University of Amsterdam Masters in Business Economics
Master Thesis 2012/ 2013
Changes in explanatory factors of stocks, bond and unsecuritized real
estate on US REIT returns during- and post- global financial crisis period
of 2007/ 2008 as compared to pre crisis period.
Tom Retera (Student ID: 5979471) 1st Coordinator: P. (Peter) van Gool FRICS Prof. Dr. 2d Coordinator M. (Marcel) A.J. Theebe Dr.
Real estate investments have long been viewed as an integral part of well-diversified portfolio investments, since they offer stable income, typically low correlation with other asset classes and inflation hedge. By implementing the Variance Decomposition Model developed by Clayton and MacKinnon (2003) this empirical research examines the changes in influence of explanatory factors of large cap- small cap growth- and small cap value stocks, bond and unsecuritized real estate returns to REIT return variability, and tests whether it changed subsequent during and after the “Global Financial Crisis” in 2007 - 2008. The results suggests REITs lose their diversification benefits, since REIT returns are far less sensitive to information, specific to the REIT market, after the financial crisis as compared to the pre crisis period. Meaning that REIT returns have become more driven by the same underlying state variables as the other asset classes in the post crisis period. Main driver are the large cap stock related factors which has become more strongly related to REITs in the post crisis period as compared to pre crisis period.
Keywords:
Diversification, Global Financial Crisis, Mixed- Portfolio, Multi-Factor Model, Real Estate Factor, Real Estate Investment Trust, REIT, Rolling Regression, Variance Decomposition,
Document Master Thesis,
Subject Influence of the Global Financial Crisis on REIT Returns; Case of United States
Datum 24th of October
Name student T. (Tom) Retera BSc
Student number 5979471
E-mail address t.p.retera@gmail.com
Telephone number +31 (0) 6 44 88 61 44
Living address Singel 440, 1017 AV Amsterdam
University University of Amsterdam
Education Masters in Business Economics
Master variant Master Real Estate and Finance
Name first Supervisor P. (Peter) van Gool FRICS Prof. Dr.
Name second Supervisor M.A.J. (Marcel) Theebe Dr.
Table of Contents 1. Introduction ... 5 1.1. Relevance/ Motive ... 6 1.2. Research Questions ... 7 1.3. Hypothesis ... 8 1.4. Method ... 9 1.5. Thesis Structure ... 10 2. Background ... 12
2.1. The Structure of REITs ... 12
2.2. Types of REIT ... 14
2.3. Pre-Crisis Period ... 15
2.4. Start and post-crisis period ... 18
2.5. Relationship Between REITs and Other Asset Classes ... 21
2.6. Literature related to the Variance Decomposition Model ... 24
3. Research Methodology ... 27
3.1. Data ... 27
3.2. Variance Decomposition Model, Sub-Periods ... 30
3.3. Rolling Regression... 33
4. Analysing the Performance ... 35
4.1. Time Variation in Factor Sensitivity ... 35
4.2. Relative contribution to the volatility of REIT returns ... 40
4.3. Time Paths of Relative Contribution ... 43
4.4. Stacked Area Time Path ... 47
5. Discussion of Results ... 50
5.1. Contribution of Idiosyncratic Risk on the variability of REIT returns ... 50
5.2. Contribution of Large Cap Stock on the variability of REIT returns ... 51
5.3. Contribution of Small Cap Stock on the variability of REIT returns ... 53
5.4. Contribution of Bonds on the variability of REIT returns ... 54
5.5. Contribution of Unsecuritized Real Estate on the variability of REIT returns ... 54
6. Conclusion ... 56
7. References ... 62
8. Appendices ... 66
8.1. Preliminary research ... 66
8.2. Unsmoothing NCREIF, Example 2010-2012 ... 68
8.4. Return Series for all Asset Classes ... 70 8.5. Returns for each Asset Class, 1989-2012 ... 72 8.6. Time Rolling Window for each Asset Class, 1993-2012 ... 75
1.
Introduction
Real Estate Investment Trusts (REITs), have existed for 50 years, and were created in 1960 to provide Americans from all income levels the opportunity to invest in large-scale, diversified portfolios of income-producing real estate. It started as a passive investment vehicle, which prohibited the operation and management of properties by REITs themselves. Over the years, however, the REIT sector has grown and evolved into a viable and credible investment class due to legislative and tax code changes. (Monroe, 2009; Smotrich et al. 2012). This evolvement started in the early 90s and is known as the ‘modern REIT era’. Since then, REITs have helped to achieve more transparency and liquidity in the real estate sector, supported the US economy by effectively and efficiently moving capital into real estate markets (Geltner, Miller, Clayton and Eicholtz, 2007; Smotrich et al. 2012; reit.com, 2013).
However, the global financial crisis, which emerged in 2006 and then deepened in 2007/ 2008, had a significant impact on the performance of REIT returns. Not only did the underlying properties show a dramatic decline in returns but also in the whole financial market. Since it’s founding, the REIT industry has grown considerably. In 2012 there were 172 publicly trading REITs operating in the US, with an overall market capitalization of US$603 billion. 139 of the 172 REITs in the National Association of Real Estate Investment Trusts (NAREIT) were equity REITs and this group represented more than 90 percent of the NAREIT market capitalization (reit.com, 2013). Furthermore, over the past 10 years (2002-2012), REITs have outperformed the major indices, generating an average return of (13.9%), compared to S&P500 (5.9%), Russell 2000 Growth Stock (8.5%), Russell 2000 Value Stock (10.1%), Bonds (5.4%) and NCREIF (6.1%) (Datasteam, 2013).
The rapidly rising fund flows into the REIT sector and the outperformance of REITs compared to large cap-, small cap growth- and small cap value stock, bond and unsecuritized real estate during this period (2002-2012), made REITs, and the link between REITs and other asset classes, increasingly important and valuable.
