Unconventional monetary policy and stock market prices in a small open economy: Evidence from Sweden’s quantitative easing

Unconventional monetary

policy and stock market

prices in a small open

economy: Evidence from

Sweden’s quantitative

easing

BACHELOR DEGREE PROJECT THESIS WITHIN: Economics NUMBER OF CREDITS: 15

PROGRAMME OF STUDY: International Economics AUTHOR: Claudia Tirado Luy, Nikola Kolev

JÖNKÖPING May 2020

The relationship between Riksbank’s large-scale asset

purchases and the OMX Stockholm 30 stock market index

i

Acknowledgements

We, Claudia Tirado and Nikola Kolev, would like to take the opportunity to express a word of gratitude to a couple of individuals who supported us throughout the preparation of this thesis.

Firstly, we would like to express our gratitude to our supervisor, Anna Nordén, for her invaluable guidance, constructive criticism and words of advice during the whole process. Secondly, we are thankful for all the input that we received from our fellow classmates during the seminar sessions.

Lastly, thank you to our families and friends, who encouraged us not only during the writing of this thesis, but throughout the whole course of our degree.

__________________________ __________________________ Claudia Tirado Luy Nikola Kolev

Jönköping International Business School May 2020

ii

Bachelor Thesis in Economics

Title: Unconventional monetary policy and stock market prices in a small open economy: Evidence from

Sweden’s quantitative easing

Authors: Claudia Tirado Luy and Nikola Kolev Tutor: Anna Nordén

Date: 2020-05-18

Key terms: Unconventional monetary policy, Quantitative easing, Large-scale asset purchase, Stock market

Abstract

This thesis aims to investigate the long-term behaviour of the Swedish stock market under quantitative easing (QE) between the years 2015-2019 in comparison to an equally long period before the implementation of QE. The relationship is analysed within the framework of transmission channels of monetary policy and with considerations for previous research on the topic. By the means of an autoregressive distributed lag (ARDL) model, we conduct a regression analysis using the price level of the OMX Stockholm 30 (OMXS30), the value of Riksbank’s assets, the short-term interest rate and the industrial production index. The results show significant but weak evidence of a positive relationship between the OMXS30 index and the Riksbank’s assets value. Furthermore, we analyse the findings to provide an insight into the transmission of unconventional monetary policy to the stock market in a small open economy. Finally, we present some broad implications of our study, as well as suggestions for future research on the topic.

iii

Table of Contents

1.

Introduction ... 1

2.

Monetary policy ... 5

2.1 Previous experiences with quantitative easing ... 6

2.2 Transmission channels of monetary policy ... 7

2.2.1 Portfolio balance channel ... 7

2.2.2 Signalling channel ... 8

2.3 The Yield Curve ... 8

3.

Previous Research ... 9

4.

Hypotheses ... 11

5.

Method and Data ... 11

5.1 Model specification ... 11

5.2 Data ... 13

5.2 Stationarity test ... 13

5.3 Cointegration Testing and Long-Term Coefficients ... 14

6.

Results and Analysis ... 15

6.1 Descriptive statistics ... 15

6.2 Dickey-Fuller Generalized Least Squares Test ... 15

6.3 ARDL Model, Long-Term Coefficients and Bounds Test ... 16

6.4 Robustness Checks ... 22

7.

Discussion ... 23

8.

Conclusion ... 25

9.

References ... 27

10.

Appendix A ... 33

11.

Appendix B ... 36

iv

Tables

Table I. Description of variables used in Equation (1) ... 13

Table II. Descriptive statistics ... 15

Table III. DF-GLS test ... 15

Table IV. Model specification ... 17

Table V. Bounds test based on Pesaran et al. (2001) ... 18

Table VI. Long-term coefficients for the period 2010-2014 ... 19

Table VII. Long-term coefficients for the period 2015-2019 ... 20

Table VIII. Philips-Perron test results ... 22

Figures

Figure 1. OMXS30 index relative to the Riksbank’s value of assets ... 21

1

1. Introduction

1.1 Background

The debate on the reaction of stock markets to monetary policy announcements has gained steam following the implementation of unconventional monetary policy tools on behalf of several major central banks worldwide. The concept of unconventional monetary policy is utilized to denote the implementation of unorthodox policy instruments on behalf of the central bank. These policy tools, however, differ among countries that have taken that policy turn. The two most used policy tools are the negative interest rate policy (NIRP) and quantitative easing (QE), while in some cases both are used in combination. On the one hand, the NIRP is a measure taken by several central banks to push short-term interest rates below the zero-lower bound. Since financial markets are designed to operate under positive nominal interest rates, prior to the implementation of NIRP, economists commonly perceived 0% as the lower bound which cannot be surpassed (Keister, 2011). Although NIRP has been implemented in some countries, its effect is hardly quantifiable due to the low frequency at which policy rate cuts occur. On the other hand, a QE program involves a large-scale purchases of government bonds on a monthly basis to influence the long-term interest rates (Joyce, Miles, Scott, and Vayanos, 2012). Given the declining efficiency of further policy rate cuts approaching 0%, QE was implemented by several central banks to provide liquidity and incentivize borrowing. Besides having an impact on the inflation rate, these large-scale purchases of assets have influenced the behaviour of several stock markets in countries where such a policy has been implemented.

Previous research on countries whose central banks adopted a QE program found a predominantly positive relationship between QE and the price level of the stock market, yet there is some heterogeneity between different economies. On the one hand, Eksi and Tas (2017) examined the change in the effect of the U.S. Federal Reserve’s monetary policy after the adoption of QE and found that the policy incentivises investors to purchase equities as a substitute for lower-yielding treasury securities, thus increasing the demand for stocks and driving their price up. In a study on the stock markets in the U.S., the U.K. and Japan, Lima, Vasconcelos and de Mendonca (2016) found strong indications that QE had a positive effect on their respective equity markets. On the other hand, according to a paper

2

by Nakazono and Ikeda (2016), stock prices respond negatively to unexpected QE in an environment beyond the zero-lower bound of interest rates.

Most of the previous research has been conducted on large open economies with large capital markets. These large economies in terms of the percentage of global GDP exert a stronger influence on world price levels by conducting their own monetary policy. Unlike large economies, countries considered as small open economies are highly influenced by exchange rate fluctuations due to monetary policy shifts on behalf of large economies' central banks, such as the U.S. Federal Reserve, the European Central Bank, etc. An example of such a country is Sweden, considered by several studies as a small export-reliant economy (Stockhammar & Österholm, 2016; Neumeyer & Perri, 2005). Since Sweden is one of the few economies of this category to have implemented QE, it becomes a suitable candidate for our study together with its central bank (Riksbank). The Riksbank monetary policy authority is given a dual mandate by the Swedish parliament. The Riksbank is primarily responsible for maintaining price stability, meaning that inflation should be low and stable, close but below the target of 2% (Sveriges Riksbank, 2018a). Second, but no less important, is the Riksbank’s responsibility to ensure the overall stability of the financial system, which includes the stock market, fixed-income and exchange rate markets (Sveriges Riksbank, 2018b).

