Quantitative Easing Effect on
Bank Profitability
A study on the relationship between quantitative easing and bank profitability in
Sweden
BACHELOR DEGREE PROJECT THESIS WITHIN: Economics NUMBER OF CREDITS: 15 PROGRAMME OF STUDY: International Economics
AUTHOR: Markus Tingvall, Erik Håbäck JÖNKÖPING: May 2021
Acknowledgement
We, Markus Tingvall and Erik Håbäck would like to express our sincere gratitude to all the individuals who have contributed towards the making of this thesis.
First, we would like to thank the invaluable input from our supervisor, Rafael B. De Rezende, and his suggestions to approach QE’s effect on bank profitability through stock prices making use of an event study regression and the proposition to measure the QE shock using forward rates based on Swanson (2021) and Altavilla et al. (2018). We are truly thankful for his expertise and words of encouragement throughout the thesis process.
Secondly, we want to extend a heartfelt thank you to our dear friends Philip Stefanov, Robin Hinny and Moa Lundberg for taking their time to give us feedback and constructive criticism.
Lastly, a genuine thank you to our friends and family for their constant support.
Markus Tingvall Erik Håbäck
Jönköping International Business School May 2021
Bachelor Thesis in Economics
Title: Quantitative easing effect on bank profitability: A study on the relationship between quantitative easing and bank profitability in Sweden
Authors: Markus Tingvall and Erik Håbäck Tutor: Rafael B. De Rezende
Date: 2021-05-24
Key Terms: Quantitative Easing, Bank Profitability, Transmission Channels, Unconventional Monetary Policy
Abstract
We analyse the effects of quantitative easing (QE) on Swedish bank profitability on the four largest banks in Sweden between 2015-2021 by utilizing daily stock prices as a proxy for bank profit. Using an event study approach, we find that QE has a significant positive effect on bank profitability in Sweden as wholesale funding conditions improve. This suggests that structural differences in bank funding have an impact on the effect of QE. Furthermore, we investigate the individual effects of QE on bank profitability. We determine that QE benefits banks with higher credit losses on their balance sheet due to improvements in debt serviceability. Finally, we complement our study by investigating how QE affects debtholders of the banks through credit default swaps (CDS). We find that QE reduces prices on CDS, therefore signalling an improvement in wholesale funding
conditions. This indicates that both equity and debtholders perceive the effects of QE positively.
Contents 1. Introduction ... 1 1.1 Problem statement ... 4 1.2 Purpose ... 5 1.3 Previous Research ... 6 1.4 Disposition ... 8 2. Theory ... 9
2.1 Quantitative easing in Sweden ... 9
2.2 Transmission Channels ...10
2.3 Signalling Channel ...12
2.4 Portfolio Balance Channel ...13
2.5 Liquidity Channel ...15
2.6 Collateral Channel ...15
3. Hypothesis ... 16
4. Data and Method ... 16
4.1 Data ...16
4.1 Model Specification ...18
4.2 Expectations regarding signs ...20
4.3 Descriptive Statistics ...22
4.4 Event Study ...23
4.5 Eicker-Huber-White Standard Errors ...24
4.6 Variance Inflation Factor ...24
4.7 Redundant Fixed Effect Test ...25
5. Analysis ... 25
5.1 Variance Inflation Factor Analysis ...25
5.2 Redundant Fixed Effects Test ...26
5.3 Results and Interpretation ...26
6. Limitations ... 33
7. Conclusion ... 35
8. References ... 37
Tables
Table I. Expected Sign ... 21
Table II. Descriptive Statistics ... 23
Table III. Regressions Output ... 32
Table IV. Policy Announcements ... 42
Table V. Variance Inflation Factor ... 44
Table VI. Redundant Fixed Effect Test. ... 44
Table VII. Individual Bank Descriptive Statistics... 45
Figures Figure I. Transmission Channels of Quantitative Easing ... 11
Figure II. Problem Loans to Gross Customer Loans ... 29
Abbreviations
QE Quantitative Easing
LIRE Low-Interest-Rate Environment
NIM Net Interest Margin
LSAP Large-Scale Asset Purchase
NIRP Negative Interest Rate Policy
EA Euro Area
CDS Credit Default Swap
1
1. Introduction
Due to the Covid-19 pandemic within many countries, the recent shutdowns have put major strains on the economy as businesses suffer from significant setbacks. Following crippling demand, elevated unemployment, and tighter trade conditions, the rate of bankruptcies among companies is rising. These conditions threaten the financial system's stability, which has led to the central banks around the world to adopt more accommodative monetary policies to safeguard the economy from a deepening recession (Altavilla et al., 2018). Quantitative easing has been at the forefront of these policies, with central banks globally committing to levels of bond purchases never seen before. The QE tool has been increasingly prevalent since Japan's initial QE attempt in the early 2000s due to the long period of deflationary pressure following the burst in the real estate bubble in 1991. While real estate bubbles on the magnitude of Japan are an uncommon sight, the most recent challenge for many countries is the low-interest-rate environment (LIRE) including Sweden our country of interest. The increase in the usage of both conventional and unconventional policies to bring the economy back to normal has created such an environment which has brought about a growing concern aimed at bank profitability as banks play a crucial role in the economy but even more so during financial distress (Allen & Carletti, 2019; Carletti & Ferrero, 2017).
In principle, QE typically aims to change the term structure by lowering the long-term interest rate, consequently flattening the yield curve.1 This should, in theory, induce bank lending and therefore
spur economic activity. Nevertheless, if long-term interest rates decline, the spread between lending rates and funding rates tightens, negatively affecting the net interest margin (NIM) that banks tend to primarily view profitability through. However, accommodative policies such as QE also influence the bank lending rates and stimulate borrowing, which should positively affect bank profitability through the decreased likelihood of credit default. In addition, banks benefit from a decline in the discount rate that increases their assets' value. Combined, the impact on bank profitability may be ambiguous as the magnitude, and the directional movements of the effects depend on the balance sheet composition of the bank's which is indicated by previous studies (Altavilla et al., 2018; Bikker & Vervliet, 2017; Lambert & Ueda, 2014; Scheiber et al., 2016).
2
Low profitability exposes the banks to adverse shocks and hinders the monetary policy transmission mechanism. The transmission mechanism is a vital process in affecting demand, liquidity, and the interest rate that determine the health of an economy. An obstruction in the transmission
mechanism can lead to an inability of banks to effectively supply the economy with sufficient credit that ultimately affects bondholders, depositors, and taxpayers (Altavilla et al., 2018). Considering the numerous channels through which QE impact the economy, following previous literature, we will primarily focus on four transmission channels that connect to interest rates and bank profitability, namely the signalling channel, liquidity channel, portfolio balance channel, and collateral channel (Krishnamurthy & Vissing-Jorgensen, 2011; Gagnon et al., 2011).