1.1. Relevance/ Motive
The major financial changes during the financial crisis, together with the rapidly rising fund flows of the past generation and outperformance of REITs compared to large cap-, small cap growth- and small cap value stock, bond and unsecuritized real estate over de past 10 years (2002-2012), made the recordings of the investor’s choice of the REITs role in the mixed-asset portfolios more significant (Newell, 2011). These recordings are the information guidelines of the portfolio management (Chiang et al., 2008). As investors can purchase a part or all of a real estate portfolio through a variety of vehicles, the question of how performance metrics play a role in the investor’s selection of investment in a mixed portfolio has become more intriguing (Springer 2006).
The diversification benefits of investing in REITs were investigated by Kuhle, as early as 1987. His empirical evidence supports that, to a certain extent, REITs offer benefits of diversification for investors holdings equities. In times of increased volatility, which can be expected during periods of financial downgrading, it is expected that the correlation coefficients will move upwards (Forbes and Rigobon, 2002). Marking the question of how the performance of REIT returns can be explained through factors such as stocks, bonds and unsecuritized real estate and how it changed due to the financial crisis of major importance (Kaymaz, 2012).
1.2. Research Questions
As a reaction to the lack of transparency of the working of REIT returns, Greg Whyte, Analyst, Morgan Stanley stated: “REITs smell like real estate, look like bonds and walk like equity”. Most of the REITs cash flow and earnings are derived from real estate assets, while being traded on organized, liquid stock exchanges, and due to their high dividend payout ratios and relatively constant distributions, the share price may also be influenced by factors affecting bond returns (Anderson et al, 2005). This study aims to enlarge the transparency of the working of the REIT returns. It will be done by answering the following questions:
“Which changes are recognizable in the relationship between REITs performance and the returns of stocks, bonds and unsecuritized real estate during and after a crisis period when compared to the pre-crisis period? Did REIT returns, for example, experienced an increase in common factors with other asset classes, move more like large or small capital stocks and did the link between REITs and unsecuritized real estate or bonds increase or decrease?”
First, literature related to the functioning of REITs during and after the financial crisis is investigated in order to answer sub-question (1): What is the relationship between REITs and other asset classes during and after the crisis when compared to the pre-crisis period? Next, the method and the working of the empirical research method will be clarified, to help to answer sub-question (2): How can the link between REIT, stocks, bonds and unsecuritized real estate returns be best examined? Then the following sub-questions will be answered: (3) Which coefficients of stocks, bonds and unsecuritized real estate factors unfold from the different sub-periods? And sub-question (4) Which outcomes of stocks, bonds and unsecuritized real estate factors unfold from the different
sub-periods during the variance decomposition? With this information it will be possible to compare the coefficients and variance decomposition results for each time period by answering the question: (5) Have the explanatory factors of stocks, bonds and unsecuritized real estate on REIT returns changed during and after the crisis when compared to the period before the crisis? This question will be answered by individually comparing large cap stock, small cap growth stock, small cap value stock, bonds, direct real estate and idiosyncratic risk related factors in the pre-crisis period in comparison to the periods during and after the crisis. Finally, with all the information collected and investigated it will be possible to answer the main research question.
1.3. Hypothesis
Real estate investments are generally expected to act somewhat differently than other investments in a mixed portfolio. In previous literature the investor’s findings, regarding this subject are not in line with each other. While some arguing the concept of diversification died completely (in article of Clayton et al, 2009; Chiang et al, 2013), other state REITs having great diversification benefits in times of financial downturn (Simon and NG, 2009; Meuller and Meuller, 2003; Geltner et al, 2007). Before starting this thesis preliminary research on the correlation between the quarterly returns of REITs compared to stocks, bonds and unsecuritized real estate in the period between 1989 and 2012 was done. Showing significant changes in correlation, when comparing pre during and post crisis periods. It is shown that there is an increase in the correlation coefficients of REITs with the other asset classes when moving into the post crisis period. These correlations measure essentially how much one-asset class moves in proportion to another, and the results, explained in Appendix 8.1, suggest the influence of the financial crisis: a positive significant impact on the factors explaining REIT returns, became notable after the financial crisis occurred. This led to the following hypothesis:
“The role of stocks, bonds and unsecuritized real estate factors in explaining the REIT returns has increased in the periods during and after the financial crisis when compared to the role of these factors in explaining the REIT returns in the pre-crisis period.”
To test this hypothesis a variance decomposition was developed and implemented for REIT returns that divided the REIT return variability into components directly related to stocks, bonds and unsecuritized real estate return indices, as well as idiosyncratic effects. This method is introduced in the following paragraph.
1.4. Method
To help answer the main research question, empirical sub-questions (2), (3) and (4) need to be answered first. To do this a variance decomposition was developed and implemented for REIT returns that divides REIT return variability into components directly related to major stocks, small cap growth stocks, small cap value stocks, bonds and real estate related return indices as well as idiosyncratic (unexplained) effects. The link between REITs, financial assets and real estate returns was measured and tested in order to establish whether it has changed substantially due to the crisis. The study by Clayton and MacKinnon (2003) is the first recognized research in which the variance decomposition was used to separate the REIT return variability. It explored the investment characteristics of equity REITs within a multi-factor model, relating REIT returns to the returns of large and small capitalization stocks, bonds and direct real estate. The time periods which were examined include: the period 1989-2013, the pre-crisis sub-periods 1989-1993, 1994-2001, 2002-2007, the crisis period 2007-2008 and the post-crisis sub-period 2009-2012. By processing the rolling regression it also examines the changing nature of the return process over time. The rolling regression shows an overall
view over a period of 25 quarters. This number of quarters is taken because its ensures the outcomes to be statistical significant. In every new period the returns of a new quarter was added and the returns of the last quarter was eliminated from the sample.