Similar to other European economies, Sweden faced significant disinflationary pressures during the period 2012-2015, when the inflation rate undershot the 2% goal set by the Riksbank. To bring it back to the desired level, the Riksbank decreased its main repo rate, the interest rate at which banks can borrow or deposit funds at the Rikbank for a period of seven days (Sveriges Riksbank, 2020). The repo rate was cut to 0%, yet with little to no effect on inflation. Empirical evidence for this effect is presented by Borio and Hofmann (2017), who argue that the effect of lower interest rates on aggregate demand is diminishing when policy rates approach the zero-lower bound. Consequently, in February 2015 the Riksbank engaged in a QE program of its own with an indefinite horizon (Sveriges Riksbank, 2015). As the Riksbank cut its repo rate below 0% in February 2015, it became the first worldwide to experiment with a combination of a QE program and NIRP (Sveriges Riksbank, 2015). Five years into the experiment, the Riksbank was also the first one to exit negative territory by bringing the repo rate back to 0% (Sveriges Riksbank, 2019a). Although

3

its innovative approach has attracted global attention, previous research has not focused on the behaviour of the Swedish stock market under unconventional monetary policy.

1.2 Problem statement

Based on empirical data, there might be a trade-off between the two goals the Riksbank has. Swedish interest rates have remained “low for long” to tackle the below-target inflation, but this might have created risks for the financial system’s health. De Haan and van Den End (2018) depicts the divergence of risks for price stability and financial stability, for a low level of inflation can be combated by QE at the expense of higher risk of asset price bubbles. As the Riksbank itself noted in its Financial Stability Report for 2019, in a low interest rate environment, investors’ search for higher yields led to the high valuation of several asset markets globally (Sveriges Riksbank, 2019b). The Riksbank supports its argument with an assessment of the International Monetary Fund (2019), which suggested the possibility that the stock markets in the US and Japan are overvalued.

Therefore, a question arises whether a similar behaviour of asset prices under unconventional monetary policy is to be expected in a small open economy. According to the Bank for International Settlements (2019), demand for the U.S. Dollar, the Japanese Yen and the British Pound is significantly greater than that for currencies of small open economies. Hence, we suspect that the magnitude of the effect of monetary policy transmission would decrease the lower is the demand for a given currency. Since the stock market is a representation of market participants’ expectations regarding the future state of the economy, we can investigate them through the relationship between QE and stock prices. Furthermore, there has been some heterogeneity in different countries regarding the reaction of market participants to announcements of unconventional monetary policy. Thus, stock market participants in Sweden might have perceived neutrally or even negatively the implementation of QE.

In addition, de Haan and van Den End (2018) suggest that by encouraging risk-taking by market agents, QE may lead to equity valuations that do not reflect their fundamental value. Due to the price increase of the assets which are purchased by the central bank, the demand for close substitutes (stocks) will also rise, increasing the risk of forming financial imbalances and asset price bubbles. Asset price bubbles, in turn, could adversely affect the entire financial system (de Haan & van Den End, 2018).

4

1.3 Purpose

The purpose of the following study is to explore the gap left by previous research regarding the relationship between the value of a central bank’s assets and the returns of a relatively small stock market in a small open economy. More explicitly, we will focus on the behaviour of the OMX Stockholm 30 (OMXS30) stock market index following the implementation of QE by the Riksbank in an environment of “low-for-long” interest rate. The OMXS30 is a market-weighted price index, which consists of the thirty most actively traded stocks on the Stockholm Stock Exchange (The Nasdaq Group, 2020). Hence, we will explore the long-term relationship between the value of Riksbank’s assets (a measure of QE) and the OMXS30. The results will answer the following question:

What is the long-term relationship between the Riksbank’s quantitative easing and the Swedish stock market in the post-crisis period?

The findings will shed light on how significant the correlation between QE and the OMXS30 resulted to be in an environment of negative short-term interest rates. Furthermore, they will provide the reader with an insight into whether stock prices increased following the implementation of a QE program by the Riksbank. Additionally, the results will offer empirical evidence regarding the efficiency of monetary policy transmission in such an environment, which will be invaluable for the management boards of the central banks of other small open economies.

1.4 Outline

The remainder of the thesis is organized as follows: section two illustrates the theory behind QE and provide some countries where it was implemented. Section three presents previous research on the behaviour of stock prices under a QE program. Section four highlights the hypothesis of our study. Section five covers the selected method, which involves a stationarity testing combined with an Autoregressive Distributed-Lag (ARDL) model. Section six presents the empirical results. Section seven discusses the findings and critically relates them to the previous literature on the topic. The last section concludes our research, followed by a list of references and an appendix where data is presented thoroughly.

5

2. Monetary policy

Monetary policy, as defined by Johnson (1962), is the device the central bank employs to control the supply of money in order to satisfy the objectives of general economic theory. In other words, it aims to, but is not limited to, preserve price stability, stimulate employment, sustain economic growth and hold stable interest rates or exchange rates (Hildebrand, 2006). For monetary policy to work, the central bank will influence prices and yields of financial assets that can affect the decisions of economic agents which in turn can lead to economic development (Bernanke & Reinhart, 2004).

Monetary policies could be of either a conventional or unconventional nature depending on the goal of the central bank and the tools that it uses to attain them. Conventional monetary policies are practiced by the means of changes in the short-term interest rates (Lima et al., 2016). This is done through open market operations that include the sale and purchase of securities from the banking system, which in turn influences the banks’ level of monetary reserves in this system (Joyce et al., 2012). In an expansionary monetary policy setup, a central bank would decrease the interest rate in order to discourage commercial banks to keep reserves with the central bank. However, when these rates approach 0%, there is little room for a central bank to offer the commercial banks greater incentives to lend out money. Therefore, as interest rates are stuck at the zero-lower bound, putting the economy effectively in a liquidity trap, a central bank can guide market participants via other tools such as making commitments (Ichiue & Ueno, 2015). In such an environment, it becomes complicated for researchers to quantify the monetary policy stance of a central bank. The liquidity problems with the traditional monetary measures, can make central banks turn to the use of unconventional monetary policies. Janus (2016) highlights, as a factor of conventional monetary policy inefficiency, the fact that a very high demand for monetary reserves in the financial system makes the central bank less capable of manipulating the interest rates in the market. Unlike conventional monetary policy, unconventional monetary measures are geared towards influencing long-term interest rates rather than the usual short-term rates. This is done by changing the composition and size of a central bank’s balance sheet through the purchase or sale of assets to influence the supply of reserves and the money stock (Bernanke & Reinhart, 2004). When expanding the assets side of the balance sheet by purchasing government securities, a central bank intends to support economic recovery by

6

putting downward pressure on the longer-term interest rates and raising the price of other assets (Lima et al. 2016). With interest rates lower than they would be under conventional monetary policy, individuals and firms will experience spending incentives due to the lower cost of borrowing (Sulaiman & Imad, 2019).

