deposit

3.2.2.2. Model Specification

We have followed the model developed by Panzar and Rosse (1982), with slight adjustment. In
general the Panzar – Rosse revenue test is based on a reduced – form revenue equation relating
gross revenue to vector of input prices and other control variables. The empirical reduced- form
revenue equation of the P-R model is written as

𝑚 𝑛

ln(𝑅𝑖𝑖 ) = � 𝛽𝑖 ln(𝐼𝐼𝑖𝑖 ) + � 𝛾𝑖 𝐶𝐶𝑖𝑖 + 𝜀𝑖𝑖 … … … … … … … … … … … … … … … … … … … … … . . (3)

𝑖=1 𝑖=1

Where; R it is revenue

IP it is vector of input prices

CV it is other control variables that are expected to affect revenue

βi is elasticity of revenue to input prices.

Then Panzar and Rosse used the sum of input price elasticity’s (which is represented by H –
statistic) to reflect the competitive structure of the market (necessary condition).

17
Condition: 𝑯 = ∑𝟑𝒊=𝟏 𝜷𝒊

If H < or = 0, then the market is characterized by Monopoly

If 0 < H < 1, then the market is characterized by Monopolistic Competition

If H = 1, then the market is characterized by Perfect Competition

3.2.2.3. ESTIMATION PROCEDURE

Theoretically, a nice model is the one which passes through three necessary steps. That is, pre
estimation test, model estimation and post estimation test (Diagnostic test).

A) Pre Estimation Test (Unit Root Test)

In panel data analysis unit root test is the basic test in order to select estimation method. If the
variables have a unit root, then applying panel VAR is best method other than random/fixed
effect. Therefore, we have tested stationarity of variables using Levin–Lin–Chu unit root test.

B) Estimation of the Models

After testing for unit root of the variables, we specified and estimated the models using fixed
effect as all the variables found stationary. For robustness check, we also estimate fixed effect
and compare the results with random effect. Using the general model of regression suggested
above in equation 1 and employing Random effect, we have estimated sixteen different models
as follows.

Model A (Three models)

Model A is used to see overtime change in market structure (degree of competition) of banking
industry in Ethiopia. First we divide the sample period (2002/03 – 2014/15) into three sub
periods by making the period suitable for analysis. The sub periods include;

Period 1 – from 2002/03 quarter three to 2008/09 quarter two

Period 2 – from 2008/09 quarter three to 2010/11 quarter one

Period 3 – from 2010/11 quarter two to 2014/ 15 quarter four

18
Then, we estimated three similar but not identical models. They are similar in the sense that all
the models estimated by taking total income as dependent variable and price of inputs as a major
independent variable. They are not identical because they take different control variables in the
course of estimation. The general form for the models is;

ln(𝑇𝑇𝑖𝑖 ) = 𝛽1 ln(𝑃𝑃𝑖𝑖 ) + 𝛽2 ln(𝑃𝑃𝑖𝑖 ) + 𝛽3 ln(𝑃𝑃𝑖𝑖 ) + 𝛾 ln(𝐶𝐶𝑖𝑖 ) + 𝜑𝑖 + 𝜀𝑖𝑖 …………….(2)

Where; ln(TI it ) is Total income of bank “i” at time “t”

ln(PLit ) is price of Labor in bank “i” at time “t”

ln(PK it ) is price of capital in bank “i” at time “t”

ln(PLit ) is price of fund in bank “i” at time “t”

ln(CV it ) is control variables for bank “i” at time “t”

𝜑𝑖 is Bank fixed effect and time invariant factors

ε it is error term

β1 , β 2 , β 3 and γ are elasticity’s

This is the general model (Model A) which is used to estimate three different models.

Model B (Two Models)

The purpose of Model B is to test whether Non interest income is a substitute or complement for
Interest income. In microeconomic theory a good is substitute for another good if they have
negative and significant cross price elasticity. For instance, Coffee and Tea are substitutes,
whereas tea and sugar are complement goods. But, in banking literature, there was no priori test
that identifies the exact relationship between interest and non-interest income. The two models 3
which are used to test the claim are;

3
Two models because we divide the period in to two sub periods

19
 ln(𝐼𝐼𝑖𝑖 ) = 𝛽1 ln(𝑃𝑃𝑖𝑖 ) + 𝛽2 ln(𝑃𝑃𝑖𝑖 ) + 𝛽3 ln(𝑃𝑃𝑖𝑖 ) + 𝛾1 ln(𝑙𝑙𝑙𝑖𝑖 ) + 𝛾2 ln(𝑠𝑠𝑠𝑖𝑖 ) + 𝛾3 ln(𝑠𝑠𝑠𝑠𝑖𝑖 )
+ 𝜑𝑖 + 𝜀𝑖𝑖 … … … … … … … … . . (3)

Where; ln(liiit ) is lag of interest income of bank “i” at time “t”

ln( stait ) is share of total asset of bank “i” from the industry at time “t”

ln( sniiit ) is share of non-interest income from the total income at time “t”

From the model our interest is γ 3 . After estimating this model well and looking at γ 3 , we can

decide whether interest and non-interest income are complement or substitute.

Rule: If we get γ 3 positive and significant, then they are complements

If we get γ 3 negative but insignificant, then they are substitutes but not perfect

If we get γ 3 negative and significant, then they are substitutes.

Model C (Four Models)

Here we estimated four models in order to test the claim of Bikker etal (mis-specification of P-R
Model) which is proposed in 2006. They suggest that using scaled dependent variable instead of
un-scaled dependent variable is biased towards perfect competition. This is due to, based on their
assumption, scaled dependent variable represent price rather than revenues, and the result (H –
4
Statistic) will be price- price elasticity rather than revenue – price elasticity.

