DEGREE PROJECT
IN REAL ESTATE AND CONSTRUCTION MANAGEMENT BUILDING AND REAL ESTATE ECONOMICS
MASTER OF SCIENCE, 30 CREDITS, SECOND LEVEL STOCKHOLM, SWEDEN 2020
Effects of Quantitative Easing on the Swedish Real Estate Market, an ARDL Approach
Hallsten, Felix Valdenström, Mikael
ROYAL INSTITUTE OF TECHNOLOGY
DEPARTMENT OF REAL ESTATE AND CONSTRUCTION MANAGEMENT
Master of Science thesis
Title
Author(s) Department Supervisor Keywords
Effects of Quantitative Easing on the Swedish Real Estate Market, an ARDL Approach
Hallsten, Felix & Valdenström, Mikael Real Estate and Construction Management TRITA-ABE-MBT-20542
Song, Han-Suck
Quantitative Easing, ARDL, Stock Prices, Real Estate, Central Banks
Abstract
Quantitative easing (QE) is an unconventional monetary policy tool used by central banks to stimulate the economy in times when conventional monetary policy is not sufficient. In the wake of covid-19, central banks around the world has announced significant increases in their QE-programs. This research paper aims to find out whether quantitative easing has any statistically significant effect on the stock prices of the Swedish real estate market. Moreover, it aims to produce an indication of the direction of the real estate stock prices over the year of 2020. To those ends, a combination of statistical analysis and economic theory is used. We estimate three Autoregressive Distributed Lag (ARDL) models. For a chosen model, an out- of-sample prediction is carried out as a way to model future stock price movements. We conclude that quantitative easing indeed has a statistically significant effect on real estate stock prices in Sweden. Furthermore, we estimate that stock prices in the real estate sector will see negative movements during the second and third quarter of 2020, followed by a return to positives during the fourth quarter.
Acknowledgement
This master thesis is the final project on the master program Real Estate and Construction Management at KTH Royal Institute of Technology in Stockholm. As of submission, a two- year journey has come to an end which would not have been possible without involvement of several dedicated professors at the department.
We would like to direct a special acknowledgement to Dr. Han-Suck Song for accepting the role as our thesis supervisor and providing us with his expertise within the fields of statistics and real estate markets. In truth, his constructive feedback has without any doubt improved the quality of this research paper significantly.
Furthermore, we would also like to direct a special thanks to all students attending the opposition meetings during the course of the semester. Having access to multiple views and inputs has been truly beneficial for our understanding of the complex structures of the real estate and financial markets.
Examensarbete
Titel
Författare Institution Handledare Nyckelord
Effects of Quantitative Easing on the Swedish Real Estate Market, an ARDL Approach
Hallsten, Felix & Valdenström, Mikael Fastigheter och Byggande
TRITA-ABE-MBT-20542 Song, Han-Suck
Kvantitativa Lättnader, ARDL, Aktiepriser, Fastigheter, Centralbanker
Sammanfattning
Kvantitativa lättnader (QE) är ett redskap inom okonventionell penningpolitik som används av centralbanker för att stimulera ekonomin när konventionella metoder inte är tillräckliga. I kölvattnet av covid-19 så har centralbanker runt om i världen meddelat kraftiga ökningar i deras program för kvantitativa lättnader. Den här uppsatsen syftar till att ta reda på om kvantitativa lättnader har någon statistiskt signifikant påverkan på priserna för fastighetsaktier som handlas på öppen marknad i Sverige. Därutöver syftar den till att ge en anvisning kring i vilken riktning priserna på dessa aktier kommer röra sig under 2020. För dessa ändamål används en kombination av statistisk analys och ekonomisk teori. Vi estimerar tre Autoregressive Distributed Lag (ARDL) modeller. För en av modellerna görs en out-of- sample prediktion för att modellera framtida prisrörelser på aktiemarknaden. Utifrån våra modeller och analys kan vi konstatera att kvantitativa lättnader har en effekt på priserna för fastighetsaktier. Vidare så estimerar vi att priserna på fastighetsaktier kommer röra sig i en negativ riktning under andra och tredje kvartalet 2020, för att sedan svänga tillbaka till positiva rörelser under fjärde kvartalet.
Förord
Den här uppsatsen är det sista projektet på masterprogrammet Fastigheter och Byggande som undervisas på Kungliga Tekniska Högskolan i Stockholm under våren 2020. I och med inlämning, så lider en tvåårig resa, som inte hade varit möjlig utan support från ett flertal dedikerade professorer på sektionen, mot sitt slut.
Vi skulle vilja rikta ett varmt tack till Dr. Han-Suck Song för att han accepterade rollen som vår handledare. Han-Suck Songs engagemang och djupa förståelse inom områdena statistik och fastigheter har utan tvivel höjt uppsatsens kvalité.
Till sist skulle vi också vilja tacka samtliga studenter som har medverkat under opponeringstillfällen under kursens gång. Genom att kontinuerligt fått feedback och haft tillgång till olika aspekter har vi haft en möjlighet att förstå och redogöra för de komplexa strukturerna på fastighets och finansmarknaden.
1. Introduction 1
1.1. Background 1
1.2. Purpose 2
1.3. Research question 3
1.4. Delimitations 3
2. Theoretical Framework 5
2.1. Quantitative easing 5
2.2. Previous and current asset purchasing programs 5
2.3. How quantitative easing works 7
2.3.1. Portfolio balance channel 8
2.3.2. Signaling channel 8
2.3.3. Exchange-rate channel 9
2.3.4. Liquidity channel 10
2.3.5. Commercial papers 10
2.3.6. Overview of transmission channels 11
2.4. Effectiveness of quantitative easing as a monetary policy tool 12
2.5 Effectiveness of econometric models 13
3. Theory Supporting Choice of Variables 15
3.1. Dependent variable 15
3.1.1. OMX Stockholm Real Estate Gross Index 15
3.2. Independent variables 16
3.2.1. Quantitative easing 16
3.2.2. Risk free rate (short and long-term) 17
3.2.3. Real estate market data 18
3.2.4. Currency rates 19
3.2.5. Inflation 21
3.3. Hypothesis 22
4. Methodology 25
4.1. ARDL-model 25
4.1.1 Requirements 25
4.1.2 Advantages and disadvantages of using an ARDL model 26
4.1.3 Alternative approaches 27
4.2. Augmented Dickey-Fuller test 27
4.3. Optimal Lag Selection 28
4.3.1 Testing for autocorrelation 29
4.4. In-sample Prediction 30
4.5. Out-of-sample prediction / forecast 30
4.6. Data 31
4.6.1 Data collection 31
4.6.2. Data processing 32
5. Result 35
5.1. Testing for stationarity 35
5.2. Optimal Lag Selection 36
5.3. ARDL-models 36
5.4. ACF-test 38
5.5. Model efficiency 39
5.6. Forecast 39
6. Analysis of Findings 41
6.1. Macroeconomic Interpretations 41
6.1.1. Swedish Central Bank QE 41
6.1.2. European Central Bank & Federal Reserve QE 42
6.1.3. Macroeconomic variables 43
6.1.4. Real Estate Yield Levels 44
6.1.5. Real Estate Rent Levels 45
6.1.6. OMX Stockholm Real Estate Index (Lagged) 46
6.2. Limitations of the statistical model 47
6.3. Data reliability 48
6.4. Recommendations on future research 49
7. Conclusion 50
References 52
Data Sources 59
1
1. Introduction
1.1. Background
During recent years, the global economy has been slowing down, a trend derived from several factors such as political instability, armed conflicts, trade-wars, China’s stunted growth potential, and more recently the covid-19 pandemic (World Bank, 2020). In the subsequent low inflation climate, and with investors still repairing their balance sheets from the subprime and eurozone debt crises, the recent turmoil on the market has caused investors to be increasingly risk averse. This, among other factors, has forced several central banks and institutions to lower their growth forecasts for the coming years (International Monetary Fund, 2019).
