Commercial Bank Profitability in a Negative Interest Rate ...

i

Commercial Bank Profitability in a Negative Interest Rate Environment A study on the relationship between negative interest rates and

commercial bank profitability in Denmark

BACHELOR DEGREE PROJECT

THESIS WITHIN: Business Administration

NUMBER OF CREDITS: 15 ECTS PROGRAMME OF STUDY: International Management

AUTHOR: Albarbari, Mohammed Imad & Kipper, Lukas JÖNKÖPING May 2020

ii

Acknowledgements

We would like to take this opportunity to extend our sincere gratitude to everyone who

contributed to the making of this paper.

Our first acknowledgement is directed towards Oskar Eng, our brilliant tutor, who provided us

with great expertise, support, and words of advice in the process of writing the thesis.

Our second acknowledgement is directed towards Felix Stamm and Nicolas Rix for taking the

time to contribute with valuable feedback and constructive criticism.

We would further like to give a special thank you to our families and friends for all the support.

Mohammed Imad Albarbari

Lukas Kipper

iii

Bachelor Thesis in Business Administration

Title: Commercial Bank Profitability In A Negative Interest Rate Environment

Authors: Albarbari, Mohammed Imad and Kipper, Lukas

Tutor: Oskar Eng

Date: 2020-05-18

Key terms: “Bank profitability”, “Negative interest rates”, “Danish Commercial Banks”,

“Monetary Policy”, “Return on Average Assets”, “Net Interest Margin”.

Abstract

Background: Denmark, along with other European countries, has decided to cut its policy

interest rate into negative territory to meet macroeconomic objectives. This has historically

been thought of as impossible and impacts commercial banks significantly. As a consequence,

concerns have been raised about commercial bank profitability, which is a primary indicator of

the banking industry’s soundness.

Purpose: The purpose of this thesis is to investigate the relationship between persistently

negative interest rates and commercial bank profitability in Denmark, covering an extended

timeframe (2011 – 2018, 165 bank years, 21 commercial banks).

Method: Bank profitability is measured using the Return on Average Assets (ROAA) and the

Net Interest Margin (NIM). The thesis follows a simple form of mixed-methods approach –

quantitatively focused, followed by a supplementary qualitative study. For the quantitative part,

data is collected through the Orbis database, which provides global company data. We utilized

a Fixed Effects Model with strongly balanced panel data, covering 59% of the Danish banking

industry’s assets. Semi-structured interviews were then conducted with professionals working

in the industry to interpret the quantitative findings.

Conclusion: The findings of this study show that in the time period observed:

1. Interest rates are not correlated with the NIM;

2. The duration of consecutive negative interest rates (in years) is negatively correlated

with the NIM;

3. Interest rates are not correlated with the ROAA;

4. The duration of consecutive negative interest rates (in years) is not correlated with the

ROAA;

The duration of consecutive negative interest rates seems to be more significant since it takes

time for the profitability-reducing effect of negative interest rates to materialize. The ROAA is

not impacted by the (years in negative) interest rates, as it is mainly determined by factors under

management control.

iv

Table of Contents

Definitions ................................................................................................. viii

1 Introduction ................................................................................. 1

1.1 Background ......................................................................................................... 1

1.2 Problem ............................................................................................................... 2

1.3 Purpose & Research Questions ........................................................................... 3

2 Literature review ......................................................................... 4

2.1 Literature search .................................................................................................. 4

2.2 Frame of reference .............................................................................................. 4

2.2.1 The banking industry........................................................................................... 4

2.2.2 Commercial banks and investment banks ........................................................... 5

2.2.3 General structure of banking profitability studies ............................................... 5

2.2.4 Measures of commercial bank profitability ........................................................ 6

2.2.4.1 Overall Profitability: ROA and ROE .......................................................................................... 7

2.2.4.2 Net Interest Margin (NIM) .......................................................................................................... 8

2.2.5 Evolution of banking profitability determinants in the academic literature ....... 8

2.2.6 Bank profitability determinants........................................................................... 9

2.2.6.1 Internal determinants of bank profitability .................................................................................. 9

2.2.6.1.1 Size ......................................................................................................................................... 9

2.2.6.1.2 Capitalization ....................................................................................................................... 10

2.2.6.1.3 Credit risk ............................................................................................................................. 10

2.2.6.1.4 Liquidity ............................................................................................................................... 11

2.2.6.1.5 Diversification ...................................................................................................................... 11

2.2.6.1.6 Operational Efficiency ......................................................................................................... 12

2.2.6.2 External Determinates of bank profitability .............................................................................. 12

2.2.6.2.1 GDP Growth/Business Cycle ............................................................................................... 12

2.2.6.2.2 Market concentration ............................................................................................................ 12

2.2.6.2.3 Inflation ................................................................................................................................ 13

2.2.6.2.4 Interest Rates ........................................................................................................................ 13

2.2.6.2.5 Term Structure ..................................................................................................................... 14

2.3 Hypothesis development & independent variables ........................................... 16

2.3.1 The profitability of core banking operations ..................................................... 16

2.3.2 Overall profitability ........................................................................................... 17

3 Data, Model Specification and Methodology .......................... 19

v

3.1 Research philosophy ......................................................................................... 19

3.2 Research design ................................................................................................. 20

3.3 Mixed methods .................................................................................................. 21

3.4 Data collection .................................................................................................. 22

3.4.1 Process and sampling for the quantitative data ................................................. 22

3.4.2 Process and sampling for the supplementary (qualitative) data ........................ 24

3.5 Variable selection .............................................................................................. 25

3.5.1 Dependent variables .......................................................................................... 25

3.5.2 Independent variables........................................................................................ 26

3.5.2.1 InterestRate ............................................................................................................................... 27

3.5.2.2 ThreeM_IBOR .......................................................................................................................... 27

3.5.2.3 ThreeM_Yield ........................................................................................................................... 28

3.5.2.4 YearsInNegInterestRate/YNeg_IBOR ...................................................................................... 29

3.5.3 Control variables ............................................................................................... 29

3.6 Possible outcomes for the independent variables.............................................. 31

3.7 Method of quantitative data analysis ................................................................ 32

3.7.1 Motivating the statistical model ........................................................................ 32

3.7.2 Hausman test ..................................................................................................... 33

3.7.3 Employed model ............................................................................................... 33

3.8 Method of supplementary (qualitative) data analysis ....................................... 34

3.9 Ethical considerations ....................................................................................... 35

4 Results & Analysis ..................................................................... 36

4.1 Quantitative results............................................................................................ 36

4.2 Results overview for NIM regressions .............................................................. 36

4.3 Results overview for ROAA regressions .......................................................... 37

4.4 Robustness tests ................................................................................................ 37

4.5 Analysis of the results ....................................................................................... 40

4.5.1 Interest rates and the NIM ................................................................................. 40

4.5.2 Other factors influencing the NIM .................................................................... 44

4.5.3 Interest rates and the ROAA ............................................................................. 45

4.5.4 Other factors influencing the ROAA ................................................................ 46

5 Conclusion .................................................................................. 49

vi

6 Discussion ................................................................................... 50

6.1 Implications for bank management ................................................................... 50

6.2 Limitations and further research ....................................................................... 51

7 Reference list .............................................................................. 53

8 Appendices ................................................................................. 65

8.1 Definitions ......................................................................................................... 65

8.2 Calculation of sample coverage ........................................................................ 66

8.3 Description of supplementary sampling procedure........................................... 66

8.4 Appendix interview questions ........................................................................... 67

8.5 Appendix statistical models .............................................................................. 68

8.5.1 Generalized Method of Moments Estimators (GMM) ...................................... 68

8.5.2 Population-Averaged Ordinary Least Squares models (PA-OLS).................... 68

8.6 Hausman tests ................................................................................................... 69

8.7 Multicollinearity tests........................................................................................ 71

8.8 Descriptive statistics.......................................................................................... 75

8.9 Original stata output .......................................................................................... 78

vii

Figures

Figure 1: Denmark’s certificate of deposit rate over time. ........................................ 1

Figure 2: The nine mixed methods designs. .............................................................. 22

Figure 3: Denmark's certificate of deposit rate over time. ....................................... 23

Figure 4: Visualization of interest rates over time. .................................................. 28

Figure 5: Possible outcomes for profitability of core banking operations. .............. 31

Figure 6: Possible outcomes for overall bank profitability. ..................................... 31

Figure 7: Development of the NIM in a prolonged negative rate environment. ...... 43

Figure 8: NIM over time. ......................................................................................... 44

Figure 9: Credit Risk over time ................................................................................ 47

Figure 10: Liquidity over time. ................................................................................ 47

Figure 11: Efficiency over time. .............................................................................. 48

Figure 12: Diversification over time ........................................................................ 51

Tables

Table 1: A summary of variables commonly affecting bank profitability. ............... 15

Table 2: Dependent & Independent Variables and corresponding hypotheses ......... 18

Table 3: Dependent variables and hypotheses........................................................... 26

Table 4: The six models and their corresponding interest rate proxy and profitability

metrics. ......................................................................................................... 27

Table 5: A summary of the variables selected in the thesis. ..................................... 30

Table 6: Model Specifications. .................................................................................. 34

Table 7: Regression Results ...................................................................................... 39

Table 8: Potential effects of prolonged exposure to negative rates ........................... 43

viii

Definitions

___________________________________________________________________________

This section is a supplementary to the main section of the thesis and is taken from Appendix

8.1. We recommend having it ready while reading the thesis since it gives an overview of the

most important concepts defined throughout the paper.

Denmark’s Policy

Interest Rate

The interest rate banks receive on deposits and pay for borrowing at the

central bank of Denmark (Nationalbanken). Denmark’s central bank can

set this rate higher or lower to meet its goals. This rate moves the interest

paid and received by Commercial Banks. A lower policy interest rate

will usually result in cheaper loans (e.g. mortgages) and lower interest

rates on saving accounts. The policy interest rate is a powerful tool to

stimulate (or cool down) a country’s economy.

Core Banking

Operations

In the context of this thesis, core banking operations are defined as

accepting deposits (on which interest is traditionally paid) and lending

out a multiple of these deposits for receiving interest.

Net Interest

Margin (NIM)

Usually, the interest commercial banks pay on deposits is lower than the

interest they receive on loans. The difference between interest paid and

interest received is the Net Interest Margin. In other words, it is the

profit margin of a commercial banks’ Core Banking Operations.

Commercial

Banks

”A bank with branches in many different places that offers services to

people and businesses, for example, keeping money in accounts and

lending money” (Cambridge University Press, 2020b)

Put another way, commercial banks generally focus on Core Banking

Operations. However, commercial banks may also engage in other

business activities, like financial guarantees and derivative sales/trading

(A. N. Berger & Bouwman, 2015). Commercial banks should be

differentiated from investment banks, which generally focus on helping

companies to generate funding (e.g. issuing stocks and bonds) and

company mergers & acquisitions (Stowell, 2010).

”Passing on

Negative Rates”

Means commercial banks pass on the negative policy rate to their

customers to retain profitability. One would then likely have to pay to

deposit money at a bank.

Proxy Variable „A variable used instead of the variable of interest when that variable

of interest cannot be measured directly“ (Oxford Reference, 2020).

1

1 Introduction

______________________________________________________________________

In this chapter, the reader is firstly introduced to the topic of negative interest rates and its

possible effects on bank profitability. The importance of investigating this topic from a business

administration perspective is also highlighted. Secondly, a gap in the literature is identified

regarding the factors affecting bank profitability. Lastly, the purpose and research questions of

this paper are presented.

1.1 Background

The last decade has spawned a unique climate for European banks. The shifts in monetary policy

we have seen from various central banks can be described as unusual at best and as extreme at

worst. The central banks of countries such as Denmark, Sweden, Switzerland and even the

European Central Bank all decided to cut their interest rates to near-zero levels and, eventually,

into the negative territory (see Figure 1 for Denmark). Reasons for this measure include

meeting growth, inflation or exchange rate targets.

The policy interest rate is a part of a central bank’s monetary policy tools. It represents the cost

of borrowing as a proportion of the amount of money that is lent, deposited, or borrowed (Faure,

2014).

Simplified, a negative interests rate means, that if an individual or institution borrows money,

they pay back less than what they borrowed. Negative policy interest rates have been thought

of as impossible by researchers for a long time – economists believed in the existence of the so-

called ‘zero lower bound’ which refers to the lowest possible level (zero) that interest rate can

drop to (Buiter, 2009).

A group of firms primarily affected by central bank policy is commercial banks. They make

most of their money by lending out money at a higher rate than they pay on deposits. Due to

Figure 1: Denmark’s certificate of deposit rate over time (Nationalbanken, 2020).

2

the banks’ potential unwillingness to pass on negative rates, such monetary policy has caused

concerns about its impact on commercial bank profitability, which is a primary indicator of the

banking industry’s soundness (Bikker & Vervliet, 2018). Negative rates are unseen in history,

and since they break through the zero-lower bound, the impacts are unknown.

Currently, it seems like the negative interest rate climate is rather permanent, making the

concerns outlined even more valid. Due to the Coronavirus pandemic, many experts now further

believe we are headed for a strong recession (Goodman, 2020). The usual monetary policy move

in a recession is to lower interest rates to stimulate the economy. This is already happening in

part, as the United States Federal Reserve bank dropped their policy rate (”Target federal funds

rate”) to 0.00 on the 16th of March 2020 (Board of Governors of the Federal Reserve System,

2020). Currently (May 2020), futures contracts even seem to price in negative US policy rates

(Bolingbroke, 2020). The current negative economic outlook makes a long-term, global

negative interest rate environment therefore more likely than ever. Should that be the case,

decision makers in banks need to understand what their situation and future look like. If, for

example, continued exposure to negative interest rates were to erode core banking operations’

profitability, it is important for leaders of the financial industry to start diversifying their

business into other areas or even fundamentally change their business model.

Studying the banking industry has tremendous implications for both the economic- as well as

the business administration- field of research. In economics, a profitable banking industry has

the ability to resist negative shocks and shows stability of the financial system (Athanasoglou

et al., 2008). In the business administration field, profitability is a must for banks to survive and

grow as businesses, especially since stakeholders generally view bank performance as profits

made despite risks taken (Bikker, 2010). Studying bank profitability has therefore attracted the

attention of company management, policymakers and researchers. It has also captured our

interests, and hence, we decided to examine the topic of negative interest rates in relation to the

Danish banking industry. No other country has had negative interest rates for such a long time.

Furthermore, the effects of the low interest rates on the banks’ business seem to materialize

substantially in Denmark, more than anywhere else in the world. For example, if one deposits

money with the Danish Jyske Bank, one now has to pay interest for providing the bank funding.

This is highly unusual, since banks are usually reluctant to lower deposit rates below zero due

to the potential loss of customers (Claessens et al., 2018). According to the Financial Times,

Jyske Bank is even offering the first negative-interest mortgage since 2015 – which essentially

means that one gets paid for borrowing money (Milne, 2019).

1.2 Problem

There are very few studies that specifically investigate the relationship between interest rates

and banking profitability, making it an under-researched area (Borio et al., 2017). There exist

even fewer studies that examine the effect of persistently low (yet does not mean negative)

interest rates on bank profitability – a notable example being Claessens et al. (2018), who

suggests, that a persistently low interest rate environment continually reduces profitability.

However, we could not find a single peer-reviewed study examining the effects of a persistently

3

negative interest rate environment, as currently seen in Denmark. Indeed, negative interest rates

were only introduced in the past decade, yet several years have passed without comprehensive

academic research conducted to investigate the topic. Some institutions such as the International

Monetary Fund (Jobst et al., 2016) and the European Central Bank (Altavilla et al., 2017) have

published working papers to debate over the subject. Still, these are not peer-reviewed articles,

and one can argue their objectivity since central banks and international institutions usually

have their own agenda. Therefore, there is a gap in the literature that requires further

investigation, and the aforementioned working papers confirm the actuality and relevancy of

the topic. The lack of knowledge in this area may give rise to business and, potentially, survival

problems for banks in the future, if the negative interest rate environment keeps on for a

prolonged period of time as it is currently expected. Consequently, clarifying the neglect in the

literature of this bank profitability is imperative for a functioning industry.