2.1 Previous experiences with quantitative easing

QE practices in the USA started after the subprime crisis hit the economy. Followed by low aggregate demand and poor economic performance, the Federal Reserve of the U.S. (Fed) decided to lower the federal funds rate, their principal monetary policy instrument, to counteract the recessionary forces the country was going through. When the federal funds rate hit the zero-lower bound, the Fed utilised unconventional measures such as QE in order to recover from the economic shock (Eksi & Tas, 2017). The QE programme was unfolded in three rounds of purchases of long-term securities, due to the limited success of injecting liquidity in the American economy during the two first rounds. Consequently, the balance sheet enlarged to about three times its initial state, taking the economy back to a sound growth rate as well as sound levels of employment by reducing longer-term interest rates (Bhar, Malliaris and Malliaris, 2015).

Another programme of QE was implemented in the United Kingdom during the financial crisis since, again, conventional monetary policies could not enhance the financial market conditions such as weak money growth and low inflation (Miles, 2010). The large-scale asset purchases made by the Bank of England amounted to £200 billion, comprising of private and public assets as well as governments securities to increase the money supply. This lowered the medium to long-term security yields by about 100 basis points which subsequently influenced the financial market, by experiencing the recovery of most assets (Joyce, Lasaosa, Stevens & Tong, 2010).

Japan made use of QE in two occasions. The first one occurred when the country entered a liquidity trap during the 1990s. In fact, the prolonged stagnation of the economy made Japan a pioneer country of QE previous the crisis of 2008-09. After 2008, the Bank of Japan embarked for the second time a programme of QE, with the aim of increasing the monetary base (Dragoi & Balgar, 2016). The Japanese large purchase of assets consisted of both long/short-term government bonds and risk assets, which resulted in an increase of the

7

money supply lowering the short-term interest rates and raising the inflation rates in order to meet the desirable target (Matsuki, Sugimoto & Satoma, 2015).

Sweden, our country of study, experienced QE through a large-scale purchase of government bonds since February 2015. In contrast with other countries, a remarkable feature of the Swedish QE programme was the negative levels that the repo rate, Sweden’s monetary tool, reached. Starting from February 2015, the Riksbank cut its repo rate to -0.10% to later, throughout the same year, cut it to down to -0.35%. Moreover, the goal of Sweden’s QE programme was not limited to providing extra liquidity, as in other countries. It also pursued to avoid a rapid currency appreciation as well as encourage bank lending by lowering the interest rates to boost the economy and meet their inflation target (De Rezende, 2017).

2.2 Transmission channels of monetary policy

To shed light on the relationship between QE and the stock market, it is important to understand the transmission channels of monetary policy, since it is through them that QE (or any other monetary measure) can influence stock prices (Hildebrand, 2006). Researchers mention various transmission mechanisms, but in this thesis, we mention two of them, which are relevant for our research purpose since these have an influence on investors, their perception and hence, the stock market.

2.2.1 Portfolio balance channel

The portfolio balance channel is when investors replace their investments as a result of the central bank purchasing securities of different maturities, and hence increasing the price of financial assets. In that way, investors try to find a different portfolio mix to counteract the decline in yields and thus, they increase the demand for substitute assets, such as stocks (Chebbi, 2019). Joyce et al. (2012, p. 277) argue that as “some people have to come to hold different portfolios and prices need to change to make this an equilibrium”, heterogeneity is an important feature across agents in this channel. In fact, the portfolio balance channel has its grounds on the assumption of perfect asset substitutability, explaining how the central bank’s modifications of asset ownership will influence the decisions of economic agents (Janus, 2016). The intervention of the Riksbank in the government bonds market will make

8

investors move away of holding such assets and substitute them with stocks. Thus, the higher demand for stocks, all else equal, will make their prices go up.

2.2.2 Signalling channel

The effects through the signalling channel could be transmitted to the stock market in two different directions (Chebbi, 2019) and rely mostly on the expectations of the state of the economy (D’Amico et al., 2012). If the public perceives QE as a signal of a worsening economy, investors will seek less risky assets such as government bonds, which in turn will lower their return. Thus, the demand for investing in other assets, e.g. in the stock market, will decrease and consequently also their prices. On the other hand, QE could give signals of the central bank’s commitment of keeping low interest rates for a prolonged period. This would make the long-term investments less appealing and hence, investors would try to get their exposure to stocks and other riskier assets, increasing the demand for them and rising their prices. This would make the long-term investments less appealing and hence, investors would try to increase their expose to stocks and other riskier assets, increasing the demand for them and rising their prices.

2.3 The Yield Curve

In order to grasp the effects of unconventional monetary policies, it is appropriate to introduce the term structure of interest rates with the help of the yield curve, which shows the spot yield on government bonds with different maturities (Sinclair & Ellis, 2012). Normally, the yield curve is upward sloping, meaning that the longer an investor holds a bond, the greater yield the investor will get when the bond matures. However, the yield curve can also be downward-sloping, or more commonly known as “inverted yield curve”. In such a scenario, investors demand more long-term government bonds due to their expectations for worsening economic activity in the short-term.

According to Sinclair and Elis (2012), the yield curve can be observed by dividing it in three segments. The very short end of the curve represents the short-term policy rates set by the central bank. At maturities of 3 months or less, investors are indifferent between investing in close-to-maturity government bonds or depositing the money (indirectly) at the central bank. These two segments depict the short term of the yield curve and in general are

risk-9

free options. The short end of the yield curve will remain close to the policy rate set by the central bank (Sinclair & Elis, 2012). On the other hand, the very long end of the yield curve depicts the long term. This part of the yield curve reflects investors’ expectations about long-term variables such as expected inflation and real interest rates. In other words, the long end of the curve shows the long-term horizon beliefs of the economy.

The yield curve is thus highly relevant for our thesis since it encompasses both the short-term and the long-short-term interest rates in the economy. More explicitly, it is a tool that can assist to represent the Riksbank’s goal with the implementation of QE. Through the purchase of long-term government bonds (i.e. QE), the Riksbank influences the very long end of the yield curve. According to Chadha and Waters (2014), the yield curve gives the benefit that, even with near zero (or zero) short-term interest rates, it can represent the changes of longer-term interest rates over time, which reflect market prospects about the future path of monetary policy. In such way, the Riksbank provides the economic agents with a signal that short-term rates will remain low in the future. In turn, the sentiment for low future interest rates incentivizes borrowing and economic activity.

3. Previous Research

Studies conducted prior to the 2008-09 financial crisis have already shown evidence that monetary policies play a role in the stock market behaviour. Taking into account only unanticipated interventions by the central bank, Bernanke and Kuttner (2005) found relatively strong and consistent stock market reactions due to monetary policy actions. However, it is difficult to demonstrate that the effect on the stock market is solely due to monetary policy. Stock market prices reflect all the available information regarding the market such as listed companies’ projected earnings, dividend payments etc. However, these are rarely accounted for in the body of literature, which is what makes it complicated to assess the extent of the effect of monetary policy on stock market prices.