As central banks all over the world have recognized the low inflation and weak development of the economy, they have tried to stimulate it through various monetary policies. The most common instrument used by the central banks is adjustments of the repo rate. However, in situations where aforesaid conventional monetary policy is not enough, some central banks opt into unconventional policies such as injecting liquidity directly into the financial system by purchasing securities on the open market. This is known as quantitative easing (QE).
The central banks around the world has adopted different strategies regarding their quantitative easing programs. The European Central Bank (ECB) has gradually reduced their open market purchases bringing it to a complete halt in September 2018 but elected to reinstate their QE program a year later (Hartmann & Smets, 2018). The Swedish Central Bank has attempted to phase out their QE and instead rely on the repo rate as the main tool for monetary policies. However, in the wake of Covid-19 and the financial distress it causes, most leading central banks has greatly expanded their QE programs, both in terms of volume and type of securities being bought, to stimulate the economy. For example, the Swedish Central Bank has announced their intention of buying corporate bonds and FED is contemplating whether equities should be included as well. Some central banks have even combined QE with even more unconventional monetary policies that has yet to be tested in a larger scale, such as helicopter money.
Even though quantitative easing is a new concept, several studies has been conducted to map out how these unconventional monetary policy programs affect the economy. However, only one research paper has been found that investigates the effects of QE on the real estate market.
The findings of the paper indicated that both the equity and debt sides of the real estate market in the U.S. has seen upswings because of quantitative easing (Gabriel & Lutz, 2015).
2 With increased globalization stemming from technological advancements and eased locomotion of people and capital, it is not solely the domestic central bank’s monetary policies that will affect the respective markets of each nation. Financial systems all over the world are now closely intertwined and connected, therefore, changes in one central bank’s monetary policies, often have a reach that supersedes its national or regional border (Bordo & Taylor, 2017). Consequently, a question is raised of how the markets react to conflicting views on monetary tools such as quantitative easing between different central banks. Figure 1 below showcases the value of the Swedish Central Bank, ECB and FED’s quantitative easing program as of March 2020 in terms of billions of USD.
Figure 1. Historical Quantitative Easing in USDbn(1)
Source: SCB, ECB, FED
Notes: (1) FX-rates as of 02/04/2020
1.2. Purpose
The aim of this thesis is to investigate how the monetary policy strategies of the Swedish Central Bank, European Central Bank and the Federal Reserve will affect the Swedish real estate market using the broad OMX Stockholm Real Estate Gross Index as a proxy for the market. We suggest that changes in QE policies, which ultimately is a change in how central banks plan to purchase bonds on the market, might have a very direct and visible impact on markets in close collaboration with the debt market. Real estate as an investment opportunity is known for being relatively high levered. Furthermore, the real estate market is unique in its reach. Changes on the market can easily escalate to a nation-wide crisis since it is not only real estate investors which are impacted by the changes, every homeowner will be impacted by changes on the market as well. Moreover, in the Swedish Central Banks Financial Stability report of 2019, the Swedish banks’ lending to the commercial real estate sector is said to pose one of the greatest threats towards financial stability in Sweden (Swedish Central Bank, 2019).
3 Finally, the real estate market is, in most countries around the world, a market encompassing great asset values. Meaning that even small changes on the market will have a significant impact on the countries’ asset bases.
Therefore, we believe that investigating the impacts of QE on the real estate market in Sweden could be of great importance to a wide audience, everyone from distinguished real estate investors and their tenants to people trying to start their housing career by purchasing their first home.
1.3. Research question
The use, effects and effectiveness of quantitative easing as a monetary policy tool has previously been analyzed by several different institutions, organizations and researchers, mainly in the U.S., Japan and England. Effects on the economy, and thereby also the effectiveness of QE, has been studied in minutiae, but few studies have tried to investigate the effects it has on the real estate markets around the world. At the time of writing this paper, no other research has been found on the effects of quantitative easing on the Swedish real estate market. This realization, coinciding in time with the massive quantitative easing programs launched by central banks around the world in the wake of covid-19, led to two research questions:
RQ 1. Is there any statistically significant effect from quantitative easing on stock prices in the Swedish real estate market?
RQ 2. Given the recent announcements of expanding quantitative easing programs, how will the prices of real estate stocks move during 2020?
1.4. Delimitations
To be able to grasp, and conduct a thorough analysis of the problem at hand, several delimitations has been required. The focus in this research will be on listed real estate companies in Sweden. The reasons for this are primarily that Sweden is renowned for being one of the most transparent countries in the world which makes us believe that credible data could potentially be relatively easy to acquire. Especially important is the “Policy of Public Access” (Sw. Offentlighetsprincipen) stating that public entities are obliged to share information with the general public. Since data regarding monetary policies are mostly covered by public entities, this further strengthens our belief that gathering data will not pose as big of a challenge if the analysis is conducted in Sweden. Furthermore, Sweden is also interesting since the country's central bank has a competing view of how to exercise monetary
4 policies compared to ECB. This could potentially open up for interesting comparisons and discussions.