1.3 Purpose & Research Questions

To address the gap identified within the literature, this paper aims to investigate the relationship

between persistently negative interest rates and bank profitability in the Danish banking

industry. The study will examine an extended timeframe that represents the ultra-low interest

rate environment of Denmark (2011-2018, 21 banks, 165 bank years). Exploring the tension

between these two variables is vital since interest rates have been historically low in the past

decade, and the negative territory could become the norm in future.

The above discussion leads to the following two main questions:

• Do interest rates affect Danish commercial bank profitability in a (near-) negative

interest rate environment? If yes: Is the effect positive or negative and how strong is it?

• How does it affect Danish commercial bank profitability when interest rates stay in the

negative territory for a prolonged period of time?

The remainder of this paper is structured as follows: Section 2 provides an overview of the

literature on the determinants of bank profitability. Section 3 describes the method and presents

the data of the Danish banks. Section 4 reports the empirical results and analyzes their

implications. Finally, Section 5 concludes the paper and Section 6 discusses additional findings

and suggests further research.

4

2 Literature review

______________________________________________________________________

The purpose of this chapter is to provide a theoretical base and gather previous research on

banking profitability. The chapter begins with an overview of the banking industry and explains

different types of banks. Secondly, several banking profitability metrics will be discussed and

compared. Thirdly, factors determining bank profitability will be categorized and compared.

Finally, and based on the previous parts, the hypotheses tested in the paper will be developed.

2.1 Literature search

To obtain a comprehensive overview of prior research on the determinants of bank profitability,

the procedure for a systematic literature search provided by Collis & Hussey (2014) was

followed. The purpose was to create a conceptual framework and identify a particular research

gap accordingly. Google Scholar, Web of Science, and Primo Search were initially used to

search for journals and articles of highest relevance. A combination of keywords was searched

for before being able to identify the most related keywords such as: macroeconomic*, interest

rate*, financial performance of bank*, and bank profitability. Because this combination of

keywords provided us with hundreds of articles, we decided to narrow down our topic towards

interest rates only and their link to banking profitability. Using Primo Search and taking

advantage of its advanced search option, we specified the terms as follows:

Title – Contains – Interest rate*

AND – Title – Contains – Bank* profit*

OR – Title – Contains – Financial performance bank*

Only peer-reviewed articles were selected, and no specific time-period was taken into

consideration. The motivation of this was to get a comprehensive understanding of the subject

from credible and trustworthy sources. However, because financial institutions engage in

research to keep their knowledge updated as well as forecast future business conditions, they

often publish working papers about current topics. We found that some of these working papers

are useful and, hence, are used (with caution) in our literature review.

2.2 Frame of reference

2.2.1 The banking industry

Banks are financial institutions that provide banking and other financial services. They have an

intermediation function based on raising funds, either from private depositors or wholesale

funding sources, to provide loans to borrowers or to finance other investments (Idiab et al.,

2011). One of the most prominent theories explaining how banks work is the Financial

Intermediation Theory (Werner, 2016). In imperfect markets, savers and investors are unable

to trade directly with each other in an optimal way, mainly because of informational

asymmetries. Therefore, financial intermediaries, specifically banks, act as agents and delegate

5

monitors to identify investors of behalf of savers because of their comparative informational

advantage over the other parties. Consequently, the theory suggests that financial intermediaries

are there to reduce transactions costs and information asymmetries (Scholtens & van Wensveen,

2003) by collecting deposits and then lending these out (Werner, 2016). They also have other

functions such as “bridging the maturity mismatch between savers and investors and facilitate

payments between economic parties by providing a payment, settlement and clearing system”.

Banks come in several different forms, depending on their activities. The most relevant types

are the following two:

2.2.2 Commercial banks and investment banks

Commercial banks typically act as a classical intermediary institution that accepts deposits from

savers and extends credit to borrowers. They have three primary functions: 1) holding time

deposits (saving accounts), 2) holding demand deposits (checking accounts), and 3) issuing

credits mainly to corporates, but also to individuals (Mizruchi, 2001). The importance of

commercial banks, through their intermediation role, lies with benefiting “the financial markets

by reducing transaction costs, spreading risk and realizing economies of scale and

specialization” (Clews, 2016). Commercial banks are highly regulated because savers, when

depositing money, place significant trust in these commercial banks. Failure in managing credit

exposure and ensuring the safety of the depositors’ funds represents a real risk (A. N. Berger &

Bouwman, 2015; Clews, 2016).

Investments banks, on the other hand, focus mainly on corporate-, investment- and government-

related clients. Typical investment banking services are security underwriting (e.g. Initial Public

Offerings, Debt Issuances) and strategic advisory services (Stowell, 2010). Therefore, the most

important differentiation between commercial and investment banks is that commercial banks

mainly focus on lending activities, which is peripheral for investment banks.

Given these widespread activities as well as the multiple stakeholders involved – such as

individuals, investors, economists, policymakers, managers, employees – the significance of

bank performance becomes visible and tangible. Bank performance can be measured in various

ways. Examples include efficiency, reliability (Bikker, 2010), cost structure, size and loan

portfolio composition (Arshadi & Lawrence, 1987). Generally however, stakeholders perceive

performance in terms of profitability regardless of risk taken (Bikker, 2010), which makes this

one of the most important performance measures.

2.2.3 General structure of banking profitability studies

When assessing the profitability of a banking industry, there is a quite specific structure most

studies follow. Generally, data points of several banks are gathered over a multi-year period.

These data points are then put into a model, which can be (very generally) described as follows:

6

The dependent variable is therefore a profitability metric, which is determined by several

factors. β is the coefficient for each factor (the amount by which a one-unit increase in the factor

influences the profitability metric) and 𝜀 is the error term.

For now, the following section will describe the left side of this equation (most relevant

profitability metrics) and the section after that will describe the right side (most relevant factors

influencing bank profitability). Near the end of the literature review, a summary table will give

a concluding overview over which studies found which factors to be influencing which

profitability metrics. All the studies referenced in the table use the abovementioned structure.

2.2.4 Measures of commercial bank profitability

There are two main issues with widespread business profitability measures, like net income and

EBITDA. Firstly, they are absolute and do not take into account bank size, asset base or

deposits. Secondly, they are inflation variant, meaning the same amount of net income has a

different real value in 2011 than it has in 2018. Both attributes reduce comparability and skew

results.

These considerations caused researchers examining bank profitability to use inflation invariant

and size-dependent metrics – which can further be grouped into two categories: overall

profitability metrics and margins on core banking operations.

Overall profitability measures how profitable a bank is as a whole. This category of metrics

considers income from core banking operations and other diversified sources of income (like

trading, fee-based services, etc.). Overall profitability is usually measured by Return on Assets

(ROA) or Return on Equity (ROE). The second category of profitability metrics is the margin

on core banking operations - excluding all other sources of income. This metric is usually

measured through the Net Interest Margin (NIM).

The literature distinguishes between these categories for two main reasons. Firstly, a bank can

have a slim NIM, yet be highly profitable overall (e.g. when a bank shifts away from possibly

unprofitable core banking operations and diversifies into more profitable fields) or vice versa.

Furthermore, a certain factor (e.g. GDP growth) can influence NIM and overall profitability

differently. It is therefore common practice to observe the effect of the factor of interest (like

GDP growth) on both NIM and overall profitability and then compare the results (e.g. Dietrich

& Wanzenried, 2011).

7

2.2.4.1 Overall Profitability: ROA and ROE

The Return on Assets (ROA), expressed in percent, generally describes how much net income

a company generates for each Danish Crown (DKK) of assets a company had in a specific year.

Return on Equity, also expressed in percent, describes how much net income a company

generates for each DKK of shareholders’ equity a company had in a specific year. Both are

standard metrics to measure overall business profitability and are used in the corporate finance

literature. There has been some discussion on which metric reflects the overall bank profitability

more accurately:

Athanasoglou et al., (2008) state, that ROA should be preferred over ROE since it considers the

risk taken on through leverage.1 They argue that the risk taken on through leverage is one of the

central considerations of the core banking business; therefore, the overall profitability metric

should reflect this factor. Several scholars argue that ROA is further favourable since it

describes the ability of bank management to generate profit based on a bank’s asset base and

cannot be influenced by high equity stakes (Menicucci & Paolucci, 2016; Rivard & Thomas,

1997). This is relevant since ROE can be influenced by shareholders increasing/decreasing their

investment in a bank - such fluctuations could skew ROE. These considerations made ROA

widely recognised as the key overall profitability indicator for banks (Athanasoglou et al., 2008;

Dietrich & Wanzenried, 2011; Golin, 2001).

Although Return on Assets (ROA) and Return on Average Assets (ROAA) are used almost

interchangeably in the banking profitability literature, they are not the same. ROA measures the

company’s assets year-end, while ROAA takes the average of a company’s asset base in a

specific year. Taking into consideration the fluctuations in asset balances throughout the year

is generally seen as more accurate than measuring assets year-end (Ayadi & Boujelbene, 2012;

Dietrich & Wanzenried, 2011). Jewell & Mankin (2011) further write, that ROAA has two main

1 Definition of leverage: “The relationship between the amount of money that a company or

organization owes and the value of the company or organization“ (Cambridge University Press,

2020c). In the case of commercial banking a high leverage means, that a bank gives out a large

sum of loans for each DKK in deposits they hold – therefore increasing the banks‘ risk.

8

benefits over ROA: Firstly, ROAA preserves the basic matching principle of accounting.

Secondly, ROAA is less affected by random changes in total assets.

2.2.4.2 Net Interest Margin (NIM)

Traditionally, banks make money by lending out money at a higher interest rate than the interest

rate they pay a depositor (which we refer to as the “core banking operations”). The net interest

margin (NIM) describes the difference between the interest income of loans and the interest

paid to depositors, divided by a bank’s asset base. It is a metric specific to banks and measures

the profitability of core banking operations. Generally, the interest customers pay on loans and

receive on deposits is strongly influenced by the policy interest rate. The policy interest rate

dictates, for which interest rate the bank can make short-term loans from the central bank. This

rate then influences the rate for which the bank’s customers can borrow and deposit. Simplified,

the interest rate a customer receives on deposits will be somewhat lower than the policy interest

rate. In contrast, the interest paid on loans will be somewhat higher than the policy interest rate

– the policy rates are therefore generally passed on to customers. The spread between these two

rates is the profit of the bank (Choudhry, 2017).

The above section describes the essential metrics one needs to understand how banking

profitability is determined. Now a focus will be on the variables that determine the banking

profitability metrics according to the literature (the right side of the equation introduced under

2.2.3 General Structure of banking profitability studies).

2.2.5 Evolution of banking profitability determinants in the academic literature

Early researchers started by grouping possible determinants into internal and external factors

(Short, 1979). As the name implies, internal determinants are micro or bank-specific factors

that are associated with bank management. External determinants are industry-specific and

macroeconomics factors that can affect the performance of financial institutions, even though

they are not related to bank management. In his paper, Short (1979) considered banks’ profit

rates as the appropriate measure of their performance. After Short’s pioneering study, Bourke

(1989) conducted one of the most impactful studies in the research field. He also differentiates

between internal and external determinants of overall bank profitability and settles for several

independent variables, including capitalization and market concentration.

While the determinants have been widely discussed, Bourke’s approach of comparing a large

sample of banks over an extended timeframe using both internal and external determinants is

one of the most critical research pieces. His methodology has been replicated multiple times

(e.g. Molyneux & Thornton, 1992). The general idea of explaining banking profitability in such

a way has been applied under various circumstances, for example spanning different business

cycles (Dovern et al., 2010) or examining the 2008 crisis (Athanasoglou et al., 2008).

9

2.2.6 Bank profitability determinants

The interest in identifying the determinants of commercial bank profitability has led researchers

to test an abundance of possible variables. Still, there are only a few variables that can be

reasonably expected to influence bank profitability. Since there is no standard of which

variables to include in a banking profitability model, it is up to every researcher to examine

which variables could impact profitability. To analyze which variables impact profitability, a

table was created with possible determinants on the left and its effect on bank profitability in

the middle. Studies supporting this finding are found on the right side. Determinants that have

weak empirical support were omitted from the table. A summary of this analysis can be found

in Table 1. Based on the overarching theme in the literature, we will now discuss the most

important profitability determinants, grouped into internal- and external factors.2

2.2.6.1 Internal determinants of bank profitability

Studies that focus on internal determinants examine variables such as bank size, capital and

liquidity, credit risk and business diversification (see e.g. Alper & Anbar, 2011; Athanasoglou

et al., 2008; Dietrich & Wanzenried, 2011). This variable group usually determines overall

profitability (ROA, ROAA, ROE, ROAE) more than NIM. These same variables will also be

used as control variables in our study.

2.2.6.1.1 Size

Bank size is used to account for the potential of economies, or diseconomies, of scale in the

banking sector, and is usually measured through the logarithm of total assets (see e.g. Al-Jafari

& Alchami, 2014; Athanasoglou et al., 2008; Tan & Floros, 2012).

Boyd & Runkle (1993) linked the size of banks to the concept of economies of scales in what

they called the modern intermediation theory. Briefly, the theory suggests that large banks, or

financial intermediaries, can benefit from their ability to access a high number of lenders and

borrowers. This usually translates into more diversification – which in turn likely leads to a

reduction of contracting costs as well as to a reduction of risks. As a result, the modern

intermediation theory predicts that large banks are more cost-efficient and less likely to fail than

small banks. This might make large banks more profitable. The positive link between size and

profitability has been generally confirmed by the literature (see e.g. Demirguc-Kunt and

Huizinga, 1998; Alper & Anbar, 2011; Borio et al., 2017; Pervan et al., 2015; Tan & Floros,

2012), with some studies finding an insignificant relationship (Nessibi, 2016). However, the

2 It should be noted, that some studies find a certain variable to be significant, while other ones

do not. The conflicting results between different studies on the same variables is expected as

studies differ in time period and countries examined, data used, statistical methods chosen, and

combination of control variables used.

10

heightened profitability of large banks seems to be eroded during financial crises and times of

economic uncertainty (Campmas, 2020; Dietrich & Wanzenried, 2011).

2.2.6.1.2 Capitalization

Bank capital is defined as “the funds – traditionally a mix of equity and debt – that banks have

to hold in reserve to support their business” (Jenkins, 2010). The variable is measured as the

ratio of equity over assets (Claessens et al., 2018; Demirguc-Kunt & Huizinga, 1999).

Research suggests that higher capitalization is associated with:

• Higher overall profitability (Abreu & Mendes, 2001; Athanasoglou et al., 2008; A. N.

Berger & Bouwman, 2013; Claessens et al., 2018; Demirguc-Kunt & Huizinga, 1999;

Nessibi, 2016)

• Higher NIM (Claessens et al., 2018; Demirguc-Kunt & Huizinga, 1999)

• Enhanced performance through higher survival probabilities and higher market shares

(A. N. Berger & Bouwman, 2013)

The positive correlation between profitability and capitalization exists because better-

capitalized banks will:

• Get access to less risky, and thus cheaper, funding (Bourke, 1989)

• Face lower expected bankruptcy cost, leading to lower funding cost and higher interest

margins (Abreu & Mendes, 2001)

• Have a safety net in case of adverse developments (Bikker & Vervliet, 2018)

A notable exception is the study of Alper & Anbar (2011), who found that capitalization had

no impact on bank profitability in Turkey.

2.2.6.1.3 Credit risk

Risk is defined as “an uncertain but possible event which can cause some losses. It corresponds

only to negative deviations from the expected outcome, a positive one would be considered as

an opportunity” (Campmas, 2020). There are several types of risks that banks are subject to and

need to take into consideration, e.g. market risk, operational risk and solvency risk. However,

credit risk is considered to be the most important one (Gilchrist & Mojon, 2018).

Credit risk is usually measured through the ratio of loan loss provisions over gross loans; in

other words: the expenses set aside for defaulting loans per total loans. A higher ratio therefore

equals a lower credit risk. According to Campmas (2020), increases in relative Loan Loss

Provisions (a decrease in credit risk) decreases overall profitability because these expected loan

losses are deducted from the banks’ income. By contrast, Dietrich and Wanzenried (2011)

reported that Loan Loss Provisions Per Gross Loans does not have a statistically significant

effect on overall bank profitability during normal economic conditions. The determinant

11

seemed to increase in significance during the years of the financial crisis, and as a result, hurt

profitability.