1.5. Thesis Structure
Chapter 2 starts with an overview of the existing literature. It explains the structure and types of REITs, followed by the evolution of REITs, in which the crisis is the main focus. This chapter also contains a discussion of previous studies done on the relationship between securitized real estate and other asset classes, and finally, the results of related studies that used the variance decomposition theory as their methodology are discussed to answer question (1). Chapter 3 answers sub-question (2) and explains the data and the methodology that have been used in this study. An analysis of the performance and answers to sub-question (3) and (4) are discussed in Chapter 4. In Chapter 5, the results, which were presented in Chapter 4, also stated in the Appendices, are discussed and the most plausible explanation of the changes in the explanatory factors is given. In addition, the similarity of the findings when compared to earlier research concerning the behavior of explanatory factors of stocks, bonds and unsecuritized real estate during and after the financial crisis, are discussed in order to answer sub-question (5). Then, in Chapter 6 the conclusion and the answer to the main question is formulated, as well as the recommendations for the portfolio manager on how to react to these findings.
Since the REIT sector started in the US, United States, and has provided us with a broad set of data, the United States is used as reference. The second reason is for using the US as reference, is cause the global financial crisis started in this country and afterwards
spread out over the whole globe. And not al least, using the US data makes it possible to compare figures with earlier research done by Clayton and MacKinnon (2003) and Anderson et al. (2005). Furthermore, this study focuses on equity REITs, as this is the most interesting group in comparison to the other classes of assets. All data return series are explained in detail in Paragraph 3.1.
2.
Background
The review of the relevant literature and studies gives a clear picture of the theoretical background of REITs and provides an opportunity to form expectations with respect to the empirical research, which will be elaborated on in the following chapters. The role of REITs in a mixed asset portfolio, and the question whether REITs provide exposure to unsecuritized real estate and to stocks or bonds have been discussed by many academics. In order to analyze previous studies that investigated the role of REITs in mixed asset-portfolios, this chapter starts with an explanation of the structure of REITs in Paragraph 2.1, followed by a summary of the types of REITs in Paragraph 2.2. Next, to get a clearer picture of the changes in regulations that have influenced the REIT returns during their existence, the chronological time schedule of the development of the REIT market from 1960 until 2006 - the pre-crisis period - is described in Paragraph 2.3. Then the crisis and post- crisis period corresponding to the financial world of the US, from 2007 until 2012, are described in Paragraph 2.4. Both paragraph 2.3 and 2.4 use information stated in the official website of NAREIT (www.reit.com). Paragraph 2.5 follows with a review of previous studies, which investigated other asset classes in a mixed portfolio and their correlation to and movement within the REIT returns during the different sub-periods. Finally, Paragraph 2.6 reviews the articles of Clayton and MacKinnon (2003) and Anderson et al. (2005), who both used the same multi-factor model, a variance decomposition, in their investigation of the relationship between REIT returns and other asset classes as used in this study.
2.1. The Structure of REITs
In this study the structure of the REIT is reviewed in detail in order to rule out any misunderstandings. REITs are exempt from corporate income tax as compared to most
other types of stock. This exemption from corporate level taxation enables REIT shareholders to avoid double taxation of corporate income that characterizes most stocks. The rationality of this exemption is that REITs are viewed as investment vehicles. These favorable tax regulations have other costs in the form of a couple of regulatory constraints. Geltner et al. (2007) describe four “tests” a company must pass in order to qualify as a REIT:
1) Ownership test: A REIT cannot be a closely held corporation, in the sense that five or fewer individuals (and certain trusts) may own more than 50 percent of the REIT's stock, and there must be at least 100 different shareholders. This is known as the “five-or-fewer” rule. With the “look-through” provision enacted in 1993, pension funds are considered for the purpose of this rule to represent as many owners as there are members of the pension plan. These, in effect institutional investors, are not limited by the five-or-fewer rule.
2) Asset test: Seventy-five percent or more of the REITs total assets must be real estate, mortgages, cash or federal government securities, and 75 percent or more of the REIT’s annual Gross Income must be derived directly or indirectly from real property (including mortgages, partnerships and other REITs). Since 2001 REITs have been allowed to form and own a taxable REIT subsidiary (TRS)1 that allows them to engage in activities and/or services to tenants that were previously not permitted by the IRS (Internal Revenue Services)2 under REIT rules. However, no more than 20 percent of its assets can consist of stocks of a TRS
3) Income Test: REITs must derive their income from primarily passive sources like rents and mortgage interest, as distinct from short-term trading or sale of property assets. They cannot use their tax status to shield non-real-estate income from corporate taxation.
1
TRS, The taxable REIT subsidiary, created in the REIT Modernization Act of 1999 that went info effect in 2001 2
IRS, Internal Revenue Service, Revenue Service of the United States federal government
A REIT is subject to a tax of 100 percent on net income from "prohibited transactions", such as the sale or other disposition of property held primarily for sale in the ordinary course of its trade or business. However, if the REIT sells property it has held for at least four years and the aggregate adjusted basis of the property sold does not exceed 10 percent of the aggregate basis of all assets of the REIT as from the beginning of the year, then no prohibited transaction is deemed to have occurred.
4) Distribution Test: At least 90 percent or more of the REIT’s annual taxable income must be distributed to shareholders as dividends each year.