In the case of USA, Corbet, Dunne and Larkin (2019) show that the stock market reacted positively to QE overall. In their study, they focus on the volatility of the American stock

10

index S&P 5001 _{and find that market returns tend to rise after each QE announcement made}

by the Fed. Changes in volatility in the financial markets were also found and Corbet et al. (2019) saw that it stabilized after some hours as investors adapted their trading strategies to the new information available. Focusing also on volatility, Bhar et al. (2015) concludes that even though QE in the USA targeted the long-term interest rates, it did not have an expected effect on long-term securities; rather it had an impact on the stock market mainly during times of high volatility. In line with this argument, Olsen (2014) and Ashraf, Hassan and Hippler (2017) presented in their respective studies that the stock market experienced a measurable rise in prices compared to periods when QE was not implemented.

Lima et al. (2016) obtained evidence of a positive impact of QE on the stock markets of the USA, UK and Japan; by investigating the effectiveness of QE in increasing the market share after the subprime crisis. As they conclude, the increase of the money supply in the USA affected positively the stock market but they also add that it was partly due to higher industrial production and an appreciated exchange rate. For the Japanese stock market, there were also strong indications that QE led to an increase on the stock returns. Finally, the stock market of the UK showed a positive relationship with QE as well, though the exchange rate also played a role in this impact as it happened in the USA.

There are, however, some research studies suggesting that unconventional monetary policies negatively affect the stock market. QE announcements, as studied by Nakazono and Ikeda (2016), tend to worsen stock returns. This is explained by the fact that the lower interest rates that accompany QE may trigger pessimistic expectations of the economic outlook and therefore, stocks diminish in value. This is in line with what Kontonikas, Macdonald and Saggu (2013) previously presented regarding negative stock returns due to QE interventions during the crisis.

The scarce but growing body of previous literature provides us with a strong foundation for our thesis. First, given the predominantly positive relationship between QE and stock returns, it is helpful for the establishment of null and alternative hypotheses. Second, the choice of our control variables is based on those variables that proved most efficient

1 _{The S&P 500 is a capitalization-weighted index, which follows the prices of the 500}

11

according to previous research. Third, we incorporate a theoretical framework based on the transmission channels of monetary policy which is commonly utilized in previous studies. Lastly, the existing body of literature, focused mainly on large open economies, allowing us to compare previous findings to the results of a study on a small open economy such as Sweden.

4. Hypotheses

Our main goal is to examine the long-run relationship between an expansion of the Riksbank’s balance sheet and the price level of OMXS30. Based on Lima et al. (2016) who conducted a similar research on economies which experienced large-scale QE, namely the United States and Japan, we would expect a similar positive relationship between the growth of Riksbank’s assets and the increase in stock market prices. Since the main policy rate is at record-lows, the yield on treasury securities gets extremely low or even negative and investors turn to equities in search for higher returns. Thus, we expect that there is a significant long-run relationship after the implementation of QE. More explicitly, we will test the following hypotheses:

H0: There is no statistically significant long-term relationship between the value of Riksbank’s assets and the

price level of OMXS30

H₁: There is statistically significant long-term relationship between the value of Riksbank’s assets and the

price level of OMXS30

5. Method and Data

5.1 Model specification

To test whether there is a statistically significant long-term relationship between QE and the price level of OMXS30, we use the models presented by Liu and Asako (2013) and Lima et al. (2016) and adjust them to construct the following empirical model:

12

In the model stated above, omxt stands for the price level of the OMXS30, c is the constant,

assetst indicates the value of Riksbank’s assets, ipit shows the level of industrial production as a proxy of GDP, tratet represents the yield on a 3-month treasury security, and εt denotes the

error term.

Unlike Lima et al. (2016), we use the size of the value of Riksbank’s assets, as a measure of the scale of QE, as proposed by Al-Jassar and Moosa (2019). Their study, with a focus on the implementation of QE in the United States, finds that the expansion of the Federal Reserve’s balance sheet was not matched by an equivalent or proportional rise in monetary aggregates. According to Al-Jassar and Moosa (2019), the banks who sold government bonds to the Federal Reserve kept the additional liquidity as reserves, rather than lending it out to the public. This view is reinforced by Eksi and Tas (2017), who also mention the use of the size of a central bank’s balance sheet as a good proxy for measuring the stance on monetary policy. Hence, we perceive the value of Riksbank’s assets as a more accurate proxy of QE and we incorporate it in our empirical model.

We make use of two control variables to account for the influence of other monetary policy instruments and the real economic activity on the price level of the stock market. First, we use ipit to control for economic growth and the business cycle. This indicator is widely used in previous research as a proxy for Gross Domestic Product (GDP) since it is available on a monthly basis, which makes it easier to collect a higher number of observations. The second independent variable introduced in the empirical model above is the tratet. The variable stands for the yield on a three-month treasury bill. The yield is mainly determined by the supply of liquidity from the Riksbank, the demand for liquidity by the banking sector and credit risk (Hydén, 2018). As such, it closely follows the values of the repo rate. Hence, for the purpose of this thesis, we use the yield on the three-month treasury bill as a more dynamic representation of the official repo rate.

The expected signs of Equation (1) are as follows: First, we expect a positive sign for the coefficient of assetst. As the Riskbank purchases government bonds through its QE program, the market will acquire other kind of assets, such as stocks, raising their price. As regard tratet, we would expect a negative relationship between the yield on a 3-month treasury bill and the OMXS30. Since an interest rate cut is generally perceived as an expansionary measure, if properly transmitted through the signalling channel such action would lead to a rise in stock

13

prices. Lastly, for ipit, we expect a positive sign; as the monthly output of industrial production expands, the expected return of the given equities should increase, and their price should follow accordingly.

5.2 Data

The data collected for this research has been extracted in order to construct a time-series of 120 observations, extended between the periods from the January 1, 2010 to December 31, 2019. All values considered in this thesis are on a monthly basis. For the stock market index, we extracted the values of the OMXS30 from Yahoo Finance. Data on the value of assets of the Riksbank and the industrial production index was retrieved from the Riksbank’s website and Statistics Sweden, respectively. Data on the short-term interest rate as measured by the three-month Swedish treasury bill was sourced from the Riksbank Statistics database. The aforementioned variables are described with their respective sources in Table I.

Table I. Description of variables used in Equation (1)

Variables Description Source

omxt The price level of the OMX Stockholm 30 in index points Yahoo Finance

assetst The value of Riksbank’s assets in SEK Riksbank

ipit The level of industrial production measured in index points Statistics Sweden

tratet The yield on a 3-month treasury security measured in % Riksbank Statistics

5.2 Stationarity test

Since we are working with time-series data, our variables may exhibit a trending behaviour which would make our analysis and conclusions misleading due to spurious regressions. Therefore, it is important to test them for stationarity, which implies that their mean, variance and autocovariance do not depend on the specific time period (Levendis, 2018).

In order to avoid spurious regressions, it is necessary to verify for stationarity of the variables by the means of unit root tests. Our explanatory variables are of a trend character and hence, we expect them to be nonstationary. To test whether the time series are stationary, we conduct a Dickey-Fuller generalized least squares method (DF-GLS). According to Elliott, Rothenberg and Stock (1996), the DF-GLS method is an improved version of the augmented

14

Dickey-Fuller test, which filters the deterministic components of the series. To check the robustness of our results we apply a Phillips-Perron test, proposed by Perron (1988) on all series under investigation.