A sector delimitation has been layered on top of the geographical delimitation. When analyzing the impact of different QEs, the commercial real estate market and the companies operating there will be in the center. The reasoning behind choosing specifically real estate is its unique ties to the debt market and the society. A more elaborate motivation of the choice of sector can be found in sector 1.2 Purpose.
The last delimitation that has been exercised in the research is the choice of comparison institutions. When choosing which institutions to include, the choice criteria has been that the monetary policy of the chosen institution is expected to affect the Swedish real estate market.
Thus, the inclusion of the Swedish central bank is given. However, as described earlier, monetary policies of some institutions have effects superseding their national or regional borders. Consequently, we suggest that due to the size and interconnectivity of the ECB and the FED, monetary policy changes carried out by them have ripple effects across the world economy reaching Sweden and must therefore be included in our research.
5
2. Theoretical Framework
2.1. Quantitative easing
Before the financial crisis in 07/08, the foundation that the central banks’ monetary policies rested upon appeared sound and sturdy. The central banks used their respective repo rates to keep inflation low and stable, and generally to a wide success one might add. However, the monetary policies of that time were limited to inflation targeting and had less focus on financial stability. Consequently, when the asset market bubble burst, central banks found themselves in a situation where interest rates could not be lowered anymore and the conventional monetary policies proved insufficient; they found themselves in the liquidity trap as introduced and theorized by Hicks (1937), and popularized by Krugman et al. (1998). This meant that they had to look at other options for trying to bring the economy out of the following recession; they turned to unconventional monetary policies. (Joyce et al., 2012) There are many forms of unconventional monetary policies, such as negative interest rates or adjusted inflation targets. The most common one however is known as quantitative easing, which is defined by Blinder (2010) as “changes in the composition and/or size of a central bank’s balance sheet that are designed to ease liquidity and/or credit conditions”. In more practical terms, when the literature refers to quantitative easing it means the central banks’
purchases of securities on the open market. Although some central banks have been using quantitative easing since the start of this millennium (Bank of Japan), the unconventional policy first gained track after the subprime crisis.
2.2. Previous and current asset purchasing programs
The Swedish Central Bank was one of many central banks that officially launched their QE in early 2015 as a way to prevent sub-zero inflation in the eurozone economy which was still suffering as a result of the European debt crisis (Swedish Central Bank, 2020a), increased competition from developing countries and declining domestic retail markets due to the popularity and globalization of e-commerce. Although, the Swedish Central Bank had previously bought a small portfolio of assets in 2012, with part of the motive being that they wanted to be prepared for the larger scale asset purchases that were being considered and would later follow. As of March 2020, the Swedish Central Bank’s balance sheet contained securities bought for monetary policy purposes of approximately SEK 380 bn. Historically, the Swedish Central Bank has limited their QE to government bonds in order to minimize risk.
However, in the wake of covid-19, they have announced that they will purchase assets worth
6 an additional SEK 300 bn during 2020, which may include municipal and corporate bonds as well as Mortgage-Backed Securities if necessary (Swedish Central Bank, 2020b).
The European Central Bank has been purchasing securities since 2009 in the following programs (Hartmann & Smets, 2018):
● Covered Bond Purchase Programme (CBPP) - 2009
● Securities Markets Programme (SMP) - 2010
● Covered Bond Purchase Programme 2 (CBPP2) - 2011
● Outright Monetary Transaction programme (OMT) - 2012
● Asset-Backed Securities Purchased Programme (ABSPP) - 2014
● Covered Bond Purchase Programme 3 (CBPP3) - 2014
These earlier asset purchasing programs are not defined as quantitative easing by the ECB.
These purchases were made with different intentions than to inject liquidity into the system, and in some cases, they were sterilized from the ECB’s balance sheet. The ECB officially launched their QE operations in early 2015, in the wake of the eurozone debt crisis and a prolonged period of low inflation in a low interest rate climate. These new types of purchases were made to mitigate the risk of deflation, and injected liquidity into the financial system with the intention of increasing spending and growth, and thus increase inflation (ibid.). In response to the covid-19, the ECB announced that they will be extending their quantitative easing efforts by an additional EUR 870 bn, and that they are making available as much as EUR 3,000 bn (European Central Bank, 2020a). As of late March 2020, the ECB’s balance sheet contained securities bought for monetary policy purposes of approximately EUR 2,700 bn.
Similar to the ECB and the Swedish Central Bank, the Federal Reserve initially launched their quantitative easing in a period where their repo rate (federal funds rate) was close to zero in combination with a weakened economy. For the FED, this happened in the wake of the financial crisis of 2008. Starting in December that year, they launched their first Large-Scale Asset Purchasing Programme (LSAP1) which lasted about 18 months. They then continued to increase the size of their balance sheet until March 2014 through reinvestments of principals, a maturity extension programme and a second and third large-scale asset purchasing programme (LSAP2 and LSAP3). In 2017 however, they started carrying out balance sheet normalization, by reducing the reinvestments of principals from existing holdings (Federal Reserve Bank of New York, 2020).
As of late 2019, the FED started increasing the size of their balance sheet again, in order to find the proper amount of cash reserves required to meet the banks demand for cash.
Moreover, in response to the covid-19, additional expansions of the balance sheet were
7 announced as the FED first communicated that they would be purchasing an additional USD 700 bn worth of treasuries and agency mortgage-backed securities Federal Reserve, 2020a).
This amount limit would later be replaced by the phrasing “... in the amounts needed to support smooth market functioning and effective transmission of monetary policy to broader financial conditions.”, meaning that there would basically be no limit on the amount of purchases (Federal Reserve, 2020b). Figure 2 below depicts the central banks’ quantitative easing programs relative to the gross domestic product of the respective region, starting from year-end 2006 and onwards as well as a forecasting period of one year. The forecasting period assumes maturity extensions, straight-line purchases of assets and uses GDP projections from the National Institute of Economic Research (Sw. Konjunkturinstitutet). Since the actual effect of covid-19 on the global economy is yet to be unveiled, the underlying variables for the forecasting period is subjected to major uncertainties. Thus, it should be interpreted as a rough estimate.
Figure 2. Quantitative Easing / GDP - ratio
Source: SCB, ECB, FED, National Institute of Economic Research (Sw. Konjunkturinstitutet)
2.3. How quantitative easing works
When Central Banks conduct open market purchases, i.e. purchases government or commercial bonds on the open market, it will affect markets and the economy through different channels. Since quantitative easing is a relatively new concept, no general consensus for labels on these channels exists and some institutions have divided/merged channels together. Below, the channels will be defined based on labels given by the Swedish Central Bank.