2.2. Types of REIT
REITs can be classified into three main groups: equity REITs, mortgage REITs and hybrid REITs. Equity REITs (EREITs) are engaged in the acquisition, management, building, renovation and sale of real estate and is favored by investors, because of the fact that it offers the greatest potential for reward. It therefore represents about 90 percent of all REITs, with a total market capitalization (12/31/2012) of US$603bn and consists of 139 EREITs (Figure 2.1, Table 2.1-2.2). The assets can be bought directly or in a joint venture with other companies. Mortgage REITs (MREITs) purchase mortgage obligations, some of them also borrow money from banks and re-lend it at a higher rate. With US$59bn it only accounts for the other 10 percent of the market capitalization (Table 2.2). Hybrid REITs (HREITs) consist of both EREITs and MREITs, and since 2010 there are zero HREITs (Liang et al. 1990; Table 2.2). This study mainly focuses on equity REITs, since it is the most interesting group with regards to its relationship to other assets.
Further, REITs are classified by the tyrospe of properties owned. The NAREIT Index is divided into different property types. The market leader is the fund focused on retail, industrial/ office and residential sectors. In Figure 2.2 the companies are listed by sector.
Figure 2.1, Types of US Listed REITs Figure 2.2, Listed US REITs by Property Type
(Since 12/31/2012) (Since 12/31/2012)
Source: Smotrich et al. 2012 Source: Smotrich et al. 2012
2.3. Pre-Crisis Period
In the last half century the REIT market has become a global investment vehicle and the REIT structure has continuously changed over time. Five years after the creation of the REIT investment vehicle in the US, the first REIT: Continental Mortgage Investors was listed on the New York Stock Exchange. It did not take long for other countries to start implementing the REIT regime. First the Netherlands (1969), quickly followed by Australia (1971), launched the REIT model into the Western World.
In the time just after the start of REITs in the US, the most important changes that have influenced the REIT performance took place. The first was the Tax Reform Act of 1986, signed by President Reagan. This led to the following consequences; the limited partnership was not attractive for investors and passive owners were not allowed to take advantage of tax deduction any longer. Therefore, REITs were now allowed to be
internally advised and managed and they were permitted to be more active, similar to industrial companies.
Table 2.1, Number of REITs Tracked by NAREIT, 1971-2012
Source: NAREIT, as of 12-31-2012
Table 2.2, Growth of Market Cap for REITs, 1971-2012 ($US bn.)
Source: NAREIT, as of 12-31-2012
From 1992 until 1997, the market capitalization grew tremendously, from US$9bn to nearly US$128bn (Table 2.2). This is also known as the REIT boom - the time period in which the Modern REIT Era started. The REIT boom started with the Omnibus Reconciliation Act of 1993, which changed the Five-or-Fewer Rule, mentioned in the structure of REITs in Paragraph 2.1, to make it easier for pension plans to invest in
REITs. This Act was followed by the Initial Public Offering (IPO) boom of 1994, in which 95 capital constrained, private real estate firms went public as REITs (Table 2.1). The reason for these real estate firms to go public was in order to raise funds on Wall Street so that they could re-finance maturing mortgages and take advantage of buying opportunities in the depressing direct real estate market. During the period of the Reform Act and the IPO boom things changed rapidly. In 1985 the number of REITs were 82, in 1992 there were 142 and in 1997 it rose to 211. The market capitalization increased in this period form US$13bn in 1991 to US$ 141bn in 1997 (Table 2.2). In the late 1990s this unlimited growth came to a halt. The REITs market capitalization dropped for the first time since 1990 (Table 2.2). After this period of decline the REIT market capitalization started to grow again in 1999 due to the REIT Modernization Act, which was passed, reducing the minimum distribution requirement from 95 percent down to 90 percent of the taxable income of REITs. Another change occurred in 2001 when the ratings agency, Standard & Poor’s, added REITs to its indices, namely the S&P 500. This marked the new start of the rising market capitalization of REITs. From 2000 until 2006 the REITs share price performance was strong, which can largely be associated to the booming real estate markets, changes in regulations and the inclusion in the Standard & Poor’s indices (Edwards, 2011). Lately it can be characterized as a consolidation of the industry. The number of REITs slightly decreased in favor of a higher market capitalization. In the late 1990s the number of REITs were 203 and the market capitalization US$124bn, this market capitalization increased dramatically to US$438bn in 2006 combined with a small decrease in the number of REITs to 183 (Table 2.1-2.2).
2.4. Start and post-crisis period
After 2006 the increasing performance of the financial sector came to a halt. The mortgage crisis that emerged in 2006, deepened in 2007 (Kaymax et al., 2012). In August 2007 investment bank BNP Paribas announced they would not be able to take money out of three hedge funds that specialized in US mortgage debt. In the months which followed the rate at which banks lend to each other rose to 6.8 percent, the highest level since December 1998. The median home prices fell from the peak in 2006 of US$246.500 to US$216.700 in January 2009 (US Department of Commerce).
Two months after the announcement by BNP Paribas the first major losses began to emerge. At first it started with Swiss BANK UBS, which lost US$3.4bn, Citigroup (US) stated losses of US$40bn and Merrill Lynch (US) lost US$7.9bn. At the end of 2007, the European bank injected US$500bn into the commercial banks, and George W. Bush, US President at that time, outlined plans to help millions of homeowners facing foreclosure. The US Federal Reserve and the Bank of England also cut interest rates. Despite these efforts 2008 started with the downgrading of a number of insurers, who specialize in insuring bonds by rating agency Standard and Poor’s (Elliott, 2011). This led to the failing of the New York Global Investment bank, The Bear Stearns, which was subsequently sold to JP Morgan Chase in March 2008.
