Determinants of Interest Rate Spreads in Uganda's Commercial Banking Sector

DETERMINANTS OF INTEREST RATE SPREADS IN UGANDA’S COMMERCIAL BANKING SECTOR: A PANEL DATA ANALYSIS

ROBERT NABENDE

A DISSERTATION SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE AWARD OF A MASTER OF ARTS DEGREE IN ECONOMICS OF MAKERERE UNIVERSITY

DECEMBER, 2018

DECLARATION

CERTIFICATION

ACKNOWLEDGEMENTS

I acknowledge African Economic Research Consortium (AERC) for organising and sponsoring my studies at the Joint Facility for Electives (JFE) at Kenya School of Monetary Studies in which I took Econometrics and Monetary Economics, and these enriched the empirical techniques applied in this study. I express my sincere appreciation to my supervisors: Dr.

Ibrahim Mukisa and Assoc. Prof. Edward Bbaale for their positive and encouraging approach towards this research. Their supportive attitude and quick responses to any questions concerning this study have been invaluable. I also thank the faculty at the School of Economics,

Makerere University for their valuable advice during the course of my undergraduate and graduate studies. Am specifically grateful to Assoc. Prof. Bruno Lule Yawe, Dr. Joweria Teera,

Dr. Ibrahim Mukisa, Dr. John Bosco Nnyanzi, Dr. Francis Wasswa, and Dr. Asumani Guloba for introducing me to Advanced Economics at graduate level. I also acknowledge my JEF lecturers: Prof. Nathan Okurut of University of Botswana, Dr. Dianah Ngui Muchai of

Kenyatta University, Dr. Jean Marcelin Bosson Brou of University Felix Houphouet Boigny, and Dr. Tom Mwebaze of Makerere University. Support and encouragement from my fellow graduate students at Makerere University: Abel Egesa, Brian Musumba, Francis Muhoozi,

Innocent Okoth, Millicent Aswata, Perez Nicholas Ochanda, Ronald Mugobera, and William

Ssemyalo is acknowledged. The 2016 JFE colleagues from University of Botswana, University of Gold Coast, Ghana, University of Malawi, University of Mauritius, and University of

Zimbabwe are acknowledged. To my family members, am grateful for your support and encouragement. Above all I thank God for good health and wisdom which made the studies possible. Notwithstanding the above acknowledgements, I am solely responsible for errors and/or omissions contained in this dissertation.

iii

DEDICATION

To my: Grandfathers, James Nalwooli (RIP) and Clement Nakhosi (RIP);

Grandmothers, Deborah Khabuya Nalwooli and Pulukelia Namakoye Nakhosi;

Father, Naphtali Wakharere;

Mother, Margret Namono Wakharere;

Siblings;

Cousins and friends.

TABLE OF CONTENTS DECLARATION...... i CERTIFICATION ...... ii ACKNOWLEDGEMENTS ...... iii DEDICATION...... iv LIST OF TABLES ...... viii LIST OF FIGURES ...... ix ACRONYMS ...... x ABSTRACT ...... xi CHAPTER ONE ...... 1 INTRODUCTION...... 1 1.1. Background to the study ...... 1 1.2. Statement of the problem ...... 4 1.3. Objectives of the study ...... 5 1.3.1. Overall objective of the study ...... 5 1.3.2. Specific objectives of the study ...... 5 1.4. Hypotheses of the study ...... 6 1.5. Scope of the study ...... 6 1.6. Significance of the study ...... 7 CHAPTER TWO ...... 8 OVERVIEW OF UGANDA’S BANKING SYSTEM ...... 8 2.0. Introduction ...... 8 2.1. Pre-financial reform policies ...... 8 2.1.1. Interest rate policy ...... 9 2.1.2. Establishment of state-owned commercial banks, and administered lending programmes ...... 10 2.1.3. Private sector banks ...... 12 2.1.4. Prudential banking sector regulation and supervision ...... 13 2.1.5. Summary ...... 14 2.2. Financial sector reforms ...... 14 2.2.1. Financial liberalisation ...... 15 2.2.2. Restructuring of state-owned banks ...... 16 2.2.3. Strengthening of prudential regulation and supervision ...... 17 2.2.4. Summary ...... 18 2.3. Structure of Uganda’s banking sector ...... 18

2.3.1. Financial development ...... 19 2.3.2. Efficiency...... 22 2.3.3. Competition and concentration...... 24 2.3.4. Financial soundness indicators ...... 25 2.4. Summary to the chapter...... 28 CHAPTER THREE ...... 29 LITERATURE REVIEW ...... 29 3.0. Introduction ...... 29 3.1. Conceptualisation of interest rate spreads ...... 29 3.2. Theoretical literature on interest rate spreads ...... 30 3.3. Empirical literature ...... 33 3.4. Summary of the literature and research gap ...... 42 CHAPTER FOUR ...... 43 METHODOLOGY ...... 43 4.0. Introduction ...... 43 4.1. Theoretical framework ...... 43 4.2. Empirical model specification ...... 49 4.3. Data sources ...... 55 4.4. Estimation techniques ...... 56 4.5. Panel unit root tests ...... 57 CHAPTER FIVE ...... 58 PRESENTATION, INTERPRETATION, AND DISCUSSION OF RESULTS ...... 58 5.0. Introduction ...... 58 5.1. Data characteristics ...... 58 5.1.1. Descriptive statistics ...... 58 5.1.2. Pairwise correlation matrices of the variables ...... 62 5.1.3. Panel unit root tests ...... 64 5.2. Presentation, interpretation, and discussion of regression results ...... 65 CHAPTER SIX ...... 73 CONCLUSIONS AND RECOMMENDATIONS ...... 73 6.0. Introduction ...... 73 6.1. Conclusions of the study ...... 73 6.2. Policy recommendations ...... 74

6.3. Limitations of the study and recommendations for further research ...... 76 BIBLIOGRAPHY ...... 77 APPENDICES ...... 86 Appendix A: List of licensed commercial banks, credit institutions, and MDIs ...... 86 Appendix B: Financial inclusion indicators ...... 87 Appendix C: Derivation of the first order conditions under the Ho and Saunders bank dealership model ...... 88 Appendix D: Panel unit root tests-PP tests...... 90

vii

LIST OF TABLES

Table 1.1. Financial intermediation across countries (% average 2005-2015) ...... 2

Table 2.1. Number of licensed branches/outlets ...... 20

Table 2.2. Mobile money operations ...... 21

Table 2.3. Financial development indicators (2005-2015) ...... 22

Table 2.4. Commercial bank assets in UShs. billion ...... 25

Table 4.1. Data sources ...... 55

Table 5.1. Descriptive statistics ...... 59

Table 5.2. Correlation matrix of bank specific variables ...... 63

Table 5.3. Correlation matrix of market specific and macroeconomic variables ...... 63

Table 5.4. Panel unit root tests: ADF-Fisher type tests ...... 64

Table 5.5. Dynamic panel regressions: two-step system GMM ...... 66

viii

LIST OF FIGURES

Figure 2.1. Interest rates and consumer price inflation: 1981-1994 ...... 10

Figure 2.2. Efficiency in Uganda's Financial System ...... 23

Figure 2.3. Structure of interest rates (%) ...... 23

Figure 2.4. Market concentration and competition in Uganda's banking sector ...... 24

Figure 2.5. Commercial bank income in billion: 2005-2016 ...... 26

Figure 2.6. Decomposition of commercial banks' interest income: 2005-2016 ...... 27

ACRONYMS

ADF Augmented Dickey Fuller test

CBA Commercial Bank of Africa

CCA Caucasus and Central Asia

DFCU Development Finance Company of Uganda

EAC East African Community

GDP Gross domestic product

GMM Generalised method of moments

HHI Herfindahl-Hirschman index

LLC Levin, Lin and Chu unit root test

MDIs Microfinance deposit taking institutions

NPART Non-Performing Assets Recovery Trust

OLS Ordinary least squares

SSA Sub-Saharan Africa

UCB Uganda Commercial Bank

ABSTRACT

Liberalisation of financial markets is associated with a reduction in interest rate spreads.

However, interest rate spreads have persistently remained high in Uganda despite the financial liberalisation that was undertaken by the government in the early 1990s. High interest rate spreads point to the low level of efficiency and lack of competitiveness in the financial sector.

The study assesses determinants of interest rate spreads in Uganda’s commercial banking sector. The analysis is based on the Ho and Saunders (1981) model and its subsequent extensions by Maudos and Fernandez de Guevara (2004). Panel data estimation techniques, specifically system generalised method of moments (GMM), are applied. We find that among the bank specific factors, interest rate spreads increase with increase in credit risk, liquidity risk, and capital adequacy ratio. Bank size is shown to be negatively related to interest rate spreads. For banking industry specific factor, foreign bank participation in the loans markets is associated with higher interest rate spreads. And for macroeconomic factors, high inflation rates are shown to be associated with high bank spreads, whilst high real gross domestic product (GDP) growth rates and broad money supply to GDP (M2/GDP) are associated with lower interest rate spreads. Going forward, banks and government should devise mechanisms to encourage loan repayment, and banks should be encouraged to reduce on holding excess liquid assets. At a macro-level, Bank of Uganda should maintain its stance on curbing inflation.

Economic growth and financial development should as well be encouraged.

Key Words: Interest rate spreads; bank specific factors; industry specific factors; macroeconomic factors; dynamic panel; Uganda’s commercial banking sector.

JEL Classifications: C23; E43; E44; G21; L11

CHAPTER ONE

INTRODUCTION

1.1. Background to the study

Proponents of financial market liberalisation argue that removal of financial market restrictions in developing countries reduces the cost of financial intermediation.1 Consequently, the reduction in the cost of intermediation translates into increased credit demand and supply, thus fostering financial development and economic growth (Beck & Hesse, 2009; Claessens, Demirguc-Kunt & Huizinga, 2001; Crowley,

2007; McKinnon, 1973; Shaw, 1973). This argument is premised on the understanding that financial liberalisation—which is often characterised by an increase in the number of local and foreign financial institutions—enhances competition and financial deepening, leading to lower interest rate spreads and thus efficiency in the financial sector.

In that regard, Uganda embarked on financial market liberalisation, as part of the country’s wider financial reform programme, in the early 1990’s to stimulate competition and enhance efficiency in financial markets. Liberalisation of financial markets was effected through a number of reforms which were implemented for over a decade. These reforms included the easing of entry requirements and privatisation of Uganda Commercial Bank (UCB)—the largest and government owned commercial bank—in 2002. As a result, there has been an increase in the number of financial institutions as well as a reduction in the level of government involvement in the banking sector (Beck & Hesse, 2009; Mugume, Apaa & Ojwinya, 2009;

Nampewo, 2013).2 Other reforms included liberalisation of interest rates in 1994; removal of credit ceilings and direct credit facilities towards crop finance; removal of restrictions on dealing in foreign exchange and treasury bills; and introduction of shilling interbank money market and the rediscount

1 The cost of financial intermediation is often proxied by interest rate spreads—the difference between the interest rate charged to borrowers and the rate paid to depositors. 2 As of June 2017, there were 24 banks, 5 microfinance deposit taking institutions (MDIs), and 4 credit institutions. See Appendix A for the list of financial institutions. 1 facility in 1993 (Egesa, 2010; Nampewo, 2013; Nannyonjo, 2001). The reforms have undoubtedly made

Uganda’s financial sector one of the most liberalised financial sectors in Sub-Saharan Africa (SSA).

However, interest rate spreads—a measure of competition and efficiency—have remained high in

Uganda’s banking sector relative to the regional and global averages despite liberalisation of the financial markets. Between 2005 and 2015, Uganda’s interest rate spreads averaged 10.4 per cent compared to averages of 8.5 per cent, 7.4 per cent, and 8.6 per cent in other East African Community (EAC) countries of Kenya, Tanzania, and Rwanda respectively.3 Furthermore, the average interest rate spreads were higher than average spreads of 7.5 per cent in SSA, 8.1 per cent in low-income countries, and 5.7 per cent world average during the same period. Similarly, Uganda’s net interest margins—banks' net interest revenue as a ratio of total interest earning assets—have consistently been higher than the regional and global averages

(see Table 1.1).4

Table 1.1. Financial intermediation across countries (% average 2005-2015)5

Interest rate Net interest Bank Private sector

spread margin deposits/GDP credit/GDP Uganda 10.4 11.3 14.3 10.2 Kenya 8.5 8.2 37.1 29.4 Tanzania 7.4 7.6 17.6 10.8 Rwanda 8.6 9.7 13.6 13.5 Burundi − 8.1 18.3 15.5 SSA 7.5 6.9 − 54.0 Low Income Country 8.1 6.0 − 15.5 World Average 5.7 − − 123.8 Source: Data are from the June 2017 Updated Version of Financial Development and Structure dataset of Demirguc-Kunt et al. (2017)

Interest rate spreads are of concern to policymakers because they, as aforementioned, reflect the cost of financial intermediation. High interest rate spreads point to the low level of efficiency and lack of competitiveness in the financial sector, which adversely affect domestic real savings and investment, and

3 Data on interest rate spreads for Burundi and South Sudan is not available. 4 Net interest margins are an ex-post definition of interest rate spreads. 5 Interest rate spread is the difference between the lending rate and the deposit rate. Net interest margin is the net interest revenue relative to total earning assets. Bank deposits/GDP is total deposits in deposit money banks as share of GDP. Private credit/GDP is total claims of financial institutions on the domestic private non-financial sector as share of GDP.

2 thus undermine the level of economic growth (Almarzoqi & Naceur, 2015; Bank of Uganda, 2015;

Folawewo & Tennant, 2008; Mugume et al., 2009; Mujeri & Younus, 2009). More specifically, real savings and investment are adversely affected by low deposit rates and high lending rates that characterise the financial sector under situations of high interest rate spreads. Low deposit rates, on one hand, discourage deposits hence limiting resources available to finance bank credit. High lending rates, on the other hand, discourage borrowing especially for long term investment. This argument, if put into context, suggests that Uganda’s low bank deposit to gross domestic product (GDP) and private sector credit to

GDP ratios relative to other EAC countries are partly attributed to low deposit rates and high lending rates respectively.6

Relatedly, high interest rate spreads often lead to channelling of deposits into less productive investments in the economy (Crowley, 2007). Due to adverse selection and moral hazard, high lending rates attract high risky borrowers which in turn exacerbates credit risk in the banking sector (Stiglitz & Weis, 1981).

This argument could be inferred for Uganda’s banking sector whose credit risk—as measured by the level of non-performing loans—is still high relative to the regional and global averages. For instance, Uganda’s banking sector non-performing loans stood at UShs. 803.9 billion, translating into a non-performing loans to total gross loans ratio of 8.3 per cent for the year ending June 2016 (Bank of Uganda, 2016). This ratio was higher than the SSA and global average non-performing loans to total gross loans ratios of 6.3 per cent and 4.4 per cent respectively (World Bank, 2017). Moreover, the value of Uganda’s banking sector non-performing loans in 2016 accounted for 165.5 per cent of the total banking sector profits for the year

(Bank of Uganda, 2016). Strictly speaking, such high levels of non-performing loans undermine banking sector profitability and stability as well as discourage credit extension to the private sector.

From the preceding discussion, it is clear that high interest rate spreads have adverse effects on the level of financial development and growth prospects of the economy. Moreover, there has been broad voicing of concerns about the high interest rate spreads in both public and policy fora. Some sections of the public

6 See Table 1.1 for bank deposit to GDP and private sector credit to GDP ratios

3 have called for government regulation of both deposit and lending rates. Such calls have largely been influenced by the enactment of an Act of Parliament—Banking (Amendment) Act, 2016—to regulate interest rate spreads in Kenya’s banking sector. In spite of such calls, there is thin empirical literature on the determinants of interest rate spreads in Uganda to guide policy direction. Notable studies include Beck and Hesse (2009), Mugume et al. (2009), Nampewo (2013), and Nannyonjo (2001). These studies generally use either time series data analysis (Nampewo, 2013), or static panel data analysis (Beck &

Hesse, 2009; Mugume et al., 2009; Nannyonjo, 2001).

Studies that use time series data analysis—though account for variation of interest rate spreads over time— do not account for variation of interest rate spreads across banks, whilst studies that use static panel data analysis—though account for variation of interest rate spreads over time and across bank—do not account for the impact of interest rate spreads in the previous periods on the current spreads. Yet empirical evidence has shown that interest rate spreads significantly vary across banks and depend on spreads in previous periods (see for example: Carbo & Rodriguez, 2007; Folawewo & Tennant, 2008). Moreover, the existing literature is not certain on the determinants of interest rate spreads given that some studies attribute high interest rate spreads to bank specific, industry specific, and macroeconomic variables (Beck

& Hesse, 2009; Mugume et al., 2009; Nannyonjo, 2001), whilst other studies attribute high interest spreads to macroeconomic variables (Nampewo, 2013). It is against such a background that the current study examines determinants of interest rate spreads in Uganda’s commercial banking sector.

1.2. Statement of the problem

Research shows that liberalisation of financial markets improves competitiveness and efficiency of the financial sector, thus leading to reduction in interest rate spreads. Consequently, the reduction in spreads increases credit demand and supply, which in turn fosters financial development and economic growth

(Claessens et al., 2001; McKinnon, 1973; Shaw, 1973). However, Uganda’s interest rate spreads have persistently remained high compared to the regional (EAC and SSA) and global averages in spite of the financial liberalisation. The combination of low deposit rates and high lending rates that has kept Uganda’s

4 interest rate spreads high has been detrimental to the quest for financial development and economic growth. More specifically, low deposit rates have discouraged saving, whilst high lending rates have discouraged borrowing. The available literature on interest rate spreads in Uganda—such as Beck and

Hesse (2009), Mugume et al. (2009), Nampewo (2013), and Nannyonjo (2001)—is not current and it does not consider the impact of interest rate spreads in the previous periods on current spreads as reported in some literature. If financial liberalisation leads to lower interest rate spreads—but to the contrary Uganda’s interest rate spreads are still high—then more should be known about the determinants of interest rate spreads in Uganda’s financial sector so as to inform policy formulation. Therefore, the purpose of this study is to examine the determinants of interest rate spreads in Uganda’s commercial banking sector.

1.3. Objectives of the study

1.3.1. Overall objective of the study

To analyse determinants of interest rate spreads in Uganda’s commercial banking sector.

1.3.2. Specific objectives of the study

The specific objectives of the study are:

(i) To analyse the effects of bank specific factors—including credit risk, liquidity risk, operating

costs, return on assets, capital adequacy ratio, bank size, and non-interest income—on interest rate

spreads in Uganda’s commercial banking sector;

(ii) To assess the effects of banking industry specific factors—including Herfindahl-Hirschman

indices (HHI) and foreign bank participation—on interest rate spreads in Uganda’s commercial

banking sector; and

(iii)To find out the effects of macroeconomic factors—including inflation rate, 91-day treasury bill

rate, exchange rate volatility, real GDP growth rate, and broad money supply to GDP (M2/GDP)

—on interest rate spreads in Uganda’s commercial banking sector.

1.4. Hypotheses of the study

The testable hypotheses of the study are that:

(i) Bank specific factors: credit risk, liquidity risk, operating costs, return on assets, and capital

adequacy ratio positively affect interest rate spreads, whilst bank size and non-interest income

have a negative impact on interest rate spreads.

(ii) Industry specific factors: HHI and foreign bank participation in the loans market are positively and

negatively related to interest rate spreads respectively.

(iii) Macroeconomic factors: inflation rate, 91-day treasury bill rate, and exchange rate volatility are

directly proportional to interest rate spreads, whilst real GDP growth rate and M2/GDP are

inversely proportional to interest rate spreads.

1.5. Scope of the study

The study covers 24 commercial banks licensed and regulated by Bank of Uganda, the Central Bank of

Uganda, as of 2015 for the period 2005−2015.7 New banks that were licensed during the period of study are included in the study, whilst banks that exited the industry are excluded. The inclusion of only commercial banks in the study does not negate the fact that Uganda’s formal banking system is composed of other financial institutions. In fact, Uganda’s formal banking system contains commercial banks (Tier

1), credit institutions (Tier 2), and microfinance deposit-taking institutions (MDIs) (Tier 3). However, banks still play a dominant role as far as intermediated funds are concerned. On the other hand, Tier 2 institutions are specialised financial institutions whose spreads might not be comparable to those of banks

(Beck & Hesse, 2009). Furthermore, the role of MDIs is negligible in financial intermediation and they are not homogeneous to commercial banks for them to be included in the same panel. Thus, the study focuses on only commercial banks.

7 See Appendix A for the list of commercial banks

1.6. Significance of the study

As noted in section 1.1, there is thin empirical research on the determinants of interest rate spreads in

Uganda. The available studies generally use either time series data analysis (Nampewo, 2013), or static panel data analysis (Beck & Hesse, 2009; Mugume et al., 2009; Nannyonjo, 2001). More specifically,

Nampewo (2013) uses an Error Correction Model on macroeconomic variables to examine the determinants of interest rate spreads for the period 1995-2010. Though this study accounts for variation of interest rate spreads over time, it does not account for variation of interest rate spreads across banks.

Consequently, the impact of bank behaviour is not considered in the examination of the determinants of interest rate spreads.

For the studies that use panel data, Beck and Hesse (2009) analyse determinants of interest rate spreads using Pooled Ordinary Least Squares (OLS) and Fixed Effects model on a panel of 16 banks between the second quarter of 1999 and the second quarter of 2005, and Mugume et al. (2009) use Random Effects

Model on a panel of 14 banks for a period of 13 years (1995-2007). In addition, Nannyonjo (2001) uses

Fixed Effects Model on data of 17 banks for the period 1994-1998. All these studies—though account for variation of interest rate spreads over time and across bank—do not account for the impact of interest rate spreads in the previous periods on the current spreads. Yet empirical evidence has shown that interest rate spreads in a given period also significantly depend on spreads in the previous period (Carbo & Rodriguez,

2007; Folawewo & Tennant, 2008).