Having this we estimated, to test their claim, four models as follows;

ln(𝐼𝐼𝑖𝑖 ) = 𝛽1 ln(𝑃𝑃𝑖𝑖 ) + 𝛽2 ln(𝑃𝑃𝑖𝑖 ) + 𝛽3 ln(𝑃𝑃𝑖𝑖 ) + 𝛾 ln(𝑠𝑠𝑠𝑠𝑖𝑖 ) + 𝜀𝑖𝑖 … … … … … … . . … … … (4)

ln(𝐼𝐼𝐼𝐼𝑖𝑖 ) = 𝛽1 ln(𝑃𝑃𝑖𝑖 ) + 𝛽2 ln(𝑃𝑃𝑖𝑖 ) + 𝛽3 ln(𝑃𝑃𝑖𝑖 ) + 𝛾 ln(𝑠𝑠𝑠𝑠𝑖𝑖 ) + 𝜀𝑖𝑖 … … … … … … … . … (5)

ln(𝑇𝑇𝑖𝑖 ) = 𝛽1 ln(𝑃𝑃𝑖𝑖 ) + 𝛽2 ln(𝑃𝑃𝑖𝑖 ) + 𝛽3 ln(𝑃𝑃𝑖𝑖 ) + 𝛾1 ln(𝑛𝑛𝑖𝑖 ) + 𝛾2 𝑑1 + 𝛾3 𝑑2 + 𝜀𝑖𝑖 … . . … (6)

ln(𝑇𝑇𝑇𝑇𝑖𝑖 ) = 𝛽1 ln(𝑃𝑃𝑖𝑖 ) + 𝛽2 ln(𝑃𝑃𝑖𝑖 ) + 𝛽3 ln(𝑃𝑃𝑖𝑖 ) + 𝜀𝑖𝑖 … … … … … … … . … … … … … … … (7)

4
For mathematical proof you can refer the work of BIkkeretal 2006

20
Where; nl it is the ratio Non – performing loan to total loan of bank “i” at time “t”

d1 is dummy for representing type of the bank. That is;

d1 = 1, if the bank is public bank

= 0, if the bank is private bank

d2 is dummy for representing credit cap

d2 = 1, if there is a credit cap in the period

= 0, if there is no credit cap

First, the result from equation (4) is compared to that of equation (5). The rule is; If we get
different H-statistics, we can confirm the claim of bikker etal especially if the latter H – statistic
is biased towards competitive environment. Moreover, we again compare the result from
equation (6) and equation (7) so as to make the conclusion more concrete. The rule is the same
with the above.

Model D (Three Models)

Three models are estimated for two purposes. First, it is used to see whether competition is lower
or higher among private banks alone by dropping public banks from the sample. Second, it is
used; how competition among private banks is changing. The estimated models include;

ln(𝑇𝑇𝑖𝑖 ) = 𝛽1 ln(𝑃𝑃𝑖𝑖 ) + 𝛽2 ln(𝑃𝑃𝑖𝑖 ) + 𝛽3 ln(𝑃𝑃𝑖𝑖 ) + 𝛾1 ln(𝑙𝑙𝑙𝑖𝑖 ) + 𝛾2 ln(𝑠𝑠𝑠𝑖𝑖 ) + 𝛾3 ln(𝑠𝑠𝑠𝑠𝑖𝑖 )

+ 𝜀𝑖𝑖 … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … . . (8)

ln(𝑇𝑇𝑖𝑖 ) = 𝛽1 ln(𝑃𝑃𝑖𝑖 ) + 𝛽2 ln(𝑃𝑃𝑖𝑖 ) + 𝛽3 ln(𝑃𝑃𝑖𝑖 ) + 𝛾1 ln(𝑙𝑙𝑙𝑖𝑖 ) + 𝛾2 ln(𝑠𝑠𝑠𝑖𝑖 ) + 𝛾3 ln(𝑠𝑠𝑠𝑠𝑖𝑖 )

+ 𝜀𝑖𝑖 … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … . . (9)

ln(𝑇𝑇𝑖𝑖 ) = 𝛽1 ln(𝑃𝑃𝑖𝑖 ) + 𝛽2 ln(𝑃𝑘𝑖𝑖 ) + 𝛽3 ln(𝑃𝑃𝑖𝑖 ) + 𝛾1 ln(𝑠𝑠𝑠𝑖𝑖 ) + 𝛾2 ln(𝑛𝑛𝑖𝑖 ) + 𝜀𝑖𝑖 … … . . (10)

21
Equation (8) is the model which used to compute degree of competition by taking both private
banks and public banks into account. Whereas equation (9) & (10) are models used to compute
degree of competition among private banks only by dropping public banks from the sample.

After estimating the models, first we compare (8) & (9) to see whether there is a difference in
degree of competition when we drop public banks from the sample. Then (9) & (10) are
compared to see over time change in competition among private banks.

Model E (Three Models)

This model is estimated to see the effect of Non- Performing Loan (nl) on banks’ ability to
generate income. It is done in two ways. By taking total income as a dependent variable, two
models are estimated to see the effect of nl on total income and overtime change in the effect of
nl on total income, if there exists.in addition to these models one additional model is estimated to
see the effect of nl on interest income. The models are,

ln(𝑇𝑇𝑖𝑖 ) = 𝛽1 ln(𝑃𝑃𝑖𝑖 ) + 𝛽2 ln(𝑃𝑃𝑖𝑖 ) + 𝛽3 ln(𝑃𝑃𝑖𝑖 ) + 𝛾1 ln(𝑙𝑙𝑙𝑖𝑖 ) + 𝛾2 ln(𝑛𝑛𝑖𝑖 ) + 𝜀𝑖𝑖 … … … (11)

ln(𝑇𝑇𝑖𝑖 ) = 𝛽1 ln(𝑃𝑃𝑖𝑖 ) + 𝛽2 ln(𝑃𝑃𝑖𝑖 ) + 𝛽3 ln(𝑃𝑃𝑖𝑖 ) + 𝛾1 ln(𝑛𝑛𝑖𝑖 ) + 𝛾2 ln(𝑠𝑠𝑠𝑠𝑖𝑖 ) + 𝛾3 𝑑2

+ 𝜀𝑖𝑖 … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … . (12)

ln(𝐼𝐼𝑖𝑖 ) = 𝛽1 ln(𝑃𝑃𝑖𝑖 ) + 𝛽2 ln(𝑃𝑃𝑖𝑖 ) + 𝛽3 ln(𝑃𝑃𝑖𝑖 ) + 𝛾1 ln(𝑠𝑠𝑠𝑖𝑖 ) + 𝛾2 ln(𝑛𝑛𝑖𝑖 ) + 𝜀𝑖𝑖 … … . . (13)

Our interest in this model is γ 2 from (11) & (13) and γ 1 from (12).

Model F (One Model)

The model is used to test whether NBE policy have positive or negative impact on banks’ ability
to generate income thereby on degree of competition. Basically, we are interested to see the
effect of credit cap on income generation capacity of banks. And we can compare the effect with
that of NBE bill, as it will have serious policy implication. The model is;

ln(𝑇𝑇𝑖𝑖 ) = 𝛽1 ln(𝑃𝑃𝑖𝑖 ) + 𝛽2 ln(𝑃𝑃𝑖𝑖 ) + 𝛽3 ln(𝑃𝑃𝑖𝑖 ) + 𝛾𝛾2 + 𝜀𝑖𝑖 … … … … … … … … … … … … (14)

Our interest is the coefficient of d2. If we get γ negative, credit cap negatively affected income
generating capacity of banks.