The International Monetary Fund (IMF), which oversees the global economy, warned that the losses from the credit crunch could run up to US$1tn or higher. In the following months the first casualties were made public. In September 2008 the notion that all banks were “too big to fail”, no longer held true. The Wall Street Bank Lehman Brothers became the first major bank to collapse since the start of the global financial crisis. In
reaction, Western governments injected vast sums of capital into their banks in order to prevent them from collapsing. Others were rescued through a take-over- Merrill Lynch; a
Figure 2.3, Daily amount ($ billions) and daily fed funds rate
Source: http://www.voxeu.org/article/what-happened-us-interbank-lending-financial-crisis
bailout - Washington Mutual; a rescue pack - Fannie Mae and Freddie Mac, or they were nationalized- Bradford & Bingley (UK), Glitnir (Iceland), Dexia (Belgium), ABN (The Netherlands). Not only did the US and Europe show severe financial downgrading, but in November 2008 China also showed the first signs of the economic crisis and it set out a two-year US$586bn economic stimulant package to help boost their economy. However, it was too late to prevent the global economy from going into a free-fall, and the Eurozone officially slipped into a recession, followed by the US a month later. At that time interest rates were cut down to as low as possible, more and more extensive fiscal stimulant packages of varying sizes were announced and electron money was created through quantitative easing. The world leaders committed themselves to a US$5tn fiscal expansion in order to boost jobs and growth as well as to reform the banks.
Figure 2.4 Total Return Index of All Equity REITs, index starting at 100 basis points in quarter 1 of 1972
In 2009 it seemed as if these worldwide mature loans, bailouts and rescue plans paid off. Looking back it is hard to believe that in 2008, at the height of the global recession, the market crashed and market analysts questioned if it would survive (Lehman and Roth, 2012). And less than two years after the height of the recession, the market has started to recover gradually. The total returns of public equity REITs in 2011 increased with 8.28 percent, compared to 2.11 percent gain for the large cap stock (S&P 500) during the same year3. The market capitalization of REITs increased to US$450bn in 2011, the highest since 2006, when it reached US$438bn (reit.com; Table 2.2). Meanwhile the REITs debt reduced, and since September 2011 the listed US REIT industry’s ratio of debt divided by total market capitalization stayed put at 42 percent, a historical average. Since 2011 the US REITs continued to, on average, perform better than the other asset classes in all quarters, quarter 3 of 2012 shows an exception, in this quarter the NAREIT all equity return series shows a steep incline in total return to 0.16 percent.
3
Year-to-date annual total return of FTSE All Equity REIT Index and S&P500
In hindsight, it seems that the REIT returns were negatively affected by the global financial crisis of 2008-2009, just like all sectors, but since the market has reached an all-time low the REIT returns have been one of the best performing asset classes (Wilshire, 2012; Lehman and Roth, 2012).
Table 2.3 Total Return Index, Since 12/31/2011
Source: Datastream 2013
2.5. Relationship Between REITs and Other Asset Classes
Most investors combine their investment in REITs with an investment in other asset classes in a diversified portfolio. Therefore it is interesting to know how REIT returns tend to perform in comparison to these sectors (Springer, 2006). Kuhle (1987) was the first to report empirical evidence indicating that, to a certain extent, REITs could offer benefits of diversification for investors holding equities. These well-recognized findings led to many follow-up research.
However, the results of these follow-up researches were somehow conflicting. Peterson et al. (1997) mentioned the high correlation between REIT returns with the returns on both bonds and stocks, and it was confirmed by Ling and Naranjo in 1999. Also Mull and
Soenen (1997) questioned the diversification benefits between REITs and stocks and found a positive correlation between the yield of REITs and stock investments. Similar results have been presented by Liang and McIntoch (1998) and He et al. (2003). Looking at the different stock investments, Lee and Stevenson (2005) provided evidence of a rather strong relationship between REITs and small cap stocks. Former empirical research by Clayton and MacKinnon (2001) show that the relationship between REITs and other assets are indeed similar in some cases and that their influence varies over time. Their results indicated that, up to 1992, REITs have been closely linked to small cap stocks and, after 1992, the relationship to the unsecuritized real estate market became more predominant. As direct real estate underlies REIT properties it is expected that they are linked to each other and influenced by the same state variables (Gilberto, 1993; Clayton and MacKinnon, 2003). Research studying the relationship between REITs and unsecuritized real estate, indicated that the empirical research is mixed. Pagliari and Webb (1995) found a weak statistical link. While Gilberto (1990) found that NAREIT - public market returns - and NCREIF -private; unsecuritized market returns - are positively related to each other after the effects of the financial market had been removed. Gilberto (1990) also claims that these results could be misleading due to the smoothed returns series of unsecuritized real estate. Barkham and Geltner (1995) investigated this lagging between securitized and unsecuritized real estate and found that unsecuritized real estate follows securitized real estate with a lag of more than one year. They argued that this lagging happens because securitized real estate has a higher trading density, it is more liquid and has microstructure advantages. All the above-mentioned studies studied the pre crisis period. The following studies described are more recent and includes the crisis period.
Basse and Friedrich (2009) found evidence which supports the fact that the financial crisis has caused a structural change in the relationship between REITs and common equities, and concluded that investment in REITs had become more risky during the crisis period. However, Simon and Ng (2009) analyzed the potential of REITs to provide protection against severe stock market downturns in the US and find results indicating that the crisis has had a neglected influence on REIT’s potential to provide protection against severe stock market losses. And state therefore that REITs seem to provide protection against stock market downturns. This is in line with the findings of Mueller and Mueller (2003) who stated that the inclusion of real estate in a mixed asset portfolio moves up the efficient frontier, leading to higher returns, even in times of market uncertainty. Geltner et al. (2007) suggests that REITs may be of greater diversification benefits for investors whose portfolio manly excised out of large cap stock. Others argue whether the concept of diversification has died as a useful concept in the article of Clayton et al (2009), Clayton et al. (2009) find this statement exaggerated. And note that it is crucial to put the crisis period into perspective, because the underlying drivers of strategic asset allocation are structured to meet long-term goals. Looking at post crisis period outcomes of the research done by Chiang et al. (2013) it’s shown that there is a higher positive relationship between the REIT market and the stock market since the crisis unfolded. They also found an increase in the correlation coefficients which doubled in the time from before the crisis until after the crisis.