5.3 Cointegration Testing and Long-Term Coefficients

According to Engle and Granger (1987), a linear combination of two or more non-stationary time series may be stationary. If such a linear combination is feasible, the variables are said to be cointegrated. The corresponding interpretation is that there is a long-run equilibrium relationship among the variables.

If the variables are found to become stationary after taking their first difference using the DF-GLS test, we can proceed to cointegration testing. For this purpose, we will use an Autoregressive Distributed-Lag (ARDL) model, as proposed by Pesaran, Shin, Smith, Hendry, and Pesaran (2001), with an automatic selection of lags based on both Akaike and Schwarz information criteria. According to Pesaran et al. (2001), previously developed cointegration techniques such as Engle and Granger (1987) and Johansen (1991) can be applied to cases in which the underlying variables are integrated of order one. Thus, an important advantage that the procedure proposed by Pesaran et al. (2001) is that it is applicable to time series regardless of their order of integration, be it purely I(0), purely I(1) or mutually cointegrated. According to Pesaran and Shin (1998), another feature of the ARDL model is that it gives consistent and reliable results even with a small sample size. This is very convenient for our thesis since we examine a relatively low number of observations (60 for each period).

As previously noted, we will divide the data into two samples of equal length: the first one indicates the period prior the use of QE (January 2010-December 2014, 60 observations) and the second one is to refer to the period after the introduction of QE (January 2015-December 2019, 60 observations). In such way, we can test if there is any difference in the statistical significance of the long-term relationship between assetst and omxt before and after the introduction of such a policy. Furthermore, we can observe any changes in the relationship of the other explanatory variables and the level of the stock market before and after the adoption of QE.

15

6. Results and Analysis

6.1 Descriptive statistics

Before we begin with the analysis of our empirical evidence, we briefly introduce the main statistics of the variables used in this study. The following table showcases the minimum, maximum and average values of each time series, as well as their respective standard deviation:

Table II. Descriptive statistics

Variables Minimum value Maximum value Mean Standard deviation

omxt 910.1700 1771.850 1353.314 236.2708

assets_t 315810.0 969911.0 621129.0 226978.7

ipit 94.7000 111.9000 103.2892 4.7052

tratet -0.7938 1.9077 0.2004 0.8495 The OMXS30 recorded its lowest value in September 2011 and reached a peak in December 2019. The value of Riksbank’s assets was lowest in April 2011 and highest in February 2019. The industrial production index reached a minimum value in January 2015 and a maximum in August 2019. The short-term interest rate was lowest in November 2016 and highest in May 2011.

6.2 Dickey-Fuller Generalized Least Squares Test

We start our analysis by conducting a unit root test to check the stationarity of the series under investigation. For that purpose, we use a DF-GLS method and obtain the following results:

Table III. DF-GLS test

Variable Level 1st difference (coefficient)

assetst -0.0019 -0.4237***

omxt 0.0029 -1.0949***

ipit -0.0227 -0.5301***

tratet -0.0068 -0.4536***

16

As the results presented above clearly show, all series under consideration are non-stationary in their level form. However, having taken the first difference, all four series become stationary. The first difference of the variables which we use in our regression analysis are stationary at the 1% level, which makes the results of our unit root test highly significant. This is to show that the order of integration of the variables is less than two, which is a requirement to conduct a cointegration testing (Pesaran et al., 2001). This result is a very important feature for the core of our regression analysis. The fact that the variables are integrated of order 1, i.e. I (1), allows us to proceed with a cointegration and bounds testing procedure, based on Pesaran et al. (2001).

6.3 ARDL Model, Long-Term Coefficients and Bounds Test

Since we have identified that all the variables are integrated of order 1, we can proceed to the core of our regression analysis, namely the construction of our ARDL model. We focus our further analysis on the output of a cointegration test based on Pesaran et al. (2001) on the following equation:

omxt=f(assetst, ipit, tratet) (2)

Therefore, on the basis of the equation (2) we will test the null hypothesis of no levels relationship i.e. that there is a significant long-term equilibrium relationship between the price level of OMXS30 and the value of Riksbank’s assets (a measure of QE), controlling for external factors using the industrial production index and the yield on a three-month treasury bill. In order to quantify the relationship between the independent variables and the dependent variable we use an ARDL model with a bounds testing approach.

First, we select the best model through an automatic lag selection based on both the Akaike (AIC) and Schwarz (SIC) information criteria. In this way, the models that are chosen are the ones with the lowest values of AIC and SIC. As throughout the whole study, we divide the observations into pre-QE and post-QE periods, and we obtain the following models:

17

Table IV. Model specification

Information Criterion 2010-2014 2015-2019

AIC (2,1,4,1) a _(2,0,0,0) a

SIC (1,0,0,0) a _(1,0,0,0) a

Notes: The values in parenthesis show the number of lags of the variables. a _{No serial}

autocorrelation of the residuals.

As it can be observed from the output in table IV, for the period prior to the introduction of QE, according to AIC the most suitable model is the one with two lags of omxt and one, four and one lag of assets_t, ipi_t and trate_t respectively. This model is the one that minimizes the value of AIC and has no serial autocorrelation of the residuals. When we choose a model subject to the minimization of SIC, the model that fits best is the one with one lag of omxt and no lags of the independent variables. For the period following the introduction of QE, according to AIC, the most suitable model is one with two lags of omxt and one, four and one lag of assetst, ipit and tratet respectively. As per SIC, the model that minimizes the SIC value is the one with one lag of omxt and zero lags of the independent variables. The difference between these two lag-selection criteria is the methodology behind AIC and SIC, and namely the fact that SIC tends to favour simpler models than those chosen by AIC (Kass & Raftery, 1995).

Having selected the models that provide the best fit, we proceed to a cointegration testing using the bounds test approach proposed by Pesaran et al. (2001). The test will determine whether there is a cointegrating relationship between omxt, on one side, and the independent variables, on the other. We establish a null hypothesis of no cointegration and an alternative hypothesis of a presence of cointegration. For the purpose of testing this, we use an F-test with critical values based on Pesaran et al. (2001). If the F-statistic falls below the lower bound critical value, we cannot reject the null hypothesis. Alternatively, if the F-statistic falls above the upper bound critical value, we reject the null hypothesis of no cointegration. In case the F-statistic falls in between the critical values, the result is inconclusive. Further data regarding the critical values can be found in Appendix A.

18

Table V. Bounds test based on Pesaran et al. (2001) Cointegration coefficient

/Information Criterion

2010-2014 2015-2019

AIC 3.7003** 4.4375**

SIC 2.0712 4.1378**

Notes: The values in the table represent the F-statistic for the hypothesis test of the absence

of cointegration. The critical values are obtained from Pesaran et al. (2001). ** denotes significance at 5%.