8 To fully understand how the purchases and the effects on the market are intertwined, it is important to acknowledge the general formula for how long-term bond rates are computed:
𝐿𝑜𝑛𝑔 𝑡𝑒𝑟𝑚 𝑔𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑏𝑜𝑛𝑑 𝑟𝑎𝑡𝑒𝑠 = 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑠ℎ𝑜𝑟𝑡 𝑡𝑒𝑟𝑚 𝑏𝑜𝑛𝑑 𝑟𝑎𝑡𝑒𝑠 + 𝑝𝑟𝑒𝑚𝑖𝑢𝑚 (1) The long-term government bond rates are mirrored by investing in short term government bonds on an ongoing basis over the maturity of the long term. There are, however, several reasons why the rate of the long-term bond diverge from ongoing investments in short term bonds, to compensate for the additional risk a premium is added. Short term interest rates are highly correlated with monetary policy decisions, therefore changes in monetary policies and/or changes in expectations on future monetary policies will also impact long term interest rates (Swedish Central Bank, 2015).
2.3.1. Portfolio balance channel
The portfolio balance channel explains how Central Banks decision of buying government bonds will affect other asset classes and maturities. When the government purchases long- term government bonds, the supply of bonds on the market will fall. Given that some investors are reluctant to switch to other assets (preferred habitat investors), the premium on long-term bonds will be compressed (Swedish Central Bank, 2015). Quantitative easing’s effect on the premium variable in the equation for long-term bond rates is sometimes referred to as a channel of its own (the premium channel).
When the return on government bonds is reduced, investors are more inclined to seek returns from other asset classes. This means that the yield compression of government bonds will put pressure on yields for other assets as well (Gern et al., 2015). The effect of the portfolio balance channel is determined by how the seller of government bonds react after the sale. Proceeds from the sale will, if the seller considers money as an imperfect substitute, reinvest the money in other asset classes e.g. government bonds with different maturities, corporate bonds or equities. The seller in that transaction will face the same decision as the first seller, and the rebalancing of portfolios will continue until investors are indifferent regarding money supply and assets. Ultimately the rebalancing process will increase asset prices and reduce yields on the market which means lower cost of borrowing for firms and households which stimulates inflation and real activity (Joyce et al., 2017).
2.3.2. Signaling channel
The signaling effect is connected with the balance sheet risks of central banks. To purchase government bonds, the central bank creates electronic money which they deposit into the
9 Swedish banks’ accounts in the central bank’s system. The money is used by the banks to purchase bonds on behalf of the Central Bank which will inflate the banks’ surplus (which the Swedish Central Bank pays interest on)1. If the interest they pay on their surplus is higher than the return they receive on the government bonds purchased on the open market, a loss will incur (Swedish Central Bank, 2015).
Therefore, when the central bank purchases bonds, it materializes as a forward guidance that the central bank will opt into an expansive monetary policy during the maturity of the bond.
Having the long-term bond rate formula in mind, this will lower short term bond rate and in turn also the long-term bond rate. Since the credibility of a future expansive monetary policy is strengthened, the signaling channel addresses inflation and growth (Krishnamurthy &
Vissing-Jorgensen, 2011). The belief in an expansive policy reduces market uncertainty and volatility, and thus creates a safer investment climate. This is sometimes referred to as the confidence channel.
2.3.3. Exchange-rate channel
As previously mentioned, quantitative easing can be interpreted as the central bank communicating that expansive monetary policies will continue, keeping the repo-rate low. By doing so, mechanisms that is directly linked to the repo-rate will be impacted, for example the exchange rate (European Central Bank, 2017). A country’s exchange rate is dependent on the current and expected trajectory of the risk-free rate since the return on government bonds is correlated with the risk-free rate. If the repo-rate is low, the return on government bonds will be low which in turn reduces the demand for the local currency (Swedish Central Bank, 2018a).
Having a relatively weak currency will not affect a domestic investor that deploys capital in a domestic asset. However, foreign investors will be more inclined to deploy capital in the country in question since they are able to buy assets cheaper than before. The outcome is therefore that asset prices increase due to the increased demand for assets, stemming from a low exchange rate. The relationship is usually referred to as the exchange-rate channel when talking about the effects of quantitative easing (ibid).
1 The interest rate paid is determined by the repo rate, currently the repo rate stands at 0% which means that the Swedish Central Bank doesn’t pay any interest on the banks’ overnight deposits
10
2.3.4. Liquidity channel
As mentioned in the previous section, purchases of bonds are conducted using newly created electronic money which will inflate the volume of capital available. This means that the quantitative easing program will increase the liquidity in the system. In extension, this will impact pricing of liquid assets such as Treasuries. Liquid assets have a price premium compared to illiquid assets due to having the option of being readily converted into cash. When the overall liquidity in the system is augmented, investors willingness to pay extra for liquid assets is diminished. This implies that quantitative easing, through the liquidity channel, will reduce prices and increase yields on liquid assets such as Treasuries (Krishnamurthy &
Vissing-Jorgensen, 2011).
The Swedish Central Bank and Bank of England argues that the liquidity channel also materializes in banks being more inclined to lend money to corporations and households.
Increased lending is sometimes referred to as a channel of its own, labeled the Bank Lending channel. When the liquidity in the banking sector increases, the possibility for a bank to end up with a deficit in their overnight deposit account in the Swedish Central Bank is reduced. If a bank has a deficit in their account, they will borrow from a bank with a surplus which they pay interest on. This means that when banks have access to more capital, the chance of being liable to pay interest on a deficit is reduced which in extension causes banks to increase lending (Swedish Central Bank, 2015).
2.3.5. Commercial papers
An alternative to buying government bonds, the central bank can opt into purchasing commercial papers. In contrast to e.g. the Federal Reserve, European Central Bank and Bank of England, the Swedish Central Bank has not yet adopted this strategy but have said that they will do it in the new QE-program launched in the wake of the covid-19 virus if necessary.
Buying commercial papers means a direct intervention in asset classes with a higher risk. It results in a positive financing effect since interest rates will fall on the specific market, and secondly the central bank will act as a stable buyer for the issuers of the papers. Actively purchasing commercial papers therefore requires more in-depth research of how different markets are interlinked and how it affects the economy.
By being able to secure the buy-side of the instruments, purchasing commercial papers can be used to stabilize certain markets that is deemed essential for the society (Swedish Central Bank, 2018). A frequently used example is when the Federal Reserve started to purchase mortgage backed securities (MBS) during the subprime crisis in 2008 when that market was
11 trembling. If the market would halt completely, it could potentially have severe consequences for credit accessibility amongst households in the US (Fawley and Neely, 2013).