When studying related literature regarding the co-integration between the two asset classes it shows that the empirical evidence is mixed. Glascock et al. (2000) reported empirical evidence indicating that co-integration has become a relevant phenomenon, however Wilson and Okunev (1999) have not been able to show that co-integration exists in the same time period. More recently, Westerheide (2006) tested the co-integration of
REITs versus stocks and bonds and the results indicated that REITs form an asset class on their own.
2.6. Literature related to the Variance Decomposition Model
This study uses the multi-factor which was first used by well-known research specialists Clayton and MacKinnon (2003). In 2005, Anderson et al. did a follow-up research which included the division of small cap stocks into growth and value cap stocks. The time period used by Clayton and MacKinnon (2003) spans form 1979-1998, and was divided into sub-periods of seven years. When comparing the mean returns and standard deviation of the sub-periods, it shows that the 1992-1998 period, in which the REIT boom took place, is characterized by lower volatility in all asset class returns. They further stated that the REIT market is highly correlated with the small cap stocks and essentially uncorrelated with direct real estate returns. After 1992 the correlation of the return series with the NAREIT return changed considerably. The correlation between REITs, stocks and bonds returns dropped dramatically as time went by.
Thereafter, they applied the variance decomposition. For the full sample period it shows state variables that drive large cap stock and bonds returns as the primary drivers of REIT returns. Clayton and MacKinnon (2003) showed that REIT returns lost their sensitivity to large cap stock over time and this was confirmed by Anderson et al. (2005). According to Clayton and MacKinnon (2003), the large cap stock showed a decline of 65 percent over a period of 21 years (1979-1998) while Anderson et al. (2005) showed a decline of 45 percent to 8 percent in the period from 1978- 2003, both with large cap stock as numeraire or base. Clayton and MacKinnon (2003) showed a bond returns relationship with NAREIT that declines over time, were the results of Anderson et al. (2005) showed that this relationship is almost negligible during all periods. Conversely to previous
factors, Clayton and MacKinnon (2003) stated that the small capital stock had an increased sensitivity to the REIT factor. In their follow-up research Anderson et al. (2005) made the distinction between small capital value and small capital stock returns. Their results showed that the relationship between REITs returns and small capital stocks is primarily due to small capital value stocks, while the small capital growth stock has a negligible impact on the volatility of REIT returns.
The unsecuritized real estate factor, in the research of Clayton and MacKinnon (2003), shows a significant relationship with REIT returns in the 1992-1998 sub-period, as prior to this period this relationship was neglectable. Furthermore, Anderson et al. (2005) confirmed the increased role of unsecuritized real estate over time. However, even the increased factor explains less than 5 percent of the REITs variability. They argued that the influence could be underestimated because the monthly relationship between public and private real estate moves too slowly in order to be a significant contributor to REIT volatility (Anderson et al, 2005). The idiosyncratic factor or unexplained factor suggests that common factors of REITs with stocks and bonds diminished over time. This is because its contribution grew from 14 percent in 1979-1984 to 63 percent in 1992-1998, while at the same time the overall model fit declines from nearly 86 percent to less than 35 percent over three periods. This also indicates the difficulty to try and explain the US REIT returns by focusing on the factors of stocks, bond and unsecuritized real estate.
According to Clayton and MacKinnon (2003), the regression results of the time paths of the relative contributions of S&P 500, Russell 2000, direct Real Estate and Idiosyncratic factors to NAREIT returns show the emerging of the small cap stock factor in the late 1980s, closely followed by the real estate factor in the early 1990s. After having reached the top in 1991, the small cap stock factor decreased to nearly zero percent in 1994 but
increased dramatically in the last year of the investigation done by Clayton and MacKinnon (2003). In contrast to the increase of small cap stock in the last year the S&P 500 stock showed a dramatic decline from 1995 -1998. Beginning at a starting point of 80 percent in 1982 the large cap stock factor declined steadily to 50 percent in 1994, then collapsed to approximately 5 percent, and remained at this level until 1998. Opposite to the movement of the large cap stock factor, the idiosyncratic risk began at a low level of 10 percent and increased steadily to 30 percent in 1994, with a rapid increased between 1994 and 1998 to a level of 70 percent.
Clayton and MacKinnon (2003) indicated three reasons why there is a lower contribution of the large stock factor and the idiosyncratic risks factor. The first reason relates to the institutionalization of stock ownership and technology. Volatility of individual stock returns has increased because firm-specific information is released more frequently. Hence, investors can act faster on this information. Besides the market is more dominated by institutional investors, who have a higher turnover in shares and this also led to increased volatility because of the “herding” behavior among these investors (Campbell, 2001). The second reason, given by Clayton and MacKinnon (2003), is the cyclic patterns of the factor variance proportions. They claimed that the magnitude of the changes of the diminishing large capital stock factor as well as the increase in idiosyncratic risk in the 1980s was much higher during the 1992 period and thus suggested that the structural changes played a huge role in explaining the shift in explanatory power. Finally, another explanation for the increasing idiosyncratic portion of the variability of REIT returns in the 1990s could be because the idiosyncratic risk is, in part, picking up the omitted consumption variable. Ling and Naranjo (1997, 1999) argued that the growth rate in real consumption per capita is an important driver of REIT returns.
3.