22
C) Post-Estimation (Diagnostic Check)

In panel data models, mostly, cross-sectional interdependence test has to be undertaken to

confirm the fitness of estimated model. Hence, after estimating the models, the researcher
checked for the validity of the models by testing cross-sectional interdependence using Breush –
Pegan Lagrangian Multiplier test.

23
 CHAPTER FOUR

4. DATA ANALYSIS AND PRESENTATION

4.1. DESCRIPTIVE ANALYSIS

4.1.1. OVERVIEW OF ETHIOPIAN BANKING INDUSTRY

The history of modern banking system of Ethiopia dated back to Emperor Minilik II when
agreement is reached between the emperor and Mr.MaGillivray, representative of the British
owned National Bank of Egypt, on the establishment of Bank of Abyssinia in 1906. Then after
Emperor Haile Selassie come to power Bank of Abyssinia is legally replaced by Bank of
Ethiopia and became the first indigenous banks for Ethiopia as well as Africa. Bank of Ethiopia
have dual role both as commercial bank and central bank until it split into National Bank of
Ethiopia (as a central bank) and Commercial Bank of Ethiopia in 1963. During the H/Selassie era
the banking sector was open to foreign banks with requirement of at least 51 percent
domestically owned. As a result there was 1 state owned, 1 private owned and 2 foreign owned
commercial banks operating in the country until nationalized by communist government in 1974.
The communist government merging three private owned banks in 1976 to make Addis Bank
and further merge Addis Banks with state owned commercial banks in 1980 to make one state
owned commercial banks. Following the downfall of Derg regime banking sector is liberalized
for private sector as a result currently there are 1 central bank, 18 commercial banks (of which 2
of them are government owned) and one special bank operating in the country, Development
Bank of Ethiopia. However, still the sector is closed for foreign banks (National Bank of
Ethiopia website).

The current banking system of Ethiopia is dominated by public owned bank, Commercial Bank
of Ethiopia. Commercial Bank of Ethiopia (CBE) has a share of around 67 percent of
commercial banks total assets and total deposits each, 62 percent of total loan extended by
commercial banks, and 37 percent of total commercial banks capital as of June 2014. Moreover,
CBE dominate the sector in terms of branch expansion. The total number of bank branch as of
June 2014 is 2,176 of which 856 branches are owned by CBE (39 percent of total commercial
banks branch). In addition, from total bank account holder of 13,044,228, CBE has a share of

24
around 63 percent (8,187,070 account holders). This is mainly due to current government
development model, democratic developmental state; in which government highly participates in
the production as well as distribution of selected goods and services, which government believe
essential for economic growth and in which private sector fails to do it with expected efficiency
and magnitude. As a result during the same period from total banking sector credit, 64 percent is
goes to government and public enterprise whereas the rest goes to other sectors (private and
cooperative). However, in terms of deposit only 18 percent of bank deposit is come from
government and public enterprises.

The other peculiarity of Ethiopia banking system is the high profitability of the sector mainly due
to ample potential of the sector and the restriction of foreign banks that give domestic banks the
opportunity to earn high profit. During the last four fiscal years (2009/10-2013/14) income after
tax of commercial banks observed to be greater than 2 percent of their total assets.

Table 3: Ratio of selected financial performance indicators

(Ratio’s are expressed in percentage)

Particulars 2009/10 2010/11 2011/12 2012/13 2013/14

Income after tax to total assets 2.01 2.03 1.93 2.18 2.23
Total expense to total income 47.37 51.53 40.60 46.77 53.09
Source: Authors computation based on National Bank of Ethiopia database

Total expense to total income ratio is calculated to look at the amount of expense incurred to
produce 100 unit of income (since ratio is calculated in percentage). On average, the ratio of total
expense to total income for the last four years is found to be 47.9. Implying, on average 47.9 Birr
of expense is required to earn 100 Birr income. Even the recent data for 2013/14 shows ratio of
total expense to total income is 53.09 implying commercial banks outlay Birr 53.09 to produce
output equal to Birr 100. This in turn shows us how much the sector is profitable.

25
4.1.2. CONCENTRATION IN ETHIOPIAN BANKING INDUSTRY

Ethiopian banking industry can be characterized by high concentration. If we take Top three5
banks concentration ratio, though it is declining over time still there is high concentration on
Asset and Deposit (Table 4).The concentration is declining because the number of banks joined
to the industry was increased as the sector is highly profitable. Bank entry in Ethiopia has been
historically fairly easy as the Minimum paid-up capital (ETB 70 Million) was relatively low if
we compare it with the new requirement (ETB 500 Million). In early period lack of skilled
manpower was the problem rather than financial requirement. But, concentration by itself will
not tell us the level of competition in the industry since the industry can be characterized by
competition in the presence of high concentration if the market is contestable.

Table 4: Top Three banks Concentration ratio

Period Asset Deposit Loan

2001/02 0.91 0.91 0.81
2002/03 0.90 0.89 0.80
2003/04 0.89 0.88 0.78
2004/05 0.87 0.87 0.76
2005/06 0.85 0.84 0.73
2006/07 0.85 0.82 0.68
2007/08 0.80 0.80 0.65
2008/09 0.78 0.78 0.61
2009/10 0.78 0.74 0.55
2010/11 0.74 0.72 0.56
2011/12 0.75 0.73 0.56
2012/13 0.80 0.75 0.55
2013/14 0.76 0.74 0.56
2014/15 0.73 0.73 0.54
Average 0.82 0.80 0.65
Source: Own computation based on National Bank of Ethiopia data base

4.1.3. COMMERCIAL BANKS INCOME COMPOSITION IN ETHIOPIA

Generally, Income of a bank is divided into two: Interest (II) and Non-interest income (NII). The
first one is generated by giving traditional banking activity which is from disbursing a loan to the
economy. Whereas the latter includes the income that is earned from forex, commission and
charge this is beyond the traditional banking activity.

5
The banks include Commercial Bank of Ethiopia, Awash International Bank and Dashen Bank

26
As it is shown under Table 5 & Fig 2, Generally, The share of interest income was much higher
than Non-interest income in early times. But the trend shows that share of NII is increasing
overtime and starts to decline after it reaches the pick on the year 2011/12. If we look at the
share by type of banks: on average, public banks Non-interest income was higher than that of
private banks. But the share of Non-interest income at private banks starts to dominate onwards
the year 2010/11 until 2011/12. Though the share of NII at private banks starts to decline after
2012/13, still the share at private banks is higher than that of public banks.