2.3.6. Overview of transmission channels
The goal of quantitative easing is to stimulate economy growth and inflation through increased spending and income. The theory suggests that this is achieved through a combination of several channels of effects. These effects can be divided into three categories: behavioral effects, pricing effects and money supply effects. Behavioral effects are seen through the signaling channel which states that the QE indicates to the public that the Central Bank will have an expansive monetary policy for at least the duration of the bonds purchased and thus reduces market volatility. The effect on pricing is also seen through the signaling channel as well as through the portfolio balancing and the liquidity channels. Relative asset prices increase, and yields compress as the asset purchasing from the central bank creates a manufactured demand. The domestic currency is weakened which increases demand from international investors and thus drives prices up through the exchange rate channel. The asset purchasing is made with newly created money which increases the total money supply and the liquidity in the financial system. Contradictory, the increased liquidity also pushes prices down on liquid assets, as the increased liquidity reduces the liquidity premium paid for liquid assets.
However, the overall outcome of these effects is suggested to be increased inflation and growth of the economy through increased spending and income. Figure 3 below illustrates the transmission channel and their effects:
Figure 3. Mapping of transmission channels
Source: G. Haldane et al., 2016 (Modified by the authors of this research paper)
12 It is however important to have in mind that even acknowledging all transmission channels, it is hard to capture the entire effect of quantitative easing since it is subjected to interference from other policies both domestically and internationally. The issue is highlighted in a statement originating from the report “The financial market impact of quantitative easing in the United Kingdom” by Joyce et al,. (2011)
“Judging the impact of QE in stimulating the macroeconomy is difficult, as the transmission mechanism may be subject to long lags, and it is hard to measure the specific contribution of the Bank of England Monetary Policy Committee’s (MPC’s) asset purchases, given the influence of other policy measures and other economic developments in the United Kingdom and internationally”
2.4. Effectiveness of quantitative easing as a monetary policy tool
During recent years, the financial markets has been stimulated by high economic growth in combination with low interest rates. This has materialized in high returns, especially for sectors which are dependent on external sources of funding. In an article written by Louis Landeman, the head of Credit Analysis at Danske Bank, he emphasizes that bank loans, which has acted as the core of external financing in Sweden, hasn’t been sufficient to cover the needs of funding for the growing number of large real estate companies in Sweden. Instead, real estate companies have complimented bank loans with issuance of preference shares and bonds, both in SEK and Euro. By year-end 2018, real estate companies represented 40% of the total corporate bond market in Sweden (Landeman, 2018).
The recent development of Covid-19 has caused a rapid decline in asset pricing which has forced an unprecedented expansion of central banks' QE-programs, both in terms of volume and type of securities bought. For example, the Swedish Central Bank has announced that corporate bonds will be purchased in addition to government bonds, if necessary. The instability on the market has made investors more risk averse, redirecting investments from equities to for example bonds and gold. Due to an increased demand for bonds, bond-yields has fallen to historical lows resulting in a cheaper source of funding (CBRE, 2020). Therefore, analyzing the effect of quantitative easing on Swedish real estate companies is highly relevant due to their exposure to the bond-market and their strong historical growth on the stock market.
The use, effects and effectiveness of quantitative easing as a monetary policy tool has previously been examined by several different institutions, organizations and researchers;
mainly in the U.S., Japan and England. The majority of the research conducted has either had
13 a macro perspective, focusing on the effects on the economy as a whole, or a financial market perspective, focusing on variables such as bond pricing and yields.
The effects on the economy as a whole has been documented by several studies, among them Kapetanios et al. (2012) provided evidence on the macroeconomic impact of Bank of England’s first round of quantitative easing in 2009 by measuring effects on GDP and inflation; the study suggests that QE had an effect on the level of real GDP by 1.5 % and on CPI inflation by 1.25
%. Two similar studies with similar results have been conducted in Japan, where Bowman et al. (2015) concludes a weak positive response on GDP from QE and where Matousek et al.
(2019) in extension concludes a strong positive response to both GDP and inflation from the QE.
On the financial market side, Bernanke et al. (2004) concludes through event studies using VAR-based term structure models that both the changes in relative asset quantities and the expectation of such changes have generated lower yields during the period of quantitative easing in the U.S. and Japan, although the evidence from Japan is less convincing than that from the U.S. from a statistical standpoint. Moreover, in England, several different studies have documented the effects of the QE; among them Meier (2009) and Joyce et al. (2011) which supports the evidence proposed by Bernanke et al. (2004), that long-term government bond yields decline as a result of the QE.
There exists a large research gap when it comes to studies that try to isolate the effects from quantitative easing on the real estate market, which is surprising to us considering how closely linked it is to the bond and debt markets. One previous study on the effects of quantitative easing on real estate markets by Gabriel and Lutz (2015) shows that in the U.S., both the equity and debt sides of the market have seen upsides as a result of the QE launched in the wake of the subprime crisis. Housing and mortgage interest rates dropped, the costs to insure commercial real estate debt and subprime mortgage debt also fell, and the overall housing distress was reduced. On the equity side, which is most relevant to this study, stock market indices related to homebuilding and Real Estate Investment Trusts showed abnormal returns following the QE (Gabriel & Lutz, 2015). This is the only previous study on the effects of QE on real estate that has been found.
2.5 Effectiveness of econometric models
Econometric models are used by companies and government entities to understand relationship and forecast future events. The efficiency of modelling has been a highly discussed topic among prominent researchers since they act as a simplified and subjective version of a very much complex reality (Ouliaris, 2020).
14 Concerns regarding econometric models echoes through the famous research paper “The Financial Instability Hypothesis” by researcher Hyman Minsky (1992). He states that economic models are efficient when history repeats itself. However, due to increased government interventions and a financial system which keeps getting more complex, market behavior is undergoing a constant change (Minsky, 1992). Furthermore, in all practical cases the true model of an empirical relationship is unknowable, an issue which is further intensified by relevant data being a scarce resource. In reality, this means that all models are wrong, but they keep being used since they could contain an indication of what the actual relationship could look like and since there are no perfect alternative approach (Phillips, 2003). The fact that econometric modelling is used even though it can be considered a flawed science is captured in the following quote from the famous economist John Maynard Keynes:
“It is better to be roughly right, than precisely wrong” - John M. Keynes
15
3. Theory Supporting Choice of Variables 3.1. Dependent variable
3.1.1. OMX Stockholm Real Estate Gross Index
When deciding on a suitable index to measure effects of quantitative easing on Swedish real estate companies, the choice was between an index measuring changes of underlying property value or a stock index. The issue with an underlying property value index is that real estate companies report property value on a quarterly basis, while stock data is available on a daily basis and have generally been computed during a longer period of time. Thus, to ensure that we work with high quality data, a real estate stock index was chosen.