Finally, it should be noted that the effect of the ratio on the NIM is not entirely clear: Dietrich

& Wanzenried (2011) find a positive correlation, while Campmas (2020) finds a negative

correlation.

2.2.6.1.4 Liquidity

The liquidity of a bank refers to the banks’ ability to meet obligations (lending and investment

commitments or deposit withdrawals) as they fall due (Lartey et al., 2013). Liquidity can be

measured by dividing net loans by total assets, which indicates the amount of the bank’s total

assets that are tied up in loans (Munteanu, 2012). The higher the ratio, the higher the liquidity

of the bank.

There is a conflict between the researchers on whether liquidity positively or negatively affects

banking profitability. Bourke (1989), as well as Duraj and Moci (2015), argue that there is a

positive relationship between liquidity and bank profitability as banks, by holding more liquid

assets, increase their ability “to absorb any possible unforeseen shock”. On the other hand,

Molyneux and Thornton (1992), as well as Goddard et al. (2004), found a negative relationship

between the two variables as these researchers believed that liquidity, usually, represents an

expense to the bank. This insight was predicted by ‘conventional wisdom’ of Bourke (1989)

who, however, failed to provide evidence of the negative relationship in his research.

2.2.6.1.5 Diversification

Diversification is the extent to which a bank diversifies its income streams away from core

banking operations into other business areas. It is therefore measured as (non) interest income

per operating income (e.g. in Campmas, 2020; Dietrich & Wanzenried, 2011).

The previously explained modern intermediation theory proposed by Boyd & Runkle (1993)

shows that higher diversification can lead to a reduction of risks taken and a lower probability

of failure. Additionally, higher diversification tends to have positive effects on business stability

as it improves the allocation of resources through internal capital markets3 (Stein, 1997).

Furthermore, in their working paper about the anatomy of bank diversification, Elsas et al.

(2006) used panel data from nine countries over seven years to examine how revenue

diversification affects bank value. The researchers found that revenue diversification increases

bank value by enhancing bank profitability. The latter is enhanced by 1) higher margins from

non-interest businesses, and 2) lower cost-to-income ratios. On the same line, Dietrich &

Wanzenried (2011) explained that larger banks are more profitable due to their ability to enjoy

higher product and loans diversification possibilities. The insights from these researchers show

3 I.e. markets where corporate headquarters allocate capital to their business units, as opposite

to external markets such as banks, finance companies, and stock market (Gertner et al., 1994)

12

that diversification is essential for banks which aims to reduce their risks and enhance both their

profitability and value.

When it comes to net interest margin, Campmas (2020) argues that diversification lowers NIM

since it leads to a higher relative share of non-interest income.

2.2.6.1.6 Operational Efficiency

Operational efficiency, measured as cost-to-income ratio, is defined as “the operating costs

(such as the administrative costs, staff salaries, and property costs, excluding losses due to bad

and non-performing loans) over total generated revenues4” (Dietrich & Wanzenried, 2011). The

ratio focuses on non-interest costs because those are considered to be mostly influenced by bank

management. The ratio does not include the bad and non-performing loans as they reflect the

quality of a bank’s credit portfolio rather than its performance (Tripe, 1998). The empirical

studies observing a positive relationship between efficiency and overall bank profitability are

plenty (Athanasoglou et al., 2008; Dietrich & Wanzenried, 2011; García-Herrero et al., 2009;

Tan & Floros, 2012), while some also find a positive correlation with the NIM (García-Herrero

et al., 2009).

2.2.6.2 External Determinates of bank profitability

Since banks are heavily dependent on the economic climate they operate in, several authors

decided to focus on important macroeconomic variables in their research. We will give an

outline of the most significant external determinants of bank profitability found in the literature.

2.2.6.2.1 GDP Growth/Business Cycle

Dietrich and Wanzenried (2011) Kanas et al. (2012), Trujillo‐Ponce (2013) and many others

measured the effect of the business cycle on bank profitability. The empirical results of their

research suggested that bank profitability and the business cycle, approximated by real GDP

growth, seems to have a positive relationship, and they concluded that profitability is pro-

cyclical. One explanation of the profit pro-cyclical feature is that, during cyclical upswings,

demand for lending increases which leads up to a more profitable business (Dietrich &

Wanzenried, 2011).

2.2.6.2.2 Market concentration

Market concentration, or market structure, is measured by the Herfindahl–Hirschman-Index, or

HHI. This index is defined as “the sum of the squares of the market shares of all the banks

within the industry, where the market shares are expressed as fractions” (Dietrich &

Wanzenried, 2011). The effect of market concentration on bank profitability is unclear for two

reasons. One the one hand, higher market concentration lowers competition and allows banks

to charge higher (lower) interest rates on loans (deposits). Such collusion would lead to a

4 In this paper (along with many others) operating income is used instead of revenue.

13

positive relationship between market concentration and profitability. One the other hand,

tougher competition might be expected to be the reason behind a higher bank concentration,

suggesting a negative relationship between the two variables (Campmas, 2020; Dietrich &

Wanzenried, 2011b; Goddard et al., 2004).

2.2.6.2.3 Inflation

Inflation is another critical determinant of bank profitability. Demirguc-Kunt and Huizinga

(1999) findings about the positive relationship between inflation and bank profitability seem to

be valid in many countries such as China (Tan & Floros, 2012a), the United States (Kanas et

al., 2012) and Spain (Trujillo‐Ponce, 2013). Other researchers, on the contrary, found a negative

relationship in a 47-country study (Claessens et al., 2018).

2.2.6.2.4 Interest Rates

Interest rates usually refer to interest rates the banks are subjected to. This can be measured in

several ways. The most common one is the policy interest rate set by the central bank of the

respective country (e.g. in Campmas, 2020; Short, 1979). Other measurements include the

Interbank Overnight Rate (e.g. in Borio et al., 2017) and the yield on short-term sovereign debt

(e.g. in Claessens et al., 2018). For a detailed discussion of these proxies, please refer to the

3.5.2 Independent variables.

The link between banking profitability and interest rates is sometimes seen as a “by-product”

in the literature and not the specific focus of research (Borio et al., 2017). The studies that use

interest rates as a profitability predictor have conflicting results. Some find a positive

relationship (Europe: Molyneux & Thornton, 1992; International: Borio et al., 2017; Campmas,

2020; Demirguc-Kunt & Huizinga, 1999), while some find that lower interest rates do not lower

overall bank profitability (Altavilla et al., 2017; Claessens et al., 2018).

The abovementioned conflict is especially interesting since most available research focuses on

the link between interest rates and profitability in what might be considered a (relatively) high-

interest rate environment today. A negative interest rate environment can have unique effects

on banking profitability since some research suggests that the interest margin compressing

effect of low interest rates gets a lot stronger the lower the interest rate (Borio et al., 2017).

Additionally, Claessens et al. (2018) show in a multi-country study carried out between 2005

and 2013, that not only do low interest rates depress overall profitability, but the effect gets

stronger over time.

Because negative interest rates only arrived in Europe the past decade, it is still unclear whether

this likely strong compression of net interest margins can be offset through other

activities/factors. A recent working paper by the European Central Bank claims that the

potentially NIM compressing effect of low interest rates is offset by an increase in credit quality,

amongst others (Altavilla et al., 2017). A recent study by Campmas (2020) supports this, as she

finds that the impact of interest rates is more substantial on the NIM than on overall profitability.

However, the study’s panel only lasts until 2015, which is relatively short for investigating

14

negative interest rates, considering they have only been introduced in 2014 (ECB) and 2012

(Danish Nationalbanken), respectively. In 2015, the Danish interest rate dropped significantly

into the negative territory when it was lowered to -0,75% - the lowest rate a country has seen

so far (Nationalbanken, 2020).

2.2.6.2.5 Term Structure

The term structure of interest rates, also sometimes known as the yield curve, represents the

“interest rates on debt securities and how these rates vary with respect to varying dates of

maturity” (Knopf & Teall, 2015). In other words, the term structure describes how much more

yield investors demand for lending their money for a longer period versus a shorter period. It is

therefore usually calculated by subtracting the yield of short-term government bonds (e.g. 2-

year bonds) from longer-term government bonds (e.g. 5-year bonds) – e.g. in Garcia &

Guerreiro (2016). Generally, the term structure of interest rates is a frequently used indicator to

determine investor sentiment and the macroeconomic outlook. A reasonably high term structure

is a sign that investors have a positive outlook on the economy (Choudhry, 2017).

The literature on banking profitability has not given substantial attention to the term structure

of interest rates (Borio et al., 2017), and hence there are comparatively fewer studies available.

The available studies generally find a significant positive correlation between Term Structure

and Profitability (Borio et al., 2017; Claessens et al., 2018; Dietrich & Wanzenried, 2011;

Garcia & Guerreiro, 2016).

15

Table 1: A summary of variables commonly affecting bank profitability.

16

2.3 Hypothesis development & independent variables

Following the research questions stated in the purpose section, two independent variables are

of interest: The first independent variable will be the interest rate a bank is subjected to. The

second independent variable will be the duration for which banks are consecutively

subjected to negative interest rates in years. More details and motivation follow in 3.5.2

Independent variables.

2.3.1 The profitability of core banking operations

Lower policy interest rates are generally expected to lower core banking profitability (which is

essentially always measured as Net Interest Margin - Bikker & Vervliet, 2018; Claessens et al.,

2018). However, due to the newness of the negative interest rate policy, it is not entirely clear

if this effect holds when rates drop into below zero. It would be plausible since banks might be

reluctant to lower interest paid to depositors below zero, even when the policy interest rate is

negative. Reasons for this effective lower bound may be the danger of losing customers to

competition or customers switching to cash savings (Claessens et al., 2018). A bank might

17

choose to accept a lower interest margin (i.e. not passing on the full effect of lower policy

interest rates) to keep its customers for future cross-selling opportunities (Berger et al., 1993).

While there seems to be a reluctance to pass on negative rates, it can still happen if banks deem

it necessary to stay afloat (see the example of Jyske Bank in Milne (2019)). Should banks decide

to pass on the negative rates, this would have the potential to widen the NIM and therefore

increase core banking profitability.

There is a further lack of information when the long-run effects of negative interest rates are

considered. One study suggests that an enduring low interest rate environment will further

compress the NIM (Claessens et al., 2018). However, no study is available for the effect of

negative interest rates. It is therefore unclear whether this effect will hold in an ongoing negative

interest rate environment. It is very plausible that Net Interest Margins will increase to positive

policy rate levels over time, once banks take the step to lower the interest rates on customer

deposits below zero. However, this would contradict the effect observed by Claessens et al.

(2018). Several further questions arise through this vacuum of information: Does NIM

compress further, extending the effect of Claessens et al., (2018)? If so, how strong is the

compression caused by negative rates? Could core banking operations even become

unprofitable? Or does NIM recover over time as soon as banks start to partly pass on negative

interest rates to customers? If so, at which pace? These questions have not been addressed

adequately by the literature yet, which brings us to the following set of hypotheses:

H1: The interest rate is not correlated with core banking operations’ profitability

H2: The duration of consecutive negative interest rates in years is not correlated with core

banking operations’ profitability

2.3.2 Overall profitability

As outlined in 2.2.6.2.4 Interest Rates, the effect of interest rates on overall profitability is still

unclear. Based on the literature, it is very difficult to predict if negative interest rates will

influence overall profitability and whether this effect is positive or negative. Negative interest

rates may influence overall bank profitability strongly. If negative interest rates would compress

the NIM significantly, and banks are dependent on a wide Net Interest Margin, overall

profitability would suffer. In this case, a NIM decline would lead to a decline in overall

profitability.

On the other hand, a declining NIM would not influence overall bank profitability if banks

diversified their business into other (profitable) fields. If negative interest rates lead to a

widening of the NIM, overall profitability might increase if banks are still relying heavily on

core banking operations. A widening NIM might not influence overall profitability if banks

shift away from core banking operations. It is therefore essential to note that a negative interest

rate environment can affect NIM and overall profitability differently. To bring clarity, we

specify the following hypotheses:

H3: The interest rate is not correlated with overall profitability

18

H4: The duration of consecutive negative interest rates (in years) is not correlated with overall

profitability

Table 2: Dependent & Independent Variables and corresponding hypotheses

19

3 Data, Model Specification and Methodology

______________________________________________________________________

This section begins with a description and justification of our research philosophy, the research

design, the mixed-methods approach, and the data collection process. Moreover, the variable

selection is described, followed by a summary table. The section ends with the methods used

for analysing both quantitative and qualitative data.

3.1 Research philosophy

The finance and banking literature is usually employing a positivist paradigm. Positivism sees

the world as a set of observable facts, which can be interpreted through reason and logic (Ayer,

1959). Therefore, a positivist paradigm goes hand in hand with quantitative data. If one looks

at the very nature of banking, the positivist preference of previous researchers can be

understood. Banking, in its essence, deals with taking deposits and other assets and lending

them out at a maximized profit, while minimizing the probability of defaulting loans. Due to

this number-driven nature of financial institutions, the literature takes on similar tendencies and

emphasises large samples and statistically derived truths about the firm and the industry.

Still, we perceive pure positivism to come with limitations - the purely quantitative focus and

the exclusion of qualitatively derived findings are two of them. Therefore, we will come from

a similar, yet not identical philosophy. We will adopt a view that became popular with other

technical fields, like information system research. Scholars in these fields had become aware of

positivism’s limitations and started to argue for a ‘mixed methods’ approach to create better

findings and, consequently, employed post-positivism (Miles & Huberman, 1994). Post-

positivism argues for “methodological pluralism”, i.e. selecting quantitative and qualitative

methods that address the research question best (Wildemuth, 1993). It also argues that

quantitative findings do not reflect absolute truth, but instead have to be understood in the

context of the research and the researcher (Guba, 1990).

We believe that such a paradigm is particularly relevant for our thesis due to the following

reasons. Even in positivist studies about banking profitability, results differ widely at times. We

pointed out some conflicting effects of variables on banking profitability in the literature review.

The post-positivist paradigm allows us to see that each researcher brings his/her own biases to

the table (reflecting in the choice of methods, sampling sources, control variables used, …) and

therefore results differ. Studies with differing results are not necessarily “wrong” in any of these

dimensions. Instead, there is a spectrum of acceptable practices which may bring different

results.

The statistical methods used, for example, can differ among Ordinary Least Squares models

(e.g. Bourke, 1989), Generalized Method of Moments estimators (e.g. Athanasoglou et al. 2008)

or Fixed Effects Models (e.g. Garcia & Guerreiro, 2016). Every author had good motivations

on why to use the specific model. However, selecting a model presents a certain bias in itself

due to each model’s shortcomings – none of them is perfect. While classical positivists might

argue that there would be one superior way to make sense of the data, post-positivists argue that

20

there is a spectrum of acceptable practices and the unavoidable researcher bias should be taken

into account. To mitigate this bias, one should triangulate and refer to other, qualitative methods

to have credible results. Through adding a qualitative dimension, we as researchers are forced

to look at our research questions and results from a different perspective. We believe that this

additional step in thinking will reduce our researcher bias and allow us to make sense of the

quantitative data better.

Post-positivism states that researchers should not limit themselves to quantitative nor qualitative

data. Instead, researchers should select (a mix of) data sources/analyses that suits their purpose

and research questions best (Guba, 1990). Our research questions will be primarily answered

by a rigorous analysis of available financial data. While no model is perfect, it is possible to

build a robust, statistical model, which is grounded in theory. According to post-positivism,

however, the sole analysis of quantitative data by the authors is biased and does not answer the

question adequately. While our quantitative analysis will give us an overview of the overall

effects for the industry, the interpretation and analysis will be helped through interviews with

industry professionals. This combination of methods and data sources further mitigates author

bias and the distance between research and real-world business.

3.2 Research design

To optimally test the effect of (prolonged) negative interest rates, we will use a deductive

approach. The deductive approach is “a study in which a conceptual and theoretical structure is

developed and then tested by empirical observations; thus, particular instances are deduced

from the general inferences” (Collis & Hussey, 2014).