Therefore, as a major contribution to literature, the study uses dynamic panel estimation techniques and recent data from all banks, except Exim Bank-Uganda, to examine determinants of interest rate spreads in

Uganda’s banking sector. The use of dynamic panel estimation techniques is intended to capture the impact of interest rate spreads in the previous periods on the current spreads. By doing something new, the results of this study can invaluable benefit policy makers, more so in the face of the current critical debate of regulated versus liberalised interest rates.

CHAPTER TWO

OVERVIEW OF UGANDA’S BANKING SYSTEM

2.0. Introduction

This chapter provides a background to policies and developments in Uganda’s banking sector in order to put the study into perspective. Section 2.1 discusses the effects of pre-financial liberalisation policies on

Uganda’s financial sector. Financial reforms are discussed in section 2.2. The current structure of

Uganda’s banking sector is discussed in section 2.3, and a brief summary and conclusions to the chapter are highlighted in section 2.4.

2.1. Pre-financial reform policies

Uganda’s commercial banking sector dates back to 1906 when the first commercial bank—the foreign- owned National Bank of India which was later renamed Grindlays Bank—was established. The bank established its first branch in Entebbe, the country’s administrative capital by then, and the second branch four years later in Kampala. Other banks that were established, before independence in 1962, included Standard Bank in 1912; Barclays Bank in 1927; Bank of Baroda and Bank of India in 1953; and the Bank of the Netherlands, which later merged with Grindlays Bank, in 1954. Just like Grindlays

Bank, all these banks were foreign-owned (Bategeka & Okumu, 2010; Beck & Hesse, 2009, Mugume,

2010).

Foreign-owned banks in Uganda, as it was the case in other newly independent African countries, were widely criticised for only lending to foreign owned companies and those owned by non-African resident communities. The widespread criticism of these foreign banks’ lending policies strengthened the belief, by then, that government intervention was necessary if the banking sector was to be instrumental in the development of local enterprises and the economy (Brownbridge, 1998; Mugume,

2010). As a result, successive government after independence pursued interventionist financial policies that were intended at controlling the banking sector, ostensibly for promotion of lending to local

8 enterprises and other non-commercial objectives. The policies included regulation of interest rates, establishment of state owned banks, government purchase of shares in the foreign banks, and establishment of various administered lending programmes (Beck & Hesse, 2009; Brownbridge, 1998;

Mugume, 2010). These policies and their impact on the financial sector are the subject of discussion in the succeeding sub-sections: 2.1.1, 2.1.2, and 2.1.3. To give perspective to banking operations, the performance of banks and prudential regulation during the pre-financial reforms period are discussed in sub-sections 2.1.4 and 2.1.5 respectively.

2.1.1. Interest rate policy

Before 1992, the level and structure of interest rates were determined by the Bank of Uganda. The policy of interest rate controls was based on the belief that the cost of credit had to be kept low to encourage investment and to subsidise favoured borrowers (Mugume, 2010). In addition, exchange controls were also used, most notably, to force residents to invest their savings in domestic assets. However, regulation of interest rates substantially limited the role of market mechanisms in mobilisation of deposits and allocation of such deposits to investors. The policy also kept the nominal interest rates well below the rate of inflation for most of the 1970s and 1980s. Inflation, for instance, averaged 103 per cent during 1981-

90, while nominal lending rates for commerce averaged 31 per cent and time deposit rates averaged 24 per cent (see Figure 2.1). Thus, the real interest rates were substantially negative, on average, during the

1970s and 1980s given the high inflation rates (Brownbridge, 1998; Mugume, 2010).

Negative real returns on monetary assets contributed to the steep decline in bank deposits and thus the financial depth of the economy. Indeed, between 1970 and 1990, the M2/GDP ratio fell from 18 per cent to 7 per cent, while the share of bank deposits in M2 fell from 65 per cent to 59 per cent (Brownbridge,

1998). In addition, there was portfolio shift from financial to tangible assets and substitution of domestic formal financial assets for foreign currency denominated assets, which subsequently spurred capital flight, and constrained the availability of credit (Kasekende & Atingi-Ego, 2003). Moreover, the attractiveness of bank deposits for the Ugandan public was further eroded by inefficient payments system operated by

9 banks, hence cash was needed to effect most transactions (Harvey, 1993). Relatedly, public confidence in holding of financial assets is also likely to have been undermined by the demonetisation exercise in 1987, which imposed a tax of 30 per cent on holdings of currency, bank deposits, and some other financial assets in an attempt to reduce liquidity in the economy. However, it should be noted that most of the damage in terms of the reduction in financial depth had already occurred by time of demonetisation (Brownbridge,

1998).

Figure 2.1. Interest rates and consumer price inflation: 1981-1994

200.0% Lending Rate: Commerce

180.0% Lending Rate: Agricultural Development 160.0% Time Deposit Rate: 7-12 Months

140.0% Bank Rate

120.0% Inflation

100.0%

80.0%

60.0%

40.0%

20.0%

0.0% 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994

Source: Values adapted from Brownbridge (1998: 19) Note: Bank Rate is the rate at which the Bank of Uganda lends to commercial banks

2.1.2. Establishment of state-owned commercial banks, and administered lending programmes

As far as establishment of public banks is concerned, an Act of Parliament—the Uganda Commercial

Bank Act, 1965—was enacted to transform the then Uganda Credit and Savings Society into the first state- owned bank, Uganda Commercial Bank (UCB). UCB was followed with the formation of Cooperative

Bank, another state-owned bank, in 1972. These banks aggressively expanded their branch network in a move which was mostly based on political rather than commercial grounds. This move was further

10 fostered in the 1970s when foreign banks were forced to close their upcountry branches or sell them to

UCB, and all government business was transferred from foreign banks to UCB (Beck & Hesse, 2009;

Nampewo, 2013). Consequently, UCB and Uganda Cooperative Bank virtually acquired monopoly on banking everywhere outside the Capital, Kampala. In fact, by the early 1990s, UCB had 190 of the 237 bank branches in the country, while Cooperative Bank with 24 branches had the second largest branch network. (Brownbridge, 1998; Mugume, 2010).

UCB and Cooperative Bank lending policies were heavily influenced by the government. Both banks were largely expected, given the reasons for their formation, to provide credit to priority sectors of the economy to facilitate government development programmes. For instance, a variety of administrated lending programmes including, among others, the administered agricultural and rural lending schemes were established in the 1960’s to promote the development of the agricultural sector and rural areas respectively (Mugume, 2010). However, government intervention negatively impacted on the performance of these state owned banks in particular and the financial sector at large. These banks were severely mismanaged, partly, due to lack of competition in financial sector, availability of automatic liquidity support from Bank of Uganda, politicisation of their management and lending, and the lack of proper accounting procedures, thus leading to their insolvency. Moreover, due to a policy of directed credit, government guaranteed loans were often regarded as grants, thus negatively affecting repayment rates which further exacerbated the volume of non-performing loans (Harvey, 1993). For instance, UCB’s non-performing loans accounted for around 75 per cent of its total loan portfolio, and its loan recovery rates on the administered lending schemes were below 50 per cent (Brownbridge,

1998).

In addition, redistributive lending policies made credit available to politically favoured classes of borrowers at concessionary interest rates, leading to the politisation of the banking sector. The lack of competitiveness in the financial system also resulted into excessive rent seeking and low levels of innovation, which severely impaired the credibility of financial institutions and limited the ability to

11 render quality services respectively (Mugume, 2010). Overall, the poor performance of state owned banks is attributable to the government interventionist financial policies that led to their gross mismanagement, though protracted economic crisis, disruption caused by war, and the weak legal system made the business climate for the banks in Uganda very difficult and undoubtedly contributed to their insolvency.

2.1.3. Private sector banks

The dominance of the foreign banks—namely Bank of Baroda, Libyan Arab Bank, Standard Chartered, and Grindlays—was sharply reduced in the 1970s through the compulsory government acquisition of 49 per cent stake in the banks and takeover of most of their branches. Foreign banks’ operations were mainly confined to Kampala and their share of commercial bank deposits fell to 30 per cent (Brownbridge, 1998;

Mugume, 2010). Contrary to state owned banks though, foreign banks’ asset management remained conservative: their liquidity ratios remained high and their lending was generally concentrated on larger companies, on trade finance, and syndicated loans to the crop marketing boards, such as Coffee Marketing

Board, to finance crop purchasing (Brownbridge, 1998).

In addition to state owned and foreign banks, local private sector investors—mainly business people with interests in trading, manufacturing, agriculture, and real estate—established local commercial banks during the second half of the 1980s and the early 1990s. Some of the locally owned private sector banks started as credit institutions before converting to banks. Their lending was mainly short term and focused on traders and manufacturers. Moreover, unlike foreign banks, the banks extended credit services to small scale businesses, and offered longer opening hours (Brownbridge, 1998).

However, local banks faced a number of difficulties in establishing their presence in the banking sector.

Lending was constrained by shortage of suitable collateral other than real estate in Kampala. In addition, loan recovery was hard given the weaknesses and delays in the court system. There was also severe scarcity of qualified and experienced staff in the 1980’s and early 1990’s (Brownbridge, 1998). The banks also faced liquidity problems, partly, because of the shallow depth of the financial system and the less developed interbank market. In fact, most local private banks were insolvent and required large overdrafts

12 from Bank of Uganda to remain liquid in the 1990’s (Beck & Hesse, 2009; Nampewo, 2013). Due to financial distress, Bank of Uganda, using its powers under the 1993 Financial Institutions Statute, closed the Teeffe Bank in 1994, and in April 1995 took over the Nile and Sembule Banks. Other local private banks including Greenland Bank, Gold Trust Bank, and Trust Bank were also closed during the period

1998 to 2000.

The financial fragility of the local banks was, largely, the result of bad debts arising from imprudent credit practices such as insider lending, lending without proper security, and very weak internal controls.

Imprudent management in the distressed banks was exacerbated by the lack of sufficient banking experience by managers and boards of directors, and limited separation of roles of managers and shareholders, hence the pressure to extend loans to the businesses of the latter. The banks were also severely undercapitalised, in part, because of the acute shortages of capital and absence of capital markets in Uganda by then (Brownbridge, 1998; Mugume, 2010).

2.1.4. Prudential banking sector regulation and supervision

Prudential regulation and supervision were not accorded much emphasis by the government and Bank of

Uganda in particular during the pre-reform period. Regulatory and supervisory capacities in Bank of

Uganda were too weak for it to discharge many of the functions it was assigned by the 1969 Banking Act.

Returns provided by banks to Bank of Uganda were inadequate for offsite inspection, while onsite inspections were not conducted regularly. In addition, bank supervision was complicated by differences in accounting and auditing practices among the banks (Brownbridge, 1998; Mugume, 2010). Furthermore,

UCB consistently ignore regulations, such as the liquidity ratios at the watch of Bank of Uganda. UCB defiance could, in part, be attributed to the government financial policy direction at the time. In fact, it could have been difficult for the central bank to overrule the activities of state owned banks through regulation and supervision when the latter were claiming to be acting in interest of government development policies. Moreover, supervision of government banks to prevent their failing could have been unnecessary given that the government was expected to bail them in case of financial distress. Relatedly,

13 prudential regulation and supervision were also neglected by foreign banks, in part, because they were subsidies of well-established and reputable foreign banks which had their own prudential management rules (Mugume, 2010). As a consequence, inefficient regulation and supervision encouraged imprudent management of banks, especially among local banks, which led to the fragility of the banking sector.

2.1.5. Summary

In a nut shell, government’s policies were intended at having a directed rather than regulated financial system. These policies had disastrous effects on the financial sector during the pre-reform period.

Financial assets were disregarded by the public in preference for material assets, leading to shallowing of the financial sector. The balance sheets of the financial institutions deteriorated as their capital bases eroded in the face of large loan losses. The scope of financial services remained severely restricted, which in turn led to the concentration of financial services in the hands of a few commercial banks, the largest being the UCB. Moreover, the inefficient regulation and supervision, did not save banks from imprudent management and the subsequent fragility of the financial sector.

2.2. Financial sector reforms

Due to the increasing economic difficulties, and influenced by the worldwide trend toward adjustment policies, the government embarked on financial reforms as part of its wide-ranging structural reform programmes. The implementation of financial sector reforms started in 1991 with the support of a World Bank and

International Monetary Fund (IMF). The reforms had three major elements: institutional reforms to the

Bank of Uganda and state owned banks, legislative changes to the banking laws and the Bank of Uganda

Act, and financial liberalisation. Financial reforms in Uganda, as in other developing countries, were specifically aimed at strengthening monetary policy, boosting deposit mobilisation, stimulating competition in financial markets, enhancing the efficiency in the financial sector, restructuring insolvent banks, improving prudential regulation and supervision, and promoting the diversification of financial markets (Brownbridge, 1998). This section discusses financial reforms that more directly relate to the

14 banking sector. Such reforms include financial liberalisation, restructuring of state owned banks, and strengthening of banking sector prudential regulation and supervision.

2.2.1. Financial liberalisation

Financial liberalisation was effected through a number of reforms that were aimed at increasing competition and efficiency in the financial sector. To start with, restriction to entry into the sector was lifted, leading to establishment of a number of local private financial institutions. The entry of the local banks injected some competition into banking markets though it was mainly confined to retail banking markets in Kampala. However, further growth in the financial sector was limited by distress that afflicted some banks in the sector. In fact, the banking sector experienced a systemic crisis from 1994 to 2003 due to lack of bank capital in the system which forced Bank of Uganda to later place a moratorium on entry of new financial institutions in 1996 (Beck & Hesse, 2009; Clarke, Cull & Fuchs, 2006; Nampewo, 2013).

In addition, there was closure, restructuring, takeover, and merger of some distressed banks. These measures, in addition to strengthening of regulation and supervision by Bank of Uganda, have restored stability in the banking sector.

In addition, a treasury bill auction was introduced in 1992, and interest rates were further liberalised in

1994 when the formal link with treasury bill rates was removed. The liberalisation of interest rates enabled positive real lending rates to be achieved without a sharp rise in nominal interest rates. For instance, lending rates for commerce fell from 41 per cent in December 1991 to 22 per cent in December 1994, while real savings and real time deposit rates were positive in 1992/93 when inflation fell sharply

(Brownbridge, 1998; Nannyonjo, 2001). Real interest rate, especially lending rates, have remained positive to date. Furthermore, there was, as part of financial liberalisation, the removal of restrictions on foreign exchange markets and trading in treasury bills that were previously imposed on commercial banks’ operations and asset holdings. Moreover, credit limits that were previously imposed by government were also removed. Removal of these restrictions led to greater diversification of bank operations and assets

(Nampewo, 2013; Nannyonjo, 2001).

2.2.2. Restructuring of state-owned banks

Restructuring of state owned UCB was a key component of the financial reforms given its near monopoly status in the market. A major restructuring programme for the UCB was initiated in 1992 with the aim of restoring the bank to commercial viability. The programme involved restructuring the UCB’s balance sheet to restore the bank to solvency; reducing operating costs to prevent further operating losses; re- organising lending procedures and internal controls to ensure that bad debts incurred from new lending were kept at acceptable levels; and insulating the bank from political pressure that had compromised commercial principles (Brownbridge, 1998).

A Non-Performing Assets Recovery Trust (NPART) was established in 1995 to which UCB’s non- performing loans were transferred. NPART was given extensive legal powers to pursue recovery of these bad debts, although most of the loans were not recovered. In return for the transfer of the non-performing loans, UCB received government bonds with a face value equivalent to the loans transferred. The government also injected almost UShs. 10 billion (approximately US$10 million by then) of additional equity capital into UCB. Hence the cost of restructuring UCB’s balance sheet, not taking into account any loans recovered by NPART, was approximately UShs. 90 billion (around UShs. 80 billion of non- performing loans plus the capital injection): this was equivalent to about $100 million or about 2 per cent of GDP. Moreover, efforts were also made to reduce the operation costs of the bank. Staff levels were significantly reduced by 48 per cent by 1995, while the branch network, which had totalled 190 branches, was cut to 85 branches and 53 agencies. As a consequence, UCB reduced operating expenses by about 25 per cent in nominal terms and generated an operating profit in 1994/95. The management of UCB was also reorganised with the appointment of a new board of directors and foreign technical advisors to the senior management (Brownbridge, 1998; Mugume, 2010).

UCB was subsequently privatised to a Malaysian investor after recapitalisation in 1998. However, subsequent insider trading and imprudent lending caused deterioration of the bank's loan portfolio and in

1999 Bank of Uganda intervened and renationalised it. In 2002, the South African Standard Bank (Stanbic

Bank) acquired 80 per cent of UCB's shares, with the remaining 20 per cent held by government for its

(UCB) employees. On the other hand, Cooperative Bank was closed in 1999 due to under capitalisation, imprudent management, and accumulation of non-performing loans (Beck & Hesse, 2009; Clarke et al.,

2006; Nampewo, 2013). The privatisation of UCB and closure of Cooperative Bank has significantly reduced government involvement in commercial banking.

2.2.3. Strengthening of prudential regulation and supervision

Following the mismanagement of banks especially government owned banks and the subsequent under capitalisation of the banking sector in the 1980s and 1990s, strengthening of prudential regulations and supervision was crucial for the stability of the financial sector. This process involved the enactment of new legislation to replace the 1969 Banking Act, and upgrading supervisory capacities in the Bank of

Uganda. The Banking Act, 1969 had become outdated and its provisions were deficient in respects of prudential regulation and supervision (Brownbridge, 1998). To that effect, government enacted the

Financial Institutions Statute, 1993 to rectified most of the defects in the Banking Act, 1969.8 The Statute gave Bank of Uganda more independence in licensing and regulating financial institutions and was more flexible than the 1969 Act as it gave Bank of Uganda authority to issue prudential regulations in relation to, among others, capital adequacy, liquidity, and data to be supplied by banks for supervisory purposes.

In addition, the Statute specified the minimum paid up capital to set a bank; specified minimum ratios for core capital and total capital to risk adjusted assets following the Basle Acord at 4 per cent and 8 per cent respectively; and imposed restrictions on insider lending, large credit exposures and investment in non- bank business such as real estate. In addition, exposure to a single borrower or group of borrowers with a common interest was also restricted. (Brownbridge, 1998; Mugume, 2010; Nampewo, 2013; Nannyonjo,

2001). Furthermore, the Bank of Uganda Statute, 1993 was also enacted. The new central bank statute clarified the role of Bank of Uganda as the regulator and supervisor of the banking system and provided

Bank of Uganda with a range of options of dealing with financial institutions acting imprudently and those

8 The Financial Institutions Act, 1993 was replaced with the Financial Institutions Act, 2004 which has since then been amended with latest being the 2016 amendments

17 that become insolvent. In addition, the MDIs Act of 2003 provided a framework for regulation and supervision of MDIs (Mugume, 2010).

In addition to the legislative changes, supervisory capacities at the Bank of Uganda have been strengthened. The Bank Supervision Department has been re-organised into two separate departments dealing with banks and other financial institutions. The submission of regular information from financial institutions to the Bank of Uganda has improved and onsite inspections are undertaken regularly.

Moreover, Bank of Ugandan authorities have also significantly strengthened bank regulation and supervision, through the introduction of a risk-based approach to bank supervision, and tightened loan classification and provisioning standards which have made the banking sector less fragile. The clean-up of some small and weak banks in the late 1990’s and early 2000’s through takeovers, mergers and in some instances closure also helped to make the banking system well capitalised, profitable, and resilient (Beck

& Hesse, 2009; Mugume et al., 2009).

2.2.4. Summary

Overall, financial reforms have improved the performance and stability of the financial sector. This has largely been achieved through the privatization of UCB, one of the key stability risks that faced the system due to large non-performing assets and imprudent management; the clean-up of some weak banks from the banking system; improvement in banking supervision with introduction of risk-based approach and passage of the new

Financial Institutions Act; and the presence of reputable banks that appear to be well capitalized, profitable, and resilient. With a sufficiently strong capital base, profits, effective management, good corporate governance, and well-designed systems and controls, the system is well placed to provide a growing contribution to the development of the economy.

2.3. Structure of Uganda’s banking sector

Financial structure is defined in terms of the aggregate size of the financial sector, its sectoral composition, and a range of attributes of individual sectors that determine their effectiveness in meeting users’ requirements. Indicators of financial structure include system-wide indicators of size, breadth, and

18 composition of the financial system; indicators of key attributes such as competition, concentration, efficiency, and access; and measures of the scope, coverage, and outreach of financial services (World

Bank & International Monetary Fund, 2005). This section discusses the structure of Uganda’s banking sector in relation to these indicators. In addition, sub-section 2.3.5 discusses the performance of Uganda’s banking sector.

2.3.1. Financial development

Uganda’s financial sector, in comparison to the financial pre-reform period, has substantially developed, more especially in terms of the number of financial institutions. Currently, the formal banking system comprises of 24 commercial banks (Tier 1), which forms a dominant part of the financial system in terms of intermediated funds; four credit institutions (Tier 2); and five MDIs (Tier 3).9 In addition to the banking sector, the formal financial system also consists of a pension fund—National Social Security Fund, 29 insurance companies, three development banks, 246 forex bureaus, a stock exchange, and four mobile money platforms (Bank of Uganda, 2017). On the other hand, the informal financial sector constitutes a wide range of institutions such as moneylenders, savings and credit cooperative associations, rotating savings and credit associations, accumulating savings and credit associations, and microfinance institutions which are not regulated by Bank of Uganda.