Firstly, the signalling channel allows the Riksbank to influence the expectations of market
participants by providing forward guidance; for instance, by publishing the planned repo rate path, interest rates expectations may stay anchored to the target rate.2 In addition, the Riksbank can show
commitment towards lower interest rates for the foreseeable future through large-scale asset purchases (LSAP) that are unconventional in nature, thus sending a credible signal that lower interest rates are here to stay (Alsterlind et al., 2015). Secondly, the liquidity channel allows the central bank to purchase bonds from commercial banks, increasing the liquidity in the banking system enables the bank to take on more risk, leading to a potential shift in the investment strategies of the banks. The third channel, the portfolio balance channel, allows the central bank to impact a broader part of the yield curve segment by relying on the imperfect asset substitutability assumption. As bonds differ in both maturity and credit risk, when the central bank reduces the supply of the preferred bond, investors will look for substitutes. This "contagion" effect affects the risk premia and bond yields on these substitute securities. Finally, the Riksbank can create a shortage of securities through purchase of a certain type of asset that is used as collateral by banks in the interbank market, which consequently lowers the yield on a said asset through supply and demand forces. This is because banks are willing to hold the bonds for a lower yield ensuring that they possess sufficient collateral. Therefore, through the collateral channel, the Riksbank can put downward pressure on short-term yields.
3
While the signalling channel may affect any part of the yield curve, the collateral channel mainly affects the short-term interest rates. This contrasts with that of the liquidity and portfolio balance channel which alters the longer-term interest rates. Collectively, the channels bring about new demand for loans due to favourable environments for the borrower, including reduced credit losses, translating into improved bank profitability.
In this paper we analyse the impact of QE on Swedish banks profitability. We focus our study on Sweden, a small open economy with prolonged periods of low-interest rates and negative interest rate policies (NIRP). Following previous literature, we use daily stock prices as a proxy for bank profitability as current market values reflect expected future profitability (Altavilla et al., 2018; De Rezende, 2017; Swanson, 2021). With the use of an event study regression approach, we capture and isolate QE effects and then investigate changes in the stock prices following a QE announcement made by the Riksbank. More specifically, since QE mainly operates on the long end of the yield curve, we construct a measure to isolate the effect of QE based on the forward rates of the five- and ten-year government bonds. As the Riksbank announcement contains multiple policies
simultaneously, such as QE and changes in repo rate, we establish a set of control variables for the other monetary policy instruments by capturing the short and mid-term segment of the yield curve. Similarly, we also introduce a set of control variables for changes in long-term foreign interest rates that affect Swedish banks profitability. We complement this study by decomposing QE impact on individual banks using interaction variables. Finally, in line with Altavilla et al. (2018) research, we use CDS to reflect the perceived riskiness of a bank. Therefore, we run an additional regression using CDS as the dependent variable to determine the robustness of our first regression and, more importantly, the impact QE has on debtholders, thus capturing the influence for all stakeholders of the commercial banks.
Our study contributes to the literature on unconventional monetary policy (UMP) as most previous studies are conducted in the US, Japan or the EA which primarily features large open economies with vastly different banking system. Moreover, there is a lot of ambiguity in previous studies on the QE effect on bank profitability partly due to researchers focus on the whole set of UMPs, which QE is only one of the many tools' central banks use. Finally, Sweden have implemented various UMPs and is the only Scandinavian country that has performed QE for an extended period of time.
4
In addition, it is a small open economy with different bank funding system making it unique amongst other small open economies.
We find that there is a clear positive relationship between QE and bank profitability as lower interest rates flatten the yield curve improving wholesale funding conditions. The discrepancies in our findings are attributed to Swedish banks being predominantly wholesale funded in contrast to EA and the US reliance on deposit funding. The implication is that structural differences in funding are an important factor to take into consideration when implementing QE. Second, banks with higher credit losses benefit more from QE due to a reduction in non-performing loans as borrower’s ability to pay their debts improves. Quantitative easing ceteris paribus may therefore be more impactful in countries where a large portion of banks experience significant credit losses on their balance sheets. The final contribution of our paper highlights that QE decreases CDS, signalling an improvement in wholesale funding condition and reduced credit risks. Our findings point to the fact that QE policy benefits both debtholder and bank shareholders which is of importance since commercial banks largely fund their operations via debt.
1.1 Problem statement
In the current literature, only a few studies examine the relationship between QE and bank profitability. Even fewer papers address bank profitability in Sweden, and these typically focus on the effect of a low-interest-rate environment on profitability. Although, research on UMP has been rising in recent times, most papers focus on macroeconomics issues, for instance, the interplay between QE and transmission channels (Lambert & Ueda, 2014). In addition, most research on the effect of QE on bank profitability has been conducted in the US, where the overall findings
conclude a negative relationship as the NIM compress. Therefore, it is not likely that the findings in countries as the US and Japan will be relevant to Sweden as these countries are large open
economies with vastly different policy and banking system.
Perhaps the best candidate to compare studies on QE with is the Euro Area (EA), but even then, there are stark differences in the way the banks' balance sheet is comprised. Swedish banks tend to have lower credit losses than their European counterparts, and the banks differ in their method of
5
obtaining financing, with the former favouring wholesale funding and the latter deposit funding (De Rezende & Laséen, 2018; Madaschi & Nuevo, 2017).3 This may lead to a significant difference in the
way QE effect bank profitability as wholesale funding fluctuates a lot, in particular during financial instability. Even amongst the EA, the findings differ; Altavilla et al. (2018) find a positive
relationship between QE and bank profitability in part due to the improvement in macroeconomic condition, in contrast, Demetriz and Wolff (2016) argue that bank profitability decline following QE since it compresses NIM. Since there are apparent discrepancies in the findings and the funding method differ, we must question the appropriateness of previous research for our study.
It is important to understand the innerworkings of QE impact on bank profitability since it is one of the most widely used tools during financial stress. Furthermore, as banks business model relies on asset transformation, bank profitability plays a large role for financial stability. However, the gap in literature allows us to investigate the QE effect on bank profitability in a small open economy with a negative interest rate policy (NIRP), namely, Sweden. To our knowledge, we are the first to provide insight into how QE affects Swedish banks' profitability.
1.2 Purpose
This study aims to address the gap in previous literature regarding the effects of QE on Swedish bank profitability. With the recent increase in coverage of QE in media and the prevalence of unconventional policy by central banks around the world, the authors of this study find our research to be timely (Winck, 2020). Therefore, to investigate the relationship between QE and bank
profitability, we formulate the following question:
Does quantitative easing by the Riksbank increase bank profitability in Sweden?
This study will give further evidence on the existence of a relationship between QE and bank profitability, thus adding to the already scarce pool of literature on QE in Sweden. Furthermore, policymakers looking to understand the benefits and consequences will now better understand the implication for financial stability when considering new QE policies. This study is also of interest to investors attempting to adjust their trading strategies as they will better comprehend QE impact on 3 Wholesale funding is a way for banks to fund their operations through issuances of various securities to financial
6
bank stock prices which allows them to position themselves accordingly. Finally, we hope that our study can serve as a starting point for future research as we are among the first to study QE effect on bank profitability in Sweden.
1.3 Previous Research
For the analysis of UMP's effect on bank profitability, a clear majority of research is conducted within the US. Lambert and Ueda (2014) establish either a negative or ambiguous relationship between UMP and bank profitability. These findings were discovered by studying the UMP measured by the change in the ratio of central bank assets to GDP and its effect on profitability reflected by three different ratios for bank profit. Mamatzakis and Bermpei (2016) building on the previous research (Lambert & Ueda, 2014; Montecino & Epstein, 2014), propose an alternative approach by accounting for the deposit insurance. In contrast, they found that there is a significant negative relationship between UMP and bank performance. Furthermore, banks with low levels of asset diversification and a greater reliance on deposit funding were unable to mitigate the extent of the negative impact from UMP and experienced more significant losses. In addition, the adverse impact of UMP is more prominent under higher levels of asset purchases.