Research Methodology
This research uses the Variance Decomposition Model to explain the sensitivity of securitized real estate returns to stocks, bonds and direct real estates. This model was used by Clayton and MacKinnon (2003) and Anderson, Clayton, MacKinnon and Sharma (2005) and is a highly acknowledge research. This chapter starts by describing, in Paragraph 3.1, the data used in this investigation and gives detail of the return series derived from DataStream. Next, in Paragraph 3.2, the Variance Decomposition Model is explained and used to answer sub-questions (2) and (3) of this research. After the explaining of the Variance Decomposition Model the rolling regression is introduced in Paragraph 3.3 which processes the sensitivity over successive time variations of 25 quarters.
3.1. Data
The six variables used in this research are REITs, large cap stock, small cap growth stock, small cap value stock, bonds and unsecuritized real estate. To implement the Variance Decomposition Model, all total return series from all asset classes have to be established. The period that will be taken into consideration is 1989 (Q1) up to 2012 (Q4). To outline this period, it is divided into sub-periods with the aim to get better results in order to establish how the crisis effected the REITs. The sub-periods that are taken into consideration have a term of at least 8 months and include the pre-crisis sub-periods 1989-1993, 1994-2001, 2002-2007, the crisis period 2007-2008, and the post-crisis sub-period 2009-2012. The reason why this research begins with data from 1989 is due to the fact that in this year all the asset return series are available. The reason for dividing the sub-period into the three periods as described above, is because the middle period, 1994-2000, corresponds to what is widely considered as the ‘modern REIT-era” (Anders et al,
2005, Clayton and MacKinnon 2003). The remainder of the pre-crisis period is divided into the sub-periods 1989-1993 and 2001-2006.Furthermore, quarterly returns series are used, as monthly unsecuritized real estate returns are not available. The data used in this study is primarily the same data used by Clayton and MacKinnon (2003) and Anderson et al. (2005). However, there are some notable differences. Table 2.3 shows the differences after which they are discussed.
Regarding the REIT and large cap stock variable the same return series have been used as in the two studies mentioned. The input for REITs consist of the ‘all equity’ REIT total return index series, subtracted from the NAREIT (2013). This National Association of Real Estate Investment Trusts has been recording REITs performance since 1972 and is considered as a benchmark for the industry and can be compared with other asset classes return series. The NAREIT elaborates on four different indices for different REIT categories: the REIT index, the equity REIT index, the mortgage REIT index and the
hybrid REIT index. This study focuses on the equity4 REIT indices, as this is the most interesting group when compared to the other asset classes. The input for large cap stock is the set of the total return of the S&P 500 index. The Standard and Poor’s 5005 is a stock market index based on the market capitalization of the 500 leading companies publicly trading on the US stock market. Next, the small capitalization stocks. In the investigation done by Clayton and MacKinnon (2003) they used the small cap stock returns series from the Russell 2000 stock return index. However, in this study, and in the study of Anderson et al. (2005) a distinction has been made between small cap growth and small cap value stocks. Instead of using the Wilshire return series, used by Anderson et al. (2005), the Russell 2000 return series, used by Clayton and MacKinnon, was used for both the small growth and small value cap stock. The reason for not using the Wilshire returns is due to the unavailability of the Wilshire returns after 2005. Furthermore, the Russell 2000 index includes REITs, while the Wilshire return series does not. Next, the bond returns, Clayton and MacKinnon (2003) and Anderson et al. (2005) subtracted this from the Lehman Brothers all maturity bond return index. The data of all maturity bond returns are no longer available because of the collapse of the Lehman Brothers. Therefore the US all maturity government total return indices has been used for bond returns. The Measuring of returns for unsecuritized real estate is more problematic, since private market return data is only available annually. The most widely used benchmark for unsecuritized real estate returns is the NCREIF property index (Clayton and MacKinnon 2003). The NCREIF is a quarterly return series extending back to 1978. However, before using this return series Clayton and MacKinnon (2003) used the transaction value, employed by Fisher and Geltner (2000) to unsmooth or de-lag the NCREIF index. Investigation done by Anderson et al. (2005) used an alternative
4
Equity REITs (EREITs) are engaged in the acquisition, management, building, renovation and sale of real estate and is favored by investors, because of the fact that it offers the greatest potential for reward (Geltner et al, 2007)
5
As of June 30, 2010, the S&P 500 includes 14 REITs
approach and used data from Green Street. Unfortunately, the Green Street data is no longer available. The method used in this research to unsmooth the NCREIF was taken from CAIA Level II, Black et al. 2009, equation (1) and contains auto-correlation for unsmoothing (Appendix 8.2).6
𝑹𝑹𝒕𝒕,𝒕𝒕𝒕𝒕𝒕𝒕𝒕𝒕= (𝑹𝑹𝒕𝒕,𝒕𝒕𝒕𝒕𝒓𝒓𝒓𝒓𝒕𝒕𝒕𝒕𝒕𝒕𝒓𝒓− 𝝆𝝆(𝑹𝑹𝒕𝒕−𝟏𝟏,𝒕𝒕𝒕𝒕𝒓𝒓𝒓𝒓𝒕𝒕𝒕𝒕𝒕𝒕𝒓𝒓)/(𝟏𝟏 − 𝝆𝝆) (𝟏𝟏)
After the data is collected it can be placed into the Variance Decomposition Model of Clayton and MacKinnon (2003) in order to resolve the cross influences in returns of various asset classes. An explanation of this model is given in the following paragraph.