In terms of outreach, the commercial banking sector, with the takeover of Crane Bank by DFCU Bank in

2016, registered a decline in branch network and number of automated teller machines (ATMs). Moreover, other banks have been rationalising their branch and ATM operations, thus shifting to the more efficient alternative channels such as mobile banking, in order to minimise operational costs. Consequently, the total number of bank branches stood at 546 as of June 2017 compared to 566 branches at the end of June

2016. Likewise, the number of ATMs decreased from 862 to 818 in the same period. There is also a sizeable number of MDIs and credit institutions branches, though the number of MDIs branches reduced

9 See appendix A for the list of Tier 2 and 3 financial institutions.

19 in 2013 when Finance Trust Microfinance was granted a license by Bank of Uganda to operate as a commercial bank (see Table 2.1).

Table 2.1. Number of licensed branches/outlets

Category 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Banks branches 301 363 393 455 455 504 558 570 566 546 Bank ATMs 480 598 637 714 768 813 834 860 818 MDIs 96 76 83 98 99 70 70 76 78 - Credit Institutions branches 27 45 42 44 47 52 55 57 61 - Source: Bank of Uganda Annual bank supervision reports (various issues); Bank of Uganda Financial stability reports (various issues); and Bank of Uganda Annual reports (various issues). Note: Information for MDIs and credit institutions for 2017 has not yet been released

Relatedly, there is increased sharing of ATM services in the banking sector. For instance, Interswitch

Uganda provides transaction-switching and electronic payments processing services, through connected

ATMs of subscribing financial institutions, and selected mobile money platforms. By the end of June

2017, 15 financial institutions (11 commercial banks, three credit institutions, and one MDI) were connected to Interswitch and sharing their network of 409 ATMs. In addition, banks also share ATMs through VISA payment platform. The increase in sharing of ATM infrastructure among the banks is likely to ease access to financial services as well as reduce the costs incurred by financial institutions (Bank of

Uganda, 2017).

Furthermore, the mobile money platforms have been growing in terms of number of subscribers, agents, transactions and value of transactions. Currently, there are 22.9 million subscribers up from just over half a million in 2009 when the service was introduced. The number of transactions have increased from about

28.8 million in 2010 to over 1.1 billion in 2017, while the value of transactions has increased from UShs.

133 billion in 2009 to over UShs. 52.8 trillion, a value more than half of Uganda’s GDP in 2017 (see Table

2.2). As far as the number of mobile network operators are concerned, there are currently four mobile network operators providing mobile money services: MTN Uganda through MTN Mobile Money, Airtel

Uganda through Airtel Money, Africell Uganda through Africell Money Uganda, and Uganda Telecom through MSente. There are also non-mobile payments providers, such as M-Cash, Ezee Money, and

Micro-pay. However, the mobile money industry is dominated by two players: MTN Uganda and Airtel

Uganda. The concentration of mobile money transactions among these two players is largely attributed to their greater customer base (Bank of Uganda, 2017).

Table 2.2. Mobile money operations

2009 2010 2011 2012 2013 2014 2015 2016 2017 No. of customers 0.552 1.683 2.299 5.663 12.118 17.6 19.5 19.6 22.9 (million) No. of transactions 28.8 87.5 241.7 399.5 445.7 566.4 809.1 1,111 (million) Value of transactions 0.133 0.927 3.75 7.2 16.8 22.2 26.5 37.4 52.8 (trillion) Source: Bank of Uganda, Annual bank supervision reports (various issues); Financial stability reports (various issues); and Annual reports (various issues).

Services offered on mobile money platform have widen from sending and receiving money to payment of utilities, services, taxes, among other transactions. Moreover, the platform has also introduced savings and credit facilities in partnership with commercial banks. Savings and loans services on mobile money were introduced in 2015 by Commercial Bank of Africa (CBA)-Uganda in partnership with MTN-

Uganda—the biggest telecommunication company in the country—dubbed Mokash. This service allows

MTN mobile money subscribers to access savings and loans services of CBA through their phones after registering for Mokash. Centenary Bank and Finance Trust Bank are also offering similar microloan facilities on phones. The wide use of mobile money has increased financial inclusion especially among people in the rural areas and informal sector who were hitherto excluded from the formal financial system.

There have also been other promising developments in Uganda’s banking sector of recent. These include the introduction of agency banking, Islamic banking, and bancassurance which resulted from the 2016 amendments to the Financial Institutions Act, 2004. Agency banking enables financial institutions to sell their products through agents which in turn narrows the gap in access to financial services. Bancassurance enables insurance companies to sell their insurance products through banks; this will consequently increase diversification of banking activities. In addition, Islamic banking allows the provision of banking

21 services basing on Islamic doctrines such as no interest charges on loans. So far one commercial bank— the Libyan Tropical Bank-Uganda—has taken up the initiative to implement Islamic banking. These innovations will go a long way in enhancing financial sector development in the country.

Despite the relatively large number of financial institutions and impressive innovations, Uganda’s commercial banking sector is less developed with low levels of intermediation. For instance, the bank deposit/GDP ratio, bank credit to deposits ratio, private sector credit to GDP and liquid liabilities to GDP averaged at 14 per cent, 72 per cent, 10 per cent and 18 per cent respectively between 2005 and 2015 (see

Table 2.3). As noted in section 1.1, this performance compares unfavourably with the regional (EAC) performance. This low level of financial development undermines financial intermediation and economic growth.

Table 2.3. Financial development indicators (2005-2015)

Indicator 2005-2010 2011 2012 2013 2014 2015 (Average) Bank deposits/GDP (%) 13.39 14.86 14.91 15.49 15.78 15.76 Bank credit/GDP (%) 8.845 11.57 12.43 12.42 12.37 12.56 Bank credit/Deposits (%) 66.29 77.89 83.32 80.17 78.39 79.65 Liquid liabilities/GDP (%) 17.18 18.42 18.43 18.94 19.23 19.29 M2/GDP (%) 21.01 22.2 20.2 20.5 21.7 21.9 Source: Data are from the June 2017 updated version of financial development and structure dataset of Demirguc-Kunt et al. (2017) 2.3.2. Efficiency

There is inherent inefficiency in the sector characterised with high bank spreads, margins, and overhead costs (see Figure 2.2). The inefficiency in Uganda’s banking sector has limited the level of financial intermediation. Specifically, low deposit rates have discouraged bank deposits while high lending rates have discouraged borrowing, especially on long term basis. This, as noted in section 1.1, has undermined the level of economic activity and growth in the country. As far as overhead costs are concerned, the sector faces high operational expenses which could probably explain the high interest spreads. Smaller banks with a wide branch network are worst affected in terms of operational costs.

Figure 2.2. Efficiency in Uganda's Financial System

14.00%

12.00%

10.00%

8.00%

6.00%

4.00%

2.00%

0.00% 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015

Interest rate spreads Net interest margins Overhead costs/total assets

Source: Annual financial statements of commercial banks; Bank of Uganda Financial stability reports, various issues.

In terms of the structure of interest rates, there a wide divergence between the lending rates and deposit rates (see Figure 2.3). Moreover, there is competition for credit between the government and the private given the high treasury bill rates. With high treasury bill rates offered by government on its securities, banks also invest a significant proportion of their assets in government securities. Moreover, banks also lend mostly to business and trade financing and majorly on short term basis. This type of lending has limited the long term investment in the country.

Figure 2.3. Structure of interest rates (%)

0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Central Bank Rate Rediscount Rate Lending Rate 91-day Treasury Bill Yield 364-day Treasury Bill Yield Time deposit rate

Source: Bank of Uganda (2017)

2.3.3. Competition and concentration

Though the HII (loans and deposits) as measures of market concentration have been declining over time as depicted in Figure 2.4, there is still moderately high concentration in Uganda’s banking sector given that HHI are still above 1000. The reduction in HHI could be attributed to entry of new commercial banks into Uganda’s financial system, increasing the number from 16 in 2005 to 25 in 2015. Furthermore, from

Figure 2.4, there is more concentration in the deposit market compared to the loans market. Moreover,

Uganda’s banking sector is dominated by foreign banks which control over 85 per cent of the banking sector assets (Demirguc-Kunt et al., 2017). In addition, three banks (Stanbic Bank, Standard Chartered

Bank, and DFCU Bank), identified by Bank of Uganda as domestic systemically important banks, accounted for 42.8 per cent, 41.6 per cent, and 75.8 per cent of total bank assets, loans, and net profit after tax respectively as of the end of June 2017 (Bank of Uganda, 2017). This level of concentration shows limited competition in Uganda’s banking sector.

Figure 2.4. Market concentration and competition in Uganda's banking sector

2500

2000

1500

1000

500

0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 HHI (deposits) HHI (loans)

Source: Annual financial statements of commercial banks; Bank of Uganda Financial stability reports, various issues

2.3.4. Financial soundness indicators

The trend of key financial soundness indicators on asset quality, earnings and profitability, capital adequacy, liquidity, and market sensitivity indicate that the sector is sound and stable, more especially after the implementation of financial reforms. Commercial bank assets have increased over time to an average of 32 trillion as of December, 2016 up from about 4.6 trillion in 2005 (see Table 2.4). The assets are slightly over 30 per cent of GDP. This is a low ratio compared to the regional asset/GDP ratios, thus showing the low level of financial deepening in the country. In terms of quality, the indicators of banks’ asset quality of recent indicate build-up of credit risk. Improvement in asset quality has generally been hampered by significant write-offs and further deterioration of existing non-performing loans. The ratio of non-performing loans to total gross loans has been relatively high, averaging at 4.2 percent between

2005 and 2016.

Table 2.4. Commercial bank assets in UShs. billion

Average 2011 2012 2013 2014 2015 2016 (2005- 2010) Cash 259.4 507.8 481.0 567.5 646.0 765.9 741.6 Balances with BOU 432.0 793.1 1,055.3 1,635.1 2,073.9 2,002.6 2,581.7 Advances 6,016.1 14,285.2 15,760.7 15,989.7 17,345.7 21,738.5 21,981.4 Securities 1,463.0 2,376.5 2,769.1 3,333.7 4,277.2 4,188.1 5,020.9 Fixed assets 296.0 407.5 476.2 539.6 774.0 895.0 957.3 Others 281.5 455.3 569.3 665.4 680.7 727.8 763.9 Total assets 8,747.9 18,825.5 21,111.6 22,731.1 25,797.5 30,317.9 32,046.9 Source: Bank of Uganda Note: BOU is Bank of Uganda

As far as earnings and profitability of the banking sector are concerned, the income of Uganda’s banking industry have been increasing over time as depicted in Figure 2.5 Most of commercial bank income, given the limited level of diversification of the banking sector, is derived from the interest income; that is interest income from bank advances, government securities, deposits abroad, and other interest earning assets.

Non-interest income including, among others, charges, fees, and commissions; foreign exchange income; and income form off balance sheet transactions is limited. Thus Uganda’s banking sector relies on the traditional banking activities for most of its income.

Figure 2.5. Commercial bank income in billion: 2005-2016

3000

2500

2000

1500

1000

500

0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Interest income Non-interest income

Source: Annual financial statements of commercial banks; Bank of Uganda Financial stability reports, various issues

Furthermore, over 70 per cent of the interest income is derived from advances (loans) by commercial banks. From Figure 2.6 it is also evident that commercial banks in Uganda invest a significant share of its assets in government securities. Interest income from government securities take the second largest share of interest income in Uganda’s commercial banking sector. In fact, between 25 per cent and 30 per cent of the interest income is derived from investment in government securities. Increased investment by banks in government securities, according to theory, crowds out the private sector, thus leading to high lending rates. However, one could also argue that the banking sector has surplus liquid assets over and above its short term financing requirement, and that there is limited effective credit demand due to lack of collateral.

As such, investment in government securities by banks is a viable option to manage their liquidity profitably. In terms of profitability, banks’ aggregate net after-tax earnings averaged 16.7 percent between

2005-2016. However, the long-term trends of return on assets and return on equity show that the profitability of Uganda’s banking industry has generally been declining.

Figure 2.6. Decomposition of commercial banks' interest income: 2005-2016

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0% 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Advances Government securities Deposits abroad Others

Source: Annual financial statements of commercial banks; Bank of Uganda Financial stability reports, various issues

In terms of liquidity, banks in Uganda maintain adequate liquidity buffers well above the regulatory minimum requirements, thus keeping liquidity risk low. For instance, the liquid assets-to-total deposits ratio rose averaged 46.2 percent in the period 2005-2015, more than double the regulatory minimum requirement of 20.0 percent. The build-up of excess liquidity in the banking industry partly reflects the low level of lending by commercial banks.

Turning to capital adequacy, the banking sector has been well capitalised since the implementation of financial reforms. All commercial banks meet the minimum core and total capital adequacy ratios of 6.

The industry’s aggregate tier one capital adequacy ratio and the total regulatory capital adequacy ratio averaged at 14.8 percent and 20.5 percent respectively between 2005 and 2016. The leverage ratio (ratio of regulatory tier 1 capital to total assets plus off-balance sheet items), which is another indicator of banks’ capital adequacy, averaged at 10.0 percent between 2005 and 2016. Overall, the capital adequacy ratios show a well-capitalised and stable banking sector.

Finally, as far as market sensitivity is concerned, the banking sector is highly exposed to the foreign exchange markets. For instance, the forex exposure to regulatory tier 1 capital, forex loans to forex deposits, and forex assets to forex liabilities averaged at −2.8 per cent, 60.4 per cent, and 101.37 per cent respectively. Moreover, foreign currency loans account for over 45 per cent of total banking sector loans.

Thus the operations in the forex market have a great impact on the performance of banking sector. The market sensitivity to forex market operations is exacerbated by the fact that most banks are foreign owned.

See Appendix B for financial stability indicators.

2.4. Summary to the chapter

The interventionist policies pursued by government during the pre-financial reform period had disastrous effects on the financial sector. These effects were exacerbated by the inefficient regulation and supervision, leading to imprudent management and the subsequent fragility of the financial sector. On a positive note though, the financial reforms that were implemented from the early 1990’s have improved the performance and stability of the financial sector. UCB, one of the key stability risks, has been privatised; small and weak banks have been cleaned-up; and bank supervision and regulation have been improved through the introduction of the risk based approach to ban regulation and regulation. As a consequence, there is a presence of reputable banks that appear to be well capitalized, profitable, and resilient; thus a sound and stable financial sector. Moreover, the financial sector has substantially developed, more especially in terms of the number of financial institutions and outreach. In addition, outreach of financial service has been enhanced with the introduction of mobile money in 2009. Notwithstanding these achievements, Uganda’s financial sector is still relatively underdeveloped and inefficient in comparison with Uganda’s regional peers.

CHAPTER THREE

LITERATURE REVIEW

3.0. Introduction

This chapter reviews literature on interest rate spreads. Specifically, the chapter deals with conceptualisation of interest rate spreads in literature, theoretical and empirical literature on interest rate spreads, and it concludes with the summary of the reviewed literature.

3.1. Conceptualisation of interest rate spreads

There are different definitions of interest rate spreads. Such definitions are generally categorised into ex- ante and ex-post definitions. The ex-ante interest rate spreads, on one hand, are measured using the bank’s lending and deposit rates. For instance, Sologoub (2006) defines ex-ante bank spread as the difference between the weighted average lending and weighted average savings and time deposit rates for individual commercial banks. On the other hand, ex-post interest rate spreads (interest rate margins) are measured using the bank’s interest revenue and interest expenses. Ex-post interest rate spreads (bank margins) are further categorised into narrow and wide definitions, within which six different definitions of margins are derived basing on the coverage of banks’ assets and liabilities. Specifically, narrowly defined margins are based on the bank’s loans and advances, while broad definitions of margins take into account larger share of the bank’s assets and liabilities compared to the narrow definitions (Brock & Rojas-Suarez, 2000;

Chirwa & Mlachila, 2004; Mugume, et al., 2009). Narrow and wide definitions of bank margins are presented in the following equations.

(a) Narrow definitions of interest rate margins

 NMARGIN1=(Interest received on loans only)/Total loans−(Interest paid on deposits

only)/Total deposits

 NMARGIN2=(Interest received)/Total loans−(Interest paid)/Total deposits

 NMARGIN3=(Interest plus commission received)/Total loans−(Interest plus commission

paid)/Total deposits

(b) Wide definitions of interest rate margins

 WMARGIN1=(Interest received−Interest paid)/Total assets

 WMARGIN2=(Interest received)/All interest earning assets−(Interest paid)/All interest bearing

liabilities

 WMARGIN2=(Interest plus commissions received)/All interest earning assets−(Interest plus

commissions paid)/All interest bearing liabilities

Note that NMARGIN is the narrow interest margin, whilst WMARGIN is the wide interest margin.

In comparison with ex-ante interest spreads, ex-post bank spreads could understate the actual spread since the spread is calculated using interest revenue and interest expenses. For instance, the measure will under state the bank spread in case of a large proportion of the bank’s loans are non-performing loans which might actually have been contracted at high ex-ante lending rates. Thus, a reduction in bank margins may not necessarily signal improved bank efficiency. It could actually be due to a high loan default rate which may reflect a deterioration of banks’ effectiveness in monitoring borrowers (Nannyonjo, 2001). As such, the ex-ante interest rate spreads are often recommended as a better measure of intermediation efficiency.

3.2. Theoretical literature on interest rate spreads

There are a number of theoretical literature that explain sources of interest spreads and margins. This literature traces the sources of high interest spreads and margins to risks and uncertainties, market and banks’ characteristics, the banks’ capital structure, macroeconomic environment, and regulation. Selected theoretical literature for the study include the risk based hypothesis, small financial system view, the market structure view, macroeconomic view, and capital structure hypothesis.

The risk-based hypothesis of interest rate spreads focuses on the risks lenders take and the compensation for these risks as part of the spread and margin. Such risks are largely attributed to information asymmetry in the financial markets. Thus, interest rate spreads compensate the risks resulting from the deficiencies in the contractual and informational frameworks and the resulting inability of lenders to perfectly ascertain the creditworthiness of the borrower and her project ex-ante and subsequently monitor the implementation ex-post (Crowley, 2007). This gives rise to adverse selection and moral hazard (Stiglitz & Weiss, 1981).

In turn, information asymmetry increases the credit and liquidity risks faced by financial institutions thus leading to increase in lending rates. Furthermore, other risks associated with higher spreads in the banking system according to the risk based view include interest risks, exchange rate risks, and operational risks.

The small financial system view focuses on the fixed cost component of financial service provision and the resulting scale economies. Such scale economies arise on different levels. Processing an individual payment or savings transaction entails costs that are—at least in part—independent of the value of the transaction. The branch network, computer systems, and legal accounting services create fixed costs at the level of financial institutions, and fixed costs even arise at the level of the financial system—such as regulatory costs—which are up to a point independent of the number of institutions regulated and supervised (Beck & Hesse, 2009). According to this hypothesis, small financial systems such as Uganda are not able to exploit these scale economies and therefore face higher costs and interest rate spreads.

Generally, larger banks are expected to incur lower per unit fixed costs and thus lower spreads, whilst banks with a relatively larger branch network and smaller average deposit and loan amounts are expected to incur higher costs and thus higher spreads (Beck & Hesse, 2009; Brock & Rojas-Suarez, 2000;

Demirguc-Kunt & Huizinga, 1999; Randall, 1998).

The market structure view focuses on the competitiveness and ownership structure of the banking market.

General economic theory suggests that more competitive systems have more efficient banks with lower spreads and margins, as well as deeper and broader banking markets. Competition, however, is not necessarily the same as market structure (Claessens & Laeven, 2004). In the case of Uganda, Hauner and

Peiris (2005) show that competitiveness has increased over the past years in spite of simultaneous increases in concentration. According to the market structure view, foreign bank entry can also improve banking system efficiency and thus reduce interest rate spreads, but can also result into higher spreads and margins if accompanied by higher concentration and lower competitiveness (Demirguc-Kunt & Huizinga,

1999).

The macroeconomic view sees interest rate spreads as being driven by instability in the macroeconomic environment of an economy. Such macroeconomic parameters include, among others, inflation, exchange rates, treasury bill rates, level of financial sector development, and economic cycles. Inflation can affect spreads if monetary shocks are not passed through to the same extent to deposit and lending rates or adjustment occurs at different speeds (Mugume et al., 2009). Changes in the treasury bill rate proxies the volatility in money market interest rates and is thus used as a benchmark in interest setting by banks. As such, higher treasury bill rates are often associated with higher bank spreads (Mugume et al., 2009;

Nampewo, 2013). In addition, volatility in the exchange rate is an important determinant of bank spreads especially for foreign-owned banks that often have a share of their assets in foreign-currency accounts overseas. Currency depreciation is generally associated with high bank spreads (Folawewo & Tennant,

2008). Real GDP growth rates is also reported to affect lending rates as the creditworthiness of borrowers varies over the business cycle with periods of recessions being associated with higher default rates and thus higher bank spreads (Saunders & Schumacher, 2000). Furthermore, financial sector development is generally associated with lower bank spreads (Demirguc-Kunt & Huizinga, 1999).

Last but not least, the capital structure hypothesis posits that the capital structure of banks can contribute to the banks’ spreads and margins. The level of capital that banks hold to cushion themselves against risks can result into higher spreads (Saunders & Schumacher, 2000). In particular, holding capital in excess of the regulatory minimum for insuring against credit risk turns out to be relatively more expensive than debt because of differential taxation (Chirwa & Mlachila, 2004). The capital costs may be offset by raising

32 banks spreads. Regulatory environment also tends to influence bank spreads. Higher reserve requirements can result into wider bank margins (Barajas, Steiner & Salazar, 1999; Saunders & Schumacher, 2000).