In the EA, the overall effect of UMP on bank profitability is mostly positive, according to the prevailing literature within the field. Altavilla et al. (2018) establish a positive relationship by constructing a regression using daily financial data to analyse changes within bank stock prices following an announcement of UMP. However, since UMP is a mixture of policies, including QE, they did not account for the isolated effects of QE. This can be seen by observing the chosen announcements and the choice of regression which is constructed to capture short-term and long-term effects on the yield curve that is associated with UMP. Therefore, there is a high likelihood that other instruments of monetary policy such as forward guidance has an impact on their results as their primary purpose is to capture the effect of UMP on bank profitability rather than QE itself.
Focusing only on the effect of QE on bank stock prices which is used as a measure for profitability, Philippe et al. (2016) discovered that QE had an overall positive impact on bank profitability. They reasoned that appreciation of sovereign bonds on the bank's balance sheets outweighs the negative
7
impact of a flattened yield curve caused by the change in the long-term structure from QE.
Complementing these findings, Varghese and Zhang (2018) found an overall positive impact of QE on bank profitability but with weak statistical significance in the EA due to a partially offsetting effect through a reduction in the interest margins caused by NIRP and prolonged LIRE.
Interestingly, both studies found that expected inflation rose, which might suggest that QE leads to improvements in economic activity and growth.
In contrast, the QE impact on banks in the paper by Demertiz and Wolff (2016) using trend analysis suggest that if deposit rates are close to zero, bank margins will be squeezed, thus QE harm bank profitability. However, these relationships might not necessarily hold for Sweden as Swedish banks' funding consists primarily of wholesale funding instead of deposit funding for the EA banks (De Rezende & Laséen, 2018). When interest rates fall, the lending rate decreases. Furthermore, the asymmetry in decline between the funding rate (FR) and the deposit rate (DR), with the former decreasing more, results in a squeeze in the interest margin. The reason is that because the interest rate on funding, in this case, deposits are already low, banks are reluctant to reduce deposit rates to negative territory fearing liquidity shortage. Consequently, bank profitability is hurt as only the lending rate fall. The effect this has on bank profitability largely dependents on the proportion of banks liabilities that are deposit funded (Riksbank, 2016) – We illustrate a simplified equation of this relationship below.4
𝑆𝑤𝑒𝑑𝑖𝑠ℎ 𝑏𝑎𝑛𝑘𝑠: ↓ 𝐿𝑅 − ↓ 𝐹𝑅 = 𝑁𝑜 𝑐ℎ𝑎𝑛𝑔𝑒 𝐸𝐴 𝑏𝑎𝑛𝑘𝑠: ↓ 𝐿𝑅 − 𝐷𝑅 = 𝐿𝑜𝑠𝑠
Considering the scarcity in the literature on QE studies, investigating the previous literature on the effects of NIRP on Swedish bank stock prices may yield promising results (De Rezende & Laséen, 2018; Eggertson et al., 2019). While NIRP and QE vary in their transmission, they share the similarity of reducing interest rates. Moreover, because of their resemblance, central banks tend to implement them in tandem to affect the economy more significantly.
4 LR: Lending rate is the rate banks charge for lending out money for instance to households and corporations.
FR: The rates banks are charged when borrowing for instance issuance of bonds and the deposit rate. DR: The deposit rate is the rate banks pay to investors depositing money at the bank.
8
Most of the literature concerning negative interest rates within Sweden argue that there is a positive impact on bank profitability. For instance, Madaschi and Nuevo (2017) discovered that bank profitability had improved by looking at changes within Swedish bank data during the time of negative interest rates. They found that the improvement stems from two parts. Firstly, the NIM stayed constant caused by positive lending growth and a reduced interest expense due to lower wholesale funding cost that comes from the fact that these costs hovered closely around the monetary policy rate, which was below zero. Furthermore, the increase in realized and unrealized securities gains is the second factor brought about by reductions in interest rate, which appreciated the value of the banks' holdings of debt securities.
However, their empirical analysis comes into question due to granularity in the data, and no
econometric analysis was implemented. De Rezende and Laséen (2018) address these issues using a non-linear local projection method. They conclude that the implementation of NIRP increased stock prices; in other words, lower interest rates boost bank profitability. Hence, their findings
complement the conclusion by Madaschi and Nuevo (2017). Furthermore, the positive impact of lower interest rates following NIRP may therefore suggest that outright QE leads to the same outcome on bank equity price.
1.4 Disposition
This study is organized in the following manner: Section 2 goes into detail on the theory behind QE in Sweden, how the transmission channels work and how it impacts bank profitability. Section 3 illustrates our a priori expectations for the results and the hypothesis of our thesis. The start of section 4 provides the data used, the chosen model for our regressions and descriptive statistics for our panel data while the latter part describes the methodology of our model and the application necessary for consistent results. Section 5 present the analysis, interpretation, and result from our regressions. Section 6 offers limitations of our study. Section 7 consists of the conclusion and suggestions for further research. We conclude our paper with a list of references and an appendix where formulas and data presented in more detail.
9
2. Theory
2.1 Quantitative easing in Sweden
The financial crisis of 2008 left many economies in disarray, and Sweden was no exception. By late 2008 the annual real gross domestic product had fallen by 15%, primarily due to falling demand in the export sector. Unlike other countries such as Japan, England, and the USA, Sweden did not initiate QE in 2008 but favoured credit easing (CE). The distinction between QE and CE can be somewhat blurry due to the similarity in the objective, namely, to affect interest rates. Credit easing policy seeks to increase funds in specific sectors, in contrast to QE that aims to bring liquidity to the whole economy via banks (Anderson, 2012).
In July 2009, the Riksbank announced it would initiate its version of QE by auctioning large amounts of fixed-rate low-interest loans instead of the standard large-scale asset purchase (LSAP). This version of QE, while unorthodox, proved effective in reducing long-term interest rates by 0.2% to 0.4% percent (Anderson, 2012). The policy's initial success was short-lived as the financial crisis brought negative spillover effects from the EA. In the end, the Riksbank had no other choice than to respond to the deflationary pressure by lowering both its target and the repo rate.
October 2014 marked a historic event as interest rates hitting the zero-lower bound. Until this date, as many other countries in the wake of the financial crisis Sweden found itself in a prolonged period of the low-interest-rate environment (LIRE). The current situation was reminiscent of that of Japan, and with rising concerns of a liquidity trap, the Riksbank decided to act swiftly. Shortly after that, at the beginning of February 2015, following the European Central Bank (ECB) announcement of QE earlier in January, Sweden initiated its own QE by purchasing large quantities of government bonds. In contrast to other countries, increasing liquidity was not the main priority for the Riksbank; instead, the focus was on stimulating the economy by lowering interest rates through various markets and subsequently boosting bank lending to get back to the inflation target of 2%.
Quantitative easing also played an important role in halting the appreciation of the Swedish Krona that had been a going concern. At the same time, a NIRP was employed, and the repo rate was cut down to -0.10%. Later the same year, it decreased further to -0.35%, signalling the Riksbank willingness to pursue expansionary policy (De Rezende, 2017).