We decided to study this effect using Denmark due to the following reasons. Firstly, Denmark

presents the most interesting case for studying the business effects of extremely loose monetary

policy, since interest rates reached a lower level, and have been in negative territory for the

most prolonged period of any country. Secondly, Danish commercial banks provide richer data

than many other countries.

The best way to objectively assess the interest rates’ effect on profitability is through analysing

a large sample (n) over a relevant time period (t). We aim to collect as many observations of

commercial banks in Denmark as possible and make them our sample. After that, different

statistical methods will be reviewed, and an adequate model will be built in Stata, including

variables for interest rates and control variables for other relevant profitability determinants.

Stata is used since it provides a comprehensive, easy-to-use set of statistical analysis tools and

is readily available in the JU Computer Lab.

To round off the study, we will talk to active or former bankers in the Kingdom of Denmark to

get their view on current (interest rate) conditions and its implications on their business.

Therefore, we will conduct interviews and leverage them to interpret our quantitative results.

This method of supplementation would allow us to gather insights into the businesses affected

by monetary policy decisions. There is support from the academic community that this practice

is productive: Rossman & Wilson (1985) suggest, that linking qualitative and quantitative data

21

allows researchers to confirm findings, develop analysis, provide richer detail and generate new

lines of thinking.

3.3 Mixed methods

Our motivation for the use of mixed methods is that we want to obtain a fuller picture and a

more in-depth understanding of the effect of negative interest rates on banking profitability.

Since the phenomenon is so new, some impacts of the negative rates might not be reflected in

our quantitative data, which lasts until 2018.

Johnson, Onwuegbuzie and Turner (2007) defined mixed methods research as “the type of

research in which a researcher or team of researchers combines elements of qualitative and

quantitative research approaches (e.g., use of qualitative and quantitative viewpoints, data

collection, analysis, inference techniques) for the broad purposes of breadth and depth of

understanding and corroboration.”

Johnson and Onwuegbuzie (2004) provided nine mixed methods designs from which authors

can choose when designing their study (see Figure 2). To be able to choose appropriately, the

researchers need to answer two questions: 1) Would they operate primarily within one dominant

paradigm (quantitatively vs qualitatively dominant)?, and 2) Would they carry out the phases

of the study concurrently (i.e. at the same time) or sequentially (i.e. one after another)? Johnson

and Onwuegbuzie (2004) stated that the findings of both quantitative and qualitative studies

must be integrated at some point.

We adopted the view of Johnson and Onwuegbuzie (2004) and decided to conduct a study that

has a quantitative dominant status and that is sequential. Therefore, we ended up on the fourth-

quadrant design where quantitative dominant study is conducted first, and then followed by a

qualitative study (QUAN –> qual).

The goal here is not to conduct an in-depth qualitative study, but rather to take the simplest form

of mixed-methods approach: Enhancing analysis and reporting of the main quantitative study

with supplementary qualitative data. In line with Bazeley (2018), the purpose of supplementary

data in its simplest form is to “[use] the alternative data for illustrative purposes, or to contribute

to an explanation, or to contextualize information arising in the primary source.” In this case, it

is justified that the supplementary qualitative part would be unable to be a study on its own, but

merely supports the primary (quantitative) analysis (Bazeley, 2018).

22

Figure 2: The nine mixed methods designs of Johnson and Onwuegbuzie (2014).

3.4 Data collection

3.4.1 Process and sampling for the quantitative data

We considered several high-quality data sources for our quantitative analysis and ended up

trusting the Orbis database of public and private companies by Bureau van Dijk (BvD)/Moody’s

Analytics as a secondary data source (Bureau Van Dijk, 2020). Since BvD’s data is generally

seen as credible by researchers and industry professionals, it is used for the vast majority of

banking profitability studies referenced. After checking several countries in Orbis for the

availability of rich data, the Danish banking industry seemed to be well represented in BvD’s

database. To assess the quality available for each country, we started a new search query where

all the following conditions had to be fulfilled: (1) the company had to be active, (2) the

company had to be a bank (3) it had to be a legal entity in the selected country.

The query for Denmark spawned 100 banks (all types), 83 of them including at least some

financial company information. After the query, we selected 34 metrics per company per year

for the timeframe of 2011 to 2018. These datapoints included information like ROAA, interest

income & expense and the number of employees. This dataset was then downloaded into excel

and non-commercial banks were excluded at this stage. We selected commercial banks for three

reasons: 1) Commercial banks account for most of the Danish banking industry measured by

assets. 2) Commercial banks are important contributors to economic growth, since they provide

investors with funds to borrow and increase financial deepening in a country (Otuori, 2013). 3)

Since Investment Banks make significant parts of their revenue by security underwriting rather

than with core banking operations, it is less likely that they are as heavily affected by negative

interest rates as commercial banks are.

23

That followed an analysis for consistency and integrity: If a bank lacked a primarily relevant

metric (in other words: a metric being used in this study), it was excluded from the sample. The

data was then imported into Stata and specified as panel data using the xtset command and

grouped by bank. The bank is therefore the unit of observation.

This study includes a strongly balanced panel on 21 Danish commercial banks, which accounts

for the vast majority of the Danish commercial banking industry and for 59% of the entire

Danish banking industry (measured by assets – Appendix 8.2).

Cross-sectional data (where t=1) would not allow us to answer our research question adequately,

since it observes units at only one point in time. Time-series data (where n=1) would vastly

decrease the reliability of our results, since it is not meaningful to infer from one bank observed

to the overall population. Therefore, panel data5 (where n>1 and t>1) is most suited for our

study, since it is much “richer” compared to cross-sectional or time series data. Having multiple

units observed over a long timeframe allows us to conduct relevant statistical analyses and have

a large enough sample for the population.

According to Hsiao (2007), panel data further reduces the collinearity among explanatory

variables and increases the efficiency of estimators. However, richer estimation methods than

OLS regressions are often needed to make sense of panel data with all its pitfalls.

The primary period of interest is from 2013 onwards, because the average policy interest rates

have been negative since then, as seen in Figure 3. However, we decided to include data from

2011 and 2012 to have two years of positive interest rates for comparison to the years of

negative interest rates. Data before 2011 was excluded due to the strong effect of the financial

5 “Panel data are repeated measurements at different points in time on the same units” (Cameron

& Trivedi, 2009). Also known as cross-sectional time-series data.

Figure 3: Denmark's certificate of deposit rate over time (Nationalbanken, 2020)

24

crisis on the banking industry in these years. Including these years could skew our results, since

we believe the crisis years are neither an accurate representation of the normal banking

industry’s functioning nor relevant to our research. The period studied is therefore 2011 until

2018,6 observing 8 years of ultra-low interest rates, 6 years of negative interest rates and a total

of 165 bank years.7

Bank-independent variables were downloaded from the available source, we judged most

trustworthy based on general reputation and closeness to official institutions. We could only

examine sources that were available to us at low cost or through university access. For interest

rate proxies this equates to: Denmark’s Nationalbank’s certificate of deposit rate (InterestRate)

was therefore downloaded directly from the Nationalbanken Website (Nationalbanken, 2020);

the three month yield on Danish government bonds (ThreeM_Yield) was downloaded from a

media company providing historical financial data to traders, Fusion Media Ltd. (2020a); the

three month interbank overnight rate (ThreeM_IBOR) was downloaded from the ECB statistical

data warehouse (European Central Bank, 2020). The data to calculate the interest rate term

structure (TermStructure) was also downloaded from the financial data provider Fusion Media

Ltd. (namely the yields on 5- and 2-year Danish sovereign bonds – Fusion Media Ltd., 2020b;

Fusion Media Ltd., 2020c).

A problem that arises when downloading interest rate data on a yearly basis is, that often year-

end values are provided (i.e. the interest rate on Dec-31). This is an issue, since the year-end

interest rate does not accurately represent the average interest rate throughout the year (i.e. the

interest rate banks have actually been subjected to). To avoid this pitfall, each of these variables

was downloaded with monthly intervals and then averaged for each year by the authors. This

method takes into account intra-year changes of interest rates and quite accurately represents

the interest climate a bank was operating in. We use the Annual Percentage Rates (APR) for all

rates to enhance comparability in line with the literature.

Yearly GDP growth (GDP) was downloaded from the OECD website (OECD, 2020).

3.4.2 Process and sampling for the supplementary (qualitative) data

We believe that the analysis of our quantitative data would benefit from leveraging qualitative

data directly from the industry. We further believe, that the people working directly with core

banking operations can offer the most valid industry insights. To capture the thinking ideally,

semi-structured interviews seem appropriate. They have the benefit of allowing us to aim for

specific topics, while maintaining enough conversational flexibility to capture each banker’s

unique insights. This allows us to ask follow-up questions if a certain point the interviewee

6 2018 was the last year we were able to obtain data for.

7 The bank years observed is 165 instead of 168 since some banks were lacking some control

variable data. This deviation is normal, even in strongly balanced panels and slightly reduces

the number of bank years observed.

25

makes seems particularly relevant. Interviews are further preferred over other qualitative data

collection methods, since they are most ideally suited for our purpose: Capturing individual,

historical insights on the implications of the negative interest rate environment.

To find interviews, “cold” E-Mails and LinkedIn InMail’s were sent out to 130-150 industry

practitioners, more senior employees of larger banks preferred. Due to space limitations and the

supplementary nature of this data, a more detailed description of the sampling procedure can be

found in Appendix 8.3. Since every (potential) participant was chosen based on our judgement,

this sample is purposive.

In the end, three interviews were conducted. The interviewees included a Private Banker of a

large Nordic bank, Interviewee 1; a Senior Banking Analyst of Denmark’s Nationalbanken,

Interviewee 2; and the CEO of a mid-sized Danish bank, Interviewee 3. The structured part of

the interview questions can be found in Appendix 8.4.

Due to the ongoing Coronavirus situation during the writing process, interviews were conducted

via telephone. The main benefit of having technology-supported interviews, is that both

interviewer and interviewee can be in surroundings they feel comfortable with (home/office).

A comfortable and familiar setting is important for a free flow of information.

3.5 Variable selection

3.5.1 Dependent variables

We fundamentally agree with the argument of using inflation-invariant and size-dependent

profitability indicators, noted by previous researchers in the field. Besides that, it is imperative

to 1) include at least one metric measuring the profitability of interest rate-sensitive core

banking operations, and 2) include at least one metric measuring the overall business

profitability. The only inflation-invariant, size-dependent profitability metric measuring the

margin on core banking operations known to us is the Net Interest Margin (NIM).

Additionally, we will use the Return on Average Assets to measure overall profitability. Not

only is ROAA inflation-invariant and size-dependent, but it also considers the risk taken on

through leverage and cannot be influenced by changes in equity stakes. Due to these reasons,

we agree with the literature’s sentiment that the return on (average) assets is the key overall

profitability indicator for banks (Athanasoglou et al., 2008; Dietrich & Wanzenried, 2011;

Golin, 2001). The ROAA is preferred over ROA due to the accuracy gains ROAA brings.

We use both NIM and ROAA as our profitability measures, making our hypotheses

synonymous to:

26

Table 3: Dependent variables and hypotheses.

3.5.2 Independent variables

As outlined in the Hypothesis Development section, there is a significant gap in the literature

on how a negative interest rate environment affects banking profitability. There are unknowns

both for the correlation between Interest Rates and Banking Profitability in this environment,

as well as unknowns on the long-term effects of a rate environment below zero. The following

section will firstly provide a detailed discussion of interest rates as an independent variable and

then conclude with how the negative rate environment is going to be incorporated into the study.

Banks are subjected to numerous different interest rates. The interest rate banks receive on their

deposits at the national bank is different than the rate they receive on deposits in the interbank

overnight market. This happens, because the different rates are influenced by different factors.

Using several proxies for interest rates allows us to capture the interest rates banks are truly

subjected to.8 The specified model is then run once with each proxy applied and the results are

compared (control variables kept constant). We use all interest rate proxies known to us that

could influence bank profitability and that have been used in the literature. Since we have a

total of three proxies for interest rates, this will result in three separate models being run for

each profitability metric (Table 4). Using several relevant proxies for the independent variable

increases the robustness of our results. If there is the same conclusive result for all the proxies

it is less likely that the obtained result is due to chance.

8 A proxy variable is defined as „A variable used instead of the variable of interest when that

variable of interest cannot be measured directly“ (Oxford Reference, 2020). Since our goal is

to measure the interest rates banks are subjected to, it is meaningful to use several types of

central bank- and market-interst rates that have the potential to influence banking profitability

as proxies.

27

Table 4: The six models and their corresponding interest rate proxy and profitability metrics.

In the following, we will explain the proxies used in more detail and give reference to previous

literature on them.

3.5.2.1 InterestRate

Our first proxy is usually referred to as the policy interest rate. This is the interest banks receive

for very short-term deposits at the Danish national bank. The certificates are zero-coupon

securities, sold on the last banking day of the week and usually settled 7 days later. Setting the

interest rate on these very short-term deposits is an important tool of Nationalbanken’s

monetary policy (Nationalbanken, 2020). This directly influences the interest rates charged and

given by commercial banks. This proxy is widely used, since it can only be controlled by central

banks and has an effect on the commercial banking sector’s deposit and lending rates (e.g. in

Campmas, 2020; Short, 1979). However, it does not represent the commercial bank deposit and

lending rates exactly, but rather influences them very strongly.

3.5.2.2 ThreeM_IBOR

The Three-Month Interbank Offered Rate (or Copenhagen Interbank Offered Rate – CIBOR)

also represents the short-term borrowing cost of a bank. This rate represents the interest a bank

has to pay, if it wants to borrow money from another bank for three months. It also

approximately represents the rate a bank can expect to receive if it lends out money to other

banks for the same period (Eurostat, 2020). The rate is administrated by the Danish Financial

Benchmark Facility (DFBF) but set by the banking industry. Every day, the DFBF receives

submissions of 12 or more Danish banks (”panel banks”), in which the banks estimate their own

borrowing cost (Danish Financial Benchmark Facility, 2019). What makes this rate interesting,

is that it gives a view from the industry’s inside. Essentially, it is a form of self-assessment,

representing what banks think their borrowing cost would be. Interbank Offered Rates are

generally subject to more fluctuations due to external influences (compared with the rather static

policy rates). The interesting background of this proxy also led to its use in the banking

profitability literature (e.g. in Borio et al., 2017)

However, the industry-dependency of IBOR’s can be a downside. The London Interbank

Offered Rate particularly has been under attack in the ”LIBOR Scandal” of 2012. It came to

light, that banks were colluding and manipulating their LIBOR submissions to benefit their own

28

trading positions. Furthermore, the LIBOR submission was often lower than the banks’ actual

borrowing cost - so they could appear in good financial health (McConnell, 2013).

While there is no evidence to our knowledge that CIBOR was inaccurate or even manipulated,

the strong industry-dependency of this rate should be kept in mind throughout the study.

3.5.2.3 ThreeM_Yield

The Three-Month Danish Sovereign Bond Yield is the anticipated interest rate one would

receive, if one were to buy a three-month Danish bond on the market. The market Yield to

Maturity (YTM) of short-term bonds move quite in line with current policy interest rates, since

the Danish government can set the interest rate it pays on new bond issuings. The interest

received on sovereign debt is usually referred to as the ”risk-free” interest rate for investors –

an important benchmark for available risk-free return in the market. This benchmark is used in

various industries to determine investment decisions (for example in the Capital Asset Pricing

Model), as well as in the banking profitability literature (e.g. in Claessens et al., 2018).

This rate also is more dynamic than the regular policy rates, since it is influenced by market

conditions and investor sentiment. For example, in a stock market downturn, investors usually

start buying safe assets, like sovereign bonds. The increased demand for bonds then lowers the

bonds’ market YTM, since investors are willing to pay more for bonds with the same interest

rate.

As can be seen below, the three rates usually move in very similar directions.

Figure 4: Visualization of interest rates over time. All values in %.

-1.5

-1

-0.5

0

0.5

1

1.5

2

20

11

20

12

20

13

20

14

20

15

20

16

20

17

20

18

Ultra-Low Interest Rates in 2011-2018

InterestRate ThreeM_IBOR ThreeM_Yield

29

3.5.2.4 YearsInNegInterestRate/YNeg_IBOR

A main point of interest in this study is the question whether negative interest rates have a

different effect on banking profitability if they endure. The conflict surrounding this question

is described in the literature review and hypothesis development section.