3.3. Empirical literature

Empirical literature that examine the determinants of interest rate spreads generally uses variables that are categorised into: bank specific factors—factors that affect individual banks and whose magnitude vary from bank to bank; banking industry specific factors and regulation—factors that affect the entire banking sector; and macroeconomic variables—factors which affect the entire economy. Some studies focus on only macroeconomic environment and in some cases industry specific factors in assessing determinants of bank spreads (see for example: Folawewo & Tennant, 2008; Nampewo, 2013). Whilst other studies, and by far most studies, consider all the three categories of factors in analysing determinants of bank spreads. Selected empirical literature on determinants of interest rate spreads is presented below.

3.3.1. Bank specific factors and interest rate spreads

There are a number of bank specific variables that influence the behaviour of banks in the intermediation process. These, among others, include credit risk, liquidity risk, bank size, operating costs, return on assets, non-interest income, and capital adequacy ratio. These factors determine the lending and deposit rates of banks during financial intermediation. Consequently, the factors affect bank spreads and margins, and their effect in some cases vary from study to study as the literature below shows.

To start with, Ahokpossi (2013)—using a sample of 456 banks in 41 SSA countries for the period 1995-

2008—shows that credit risk is positively related to interest margins. The study measures credit risk as a ratio of non-performing loans to total bank assets. Likewise, running pooled and fixed effects regressions of a panel of 44 Kenyan banks for the period 2000-2009, Tarus, Chekol, and Mutwol (2012) show that credit risk has a positive and significant effect on net interest margins. In addition, Mugume et al. (2009) show that provisioning for bad debts is directly proportional to interest margins in Ugandan banking sector. Thus, an increase in credit risk translates into higher spreads. This argument hinges on the fact that higher credit risk resulting from high default rates reduces bank profitability due to loan-loss provisions.

Conversely, the cost of loan-loss provisions in most cases is transferred to customers by banks either in form of high lending rate and/or low deposit rates to compensate for lost profits, and thus leading to high spreads (Chirwa & Mlachila, 2004; Grenade, 2007; Jamaludin, Klyuev & Serechetapongse, 2015;

Mugume et al., 2009). As such, it can be noted that high interest spreads are a risk premium for lending to high risk borrowers.

However, following the Ho and Saunders (1981) model in which the spread is decomposed into a pure spread and the remaining component is explained by market structure, regulation and idiosyncratic bank factors, Mannasoo (2012) shows that credit risk plays a minimal role in determination of interest margins in Estonia. Additionally, Nannyonjo (2001) notes that default risk is not important in banks’ decisions to rise ex-ante spreads though it reduces the net interest margins in Uganda. Moreover, results by Crowley

(2007) are inconclusive as whether loan quality has significant or insignificant direct effect on the spreads in English-Speaking African Countries. In regressions of adjusted interest spreads and net interest margins, the coefficient for loan quality is positive and significant. Yet loan quality is not significant in any of multivariate regressions. The study further shows that provisioning for bad debts has no effect on adjusted interest rate spreads, but a nearly positive significant effect on net interest margins. Furthermore,

Brock and Rojas-Suarez (2000) show that non-performing loans are not in general associated with higher spreads in Latin America which contrasts the case of industrialised countries. Notwithstanding the significance of credit risk in determining bank spreads, all studies suggest that increase in credit risk generally leads to an increase in spreads.

The other bank specific factor that is reported to affect interest spreads is liquidity risk. Higher bank liquidity is associated with lower interest margin (Ahokpossi, 2013; Beck & Hesse, 2009; Islam &

Nishiyama, 2016; Manamba, 2014; Mannasoo, 2012). High liquidity ratios are used by banks to safeguard against sudden withdrawals by customers, which can greatly compromise banking sector stability in addition to reducing the share of deposits that can be used for lending, thus leading to an increase in ex- ante spreads. Furthermore, Nannyonjo (2001) shows that ex-ante spreads reflect interest rate risk, liquidity

34 risk and insolvency premiums. However, the study notes that these risk factors on the other hand reduce the net interest margins of banks. Overall, high liquidity risk faced by banks is associated with high bank spreads.

The size of the bank is also reported in literature to have a significant impact on interest spreads and margins. Beck and Hesse (2009) find that smaller banks in Uganda have higher margins than bigger banks.

In addition, Willmott (2012) shows that banks with greatest market share offer, on average, the lowest lending rates, whilst banks with the smallest market share offer the highest lending rates in Uganda.

Moreover, Almarzoqi and Naceur (2015) recommend for consolidation of the banking sector in Caucasus and Central Asia (CCA) countries given that larger banks were associated with lower spreads. This recommendation could relate to economies of scales that large banks tend to enjoy relative to smaller banks. However, these findings contradict other studies that attribute high spreads to large banks. For instance, Afanasieff, Lhacer and Nakane (2002) find that the larger the bank, the higher the spread in

Brazil. Similarly, Were and Wambua (2015) show that, on average, big banks have higher spreads compared to small banks in Kenyan commercial banking sector. Whilst Gambacorta (2004) shows that bank size is irrelevant in influencing bank margins in Italy. Generally, most literature attribute bank size to lower interest spreads and margins especially in small economies given the role of economies of scale in bank operations.

Furthermore, operational efficiency—as measured by the cost of operations—also influences bank spreads. Costs in the intermediation process relate to screening and monitoring borrowers, and processing savings and payment services. Using the Ho and Saunders (1981) model and its subsequent extensions,

Almarzoqi and Naceur (2015) show that operational efficiency is by far the most important driver of interest spreads in CCA countries since higher operating costs were reflected in higher interest spreads.

This conclusion in line with, among other studies, Afanasieff et al. (2002), Brock and Rojas-Suarez (2000),

Grenade (2007), Mugume et al. (2009), Mujeri and Younus (2009), Siddiqui (2012), and Tarus et al.

(2012). In addition, Beck and Hesse (2009) find evidence in Uganda that banks targeting the low end of

35 the market incur higher costs and therefore have higher margins. Additionally, all measures of costs have positive coefficients in Crowley (2007) though not significant. Moreover, Nannyonjo (2001) shows that higher overhead costs are not reflected in high ex-ante spreads, even though they were associated with higher net interest margins. By and large, high interest spreads show the additional cost that banks incur to perform intermediation activities.

Return on assets is another bank specific variable that is reported in the literature to have significant impact on bank spreads. For instance, using an annual panel of 22 banks in Pakistan, Siddiqui (2012) shows that the spread increases with increase in return on assets in all the regressions (pooled, fixed and random effects regressions). This is consistent with findings of Were and Wambua (2015) for the case of the

Kenyan banking sector. Furthermore, Brock and Rojas-Suarez (2000) shows that increase in the capital/asset ratio increases the spread in Bolivia and Colombia though not statistically significant in the other three Latin American countries (Chile, Mexico, and Peru) included in the sample. In Uganda, Beck and Hesse (2009) show that an increase in return on assets leads to an increase in both spreads and margins and they conclude that more profitable banks also, on average, charge higher spreads and earn higher margins. Conversely speaking, a high return on assets implies high bank spreads especially in developing countries where interest income accounts for much of the bank profits given the limited diversification of banking activities.

Non-interest income also plays a significant role in determining bank spreads. Using a bank profit maximisation model based on empirical industrial organisation approach, Mujeri and Younus (2009) show that the higher the non-interest income, the lower the interest spreads in the Bangladesh banking sector.

They also show that non-interest income is more significant in determining spreads for foreign owned banks compared to local banks. This could be due to the fact that banking activities of foreign banks are more diversified than those of local banks. In addition, Almarzoqi and Naceur (2015), and Carbo and

Rodriguez (2007) show that banks that rely heavily on revenue from non-traditional business have lower interest margins given that such non-interest income somewhat compensates for the lower margins from

36 traditional bank activities. Generally, as non-interest income declines, banks raise lending rates to compensate for the loss in income.

Last but not least, capital adequacy ratio also determines the setting of interest rate spreads by banks. It is usually associated with higher interest spreads and margins (Ahokpossi, 2013; Crowley, 2007). However, capital adequacy ratio is not statistically significant in explaining net interest margins in CCA countries except in Armenia (Almarzoqi & Naceur, 2015). Nonetheless, it has a positive sign thus alluding to the fact that an increase in the capital adequacy ratio leads to an increase in bank spreads. Overall, a high capital adequacy ratio is associated with high bank spreads given that under the assumption of risk aversion, shareholders usually demand higher returns on their additional equity.

In a nutshell, bank specific variables affect bank behaviour during the process of financial intermediation.

As such, they affect interest spreads of banks as the above discussion shows. However, the direction, magnitude and significance of the effect vary from variable to variable and in some cases from study to study even within a single country. Moreover, the impact of bank specific variables on interest spreads also vary across countries with notable differences between developed and developing countries.

3.3.2. Industry specific factors and interest rate spreads

Industry specific factors affect the entire banking sector and as such have a bearing on interest spreads.

Such factors relate to bank concentration, market structure. and bank regulation. As far as market concentration is concerned, Ahokpossi (2013) finds that concentrated bank markets—as measured by

HHI—are positively related to interest margins and as such he recommends for policies that promote competition and reduce market concentration so as to lower interest margins in SSA. These findings are in line with, among others, findings of Almarzoqi and Naceur (2015), Grenade (2007), Jamaludin et al.

(2015), and Ramfall (2001) who attribute high bank spreads in CCA countries, Eastern Caribbean

Currency Union, Pacific island countries and Mauritius respectively to high market power of banks. In addition, Manamba (2014) attributes high interest spreads in Tanzanian banking sector to lack of competition among financial institutions, whilst Nannyonjo (2001) shows that lack of competition in

Uganda’s banking sector translates into high ex-ante spreads, although they were not necessarily associated with high net interest margins. Relatedly, Chirwa and Mlachila (2004) attribute the high spreads in Malawian banking sector to high monopoly power of some banks. Generally speaking, all these findings are in agreement with a pioneer paper on the relationship between market concentration and bank spreads by Demirguc-Kunt and Huizinga (1999) that shows that a lower market concentration ratio leads to lower bank margins and profits.

However, contrary to above studies, Carbo and Rodriguez (2007) show that increasing market power in

European banking sector is associated with decreasing interest margins. Additionally, Beck and Hesse

(2009), and Mugume et al. (2009) show that market concentration—as measured by HHI—is negatively related to bank margins in Uganda’s banking sector. This could suggest that there is little collusive behaviour in the pricing of financial products in Uganda’s banking sector. Furthermore, Crowley (2007) and Tarus et al. (2012) show that market concentration has a negative impact on bank spreads in English-

Speaking African Countries and the Kenyan banking sector respectively. Equally important, high concentration among large banks is found to not necessarily be associated with increased or reduced efficiency in the Kenyan banking sector (Beck et al., 2010). Overall, market concentration is associated with high bank spreads though there are a few exceptions in literature.

As far as bank regulation is concerned, Gelos (2006) shows that higher interest spreads in Latin America are partly due to large reserve requirements than in other regions. The study uses bank and country level data from 85 countries, including 14 Latin American economies. Likewise, Chirwa and Mlachila (2004), and Grenade (2007) shows that the spread in Malawi and Eastern Caribbean Currency Union respectively increases with increase in reserve requirements. Additionally, Afanasieff et al. (2002) and Crowley (2007) also show that the required reserve ratio positively influences bank spreads though the coefficients were not significant. Furthermore, a comparative analysis by Demirguc-Kunt and Huizinga (1999) shows that reserves have a more pronounced impact on margins and profitability in developing countries than in developed countries. Generally, reserve requirements, especially in developing countries, act as a tax on

38 banks and as such they translate into higher spreads (Brock & Rojas-Suarez, 2000; Folawewo &Tennant,

2008; Islam & Nishiyama, 2016; Nannyonjo, 2001; Ramfall, 2001).

Additionally, other forms of prudential bank regulations and institutional quality also influence interest spreads in the banking sector. Indeed, Crowley (2007) shows that the quality of regulatory regime has significant effect on lowering spreads in univariate regressions of English-speaking African countries.

Equally important, Gelos (2006) notes that a less supportive legal environment contributes to larger intermediation costs. Thus, high institutional quality is generally associated with lower spreads (Beck &

Hesse, 2009; Fofack, 2016; Jamaludin et al., 2015; Rebei, 2014).

Lastly, the share of bank ownership is also reported in the literature as one of the industry specific variables that determine interest spreads. Specifically, foreign bank ownership is found to have a positive relationship with bank spreads in English-speaking African countries (Crowley, 2007). Likewise, Beck and Hesse (2009) show that foreign bank share in the loans market is positively related to interest margins and spreads. The association of foreign banks with high spreads could be because foreign banks in developing countries do not face the same competitive pressure they face in developed countries.

However, on the contrary, Ahokpossi (2013) shows that foreign ownership leads to lower interest margins though the coefficient is not significant. Generally, foreign bank market share is associated with lower bank spreads and margins in developed countries but higher spreads and margins in developing countries

(Demirguc-Kunt & Huizinga, 1999). In summary, industry specific factors and bank regulation affect the environment in which banks operate and as such they have an effect on interest spreads, though, the literature is not conclusive on their impact on interest spreads and margins.

3.3.3. Macroeconomic factors and interest rate spreads

The prevailing macroeconomic environment in an economy also affects bank spreads. Some of the macroeconomic variables that affect interest spreads according to literature include, among others, inflation, real GDP growth rate, treasury bill rates, exchange rate volatility, and M2/GDP. Inflation has a significant impact in raising interest spreads and margins, though the extent of the impact differs from

39 country to country or across market segments (Ahokpossi, 2013; Almarzoqi & Naceur, 2015; Beck &

Hesse, 2009; Chirwa & Mlachila, 2004; Mugume et al., 2009; Nannyonjo, 2001; Tarus et al., 2012). This is because inflation leads to decrease in real interest rates and as such banks tend to set wide spreads to compensate for the loss. However, Afanasieff et al. (2002) find that inflation rate negatively affects the pure spread in Brazil, which actually contradicted a priori expectations. Similarly, Crowley (2007) shows that inflation has a negative effect on inflation adjusted spreads and inflation adjusted net interest incomes in English Speaking African Countries. Additionally, Jamaludin et al. (2015) conclude that inflation has just marginal impact on spreads in Pacific island countries. Notwithstanding the contradicting findings in literature, inflation, especially hyperinflation, translates into high bank spreads.

In addition, studies show that real GDP growth has a negative impact on interest spreads (Beck & Hesse,

2009; Crowley, 2007; Islam & Nishiyama, 2016; Mugume et al., 2009; Tarus et al., 2012). This is because a slowdown in economic growth puts an upward pressure on bank spreads more especially through its impact on banks’ profitability given the fact that asset quality deteriorates with increase in the default rate which is more prevalent in situations of economic slowdown. Furthermore, slow growth is characterised with low savings in an economy which constrains mobilisation of investible resources by banks hence leading to wide spreads. However, in contrast, Afanasieff et al. (2002) and Grenade (2007) show that spreads increase with increase in real GDP growth. Moreover, Ahokpossi (2013), and Were and Wambua

(2014) show that spreads are not sensitive to economic growth. As far as the size of the economy is concerned, Crowley (2007) finds that an increase in the size of the economy significantly increases adjusted net interest margins, whilst Jamaludin et al. (2015) finds that the size of the economy is negatively correlated with spreads. By and large, real GDP growth rates and the size of the economy are negatively related to interest spreads especially in small economies though there are contradicting findings in the literature.

Furthermore, an increase in the treasury bill rate leads to an increase in the interest spreads in Uganda’s banking sector (Mugume et al., 2009; Nampewo, 2013). This is consistent with Beck and Hesse (2009)

40 who attribute the large proportions of the high interest margins in Uganda to high treasury bill rates in both their international comparison and country panel analysis. As a matter of fact, high treasury bill rates are an incentive to banks to invest more of their deposits in risk free government instruments compared to loans which are characterised by high credit risk. This translates into high bank spreads in the banking sector. However, Folawewo and Tennant (2008) though not significant curiously show that treasury bill rates are negatively related with interest spreads. Overall, high treasury bill rates lead to high bank spreads since banks especially in developing countries are more attracted to relatively risk free government instruments compared to loans given the associated credit risk.

Similarly, Mugume et al. (2009) and Nampewo (2013) show that exchange rate volatility is also significantly and positively related to interest margins in Uganda. This is because uncertainty in the foreign exchange market affects the profitability of banks especially foreign owned banks (which control over 80 per cent of banking sector assets). These findings are consistent with those of Folawewo and Tennant

(2008) who consider 41 SSA countries in their study. Relatedly, Beck and Hesse (2009) attribute high bank margins to exchange rate depreciation. Different from the above findings, variability of the exchange rate is found to be insignificant in explaining interest margins in English-speaking African countries by

Crowley (2007) though the coefficient is positive. Conversely speaking, variability in exchange rates leads to higher premium demands by banks especially foreign owned ones thus leading to higher bank spreads.

Finally, M2/GDP growth is found to be significantly and negatively related to interest spreads but only indirectly through its effects on inflation (Crowley, 2007). Indeed, Nampewo (2013) conclusively shows that M2/GDP is indirectly related to bank spreads in Uganda. Given that M2/GDP is an indicator of financial sector development, M2/GDP growth leads to lower interest spreads especially in developing countries like Uganda where financial markets are not well developed. To sum up, uncertainty in the macroeconomic environment is associated with high interest spreads and margins.

3.4. Summary of the literature and research gap

In conclusion, there are a number of empirical studies on the determinates of interest rate spreads focusing on different sets of factors (bank-specific, industry-related, and/or macroeconomic factors). However, most studies note that spreads and margins are significantly affected by bank specific factors than any other category of factors. Nonetheless, for an in-depth understanding of factors that influence spreads, there is always need to focus on all the three categories of factors during empirical analysis. As far as the research gaps are concerned, studies on determinants of interest rate spreads in Uganda, as aforementioned in chapter one, do not consider the impact of interest rate spreads in the previous periods on current spreads. The studies, with the latest covering data up to 2007, do not used the recent data yet the banking sector has seen a number of developments. Moreover, the studies are not conclusive on the direction of the determinants of interest spreads. To that effect, this study seeks to close these research gaps.

CHAPTER FOUR

METHODOLOGY

4.0. Introduction

This chapter discusses the methodology used in the study. It discusses the theoretical framework and econometric model specification. It also deals with data description and data sources of the variables used in the study. It further highlights the estimation techniques, and panel unit root tests used in the study.

4.1. Theoretical framework

Theoretical frameworks that are often used to analyse bank spreads are based on banks’ balance sheet

(Demirguc-Kunt & Huzinga, 1999) or on behavioural assumptions of the banking firm (Barajas et al.,

1999; Brock & Rojas-Suarez, 2000; Ho & Saunders, 1981; Randall, 1998; Saunders & Schumacher,

2000). In the Ugandan context, Mugume et al. (2009) analyse the determinants of interest rate spreads basing on the behavioural assumptions of the banking firm, whilst Nannyonjo (2001) specifically follows the Ho and Saunders (1981) bank dealership model. In addition, Nampewo (2013) follows the McKinnon and Shaw (1973) paradigm in the analytical framework of determinants of interest rate spreads in Uganda.

For purposes of this study, we follow the Ho and Saunders (1981) bank dealership theoretical framework given that it explains bank behaviour during the intermediation process.

Ho and Saunders (1981) show that the pure spread depends on the size of banks’ transactions, the market structure of the banking industry, the volatility of interest rates, and the degree of managerial risk aversion.

The model has been modified by different authors to include other factors that affect interest rate spreads and/or use alternative variable definitions. For instance, McShane and Sharpe (1985) proxies the volatility of the deposit or lending rates by the volatility of money market interest rate, whilst Allen (1988) considers deposit and loan heterogeneity in the model and shows that pure interest margins may be reduced by diversification of bank services and products. In addition, Angbazo (1997) extends the Ho and Saunders model to include credit risk and its interaction with interest rate risk. Furthermore, Maudos and Fernandez

43 de Guevara (2004) extends the model to explicitly take into account banks’ operating costs and use the

Lerner’s index as a measure of competition besides the Herfindahl indices. Moreover, Carbo and

Rodriguez (2007) includes both traditional and non-traditional activities in the model to anlyse the effect of specialisation on bank spreads, and Maudos and Solis (2009) incorporate all the developments in one model.

The study, more specifically, follows the Ho and Saunders (1981) model as extended by Maudos and

Fernandez de Guevara (2004) because the model clearly explains the theoretical determinants of bank spreads though it does not include non-traditional banking activities—such as revenue from commission, brokerage fees, capital gains, dividends, and income from foreign exchange transactions, among others— in the calculation of these bank spreads as in Maudos and Solis (2009). However, this is not a challenge given that this study focuses on only traditional bank activities in calculation of interest rate spreads just like in Maudos and Fernandez de Guevara (2004). The Ho and Saunders (1981) model as extended by

Maudos and Fernandez de Guevara (2004) is as follows.

A bank is assumed to be risk-averse agent during the financial intermediation process. The model assumes a single planning period in which the bank sets interest rates at the beginning of the period (such rates are kept constant for the entire period) before any deposits or loans are made. Due to asymmetric information, banks have to set interest rates on loans (푟퐿) and deposits (푟퐷) optimally so as to minimise the risk arising from the uncertainty of interest rates in the money markets to which they have to resort in the event of excessive demand for loans or excessive supply of deposits. As such, they set their interest rates as margin relative to the interest rate of the money market (r), that is:

푟퐷 = 푟 − 푎

푟퐿 = 푟 − 푏 (4.1) a and b are the margins relative to the money market interest rate set by the banks for deposits and loans respectively. Hence, the unit margin or spread 푆 can be expressed as follows:

푆 = 푟퐿 − 푟퐷 = (푎 + 푏) (4.2)

The intuition of the model is that if the bank receives a new deposit before any new demand for loans, the bank will temporarily invest the funds received in the money market at an interest rate r and thus face a risk of reinvestment at the end of the period if money market interest rates fall. Similarly, if a new demand for loans reaches the bank before any new deposit, the bank will obtain the funds in the money market, and will therefore face a refinancing risk if interest rates rise. In addition, the return on loans is uncertain due to the possibility of default on some loans. Thus, the bank will apply a margin to loans (b) and deposits

(a) that will compensate for both the interest rate and credit risks.