10
While the Riksbank has used both types of unconventional monetary policy and they share similarities that makes it difficult to distinguish the relative effect of the two. One thing that sets apart from QE is the difference in the transmission channels used since NIRP involves the cut the repo rate and not purchases of specific assets; for instance, the collateral channel will not be present when NIRP is performed. Moreover, with QE, the Riksbank can target a certain asset that is
especially beneficial during distress as these securities will be in low demand. Previous research also shows that QE is more potent at effecting prices on long-term securities, which may give some inclination towards the preference of the Riksbank use in unconventional monetary policy
(Dell’Ariccia et al., 2017).5 Recent work on the Riksbank QE success has been researched by several
scholar (Casola & Stockhammar, 2021; De Rezende & Ristiniemi, 2020; De Rezende, 2017;
Eidestedt et al, 2020; Grimaldi et al, 2021; Gustafsson & Brömsen, 2021; Melander, 2021; Shamloo & Diez de los Rios, 2017). The findings by De Rezende et al. (2015) point to QE's success in that purchases of bonds did indeed lower bond yields, and the Swedish Krona experienced more significant depreciation than in the absence of QE.
2.2 Transmission Channels
The Riksbank conducts QE to stimulate economic activity by purchasing securities from financial institutions, among them banks, with the incentive that changes in the interest rate through the various transmission channels will spur lending, inflation, and growth within the economy.6 The
following image illustrates a simplified overview of the essential channels that translate interest rates into profitability for banks:
5 The use of UMP is mostly on a case-by-case basis as such for more insight on the preferences of the Riksbank we
recommend reading the speech given by Stefan Ingves, governor of the Sveriges Riksbank (Riksbank, 2020). One can also read the Accounts of Monetary Policy Report and Monetary Policy Report.
6 The transmission channels for the banks are the same but the effect may differ depending on asset composition among
11
Figure I. Transmission Channels of Quantitative Easing
Note: Adapted from ‘’The Central Bank Balance Sheet as a Policy Tool: Past, Present and Future’’ by A. J. Bailey., J. Bridges., R. Harrison., J. Jones., & A. Mankodi, (2020), Bank of England Working Paper, 2020(899), p.8. Copyright 2020 by Bank of England.
12
2.3 Signalling Channel
The signalling channel or the "confidence channel" encompasses any information about the future monetary policy regarding asset purchases that economic agent can learn. Piazzesi and Swanson (2008) showed that we should expect all interest rates across maturities to be affected through the signalling channel via the expectation hypothesis when changing interest rates.7 Research by De
Rezende (2017) reaffirms this assertion and further suggests that alterations in the short-rate
expectations primarily affect the intermediate part of the yield curve in Sweden. For policy decisions, it implies that the Riksbank can influence the interest rate of the whole yield curve through UMP but in particular the mid-segment via the signalling channel.
In many instances, large asset purchases have "guided" market participants by showing the level of determination a central bank has towards policy; if perceived credible, it can decrease long-term bond yields (Joyce et al., 2011). The unconventional nature of QE can signal that a central bank is committed to keeping interest low for a more extended period and serves as an indicator of the economy's state. Moreover, by announcing that interest rates will be kept low central banks can use forward guidance to influence the expectations of market participants. A low-interest-rate
environment, in turn, affects the credit channel.8 When the discount rate decreases, the present value
of real assets and demand rises. For instance, taking out a loan and investing in real estate becomes relatively more attractive. The change in the interest rate increases bank lending and subsequently bank profitability in two ways; firstly, these assets are used as collateral for the banks, when asset value increases the bank alters its lending behaviour making it less restrictive, increasing the number of loans. Secondly, demand for loans increases as the costs of capital fall for businesses, increasing their profitability and stimulating growth. What is more, banks will seek out more profitable
investments than bonds and stocks in the form of lending to companies and households (Hopkins et al., 2009).
7 The expectation hypothesis assumes that future short-term rates are predicated on present long-term interest rates. 8 The credit channel is influenced by monetary policy and determine the amount of credit issued to the economy.
13
2.4 Portfolio Balance Channel
According to a speech given by Charles Bean, a former deputy for monetary policy at the Bank of England, policymakers emphasize the portfolio balance channel's role as a vital component in transmitting asset purchases to the broader economy (Bank of England, 2011). Essentially it works through the assumption of imperfect substitutability; that is, assets do not have an identical
substitute (Tobin, 1958). The existence of such a channel within Sweden is supported by De
Rezende (2017), who found through the application of a dynamic term structure model coupled with an event study regression that the term premia of longer maturities were effectively lowered through the portfolio balance channel. When the central bank buys a specific security, for instance, bonds, they lower the yield on these securities through adjustments in demand and supply. As the reader might predict, holders of these assets try to rebalance their portfolios by finding assets with similar credit risk and duration. We, therefore, have two subsections of the portfolio balance channel, namely the local supply channel and duration risk channel.
Local Supply channel
The reduced supply of a particular bond purchased by the central bank creates scarcity, leading to upward pressure on the price of that bond. Habitual investors whose preferences are inclined towards a specific bond, commonly long-term bonds, will increase the demand for nearby
substitutes in their search for assets with similar yield and maturity.9 This process of lowering yields
on substitutes is known as the contagion effect. It stands to reason that the central bank can target specific bonds and a broader segment of the yield curve through contagion via the local supply channel.
The Duration Risk Channel
The mechanics of the duration risk channel are based on the fact that bonds with longer duration are more sensitive to interest rate changes. Thus, investors demand a term premium for taking the extra risk of holding long-term assets relative to short-term assets, that is, for the added duration risk. 9 Habitual investors are based on the preferred habitat theory which states that investors have a certain orientation to
bonds of specific maturity and that the risk premium is what drives the decision to shift investments into a different bond.
14
Large-scale purchases may lower the aggregate duration risk and, as a result, the risk premium for investors by removing long-term securities. In principle, only the preferred-habitat investors will be left holding these assets as they are willing to forego the premium for alternative motives, for instance, retirement savings. Alternatively, some investors will concede some of the premia as their portfolio will now consist of relatively lower amounts of duration risk (Gagnon et al., 2018).
Since long-term bonds are among the bond maturities purchased under QE, it effectively reduces the outstanding volume of said assets and increases the competition for the remaining bonds. For this reason, arbitrage investors will look for other asset classes that can work as close substitutes to the bonds purchased by the central bank. These effects lead to altercations within the yield curve; more specifically, it leads to a decrease in the long-term to short-term relative bond yield
(Krishnamurthy & Vissing-Jorgensen, 2011). Accordingly, the price increase and the yield decrease depend on investors' price elasticity pertaining to their preferred assets.
Bank profitability
Since QE predominantly involves purchases of long-term maturity bonds, buying large quantities of those assets creates a contagion effect that spread to nearby substitutes decreasing their yields. The extent of contagion and the impact of QE on bank profitability is mainly determined by investors who decide to sell their assets. On the one hand, if they reinvest their money in close substitutes, the so-called rebalancing effect will be substantial carrying through to the broader economy in the form of lower interest rates. On the other hand, investors may decide to shift their investment portfolio to different securities such as corporate bonds and mortgages (Alsterlind et al., 2015). The changes to investors' portfolios should ultimately lead to economic growth due to a more favourable lending environment and a rise in bank profitability (Park et al., 2020).