More specifically, the share of NII from total income at newly established banks are higher than
that of old established banks 6 in recent periods. The share of NII at private banks (both old &
new) starts to dominate interest income onwards the year 2010/11 where there was a credit cap
policy in the specified period. The highest record was on the year 2011/12 in which newly
established banks earned 60.0 percent of their income from NII. The trend shows that public
banks are getting their income mainly from traditional banking activities (giving loan) whereas
private banks, especially newly established, earn their most income from non-primary banking
activities. Currently public banks income is mainly composed of interest income which is around
70 percent and the remaining 30 percent from NII. On the other hand, newly established banks
earn their most income in the form of NII and it is around 51.0 percent and the part 49.0 percent
is comes from interest income. Moreover, newly established banks are more responsive for NBE
policy. This is explained in the period of credit cap where their income composition shows a
quick shift from interest income to NII that is from 42.0 to 51.0 percent.

6
We have categorized banks depend on the year of establishment.

27
Table 5: Share of Non-interest income in Ethiopian Banking industry

(In percent)

Period Commercial Bank Private Banks

of Ethiopia(CBE) All Old New
2001/02 24.5 32.3 32.3 -
2002/03 33.8 27.8 27.8 -
2003/04 48.3 36.4 36.4 -
2004/05 46.4 37.7 37.7 -
2005/06 53.4 37.5 37.5 -
2006/07 53.3 33.0 37.9 3.9
2007/08 54.4 32.1 33.8 26.9
2008/09 48.1 37.9 38.1 36.8
2009/10 36.2 47.6 41.6 59.5
2010/11 43.6 53.2 52.0 54.1
2011/12 49.7 58.5 55.5 60.0
2012/13 41.6 49.3 41.3 52.9
2013/14 29.7 47.6 36.9 51.0
2014/15 27.3 46.4 35.72 50.43
Source: Own computation based on National Bank of Ethiopia data base

Figure 2: Trend in Share of Non-interest income

70.0

60.0

50.0
Commercial Bank of
in percent

40.0 Ethiopia
All private banks
30.0

20.0 Old private banks

10.0 Newly established Private

banks
0.0

Year

Source: Own computation based on National Bank of Ethiopia data base

28
4.2. ECONOMETRIC ANALYSIS

4.2.1. UNIT ROOT TEST

Before conducting the estimation of the models, we first test for stationarity of the variables
using Levin-Liu-Chu (LLC) Test. And all the variables are found to be stationary. The result of
the test is presented in Appendix 2.

4.2.2. RESULT AND DISCUSSION

4.2.2.1. Competition and Contestability Test (Model A)

The econometric model with three different specifications was run on a pooled data set of the 17
banks to obtain a general picture of the competitive structure of the Ethiopian banking industry
over the sample period of 2002/03 – 2008/09, 2008/9-2010/11, and 2010/11 – 2014/15.We also
looked at the trend of changes in competitive conditions. Appendix 3 reports the model results.
In this model we used Total income as a dependent variable.

As it is shown under table 6, H-statistic is declining overtime. It was 0.7 for the period 2002/03-
2008/09 and slightly declined to 0.6 on the period 2008/09 – 2010/ 11 and finally becomes –0.2
(= 0 with χ 2 test) for the recent period. The result shows that competition in Ethiopian banking
industry is relatively high in earlier periods. As it is presented in Table 10, Policy of NBE like
Credit cap and NBE bill purchase are the reasons for the declining trend in competition. The
coefficient of D 2 in two regressions found negative and significant showing the negative impact
of credit cap on banks power to generate income thereby on competition.

Interestingly, the coefficient of Price of fund is found to be positive and significant for the period
2002/03 – 2008/09 and 2008/9-2010/11. But its sign become negative for the period 2010/11 –
2014/15. This is explained by; if firms operate under competitive market the increase in price of
fund will increase their revenue by increasing either the price of loan or quantity of loan.
Therefore, the effect of price of fund in total income becomes positive. But, theoretically, if
banks operate under monopoly and set prices so as to maximize profits; it will choose prices such
that its gross revenue responds in the opposite direction as a change in unit costs. This means
bank could have earned high profit by shrinking even without a change in input prices. Therefore,
the increase in unit costs leads to a decrease in revenue. The negative coefficient in our case is

29
appeared when the market is under monopoly. From microeconomic theory, Total revenue is
increased by increasing either price or output depends on the demand of the product. Borrowing
this theory, when there was a credit cap banks cannot increase the amount of loan (output)
beyond some level. Moreover, as banks were operating under monopoly, they cannot increase
price of their output as much if they would operate under competitive market. These two reasons
make the coefficient of price of fund negative.

Table 6: Summary of H - Statistics from Model A

Period Value of H- Statistics

2002/03 Q III – 2008/09 QII 0.7
2008/09 Q II – 2010/11 Q I 0.6
2010/11 QII – 2014/15 Q IV -0.2
Source: Own computation based on Model A results

Now let’s combine what we get in the descriptive and econometric analysis. In the descriptive
part, we observe Concentration was higher at earlier periods though it is declining. And
according to the above finding, we reject the monopoly and perfect competition hypotheses for
the period 2002/03 to 2008/09 and 2008/09 to 2010/11 and conclude that banks in Ethiopia seem
to earn their revenues as if under the conditions of monopolistic competition. The results suggest
that a highly concentrated banking market do not seem to lead to anti-competitive conduct as
suggested by the traditional SCP hypothesis, since the current analyses consistently reject the
existence of collusive behavior. These results seem to be compatible with contestable markets
theory (CMT), if we can assume that incumbent banks set their prices close to the competitive
level because of potential competition; otherwise higher prices will attract potential entrants with
hit-and-run strategies. These results are also consistent with the expectation that liberalization
and deregulation of Ethiopian financial markets have increased the competitive conditions in
banking industry.

To conclude about the market structure, we test the hypothesis proposed by Panzar and Rosse.
The test is based on elasticity of revenue to sum of input prices. The test rejects both monopoly
and perfect competition for the first two sample periods but can’t reject the presence of
monopoly in the last sample period.

30
4.2.2.2. Interest Vs Non- Interest Income (Substitute or complement products)? (Model B)

As it is discussed in descriptive part, the share of non-interest income is increasing overtime and
in some period its share was higher than that of interest income. This raises the question that has
to be answered: is non-interest income substituting or complement with interest income? To do
this, we have estimated model B1 by using share of non-interest income (SNII) as an explanatory
variable while interest income as a dependent variable (Appendix 4: model B1).And our interest
is the coefficient of SNII.