The initial wealth of the bank is the difference between its assets–loans (L) and liabilities–deposits (D) plus net money market assets (M):

푊0 = 퐿0 − 퐷0 + 푀0 = 퐼0 + 푀0 (4.3)

퐿0 − 퐷0 is the net credit inventory (퐼0)

The operating costs of a banking firm are assumed to be a function of the deposits received (퐶(퐷)) and the loans made (퐶(퐿)) such that the cost of the net credit inventory are 퐶(퐼) = 퐶(퐿) − 퐶(퐷).

Given the above assumptions, the final wealth of the bank will be:

푊푇 = (1 + 푟퐼 + 푍퐼)퐼0 + 푀0(1 + 푟 + 푍푀) − 퐶(퐼0)

=퐼0 + 퐼0푟퐼 + 퐼0푍퐼 + 푀0 + 푀0푟 + 푀0푍푀 − 퐶(퐼0)

=퐼0(1 + 푟퐼) + 푀0(1 + 푟) + 퐼0푍퐼 + 푀0푍푀 − 퐶(퐼0)

But 푊0 = 퐼0 + 푀0

푊푇 = 푊0(1 + 푟푊)+퐼0푍퐼 + 푀0푍푀 − 퐶(퐼0) (4.4) where:

푟퐿퐿0−푟퐷퐷0 푟퐼 = is the average profitability of the net credit inventory; 퐼0

퐼0 푀0 푟푊 = 푟퐼 + 푟 is the average profitability of the bank’s initial wealth; and 푊0 푊0

퐿0 퐷0 퐿0 10 푍퐼 = 푍퐿 + 푍퐷 = 푍푃 is the average risk of the net credit inventory. 푍푀 and 푍퐿 reflect the 퐼0 퐼0 퐼0 uncertainty faced by the banks, which is of two kinds: interest rate risk, distributed as a random variable

2 푍푀 ∽ 푁(0, 휎푀), and credit risk—the profitability of the loan is uncertain and is distributed as a random

2 variable 푍푀 ∽ 푁(0, 휎퐿 ). In order to take into account the interaction between credit risk and interest rate risk, the joint distribution of the two disturbances is assumed to be bivariate normal with non-null covariance (휎퐿푀).

Banks are assumed to maximise their expected utility. The bank’s utility function is approximated using the Taylor expansion around the expected level of wealth (푊 = 퐸(푊))

1 퐸푈(푊) = 푈(푊) + 푈′(푊)퐸(푊 − 푊) + 푈′′(푊)퐸(푊 − 푊)2 (4.5) 2 where it is assumed that the bank’s utility function is continuous and doubly differentiable with 푈′ > 0 and 푈′′ < 0 hence the bank is risk averse11.

When a new deposit D, attracting a rate 푟퐷 is made and the bank does not grant an additional credit, it will invest the funds in the money market, obtaining a return (푟 + 푍푀)퐷. Given that, 푊 − 푊 = 퐿0푍퐿 + 푀0푍푀 and given the existence of operating costs in the receipt of deposits 퐶(퐷), substituting the new value of

12 the final wealth in (4.5), the increase in expected utility associated with the new deposit will be:

훥퐸푈(푊퐷) = 퐸푈(푊푇) − 퐸푈(푊)

1 2 = 푈′(푊)[푎퐷 − 퐶(퐷)] + 푈′′(푊) [(푎퐷 − 퐶(퐷)) + (퐷 + 2푀 )퐷휎2 + 2 퐿 퐷휎 ] (4.6) 2 0 푀 0 퐿푀

Similarly, if a new request for credit is made for which there is also a cost of production 퐶(퐿), the increase in expected utility would be:

훥퐸푈(푊퐿) = 퐸푈(푊푇) − 퐸푈(푊)

10 It is assumed that the deposits are an activity that is not subject to any kind of risks. Hence, 푍퐷 = 0. 11 If the bank were risk neutral, the bank would be an expected wealth maximiser. 12 See Appendix C.

2 1 (푎퐿 − 퐶(퐿)) + (퐿 + 2퐿 )퐿휎2 + (퐿 − 2푀 )퐿휎2 + 2(푀 − = 푈′(푊)[푎퐿 − 퐶(퐿)] 푈′′(푊) [ 0 퐿 0 푀 0 ] 2 퐿0 − 퐿)퐿휎퐿푀

(4.7)

Assuming that credits and deposits are made randomly according to a Poisson distribution, the probability of granting a credit or receiving a deposit is a decreasing function of the margins applied by the bank:

푃푟퐷 = 훼퐷 − 훽퐷푎

푃푟퐿 = 훼퐿 − 훽퐿푏 (4.8)

The maximisation problem is therefore as follows:

푀푎푥푎,푏퐸푈(훥푊) = (훼퐷 − 훽퐷푎)훥퐸푈(푊퐷) + (훼퐿 − 훽퐿푏)훥퐸푈(푊푙) (4.9)

The first order conditions with respect to a and b are given as:

′′ 1 훼퐷 퐶(퐷) 1 푈 (푊) 2 푎 = + − [(퐷 + 2푀0)휎푀 + 2 퐿0휎퐿푀] 2 훽퐷 퐷 4 푈′(푊)

′′ 2 2 1 훼퐿 1 퐶(퐿) 1 푈 (푊) (퐿 + 2퐿0)휎퐿 + (퐿 − 2푀0)휎푀 + 2(푀0 − 푏 = + − ′ [ ] (4.10) 2 훽퐿 2 퐿 4 푈 (푊) 퐿0 − 퐿)휎퐿푀

So the optimal interest margin S is:

1 훼 훼 1 퐶(퐿) 퐶(퐷) 푆 = 푎 + 푏 = ( 퐷 + 퐿) + ( + ) 2 훽퐷 훽퐿 2 퐿 퐷

′′ 1 푈 (푊) 2 2 − [(퐿 + 2퐿0)휎퐿 + (퐿 + 퐷)휎푀 + 2(푀0 − 퐿)]휎퐿푀 4 푈′(푊)

(4.11)

Thus, from the theoretical model in (4.11), the determinants of interest rate spreads are:

a) The competitiveness of the market structure. This depends on the elasticity of the demand for loans

and supply of deposits (β), such that the less elastic the demand for credit (or supply of deposits),

the less will be the value of β, and the bank will be able to apply high margins if it exercises

monopoly power.

b) Operating costs—C(D) and C(L). Firms that incur high unit costs will logically need to work with

higher margins to enable them cover their higher operating costs.

c) Risk aversion, expressed by the coefficient of absolute risk aversion-푈′′(푊)/푈′(푊), where on

the assumption that the bank is risk averse 푈′′(푊) < 0 then [푈′′(푊)/푈′(푊)] > 0. Obviously,

the more risk-averse banks charge higher margins.

2 d) The volatility of money market interest rates (휎푀). The more volatile the money market interest

rates, the greater will be the market risk, and thus higher margins as the banks will require a higher

premium.

2 e) The credit risk (휎퐿 ).The greater the uncertainty of the return expected on loans granted (default

risk), the greater will be the spread.

f) The covariance or interaction between interest rate risk and credit risk 휎퐿푀.

g) The average size of credit (퐿 + 2퐿0) and deposit (퐿 + D). The unit margins are an increasing

function of the average size of operations (deposits and loans). This is because for a given value

of credit risk and market risk, an operation of greater size would mean a greater potential loss, so

the bank will require a greater margin.

In addition to the variables in the theoretical framework, empirical literature shows the following variables as other bank specific determinants of interest rate spreads. As such the study includes them in the empirical model specification.

h) Capital adequacy ratio. The changes in capital requirements have a direct impact on the bank’s

optimal interest margin (Ahokpossi, 2013; Crowley, 2007). An increase in the capital adequacy

ratio is shown to increase interest rate spreads under the reasonable assumption of risk aversion.

i) Non-interest income. Non-interest income supplements revenue from the traditional banking

activities. Banks with relatively high non-interest income tend to charge lower spreads to maintain

their market power (Carbo & Rodriguez, 2007; Maudos & Solis, 2009).

With the aim of controlling for the effects of macroeconomic factors on interest spreads, the following macroeconomic variables are often used in empirical studies, and are also included in the empirical model.

j) Inflation rate. Hyperinflation reduces real interest rates and thus profitability of banks. To that

effect, inflation is directly proportional to interest rate spreads (Beck & Hesse, 2009; Chirwa &

Mlachila, 2004; Nampewo, 2013).

k) Real GDP growth rate. GDP growth rates are a reflection of the level of economic activity in the

country. High levels of economic activity are linked to lower interest rate spreads (Beck & Hesse,

2009; Gelos, 2006; Nampewo, 2013).

l) Exchange change rate volatility. This mostly affects foreign owned banks given that their

profitability depends on the variations in exchange rates. Exchange rate depreciations are

associated with higher interest rate spreads (Folawewo & Tennant, 2008).

m) M2/GDP. It shows the level of development of the financial sector. Low level of M2/GDP is

indicative of a less developed financial system. As such, low level of M2/GDP is associated with

higher interest rate spreads (Folawewo & Tennant, 2008).

Generally, the above variables are broadly categorised into bank specific characteristics, market structure and regulation, and macroeconomic factors. As a modification to Ho and Saunders (1981) model as extended by Maudos and Fernandez de Guevara (2004), the empirical model includes all these three categories of factors that affect interest rate spreads.

4.2. Empirical model specification

To estimate the marginal impact of the determinants of interest rate spreads, we model interest rate spreads as a linear function of bank-specific characteristics, banking industry specific variables, and macroeconomic conditions. As in Beck and Hesse (2009), Chirwa and Mlachila (2004), Maudos and Solis

(2009), Mugume et al. (2009), Rebei (2014), and Were and Wambua (2014), among other studies, the interest rate spreads function is given as:

푗 푘 푙 퐼푅푆푖푡 = 푓(푋푖푡, 훸푡 , 푋푡, 푢it) (4.12)

푗 푘 where 퐼푅푆푖푡 is the interest rate spread of bank i in period t, 푋푖푡 denotes bank-specific variables, 푋푡 denotes

푙 bank industry specific variables, 푋푡 denotes the macroeconomic factors and 푢푖푡is the disturbance term.

We use a dynamic panel model to estimate determinants of the interest rate spread. This is because interest rate spreads have dynamic relationships with their determinants (Folawewo &Tennant, 2008; Maudos &

Solis, 2009). Under dynamic panel estimation, the lagged dependent variable—interest rate spreads in period 푡 − 1—is included among the regressors. Thus, the dynamic panel estimation equation is specified as follows:

7 2 5 푗 푘 푙 퐼푅푆푖푡 = 훼 + 훿퐼푅푆푖푡−1 + ∑ 훽푗푋푖푡 + ∑ 휂푘훸푡 + ∑ 휙푙푋푡 +푢푖푡 (4.13) 푗=1 푘=1 푙=1

푢푖푡 = 휇푖 + 휆푡 + 푣푖푡 (4.14)

2 2 2 with 휇푖 ∼ IID(0,휎휇 ), 휆푡 ∼ IID(0,휎휆 ) and 푣푖푡 ∼ IID(0,휎푣 ); 푖 = 1, … ,24, 푡 = 1, … , 푇푖

퐼푅푆푖푡−1 is interest rate spread of bank 푖 in period 푡 − 1.

The error term, 푢푖푡, in (4.14) is composed of the time specific effects (휆푡), individual bank specific effects

(휇푖) and the remainder error term (푣푖푡). Since the panel is unbalanced, we use 푇푖 instead of the exact T in the notations above. The maximum T in the panel is 11.

IRS in equation (4.13) is interest rate spreads—the dependent variable of this study. Following Brock and

Rojas-Suarez (2000), Chirwa and Mlachila (2004), and Mugume et al. (2009), we measure interest rate spreads as the difference between the ratio of interest received to total loans and the ratio of interest paid to total deposits for individual commercial bank in a given year. Ideally, interest rate spreads would have been defined as the difference between the weighted average lending and weighted average savings and time deposit rate for individual commercial banks (Sologoub, 2006). However, data on ex-ante interest rates is not available given that annual bank reports and financial statements of commercial banks do not report such information.

The explanatory variables of the study in (4.13), as aforementioned, are composed of bank-specific characteristics, banking industry specific variables, and macroeconomic factors. We expand the vectors of the three categories of independent variables for their description and measurement. The expansion of the vector of bank specific characteristics in (4.13) gives:

7 푗 ∑ 훽푗푋푖푡 = 훽1퐶푅푖푡 + 훽2퐿푅푖푡 + 훽3퐵푆푖푡 + 훽4푂퐶푖푡 + 훽5푅푂퐴푖푡 + 훽6푁퐼퐼푖푡 푗=1

+ 훽7퐶퐴푅푖푡 (4.15) with 훽1, 훽2, 훽4, 훽5, 훽7 > 0 and 훽3, 훽6 < 0

CR is the credit risk. As in Ahokpossi (2013), we measure credit risk as the ratio of non-performing loans to total loans. Credit risk shows the possibility of loan default by bank customers. As credit risk increases, financial institutions are expected to raise their lending rates to compensate for lost interest revenue due to loan default (Chirwa & Mlachila, 2004; Grenade, 2007; Jamaludin et al., 2015; Mugume et al., 2009).

As such, a positive relationship between credit risk and interest rate spreads is expected in this study.

LR is the liquidity risk. It shows the possibility of a bank being unable to meet its short-term financial demands of its customers as well as its short-term expenses. Following, among others, Ahokpossi (2013), and Beck and Hesse (2009), liquidity risk is measured as a ratio of liquid assets to deposits and short term financing. Financial institutions with high liquidity risk tend to keep high levels of reserves over and above the regulatory reserve requirement or borrow emergency funds at high costs in the interbank market. This is a self-imposed cost to banks for prudential reasons or as a result of regulation (reserve requirement).

Thus banks charge a liquidity premium to compensate the liquidity risk, leading to higher spreads (Beck

& Hesse, 2009; Islam & Nishiyama, 2016; Manamba, 2014; Mannasoo, 2012). A priori a positive relationship is expected between liquidity risk and interest rate spreads.

BS is the bank size. Following Beck and Hesse (2009), and Almarzoqi and Naceur (2015), it is measured using total assets of individual banks in each year. A logarithm of the total asset is taken to reduce the

51 magnitude of the figures whilst maintaining the properties of the variable. Bank size is used to test the existence of economies of scale. Bigger banks are expected to attract large pool of deposits, leading to favourable lending rates in addition to the large volume of loans (Beck & Hesse, 2009). Premised on this, the study expects a negative relationship between bank size and interest rate spreads.

OC is operating costs. As in Mugume et al. (2009), it is measured as a ratio of operating costs to total assets. A higher cost of financial intermediation will drive up interest rates on loans whilst depressing interest rates on deposits hence wide interest rate spreads (Afanasieff et al., 2002; Grenade, 2007; Mugume et al., 2009). Thus, the study expects operating costs and interest rate spreads to be directly proportional.

ROA is the return on assets. It is used as an indicator of how profitable a bank is relative to its total assets.

As in Siddiqui (2012), return on assets is calculated as net income divided by average total assets of the bank. A priori, it is expected that return on assets positively affect interest rate spreads given that more profitable banks, on average, also charge higher spreads and earn higher margins (Beck & Hesse, 2009).

NNI is the non-interest income. Just as Maudos and Solis (2009), it is measured as the ratio of non-interest income to total assets. It measures the contribution of non-core business activities: commission, brokerage fees, capital gains, dividends, and income from foreign exchange transactions, among others, towards profitability of a bank. Banks with diversified and stable revenue sources are expected to influence the pricing of loan products and therefore may charge lower margins due to subsidisation of traditional banking activities (Carbo & Rodriguez, 2007; Maudos & Solis, 2009; Mujeri & Younus, 2009). Therefore, a negative relationship between non-interest income and interest rate spreads is expected.

Finally, CAR is capital adequacy ratio. It is measured as the total shareholder’s equity divided by risk weighted assets. As noted in the theoretical framework, the changes in capital requirements have a direct impact on a bank’s optimal interest rate spreads. An increase in the capital adequacy ratio is shown to increase interest rate spreads under the reasonable assumption of risk aversion (Ahokpossi, 2013).

The expansion of the vector of banking industry specific variables in (4.13) gives:

2 푙 ∑ 휂푙훸푡 = 휂1퐻퐻퐼푡 + 휂2퐹푂푅퐸퐼퐺푁푡 푘=1

(4.16) with 휂1 > 0 and 휂2 < 0

HHI is the Herfindahl-Hirschman index (deposit and loan). It measures market concentration in the banking sector. As in Mugume et al. (2009), the HHI is computed as the sum of the square of the market share (loans or deposits) of each bank. A HHI of below 0.01 indicates a highly competitive market, whilst a HHI of between 0.01and 0.1 indicates an unconcentrated market. Furthermore, a HHI of between 0.1 and 0.18 indicates a moderately concentrated market and a HHI of above 0.18 indicates a highly concentrated market. The calculation of HHI takes into account the relative size and distribution of the banks in a market and approaches zero when a market consists of a large number of banks of relatively equal size. A highly concentrated market promotes collusive behaviour among banks, particularly in their pricing behaviour (Ahokpossi, 2013; Almarzoqi & Naceur, 2015; Jamaludin et al., 2015). Hence, a positive relationship is expected between HHI and interest rate spreads.

FOREIGN is the foreign bank participation. As in Beck and Hesse (2009), it measures the market share of foreign banks in the loans market. Generally, increased foreign bank participation is associated with lower bank spreads (Ahokpossi, 2013). As such, a negative relationship is expected.

The expansion of the vector of macroeconomic factors in (4.13) yields:

5 푚 ∑ 휙푚푋푡 = 휙1퐼푁퐹푡 + 휙2푅퐺퐷푃푡+휙3푇퐵푅푡+휙4퐸푅푉푡 + 휙5푀2/퐺퐷푃푡 푚=1

(4.17) with 휙1, 휙3, 휙4, > 0 and 휙2, 휙5 < 0

INF is inflation. It is measured as the annual change in consumer price index. A higher rate of inflation is expected to lead to higher interest rate spreads given that it reduces real interest rates (Almarzoqi &

Naceur, 2015).

RGDP is the annual real GDP growth rate. As output growth slows down during the business cycle, creditworthiness deteriorates. Other things being equal, this is likely to be reflected in the higher bank loan rates, leading to higher interest rate spreads (Beck & Hesse, 2009; Gelos, 2006; Nampewo, 2013).

TBR is the treasury bill rate. It is used to measure interest rate risk in the money market. As in Mugume et al. (2009), it is measured as a 91-day annualized treasury bill rate. Unlike loans, treasury bills are risk free given that they are backed by government. This makes them one of the safest forms of investment in the world. An increase in the treasury bill rate induces banks to invest in these short term instruments instead of loans. This reduces the amount of reserves available for advancing credit to the public hence high interest rate spreads (Nampewo, 2013). To that effect, a positive relationship is expected between interest rate spreads and treasury bill rates.

ERV is exchange rate volatility. It is used to measure the external macroeconomic instability. Following

Folawewo and Tennant (2008) exchange rate volatility for each year is calculated as the standard deviation of the percentage change in the real UShs/US$ exchange rate for the preceding three years. Since increased macroeconomic instability heightens the risk faced by commercial banks, exchange rate volatility is expected to be positively correlated with interest rate spreads (Beck & Hesse, 2009; Folawewo, &

Tennant, 2008).

Last but not least, M2/GDP is broad money supply to GDP. It is used as a measure of the level of financial development. It captures the degree of monetisation in the financial system of an economy. A lower level of monetisation of the financial system may reflect lower level of efficiency in intermediation process, thus leading to higher spreads (Crowley, 2007; Folawewo & Tennant, 2008; Nampewo, 2013). In Uganda, the M2/GDP ratio shows an increasing trend. Therefore, a negative relationship between M2/GDP and interest rate spreads is expected.

4.3. Data sources

We obtained the data on bank specific and industry specific variables from the audited financial statements of the 24 commercial banks included in the study for the period 2005-2015.13 The financial statements are obtained from Bank of Uganda. For macroeconomic variables, data is collected from Bank of Uganda,

Uganda Bureau of Statistics, and World Bank Development Indicators database.