Since Swedish banks rely on covered bonds to fund 70% of their mortgages, a shift in investor demand towards such assets would ceteris paribus increase bank profitability. Moreover, banks themselves also attempt to change their balance sheets as the term premium on long-term bonds fall through the issuance of new riskier loans. Previous findings from the US support the idea of shifts within holdings of investors when the FED buys government bonds (Goldstein et al., 2018).
15
Similarly, this behaviour applies to insurance and fund management companies as they alter their investment portfolio in response to QE (Joyce et al., 2014).
2.5 Liquidity Channel
The purpose of the liquidity channel is simple: to increase liquidity in the banking system to
incentivize bank lending. When the Riksbank decides to purchase bonds, the bond seller, namely the commercial bank, will increase the volume of balance reserves on their asset side and compress the difference in the liquidity premia between illiquid and liquid assets (Krishnamurthy & Vissing-Jorgensen, 2011).
The so-called balance reserves are more liquid than long-term securities, therefore reducing the individual risk banks face in the interbank market. In other words, the likelihood that a bank is finding itself in an excessive deficit on its account is reduced. As a result of QE, the yield curve flattens as the liquidity premium on bonds fall, and with the overall increase in liquidity, banks can afford to lend out more money (Alsterlind et al., 2015). In addition, when liquidity increases, the aggregate duration of the portfolio of the bank system fall. The decline is due to the requirement for banks to deposit their excess liquidity at the central bank.
To compensate, similarly to the portfolio balance channel, for the lower duration, banks start purchasing securities with long-term duration in what is called the reserve induced effect
(Christensen & Krogstrup, 2019). Quantitative easing can therefore also cause changes in the long-term yield through the reserve-induced effect as the increase in demand decrease interest rates. Importantly the liquidity channel is more prevalent in times of financial distress.
2.6 Collateral Channel
Banks often borrow from one another in the interbank market to fund their operations. In the process of borrowing from other banks, they must have some collateral. One type of collateral is the liquid short-term government bond which becomes highly sought after. Accordingly, if central banks
16
decide to purchase these types of bonds, they will have to pay a premium on these assets in order for banks to let go of their collateral.
These purchases will lead to scarcity in the interbank market, resulting in lenders accepting a lower yield as the short-term asset has a fundamental value working as collateral in the interbank market.10
Hence, through QE, the central bank can affect the short-term end of the yield curve via this collateral channel. If central banks decide to lower short-term yields, the banks may therefore be more profitable as credit conditions improve, leading to an increase in credit issued to the economy (Melander, 2021).
3. Hypothesis
This study aims to investigate if a link exists between stock price and quantitative easing on the announcement day of such a policy. Our a priori expectations based on previous findings in the EA is that a flattening of the yield curve will stimulate borrowing (Altavilla et al., 2018; Demertiz & Wolff, 2016; Philippe et al., 2016; Varghese & Zhang, 2018). Therefore, we should expect
comparable results; explicitly, there should be a positive relationship between quantitative easing and bank profitability. With this, the following hypothesis is formed:
𝐻ₒ: 𝑇ℎ𝑒𝑟𝑒 𝑒𝑥𝑖𝑠𝑡𝑠 𝑛𝑜 𝑠𝑖𝑔𝑛𝑖𝑓𝑖𝑐𝑎𝑛𝑡 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 𝑟𝑒𝑙𝑎𝑡𝑖𝑜𝑛𝑠ℎ𝑖𝑝 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑞𝑢𝑎𝑛𝑡𝑖𝑡𝑎𝑡𝑖𝑣𝑒 𝑒𝑎𝑠𝑖𝑛𝑔 𝑎𝑛𝑑 𝑏𝑎𝑛𝑘 𝑝𝑟𝑜𝑓𝑖𝑡𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝐻1: 𝑇ℎ𝑒𝑟𝑒 𝑒𝑥𝑖𝑠𝑡 𝑎 𝑠𝑖𝑔𝑛𝑖𝑓𝑖𝑐𝑎𝑛𝑡 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 𝑟𝑒𝑙𝑎𝑡𝑖𝑜𝑛𝑠ℎ𝑖𝑝 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑞𝑢𝑎𝑛𝑡𝑖𝑡𝑎𝑡𝑖𝑣𝑒 𝑒𝑎𝑠𝑖𝑛𝑔 𝑎𝑛𝑑 𝑏𝑎𝑛𝑘 𝑝𝑟𝑜𝑓𝑖𝑡𝑎𝑏𝑖𝑙𝑖𝑡𝑦
4. Data and Method
4.1 Data
In this study, data is collected to construct an event study using a panel data approach by observing Swedish banks examined between the period 2015-2021. Our time series sample builds upon previous research done by De Rezende and Ristiniemi (2020). They observed important monetary policy announcements made by the Riksbank from the introduction of QE policy on February 12,
17
2015, to April 24, 2020, totalling 38 announcements during the period. Furthermore, starting from where their previous research ended in April, we observed 16 additional policy announcements up to March 15, 2021; hence our time series extends to 54 observations in total. A short sample of the announcements is given in Table IV.11
We use the closing stock and credit default swap prices of the banks collected from Yahoo Finance and Refinitiv Eikon, respectively. Data for the graphs were obtained through S&P global and Refinitiv Eikon databases. The cross-sectional data consists of 4 Swedish banks namely: Nordea (NDA), Svenska Handelsbanken (SHB), Skandinaviska Enskilda Banken (SEB) and Swedbank (SWED) together represent around 70% of the entire Swedish credit market.12 As we capture a large
part of the banking system and the banking sector in Sweden has the value of 16 in the HHI index which is on the average compared to other banking sectors in the European Union for instance Netherlands, Germany and Denmark (Næss-Schmidt et al., 2019). We determine that the data is sufficient as a representative of the banking system. Therefore, it is no immediate concern that competition is distorted which could affect the transmissions channel and subsequently our findings (Kim et al., 2021).
The banks were chosen based on the criteria that it should be listed on the stock market by the start of QE in February 2015. The condition is essential to make the sample consistent across time between the banks and ensure that all announcement dates are included from the initiation of QE. Data on the bond yields for Sweden, Europe, and the United Kingdom was extracted through the respective central bank of each country, while the United States bond yields were gathered from the treasury. Three observations on the announcement day were missing, one on bond yields and another for the treasury bill. The US market was closed on the 4th of July 2017 since it was a federal holiday. Similarly, on the 8th of May 2020, there was a holiday in the UK. For the data on Swedish T-bill, due to lack of trading frequency in the exchanges, there was no rights issue from August to October in 2018. As a result of our missing data, the panel becomes unbalanced.