YearsInNegInterestRate therefore measures the consecutive years in which InterestRate and

ThreeM_Yield have been negative. The policy rate and the Three-Month Yield share this

variable, since they went and stayed below zero simultaneously.

ThreeM_IBOR however only became negative in 2015 and stayed below zero ever since,

therefore we created a variable measuring the number of consecutive negative years for our

IBOR regressions called YNeg_IBOR.

3.5.3 Control variables

The control variables selected for our study mirror the variables found significant during the

literature analysis. As discussed in 2.2.6 Bank profitability determinants, only a handful of

variables can be reasonably expected to impact bank profitability. As far as control variables

go, our study therefore aims to include all the variables that can be expected to impact bank

profitability.

It should further be noted that the proxies for the determinants are quite homogenous within the

literature. For example, it is likely that credit risk influences profitability. Credit risk is

essentially always measured as Loan Loss Provisions over Gross Loans. Our study therefore

mirrors these commonly agreed upon measurements/proxies.

Several other variables were previously included in the model but excluded due to

multicollinearity issues. These variables include the Herfindahl-Hirschman Index of market

concentration and the CPI as an inflation measure. This issue is discussed in 4.4 Robustness

tests.

30

Table 5: A summary of the variables selected in the thesis.

31

3.6 Possible outcomes for the independent variables

The following charts should give a brief overview of possible outcomes for the study. The first

level of the graph resembles banking profitability before the introduction of negative interest

rates. On the second level, the correlation between interest rates and the profitability metric is

examined. Three outcomes are possible: interest rates can either have a positive, a negative or

no correlation with profitability. On the third level we examine whether negative interest rates

influence profitability differently over time. This level has the same three potential outcomes:

an enduring negative interest rate environment can also either have a positive, negative or no

correlation with profitability.

Figure 5: Possible outcomes for profitability of core banking operations.

Figure 6: Possible outcomes for overall bank profitability.

32

3.7 Method of quantitative data analysis

3.7.1 Motivating the statistical model

As mentioned in the chapter about research philosophy, there is a diversity of statistical methods

used in the literature. The banking profitability literature generally uses three types of models,

it is rare to find other methodologies. These methodologies are: population-averaged Ordinary

Least Squares models (PA-OLS – e.g. Bourke, 1989; Demirguc-Kunt & Huizinga, 1999; Duraj

& Moci, 2015; Molyneux & Thornton, 1992; Short, 1979), Fixed/Random effects Models (e.g.

Athanasoglou et al., 2008; Ben Naceur & Goaied, 2008; Garcia & Guerreiro, 2016;

Konstantinos, 2012; Pasiouras & Kosmidou, 2007) and Generalized Method of Moments

Estimators (GMM - e.g. Athanasoglou et al., 2008; Bikker & Vervliet, 2018b; Konstantinos,

2012; Trujillo‐Ponce, 2013).

We decided against PA-OLS and GMM for various reasons outlined in Appendix 8.5. A

detailed discussion of these two models can be found there. Instead, we utilized a Fixed effects

Model since it allows regressors to be correlated with the time-invariant individual error 𝛼𝑖. We

believe that this is an important issue in our model, since there are likely individual differences

between the banks, we did not account for. These would be unobserved abilities of the bank to

generate profit, for example a high customer goodwill. These unaccounted individual factors

(errors) may be correlated with our regressors – a bank with high goodwill from its customers

may be able to sell more services to its customers (i.e. higher diversification). This can then

again influence the bank’s ability to generate profits. In a Fixed effects model, this limited form

of endogeneity is allowed and is accounted for by the time-invariant individual error 𝛼𝑖. The

correlation between the time-invariant individual error (𝛼𝑖) and the regressors needs to be close

to 0 for a Random effects or PA-OLS model to be accurate. Please refer to the results section

to see whether this argument confirmed by the Stata output.

For a Fixed effects Model to be appropriate, regressors need to be high in within variation

(variation over individuals and time), rather than in between variation (variation across

individuals - Cameron & Trivedi, 2009). The variables closely examined in this study (interest

rates and duration of negative rates) are high in within variation, since they change over time

and do not differ between banks.9 This makes a Fixed effects model more appropriate.

If we were to include dummies on the other hand (e.g. for categorizing banks in small, medium

and large), a Random effects model would be appropriate. These dummies would be low in

within variation (do not change over time), and high in between variation (do vary across

individuals).

A Random effects model further assumes that 𝛼𝑖 is (as the name implies) random. This is quite

a strong assumption, which we think is unlikely to be filled in the case of bank panel data.

Banking profitability is likely determined by many more individual factors than merely the

9 For descriptive statistics please refer to Appendix 8.8.

33

independent variables in our model. This is, because not every dimension of a bank’s profit

generating activities can be accurately quantified. For example, personal connections of

relationship managers might be an individual difference, which we are unable to account for. A

Fixed effects model solves this issue.

However, a Random effects model does come with upsides. Besides the ability to measure time-

invariant variables, a Random effects model (if appropriate) leads to more efficient estimates

and generally lower standard errors (Allison, 2009). Yet, the result may be biased if a Fixed

effects model would in fact be appropriate.

3.7.2 Hausman test

The Hausman test is a standard statistical procedure to decide between a Fixed and Random

Effects model. H0 is, that individual effects are random and both fixed and random Effects

model are consistent. H1 is, that the model coefficients diverge systematically between Fixed

and Random Effects Model. If we were to accept H0 at the 5% Level, we would use the random

effects model due to the efficiency gains it provides. If we reject H0 however, the less restrictive

Fixed Effects model is more appropriate due to the inconsistency between the models (Cameron

& Trivedi, 2009).

Our Hausman test is conducted in Stata with the sigmamore option, which estimates the

disturbance variance for the efficient estimator and then bases both covariance matrices on that

estimation (Cameron & Trivedi, 2009). H0 can be rejected at the 1% Level for both ROAA and

NIM. This result further confirms that a Fixed Effects model is most appropriate (see Appendix

8.6).

3.7.3 Employed model

Based on the discussion above, we decided to specify the following Fixed Effects model. The

corresponding Variable explanation can be seen in the table below. The model will be run using

Stata’s command for Fixed effects regressions (xtreg, fe).

34

Table 6: Model Specifications.

Through this model, the coefficients with their respective robust standard errors for the

independent- and control-variables will be determined. In line with general statistical practice,

coefficients will be seen as significant if they reach a significance level of at least α=0.05

(Freeman et al., 2017).

3.8 Method of supplementary (qualitative) data analysis

Whereas quantitative analysis deals with clear conventions that the researcher can use, the

methods of analysis in qualitative data are not well formulated. The latter has fewer guidelines

or standardized procedures against self-delusion, unreliable or invalid conclusions (Miles,

1979). Therefore, we decided to follow the Flow Model of Miles and Huberman (1994) who

defined analysis as consisting of the following three activity flows:

35

1. Data Reduction: “the process of selecting, focusing, simplifying, abstracting, and

transforming the data that appear in written-up field notes or transcripts” (Miles &

Huberman, 1994, P. 10). After transcribing the data, we directly started this process by

data selection, summarizing and paraphrasing. As it was suggested by the researchers,

the process continued until the report was complete.

2. Data Display: “an organized, compressed assembly of information that permits

conclusion drawing and action” (Miles & Huberman, 1994, P. 11). This stage is very

important as a decision can be made here whether to draw justified conclusions or

analyse further, depending on the understanding of the display. Displays are “designed

to assemble organized information into an immediately accessible, compact form” and

can include any type of matrices, graphs, charts, etc. In this step, the data was organized

in graph-format using Excel when possible.

3. Conclusion Drawing and Verification: at this final stage of qualitative data analysis, we

started to make sense of the data by “noting regularities, patterns, explanations, possible

configurations, causal flows, and propositions” (Miles & Huberman, 1994, P. 11). We

also made sure to follow the advice of the researchers that the conclusions may not

appear directly, and even if they do, they are more likely to be preliminary and

ambiguous; final conclusions come out only when data collection is over. The next step

was to verify these conclusions and test for their validity. This was done by conducting

short follow-up interviews or contacting the interviewees through E-Mail.

3.9 Ethical considerations

Throughout the process of conducting and writing our thesis, research ethics were taken into

consideration. Research ethics “relates to questions about how we formulate and clarify our

research topic, design our research and gain access, collect data, process and store our data,

analyse data and write up our research findings in a moral and responsible way” (Saunders,

Lewis and Thornhill, 2007, P. 178). Therefore, we followed the ethical guideline of Bell and

Bryman (2007), as cited in Collis and Hussey (2014), which requires the researchers to:

1. Inform the involved participants about the aim and nature of research;

2. Avoid any potential harm to the participants through the research process;

3. Respect their dignity, cause them no discomfort and protect their privacy;

4. Ensure the data collected in the study remain confidential;

5. Avoid deception;

6. Avoid misunderstanding, misinterpreting, and falsely reporting the findings;

7. And finally, declare aby personal or professional issues that may have had an influence

on the research.

36

4 Results & Analysis

______________________________________________________________________

This section will firstly give an overview over key metrics of the employed models. Secondly,

results will be objectively summarized, followed by robustness tests. Thirdly, the determinants

of both NIM and ROAA will be discussed with a focus on the effect of (continuously) negative

interest rates. To ensure structural consistency, determinants unrelated to interest rates will

also be discussed here.

4.1 Quantitative results

In Table 7 the results of the models are shown.10 The table has a two-way split: the upper part

describes the core banking operations’ profitability (NIM) models and the lower part describes

the overall profitability (ROAA) models. For each of the two dependent variables, three models

are available, with the interest rate proxies outlined in 3.5.2 Independent variables.

The NIM and ROAA could be well explained by our specified models. The F-Test shows a high

significance for all of our models and the within R2 is quite high for banking profitability studies

- ranging from 0.7527 (NIM) to 0.8629 (ROAA). Overall we can therefore conclude that our

independent variables were appropriately selected and explain around 75% of the Danish

commercial banks’ NIM and around 86% of the Danish commercial banks’ ROAA.

As explained in the methodology section, we believed a Fixed effects model to be appropriate

because we suspected a correlation between the time-invariant individual error (𝛼𝑖) and the

regressors. This suspicion turned out to be correct, the table shows that that this correlation

averages to -0.5537 (NIM models) and to -0.6122 (ROAA models). Our model choice is

therefore confirmed by the regression results.

4.2 Results overview for NIM regressions

The models paint a clear picture of which variables predict core banking profitability. The most

influential predictor of the NIM by far is Size (LogSizeDKK). Significant at the 5% level in

every model, an increase of one unit in the Logarithm of total assets decreases NIM by -0.6290

(±0.2421).11 The second biggest predictor of NIM is TermStructure: it is significant at the 5%

Level in the ThreeM_IBOR model, decreasing NIM by 0.2485 (±0.0995) for each one percent

increase in the slope of the yield curve. Further important predictors are the Years in Negative

Interest Rate (-0.1093±0.0359) and the Credit Risk (0.0909±0.0265). Significant, although less

impactful NIM predictors include Diversification (0.0111%±0.005) and Liquidity

(0.0076±0.0036).

10 Unadulterated Stata output can be fond under Appendix 8.9.

11 The number in brackets always denotes the robust standard errors for a 95% confidence

interval unless otherwise stated.

37

Interestingly, the interest rate proxy was not a significant predictor of NIM in any of the models

at the 5% level. It should be noted, that ThreeM_Yield predicts the NIM at the 10% level (-

0.1001±0.0560), however. The other two variables found to be insignificant are GDP Growth,

Efficiency and Capitalization; the last one being significant at the 10% level in every NIM

model (0.0484±0.0254).

4.3 Results overview for ROAA regressions

The specified ROAA models also allow us to assess its predictors. The independent with the

biggest impact on overall profitability is by far credit risk with ROAA decreasing by -

0.6228±0.0605 for each percentage point increase of Loan Loss Provisions per Gross loans.

Two weaker predictors of overall profitability can be identified: Efficiency (-0.0592±0.0098)

and Liquidity (-0.0237±0.0046). No other factors were found to be significant at the 5% Level.

Diversification is significant at the 10% Level. Each percentage point increase in the ratio of

Non-Interest Income Per Operating Income decreases ROAA by -0.0139%±0.0074. This would

indicate that a more diversified bank is slightly less profitable if one accepts the 10% Level of

significance.

4.4 Robustness tests

Our models pass the goodness of fit test, with Prob > F 0.000 for all our models. For ROAA,

all our models have a very high R2 value of 0.8620-0.8629. For NIM, the R2 values are between

0.7527 and 0.7581. These results further prove the explanatory power of our model and are

based on the within-estimation of R2. As mentioned previously, we believe this to be the most

important metric, because interest rates and the years in negative interest rates are high in

within-correlation.

If errors are not normally distributed, the standard errors of our model would be incorrect.

Stata’s vce(robust) option allows us to control for this issue and obtain cluster robust standard

errors (Cameron & Trivedi, 2009).

Our model was inspected for multicollinearity using the variance inflation factor (vif) test. As

a rule of thumb, multicollinearity is likely a problem when the vif for an independent variable

exceeds 10. In the original model, both the Herfindahl-Hirschman Index of market

concentration (HHI) and the CPI as an inflation measure were included. However, the variance

inflation factor for these independents exceeded 10. Strong multicollinearity of these two

variables with GDP was subsequently identified. Since we believe GDP to be the most relevant

predictor of banking profitability of the three (as seen in e.g. Pervan et al., 2015), we kept GDP

and dropped HHI/CPI. According to the variance inflation factor calculations after these

adjustments, problematic cases of multicollinearity are unlikely. The vif test and a correlation

matrix can be found in Appendix 8.7.

We avoided the omitted variable bias to the best of our knowledge through rigorous analysis of

the available data and the literature. TermStructure, if excluded, would have presented such a

bias. Its case will be discussed in the next section.

38

The fact that we built a total of six models and took into account three measures for interest

rates and two measures for profitability further strengthens our argument. The consistent results

for all three interest rate proxies makes it unlikely that the results are due to chance. The further

differentiation between two relevant profitability metrics allows us to make stronger inferences

on the banking industry.

39

Table 7: Regression Results

40

4.5 Analysis of the results

4.5.1 Interest rates and the NIM

H1: The interest rate is not correlated with the NIM

All our models fail to reject H1 at the 5% level. We could therefore not find evidence that there

is a correlation between the interest rates Danish banks are subjected to and their Net Interest

Margins. It is worth noting that the ThreeM_Yield model finds a negative correlation between

interest rates and the NIM, significant at the 10% level. This correlation describes that a one

percent absolute decrease in Danish government bond yields increases the Net Interest Margin

by approximately 10 basis points (0.1001±0.056). However, due to the insufficient statistical

significance of this correlation and the fact that it was only observed in one model, we dismiss

this result.

It is worth mentioning that all interest rate proxies were found significant at the 5% level before

including the TermStructure variable. Following the literature, we initially thought this control

variable to be of little relevance. According to our models, however, the term structure of

interest rates is found to be vastly more significant than the interest rates themselves. To avoid

omitted variable biases, we suggest including a control variable for the term structure when

examining banking profitability.

As shown in the literature review, several studies found positive correlations between interest

rates and the NIM (Borio et al., 2017; Campmas, 2020; Claessens et al., 2018; Demirguc-Kunt

& Huizinga, 1999). Our study suggests that such direct correlations are non-existent once a

negative interest rate environment is entered, and the term structure and the years in negative

interest rates are taken into account.

One explanation could therefore be, that previous studies find interest rates to affect the NIM

because they did not control for the Term Structure of interest rates (as it was initially the case

in our study). Furthermore, the interest rate level itself might not be as significant as the duration

of continued exposure to low/high interest rates. Only one study known to us (Claessens et al.,

2018) includes a variable to measure the effect of continued exposure to low rates. There is

therefore still no consensus on whether low interest rates themselves or continued exposure to

them cause lower profitability. Our results suggest the latter is true in a negative interest rate

environment in the medium-term.