The output in table V shows that in three out of the four models, the null hypothesis of no cointegration is rejected, i.e. there is significant evidence for cointegration of the independent variables and the dependent variable, in this case the OMXS30. This is to suggest that, referring to Engle and Granger (1987), the linear combination of the non-stationary time series which we study is stationary. Hence, we can use these three models to derive reliable long-term coefficients for the relationships between the independent variables and the dependent variable. The presence of cointegration between the value of Riksbank’s assets (a measure of QE) and the OMXS30 is consistent with Lima et al. (2016), which finds cointegration of similar variables for other countries, namely the United States, the United Kingdom and Japan.

This result is essential for the purposes of this study for two main reasons. First, the presence of cointegration eliminates the possibility that the relationship between the variables under investigation is spurious. In other words, the relationships which we analyze in the following paragraphs are due to neither coincidence nor the presence of an unobserved factor. Second, the existence of cointegration indicates that a long-term equilibrium relationship is detected between the variables. This allows us to proceed to the next step of our regression analysis, namely the analysis of long-run coefficients of the independent variables.

19

Table VI. Long-term coefficients for the period 2010-2014 AIC

assetst -0.0024**

ipi_t -58.1707**

tratet -327.6850

Notes: ** denotes significance at 5%.

Since no cointegration among the variables was found in the model based on SIC, we cannot analyze the long-run coefficients of the given output in table VI. Thus, we will focus our analysis on the coefficients given from the cointegration equation based on AIC instead. First, a negative relationship is found between assets and omx. Hence, the markets reacted negatively in the long run to an expansion of the Riksbank’s balance sheet, or positively to a sale of assets on behalf of the Riksbank. Here, it is important to note that no large-scale purchase nor sale of government bonds was done by the Riksbank prior to the implementation of QE. Thus, it is the management of other assets such as foreign exchange reserves and other conventional monetary instruments which caused fluctuations of the value of Riksbank’s assets. Furthermore, it is essential for the purpose of this analysis to take into consideration the period under investigation. During the 2008-09 financial crisis, the Riksbank injected large amounts of liquidity to support the markets and consequently, in the period 2010-2012, engaged into removing the excess liquidity as it was no longer needed to support the economy. At the same time, the stock market was recovering from the crisis and trended upwards and perceived the shrinkage of the balance sheet as a positive signal for the future economic outlook.

Second, the coefficient for the industrial production index carries a negative sign, meaning that it is negatively correlated to the price level of the OMXS30. Although the result does not satisfy our expectation, it does not go without explanation. The reason might be the nature of the stock market, which reflects future economic performance, and the fact that the industrial production index reflects current performance. Hence, when companies are

20

expected to perform better in the following months, the stock market can price that in immediately, even though that is not yet reflected by the industrial production data.

Third, in line with our expectations, the sign of the long-run coefficient for trate is negative. In the period prior to the introduction of QE, short-term interest rates were the main monetary policy tool for the Riksbank. This is to suggest that monetary transmission through adjustments of short-term interest rates functioned well in the pre-QE period. In other words, the stock market reacted accordingly to both monetary easing and tightening on behalf of the Riksbank.

Table VII. Long-term coefficients for the period 2015-2019

AIC SIC

assets_t 0.0040 0.0024*

ipit -80.2779 -44.4540

tratet 1475.152 938.3232**

Notes: *, ** denote significance at 10% and 5% respectively.

In the period following the introduction of QE, the relationship between assets and the omx is rather ambiguous given the results of the regression analysis in table VII. The coefficient for assets becomes positive but weakly significant, and that is only in the model based on SIC. Hence, we argue that the signal that the Riksbank sent out to investors when it engaged in QE was not perceived unanimously by all market participants. One reason behind this might be an inefficiency of monetary policy communication through the signaling channel on behalf of the Riksbank. It seems to have shifted investment expectations among investors who in turn became predominantly pessimistic during the months after QE. Another cause for the decreased statistical significance could be attributed to the inability of investors to further expose their portfolios towards the stock market. Since the yield on short-term government bonds was already approaching 0% prior to the implementation of QE, we suspect that market participants’ exposure to fixed-income securities might have been significantly reduced even prior to QE. Thus, we argue that the portfolio balance channel has lost to a great extent its efficiency as there were no more gains to be captured by asset substitution between stocks and fixed-income securities.

21

Figure 1. OMXS30 index relative to the Riksbank’s value of assets

Notes: On the left-hand side omx is measured in index points and on the right-hand side

assets is measured in SEK. The horizontal axis shows the years between 2010-2019.

Graphically in figure I, we can see that the omx and assets did not behave according to our expectations when observing the immediate effect of the latter variable. Overall, the stock market seemed to move towards an appreciation before the implementation of QE; and when this was announced, in the beginning of 2015, the market did not react positively. Although this short-term effect is not within the scope of our research, we can clearly see an unexpected downward correction of the OMXS30. The announcement of further expansionary measures might have deceived economic agents, giving signals of worsening economic conditions and thus, investors opted for decreasing their exposure to stocks. When it comes to the long-term coefficient for ipi, it maintains a negative sign but becomes statistically insignificant after the introduction of QE. This result suggests a decoupling of stock market prices and real economic output. Such divergence might occur due to influence on stock prices from non-fundamental factors such as changes in monetary policy. Hence, the unconventional monetary policy practiced by the Riksbank in this period could be one of the reasons behind this occurrence.

As regards the relationship between the trate and omx, the result is rather surprising and inconsistent with previous research. The positive relationship of short-term interest rates below the zero lower bound and the valuation of the stock market point to an innovative finding: adjustments of the repo rate by the Riksbank become inefficient below the 0% level. If we look to the problem through the prism of transmission channels, it turns out that

22

negative short-term interest rates have given a negative signal to market participants regarding the future state of the economy. Another possible argument for the positive relationship might be found taking a more fundamental approach to stocks’ pricing. Negative rates on deposits are found to hurt the margins of banks and other financial institutions, which in turn has an adverse effect on their earnings. According to Molyneux and Reghezza (2019), who conducted a study on a total of 7,359 banks from 33 OECD Member States, including Sweden, banks in countries who adopted a NIRP experienced a decline in net interest margins and return on assets of 16.41% and 3.06%, respectively, compared to countries who did not engage in such policy. As shown in Appendix B, the OMXS30 stock market index is comprised of several banks and financial institutions, whose stock prices are expected to decrease if their future earnings decrease. Hence, the price level of the OMXS30 will be weighed on by these companies. Nevertheless, this problem falls beyond the purpose of this thesis and shall be examined through further research.

6.4 Robustness Checks

To examine the validity and reliability of our previous results, it is necessary to conduct robustness checks that test the hypotheses previously stated in our method. We do this by the means of an alternative method, namely the Phillips-Perron test.

To test whether the time series are stationary, we conduct a Phillips-Perron test, proposed by Perron (1988), on each series and obtain the following results:

Table VIII. Philips-Perron test results

Variable Level 1st _difference

omxt -1.0214 -12.1591***

assets_t -0.5554 -9.3138***

ipit 0.1468 -17.4047***

tratet -0.8254 -6.2014***

23

Since we find both our dependent and independent variables to be non-stationary in their level form, we proceed with the test of first differences. Thus, we conduct the Philips-Perron test on this new form to assess if they become stationary, as done in our method section. The results from table VIII suggest that after taking the first difference of each of the time series, they all become stationary. Therefore, we can conclude that all series that we observe are I (1). This finding confirms the previously presented output from the DF-GLS test.