11 The increase in the amounts of announcements relative to previous year is due to the COVID-19 pandemic.
12 The specific share that each bank hold of the Swedish credit market follows: NDA 12%, SHB 21%, SEB 14%, SWED
18
Table IV. Monetary policy announcements by the Riksbank
Mar 16, 2020 Riksbank to offer increased loans to banks on favourable terms. During 2020, it also intends to buy government, municipal and mortgage bonds for up to an additional SEK 300 billion
Mar 19, 2020 Riksbank will enable loans in US dollars to ensure its continued supply for Swedish companies. Limit rules for mortgage bonds to be used as collateral are removed to facilitate lending to banks. Riksbank also intends to buy commercial paper within its SEK 300 billion asset purchase program Mar 20, 2020 In addition, to government, municipal and covered bonds, Riksbank to purchase also securities issued by
nonfinancial corporations within its SEK 300 billion asset purchase program
4.1 Model Specification
Using Swanson (2021) and Altavilla et al. (2019) as a blueprint, we create a high-frequency regression to determine the significance of QE announcement on Swedish banks' stock prices. As such, the following model panel data is constructed:
𝑦𝑖,𝑡 = 𝛼 + 𝛽1∆𝑄𝐸𝑖𝑡+ 𝛽2∆𝑟𝑡∗𝑢 + 𝛽3∆∅𝑖𝑡+ 𝛽4∆𝐸𝐶𝐵𝑖𝑡+ 𝛽5∆𝑈𝐾𝑖𝑡+ 𝛽6∆𝑈𝑆𝑖𝑡+ 𝜀𝑖𝑡 𝑖 = 1,2, … ,4
𝑡 = 1,2 … ,54
In the model above 𝑦𝑖,𝑡 indicates the particular percentage change in stock prices or CDS spread of
the ith bank and t indexes time of QE announcement, α is the intercept, 𝑄𝐸𝑖𝑡 is the variable of
interest displaying our measure of QE. We apply a set of control variables; 𝑟𝑡∗𝑢 denotes the
short-term surprise, ∅𝑖𝑡 represents 6-month to 2-year T-bill forward rates and 𝐸𝐶𝐵𝑖𝑡, 𝑈𝐾𝑖𝑡 and 𝑈𝑆𝑖𝑡
exhibit the particular long-term forward rate for respective area, lastly 𝜀𝑖𝑡denotes the error term.
Approximating the underlying regression presented by Swanson (2021) and drawing inspiration for the explanatory variables presented by De Rezende (2017), we incorporate these concepts into our model to form the yield curve consisting of three segments: the short-term, mid-term, and long-term segments. Using this approach, we can control for other instruments of monetary policy such as cuts in the repo rate and forward guidance employed by the Riksbank.
19
Beginning with our dependent variable, we use bank stock price as it reflects future bank
profitability. Moreover, we use the percentage change rather than the absolute change as the size of the banks matter for equity prices. Considering, the purpose of QE is to lower long-term interest rates, we can capture the effect of QE shock in 𝛽1. We then isolate the impact of QE to avoid biases
by controlling for different segments of the curve as outlined previously. The long end segment of the yield curve is constructed using forward rates with the 5- and 10-year government bonds.13 The
short-run surprise controls for the immediate shock of QE announcements by using STINA 1-Month rates (De Rezende, 2017). This represents the change in the short-term portion of the yield curve. To control for the effects of forward guidance, we encapsulate the intermediate section of the curve by deriving a forward rate from the collation of 6-month and 2-year government bonds, and ECB are computed using 5- and 10-year government bond data from the countries to control for fluctuations in foreign long-term interest rates that may affect Swedish banks.
Finally, we want to investigate how QE impacts banks accounting for individual characteristics. To do this, we create three dummy variables and let SHB be the reference group therefore bypassing the dummy variable trap, we then allow each dummy variable to interact with our QE variable to determine if there are differences in the effect on QE on each using the following equation:14
𝑦𝑖,𝑡 = 𝛼 + 𝛽1∆𝑄𝐸𝑖𝑡+ 𝛽2∆𝑟𝑖𝑡∗𝑢 + 𝛽3∆∅𝑖𝑡+ 𝛽4∆𝐸𝐶𝐵𝑖𝑡+ 𝛽5∆𝑈𝐾𝑖𝑡+ 𝛽6∆𝑈𝑆𝑖𝑡+ 𝛽5∆𝑄𝐸𝑖𝑡𝐷𝑁𝐷𝐴
+ 𝛽6∆𝑄𝐸𝑖𝑡𝐷𝑆𝐸𝐵+ 𝛽7∆𝑄𝐸𝑖𝑡𝐷𝑆𝑊𝐸𝐷+ 𝜀𝑖𝑡
Hitherto, our study consists of stock prices as a proxy that captures the perceived future bank profitability by market participants. Assuming the efficient market hypothesis holds, all available information about a stock should immediately be reflected within its price, in other words when market participants receive new information, they should react by incorporating this news into the stock price. This assumption allows us to capture the effect of monetary policy around the given 1-day event window which is a necessity for our study (Bernanke & Kuttner, 2005). Therefore, the usage of high-frequency data should make intuitive sense since if the efficient market hypothesis does not hold then the market would not react “efficient” following a new announcement and we would
13 In the calculation of the forward rates, we used the formula proposed by Gurkaynak et al. (2004). The formula can be
found in the appendix.
20
be unable to capture the causal relationship between QE and bank profitability (Gürkaynak et al. 2005b).
While the share price is important for shareholders, the equity of European banks makes up a meagre 5% of the total amount of banks' assets. At the same time, the level of debt is a lot more significant as bank operations tend to be reliant on debt financing (Altavilla et al., 2018). Since this is also true for Swedish banks, we follow Altavilla et al. (2018) measure of CDS, which allow us to gauge investors' sentiment regarding the bank's credit risk. Therefore, we run a second identical regression to highlight the impact of QE on CDS and its implication for debtholders; in addition, it will also serve as proof of robustness.
4.2 Expectations regarding signs
The expectations are mostly straightforward as this study is based on the idea that QE lowers long-term interest rates leading to a higher value in stock prices. We postulate that our variable will be negative. As the Riksbank implement QE and lowers interest rates, banks' stock prices should increase, and therefore we have a negative relationship which should make sense intuitively. Regarding ∆𝑟𝑡∗𝑢 we also expect the sign to be negative as decreasing short-term interest rates will
make the yield curve steeper, which is in line with previous studies (Altavilla et al., 2018). Banks will therefore profit when the interest margin increase between short-term borrowing and long-term lending.
For the ∆∅𝑖𝑡 we should expect the sign to be either positive or negative as changes in the midterm
yield may be heavily influenced by the balance sheet composition. For the foreign country's coefficient, we apply similar reasoning but with some added complexities. There is a lot more ambiguity regarding how foreign long-term interest rates affect Swedish banks' stock prices. For instance, through foreign exchange rates and the number of securities denominated in foreign currency, the result could be positive or negative depending on the securities held by the bank.
21
As QE yet again lower interest rate, it improves the lending environment reducing credit losses, therefore, as CDS is a measure of the perceived risk of default, falling interest rates will decrease CDS spreads. We expect there to be a negative relationship between QE and CDS. Thus, the QE coefficient for the CDS regression should have the opposite sign to the one used to measure stock prices, that is, positive. In fact, the two regressions should have the opposite signs on all coefficients. If this is indeed the case, it will be a sign of robustness.