Nasdaq Nordic computes sector indices with a geographical delineation, the OMX Real Estate GI and OMX Real Estate PI indices. Comparing a Gross Index with a Price Index, the gross index acknowledges both difference in stock price and dividend payouts, while a price index solely focuses on performance of the stock price. Thus, to get a better view of real estate companies, the dependent variable in the statistical analysis will be the gross index, OMX Stockholm Real Estate GI (SX8600GI). The index tracks the performance of the 32 largest real estate companies listed on the Stockholm stock exchange. Some of the companies has listed several securities, like preference stocks and shares with different voting rights which is included in the index, making the total number of securities tracked by the index 40 (Nasdaq Nordic, n.d.).
Historical movements of the index since the start of the third quarter in 2006 can be seen in Figure 4 below. The real estate market has outperformed the broad OMX Stockholm 30 Index over the whole period. Following developments of the covid-19 pandemic in the first quarter of 2020 however, the real estate market dropped 46 % from the top to its lowest closing whilst the OMXS30 in comparison only dropped 30 %.
16 Figure 4. OMX Stockholm Real Estate GI Index (06/07/2006=100)
Source: Nasdaq Nordic
3.2. Independent variables
3.2.1. Quantitative easing
Currently, there are several central banks that has adopted quantitative easing as part of their monetary policies. As mentioned in section 1.4, this research paper has been limited to analyze the central banks that is expected to have the most influence on the Swedish real estate market.
Therefore, the central banks being analyzed is the Swedish Central Bank, The European Central Bank and The Federal Reserve.
There has been an ongoing globalization during the past decade, where countries and companies are increasingly affected by global events and policies. One could argue that large corporations, such as those real estate companies included in the OMX Real Estate GI index, are even more affected by the globalization since it is common to have foreign equity and/or debt owners.
As of April 2020, there are several central banks that has adopted quantitative easing as part of their monetary policy. In addition to the central banks included in this research article, Bank of England, Swiss National Bank, Bank of Japan and Bank of Korea has also opted into more unconventional monetary policies, with several more considering a launch in the wake of Covid-19. As previously mentioned in the delineation section, the reasoning behind selecting ECB and FED as comparable to the Swedish Central Bank is due to the raw size of their individual quantitative easing programs and their strong influence on the global economy.
17 To be able to measure the size of the quantitative easing programs, we have elected to analyze the value of monetary policy instruments held outright, similar to the methodology used by Balatti et al., (2018). The balance sheet of central banks includes a vast number of accounts which includes different operations. To ensure that the data extracted from each central bank is accurate and comparable, we have been in contact with employees at the monetary policy or data processing department for each bank.
3.2.2. Risk free rate (short and long-term)
Investment decisions and returns on the market is affected by several factors, one of them being the risk-free rate of return. A government bond of equal maturity as the alternative investment opportunity is market standard for determining the risk-free rate on a long-term basis, and the interest rate of the government bond is controlled by the central banks through their repo rates (the short-term risk-free rate) by adding a liquidity premium and expectations about the future to that rate.
The real estate sector is not exempt from being affected by the risk-free rate. How real estate appraisers value a property is by discounting the free cash flow from the underlying property.
How the discount rate is determined is through the following formula (Fastighetsnytt Förlags AB et al., 2015):
𝐷𝑖𝑠𝑐𝑜𝑢𝑛𝑡 𝑟𝑎𝑡𝑒 = 𝑟𝑖𝑠𝑘 𝑓𝑟𝑒𝑒 𝑟𝑎𝑡𝑒 + 𝑟𝑖𝑠𝑘 𝑝𝑟𝑒𝑚𝑖𝑢𝑚 (2)
As such, if the central bank lowers the risk-free rate like they did during the financial crisis of 2008-2009 and the eurozone debt crisis in 2011-2015 (see figure D) then, ceteris paribus, the value of properties will increase and vice versa. This means that when analyzing returns on the real estate market or real estate companies, the repo rate and the government bond rates are important factors to consider.
Since the government bond rates are derived from the repo rate, the two variables have a very strong correlation. Including them both in the same model is likely to cause problems with multicollinearity (Jalil et al., 2010). This can lead to erroneous inferences such as wrongful estimation and specification of the independent variables (Farrar & Glauber, 1967). The trajectory of the Swedish repo rate since 2006 can be seen in Figure 5 below:
18 Figure 5. Swedish Repo rate
Source: Swedish Central Bank
3.2.3. Real estate market data
A factor reflecting the climate on the real estate market is the yield, sometimes referred to as the capitalization rate. The yield is a measurement of how high returns the investor demand from the property, the formula for calculating property yield is as follows:
𝑅𝑒𝑎𝑙 𝑒𝑠𝑡𝑎𝑡𝑒 𝑦𝑖𝑒𝑙𝑑 = 𝑁𝑒𝑡 𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝑖𝑛𝑐𝑜𝑚𝑒 (𝑠𝑢𝑏𝑠𝑒𝑞𝑢𝑒𝑛𝑡 12 𝑚𝑜𝑛𝑡ℎ𝑠) / 𝑃𝑟𝑜𝑝𝑒𝑟𝑡𝑦 𝑝𝑟𝑖𝑐𝑒 (3) Where net operating income reflects the cash-flow from a specific property, meaning the rental income less vacancy loss and property related costs. Rearranging the formula as net operating income divided by yield, appraisers can use yields on the market to bridge cash flows from a property to a property value. This is a frequently used method among real estate appraisers, which means that fluctuations in yields and factors impacting the net operating income will affect the value of properties in Sweden, and in extension real estate companies.
The real estate index chosen for our dependent variable constitutes of 30 unique Swedish real estate companies active in different segments within the real estate market. To capture movements in the different segments, data on yields and rental levels has been extracted for offices, retail, industrial and housing. An argument can be made for also having a geographical split of yields and rental levels since variables are dependent on location. However, for this research it is not the actual values that are the most important but the movement of the variables, and it is our belief that the yields in different major cities in Sweden is highly correlated. In addition, by excluding a geographical split, the number of independent variables is lowered meaning higher degrees of freedom in the statistical analysis. As such, the yields
19 and rental levels used as independent variables in this research paper refers to the Stockholm region and can be found in Figure 6 and 7 below.