7. Discussion

The results presented above were inconclusive to show that QE, as an unconventional monetary policy, was successful in outperforming the shortcomings of conventional monetary policy with short-term interest rates below the zero-lower bound. Although we found a long-term equilibrium relationship between the variables, the long-term coefficients suggested a weak correlation between the value of Riksbank’s assets and the price level of OMXS30. Given that the long-term coefficient for assetst is significant only in one of the two selected models for the period 2015-2019, the level of statistical significance of our result does not match the unambiguous positive relationship between QE and the price level of stock market indices found in previous literature. The findings of our study are not as significant as these of Bernanke and Kuttner (2005), Corbet et al. (2019), Olsen (2014), Lima et al., (2016) and Ashraf et al. (2017), who found a clear positive relationship between QE as implemented in other countries and their corresponding stock markets.

To address the differences in the outcomes between our study and previous ones, we refer to the essential characteristic of Sweden being a small open economy, which distinguishes it from the previously studied countries. As such, the Swedish krona is not nearly as demanded as its counterparts in the U.S, the United Kingdom and Japan. According to the Bank for International Settlements (2019), the U.S. dollar, the Japanese yen and the Great Britain pound were among the four most traded currencies worldwide as of April 2019. Hence, when the supply of the currencies issued by the central banks of large open economies increases, it is met by a global demand. Unlike them, the Swedish krona is a currency with local significance, the demand for which comes mainly from the domestic market. Therefore, as the Riksbank injects liquidity through the purchase of government bonds, the money created

24

might remain in the reserves of commercial banks as market participants might not increase their borrowing.

Since we do not observe a significant appreciation of stock prices following the introduction of QE, we find no evidence to suggest that asset price bubbles were formed as a consequence of unconventional monetary policy. Our initial supposition regarding the potential for asset price bubble formation in Sweden was based on de Haan and van Den End (2018), who suggest that QE may lead to equity valuations that do not reflect their true value. This view was reinforced by the Riksbank (2019b) which stated that the search for yields has led to high equity valuations worldwide. Nevertheless, both the immediate as well as the long-term behaviour of the OMXS30 following the implementation of QE does not provide us with a reason to believe that stock prices decoupled significantly from their fundamental value. Both statistically and graphically, there is no evidence to support the notion of a large impact of the increase in the value of Riksbank’s assets on the price level of OMXS30.

Another reason for the discrepancy between the findings of our study and the results of previous research on other countries is the distinctive combination of NIRP and QE implemented by the Riksbank. Even though both policies have similar objectives, they might decrease each other’s efficiency when being transmitted in the economy simultaneously. As previously stated, some of the components of the OMXS30, e.g. the financial institutions, might have been positively affected by QE in the long run, but negatively by NIRP which makes the net effect of said policies ambiguous. Furthermore, since the NIRP brought the short end of the yield curve below 0%, through QE the Riksbank intended to flatten the curve and reduce long-term interest rates. Hence, either the volume of government bond purchases or the communication on behalf of the Riksbank might not have been optimal to compensate for the NIRP.

We are aware that our research does not come without its limitations. First, as mentioned in Eksi and Tas (2017), there is a time lag between a QE announcement and the actual purchase of securities on behalf of the Riksbank, meaning that we cannot capture immediate effects of QE. To evaluate the short-term effects of QE on the OMXS30, we suggest an event-study method similar to the ones implemented by Breedon, Chadha and Waters (2012) and Bernanke, Reinhart and Sack (2004).

25

Second, we acknowledge that the variables that are part of our thesis do not encompass all the factors that influence the price level of the OMXS30. In other words, we maintain a macroeconomic approach to the formation of stock prices, without consideration of the individual performance of the 30 companies that form the OMXS30. Although we depict potential explanations for the effect that the NIRP and QE might have had on the performance of various industries, we suggest further research to focus on the change in expected earnings following the implementation of these policies. Moreover, the OMXS30 includes different companies along the time, which makes it difficult for us to draw general conclusions of the stock market without taking into consideration the composition of the index.

Third, researchers consider it rather early to evaluate the long-term impact of QE on the Swedish economy. Although we managed to obtain reliable results through an ARDL approach, an increase in the sample size would have given us greater flexibility as regards to the model selection. Hence, our suggestion is for future studies to research the topic using a methodology based on Engle and Granger (1987) or Johansen (1991) to check the robustness of our findings. Furthermore, our thesis does not account for changes in the composition of the Riksbank’s balance sheet, which occurred under the period of investigation. Thus, further research can break down the Riksbank’s assets by components and conduct a similar study.

8. Conclusion

Following the crisis of 2008-09, the inflation rate in Sweden remained well below the Riksbank’s target of 2%, despite the conventional monetary policy measures taken by the authority. To provide further easing, the Riksbank engaged in unconventional monetary policy, implementing a QE program in the beginning of 2015 to incentivize borrowing and lift the inflation rate. Since QE involves a large-scale purchase of government bonds on behalf of the Riksbank, previous research suggests that there might be an effect on equity prices.

The purpose of our thesis was to explore the behaviour of the stock market following the implementation of unconventional monetary policy. More explicitly, we investigated whether there is a long-term relationship between the value of Riksbank’s assets and the price level

26

of the OMXS30. We provided a critical analysis of the signalling and portfolio balance channels and how central banks use them to affect the behaviour of market participants. To quantify the relationship, we conducted a cointegration testing based on an ARDL model for a five-year period prior to QE and an equally long period post-QE, controlling for the level of industrial production and the short-term interest rate. Similar to the body of literature on the topic, we found a statistically significant long-term equilibrium relationship between QE and stock prices in the period 2015-2019. However, the long-term coefficient for the value of Riksbank’s assets was significant only in one of the two models that we run. Hence, it remained ambiguous whether there is a significant correlation between the value of Riksbank’s assets and the price level of OMXS30. We argued that the low level of significance is a consequence of Sweden being a small open economy with low demand for Swedish kronor and the limited efficiency of the portfolio balance channel beyond the zero-lower bound of interest rates among others. Our findings were limited by the relatively small sample size and the omission of company-specific data.

The findings from the experience of Sweden as an innovator in the area of unconventional monetary policy can well serve the policy-making committees of the central banks of other small open economies. We encourage future researchers to implement another approach to our problem when a larger sample is obtainable. Moreover, there is a growing number of central banks that implement QE programs of their own to counter the adverse economic effects of the Covid-19 pandemic outbreak. When sufficient time series data becomes available, we suggest comparing the Swedish case to that of another small open economy which experienced QE under different circumstances.