Table 7: Coefficient of Share of Non-interest Income (SNII) in Model C

Period Coefficient of SNII Decision

Early Periods Negative and insignificant Substitute
Recent Periods Negative and significant Substitute (relatively perfect)
Source: Own computation based on result of Model C1

The table above shows that Interest and Non-interest income are Non-complement products.
More specifically, non- interest income seems a perfect substitute for interest income in recent
times than early periods. Early periods are relatively fair for banks to earn interest income and
therefore banks don’t go deep to shift their activities from interest to non-interest income.

4.2.2.3. Mis-specification of Empirical P-R Model test

This test is undertaken to check whether the claim of Bikker etal (2006) is valid or not. Moreover
it is used to argue against the result of Zeray etal (2013) that used scaled dependent variable in
their way to compute competition of banks in Ethiopian banking industry. So, we estimated 4
models in order to test the claim of Bikker etal. The results confirm that using scaled dependent
variable is biased towards competitive structure as the P-R H- Statistics become price – price
elasticity rather than Revenue- price elasticity. The Table below is the summary of the model
results. (The detail is presented under appendix 5).

31
Table 8: Summary result of H- statistics that used to test mis-specification of P-R model

Dependent Variable Value of H-statistics

Scenario I
Interest Income 0.5
Interest Income / Total Asset 0.8
Scenario II
Total Income -0.1
Total Income / Total Asset 0.4
Source: Own computation based on result of Model C2

Hence, we empirically test the claim of Bikker etal and find an evidence of using scaled
dependent variable is a biased estimation to determine the stance of banking market structure in
an economy.

4.2.2.4. Competition among Private Banks (Model D)

Here we drop public banks from the sample and see the level of competition among private
banks. Furthermore, we compared the result with model A to see the effect of public banks in the
overall banking market structure. The result, Table 10, shows that competition among private
banks is increasing overtime. This competition is endogenously comes from the existence of
Commercial Bank of Ethiopia (CBE) in the industry. This can be explained by; CBE’s recent
time activities like aggressive branch opening & introduction of new deposit mobilizing schemes
make private banks to increase their participation in different banking activities so as to stay in
the business. But in early period competition is relatively high when we look the whole sample.
For the detail you can refer Appendix 6.

Table 9: Summary of competition result among private banks and comparison with Model A

Period Value of H-statistics

Among private banks In all banks (Model A)
Early Periods 0.5 0.7
Recent Periods 0.6 -0.2
Source: Own computation based on result of Model D & A

4.2.2.5. How Non- Performing loan affects Income of Banks? (Model E)

Mostly non-performing loan is expected to decrease income of a bank. We have estimated 3

models in order to check the effect of NPL on income of a specific bank. The regression result
shows that non-performing loan have a non-negative effect on Total income and negative effect

32
on interest income (Appendix 7). As it is shown under Figure 3 NPL affects total income of a
bank positively. But the effect can be divided into to two: significantly and insignificantly. In
early times the effect was not significant but significant in current periods.

When there is NPL it is automatic for interest income to decline but the effect on total income is
depend on the magnitude of an increase in Non- Interest income. Banks doesn’t give credit for
nothing; they accept collateral before disbursing the loan. And this collateral will be sold if the
borrowers default to payback the amount he/she borrowed. This is where the effect of NPL on
Non-interest income comes on. Banks sell the collateral and record it as non-interest income in
their income statement. This way NPL always have positive effect on non-interest income. The
non-negativity effect of NPL on Total income is then depending on banks’ ability to sell the
collateral in a good price or not. If they sell with a better price that they would get from
collecting interest income then it will more than compensate the loss of interest income and by
effect increase total income significantly (positive and Significant). On the other hand, if they
sell the collateral in a price that is almost equal to the loss they incur from not collecting the
interest income then the effect of increase in non-interest income becomes insignificant.

As per the interview made with officials at different commercial banks, in early period banks
were not selling the collateral in a good price than current times. This confirmed the econometric
result that NPL have positive & insignificant effect in early times and positive & significant in
current times on Total income. Diagrammatically;

33
 Figure 3: The effect of Non-performing Loan on Interest, Non- Interest and Total Income

Non- Performing loan (Increase)

(A)

Interest income (decrease) Non-interest income (Increase)

(B) (C)

High increment: Small increment: If

If collateral sold collateral not sold
with good price with good price
(C1) (C2)

Total income (increase)

(D)= (B) + (C)

Small increment
High increment
(B) + (C2)
(B) + (C1)

34
4.2.2.6. The Effect of NBE policy on Banks Income Statement

As it is in many countries, NBE is given the mandate to license and regulate banks & other
financial institutions by Article 14 of NBE establishment proclamation N0.591/2008.
Consequently, NBE issues a directive, policy or procedure in order to control the efficient
functioning of commercial banks. And these policies have either positive or negative effect on
banks performance. In the last decade, there are three major NBE policies which are expected to
affect banks income statement and competition in the industry directly or indirectly. These
include Credit cap, NBE Bill purchase and Minimum paid-up capital required for establishing a
bank. We have tested the effect of Credit cap and NBE bill purchase on banks income statement.
The result of regression shows that both Credit Cap (D2) and NBE Bill purchase (D1) have a
non-positive effect on banks revenue, especially on interest income. (See appendix 8 for the
detail). Credit cap limit banks not to give a credit more than some boundary and hence it lowers
the amount of interest income they would earn if there is no a limit. Regarding the effect of NBE
Bill purchase (insignificant), though it is not restrictive as that of Credit Cap, it lowers the
interest income they would earn if they disburse the total loan (including the 27 %) to the sector
they want.

Table 10: The effect of Credit Cap and NBE Bill purchase on Banks Income statement

Policy Effect on Income statement

Credit cap Negative and significant
NBE Bill purchase Negative
Source: Own computation based on Model F

Model Diagnosis / Post-estimation Test

Estimation of P-R H- statistics is valid if the models we estimated above pass the diagnostic test.
In panel data analysis, cross-sectional dependence is a threat for using random/ fixed effect
models. As we estimated the models using fixed effect, we need to check cross sectional
dependence after the estimation. The result of the test shows that, all the models estimated above
have no problem of Cross-sectional interdependence.

35
 CHAPTER FIVE

5. CONCLUSION AND RECOMMENDATION

5.1. CONCLUSION

The paper tries to see the nature and trend of market structure in Ethiopian banking industry over
the period 2002/03 to 2014/15 and finds the following points.