Table I. Expected signs
Variables Regression Notation Expected Sign
Stock
Expected Sign CDs
Dependent variables
Stock Prices Credit Default Swap
Independent variables
𝑦𝑖,𝑡
Long-term rate change ∆𝑄𝐸𝑖𝑡 - +
Short-Term rate surprise ∆𝑟𝑡∗𝑢 - +
Medium-term rate change ∆∅𝑖𝑡 +/- +/-
Effects from ECB ∆𝐸𝐶𝐵𝑖𝑡 +/- +/-
Effects from UK ∆𝑈𝐾𝑖𝑡 +/- +/-
22
4.3 Descriptive Statistics
In this brief section, we describe the characteristics of our regression. We have obtained 204 observations on four Swedish commercial banks listed on the OMX Stockholm. Descriptive statistics have been created to give the reader an idea of the minimum, maximum, mean, and standard deviation values of each variable in our regression.
There are a couple of interesting features in this table. The largest and smallest value change in stock returns 611.715 and -740.7408 respectively occurred in 2020, both of which pertained to
Handelsbanken in March.15 Moreover, in the case of CDS, in March of 2020 the smallest value
change was found to be -10290 which represented Handelsbanken while the largest value was recorded in December of 2018 for Swedbank. The abnormal change in value for both stock price and CDS of Handelsbanken could be explained by the sudden shift in volatility that appeared in the financial markets around that time, including, extraordinary measures taken by the Riksbank to stabilize the economy.
Finally, the largest change in value of the CDS happened in December 2018 for Swedbank which could be explained by the decision to raise interest rates which may have induced significant uncertainty as this was the first time ever that the Riksbank had decided to increase interest rates within a negative environment since lowering it into that territory in 2015. The remaining changes for the individual banks can be viewed in Table II below.
23
Table II. Descriptive statistics
Variables Minimum value Maximum value Mean Standard deviation 𝑆𝑡𝑜𝑐𝑘 𝑃𝑟𝑖𝑐𝑒𝑖𝑡 -740.7408 611.715 12.70567 189.8985 𝐶𝐷𝑆𝑖𝑡 -10290 1546.155 -29.53209 775.9034 ∆𝑄𝐸𝑖𝑡 -20.3 21.2 0.230392 7.100748 ∆𝑟𝑡∗𝑢 -14.5 7.3 -0.499782 3.57122 ∆∅𝑖𝑡 -16.5 10.8 -0.155021 5.008477 ∆𝐸𝐶𝐵𝑖𝑡 -14.24435 17.63148 0.978974 5.166231 ∆𝑈𝐾𝑖𝑡 -2.9108 1.54185 0.102153 0.755747 ∆𝑈𝑆𝑖𝑡 -3.38 2.34 -0.02549 0.925474
Note: The Stock price and CDS is measured as a percentage change (𝑡 − 𝑡−1)/(t−1) and then expressed in basis points. All values are expressed in
basis points unless stated otherwise. ∆ denotes (𝑡 − 𝑡−1)
4.4 Event Study
The usage of event studies is a popular approach to investigate the effect of an economic event on the firms' value. It is common for researchers in finance to use an event study to observe the change in stock price due to an announcement of a merger and acquisition (Eckbo, 2008). In our research, we are interested in the behaviour of stock prices as an indication of the profitability of Swedish banks to the announcement of QE made by the Riksbank. The benefits of an event study, assuming rationality in the market, is that an announcement of QE should affect the long end of the yield curve, namely the 5-to-10-year bonds. Moreover, prior research trying to estimate the effect of QE involved the use of event study methods (Bauer & Neely, 2012; Bauer & Rudebusch, 2011; Rosa, 2012), as it allows for the capture of changes in a given time window such as changes in 10-year yields which could be observed following monetary policy announcements (Gagnon, 2016).
24
To perform an event study, we make use of the framework outlaid by Mackinlay (1997). We
determine our event window to be 1-day in conformity with previous literature (Altavilla et al., 2018; Gagnon et al., 2011; Swanson, 2021) and define the event of interest in our study announcements containing news about QE from the Riksbank. It is common for the Riksbank to bundle its monetary news in contrast to other central banks. For instance, on the 27th of October 2017 announcement, the repo rate would remain at -0.50%, and that QE purchases would continue. In contrast, on the 17th of November, the central bank announced corporate bond purchases. For our study, both announcements contain QE information and are therefore regarded as an event.
However, we acknowledge that there are potential issues when multiple announcements are made. Thus, we will elaborate on this further in the limitation section of our thesis.
4.5 Eicker-Huber-White Standard Errors
In the use of panel data, it is possible to have a heteroscedasticity problem, in short, the disturbances 𝑢𝑖 variance is not consistent across the observations. Commonly, cross-sectional analysis encounters
difficulties since the members of the population differ in magnitude, in our case, the size of each bank (Gujarati & Porter, 2008). Consequently, this may affect the variance of each observation. Suppose our regression is, in fact, suffering from heteroscedasticity. In that case, it will mean that the estimators in the standard OLS will not be the best linear unbiased estimator (BLUE), in other words, the variance of the estimator is not the minimum one and therefore is inefficient. We correct these inefficiencies using a common approach among financial research, namely, the Eicker-Huber-White standard errors (King & Roberts, 2015). The white standard errors allow us to transform the data with no heteroskedasticity by producing more precise estimates of the true variance of the parameters replacing the unknown variance with a squared residual for each observation.
4.6 Variance Inflation Factor
Since our regression consists of multiple explanatory variables and we work with time-series data, there might be an incidence of intercorrelations. The presence of multicollinearity may impact the significance of our tests. In fact, when working with time, it is not too uncommon for time series data to share a common trend over time. According to economic theory, we suspect that there is collinearity between the independent variables as monetary policy tends to have a cross border
25
effect. Due to interconnectedness, it is quite possible that, for example, UK and ECB share similar trends in interest rate movements through shared monetary policy. To determine the degree of multicollinearity, we will apply the variance inflation factor test, indicating the level of correlations between all independent variables. Variance inflation factor values exceeding 10 indicates the presence of high collinearity (Gujarati & Porter, 2008).
4.7 Redundant Fixed Effect Test
When working with panel data, a decision must be made which regression model to use: fixed effect model (FEM), random effects model (REM) or the pooled OLS. Since our regression contains only data on four cross-sections and we want to estimate six independent variables, we cannot use the REM model as the number of cross-sections need to be greater than the number of coefficients. As such, we are restricted to the FEM or the pooled OLS model. To estimate the most efficient model, we use the Redundant fixed effect test to determine whether there are fixed effects during the observed periods. In following, we state the null hypothesis: there are no fixed effects, i.e., there is no discernible specific variation between each bank over the period (Bolek & Lyroudi, 2017) versus the alternative hypothesis that there is a fixed effect between the banks.
5. Analysis
5.1 Variance Inflation Factor Analysis
Before we begin our interpretation, we start by investigating the presence of multicollinearity between our independent variables. The output table V shows that the most noticeable VIF values are 4.3257 and 4.7588 for ECB and UK, respectively. These values overall do not indicate the existence of severe multicollinearity, and the t-test will be trustworthy. On the contrary, amongst the independent variables, our a priori expectation is that multicollinearity should be relatively high and present to some degree. However, since our highest values were found in the control variables, this should not pose a significant problem for the analysis. If the VIF values are deemed too high, we recommend that future researchers use techniques such as RIDGE regression or make a weighted average by combining the variables for UK and ECB.