Figure 6. Real Estate yield
Source: Datscha
Figure 7. Real Estate rental level
Source: Datscha
3.2.4. Currency rates
Generally, a weak SEK will attract more international capital and vice versa. If the SEK is weak, international investors realize that real estate in Sweden is purchasable at a discount due to the exchange rate. This effect is observable when looking at the growing number of foreign investors of Swedish real estate during recent years when the SEK has been relatively weak.
The total volume of commercial real estate transactions in Sweden for the year 2019 was SEK 218 billion, out of which 35 % was international investors, this level has increased from approx.
20 12 % in 2010 (Savills Sverige, 2020). Meanwhile, the KIX has increased from 108 to 123 over the same period, indicating a substantial weakening of the SEK (see figure E below). The KIX- index is a geometric chain-index computed by the Swedish Central Bank which measures how competitive the SEK is compared to several other currencies (Swedish Central Bank, 2020c).
The currencies included in the index is designed to encompass our greatest trade partners, and constitutes of OECD countries, Russia, Brazil, China and India (National Institute of Economic Research, 2007).
An influx of international capital, driven by an increased demand for Swedish assets, will affect the Swedish real estate market since asset values will be inflated. Another factor to acknowledge is the cost of construction since construction companies import materials and services for their projects. If the domestic currency is weakened, then the cost of importing increases (Swedish Central Bank, 2018c), potentially decreasing profits for real estate companies. Moreover, Hutchings (1997) showed that for cross-border investors and occupiers, currency rates will have a large impact on returns and costs respectively, further pointing towards the necessity of including currency in our model.
A factor that however can interfere with the relationship between foreign investments and currency is financial crisis. During the subprime crisis in 2008, the USD significantly appreciated, meaning that SEK was relative to USD weakened while at the same time foreign investors were increasingly risk averse, meaning a negative correlation between a weakened SEK and foreign investments. There are several theories as to why the USD was strengthened during the subprime crisis. One theory is that when crisis occur, investors tend to divert capital to less risky assets with US government bonds being highly regarded. Thus, increasing the demand for USD which causes an appreciation (Federal Reserve Bank of Boston, 2018).
Another theory is that when certain American assets decline in value, like the mortgaged- backed securities during the crisis in 2008, other dollar assets can serve as substitutes in the provision of liquidity services (Rose & Spiegel, 2012). Figure 8 below depict the share of foreign investors of Swedish real estate assets in comparison to the Swedish currency index KIX.
21 Figure 8. Swedish FX-index (KIX) in relation to percentage of foreign investors
Source: Swedish Central Bank and Datscha
3.2.5. Inflation
Inflation is the general increase of overall prices for goods and services over time and its level is measured by the central banks through evaluation of the aggregate of several different price indices (Federal Reserve, 2020c). In Sweden, inflation is measured by the Consumer Price Index (CPI). The Swedish central bank provides two indices: KPI and KPIF. The difference between KPI and KPIF is that KPIF disregards the effect of changes in mortgage interest rates for households. The target inflation rate used by the Swedish Central Bank was a KPI growth of 2% until 2017 when it was changed to a KPIF growth of 2% (Swedish Central Bank, n.d.). In this research paper, KPI is used (denoted as CPI from here on) as a measurement of inflation since it has been used as the target for most of the years being analyzed. Figure 9 below illustrates the target and actual CPI growth since 2006:
Figure 9. Swedish Consumer Price Index (CPI)
Source: Statistics Sweden and Swedish Central Bank
22 All investments are affected by inflation due to its effect on the rate with which the investment is discounted over time (discount rate). A higher inflation rate results in a higher discount rate, and consequently a lower return on the investment (Fama & Gibbons, 1982). This has led to the desire for investors to hedge against expected inflation increases. An inflation-hedge is an asset class with returns that are strongly, positively correlated with inflation. Commercial real estate has long been regarded as one of these asset classes. Hence, in higher-inflation climates, real estate becomes a more attractive asset class with higher capital allocation as a consequent.
The main reason behind the effectiveness of real estate as an inflation-hedge, is the structure of the lease between owner and tenant (Peyton, 2011). Usually, the rent levels paid by the tenant are indexed to inflation, in addition, utility and property tax expenses are often transferred to the tenant in the lease. Thus, returns on real estate investments are affected by inflation in two opposite directions: positively through the effect on the net operating income stemming from the lease-structures, and negatively through effect on the discount rate. The effectiveness as an inflation-hedge is then determined by which effect is stronger. Historical data suggests that the effect on the net operating income is stronger and that commercial real estate is an effective inflation-hedge (ibid.).
3.3. Hypothesis
This section serves as a forum for the authors of this research paper to present their views on how they believe that the independent variables will affect the dependent variable. The content is mainly focused on the direction of the effect, in other words if the beta-estimate will be positive or negative.
Lagged OMXS REGI
The Autoregressive part of the ARDL-model refers to using a lagged version of the dependent variable to act as an independent variable in the regression. Whether lagged stock prices can be used as an explanatory power for future stock prices has been a widely discussed topic, sometimes referred to as technical analysis of stocks.
We believe that when there are high market sentiments with low volatility, technical analysis could deem to be a powerful tool for predicting future stock prices since prices are generally trending steadily upwards. However, in periods of high volatility and low market sentiment, technical analysis should have a lower explanatory power and be used with caution. These views are in line with several research articles, for example “Sentiment and Effectiveness of Technical Analysis: Evidence from the Hedge Fund Industry” by Smith et al., (2016).
23 Since the Swedish stock market has seen relatively high market sentiment during a majority of the time period being analyzed, the lagged version of the OMX Stockholm Real Estate Gross Index should prove to be a statistically relevant variable with a positive correlation with the dependent variable.
Quantitative Easing variables
Since the quantitative easing programs are designed to boost the market in times of a struggling economy, the hypothesis for the effect of the central banks quantitative easing programs are the same. Based on the description of the quantitative easing programs presented above, these variables are thought to have a positive effect on the dependent variable.
Risk-free rate of return
The risk-free rate of return is divided between short- and long-term interest rates. The central banks control the short-term interest rates using the repo-rate, expecting the long-term interest rates to respond accordingly (see formula 1). Therefore, both the repo-rate and the 5- year government bond rate has been included as variables in this research paper.