Table 4.1. Data sources

Variable Definition Source Dependent variable The difference between interest received divided by total Interest rate spread loans and interest paid divided by total deposits for BoU individual commercial bank in a given year

Explanatory variables (A) Bank specific variables The ratio of NPLs to total loans of each bank in each Credit risk BoU year. A ratio of liquid assets to deposits of each bank in each Liquidity risk BoU year Bank size Logarithm of each bank’s total assets in each year BoU A ratio of each bank’s operating costs to total assets in Operating costs BoU each year Net income divided by average total assets of each bank Return on assets BoU in each year The ratio of non-interest income to total assets of each Non-interest income BoU bank in each year Total shareholder’s equity divided by risk weighted Capital adequacy ratio BoU assets

(B) Banking industry specific variables The sum of the square of the market share (loans or HHI BoU deposits) of each bank in each year Foreign bank The percentage of foreign bank market share in the loans BoU participation market in each year

(C) Macroeconomic variables Inflation The annual change in consumer price index BoU BoU, UBOS, Real GDP growth rate Annual real GDP growth rate WDI Treasury bill rate Annual average 91-day treasury bill rate BoU The standard deviation of the percentage change in the Exchange rate volatility BoU real US$ exchange rate for the preceding three years M2/GDP Annual ratio of M2 to GDP WDI Note: BoU is Bank of Uganda, IMF is International Monetary Fund, UBOS is Uganda Bureau of Statistics, WDI is the World Bank World Development Indicator data

13 See Appendix A for the list of banks

4.4. Estimation techniques

Unlike static panel data models, the dynamic panel data regression described in (4.13) and (4.14) is characterised by two sources of persistence over time: autocorrelation due to the presence of a lagged dependent variable among the regressors; and bank specific effects characterising the heterogeneity among the banks. Consequently, this persistence over time renders the static panel estimation methods of pooled ordinary least squares (OLS), fixed effects, and random effects biased and/or inconsistent

(Arellano & Bond, 1991; Baltagi, 2005; Cameron & Trivedi, 2005; Hsiao, 2014).

The problems of biasness and inconsistency associated with static panel estimators in dynamic panel data have been dealt with by, among others, Anderson and Hsiao (1981), Arellano and Bond (1991), Ahn and

Schmidt (1995), Arellano and Bover (1995), and Blundell and Bond (1998) using either instrumental variables (IV) or generalised method of moments (GMM) estimation methods. The GMM estimation methods are more efficient than IV estimation methods. In fact, unlike GMM estimation methods, the IV estimation methods lead to consistent but not necessarily efficient estimates since they do not make use of all the available moment conditions (Ahn & Schmidt, 1995). As such, the study uses the GMM estimation methods given that they are more efficient than IV estimation method. Such GMM estimation methods include Arellano and Bond (1991) first difference GMM estimator, the levels GMM estimator, and Blundell and Bond (1998) system GMM estimator.

On the GMM estimation methods, Arellano and Bover (1995), Blundell and Bond (1998), and Bond

(2002) note that the first difference GMM estimator of Arellano and Bond (1991) has poor finite sample properties. It is biased downwards particularly when the time dimension (푇) is small. This is likely to be an issue in this study since 푇 = 11 and 푁 = 24. Consequently, we use the system GMM estimator to improve the precision and reduce the finite sample bias common with Arellano and Bond estimator

(Blundell, Bond & Windmeijer, 2000). Furthermore, we use the two step procedure given that the two- step system GMM estimator is asymptotically more efficient compared to one-step estimator (Arellano &

Bond, 1991; Blundell & Bond, 1998). Lastly, the systems GMM estimator may suffer from proliferation

56 of instruments which could lead to inconsistent estimates (Roodman, 2009). To reduce the number of instruments, we restrict lagged instruments to the third lags of the endogenous variable.

The consistency of the system GMM estimator depends on the validity of the over-identifying restrictions.

As such, we use Sargan test for over-identifying restrictions—the standard test for the validity of the moment conditions used in GMM estimation procedure (Blundell et al., 2000). The null is that the over- identifying restrictions are valid. Additionally, the consistence of the system GMM estimator relies on the fact that 퐸(훥푣푖푡훥푣푖,푡−2) = 0 (Baltagi, 2005). Consequently, the study uses the Arellano and Bond (1991)

Hausman-type test for first and second order serial correlation for disturbances of the first differenced equation. The null hypothesis is that there is no second order serial correlation in the disturbances of the first differenced equation.

4.5. Panel unit root tests

Though it is common to test for unit roots in time series studies, testing for unit roots in panels is recent.

Unit root tests are aimed at avoiding working with non-stationary data that leads to spurious regressions and meaningless hypothesis testing. Indeed, if the variables in the regression are not stationary, the standard assumptions for asymptotic analysis cannot hold. Thus, the study cannot validly undertake hypothesis tests about the regression estimates (Baltagi, 2005). Commonly used tests for unit root in panel data are Levin, Lin and Chu (2002) (LLC) test, Im, Pesaran and Shin (2003) test, Breitung’s test,

Combined p-value tests, Harris-Tzavalis test and Residual-Based Langrage Multiplier (LM) test (Baltagi,

2005; Cameron & Trivedi, 2005; Hsiao, 2014). Although all these tests test unit roots in panel data, the

LLC, Breitung’s, Harris-Tzavalis, and Residual-Based LM tests require strongly balanced panels, and the

IPS works on unbalanced panels with a minimum of at least 10 time periods, which is not the case with the data in this study. As such, we use Combined p-value tests (ADF-Fisher type and PP-Fisher type tests) to test for stationarity of the data since they assume individual unit root processes and do not required balanced panels (Baltagi, 2005).

CHAPTER FIVE

PRESENTATION, INTERPRETATION, AND DISCUSSION OF RESULTS

5.0. Introduction

This chapter presents, interprets, and discusses the results of the study. Section 5.1 of the chapter, specifically, gives data characteristics: descriptive statistics, pairwise correlation, and panel unit root tests.

Results are presented, interpreted, and discussed in section 5.2. Finally, section 5.3 gives the summary of the results.

5.1. Data characteristics

5.1.1. Descriptive statistics

The study uses an unbalanced panel of 24 commercial banks regulated by Bank of Uganda as of 2015 for the period 2005-2015.14 The new banks that were licensed during the period of study are included while those that exited the industry are excluded from the study. With the exception of bank size, HHI (loan and deposit), and exchange rate volatility, all variables are measured as ratios (or percentages). Bank size is measured as the logarithm of each bank’s total assets in each year; HHI is measured as the sum of the square of the market share (loans or deposits) of each bank in each year; and exchange rate volatility is measured as the standard deviation of the percentage change in the real US$ exchange rate for the preceding three years. We use STATA 13.0 to examine the data characteristics and run regression equations. Table 5.1 presents the descriptive statistics of the data used in empirical analysis.

From Table 5.1, there are 198 observations on interest rate spread with a mean of 18.43 per cent. As noted in Section 1.1, this mean interest rate spread is higher than the regional and global averages. Moreover,

18.23 per cent standard deviation of interest rate spreads from the mean points suggests that there is significant variation in interest rate spreads across banks. The minimum and maximum interest rate

14 Only Exim Bank-Uganda out of the 25 banks as of 2015 is excluded from the sample because it publishes consolidated financial statements, including its operations in other countries. Nonetheless, its share in both the loans and deposit markets is negligible in Uganda.

58 spreads during the period were −4.41 per cent and 40.86 per cent respectively. Both the minimum and maximum interest rate spreads were registered in 2012 by start-up banks in their first year of operations.

This could suggest that even among start-up banks, there is significant variations in their intermediation efficiencies even if they are established in the same year.

Table 5.1. Descriptive statistics

Standard Variable Observations Mean Minimum Maximum deviation Interest rate spread 198 0.1843 0.1823 −0.0441 0.4086 Credit risk 196 0.0226 0.0309 −0.0057 0.2007 Liquidity risk 196 0.8900 1.0466 0.2150 2.5383 Bank size 198 11.5057 0.5376 10.2626 12.5716 Operating costs 198 0.0850 0.0577 0.0214 0.5162 Return on assets 198 0.0156 0.0364 −0.1773 0.1390 Non-interest income 198 0.0401 0.0220 0.0011 0.1437 Capital adequacy ratio 146 0.3295 0.2497 0.1042 1.3619 HHI (Loan) 198 0.1176 0.0256 0.0943 0.1840 HHI (Deposit) 198 0.1226 0.0365 0.0894 0.2033 Foreign bank participation 198 0.8036 0.0272 0.7725 0.8571 Inflation 198 0.0893 0.0484 0.0399 0.1868 Real GDP growth 198 0.0627 0.0218 0.0356 0.1078 T-bill rate 198 0.1077 0.0355 0.0501 0.1583 Exchange rate volatility 198 0.0570 0.0290 0.0170 0.1034 M2/GDP 198 0.2123 0.0130 0.1932 0.2362 Source: Author’s calculations

Given that credit risk was defined as a ratio of non-performing loans to total loans, then descriptive statistics show that, on average, 2.26 per cent of the loans of the banking sector were non-performing loans during the period 2005−2015. This is a relatively low level of non-performing loans to total loans ratio. However, credit risk differ from bank to bank as the minimum and maximum level of credit risk during the period were −0.57 per cent and 20.07 per cent respectively. Negative values of the variables are attributed to recovery of loans that were formerly classified as bad debts by a few banks in some years.

Liquidity risk averaged 89 per cent with a minimum and maximum of 21.50 per cent and 2.5383 per cent respectively during the period of study. In addition, start-up banks had more liquid assets relative to deposits in their initial years of operations. Since, liquidity ratio was measured as a ratio of liquid assets to deposits of each bank in each year, the statistics point to the stability of the banking sector in Uganda.

Bank size, in terms of absolute assets, averaged at UShs. 613.5 billion for the period 2005-2015. The lowest value of assets was UShs. 18.3 billion owned by a start-up in its first year of operation, whilst the highest value of assets was UShs. 3,729.1 billion. In terms of the logged values, bank assets averaged at

11.51 with a minimum of 10.27 and a maximum of 12.57. As expected a priori, large banks over the period of study were traditional foreign banks, whilst small banks were generally banks that were started recently.

As far as operating costs are concerned, banks use on average 8.5 per cent of total assets in operating expenses. The maximum operating costs/assets ratio (51.62 per cent) was incurred by a new bank during its second year of operation, whilst the minimum operating costs/assets ratio (2.14 per cent) was incurred by a traditional foreign owned bank. This, as also noted by Beck and Hesse (2009), suggests that established banks incur relatively lower operating expenses compared to start-up banks.

The average return on assets for all banks was 1.56 per cent with a minimum and maximum return on assets of −17.73 per cent and 13.9 per cent respectively. Paradoxically, the minimum and maximum return on assets were both registered by new banks in their first year of operations. Moreover, some of the new banks made profits in their first year of operation whilst others took more than 8 years to breakeven, thus pointing to the divergence in efficiency of start-up banks.

Non-interest income averaged at 4.01 per cent during the period. The minimum non-interest income (1.1 per cent) was recorded by a start-up bank, whilst the maximum non-interest income (14.37 per cent) was earned by an established bank. Generally, statistics show that Uganda’s commercial banking sector is less diversified. Moreover, established banks have more diversified activities compared to new banks.

Capital adequacy ratio had an average of 32.95 per cent, with a minimum of 10.42 per cent and a maximum of 136.19 per cent. Given the relatively high ratio of equity to risk weighted assets, the descriptive statistics suggest that the banking sector is well capitalised and thus stable.

For industry specific factors, HHI (loan) averaged at 0.1176 with a minimum and maximum of 0.0943 and 0.1840 respectively, whilst HHI (deposit) averaged at 0.1226 with a minimum and maximum of

0.0894 and 0.2033 respectively. Thus, Uganda’s banking sector is moderately concentrated given that both the HHI (loan) and HHI (deposit) averages are within the range of HHI of moderately concentrated markets (0.1 to 0.18). Moreover, the descriptive statistics also show that there is relatively more concentration in the deposits market compared to the loans market. Whilst foreign bank participation in the loans markets averaged at 80.36 per cent. The minimum and maximum levels of foreign bank participation were 77.25 per cent and 85.71 per cent respectively. Generally, Uganda’s banking sector is dominated by foreign owned banks. The dominance is partly attributed to the financial liberalisation policy that led to entry of many foreign banks.

Turning to macroeconomic variables, inflation averaged at 8.93 per cent with a recorded minimum and maximum of 3.99 per cent and 18.68 per cent respectively. The highest level of inflation was recorded in

2011. The high inflation could probably explain the high lending rates that were charged by banks in that year. Generally, high inflation rates increase the cost of doing business and reduce the real interest rates; which often lead to increase in nominal interest rates, especially leading rates, by banks.

Real GDP growth averaged at 6.27 per cent with the highest level of economic growth (10.78 per cent) recorded in 2006, whilst the lowest level of growth (3.56 per cent) recorded in 2013. The descriptive statistics show a relatively sustained trend of economic growth over the period of study. Overall, high growth rates are associated with better performance of the banking sector and lower interest rate spreads.

The 91-day treasury bill rate averaged at 10.77 per cent with minimum and maximum rates of 5.01 per cent and 15.83 per cent respectively. The high treasury bill rates are indicative of high levels of government domestic borrowing. Holding other factors constant, governments often set high interest rates to attract more investors in purchasing government securities so as to finance budget deficits.

Exchange rate volatility averaged at 0.057 with a recorded minimum and maximum of 0.017 and 0.1034 respectively. Since each year’s exchange rate volatility was measured as the standard deviation of

US$/UShs. exchange rate for the preceding three years, the statistics suggest that there was high volatility

61 in exchange rates over the period of study. Holding other factors constant, volatility in exchange rates is associated with wider interest rate spreads.

Finally, broad money supply to GDP (M2/GDP) averaged at 21.23 per cent for the period of study with a minimum of 19.32 per cent recorded in 2005 and a maximum of 23.62 per cent recorded in 2008. Overall,

M2/GDP has been increasing over time though still low; thus pointing to the low level of development of the financial sector.

5.1.2. Pairwise correlation matrices of the variables

Multicollinearity leads to inefficient estimates though it does not violate OLS assumptions (Carl &

Praveen, 2002). Extreme multicollinearity is tested using pairwise correlation matrix. The bank specific variables have between 146 and 198 observations. Industry specific and macroeconomic variables have observations for 11 years but repeated across panels; this gives 198 observations for bank specific and macroeconomic variables. Furthermore, bank specific variables vary both across cross-sections and over time, whilst industry specific and macroeconomic variables only vary over time. As such, we run separate two pairwise correlation matrices for these groups of variables given their distinct characteristics to ascertain extreme multi-collinearity among variables. The pairwise correlation matrices of bank specific variables, and industry specific and macroeconomic variables are presented in Tables 5.2 and 5.3 respectively. Overall, severe multi-collinearity that compromises efficiency of regression results is neither detected among both bank specific variables nor industry specific and macroeconomic variables.

However, this conclusion excludes the correlation between HHI (loan) and HHI (deposits). HHI (loan) and HHI (deposits) are highly and significantly correlated (0.986) at 1 per cent. This high correlation is expected given they both measure the level of market concentration in the banking industry. As such, HHI

(loan) and HHI (deposits) are interchangeably, rather than concurrently, run in regressions to remedy collinearity.

Table 5.2. Correlation matrix of bank specific variables

IRS CR LR BS OC ROA NII CAR IRS 1.000 CR −0.013 1.000 LR 0.016 0.002 1.000 BS −0.028 −0.014 −0.220*** 1.000 OC −0.027 0.096 −0.016 −0.365*** 1.000 ROA 0.087 −0.308*** −0.080 0.371*** −0.594*** 1.000 NII −0.040 0.156** −0.217*** −0.037 0.325*** −0.056 1.000 CAR 0.1918** 0.039 0.541*** −0.559*** 0.043 −0.072 −0.181** 1.000 ***, **, and * indicates that the correlation coefficients are significant at 1 per cent, 5 per cent, and 10 per cent respectively. IRS is interest rate spreads, CR is credit risk, LR is liquidity risk, BS is bank size, OC is operating costs, ROA is return on assets, NII is non- interest income, and CAR is capital adequacy ratio. Source: Author’s calculations

Table 5.3. Correlation matrix of market specific and macroeconomic variables

IRS HHI (LOAN) HHI (DEPOSIT) FOREIGN INF RGDP TBR EVR M2/GDP IRS 1.000 HHI (LOAN) 0.293*** 1.000 HHI (DEPOSIT) 0.263*** 0.986*** 1.000 FOREIGN 0.010 0.148** 0.120* 1.000 INF −0.004 0.074 0.085 0.087 1.000 RGDP 0.024 0.620*** 0.646*** 0.116 0.401*** 1.000 TBR −0.063 −0.388*** −0.416*** 0.254*** 0.390*** −0.188*** 1.000 ERV 0.146** −0.193*** −0.218*** −0.012 −0.124* −0.696*** 0.473*** 1.000 M2/GDP −0.208*** −0.250*** −0.227*** 0.445*** −0.099 0.199*** 0.008 −0.342*** 1.000 ***, **, and * indicates that the correlation coefficients are significant at 1 per cent, 5 per cent, and 10 per cent respectively. IRS is interest rate spreads, FOREIGN is foreign bank participation, INF is inflation, RGDP is real GDP Growth, TBR is T-bill rate, and ERV is exchange rate volatility. Source: Author’s calculations

5.1.3. Panel unit root tests

As noted in Section 4.5, we use Combined p-value tests (ADF-Fisher type and PP-Fisher type tests) to test for stationarity of the data since they assume individual unit root processes and do not require balanced panels. These tests include Fisher’s inverse chi-square (P) test, inverse normal (Z) test, the logit (L*) test, and modified inverse chi-square (Pm) test. Overall, the Z test is reported to outperform other tests and is thus recommended (Baltagi, 2005). However, since all the four tests are reported in

Fisher-type unit root tests in STATA output, we use all in ascertaining the stationarity of data. In all cases, the null hypothesis is that all panels contain unit roots, whilst the alternative hypothesis is that at least one panel is stationary. Table 5.4 presents ADF-Fisher type tests whilst PP-Fisher type tests are presented in Appendix D.

Table 5.4. Panel unit root tests: ADF-Fisher type tests

Variable P Z L* Pm Interest rate spread 246.397*** −7.958 *** −13.248*** 20.893 *** Credit risk 97.037*** −3.498*** −4.104*** 5.321*** Liquidity risk 290.836*** −8.396*** −15.596*** 25.526*** Bank size 87.841*** −3.119*** −3.412*** 4.066*** Operating costs 191.623*** −4.962*** −9.236*** 14.659*** Return on assets 90.85*** −3.504*** −4.004*** 4.374*** Non-interest income 147.073*** −5.178*** −7.307*** 10.112*** Capital adequacy ratio 156.670*** −3.950*** −7.432*** 11.538*** HHI (Loan) 122.835*** −6.543*** −6.711*** 7.638*** HHI (Deposit) 137.555*** −5.590*** −7.221*** 9.140*** Foreign bank participation 99.902*** −5.037*** −4.941*** 5.297*** Inflation 220.723*** −7.692*** −12.515*** 17.629*** Real GDP growth 82.021*** −3.489*** −3.770*** 3.472*** T-bill rate 44.209*** −1.345*** −1.253*** −0.387*** Exchange rate volatility 74.183*** −3.645*** −3.476*** 2.672*** M2/GDP 220.021*** −10.125*** −12.292*** 17.557*** Note: P is the inverse chi-squared statistic; Z is the inverse normal statistic; L* is the inverse logit statistic; and Pm is the modified inverse chi-squared statistic. ***, **, and * indicates significance of the unit root statistics at 1 per cent, 5 per cent, and 10 per cent respectively. Source: Author’s calculations

All unit root test statistics for all variables are statistically significant at 1 per cent. The significance of the unit root tests, therefore, leads to rejection of the null hypothesis at 1 per cent for all variables. That is, all variables do not have a unit root or are I(0).

5.2. Presentation, interpretation, and discussion of regression results

As noted in Section 4.4, we use dynamic panel estimation techniques (two-step systems GMM) to run interest rate spreads equations. Regression results are presented in Table 5.5. We run five regressions in total. Regression (1) contains only bank specific factors as explanatory variables, whilst regressions (2) and (3) have bank specific and industry specific factors as explanatory variables. However, due to collinearity, regression (2) uses HHI (loan) to measure market concentration whilst regression (3) uses HHI

(deposit). Finally, regressions (4) and (5) have all the three categories of variables (bank specific, industry specific, and macroeconomic factors). Again, due to collinearity between HHI (loan) and HHI (deposit), regression (4) is run with HHI (loan) whilst regression (5) is run with HHI (deposit) as a measure of market concentration.

As far as general specification of the dynamic panel regressions is concerned, the Wald 휒2 statistic is significant at 1 per cent in all regressions. Thus, we reject the null hypothesis that all coefficients, in each regression, are simultaneously equal to zero. The rejection of the null implies that the models explain the variations in interest rate spreads. In addition, the Sargan test of over identifying restrictions fails to reject the null that over identifying restrictions are valid in all regressions. As such, there is no over identification of instruments in all models. Furthermore, Arellano and Bond (1981) test for second order autocorrelation in the error term component is insignificant in all regressions. That is, there is no second order autocorrelation in error term. This is in line with the assumptions of dynamic panel estimation: the error term can have first order autocorrelation but there ought to be no second order autocorrelation.