The Descriptive Analysis confirmed that Ethiopian banking industry is highly concentrated
though it is declining overtime. Looking at income composition of commercial banks, now days
newly established Private Banks earn their income mostly from non-traditional banking activity
in the form of non-interest income. Whereas public banks income is mainly composed of interest
income which shows public banks are relatively stable since the bank which depends on non-
interest income is highly vulnerable for the economy movement.

The econometric analysis found out that the Ethiopian banking industry is contestable in the
sense that though the industry has been highly concentrated there is an evidence of competition.
Moreover, it depict the competition is declining over time. The result further confirmed
competition is high among private banks. With respect to the effect of policy on banking
competition, Credit cap had a negative impact on banks income statement and hence on
competition. Non-Performing Loan have positive effect on total income and negative effect on
interest income. Moreover, the effect on total income is significant in recent periods as banks
gain more from selling the collateral than what they loose from not collecting the loan. The claim
of Bikker etal (2006) about Mis-specification of empirical P-R is correct and biased towards
competition as the elasticity becomes Price –Price if we used scaled dependent variable rather
than Revenue- Price elasticity when we use unscaled dependent variable. Regarding the effect of
ownership (dummy) on banks income statement, public banks get more income than private
banks as they have the highest market share. There is persistence in revenue of a bank as model
results show a significant effect of lagged dependent variable. Finally, the study introduced an
approach to test the relationship between interest and Non-interest income. By applying the
approach in Ethiopia banking industry, Interest and Non- interest Income has found to be non-
complement products.

36
 5.2. RECOMMENDATION

The Analysis part gives a clear picture that National Bank of Ethiopia policy, specially related to
commercial banks’ lending activity, have a negative impact on competition in the banking
industry and caution must be taken while formulating a policy. Furthermore, National Bank of
Ethiopia have to be fair enough both for private and public banks while introducing policy that
could limit part of banking business like lending activity. In analyzing bank competition, it is
good to use unscaled dependent variable rather than the extensive empirical scaled dependent
variable as the scaled dependent variable is biased towards competitive results. To empirically
test the relationship between Interest and Non-interest income in the banking industry, the study
recommend the use of share of non-interest income as an independent variable and interest
income as the main dependent variable by taking into account other control variables depending
on the structure of banking industry in that specific country.

37
 REFERENCES

Allen, N, Leora, F and Rima, T (2008), “Bank Competition and Financial Stability”, Policy
Research Working Study, The World Bank, WPS 4696.

Anthony, A, Akoena,s, Antwi-asare, T and Gockel, A (2008), “Explaining the market power of
Ghanaian banks”, South African Journal of Economics, Vol. 76:4

Baumol, W. (1982), “Contestable Markets: An Uprising in the Theory of Industry Structure”, the
American Economic Review, Vol. 72, No. 1(pp. 1-15)

Beck, T (2008), “Bank competition and stability, friends or foes?”,World bank policy research
working study, No.4656.

Bikker, J. (2004), “Competition and efficiency in a unified European banking market”, Edward
Elgar publishing limited, UK.

Bikker, J., L. Spierdijk, and P. Finnie. (2006.), “Misspecification of the Panzar-Rosse Model:
Assessing Competition in the Banking Industry." De Nederlandsche Bank Working Study, No.
114

Bing, Xu, Adrian van Rixtel and Michiel van Leuvensteijn (2013), “Measuring bank competition
in China: a comparison of new versus conventional approaches applied to loan markets”, BIS
working Studys, No 422

Boyd J and De Nicoló (2005), “The Theory of Bank Risk-Taking and Competition Revisited”,
Journal of finance, Vol.60 (pp. 1329-1343).

Bresnahan, T (1982), “The oligopoly solution concept is identified”, Economic Letters, Vol. 10,
(pp. 87- 92)

Brian, A, Lynn, L, Jeffrey, N and Tjomme, O. (2014), “Bank Competition and Financial Stability:
Evidence from the Financial Crisis”, Journal of Financial and Quantitative Analysis, Research
Collection School of Accountancy.

38
Cetorelli,N (2001), “Competition among banks: Good or bad?”, Economic perspectives, Federal
Reserve Bank of Chicago

Christos, K and Anastasia, K (2006), “Competition and Concentration in the New European
Banking Landscape”, European Financial Management, Vol. 12, No. 3 (pp. 443–482)

Claessens S and Laeven L (2004), “What Drives Bank Competition? Some International
Evidence”, Journal of Money, Credit and Banking, Vol. 36, No. 3, (pp. 563-583)
Coccorese, P (2002), “Competition among dominant firms in concentrated markets: evidence
from the Italian banking industry”, Working Study No. 89

David, H and Shanaka, J (2005),”Bank Efficiency and Competition in Low-Income Countries:

The Case of Uganda”, IMF working study, WP/05/240

De Bandt, O and Davis, P (2000), “Competition, contestability and market structure in European
banking sectors on the eve of EMU”, Journal of Banking and Finance, Vol.24, (pp. 1045–1066)

Diego A, María, S Martínez, P and Roberto, R (2010), Bank Competition in the Middle East and
Northern Africa Region, World Bank policy research working study, WPS5363.
Econometrics

Florian L (2014), “Has competition in African banking sectors improved? Evidence from West
Africa”, CERDI, Universit_ed'Auvergne.
Gutierrez de RL (2007), “Testing for competition in the Spanish banking industry: the Panzar-
Rosse approach revisited”, Documentos de Trabajo.p.0726

Haytem, A and Rashid, S (2015), “The Relationship between Banking Competition and Stability
in Developing Countries: The Case of Libya”, MPRA Study, No. 64932

Hempell, H (2002), “Testing for competition among German banks’”, Discussion Study, 04/02,
Economic Research Centre, Deutsche Bundesbank

Hondroyiannis, G, Lolos, S. and Papapetrou, E (1999), “Assessing competitive conditions in the

Greek banking system”, Journal of International Financial Markets, Institutions and Money, Vol.
9, (pp. 377–391).
39
Jackson, W (1992). The Price-concentration relationship in banking: A comment. Review of
Economics and Statistic, Vol.74 (pp. 373-376)

Jennifer M, Boaz, N, Jacob, O, Simpasa, A (2014), “Financial Sector Reforms, Competition and
Banking System Stability in Sub-Saharan Africa”, Study to be presented at the IMF/DFID
Conference on “Macroeconomic Challenged Facing Low-Income countries”, International
Monetary Fund, Washington DC, January 30 – 31, 2014

Keeley, M. (1990), “Deposit Insurance, Risk and Market Power in Banking”, American
Economic Review, (pp. 1183-1200)

Molyneux, P, Lloyd-Williams, D. and Thornton, J (1994), “Competitive conditions in European

banking”, Journal of Banking and Finance, Vol. 18, ( pp. 445–59).