The repo rate is regarded as a more conventional form of monetary policy and an instrument used by the central banks to impact the market. Being directly linked to asset valuation, interest paid on debts as well as the currency rate, we believe that the repo-rate, and in extension the 5-year government bond rate, could prove to be two of the most influential variables in the model. Based on findings presented in section 3.2.2, we believe that both the repo-rate and the 5-year government bond rate will have a negative correlation with the dependent variable. However, due to the correlation between the repo-rate and the 5-year bond rate, one can argue that a model with both variables could be suffering from problems with multicollinearity.
Inflation (CPI)
The inflation plays an important role when assessing asset values since it increases the discount rate. Therefore, inflation causes decreased asset values for all sectors including the real estate sector. At first glance, this means that it is justifiable to imagine a negative correlation between inflation and the OMXS REGI index. However, as concluded in 3.2.5.
previous research indicates that real estate can be used as an inflation hedge due to the nature of rental contracts. In most cases, a rental contract is partly or fully indexed against changes in inflation, meaning that if the inflation increases so does the rent paid by the tenant. Thus,
24 our hypothesis is that the negative impact of inflation will be partly offset by the hedge, meaning that the variable should have a marginally negative impact on the index.
Currency rate (KIX)
The importance of including a currency index has been mentioned in section 3.2.4 since it facilitates reallocation of international capital to domestic market. In this report, the KIX- index is utilized to measure the competitiveness of the Swedish Krona. If the KIX-index increases, that means that SEK is weakened in relation to the currencies of Sweden’s most important trade partners.
Based on facts presented in section 3.2.4. we believe that there should be a positive relationship between the KIX-index and the dependent variable since a weak SEK will result in increased demand from international investors while Swedish investors are not directly affected. However, some of the effect could be diluted since domestic investors will experience higher competition for assets and potentially reduced profits stemming from higher construction costs.
Rent levels
Regarding rent levels, our hypothesis is quite straightforward. We believe that if rent levels increase, that should mean that properties generate higher income which results in higher profitability for the real estate companies. Higher cash-flows could also mean that real estate companies could receive more favorable loan terms, which for a highly levered sector will significantly reduce costs. Therefore, we believe that there should be a positive correlation between rent levels and the dependent variable.
Yield levels
The arguments for why yields are an important variable is the same as for rent levels since these two variables are interlinked. However, the difference is that a decrease in yield levels will cause an increase of property values and vice versa. As such, due to the nature of yields there should be a negative correlation between yields and the dependent variable.
25
4. Methodology
This chapter provides a walkthrough for the reader of the econometric methodology, data collection and data processing part of this research paper. An autoregressive distributed lag (ARDL) model will be used, and data is collected from what the authors deem as reliable sources.
4.1. ARDL-model
Regression models are a powerful statistical tool which have been widely used for statistical analysis and prediction since the early 19th century. Regression models use two sets of variables, independent variables (input variables) and dependent variables (output variables), where the models are trying to explain the effects that the input variables have on the output variable. Regression models generally have a stationarity requirement, meaning that the variables should have constant mean and variance over time. Autoregression is a type of regression model which aims to model random time-varying processes; specifying that the output variable is linearly dependent on its own previous values (lags). This makes the models very useful in the fields of finance and economics where historical data are of great importance.
The statistical model used to analyze the collected data in this research report is an autoregressive distributed lag (ARDL) model. An ARDL-model is one where the dependent variable is partly explained by lagged values of itself, and partly by values of the input variables as well as lagged versions of these. The general formula with p lags of the dependent variable and q lags of the independent variable, abbreviated as ARDL (p,q), is as follow (Hill et al., 2018b):
yt = β0 + β1yt-1 + ... + βkyt-p + α0xt + α1xt-1 + α2xt-2 + ... + αqxt-q + et (4)
4.1.1 Requirements
An important concept in econometric research and ARDL models is order of integration. Order of integration refers to the amount of differencing required to achieve stationarity in the variable. For example, a variable which is stationary in a non-transformed state is labeled as integrated of order zero I(0), since zero differencing is required. Subsequently, a variable which is nonstationary in a non-transformed state, but that becomes stationary when taking its first difference, is labeled as integrated of order one I(1). Another important concept is cointegration, which implies that there is a stochastic trend relationship between the dependent variable (𝑦𝑡) and independent variable (𝑥𝑡). Meaning that the variables never diverge so far from each other as for the difference between the variables (𝑦𝑡 = 𝛽1+ 𝛽2𝑥𝑡+
26 𝑒𝑡 → 𝑒𝑡 = 𝑦𝑡− 𝛽1− 𝛽2𝑥𝑡) to be nonstationary (the true difference (𝑒𝑡) cannot be observed, so a test for cointegration is carried out on the residuals (ê𝑡) of the regression). Hence, if the residuals are stationary, the variables are said to be cointegrated (Hill et. al., 2018c).
As concluded in section 4.1., the general rule for regression model is to use stationary variables.
However, with ARDL models, it is possible to carry out regressions on nonstationary variables.
If the variables are integrated to the first order I(1) and cointegrated it is possible to estimate models with the level variables without encountering misleading statistical evidence of the modeled relationships. However, if the models are integrated to the first order I(1) and not cointegrated, the model must be estimated with the first differences of the nonstationary variables. Another option when modeling nonstationary time-series is to detrend the nonstationary variables before carrying out the regression. In this paper, a combination of the two latter methods has been used to produce the three models.
4.1.2 Advantages and disadvantages of using an ARDL model
The major disadvantage of the ARDL-models is that interpretations of the estimated beta coefficients cannot easily be made, due to the inclusion of more than one version of each variable. Thus, an ARDL-model is more useful as a tool of estimating the direction rather than the scale of the relationship. However, if the requirements in the previous section for using an ARDL-model is met, there are several advantages of using the model when analyzing time- series data. Below is a list of some of the most notable advantages;
● Since the regressors form a single equation in contrast to solving a set of equations, there are less problems with endogeneity since there is no correlation in the residual term (Anh et al., 2018). In other words, the residuals are assumed to be endogenous (Nkoro & Uko, 2016).
● The ARDL-model can distinguish between level data, I(0), and first-order integrated data, I(1), which models like Johansen & Juselius or Engle & Granger can’t. This means that an ARDL-model works with both types of data or a combination of both. As a result, the need for unit-root tests, such as Augmented Dickey-Fuller, is not as great when employing an ARDL-model (Anh et al., 2018).
● Furthermore, a major advantage of an ARDL-model lies in its ability to identify cointegrating vectors in a situation when there could potentially be several (Nkoro &
Uko, 2016).
● Finally, compared to for example the Johansen cointegration model, the ARDL-model does not require as big of a sample to produce accurate results (Anh et al., 2018).
However, this will not make a difference for this research paper since the timeframe