Table 5.5. Dynamic panel regressions: two-step system GMM

Interest rate spread (1) (2) (3) (4) (5) (IRS) 0.1194 0.1469 0.1343 0.0408 0.0061 IRS(L1) (0.0558)** (0.0653)** (0.0668)** (0.0778) (0.0697) 0.3659 0.4285 0.3965 0.2882 0.2809 Credit risk (0.1239)*** (0.1493)*** (0.1484)*** (0.1322)** (0.1276)** 0.0297 0.0286 0.0324 0.0340 0.0410 Liquidity risk (0.0112)*** (0.0134)** (0.0145)** (0.0206)* (0.0219)* −0.0158 −0.0463 −0.0044 −0.0217 −0.0190 Band size (0.0190) (0.0265)* (0.0280) (0.0249) (0.0249) − 0.2828 −0.2789 −0.4555 0.5615 −0.6069 Operating costs (0.0995)*** (0.1300)** (0.1430)*** (0.3009)* (0.2925)* 0.0923 0.3994 0.1369 0.0196 −0.0110 Return on assets (0.3434) (0.4746) (0.4618) (0.6669) (0.6570) 0.8968 0.8776 0.9397 0.6507 0.7055 Non-interest income (0.2214)*** (0.4252)** (0.3733)** (0.3795)* (0.3630)* 0.1107 0.1163 0.0947 0.1221 0.1268 Capital adequacy ratio (0.0231)*** (0.0220)*** (0.0131)*** (0.0281)*** (0.0269)*** −0.1004 0.6785 HHI (Loan) (0.3013) (0.7015) 0.0581 0.4690 HHI (Deposit) (0.2880) (0.4582) 0.1758 0.1782 0.4193 0.4040 Foreign bank participation (0.0836)** (0.0870)** (0.2040)** (0.2019)** 0.2500 0.2007 Inflation (0.1172)** (0.1365) −0.9171 −0.7762 Real GDP growth (0.4463)** (0.4636)* −0.0978 −0.0626 T-bill rate (0.1821) (0.1924) −0.4249 −0.4575 Exchange rate volatility (0.2666) (0.2435)* −1.4080 −1.5690 M2/GDP (0.5095)*** (0.5462)*** 0.2577 0.1822 −0.0455 0.3363 0.3304 CONS (0.1548)* (0.2609) (0.3182) (0.3329) (0.3331) Number of obs 133 133 133 133 133 Number of banks 24 24 24 24 24 Number of instruments 62 64 64 69 69 AB AR(2) [P-values] [0.1956] [0.1787] [0.1701] [0.6425] [0.9156] 휒2(53) 휒2(53) 휒2(53) 휒2(53) 휒2(53) Sargan test =15.6826 =13.9531 =18.8473 = 6.1972 =6.15796 P-values [1.0000] [1.0000] [1.0000] [1.0000] [1.0000] 휒2(8)= 휒2(10)= 휒2(10)= 휒2(15)= 휒2(15)= Wald 휒2 statistic 1861.63*** 949.84*** 1550.90*** 1666.18*** 1590.68*** Standard errors in parentheses ***, **, and * indicates significance at 1 per cent, 5 per cent, and 10 per cent respectively AB=Arellano and Bond test of autocorrelation in differenced errors Source: Author’s calculations

From the regression results in Table 5.5, on one hand, the first lag of interest rate spread (퐼푅푆푖푡−1), credit risk, liquidity risk, bank size, non-interest income, capital adequacy ratio, foreign bank participation, inflation rate, real GDP growth rate, exchange rate volatility, and M2/GDP are variables that are shown to significantly affect interest rate spreads. On the other hand, return on assets, HHI, and treasury bill rates are shown to be insignificant determinants of interest rate spreads and as such these factors are not considered in the interpretation and discussion of results.

To start with, the first lag of interest rate spreads (퐼푅푆푖푡−1) is positively related to interest rate spreads in all regressions. However, the coefficients of 퐼푅푆푖푡−1 are only significant at 5 per cent in regressions

(1), (2), and (3), and insignificant in the overall regressions, (4) and (5), that include macroeconomic variables. Nevertheless, the results imply that higher interest rate spreads in the previous period are associated with higher interest rate spreads in the current period. These findings are consistent with a priori expectations of the study and empirical literature that associate higher interest rate spreads in the previous periods with higher spreads in the current period since interest rate spreads are sticky downwards (Carbo & Rodriguez, 2007; Folawewo & Tennant, 2008).

As expected, higher credit risk is associated with higher interest rate spreads in all regressions. That is, an increase in credit risk translates into higher bank spreads. The coefficients are significant at 1 per cent in regressions with only bank specific and/or bank specific variables (regressions: 1, 2, and 3) and significant at 5 per cent in overall regressions: (4) and (5). Given that credit risk is measured as the ratio of non-performing loans to total loan portfolio, the results are in line with theory and empirical literature that associate high levels of non-performing loans (high credit risk) with higher bank spreads.

Specifically, the findings are in line with the risk based view of interest rate spreads that attributes higher bank spreads to, among other risks, higher credit risk faced by commercial banks (Crowley,

2007). Furthermore, the findings are similar to those of Ahokpossi (2013), Tarus et al. (2012), and

Mugume et al. (2009), among others. Generally, both empirical and theoretical literature acknowledge

67 that higher credit risk resulting from high default rates reduces bank profitability due to loan-loss provisions. Thus, high interest spreads are a risk premium for lending to high risk borrowers.

Liquidity risk is positively related to interest rate spreads in all regressions. Liquidity risk is significant at 1 per cent in regression (1); at 5 per cent in regressions (2) and (3); and at 10 per cent in regressions

(4) and (5). Overall regression results on liquidity risk are in line with the hypothesis of the study.

Without any contradiction, the results are in support of theoretical literature (risk based view) and empirical literature that show that high liquidity ratios are used by banks to safeguard against sudden withdrawals by customers, thus leading to high interest rate spreads (Beck & Hesse, 2009; Islam &

Nishiyama, 2016; Manamba, 2014; Mannasoo, 2012).

Results show that bank size negatively drives interest rate spreads in all regressions. The results show that an increase in the size of the bank leads to a decrease in interest rate spreads. Though this is in line with the expectations of the study, the variable is significant at 10 per cent in only regression (2) and insignificant in all other regressions. Nonetheless, all coefficients in the five regressions suggest that bank size negatively affects interest rate spreads. These results are in agreement with the small financial system hypothesis and most of empirical literature that associate larger banks with lower bank spreads (Almarzoqi & Naceur, 2015; Beck & Hesse, 2009; Willmott, 2012). The small financial system hypothesis attributes the reduction in interest rate spreads to economies of scales that large banks tend to enjoy relative to smaller banks.

The results of operating costs are mixed. Regressions (4), as expected, show that operating costs positively affect interest rate spreads, though significant at just 10 percent. These findings are consistent with the efficiency hypothesis of interest rate spreads, which attributes higher bank spreads to operational inefficiency; that is higher operational costs translate into higher interest rate spreads.

In addition, the findings are in line empirical literature that associate high interest rate spreads with high operational costs (Afanasieff et al., 2002; Mugume et al., 2009; Mujeri & Younus, 2009; Siddiqui,

2012; Tarus et al., 2012). This could be explained by the tendency by banks to transfer their higher

68 operating costs to customers in form of low deposits and/or high lending rates, thus leading to high interest rate spreads. On the contrary, though, regressions (1), (2), (3), and (5) show that operating costs negatively affect interest rate spreads. Moreover, the coefficients are significant at 1 per cent, 5 per cent, 1 per cent, and 10 per cent in regressions (1), (2), (3), and (5) respectively. There is no clear explanation to this contradiction in results in both theory and empirical literature and as such there is need for more investigation.

Results associate higher interest rate spreads with higher non-interest income in all regressions. The variable is significant at 1 per cent in regressions (1); 5 per cent in regressions (2), and (3); and 10 percent in regressions (4) and (5). In contrast to theory and a priori expectation, regression results show that non-interest income is associated with high interest rate spreads. As such, diversification of the banking sector seems to be linked with higher bank spreads. This also contrasts empirical literature such as Almarzoqi and Naceur (2015), Carbo and Rodriguez (2007), Mujeri and Younus (2009), and

Mugume et al. (2009) that associate higher non-interest income with lower bank spreads. These studies consider non-interest income as a somewhat form of compensation for lower revenues due to lower bank spreads and margins. To that effect, the conclusion is always that as non-interest income declines, banks raise lending rates to compensate for the loss in income and vice versa. However, the contradiction between the study findings and other empirical literature could partly be attributed to market concentration which is often enhanced by increment in the banks’ revenue. In fact, market concentration is found to be positively related with interest rate spreads in the study. Moreover,

Willmott (2012) notes that, on average, banks with diversified operations and that incur lower costs do not necessarily charge lower interest rates in Uganda. Furthermore, non-interest income is directly proportional to the volume of transactions in the deposits and loans markets. To that effect, the calculation of interest rate spreads using interest revenue and expenditure could, partly, explain why findings of the study associate an increase in non-interest income with higher interest rate spreads.

All regression results show that capital adequacy ratio is directly proportional to interest rate spreads in all regressions. Thus, an increase in the ratio of equity to risk weighted assets leads to an increase in interest rate spreads. The variable is significant at 1 per cent in all regressions. The findings of the study are consistent with empirical literature; that is capital adequacy ratio is positively related to interest rate spreads (Ahokpossi, 2013; Almarzoqi & Naceur, 2015; Crowley, 2007). In addition, the study findings support the capital structure hypothesis that associates higher bank spreads with banks that have higher equity. This assertion is based on reasoning that under the assumption of risk aversion, shareholders usually demand higher returns on their additional equity. Moreover, banks with higher equity are reluctant in increasing deposit rates given that they can make loans using their capital hence leading to wider spreads (Chirwa & Mlachila, 2004; Saunders & Schumacher, 2000).

Foreign bank participation in the loans market is associated with higher interest rate spreads in all regressions. The coefficients of foreign bank participation are significant at 5 per cent in all regressions containing it as an independent variable, that is regressions (2), (3), (4), and (5). These findings are consistent with findings of Beck and Hesse (2009), and Crowley (2007) that show that foreign share of bank ownership and foreign bank share in the loans market respectively are positively related to interest margins and spreads. The high interest spreads among foreign owned banks, just as in other in developing countries, could partly be due the limited competitive pressures the banks face from local banks (Demirguc-Kunt & Huizinga, 1999). In fact, over 80 per cent of Uganda’s banking sector assets are controlled by foreign banks. On the contrary to study findings, the results are in disagreement with

Ahokpossi (2013) who shows that foreign ownership leads to lower interest margins. Overall, foreign bank participation in the banking sector is associated with high interest rate spreads in Uganda.

Inflation is positively related to interest rate spreads in both regressions with macroeconomic variables.

An increase in the annual inflation rate, as expected, leads to an increase in interest rate spreads.

Generally, inflation leads to decline in the real interest rates which prompts banks to increase their nominal lending rates hence leading to wide bank spreads. However, inflation is significant in only

70 regression (4) at 5 per cent. Nonetheless, it is agreeable in all regressions that higher inflation rates are associated with higher interest rate spreads. This is in line with the macroeconomic view of interest rate spreads that postulates inflation to be positively related to bank spreads. According to the theory, inflation leads to decrease in real interest rates and as such banks tend to set wide spreads to compensate for the loss. Furthermore, the findings of the study are also consistent with empirical literature such as Ahokpossi (2013), Almarzoqi and Naceur (2015), Beck and Hesse (2009), Chirwa and Mlachila (2004), and Mugume et al. (2009). But the study results are totally in disagreement with findings of Afanasieff et al. (2002) and Crowley (2007) that show that inflation rate negatively affects bank spreads.

The level of real GDP growth, as expected a priori, is negatively related to interest rate spreads in both regressions. In addition, its coefficients are significant at 5 per cent and 10 per cent in regressions (4) and (5) respectively. The study findings are consistent with macroeconomic hypothesis that associates an increase in the real GDP growth rates with a reduction in interest spreads. Furthermore, the findings are also in line with findings on GDP growth in empirical literature reviewed (see for example: Beck

& Hesse, 2009; Crowley, 2007; Islam & Nishiyama, 2016; Mugume et al., 2009). However, findings of studies like Afanasieff et al. (2002) and Grenade (2007) that find real GDP growth to be directly proportional to bank spreads are in disagreement with the study findings.

Exchange rate volatility also negatively affects interest rate spreads in both regressions. Its coefficient is significant at 10 per cent in regression (5) but it is shown to be an insignificant determinant of interest spreads in regression (4). These findings are inconsistent with the macroeconomic view of interest rate spreads and most of the empirical literature that often associate higher exchange rate volatility with higher interest rate spreads. For instance, Beck and Hesse (2009), Crowley (2007), Folawewo and

Tennant (2008), Mugume et al. (2009), and Nampewo (2013)—contrary to the study findings—show that exchange rate volatility is positively related to interest spreads. This is hinged on the notion that uncertainty in the foreign exchange market affects the profitability of banks especially foreign owned

71 banks. However, in contradiction, the current study finds that exchange rate volatility is negatively related to interest rate spreads though the coefficients are not significant. This could probably be attributed to the preference by most banks to lend in the local currency given the volatility in the foreign exchange market. As such, it could be inferred that banks tend to lower interest rates on local currency denominated loan to induce customers to borrow in local currencies instead of highly volatile foreign currencies.

Last but not least, high levels of broad money supply to GDP (M2/GDP) are associated lower interest rate spreads. Moreover, it is shown to be a statistically significant determinant of interest rate spreads at 1 per cent in both regressions that include it (M2/GDP) as an explanatory variable. Bearing in mind that M2/GDP is used as a proxy of the level of financial sector development, the results show that high levels of financial development lead to lower bank spreads. Generally, financial sector development is associated with increase in the outreach and scope of financial services (financial deepening and widening) which increase competition in the financial markets and thus result into lower bank spreads

(Demirguc-Kunt et al., 2017). The study findings on M2/GDP are in agreement with the macroeconomic view of interest rate spreads and empirical literature such as Crowley (2007).

CHAPTER SIX

CONCLUSIONS AND RECOMMENDATIONS

6.0. Introduction

This chapter draws conclusions of the study and makes policy recommendations. Specifically, section

6.1 gives the conclusions of the study, while section 6.2 gives the policy recommendations of the study.

Furthermore, limitations to the study and recommendations for further research are discussed in section

6.3.

6.1. Conclusions of the study

In comparison to regional (EAC and SSA) and international standards, interest rate spreads have persistently remained high in Uganda despite the financial liberalisation that was undertaken by the government in the early 1990s. The high interest rate spreads reflect the high cost of financial intermediation which undermine the growth of savings, investment, employment, and consequently the country’s economic growth. Thus these adverse effects of high cost of financial intermediation justify the importance of analysing the determinants of interest rate spreads in Uganda, so as to inform policy formulation.

The current study used panel data collected from audited commercial banks’ financial statements,

Bank of Uganda, Uganda Bureau of Statistics, and World Development Indicators for the period 2005-

2015 to investigate the determinants of interest rate spreads in Uganda’s commercial banking sector.

The study tested the hypotheses that: (1) bank specific factors: credit risk, liquidity risk, operating costs, return on assets, and capital adequacy ratio positively affect interest rate spreads, whilst bank size and non-interest income have a negative impact on interest rate spreads; (2) industry specific factors: HHI and foreign bank participation in the loans market are positively and negatively related to interest rate spreads respectively; and (3) macroeconomic factors: inflation rate, 91-day treasury bill rate, and exchange rate volatility are directly proportional to interest rate spreads, whilst real GDP growth rate and M2/GDP are inversely proportional to interest rate spreads. As a major contribution

73 to literature, the study used dynamic panel estimation techniques and more recent data unlike other studies on interest rate spreads in Uganda. The use of dynamic panel estimation techniques was intended to capture the dynamics of interest rate spreads adjustment.

Results show that among the bank specific factors, interest rate spreads are significantly increased by an increase in credit risk, liquidity risk, non-interest income, and capital adequacy ratio. On the other hand, bank size is shown to be negatively related to interest rate spreads. Results on operating costs and return on assets are inconclusive. However, unlike return on assets, operating costs are shown to be significant determinants of interest rate spreads in all regressions; with four out of five regressions, contrary to the a priori expectations, showing that operating costs as associated with a reduction in interest rate spreads. As far as banking industry factors are concerned, foreign bank participation is positively and significantly related to interest rate spreads. Results on HHI are inclusive and insignificant. For the case of macroeconomic variables, results show that high inflation rates translate into high bank spreads, whilst high real GDP growth rates, high exchange rate volatility, and high levels of broad money supply to GDP (M2/GDP) are associated with lower bank spreads.

Paradoxically, 91-day treasury bill rate and exchange rate volatility are associated with lower bank spreads. However, coefficients for 91-treasury bill rate are insignificant. The lagged interest rate spreads are shown to be positively and significantly related to interest rate spreads. Overall, the results show the bank specific characteristics as the most significant determinants of interest rate spreads compared to industry specific and macroeconomic factors.

6.2. Policy recommendations

As far as bank characteristics are concerned, the study finds that increase in the volume of non- performing loans significantly increases bank spreads. As such, mechanisms that encourage loan repayment should be strengthened. This is currently being spearheaded by Credit Reference Bureau that was established under the Financial Institutions Act, 2004. However, the bureau should also consider collecting and sharing information on microfinance institutions credit operations. More often

74 borrowers obtain multiple loans in microfinance institutions whose data is not captured by Credit

Reference Bureau, which negatively affects loan repayment in commercial banks. In addition, Credit

Reference Bureau should consider using National Identification Registrations Authority data to enhance its role of credit information sharing.

It is also shown that excess liquidity and high capital adequacy ratios positively influence interest rate spreads. Moreover, treasury bill rates, contrary to theory, are shown to be negatively related to interest rate spreads; thus implying that commercial lending to government does not crowd out the private sector. To that effect, banks should be encouraged to offer credit to the public. Probably strengthening of property rights can go a long way in encouraging commercial bank lending since it could reduce the credit risk faced by banks as well increase the number of credit worthy bank customers. Indeed, given that most of the collateral in Uganda relates to land, a long term measure to encourage lending among commercial banks would be increased registration of land, and strengthening of the legal system particularly the land and commercial divisions to resolve land related and commercial cases respectively. At bank level still, the size of the bank is shown to be negatively related to interest rate spreads. As such measures that increase the assets of banks should be encouraged. Such measures could be driven towards increasing deposit mobilisation and enhancing financial inclusion, probably through agency banking.

At a macro-level, Bank of Uganda should maintain its stance on curbing inflation given that among the macroeconomic variables, it is found to lead to higher interest spreads. This is because inflation generally reduces real interest rates. Furthermore, enhancement of real growth rates of GDP can also go a long way in reducing bank spreads. This could be promoted by the continuous investment in the infrastructure to enhance the economy’s productivity as well as encourage regional integration to widen the market of the country’s products. Financial sector development is also found to be crucial in reducing interest rate spreads. As such, developments such as mobile money savings and credit facilities, agency banking, bancassurance and Islamic banking should be encouraged to widen the

75 outreach and scope of banking services. Moreover, the Bank of Uganda could restrict the circulation of large denomination notes to encourage the use of banking services among the population. This would in turn increase the resources available to financial institutions to lend to public, thus reducing interest rate spreads.

6.3. Limitations of the study and recommendations for further research

The exclusion of bank-like financial institutions (Tier 2), MDIs (Tier 3), and other financial sector players like savings and credit cooperatives not regulated by Bank of Uganda is one of the limitations of the study in explaining the variation in interest rate spreads given that they also take part in financial intermediation though on a small scale. However, this was aimed at avoiding a scenario of coming up with a seemingly unrelated panel as bank-like institutions and MDIs are not homogeneous to commercial banks and as such poses estimation challenges. In addition, the exclusion of failed banks is another challenge as they too took part in the intermediation process during the period of their existence. However, this was aimed at controlling attrition which can also pose estimation challenges in panel data. Nevertheless, these two limitations present an opportunity for further research which could model interest rate spreads, whilst catering for attrition and/or using seemingly unrelated panels.

Furthermore, future researcher could also, unlike other studies that rely on macroeconomic and industry specific variables, consider using bank level information in examining the determinants of interest rate spreads at regional level (EAC and SSA). This is because bank characteristics give more reliable information on the determinants of interest rate spreads compared to macroeconomic variables.

BIBLIOGRAPHY

Afanasieff, T. S., Lhacer, P. M. V., & Nakane, M. I. (2002). The determinants of bank interest spread

in Brazil (Working Paper No. 46). Brasilia: Banco Central do Brasil.

Ahn, S. C., & Schmidt, P. (1995). Efficient estimation of models for dynamic panel data. Journal of

Econometrics, 68, 5–27.

Ahokpossi, C. (2013). Determinants of bank interest margins in Sub-Saharan Africa. (Working Paper

No. WP/13/34). Washington DC: International Monetary Fund.

Allen, L. (1988). The determinants of bank interest margins: A note. Journal of Financial and

Quantitative Analysis, 23 (2), 231–235.

Almarzoqi, R., & Naceur, S. B. (2015). Determinants of bank interest margins in the Caucasus and

Central Asia (Working Paper No. WP/15/87). Washington DC: International Monetary Fund.

Anderson, T. W., & Hsiao, C. (1981). Estimation of dynamic models with error components. Journal

of the American Statistical Association 76, 598–606.

Angbazo, L. (1997). Commercial bank net interest margins, default risk, interest rate risk and off

balance sheet banking. Journal of Banking and Finance 21, 55–87.

Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence

and an application to employment equations. Review of Economic Studies, 58, 277–297.

Arellano, M., & Bover, O. (1995). Another look at the instrumental variables estimation of error-

component models. Journal of Econometrics, 68, 29–51.

Baltagi, B. H. (2005). Econometric analysis of panel data (3rd ed). Chichester, England: John Wiley

& Sons.

Bank of Uganda (2009). Financial stability report (Issue No. 1). Kampala, Uganda: Bank of Uganda.

Bank of Uganda (2010). Financial stability report (Issue No. 2). Kampala, Uganda: Bank of Uganda.

Bank of Uganda (2011). Financial stability report (Issue No. 3). Kampala, Uganda: Bank of Uganda.

Bank of Uganda (2012). Financial stability report (Issue No. 4). Kampala, Uganda: Bank of Uganda.

Bank of Uganda (2013). Annual supervision report (Issue No. 5). Kampala, Uganda: Bank of Uganda.

Bank of Uganda (2013). Financial stability report (Issue No. 5). Kampala, Uganda: Bank of Uganda.

Bank of Uganda (2014). Financial stability report (Issue No. 6). Kampala, Uganda: Bank of Uganda.

Bank of Uganda (2015). Financial stability report (Issue No. 7). Kampala, Uganda: Bank of Uganda.