One of the professionals interviewed, Interviewee 1, stated another explanation for the statistical

insignificance of interest rates. He noted that the lowering interest rates below zero only partly

affects the NIM. In his experience, a drop of interest rates below zero would lower the NIM

initially due to the banks’ unwillingness to pass on the negative rates to keep customers. This

view was also supported by Interviewee 2, who is an Analyst at Denmark’s central bank. He

said:

41

“There was a slow reaction from the banks. It is more like a prisoner’s dilemma for them. They

do not like to start to impose negative interest rates on private customers in fear of losing some

of them. So, therefore, every bank management would prefer that other banks start imposing

the negative interest [rates] so that they could be fast followers.”

There is some evidence that banks do not pass on low interest rates to keep customers (Berger

et al., 1993; Claessens et al., 2018). Interviewee 1, however, made a distinction between the

short-and long-run. He stated that after a certain time, banks would pass on the negative rates

more directly, widening the NIM over time. He claims that a reduction of interest rates further

below zero would then be passed on. If NIM firstly narrows when rates drop, but then widens

when rates are narrowed further, this might lead to the variable having no statistically significant

effect in one direction or the other – ultimately leading to the variable being statistically

insignificant in a linear regression model. Interviewee 2 also agreed to Mr. F’s idea, stating:

“When the banks faced negative interest rates, they considered whether they should transfer it

to the customers, or whether they should absorb it themselves. At the beginning, they started to

absorb it themselves and were very careful about when to impose them … This year [2020],

however, they started to pass on negative interest rates to ordinary customers. It was, in fact,

the second-largest bank, when taking into consideration banking activities in Denmark, which

started it, called Jyske Bank, and many others have already followed the example of Jyske.”

This leads us to the second hypothesis:

H2: The duration of consecutive negative interest rates in years is not correlated with the NIM

All our models reject H2. The InterestRate model allows us to reject H2 at the 5% level, while

the ThreeM_Yield and ThreeM_IBOR models allow us to reject H2 at the 1% level. This strong

evidence suggests that core banking operations will become less profitable the longer banks are

subjected to negative rates. Averaging all models, it shows that each consecutive year of

negative policy interest rates narrows the NIM by an expected -0.1093 (±0.0359). With 95%

confidence, we can therefore conclude that one additional year of negative policy rates will

lower the NIM between -0.0734 and -0.1452.

Claessens et al. (2018) find that NIM compresses over time in a low (but positive) interest rate

environment. They found a compression of 8 basis points per additional year of low interest

rates (an absolute reduction in NIM of 0.08). Before conducting our study, it was hypothesized

that if rates drop below zero (and remain negative long enough), banks will start passing on low

rates to their customers. This, in turn, would widen the NIM instead of narrowing it further. Our

results suggest that 5 years is not long enough to make banks pass on negative rates. During the

years observed, a similar effect as found by Claessens et al. (2018) for a low interest rate

environment can be measured. On average, our models predict a reduction of the NIM by 11

basis points (-0.11) for each year of additional negative rates. This suggests that a negative

interest rate environment has the same medium-term effects as a low interest rate environment.

However, the effect is likely to be even stronger in the negative territory.

42

The following seems to be a likely explanation for this effect: newly issued fixed-interest, multi-

year securities (like certificates of deposits - CDs)12 pay customers “too much” interest due to

the banks' unwillingness to pass on negative rates. This, in turn, could lower the NIM over time

as the more profitable fixed-interest CDs from the positive interest rate era expire and the new

(less profitable) CDs become more significant to the banks’ income. While it is very hard to

find quantitative data to verify this reasoning, Claessens et al. (2018) explain the effect

similarly. This would further explain the non-correlation between interest rates and the NIM: A

lowering of interest rates does not immediately lower profitability – what matters is prolonged

exposure to the negative rates.

This result is of a tremendous impact on the Danish banking industry. If one considers that a

yearly reduction of profitability compounds over time, the results are even more alarming. The

table below shows a hypothetical scenario of what could happen to Danish banks’ NIM if the

negative interest rate environment persists. In the table, year 0 represents the base profitability

during times of positive rates (averaged out over all banks in the sample). Year 1 then represents

the profitability after one year of negative interest rates (a reduction of 11 basis points). Year 2

shows the effects of 2 years of negative rates, and so on. After five years of negative rates, the

NIM is therefore expected to be between -16% (simple average) and -43% (weighted average)

lower than it was during positive rates, ceteris paribus. After ten years, the NIM would have

been almost completely eradicated according to our weighted average scenario, rendering core

banking operations unprofitable (all other things equal).

While this clearly is an over-simplified scenario of the multi-facetted predictors of the NIM in

the real world, it goes to show how a continuous negative interest rate environment has serious

potential to damage the banking industry.

12 Definition: “a type of investment in which customers earn interest for saving their money for

a fixed period of time [e.g. 5 years]” (Cambridge University Press, 2020a) The examples of

CDs is taken here for clarity, although this could also be true for other securities.

43

Table 8: Potential effects of prolonged exposure to negative rates

Combining the interviews and quantitative data leads us to the following conclusion: until

recently, banks’ core operations have suffered from the continued exposure to the negative

rates, more so than from a further reduction of the rates themselves. Their NIM has compressed

the longer they have been subjected to the extremely loose monetary policy measures. This was

measured through our quantitative models. After getting through this phase, our interviews

suggest that there seems to be a turnaround through managerial decisions: banks did not want

to take the lowered profitability any longer and are now passing on the negative rates. It is likely

that now, after enough time passed, the compressing effect of a negative interest rate

environment has become weaker, if not reversed completely. The profitability of core banking

operations is restored as soon as profitable loans and deposits take over the majority of the

interest income, creating a U-curve:

Figure 7: Development of the NIM in a prolonged negative interest rate environment.

While it is too early to verify this effect on a quantitative basis – after all, our sample lasts until

2018, and the qualitative results were obtained in 2020 - it is worth revisiting this conclusion

with a quantitative model in several years and see if it holds.

44

4.5.2 Other factors influencing the NIM

The below chart shows the simple and weighted average of the NIM for the observed time

period. A simple average is used to represent each bank equally, while the weighted average

gives more significance to relatively larger banks (measured by size). It is still useful to use a

simple average to assess the development for smaller firms. The four largest banks control for

over 93% of the sample measured by assets and hence dominate the weighted average curve.

The two vertical lines represent the years in which yearly-averaged interest rates dropped into

negative territory (InterestRate: 2013; ThreeM_Yield/ThreeM_IBOR: 2015).

Figure 8: NIM over time.

As explained, Size (LogAssetsDKK) was the most impactful predictor of the NIM, showing that

larger banks have a lower NIM. This can also be observed in the chart, the NIM weighing larger

banks more heavily is consistently lower than the NIM weighing smaller and large banks

equally. An increase in size might therefore not necessarily desirable for commercial banks

wanting to increase the profitability of their core banking operations.

One potential explanation can be found in the nature of the NIM formula:

If banks do not manage to increase (interest income – interest expenses) proportionally to their

Assets, NIM will inevitably suffer. Other possible explanations (e.g., that smaller, more

specialized banks can generate a higher NIM) are subject to further research.

TermStructure is further a significant predictor of NIM. An absolute increase of one percent in

the difference between the yield of 5- and 2-year Danish government bonds leads to a reduction

of the NIM by -0.2485 (±0.0995) in the ThreeM_IBOR model. Similar effects were found in

the other two models at a 10% significance level.

45

An increase of TermStructure signals that investors are willing to pay a higher premium for

debt with longer maturity. To a certain extent, this is perceived as a sign of a healthy economy.

A decrease in TermStructure, however, generally signals that investors have a negative outlook

on the economy and perceive an increased potential for a recession. A better economic outlook

is therefore associated with lower NIMs. If the economic outlook is good, investors might

decrease the “safe” amount they hold in their bank account and shift into other (more profitable)

investments like stocks. This reduction of deposits could then lead banks to increase the interest

paid on deposits to win investors back with more attractive returns. This practice would lower

the NIM. While this may sound logical, there is no absolute evidence in our study to back this

up. Therefore, we suggest that research should further explore the negative correlation between

NIM and TermStructure.

Credit risk (LLPPerGrossLoans) was also found to be a significant NIM predictor. A one

percent absolute increase in LLPPerGrossLoans (equivalent to a significant reduction in credit

risk) leads to an increase of the NIM of 9 basis points. Core banking operations therefore

become more profitable, the more banks set aside for nonperforming loans. Our results are in

line with Dietrich & Wanzenried (2011), who also found a significant positive correlation

between the NIM and LLPPerGrossLoans.

Diversification (NonInterestIncPerOpInc) and liquidity (NetLoansPerAssets) were also found

to influence the NIM. Our expectation was that a higher diversification would lower the NIM,

since diversification leads to a higher share of non-interest income (Campmas, 2020). We found

the opposite – which is surprising - but it is worth noting that the effect of diversification on the

NIM is very small with a one percent increase in NonInterestIncPerOpInc increasing the NIM

between 0.0061 and 0.0161 (95% confidence interval). One of the reasons for this effect could

be that more diversified banks also manage to charge higher interest margins to their customers.

Furthermore, increasing liquidity lowers the NIM. This is also expected, since there seems to

be a trade-off between liquidity and profitability (Goddard et al., 2004; Molyneux & Thornton,

1992). The less liquid the bank (i.e. the higher its NetLoansPerAssets), the more money it can

earn from its loans. Each one unit increase in NetLoansPerAssets increases the NIM by an

expected 0.8 basis points.

Since the effects of both NonInterestIncPerOpInc and NetLoansPerAssets are rather small and

are only peripherally related to our topic, we will omit a further discussion of these results.

GDP Growth, Capitalization and Efficiency all did not impact the NIM. Economic growth, a

high Equity/Assets ratio or a highly efficient organizational structure therefore do not directly

lead to an increase in core banking operations’ profitability.

4.5.3 Interest rates and the ROAA

H3: The interest rate is not correlated with the ROAA

We fail to reject H3 at any significance level for all our models. Our models therefore suggest

that overall banking profitability is not correlated with the interest rates banks are subjected to.

46

It seems like banks overall profitability have not been influenced by changes in interest rates

over the years. This contrasts some of the studies which did find a positive relationship between

overall profitability and interest rates (Borio et al., 2017; Demirguc-Kunt & Huizinga, 1999).

While technically interest rates are sometimes seen as a predictor of overall profitability in the

literature, it is worth noting that the effect is often a lot weaker for overall profitability as for

the NIM (Claessens et al., 2018).

As mentioned previously, Claessens et al. (2018) are the only authors which also implement an

independent variable measuring the continuity of the low interest rate environment, making our

results particularly comparable. In regards to overall profitability, our analysis extends the

results of said study: interest rates do not affect overall profitability in a negative interest rate

environment.

H4: The duration of consecutive negative interest rates (in years) is not correlated with the

ROAA

We also fail to reject H4 at any significance level for all our models. There is no evidence in

our data to suggest that the ROAA of Danish commercial banks is correlated with consecutive

years of negative interest rates. Banks seem unaffected by a continuously negative interest rate

environment.

When the European Central Bank recently examined the effects of a long-term low interest rate

environment on banking profitability in a working paper, it found no material negative effect

on banks’ overall profitability (Altavilla et al., 2017). Our results suggest that the same logic

extends into the negative interest rate environment, even in the medium term. Altavilla et al.,

(2017) argue that the potentially long-run negative effects on overall profitability are offset by

improved macroeconomic conditions the low interest rates cause. As low interest rates often act

as a monetary stimulus to the economy and banks are shown in some studies to profit from

improved macroeconomic conditions (Dietrich & Wanzenried, 2011; Kanas et al., 2012;

Trujillo‐Ponce, 2013), this is a plausible theory.

However, we believe it is important to look at other possible explanations. In the following

section, we will take a look at other trends that could influence overall profitability.

4.5.4 Other factors influencing the ROAA

As of now, banking profitability is still mostly determined by internal factors, under the control

of bank management: Credit risk, Liquidity and Efficiency all were significant predictors of the

ROAA.

Starting with credit risk, a one percent absolute increase in LLPPerGrossLoans decreases

ROAA by -0.6228%±0.0605, making riskier banks more profitable overall. Since higher

LLPPerGrossLoans means that higher expenses are recorded, a decline in ROAA is inherent to

this control variable and in line with the literature (Campmas, 2020).

47

Figure 9: Credit Risk over time

It can be observed that the industry significantly decreased LLPPerGrossLoans over time,

leading to an increase in overall credit risk. Based on the data available to us, we are unable to

say whether this is due to a conscious decision of management to artificially increase

profitability ratios or the result of other (macroeconomic) factors. This topic could however be

considered for further research.

Liquidity (NetLoansPerAssets) and efficiency (CostToIncomeRatio) were also significant

predictors of the ROAA. The higher the NetLoansPerAssets ratio, the lower the liquidity of the

bank. An increase in NetLoansPerAssets (lowering of the banks’ liquidity) was associated with

a slightly decreased profitability. This is in line with the literature’s theory that a bank holding

more liquid assets will be more profitable (Bourke, 1989; Duraj & Moci, 2015).

Figure 10: Liquidity over time.

The lower the CostToIncomeRatio, the higher the operational efficiency of the bank. As

expected, a lower CostToIncomeRatio (higher operational efficiency) was also associated with

an increase in ROAA. This result is in line with the literature’s findings (Athanasoglou et al.,

2008; Dietrich & Wanzenried, 2011; García-Herrero et al., 2009; Tan & Floros, 2012). Since

48

the effect of these variables on ROAA is rather weak and there is no recognizable industry trend,

we will omit a further discussion of these factors.

Figure 11: Efficiency over time.

The following factors were expected to influence ROAA, but were statistically insignificant:

GDP growth, Size and Capitalization.

The insignificance of economic growth can be partly explained, as the period of 2011-2018 has

quite consistent GDP growth rates. A significant effect of GDP growth is expected if one

observes banks over several business cycles (as in e.g. Dietrich & Wanzenried, 2011).

A large asset base is further not affecting overall profitability. Athanasoglou et al., (2008)

suggested that Size’s effect on overall profitability has a flat inversed U-Shape. Such a

correlation may create statistical insignificance in a linear regression model. Further research

could therefore examine the effect of size on profitability with a non-linear statistical model.

A high Equity/Assets ratio further did not impact overall bank profitability, in line with Alper

& Anbar (2011). Due to the limited space and our inability to find trends or significant effects

regarding capitalization, we will omit a further discussion of this control variable.

49

5 Conclusion

______________________________________________________________________

This section serves as a link between the research purpose of the thesis and the previously

analysed empirical results, as well as concludes the main findings.

Having confirmed a research gap in the literature, this study was conducted to investigate the

relationship between persistently negative interest rates and profitability in the Danish banking

industry.

The immediate effect of interest rates on core banking operations’ profitability in a near and

below-zero environment is statistically insignificant. Rather, a prolonged negative interest rate

environment reduces the NIM for the observed period by 11 basis points per additional year in

negative interest rates. The profitability-reducing effect of a prolonged negative interest rate

environment is severely stronger than the effect of a mere low interest rate environment. While

this causes pressure on bank management to act to restore profitability, the companies are

unwilling to pass on the negative rates immediately to remain competitive. This decision,

however, likely is the reason that core banking profitability suffers in the medium-term: The

newly issued loans and certificates of deposit (CDs) are less profitable than the securities from

the positive-interest rate era. Since many fixed-interest CDs are issued for a multi-year period,

the effect of the new, unprofitable CDs does not immediately materialize on the banks’ income

statement. Rather it takes some years as the relatively less profitable loans and CDs become a

bigger part of a banks’ core operations. This causes profitability of core banking operations to

suffer over time.

Since bank management is aware of these severe impacts, they seem to start passing on negative

rates eventually, restoring the core operations’ profitability in the long run. This is suggested

by industry practitioners, but not yet reflected in quantitative data available – as explained

above, it takes time for such changes to materialize in profitability metrics.

While core banking operations of Danish banks suffered in the negative interest rate

environment, the same cannot be said about overall business profitability. Neither interest rates

themselves, nor a prolonged exposure to negative rates influenced the ROAA. It seems highly

likely that bank management can successfully maintain overall business profitability in a

negative interest rate environment through controlling various internal factors like credit risk,

liquidity and business efficiency. Overall bank profitability may further benefit from the

improved macroeconomic conditions negative interest rates cause.