Nathan, A and Neave, E (1989), “Competition and Contestability in Canada's Financial System”,
The Canadian Journal of Economics, Vol. 22, No. 3 ( pp. 576-594)

National Bank of Ethiopia, various issues

Panzar, J and Rosse, J (1987), “Testing For "Monopoly" Equilibrium”, The Journal of Industrial
Economics, Vol. 35, No. 4 ( pp. 443-456)

Pagano, M (1993),”Financial markets and growth; an over view”, European Economic review

Rufus, A and Anthony, E (2013), Testing for Competition in the Nigerian Commercial

Banking Sector, modern economy, Vol.4, No.7 (pp. 501-511)

Samuel, F (2013), “Banking Competition in Africa: Sub-regional Comparative Studies” Working

Study No. 13/12, University of Leicester working study, No.13/12

Shaffer, S (1993), “A Test of Competition in Canadian Banking”, Journal of Money, Credit and
Banking, Vol. 25, No. 1 (pp. 49-61).

Shaffer, S (1994), “Bank competition in concentrated markets”, Business Review, Federal

Reserve Bank of Philadelphia.

40
Simpasa, A (2013), “Competition and market structure in the Zambian banking sector”,

Working Study Series No. 168, African Development Bank, Tunis, Tunisia

Schumpeter, J.A (1934), “the theory of economic development”, Harvard university press
Thierry, B and Johan, M (2005), “Competition and Efficiency in Banking: Behavioral Evidence
from Ghana”, IMF working study, WP/05/17

Xavier, V. (2001), “Competition in the changing world of banking”, Oxford review of Economic
policy, Vol.17, No.4 (pp .537 -547).

Xavier, V (2010), “Competition and stability in banking”, University of Navarra working study,
wp-852.

Yildirim, H and Philippatos, C (2002), “Competition and contestability in Central and Eastern
European banking markets”, University of Tennessee Working Study.

Zerayehu, S, Kagnew, W and Teshome, K (2013), “Competition in Ethiopian Banking Industry”,

African Journal of Economics, Vol 5(1).

41
 APPENDIXES

Appendix 1: List of sampled banks

No. Name of the Bank Year of Establishment (G.C)

1 Commerial Bank of Ethiopia (CBE) 1963
2 Construction and Bussiness Bank (CBB) 1975
3 Awash International Bank (AIB) 1994
4 Dashen Bank (DB) 1995
5 Bank of Abyssinia (BoA) 1996
6 Wegagen Bank (WB) 1997
7 United Bank (UB ) 1998
8 Nib International Bank (NIB) 1999
9 Cooperative Bank of Oromia (CBO) 2004
10 Lion International Bank (LIB) 2006
11 Zemen Bank (ZB) 2008
12 Oromia International Bank (OIB) 2008
13 Bunna International Bank (BUIB) 2009
14 Berhan International Bank (BIB) 2009
15 Abay Bank S.C (AB) 2010
16 Addis International Bank (AdIB) 2011
17 Debub Global Bank S.C (DGB) 2012
18 Enat Bank S.C. (EB) 2012
Source: National Bank of Ethiopia

42
Appendix 2: Unit Root test Results

Appendix 2A: Unit Root test results for TI, II,TITA and IITA

Appendix 2A1: Unit Root test result for total income(ti)

Levin-Lin-Chu unit-root test for TI

-----------------------------------

Ho: Panels contain unit roots Number of panels = 8

Ha: Panels are stationary Number of periods = 32

------------------------------------------------------------------------------

Statistic p-value

------------------------------------------------------------------------------

Unadjusted t -8.9846

Adjusted t* -3.0727 0.0011

------------------------------------------------------------------------------

Appendix 2A2: Unit Root test result for interest income (II)

Levin-Lin-Chu unit-root test for II

-----------------------------------

Ho: Panels contain unit roots Number of panels = 8

Ha: Panels are stationary Number of periods = 32

------------------------------------------------------------------------------

Statistic p-value

------------------------------------------------------------------------------

Unadjusted t -6.7736

Adjusted t* -2.3423 0.0096

------------------------------------------------------------------------------

43
Appendix 2A3: Unit Root test result for total income/ total asset (TITA)

Levin-Lin-Chu unit-root test for TITA

-------------------------------------

Ho: Panels contain unit roots Number of panels = 8

Ha: Panels are stationary Number of periods = 32

------------------------------------------------------------------------------

Statistic p-value

------------------------------------------------------------------------------

Unadjusted t -20.7152

Adjusted t* -14.0112 0.0000

------------------------------------------------------------------------------

Appendix 2A4: Unit Root test result for interest income / total asset (IITA)
Levin-Lin-Chu unit-root test for IITA

-------------------------------------

Ho: Panels contain unit roots Number of panels = 8

Ha: Panels are stationary Number of periods = 32

------------------------------------------------------------------------------

Statistic p-value

------------------------------------------------------------------------------

Unadjusted t -17.1854

Adjusted t* -10.9503 0.0000

------------------------------------------------------------------------------

44
Appendix 2B: Unit Root test results for Price of Labor (PL), Price of Capital (PK) &

Price of Fund (PF)

Appendix 2B1: Unit Root test result for Price of Labor (PL)

Levin-Lin-Chu unit-root test for PL

-----------------------------------

Ho: Panels contain unit roots Number of panels = 8

Ha: Panels are stationary Number of periods = 32

------------------------------------------------------------------------------

Statistic p-value

------------------------------------------------------------------------------

Unadjusted t -12.1889

Adjusted t* -6.0905 0.0000

------------------------------------------------------------------------------

Appendix 2B2: Unit Root test result for Price of Capital (PK)

Levin-Lin-Chu unit-root test for PK

-----------------------------------

Ho: Panels contain unit roots Number of panels = 8

Ha: Panels are stationary Number of periods = 32

------------------------------------------------------------------------------

Statistic p-value

------------------------------------------------------------------------------

Unadjusted t -12.4906

Adjusted t* -5.8286 0.0000

------------------------------------------------------------------------------

45
Appendix 2B3: Unit Root test result for Price of Fund (PF)

Levin-Lin-Chu unit-root test for PF

-----------------------------------

Ho: Panels contain unit roots Number of panels = 8

Ha: Panels are stationary Number of periods = 32

------------------------------------------------------------------------------

Statistic p-value

------------------------------------------------------------------------------

Unadjusted t -6.0211

Adjusted t* -2.4441 0.0073

------------------------------------------------------------------------------

Table 2C: Unit Root test results for NPL to loan, Share of nii & share of bank asset to the industry