Bank of Uganda (2016). Annual supervision report (Issue No. 8). Kampala, Uganda: Bank of Uganda.

Bank of Uganda (2016). Financial stability report (Issue No. 8). Kampala, Uganda: Bank of Uganda.

Bank of Uganda (2017). Financial stability report (Issue No. 9). Kampala, Uganda: Bank of Uganda.

Barajas, A., Steiner, R., & Salazar, N. (1999). Interest spread in banking: Costs, financial taxation,

market power and loan quality in the Colombian case, 1974–1988 (Working Paper No. 110).

Washington DC: International Monetary Fund.

Bategeka, L., & Okumu, J. L. (2010). Banking sector liberalisation in Uganda: Process, results, and

policy options (Research Report). Amsterdam, The Netherlands: Centre for Research on

Multinational Corporations.

Beck, T., & Hesse, H. (2009). Why are interest spreads so high in Uganda? Journal of Development

Economics, 88, 192-204

Beck, T., Cull, R., Fuchs, M., Getenga, J., Gatere, P., Randa, J., & Trandafir, M. (2010). Banking

sector stability, efficiency and outreach in Kenya (Working Paper No. 5442). Washington DC:

The World Bank.

Blundell, R. W., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data

models. Journal of Econometrics, 87, 115–143.

Blundell, R., Bond, S., & Windmeijer, F. (2000). Estimation in dynamic panel data models: Improving

on the performance of the standard GMM estimator. Advances in Econometrics 15, 53–91.

Bond, S. (2002). Dynamic Panel Data Models: A Guide to Micro Data Methods and Practice

(Working Paper No. CWP09/02). London; Centre for microdata methods and practices.

Bond, S., & Windmeijer, F. (2002). Projection estimators for autoregressive panel data models.

Econometrics Journal 5, 457–479.

Brock, L. P., & Rojas-Suarez, L. (2000). Understanding interest rate spreads in Latin America. Journal

of Development Economics, 63, 113-134.

Brownbridge, M. (1998). Financial repression and financial reform in Uganda (Research Report).

Geneva, Switzerland: UNCTAD, Division for Least Developed Countries.

Cameron, C., & Trivedi, K. (2005). Microeconomics. methods and applications. New York:

Cambridge University Press.

Carbo, S. V., & Rodriguez, F. F. (2007). The determinants of banks’ margins in European banking.

Journal of Banking and Finance, 31, 2043–2063.

Carl, F. M., & Praveen, K. K. (2002). The impact of collinearity in regression analysis: The asymmetric

effect of negative and positive correlations. Applied Economics, 34, 667-677.

Chirwa, W., & Mlachila, M. (2004). Financial reforms and interest rate spreads in the commercial

banking system in Malawi. IMF Staff Papers, 51(1), 96-122.

Choi, I. (2001). Unit root tests for panel data. Journal of International Money and Finance 20, 249–

272.

Claessens, S., & Laeven, L. (2004). What drives bank competition? Some international evidence.

Journal of Money Credit and Banking, 36, 563–83.

Claessens, S., Demirguc-Kunt, A., & Huizinga, H. (2001). How does foreign entry affect domestic

banking markets? Journal of Banking and Finance 25, 891–911.

Clarke, G., Cull, R., & Fuchs, M. (2006). Bank privatization in Sub-Saharan Africa: The case of

Uganda Commercial Bank. World Development, 37(9), 1506-1521.

Crowley, J. (2007). Interest Rate Spreads in English-Speaking African Countries (Working Paper

No. WP/07/101). Washington D.C: International Monetary Fund.

Demirguc-Kunt, A., & Huizinga, H. (1999). Determinants of commercial bank interest margins and

profitability: Some international Evidence. World Bank Economic Review, 3, 379–408.

Demirguc-Kunt, A., Cihak, M., Feyen, E., Beck, T., & Levine, R. (2017, August 2). Financial

development and structure dataset. Retrieved from

http://www.worldbank.org/en/publication/gfdr/data/global-financial-development-database

Egesa, A. K. (2010). Financial sector liberalization and productivity change in Uganda's commercial

banking sector (Research Paper No. 202). Nairobi: African Economic Research

Consortium.

Fofack, D. A. (2016). The Determinants of interest spreads: Empirical evidence from Central Africa

Economic and Monetary Community. Journal of Economics and International Finance 8(6),

66-78.

Folawewo, A. O., & Tennant, D. (2008, July 9-11). Determinants of interest rate spread in Sub-

Saharan African countries: A dynamic panel analysis. A paper prepared for the 13th Annual

African Econometrics Society Conference, Pretoria, Republic of South Africa.

Gambacorta, L. (2004). How banks set interest rates? (Working Paper No. 10295). Cambridge:

National Bureau of Economic Research.

Gelos, R. G. (2006). Banking spreads in Latin America (Working Paper No. WP/06/44). Washington

DC: International Monetary Fund.

Grenade, K. H. I. (2007). Determinants of commercial banks interest rate spreads: Some empirical

evidence from the Eastern Caribbean Currency Union (Bank Staff Research Paper No. 07(01)).

Basseterre: Eastern Caribbean Central Bank.

Harvey, C. (1993). The role of commercial banking in recovery from economic disaster in Ghana,

Tanzania, Uganda and Zambia. Brighton, England: Institute of Development Studies.

Hassan, K. M., & Khan, B. (2010). What drives interest rate spreads of commercial banks in Pakistan?

empirical evidence based on panel data. SBP Research Bulletin, 6(2), 15-36.

Hauner, D., & Peiris, S. J. (2005). Bank efficiency and competition in low income countries: The case

of Uganda (Working Paper No. WP/05/240). Washington DC: International Monetary Fund.

Ho, T., & Saunders, A. (1981). The determinants of bank interest margins: Theory and empirical

evidence. Journal of Financial and Quantitative Analysis, 16(4), 581- 600.

Hsiao, C. (2003). Analysis of Panel Data (2nd ed.). Cambridge: Cambridge University Press.

Hsiao, C. (2014). Analysis of Panel Data (3rd ed.). Cambridge: Cambridge University Press.

Im, K. S., Pesaran, M. H., & Shin, Y. (2003). Testing for unit roots in heterogeneous panels. Journal

of Econometrics 115, 53–74.

Islam, S., & Nishiyama, S. (2016). The determinants of bank net interest margins: A panel evidence

from South Asian Countries. Research in International Business and Finance 37, 501-514.

Jamaludin, F., Klyuev, V., & Serechetapongse, A. (2015). What drives interest rate spreads in Pacific

Islands Countries? An Empirical Investigation (Working Paper No. WP/15/96). Washington

DC: International Monetary Fund.

Kasekende, L. A., & Atingi-Ego, M. (2003). Financial liberalization and its implication for the

domestic system: The case of Uganda (Research Paper No. 128). Nairobi: African

Economic Research Consortium.

Levin, A., Lin, C. F., & Chu, C. (2002). Unit root test in panel data: Asymptotic and finite sample

properties. Journal of Econometrics 108, 1–25.

Maddala, G. S., & Wu, S. (1999). A comparative study of unit root tests with panel data and a new

simple test. Oxford Bulletin of Economics and Statistics 61, 631–652.

Manamba, E. (2014). An empirical analysis of interest spreads in Tanzania. Asian Journal of Research

in Banking and Finance 4(10), 244-269.

Mannasoo, K. (2012). Determinants of bank interest spread in Estonia (Working Paper No. 1/2012).

Tallinn: Bank of Estonia.

Maudos, J., & Fernandez de Guevara, F. (2004). Factors explaining the interest margin in the banking

sectors of the European Union. Journal of Banking and Finance, 28(9), 2259–2281.

Maudos, J., & Solis, L. (2009). The determinants of net interest income in the Mexican banking

system: An integrated model. Journal of Banking and Finance, 33(10),1920-1931.

McKinnon, R. I. (1973). Money and Capital in Economic Development. Washington, US: Brookings

Institution Press.

McShane, R. W., & Sharpe, I. G. (1985). A time series/cross section analysis of the determinants of

Australian trading bank loan/deposit interest margins: 1962–1981. Journal of Banking

and Finance, 9, 115–136.

Mugume, A. (2010). Competition and Performance in Uganda’s Banking System (Research Paper

No. 203). Nairobi: African Economic Research Consortium.

Mugume, A., Apaa, J., & Ojwiya, C. (2009). Interest rate spreads in Uganda: Bank specific

characteristics or policy changes. The Bank of Uganda Staff Papers 3(2), 5-21.

Mujeri, M. K., & Younus, S. (2009). An analysis of interest rate spread in the banking sector in

Bangladesh. The Bangladesh Development Studies, XXXII, 4, 1-33.

Nampewo, D. (2013). What drives interest rate spreads in Uganda’s banking sector? International

Journal of Economics and Finance, 5(1), 76 – 85.

Nannyonjo, J. (2002). Financial sector reforms in Uganda (1990-2000): Interest rate spreads, market

structure, bank performance and monetary policy (Unpublished doctoral thesis). Gothenburg

University, Gothenburg, Sweden.

Ramfall, P. (2001). The determinants of interest rate spread: Empirical evidence on the Mauritian

banking sector, Bank of Mauritius, Port Louis.

Randall, R. (1998). Interest rate spreads in the Eastern Caribbean (Working Paper No. WP/98/59).

Washington DC: International Monetary Fund.

Rebei, N. (2014). Determinants of interest spreads in Solomon Islands (Working Paper No.

WP/14/105). Washington DC: International Monetary Fund.

Roodman, D. M. (2009). How to do xtabond2: An Introduction to Difference and System GMM in

Stata. Stata Journal, 9(1), 86-136

Saunders, A., & Schumacher, L. (2000). The determinants of bank interest rate margins: An

international study. Journal of International Money and Finance, 19(6), 813-832.

Shaw, E. S. (1973). Financial Deepening in Economic Development. New York, US: Oxford

University Press.

Siddiqui, M. A. (2012). Towards determination of interest spread of commercial banks: Empirical

evidences from Pakistan. African Journal of Business Management, 6 (5), 1851–1862.

Sologoub, D. (2006). The determinants of Bank Interest Margins and Profitability: The case of

developing countries (Working Paper No. WP/06/221). Washington DC: International

Monetary Fund.

Stiglitz, J., & Weiss, A. (1981). Credit rationing in markets with imperfect information. American

Economic Review, 71, 393–410.

Tarus, K. D., Chekol, B. Y., & Mutwol, M. (2012). Determinants of net interest margins of commercial

banks in Kenya: A panel Study. Procedia Economics and Finance, 2, 199-208.

Were, M., & Wambua, J. (2014). What factors drive interest rate spread of commercial banks?

Empirical Evidence from Kenya. Review of Development Finance, 4, 73–82.

Willmott, B. (2012). A decomposition of Ugandan lending rates (Working Paper No.

WP/BOU/2012/07). Kampala: Bank of Uganda Research Department.

Wooldridge, J. M. (2002). Econometric Analysis of Cross Section and Panel Data. Cambridge, MA:

MIT Press.

World Bank & IMF. (2005). Financial sector assessment: A handbook. Washington DC: The World

Bank and IMF.

World Bank (2017, August 2). Bank non-performing loans to total loans (%). Retrieved from

https://data.worldbank.org/indicator/FB.AST.NPER.ZS

APPENDICES

Appendix A: List of licensed commercial banks, credit institutions, and MDIs

Commercial banks (Tier 1) (As at March, 2016) 1. ABC Capital Bank Limited 14. Equity Bank Uganda Limited 2. Bank of Africa-Uganda Limited 15. Exim Bank (Uganda) Limited 3. Bank of Baroda (Uganda) Limited 16. Finance Trust Bank Limited 4. Bank of India (Uganda) Limited 17. Guaranty Trust Bank Uganda Limited 5. Barclays Bank of Uganda Limited 18. Housing Finance Bank Limited 6. Cairo International Bank Limited 19. KCB Bank Uganda Limited 7. Centenary Rural Development Bank Limited 20. NC Bank Uganda Limited 8. Citibank Uganda Limited 21. Orient Bank Limited 9. Commercial Bank of Africa (Uganda) limited 22. Stanbic Bank Uganda Limited 10. Crane Bank Limited 23. Standard Chartered Bank Uganda Limited 11. DFCU Bank Limited 24. Tropical Bank Limited 12. Diamond Trust Bank Uganda Limited 25. United Bank for Africa (Uganda) Limited 13. Ecobank Uganda Limited Note: Exim Bank (Uganda) Limited is not included in the panel though it existed during the time of the study. The bank only published its consolidated financial statements, including its performance in other countries

Credit institutions (Tier 2) (As at 30, June-2016)

1. Mercantile Credit Bank Limited 2. Opportunity Bank Uganda Limited 3. Post Bank Uganda Limited 4. Top Finance Bank Uganda Limited

MDIs (Tier 3) (As at 30, June-2016)

1. FINCA Uganda Limited 5.0.UGAFODE Microfinance Limited 2. Pride Microfinance Limited 6.0.EFC Uganda Limited 3. Yako MDI Limited

Appendix B: Financial inclusion indicators Average (2005- 2011 2012 2013 2014 2015 2016 2010) Capital adequacy (%) Regulatory capital to risk-weighted 19.56 20.32 21.89 22.88 22.22 20.97 19.83 assets Regulatory tier 1 capital to risk- 17.62 17.92 18.79 19.87 19.68 18.58 17.31 weighted assets NPLs less specific provisions to 3.86 3.79 6.91 8.61 6.25 9.31 13.03 capital Insider loans to total capital 2.81 4.54 3.35 4.49 3.06 3 5.09 Total capital to total assets 13.61 15.42 16.66 16.5 16.29 16.54 15.46 Leverage ratio 9.38 10.44 10.56 11.1 11 11.08 9.62 Asset quality (%) NPLs to total gross loans 2.98 2.21 4.23 5.63 4.13 5.29 10.47 NPLs to total deposits 1.89 1.73 3.15 4.05 2.95 3.87 7.41 Specific provisions to NPLs 60.14 50.95 45.87 47.16 48.86 41.64 60.35 Earning assets to total assets 78.06 74.01 71.29 69.71 71.52 69.18 67.34 Large exposures to gross loans 37.64 34.62 34.64 36.31 38.29 40.95 42.43 Large exposures to total capital 117.05 120.76 104.72 105.15 113.19 123.45 133.17 Earnings and profitability (%) Return on assets 3.28 4.03 3.9 2.53 2.63 2.62 1.33 Return on equity 24.83 27.41 24.24 15.24 16.05 15.96 8.33 Net interest margin 10.30 11.69 12.81 11.54 10.95 11.32 12.81 Yield on advances 17.15 18.94 19.84 17.3 16.54 17 16.74 Cost of deposits 2.75 3.15 4.15 3.7 3.55 3.34 3.48 Cost to income 70.62 68.16 70.93 77.16 68.71 69.25 67.17 Overhead to income 50.68 43.91 40.09 46.65 39.98 41.87 29.82 Liquidity (%) Short term gap -18.73 -20.88 -17.25 -17.22 -21.16 -16.9 -6.74 Liquid assets to total deposits 48.33 37.56 41.96 42.49 43.97 46.44 51.54 Liquid assets to total assets 32.45 25.76 28.37 28.22 29.68 31.69 35.32 Interbank borrowings to total 5.83 3.52 3.28 4.08 2.55 2.65 1.99 deposits Bank-funded advances to total 60.13 74.88 70.29 63.23 62.39 64.11 63.13 deposits Market sensitivity (%) Forex exposure to regulatory tier 1 -3.69 -3.62 -0.57 -3.02 -6.86 -5.93 8.52 capital Forex loans to forex deposits 55.69 62.81 79.26 62.22 64.54 59.16 62.49 Forex assets to forex liabilities 102.76 100.19 105.01 96.76 97.05 101.75 99.16 Source: Financial Stability Department, Bank of Uganda

Appendix C: Derivation of the first order conditions under the Ho and Saunders bank dealership model Taking into account equation (4.5), the expected utility of the bank is15:

1 퐸푈(푊) = 푈(푊) + 푈′(푊)퐸(퐿 푍 + 푀 푍 ) + 푈′′(푊)퐸(퐿 푍 + 푀 푍 )2 0 퐿 0 푀 2 0 퐿 0 푀

1 = 푈(푊) + 푈′′(푊)퐸(퐿2 휎2 + 푀2휎2 + 2퐿 푀 휎 ) (C.1) 2 0 퐿 0 푀 0 0 퐿푀

When a new deposit, D, is made, the banking firm has to pay 푟퐷퐷 and operating costs 퐶(퐷), and will obtain a return (푟 + 푍푀)퐷 in the money market. In this way, the bank’s final wealth will be:

푊푇 = (1 + 푟퐼 + 푍퐼)퐼0 − (1 + 푟푑)퐷 + (1 + 푟 + 푍푀)푀0 + (1 + 푟 + 푍푀)퐷 − 퐶(퐼0) − 퐶(퐷)

=(1 + 푟푊)푊0 + 퐿0푍퐿 + 푎퐷 + (푀0 + 퐷)푍푀 − 퐶(퐼0) − 퐶(퐷) (C.2) and the expected utility after the new deposit has been made is given by the following expression:

1 2 퐸푈(푊 ) = 푈(푊) + 푈′(푊)퐸(푊 − 푊) + 푈′′(푊)퐸(푊 − 푊) 푇 2

1 = 푈(푊) + 푈′(푊)[푎퐷 − 퐶(퐷)] + 푈′′(푊)[(푎퐷 − 퐶(퐷))2 + 퐿 휎2 + (푀 + 퐷)휎2 + 2퐿 (푀 + 2 0 퐿 0 푀 0 0

퐷)휎퐿푀] (C.3)

Given the level of wealth after the arrival of the new deposit, the increase in expected utility is as follows:

1 2 훥퐸푈(푊 ) = 퐸푈(푊 ) − 퐸푈(푊) = 푈′(푊)[푎퐷 − 퐶(퐷)] + 푈′′(푊) [(푎퐷 − 퐶(퐷)) + 퐷 푇 2

2 (퐷 + 2푀0)퐷휎푀 + 2퐿0퐷휎퐿푀] (C.4)

In the same way, if the bank grants a new credit for an amount L it will receive an income 푟퐿퐿 =

(푟 + 푏 + 푍퐿)퐿, and incur operating costs 퐶(퐿) and costs of financing the granting of credits (푟 + 푍푀)퐿

Analogously to the receiving of deposits, the increase of the bank’s expected utility due to the granting of an additional credit will be:

15 푊 = 퐸푈(푊) = 퐸(푊0(1 + 푟0)+퐼0푍퐼 + 푀0푍푀 − 퐶(퐼0) = 푊0(1 + 푟0) − 퐶(퐼0) 88

1 2 훥퐸푈(푊 ) = 퐸푈(푊 ) − 퐸푈(푊) = 푈′(푊)[푏퐿 − 퐶(퐿)] + 푈′′(푊) [(푏퐿 − 퐶(퐿)) + 푇 푇 2

2 2 (퐿 + 2퐿0)퐿휎퐿 + (퐿 − 2푀0)퐿휎푀 + 2(푀0 − 퐿0 − 퐿)퐿휎퐿푀] (C.5)

Bearing in mind the probabilities of granting credits or capturing deposits reflected in equation 4.8), the problem of maximisation of (4.9) can be written:

1 2 푀푎푥 퐸푈(훥푊) = (훼 − 훽 푎) {푈′(푊)[푎퐷 − 퐶(퐷)] + 푈′′(푊) [(푎퐷 − 퐶(퐷)) + 푎,푏 퐷 퐷 2

1 2 (퐷 + 2푀 )퐷휎2 + 2퐿 퐷휎 ]}+(훼 − 훽 푏) {푈′(푊)[푏퐿 − 퐶(퐿)] + 푈′′(푊) [(푏퐿 − 퐶(퐿)) + 0 푀 0 퐿푀 퐿 퐿 2

2 2 (퐿 + 2퐿0)퐿휎퐿 + (퐿 − 2푀0)퐿휎푀 + 2(푀0 − 퐿0 − 퐿)퐿휎퐿푀]} (C.6)

The first order conditions with respect to a and b give rise to the margins of expression (4.10).

Appendix D: Panel unit root tests-PP tests Variable Fisher type test-PP tests P Z L* Pm Interest rate spread 246.397*** −7.958 *** −13.248*** 20.893 *** Credit risk 97.037*** −3.498*** −4.104*** 5.321*** Liquidity risk 290.836*** −8.396*** −15.596*** 25.526*** Bank size 87.841*** −3.119*** −3.412*** 4.066*** Operating costs 191.623*** −4.962*** −9.236*** 14.659*** Return on assets 90.854*** −3.504*** −4.004*** 4.374*** Non-interest income 147.073*** −5.178*** −7.307*** 10.112*** Capital adequacy ratio 156.670*** −3.950*** −7.432*** 11.538*** HHI (Loan) 122.835*** −6.543*** −6.711*** 7.638*** HHI (Deposit) 137.555*** −5.590*** −7.221*** 9.140*** Foreign bank participation 99.902*** −5.037*** −4.941*** 5.297*** Inflation 220.723*** −7.692*** −12.515*** 17.629*** Real GDP growth 82.021*** −3.489*** −3.770*** 3.472*** T-bill rate 44.209*** −1.345*** −1.253*** −0.387*** Exchange rate volatility 74.183*** −3.645*** −3.476*** 2.672*** M2/GDP 220.021*** −10.125*** −12.292*** 17.557*** Note: P is the inverse chi-squared statistic; Z is the inverse normal statistic; L* is the inverse logit statistic; and Pm is the modified inverse chi-squared statistic. ***, ** and * indicate significance of the unit root statistics at 1 per cent, 5 per cent and 10 per cent significance levels respectively. Source: Author’s calculations