50

6 Discussion

______________________________________________________________________

This section will firstly address the additional findings of the thesis, namely the implications for

bank management. Secondly, limitations and further research recommendations will be

presented.

6.1 Implications for bank management

Danish banks have passed on negative rates at different paces since they were introduced. When

Jyske Bank firstly passed on negative interest rates, the company strategized using a customers’

deposit balance: it initially decided to target customers with a balance above 7.5M DKK

(roughly $1.1M), according to Interviewee 2. When other banks followed the example of Jyske,

however, the latter lowered the limit drastically to target customers with deposits in excess of

700K DKK ($94K). According to Interviewee 2, the Danish people understood the reasoning

behind negative deposit rates, which is why they did not withdraw funds from Jyske. The

company then further decreased the limit to 250K DKK ($33.5K) amid the coronavirus

pandemic. For comparison: The largest bank in Denmark – Danske Bank – targeted customers

with balances above 1.5M DKK ($201K). One might expect that such different target levels

would cause people to switch banks in order to avoid negative interest rates on their savings.

However, this was not the case, as the Danish are “not that interest sensitive, as long as they are

satisfied”, according to Interviewee 2. A main contributor is the thought customers have, that

switching banks takes significant time and work. To avoid this, customers would be willing to

pay a little extra money if they are satisfied with the services offered:

“For many people, banking business is a mean not a goal in such. If you can make it easy for

your customers, they are satisfied. And many of them, in fact, are not that interest sensitive. If

they were, you would see much more movement of the customers between the different banks”

This leads us to believe that the factor that will determine a banks’ customer base is not when

it passes on negative interest rates. Rather, it is the quality of customer service offered by the

bank. Luckily, this factor remains in management control.

51

Figure 12: Diversification over time

The figure above shows that the Danish banking industry is undergoing some changes. Since

the negative interest rate introduction in 2013, Danish banks have continued to diversify into

other streams of revenue. While this figure does not tell us which revenue streams they

diversified in, it seems like bank management has grasped that the sole providing of loans is

not enough to secure profitability for the future.

6.2 Limitations and further research

We believe our paper analyzed the medium-term effect of a negative interest rate environment

on Danish bank profitability as thoroughly as possible. One limitation was the timeframe of our

panel: since the effects of the negative rate policy seem not to have fully materialized yet in the

quantitative data, this might limit the correlations to be true only in the medium-term.

Based on the interviewee’s hints that bank’s NIM is changing now, we suggest quantitively

revisiting the effects of Denmark’s negative interest rate environment on banking profitability

in a few years. Ideally, a non-linear model can be employed to capture the potentially u-shaped

curve of the NIM over time. Such correlations might not be captured in a linear model like ours.

We further encourage more researchers to examine the negative interest rate environments’

impact on banking profitability using other popular banking profitability models, namely PA-

OLS and GMM. Conducting further studies for the same environment and timeframe using

different statistical methods would give more certainty about the effect of the policy. Due to the

scope of this thesis we were limited to one method only.

A further limitation is that the potentially significant HHI and CPI had to be dropped. Further

research could use these control variables to verify whether the results found here still hold true.

Since our qualitative data was highly limited due to its supplementary nature, a very timely

research suggestion would be to go more in-depth and qualitatively capture the current effects

of the negative interest rate environment. This turning point, at which Danish banks slowly start

to pass on negative rates, would make interesting material for a (multiple) case study on what

52

happens inside these organizations. It also opens the question, what factors determine

commercial banks’ customer success in a negative interest rate environment.

As described in the analysis section, several control variables were found to be significant that

could not be fully interpreted through the data available to us. The following topics could

therefore be examined in future studies:

- The impact of specialization and diversification on banking profitability (in a negative

interest rate environment) and its implications

- Term Structure of interest rates in a negative rate environment and its implications for

banking profitability

- Reasons for the rise of commercial banks’ credit risk taking during Denmark's negative

interest rate period

Since negative interest rates are becoming a global trend, we would further suggest replicating

similar studies in other countries imposing negative policy rates.

53

7 Reference list

Abreu, M., & Mendes, V. (2001). COMMERCIAL BANK INTEREST MARGINS AND

PROFITABILITY: EVIDENCE FOR SOME EU COUNTRIES. 12.

Al-Jafari, M. K., & Alchami, M. (2014). Determinants of Bank Profitability: Evidence from

Syria. Journal of Applied Finance and Banking; Athens, 4(1), 17–45.

Allison, P. D. (2009). Fixed effects regression models. SAGE.

Alper, D., & Anbar, A. (2011). Bank Specific and Macroeconomic Determinants of

Commercial Bank Profitability: Empirical Evidence from Turkey. Business and

Economics Research Journal; Bursa, 2(2), 139–152.

Altavilla, C., Boucinha, M. & Peydró, J.-L.. (2017). Monetary policy and bank profitability in

a low interest rate environment. Retrieved February 02, 2020, from

http://dx.publications.europa.eu/10.2866/825393

Arellano, M., & Bond, S. (1991). Some Tests of Specification for Panel Data: Monte Carlo

Evidence and an Application to Employment Equations. The Review of Economic

Studies, 58(2), 277–297. JSTOR. https://doi.org/10.2307/2297968

Arshadi, N., & Lawrence, E. C. (1987). An empirical investigation of new bank performance.

Journal of Banking & Finance, 11(1), 33–48. https://doi.org/10.1016/0378-

4266(87)90020-3

Athanasoglou, P. P., Brissimis, S. N., & Delis, M. D. (2008). Bank-specific, industry-specific

and macroeconomic determinants of bank profitability. Journal of International

Financial Markets, Institutions and Money, 18(2), 121–136.

https://doi.org/10.1016/j.intfin.2006.07.001

54

Ayadi, N., & Boujelbene, Y. (2012). The Determinants of the Profitability of the Tunisian

Deposit Banks. IBIMA Business Review Journal, 1–21.

https://doi.org/10.5171/2012.165418

Ayer, A. J. (1959). Logical positivism. Free Press.

Bazeley, P. (2018). Integrating Analyses in Mixed Methods Research. SAGE Publications Ltd.

https://doi.org/10.4135/9781526417190

Ben Naceur, S., & Goaied, M. (2008). The Determinant of Commercial Bank Interest Margin

and Profitability: Evidence from Tunisia. Frontiers in Finance and Economics, 5.

https://doi.org/10.2139/ssrn.856365

Berger, A., Hancock, D., & Humphrey, D. (1993). Bank Efficiency Derived from the Profit

Function. Journal of Banking & Finance, 17(2–3), 317–347.

https://doi.org/10.1016/0378-4266(93)90035-C

Berger, A. N., & Bouwman, C. H. S. (2013). How does capital affect bank performance during

financial crises? Journal of Financial Economics, 109(1), 146–176.

https://doi.org/10.1016/j.jfineco.2013.02.008

Berger, A. N., & Bouwman, C. H. S. (2015). Bank Liquidity Creation and Financial Crises.

Elsevier Inc. https://www.sciencedirect.com/book/9780128002339/bank-liquidity-

creation-and-financial-crises

Bikker, J. A. (2010). Measuring Performance of Banks: An Assessment. The Journal of Applied

Business and Economics; Thunder Bay, 11(4), 141–159.

Bikker, J. A., & Vervliet, T. M. (2018). Bank profitability and risk-taking under low interest

rates. International Journal of Finance & Economics, 23(1), 3–18.

https://doi.org/10.1002/ijfe.1595

55

Board of Governors of the Federal Reserve System. (2020). Federal Reserve Board—Open

Market Operations. Board of Governors of the Federal Reserve System. Retrieved May

05, 2020, from https://www.federalreserve.gov/monetarypolicy/openmarket.htm

Bolingbroke, E. (2020). Futures Market Sees Negative U.S. Policy Rate by Early 2021.

Bloomberg.com. Retrieved May 16, 2020, from

https://www.bloomberg.com/news/articles/2020-05-07/negative-u-s-policy-rate-seen-

by-early-2021-in-futures-market

Bolt, W., de Haan, L., Hoeberichts, M., van Oordt, M. R. C., & Swank, J. (2012). Bank

profitability during recessions. Journal of Banking & Finance, 36(9), 2552–2564.

https://doi.org/10.1016/j.jbankfin.2012.05.011

Borio, C., Gambacorta, L., & Hofmann, B. (2017). The influence of monetary policy on bank

profitability. International Finance, 20(1), 48–63. https://doi.org/10.1111/infi.12104

Bourke, P. (1989). Concentration and other determinants of bank profitability in Europe, North

America and Australia. Journal of Banking & Finance, 13(1), 65–79.

https://doi.org/10.1016/0378-4266(89)90020-4

Boyd, J. H., & Runkle, D. E. (1993). Size and performance of banking firms. Journal of

Monetary Economics, 31(1), 47–67. https://doi.org/10.1016/0304-3932(93)90016-9

Buiter, W. H. (2009). Negative nominal interest rates: Three ways to overcome the zero lower

bound. The North American Journal of Economics and Finance, 20(3), 213–238.

https://doi.org/10.1016/j.najef.2009.10.001

Bureau Van Dijk. (2020). Orbis | Company information across the globe | BvD. Retrieved

February 24, 2020, from https://orbis.bvdinfo.com/

56

Cambridge University Press. (2020a). CERTIFICATE OF DEPOSIT | meaning in the

Cambridge English Dictionary. Cambridge English Dictionary. Retrieved March 03,

2020, from https://dictionary.cambridge.org/dictionary/english/certificate-of-deposit

Cambridge University Press. (2020b). COMMERCIAL BANK | meaning in the Cambridge

English Dictionary. Cambridge English Dictionary. Retrieved March 03, 2020, from

https://dictionary.cambridge.org/dictionary/english/commercial-bank

Cambridge University Press. (2020c). FINANCIAL LEVERAGE | meaning in the Cambridge

English Dictionary. Cambridge English Dictionary. Retrieved March 03, 2020, from

https://dictionary.cambridge.org/dictionary/english/financial-leverage

Cameron, A. C., & Trivedi, P. K. (2009). Microeconometrics using Stata. Stata Press.

Campmas, A. (2020). How do European banks portray the effect of policy interest rates and

prudential behavior on profitability? Research in International Business and Finance,

51. https://doi.org/10.1016/j.ribaf.2018.09.001

Choudhry, M. (2017). An introduction to banking: Principles, strategy and risk management

(2nd edition). John Wiley & Sons, Inc.

Claessens, S., Coleman, N., & Donnelly, M. (2018). “Low-For-Long” interest rates and banks’

interest margins and profitability: Cross-country evidence. Journal of Financial

Intermediation, 35, 1–16. https://doi.org/10.1016/j.jfi.2017.05.004

Clews, R. J. (2016). Project Finance for the International Petroleum Industry.

https://www.sciencedirect.com/book/9780128001585/project-finance-for-the-

international-petroleum-industry

Collis, J., & Hussey, R. (2014). Business research: A practical guide for undergraduate &

postgraduate students.

57

Danish Financial Benchmark Facility. (2019). CIBOR Calculation Methodology. Retrieved

March 15, 2020, from https://dfbf.dk/wp-content/uploads/2019/12/CIBOR-calc.pdf

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

and profitability: Some international evidence. World Bank Economic Review, 13(2),

379–408. https://doi.org/10.1093/wber/13.2.379

Dietrich, A., & Wanzenried, G. (2011). Determinants of bank profitability before and during

the crisis: Evidence from Switzerland. Journal of International Financial Markets,

Institutions and Money, 21(3), 307–327. https://doi.org/10.1016/j.intfin.2010.11.002

Dovern, J., Meier, C.-P., & Vilsmeier, J. (2010). How resilient is the German banking system

to macroeconomic shocks? Journal of Banking & Finance, 34(8), 1839–1848.

https://doi.org/10.1016/j.jbankfin.2009.12.001

Duraj, B., & Moci, E. (2015). Factors Influencing the Bank Profitability—Empirical Evidence

from Albania. Romanian Economic and Business Review; Brasov, 10(1), 60–72.

Elsas, R., Hackethal, A., & Holzhaeuser, M. (2006). The Anatomy of Bank Diversification.

https://doi.org/10.5282/UBM/EPUB.1167

European Central Bank. (2020). Denmark Interbank 3-month Offered Rate - Ask price or

primary activity, average of observations through period. Retrieved February 26, 2020,

from

https://sdw.ecb.europa.eu/quickview.do?SERIES_KEY=143.FM.M.DK.DKK.DS.MM

.CIBOR3M.ASKA

Eurostat. (2020). Short-term interest rates: Three-month interbank rates.

https://data.europa.eu/euodp/en/data/dataset/FJekIA3eWTjlExaBYGPgw

Faure, A. P. (2014). Interest Rates 1: What are Interest Rates? SSRN Electronic Journal.

https://doi.org/10.2139/ssrn.2542083

58

Freeman, J., Shoesmith, E., Sweeny, D., Anderson, D., & Williams, T. (2017). Statistics for

Business and Economics (4th edition.). Cengage Learning EMEA.

https://www.bokus.com/bok/9781473726567/statistics-for-business-and-economics/

Fusion Media Ltd. (2020a). Denmark 3-Month Bond Yield. Retrieved February 25, 2020, from

https://www.investing.com/rates-bonds/denmark-3-month-bond-yield

Fusion Media Ltd. (2020b). Denmark 2-Year Bond Yield. Retrieved February 25, 2020, from

https://www.investing.com/rates-bonds/denmark-2-year-bond-yield

Fusion Media Ltd. (2020c). Denmark 5-Year Bond Yield. Retrieved February 25, 2020, from

https://www.investing.com/rates-bonds/denmark-5-year-bond-yield

Garcia, M. T. M., & Guerreiro, J. P. S. M. (2016). Internal and external determinants of banks’

profitability: The Portuguese case. Journal of Economic Studies, 43(1), 90–107.

https://doi.org/10.1108/JES-09-2014-0166

García-Herrero, A., Gavilá, S., & Santabárbara, D. (2009). What explains the low profitability

of Chinese banks? Journal of Banking & Finance, 33(11), 2080–2092.

https://doi.org/10.1016/j.jbankfin.2009.05.005

Gertner, R. H., Scharfstein, D. S., & Stein, J. C. (1994). Internal Versus External Capital

Markets. The Quarterly Journal of Economics, 109(4), 1211–1230. JSTOR.

https://doi.org/10.2307/2118361

Gilchrist, S., & Mojon, B. (2018). Credit Risk in the Euro Area. Economic Journal, 128(608),

118–158. https://doi.org/10.1111/ecoj.12427

Goddard, J., Molyneux, P., & Wilson, J. O. S. (2004). The profitability of european banks: A

cross-sectional and dynamic panel analysis. The Manchester School, 72(3), 363–381.

https://doi.org/10.1111/j.1467-9957.2004.00397.x

59

Golin, J. L. (2001). The bank credit analysis handbook: A guide for analysts, bankers and

investors. Wiley.

Goodman, P. S. (2020). Why the Global Recession Could Last a Long Time. The New York

Times. Retrieved May 10, 2020, from

https://www.nytimes.com/2020/04/01/business/economy/coronavirus-recession.html

Guba, E. G. (1990). The paradigm dialog. Sage Publications, Inc.

Hsiao, C. (2007). Panel data analysis—Advantages and challenges. Heidelberg, 16(1), 1–22.

http://dx.doi.org.proxy.library.ju.se/10.1007/s11749-007-0046-x

Idiab, A. I. M., Haron, M. S., & Ahmad, S. B. H. (2011). Commercial Banks & Historical

Development. 6.

Jenkins, P. (2010). What is bank capital?. Financial Times.

https://www.ft.com/content/f98164ba-bf65-11df-965a-00144feab49a.