27

9. References

Al-Jassar, S., & Moosa, I. (2019). The effect of quantitative easing on stock prices: a structural time series approach. Applied Economics, 51(17), 1817–1827.

https://doi.org/10.1080/00036846.2018.1529396

Ashraf, A., Hassan, M., & Hippler, W. (2017). MONETARY SHOCKS, POLICY TOOLS AND FINANCIAL FIRM STOCK RETURNS: EVIDENCE FROM THE 2008 US QUANTITATIVE EASING. The Singapore Economic Review, 62(1), 27–56.

https://doi.org/10.1142/S0217590817400021

Bank for International Settlements. (2019). Triennial Central Bank Survey: Foreign exchange

turnover in April 2019. Retrieved from https://www.bis.org/statistics/rpfx19_fx.pdf

Bernanke, B., & Kuttner, K. (2005). What Explains the Stock Market’s Reaction to Federal Reserve Policy? Journal of Finance, 60(3), 1221–1257. https://doi.org/10.1111/j.1540-6261.2005.00760.x

Bernanke, B., & Reinhart, V. (2004). Conducting Monetary Policy at Very Low Short-Term Interest Rates. American Economic Review, 94(2), 85–90.

https://doi.org/10.1257/0002828041302118

Bernanke, B., Reinhart, V., & Sack, B. (2004). Monetary policy alternatives at the zero bound: An empirical assessment. Brookings Papers On Economic Activity, (2), 1–100.

Bhar, R., Malliaris, A., & Malliaris, M. (2015). The impact of large-scale asset purchases on the S&P 500 index, long-term interest rates and unemployment. Applied Economics, 47(55), 6010–6018. https://doi.org/10.1080/00036846.2015.1061646

Borio, C., & Hofmann, B. (2017). Is monetary policy less effective when interest rates are persistently low? IDEAS Working Paper Series from RePEc. Retrieved from

28

Breedon, F., Chadha, J., & Water, A. (2012). The Financial Market Impact of UK Quantitative Easing. IDEAS Working Paper Series from RePEc. Retrieved from

http://search.proquest.com/docview/2098345865/

Chadha, J., & Waters, A. (2014). Applying a macro-finance yield curve to UK quantitative Easing. Journal of Banking and Finance, 39(1), 68–86.

https://doi.org/10.1016/j.jbankfin.2013.11.008

Chebbi, T. (2019). What does unconventional monetary policy do to stock markets in the euro area? International Journal of Finance & Economics, 24(1), 391–411.

https://doi.org/10.1002/ijfe.1669

Corbet, S., Dunne, J., & Larkin, C. (2019). Quantitative easing announcements and high-frequency stock market volatility: Evidence from the United States. Research in International

Business and Finance, 48, 321–334. https://doi.org/10.1016/j.ribaf.2019.01.007

D’Amico, S., English, W., López-Salido, D., & Nelson, E. (2012). THE FEDERAL RESERVE’S LARGE-SCALE ASSET PURCHASE PROGRAMMES: RATIONALE AND EFFECTS. The Economic Journal, 122(564), F415–F446.

https://doi.org/10.1111/j.1468-0297.2012.02550.x

de Haan, L., & van Den End, J. (2018). The signalling content of asset prices for inflation: Implications for quantitative easing. Economic Systems, 42(1), 45–63.

https://doi.org/10.1016/j.ecosys.2017.08.006

De Rezende, R. (2017). The interest rate effects of government bond purchases away from the lower bound. Journal of International Money and Finance, 74, 165–186.

https://doi.org/10.1016/j.jimonfin.2017.03.005

Dragoi, A., & Balgar, A. (2016). Quantitative Easing limits. Evidence from Japan.(Report).

29

Eksi, O., & Tas, B. (2017). Unconventional monetary policy and the stock market’s reaction to Federal Reserve policy actions. North American Journal of Economics and Finance,

40, 136–147. https://doi.org/10.1016/j.najef.2017.02.004

Elliott, G., Rothenberg, T., & Stock, J. (1996). Efficient Tests for an Autoregressive Unit Root. Econometrica, 64(4), 813–836. https://doi.org/10.2307/2171846

Engle, R., & Granger, C. (1987). Co-Integration and Error Correction: Representation, Estimation and Testing. Econometrica, 55(2), 251–276. https://doi.org/10.2307/1913236

Hildebrand, P. (2006). Monetary Policy and Financial Markets. Financial Markets and Portfolio

Management, 20(1), 7–18. https://doi.org/10.1007/s11408-006-0004-8

Hydén, M. (2018). Market Pulse SEK: The whole story about Stibor. Retrieved May 14, 2020 from Nordea Markets website:

https://e-markets.nordea.com/api/research/attachment/69191

Ichiue, H., & Ueno, Y. (2015). Monetary policy and the yield curve at zero interest. Journal

of the Japanese and International Economies, 38.

International Monetary Fund. (2019). Global Financial Stability Report: Lower for Longer. Retrieved from https://www.imf.org/en/Publications/GFSR/Issues/2019/10/01/global-financial-stability-report-october-2019

Investing.com. OMX Stockholm 30 Components. Retrieved May 16, 2020 from Investing.com website: https://www.investing.com/indices/omx-stockholm-30-components

Janus, J. (2016). The Transmission Mechanism of Unconventional Monetary Policy.

Oeconomia Copernicana, 7(1), 7–21. https://doi.org/10.12775/OeC.2016.001

Johansen, S. (1991). Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models. Econometrica, 59(6), 1551–1580.

30

Johnson, H. (1962). Monetary Theory and Policy. The American Economic Review, 52(3), 335– 384.

Joyce, M., Lasaosa, A., Stevens, I., & Tong, M. (2010). The financial market impact of quantitative easing. Bank of England. Quarterly Bulletin, 50(3), 205. Retrieved from

http://search.proquest.com/docview/851892753/

Joyce, M., Miles, D., Scott, A., & Vayanos, D. (2012). Quantitative Easing and

Unconventional Monetary Policy – an Introduction*. Economic Journal, 122(564), F271– F288. https://doi.org/10.1111/j.1468-0297.2012.02551.x

Kass, R., & Raftery, A. (1995). Bayes Factors. Journal of the American Statistical Association,

90(430), 773–795. https://doi.org/10.1080/01621459.1995.10476572

Keister, T. (2011). Why Is There a “Zero Lower Bound” on Interest Rates?. Retrieved May 15, 2020, from https://libertystreeteconomics.newyorkfed.org/2011/11/why-is-there-a-zero-lower-bound-on-interest-rates.html

Kontonikas, A., Macdonald, R., & Saggu, A. (2013). Stock market reaction to fed funds rate surprises: State dependence and the financial crisis. Journal of Banking and Finance, 37(11), 4025–4037. https://doi.org/10.1016/j.jbankfin.2013.06.010

Levendis, J. (2018). Time series econometrics : Learning through replication.

https://doi.org/10.1007/978-3-319-98282-3

Lima, L., Vasconcelos, C., Simão, J., & de Mendonça, H. (2016). The quantitative easing effect on the stock market of the USA, the UK and Japan. Journal of Economic Studies, 43(6), 1006–1021. https://doi.org/10.1108/JES-05-2015-0081

Matsuki, T., Sugimoto, K., & Satoma, K. (2015). Effects of the Bank of Japan’s current quantitative and qualitative easing. Economics Letters, 133, 112–116.