Table 2C1: Unit Root test result for NPL to loan

Levin-Lin-Chu unit-root test for NL

-----------------------------------

Ho: Panels contain unit roots Number of panels = 8

Ha: Panels are stationary Number of periods = 32

------------------------------------------------------------------------------

Statistic p-value

------------------------------------------------------------------------------

Unadjusted t -7.3349

Adjusted t* -2.6169 0.0044

------------------------------------------------------------------------------

46
Table 2C2: Unit Root test result for Share of non-interest income to total income

Levin-Lin-Chu unit-root test for SNII

-------------------------------------

Ho: Panels contain unit roots Number of panels = 8

Ha: Panels are stationary Number of periods = 32

------------------------------------------------------------------------------

Statistic p-value

------------------------------------------------------------------------------

Unadjusted t -8.1212

Adjusted t* -2.4133 0.0079

------------------------------------------------------------------------------

Table 2C3: Unit Root test result for share of bank asset to the industry asset

Levin-Lin-Chu unit-root test for STA

------------------------------------

Ho: Panels contain unit roots Number of panels = 8

Ha: Panels are stationary Number of periods = 32

------------------------------------------------------------------------------

Statistic p-value

------------------------------------------------------------------------------

Unadjusted t -7.1083

Adjusted t* -2.0983 0.0179

------------------------------------------------------------------------------

47
Appendix 3 (Model A): Trend of Bank Competition

(Model A1) (Model A2) (Model A3)

(2002/03- 08/09) (2008/09- 10/11) (2010/11 – 14/15)
Ti ti ti
Pl 0.365*** 0.208** 0.429***
(6.31) (2.64) (3.53)

Pk 0.161*** 0.0571* 0.147**

(5.08) (2.11) (3.10)

Pf 0.260*** 0.371*** -0.730**

(4.45) (4.33) (-2.76)

Lti 0.499***
(11.16)

Sta 0.435*** 0.833***

(10.56) (23.75)

Snii 0.229*** 0.496***

(3.65) (7.83)

Nl 0.0760***
(6.51)

d2 -0.268+ -0.314***
(-1.74) (-3.48)

_cons 3.572*** 14.91*** 9.155***

(5.15) (14.93) (5.20)
t statistics in parentheses
+
p< 0.10, *p< 0.05, **p< 0.01, ***p< 0.001

48
Appendix 4 (Model B): Model that used to test the type of product of II & NII

(Model B1) (Model B2)

(2002/03 – 2010/11) (2010/11- 2014/15)
ii ii
Pl 0.399*** 0.316+
(7.07) (1.82)

Pk 0.149*** 0.0923+
(4.80) (1.78)

Pf 0.295*** -0.566**
(5.17) (-2.75)

Lii 0.490*** 0.684***

(11.26) (15.21)

Sta 0.445***
(11.32)

Snii -0.0493 -0.670***

(-0.79) (-3.78)

_cons 4.504*** 1.235

(6.52) (0.53)
t statistics in parentheses
+
p< 0.10, *p< 0.05, **p< 0.01, ***p< 0.001

49
Appendix 5 (Model C): Model used to test misspecification of Empirical P-R model
(Model C1) (Model C2) (Model C3) (Model C4)
ii iita Ti tita
Pl 0.720*** 0.211** 0.413*** 0.0841
(6.93) (3.15) (3.30) (1.11)

Pk 0.0804+ 0.0780** 0.155** 0.135***

(1.77) (2.66) (3.21) (4.69)

Pf -0.203 0.624*** -0.677* 0.340**

(-0.79) (3.77) (-2.47) (2.60)

Snii -0.241* -0.237***

(-2.59) (-3.94)

Nl 0.0587+
(1.76)

d1 1.504**
(2.99)

d2 -0.321***
(-3.44)

_cons 10.09*** -3.512*** 9.184*** -4.706***

(6.53) (-3.51) (5.06) (-4.52)
N 140 140 140 140
R2 0.348 0.286
adj. R2 0.257 0.186
F 16.28 12.20
t statistics in parentheses
+
p< 0.10, *p< 0.05, **p< 0.01, ***p< 0.001

50
Appendix 6 (Model D): Model to see bank completion in private banks

(Early periods) (Recent periods)

ti ti
Pl 0.523*** 0.416***
(5.27) (5.35)

Pk 0.138*** 0.159***
(3.86) (3.39)

Pf -0.156** 0.294***
(-3.25) (3.64)

Sta 0.771*** 0.305***

(21.69) (3.90)

Nl 0.0640***
(5.53)

Lti 0.633***
(13.47)

Snii 0.298***
(3.37)

_cons 9.256*** 0.463

(11.48) (0.56)
t statistics in parentheses
+
p< 0.10, *p< 0.05, **p< 0.01, ***p< 0.001

51
Appendix 7 (Model E): The effect of NPL on interest and total income
(E1) (E2) (E3)
ti ti Ii
Pl 0.245*** 0.188* 0.323***
(3.40) (2.37) (5.83)

Pk 0.111** 0.0665* 0.137***

(2.98) (2.49) (4.71)

Pf 0.0740 0.356*** 0.362***

(1.15) (4.08) (7.18)

Lti 0.893***
(30.73)

Nl 0.0361 0.0723*** -0.140***

(0.99) (6.26) (-4.59)

Sta 0.875*** 0.518***

(37.93) (12.63)

Snii 0.494***
(7.62)

Lii 0.453***
(10.73)

_cons -1.265+ 14.62*** 6.363***

(-1.75) (14.78) (8.15)
t statistics in parentheses
+
p< 0.10, *p< 0.05, **p< 0.01, ***p< 0.001

52
Appendix 8 (Model F): Model estimated to see the impact of Credit cap and NBE Bill
Purchase on Banks income statement & competition

(Model F1) (Model F2)

ti ti
Pl 0.429*** 0.388***
(3.53) (3.75)

Pk 0.147** 0.0879*
(3.10) (2.45)

Pf -0.730** 0.207+
(-2.76) (1.79)

d2 -0.314*** -0.381+
(-3.48) (-1.72)

Sta 0.788***
(18.73)

Nl 0.0813***
(5.10)

d1 -0.0994
(-0.82)

_cons 9.155*** 14.09***

(5.20) (10.21)
t statistics in parentheses
+
p< 0.10, *p< 0.05, **p< 0.01, ***p< 0.001

53
i