Jewell, J. J., & Mankin, J. A. (2011). What is Your ROA? An Investigation of the Many Formulas

for Calculating Return on Assets (SSRN Scholarly Paper ID 2155943). Social Science

Research Network. https://papers.ssrn.com/abstract=2155943

Jobst, A., & Lin, H. (2016). Negative Interest Rate Policy (NIRP): Implications for Monetary

Transmission and Bank Profitability in the Euro Area. IMF Working Papers, 16(172),

1. Retrieved May 01, 2020, from https://doi.org/10.5089/9781475524475.001

Johnson, R. B., & Onwuegbuzie, A. J. (2004). Mixed methods research: A research paradigm

whose time has come. Educational researcher, 33(7), 14-26.

Johnson, R. B., Onwuegbuzie, A. J., & Turner, L. A. (2007). Toward a definition of mixed

methods research. Journal of mixed methods research, 1(2), 112-133.

60

Kanas, A., Vasiliou, D., & Eriotis, N. (2012). Revisiting bank profitability: A semi-parametric

approach. Journal of International Financial Markets, Institutions and Money, 22(4),

990–1005. https://doi.org/10.1016/j.intfin.2011.10.003

Kiviet, J. F. (1995). On bias, inconsistency, and efficiency of various estimators in dynamic

panel data models. Journal of Econometrics, 68(1), 53–78.

https://doi.org/10.1016/0304-4076(94)01643-E

Knopf, P. M., & Teall, J. L. (2015). Risk Neutral Pricing and Financial Mathematics: A Primer.

Elsevier.

Konstantinos, A. (2012). The determinants of European bank profitability before, during and

after the financial crisis of 2008. 45.

Lartey, V. C., Antwi, S., & Boadi, E. K. (2013). The Relationship between Liquidity and

Profitability of Listed Banks in Ghana. 4(3), 9.

McConnell, P. (2013). Systemic operational risk: The LIBOR manipulation scandal. The

Journal of Operational Risk, 8(3), 59–99. https://doi.org/10.21314/JOP.2013.127

Menicucci, E., & Paolucci, G. (2016). The determinants of bank profitability: Empirical

evidence from European banking sector. Journal of Financial Reporting and

Accounting, 14(1), 86–115. https://doi.org/10.1108/JFRA-05-2015-0060

Miles, M. B. (1979). Qualitative Data as an Attractive Nuisance: The Problem of Analysis.

Administrative Science Quarterly, 24(4), 590–601. JSTOR.

https://doi.org/10.2307/2392365

Miles, M. B., & Huberman, A. M. (1994). Qualitative Data Analysis: An Expanded Sourcebook.

SAGE.

61

Milne, R. (2019). Denmark’s Jyske Bank imposes negative interest rates. The Financial Times.

Retrieved January 25, 2020, from https://www.ft.com/content/ebabea22-c31b-11e9-

a8e9-296ca66511c9

Mizruchi, M. S. (2001). Commercial Banking. In N. J. Smelser & P. B. Baltes (Eds.),

International Encyclopedia of the Social & Behavioral Sciences (pp. 2272–2276).

Pergamon. https://doi.org/10.1016/B0-08-043076-7/04269-8

Molyneux, P., & Thornton, J. (1992). Determinants of European bank profitability: A note.

Journal of Banking & Finance, 16(6), 1173–1178. https://doi.org/10.1016/0378-

4266(92)90065-8

Munteanu, I. (2012). Bank Liquidity and its Determinants in Romania. Procedia Economics

and Finance, 3, 993–998. https://doi.org/10.1016/S2212-5671(12)00263-8

Nationalbanken. (2020). Official interest rates. Retrieved February 01, 2020, from

http://www.nationalbanken.dk/en/marketinfo/official_interestrates/Pages/default.aspx

Nessibi, O. (2016). The Determinants of Bank Profitability: The Case of Tunisia. International

Journal of Finance & Banking Studies; Istanbul, 5(1), 39–50.

OECD. (2020). Gross Domestic Product (GDP). Retrieved February 25, 2020, from

https://data.oecd.org/gdp/gross-domestic-product-gdp.htm

Otuori, O. H. (2013). Influence of exchange rate determinants on the performance of

commercial banks in Kenya. European Journal of Management Sciences and

Economics, 1(2), 86–98.

Oxford Reference. (2020). Proxy variable—Oxford Reference. Oxford Reference. Retrieved

March 03, 2020, from https://doi.org/10.1093/oi/authority.20110803100351624

62

Pasiouras, F., & Kosmidou, K. (2007). Factors influencing the profitability of domestic and

foreign commercial banks in the European Union. Research in International Business

and Finance, 21(2), 222–237. https://doi.org/10.1016/j.ribaf.2006.03.007

Pervan, M., Pelivan, I., & Arneric, J. (2015). Profit persistence and determinants of bank

profitability in Croatia. Ekonomska Istrazivanja: Znanstveno-Strucni Casopis; Pula,

28(1), 284–298. http://dx.doi.org/10.1080/1331677X.2015.1041778

Petria, N., Capraru, B., & Ihnatov, I. (2015). Determinants of Banks’ Profitability: Evidence

from EU 27 Banking Systems. Procedia Economics and Finance, 20, 518–524.

https://doi.org/10.1016/S2212-5671(15)00104-5

Petter, S. C., & Gallivan, M. J. (2004). Toward a framework for classifying and guiding mixed

method research in information systems. 37th Annual Hawaii International Conference

on System Sciences, 2004. Proceedings of The, 10 pp.-.

https://doi.org/10.1109/HICSS.2004.1265614

Rivard, R. J., & Thomas, C. R. (1997). The effect of interstate banking on large bank holding

company profitability and risk. Journal of Economics and Business, 49(1), 61–76.

https://doi.org/10.1016/S0148-6195(96)00041-0

Rossman, G. B., & Wilson, B. L. (1985). Numbers and Words: Combining Quantitative and

Qualitative Methods in a Single LargeScale Evaluation Study. Evaluation Review.

Saunders, M., Lewis, P., & Thornhill, A. (2007). Research methods for business studies. Second

Impression, Dorling Kindersley (India) Pvt Ltd.

Scholtens, B., & van Wensveen, D. (2003). The Theory of Financial Intermediation: An Essay

On What It Does (Not) Explain (Research Report 2003/1). SUERF Studies.

https://www.econstor.eu/handle/10419/163455

63

Short, B. K. (1979). The relation between commercial bank profit rates and banking

concentration in Canada, Western Europe, and Japan. Journal of Banking & Finance,

3(3), 209–219. https://doi.org/10.1016/0378-4266(79)90016-5

Statista. (2020a). Denmark: Gross domestic product (GDP) 2008-2018. Statista. Retrieved

February 10, 2020, from https://www.statista.com/statistics/587205/gross-domestic-

product-gdp-in-denmark/

Statista. (2020b). Total banking assets as GDP percentage Denmark 2018. Statista. Retrieved

February 10, 2020, from https://www.statista.com/statistics/810498/total-banking-

assets-as-share-of-gdp-denmark/

Stein, J. C. (1997). Internal Capital Markets and the Competition for Corporate Resources. The

Journal of Finance, 52(1), 111–133. JSTOR. https://doi.org/10.2307/2329558

Stowell, D. (2010). An introduction to investment banks, hedge funds, and private equity: The

new paradigm. Academic Press/Elsevier.

Tan, Y., & Floros, C. (2012a). Bank profitability and inflation: The case of China. Journal of

Economic Studies, 39(6), 675–696. https://doi.org/10.1108/01443581211274610

Tan, Y., & Floros, C. (2012b). Bank profitability and GDP growth in China: A note. Journal of

Chinese Economic and Business Studies, 10(3), 267–273.

https://doi.org/10.1080/14765284.2012.703541

Tripe, D. (1998). Cost to income ratios in Australasian banking. Department of Finance,

Banking and Property, College of Business, Massey University.

Trujillo‐Ponce, A. (2013), What determines the profitability of banks? Evidence from Spain.

Accounting & Finance, 53: 561-586.

64

Werner, R. A. (2016). A lost century in economics: Three theories of banking and the

conclusive evidence. International Review of Financial Analysis, 46, 361–379.

https://doi.org/10.1016/j.irfa.2015.08.014

Wildemuth, B. (1993). Post-Positivist Research: Two Examples of Methodological Pluralism.

The Library Quarterly, 63(4), 450–468. https://doi.org/10.1086/602621

Ziliak, J. P. (1997). Efficient Estimation With Panel Data When Instruments Are

Predetermined: An Empirical Comparison of Moment-Condition Estimators. Journal of

Business & Economic Statistics, 15(4), 419–431.

https://doi.org/10.1080/07350015.1997.10524720

65

8 Appendices

8.1 Definitions

Denmark’s Policy

Interest Rate

The interest rate banks receive on deposits and pay for borrowing at the

central bank of Denmark (Nationalbanken). Denmark’s central bank can

set this rate higher or lower to meet its goals. This rate moves the interest

paid and received by Commercial Banks. A lower policy interest rate

will usually result in cheaper loans (e.g. mortgages) and lower interest

rates on saving accounts. The policy interest rate is a powerful tool to

stimulate (or cool down) a country’s economy.

Core Banking

Operations

In the context of this thesis, core banking operations are defined as

accepting deposits (on which interest is traditionally paid) and lending

out a multiple of these deposits for receiving interest.

Net Interest

Margin (NIM)

Usually, the interest commercial banks pay on deposits is lower than the

interest they receive on loans. The difference between interest paid and

interest received is the Net Interest Margin. In other words, it is the

profit margin of a commercial banks’ Core Banking Operations.

Commercial

Banks

”A bank with branches in many different places that offers services to

people and businesses, for example, keeping money in accounts and

lending money” (Cambridge University Press, 2020b)

Put another way, commercial banks generally focus on Core Banking

Operations. However, commercial banks may also engage in other

business activities, like financial guarantees and derivative sales/trading

(A. N. Berger & Bouwman, 2015). Commercial banks should be

differentiated from investment banks, which generally focus on helping

companies to generate funding (e.g. issuing stocks and bonds) and

company mergers & acquisitions (Stowell, 2010).

”Passing on

Negative Rates”

Means commercial banks pass on the negative policy rate to their

customers to retain profitability. One would then likely have to pay to

deposit money at a bank.

Proxy Variable „A variable used instead of the variable of interest when that variable

of interest cannot be measured directly“ (Oxford Reference, 2020).

66

8.2 Calculation of sample coverage

Sources: (Statista, 2020a; Statista, 2020b)

8.3 Description of supplementary sampling procedure

To conduct interviews with experts of the industry, the authors reached out to approximately

130-150 unique E-Mail addresses or LinkedIn Profiles likely associated with bankers of the

Danish banking industry as follows:

Companies from our sample were looked up on LinkedIn, which allows us to find employees

relevant to our study easily. Larger banks (by assets) generally engage in more depositing and

lending, which led us to assume that they have a bigger influence on the industry. Therefore,

employees of larger banks were contacted first. Under the “Employees” tab, 2-4 people with

mostly senior ranks and lending exposure in their work were selected for contact. The rationale

behind selecting senior people was, that they might have a better overview over the business

and can provide the more interesting insights. The duration of activity in the banking industry

was further taken as selection criteria, since we expect, that more experienced professionals can

better reflect on the effects of a multiple-year negative interest rate environment.

This approach excludes employees without a LinkedIn profile. However, we can only contact

a very small fraction of all Danish bankers in any case. Furthermore, even fewer of the bankers

contacted are willing to do an interview. Since the interviewees are therefore determined by

essentially randomness, the potential sampling error of excluding employees without a

LinkedIn profile can be disregarded as irrelevant.

The bankers selected for contact were then either contacted through LinkedIn’s InMail feature

or through E-Mails. LinkedIn’s premium plan was activated on our personal accounts to send

InMail messages. E-Mail addresses were found through services like RocketReach, Hunter.io,

but also through research on the employer’s website. The first two services provide a search

engine for publicly available data of professionals. Since the services merely index already

public data, a privacy breach is unlikely. Furthermore, only one interview request per person

was sent to ensure nobody was bothered by our practices. The sample here is quite clearly a

convenience sample, since we accept interviews with every banker we deem to be experienced

in the industry.

In line with the QUAN-> qual approach we aimed to conduct interviews after the regression

results were available. Due to Interviewee 1 allowing us a time slot pre-regression, we accepted

the offer, as it is our belief that some qualitative data pre-regression is better than not receiving

67

this particular data at all. For this reason, Appendix 8.4 shows two different set of questions:

The original set has been modified after the regressions were done.

8.4 Appendix interview questions

Pre-regression interview (Interviewee 1):

• What do you see as the main opportunities for banking today?

• What do you see as the main challenges for banking today?

• What are the macroeconomic factors that most influence your organisation?

• How did the bank react to Nationalbanken’s negative interest rate policy when it was

first introduced? What happened to your organisation since then (organisational

dimension, customer dimension)?

• Do you approve of the Nationalbanken’s decision to put interest rates in a negative

territory and keep them there? Why or why not?

• Has the bank been able to maintain its level of profitability after the introduction of the

negative interest rates? Did it change over time (our results suggests a longer time lowers

interest rates more)?

• What do you believe Nationalbanken’s monetary policy will look like in the future and

how do you prepare for it?

• In Japan, banks changed their business model due to the ongoing low interest rate

environment, do you see something similar happening to your bank?

• (Research is thin, but we expect that our statistical analysis will show that the negative

interest rate policy did not change the overall profitability of banks in Denmark or your

bank. Is there anything you think we are missing when we examine profitability of

Banks in Denmark?)

Post-regression interviews (Interviewee 2&3):

• Our data showed that a continuously negative interest rate environment lowers the main

margin of banking profitability. Do you agree with these results? Why? Why not?

• Do you expect this lower margin of profitability will hinder the growth of the industry?

• What do you think the measures or the actions need to be taken by banks in order to

keep the growth of industry?

• Jyske Bank has been the first bank to launch the negative mortgage interest rate in the

world, do you expect other banks will follow Jyske bank’s example?

• Do you think that the negative interest rate environment will keep on for a long period

of time?

• Any expectations about how Nationalbanken’s monetary policy will look like in the

future?

68

8.5 Appendix statistical models

8.5.1 Generalized Method of Moments Estimators (GMM)

GMM Estimators have often been used in the literature, since they deal with some of the

endogeneity and heteroskedasticity problems of panel data. However, GMM these benefits

come at a cost. Firstly, the GMM estimators often become imprecise when N is too small and

one wants to have robust standard errors. Secondly, the standard deviations of GMM estimators

are usually quite large compared to Fixed Effects Models (Arellano & Bond, 1991; Kiviet,

1995) . Thirdly, a large number of moment conditions leads to substantial bias (Ziliak, 1997).

Since the number of moment conditions would be quite high for our model, this would again

make GMM unsuitable. Lastly, it turns out that GMM models can suffer from a too-many-

instruments problem when applying the Hansen-Test (as suggested in Dietrich & Wanzenried,

2011). After considering these factors, we decided against applying a GMM model.

8.5.2 Population-Averaged Ordinary Least Squares models (PA-OLS)

Population averaged OLS models have also been widely used in the literature, partly due to

their simplicity. This approach pools the observations across banks and estimates a basic linear

regression model. Quite similar to regular multivariable OLS models, population averaged OLS

models for panel data estimate a linear model with a constant, parameters for each regressor

and a random error term. However, there is an important assumption population-averaged

models make, that often cannot be filled in panel data: it assumes that regressors are strictly

exogenous. This can be a problem, since it is likely that the errors in a regression model will be

correlated over time and individual (within correlation) or correlated across individuals

(between correlation). (Cameron & Trivedi, 2009) This important consideration led us to

finding a PA-OLS model unsuitable for our purpose.

69

8.6 Hausman tests

ROAA Test

70

NIM Test

71

8.7 Multicollinearity tests

Variance Inflation Tables:

72

73

74

Correlation matrix:

75

8.8 Descriptive statistics

The xtsum output is shown here with variables in the following order:

CompanyName

Year Dummy

ROAA

InterestRate

ThreeM_IBOR

ThreeM_Yield

TermStructure

YearsInNegInterestRate

YNeg_IBOR

CPI

GDP

HHI

LLPPerGrossLoans

NetLoansPerAssets

NonInterestIncPerOpInc

EquityPerTotalAssets

NIM

CosttoIncomeRatio

CompanyNum (Grouping Variable for xtset)

LogAssetsDKK

76

77

78

8.9 Original stata output

ROAA Models

79

NIM Models

80