26
5.2 Redundant Fixed Effects Test
Since we use a panel data regression, we want to investigate which of the two models is the better fit. For this purpose, we conduct the redundant fixed effects test with the null hypothesis that there is no fixed effect and pooled OLS is the suitable model. We obtain the following results posted in table VI in the appendix. Since the p-value is 0.8107, we do not reject the null hypothesis at the alpha 5% level, concluding that there are no fixed effects between the banks. We will proceed with the analysis using the pooled OLS regression.
5.3 Results and Interpretation
We can reject the null hypothesis at the 5% significance level for all our results in the first regression, meaning that there is evidence for a statistically significant relationship. Regarding how the data fit the models, it may appear to be slightly low as the R-squared value indicate that about 16% of the variation in bank profitability is explained by the model. However, it is not surprising as the number of variables that influence stock prices even within such a narrow window is large. For instance, news about bank management, macroeconomic events, and specific bank ratios, among many, may impact bank share prices. Nevertheless, our study seeks to establish a relationship and not the determinants of banks stock prices. Furthermore, our R-squared is consistent with the equivalent value found in neighbouring studies such as Ampudia and Van den Heuvel (2019) and Altavilla et al. (2018) as such we deem our value to be sufficient for the purpose of our study. Concerning the CDS regression, since it is mostly meant to verify the result found in our stock price regression, the R-squared is of little significance.
Analysing the results reported in table III indicates that the long-term rate change is highly statistically significant, and we reject the null hypothesis that there exists no significant positive relationship. We find that a ten-basis point decrease in long-term rates our measure of QE on average increase bank's stock prices by 0.467%. This result is in contrast with Demetriz and Wolff (2016) while being consistent with the research presented by Altavilla et al. (2018). Like the previous studies, lower interest rates should improve the macroeconomic outlook. Therefore, banks should benefit from the increase in loan volumes and the decrease in non-performing loans if these effects outweigh the loss in the NIM. However, in contrast to Altavilla et al. (2018), we find a more
27
substantial statistical significance of our result. The discrepancies in findings may be explained by looking at the liability side of the banks' balance sheet. Swedish banks' funding consists primarily of wholesale funding instead of deposit funding for the European Area banks and as such, they benefit relatively more from QE (De Rezende & Laséen, 2018; Riksbank, 2016).
Taking into consideration that a large part of Swedish banks wholesale funding is done through the issuance of covered bonds (Eidestedt et al., 2020). The banks benefits significantly from these bonds when interest rates fall for two reasons; the bank has to pay a lower interest rate to the investors, and the collateral value goes up, which further reduces the cost of the covered bonds.16 Therefore, lower
long-term interest following QE increases Swedish bank profitability as market participants perceive that the overall wholesale funding conditions improve. Moreover, the effects of QE on bank
profitability can differ between countries and may be positive in Sweden partly due to structural differences in bank funding. Importantly, this implies that the choice of banks' funding is an important determinant to consider before QE.
The same ten basis points decrease in short-term rate on average increase bank’s stock prices by 1.074% as the gap between short-term interest rate and long-term interest rate widen; in other words, the yield curve steepens, improving the NIM, given that the bank begins lending at the higher interest margin (Riksbanken, 2016). Lower short-term interest rates ceteris paribus decrease the cost of funding, importantly on interbank borrowing and deposits rates therefore improving short-term market funding and lending to households and companies. Furthermore, the fall in interest rates will improve customers' debt-servicing ability and increase loan volumes similar to previous segment of our analysis.
In addition, short-term securities will increase in value following a decrease in the discount rate. This benefits not only the banks holding these assets but also borrowers as the quality of their collateral rise which should further stimulate credit demand (Altavilla et al., 2018). However, with an increase in loan volumes banks are bound to make riskier loans in search for higher yields which may in the long run have severe consequences for the wider economy i.e., defaults and in the worst outcome 16 Value of collateral increase as real assets benefit from lower discount rates.
28
recessions when interest rates finally rise. In contrast since most of mortgages are tied to a variable interest rate declining interest rate should negatively impact the bank. Nevertheless, we find changes in the short-term interest rate to be the most impactful which may be attributed to the influence the repo-rate has on the NIM.
The result for the medium-term rate change might be the most ambiguous out of our controls making up the yield curve. Previous literature primarily studies the relationship between bank profitability and QE through short or long-term interest rates. This is partly due to the intermediate segment being between the other two segments. Therefore, it is hard to disentangle the particular effect on the midterm part of the yield curve. However, we hypothesize that the adverse effects of QE on bank profitability may stem from the signalling channel; in particular, forward guidance has a more considerable impact on the intermediate segment on bond yields (De Rezende, 2017;
Krishnamurthy & Vissing-Jorgensen, 2011; Swanson, 2021). Therefore, ceteris paribus, if the bank holds a substantial amount of mid-term bonds, a decline in the interest rate may disproportionally decrease yields due to forward guidance influence on the mid-term segment as the term premia is affected.
For maturity transformation to be profitable, the term premia pertaining to the long-term asset must be larger than that of the short-term since the bank funds the purchase with this difference in premia. This width of the gap is determined by the term premia on assets which means that when QE flattens the yield curve, the term premia flatten, thus reducing maturity transformation income. Similarly, this will be the case for forward guidance as it increases the confidence of investors that the future interest rate will be around a certain level and so these two effects may have an adverse impact on bank profitability in the medium-term to long-term (Dell’Ariccia et al., 2017). In the paper by Swanson (2021) who observed such effects as both QE and forward guidance positively affected bond yields with mid-term maturity.17 However, a change in QE was associated with a negative
effect on long-term bond yields. Intuitively this make sense as QE primarily target long-term interest
17 We would just like to remind the reader that for consistency we assume a decline in interest rates for our interpretation.
However, in Swanson’s paper he assumes an increase in LSAP and Forward guidance, but this will still lead to the same conclusion.
29
rates and therefore reduce bonds yields on the long end of the yield curve, which is consistent with our reasoning in the previous section of the analysis.
The findings when including a dummy interaction effect for each bank using SHB as the reference can be interpreted that on average the marginal impact of QE on bank’s stock prices is an increase by 0.73%, 0.862%, 0.162% and 0.219% for Nordea, SEB, Swedbank and SHB respectively given a ten-basis point decrease in long-term interest rates. The presented values and the signs are as expected with only one outlier, namely, Swedbank. These results suggest a difference in the effectiveness of QE on bank profitability depending on which bank is observed. Looking at figure II, we observe that Nordea and SEB overall have the highest problem loan to gross customer loan ratios for most of the period monitored while Swedbank has the lowest. 18 The differences in ratios
may be explained by the individual banks risk taking behaviour and exposure to certain sectors, in particular we observe that Nordea have a large share of foreign loans which may have contributed to the consistently high ratio of problem loans.
Figure II. Problem loans to gross customer loans
18 Problem loan is a commercial or consumer loan that is past due and include nonperforming loans, gross impaired
loans, net impaired loans and other problem loans (Definition provided by SNL).
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 2014-12-01 2015-03-01 2015-06-01 2015-09-01 2015-12-01 2016-03-01 2016-06-01 2016 -09 -01 2016-12-01 2017-03-01 2017-06-01 2017-09-01 2017-12-01 2018-03-01 2018-06-01 2018-09-01 2018-12-01 2019-03-01 2019-06-01 2019-09-01 2019-12-01 2020-03-01 2020-06-01 2020 -09 -01 2020-12-01 2021-03-01