3.2. Variance Decomposition Model, Sub-Periods
The REIT index returns can be indicated as a linear function of large capital stocks, small capital growth stocks, small capital value stocks, bonds and real estate factors. The following equation is used:
𝒕𝒕𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝑹= 𝜷𝜷𝟎𝟎+ 𝜷𝜷𝟏𝟏𝒇𝒇𝑳𝑳𝑳𝑳+ 𝜷𝜷𝟐𝟐𝒇𝒇𝑳𝑳𝑺𝑺+ 𝜷𝜷𝟑𝟑𝒇𝒇𝑳𝑳𝑺𝑺+ 𝜷𝜷𝟒𝟒𝒇𝒇𝑩𝑩+ 𝜷𝜷𝟓𝟓𝒇𝒇𝑹𝑹𝑹𝑹+ 𝒕𝒕 (𝟐𝟐)
where rREIT is the return on an index of REITs; fLS is a factor associated with large capital stock returns; fSV the return to the small capital value stock market; fSG a small capital growth stocks return factor; fB a bond market return factor and fRE the return to unsecuritized real estate, in period t. The betas represent the sensitivity of REIT returns to the five factors on the right hand side of the function, and e represents the proportion of REIT returns that is unexplained, also known as the idiosyncratic factors.
6
The formula for the correlation coefficient of a sample is used to compute the first order autocorrelation of the NCREIF total return series. It should be noted that the sample period covers the highly unusual global financial crisis in 2007-2008. Accordingly, the observed correlations may not be representative of more normal economic conditions due to the presence of outliers and their potentially disproportionate influence (Black et al, 2009).
Equation (2) is not used in the empirical analysis because the specifications of the explanatory variables have factors that are common to all of them (e.g. inflation rate, GDP, consumption). This could lead to multicollinearity and might cause the results to be biased. Therefore, the right hand side variables have to be made independent (Clayton and MacKinnon, 2003). To obtain these “pure factors” that are not correlated to each other, the regression in equation (2) is orthogonalized in first stage regressions. To start with, one of the right hand side variables is used as numeraire/base, hereafter referred to as: base. In this case the long-term stock returns is the base. In equation (3) – (6) the returns series of each asset class are modeled as a linear function of the return of the other asset classes. The residual of equation (2) where rLS, rSG, rSV, rB, are returns on large stock, small cap growth and value stocks and bonds respectively, are absent of any influence from factors affecting the other variables. These residuals represent the unsecuritized real estate factor in equation (3). In the next equation (4) the residuals can be seen as the “pure” bond return series and represents the bond factor in equation (5), (6) and (7). This process is continued until only the base variable is left as the raw return series. This base is now independent because the other factors have become independent with the use of residuals. Hence, it is possible to run the regression in equation (7), the unbiased version of equation (2).
𝒕𝒕𝑹𝑹𝑹𝑹= 𝜹𝜹𝟎𝟎+ 𝜹𝜹𝟏𝟏𝒕𝒕𝑳𝑳𝑳𝑳+ 𝜹𝜹𝟐𝟐𝒕𝒕𝑳𝑳𝑺𝑺+ 𝜹𝜹𝟑𝟑𝒕𝒕𝑳𝑳𝑺𝑺+ 𝜹𝜹𝟒𝟒𝒕𝒕𝑩𝑩+ 𝜺𝜺𝑹𝑹𝑹𝑹 (𝟑𝟑)
𝒕𝒕𝑩𝑩= 𝜸𝜸𝟎𝟎+ 𝜸𝜸𝟏𝟏𝒕𝒕𝑳𝑳𝑳𝑳+ 𝜸𝜸𝟐𝟐𝒕𝒕𝑳𝑳𝑺𝑺+ 𝜸𝜸𝟑𝟑𝒕𝒕𝑳𝑳𝑺𝑺+ 𝜸𝜸𝟒𝟒𝜺𝜺�𝑹𝑹𝑹𝑹+ 𝜺𝜺𝑩𝑩 (𝟒𝟒)
𝒕𝒕𝑳𝑳𝑺𝑺= 𝛌𝛌𝟎𝟎+ 𝛌𝛌𝟏𝟏𝒕𝒕𝑳𝑳𝑳𝑳+ 𝛌𝛌𝟐𝟐𝒕𝒕𝑳𝑳𝑺𝑺+ 𝛌𝛌𝟑𝟑𝜺𝜺�𝑹𝑹𝑹𝑹+ 𝛌𝛌𝟒𝟒𝜺𝜺�𝑩𝑩+ 𝜺𝜺𝑳𝑳𝑺𝑺 (𝟓𝟓)
𝒕𝒕𝑳𝑳𝑺𝑺= 𝝋𝝋𝟎𝟎+ 𝝋𝝋𝟏𝟏𝒕𝒕𝑳𝑳𝑳𝑳+ 𝝋𝝋𝟐𝟐𝜺𝜺�𝑹𝑹𝑹𝑹+ 𝝋𝝋𝟑𝟑𝜺𝜺�𝑩𝑩+ 𝝋𝝋𝟒𝟒𝜺𝜺�𝑳𝑳𝑺𝑺+ 𝜺𝜺𝑳𝑳𝑺𝑺 (𝟔𝟔)
𝒕𝒕𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝑹= 𝜷𝜷𝟎𝟎+ 𝜷𝜷𝑳𝑳𝑳𝑳𝒕𝒕𝑳𝑳𝑳𝑳𝑳𝑳+ 𝜷𝜷𝑳𝑳𝑺𝑺𝜺𝜺�𝑳𝑳𝑺𝑺𝑳𝑳+ 𝜷𝜷𝑳𝑳𝑺𝑺𝜺𝜺�𝑳𝑳𝑺𝑺𝑳𝑳+ 𝜷𝜷𝑩𝑩𝜺𝜺�𝑩𝑩𝑳𝑳+ 𝜷𝜷𝑹𝑹𝑹𝑹𝜺𝜺�𝑹𝑹𝑹𝑹𝑳𝑳+ 𝒕𝒕 (𝟕𝟕)
in the above equation (7), in which large cap stock is employed as the base, there will be an overemphasis on the influence of large cap stock. Therefore, as these underlying state variables contribute to returns and variance of REITs, they are attributed only to large stocks in the above equations and it makes the component of returns and variance attributed to large stocks biased upwards. To overcome the problem of over-purging the same orthogonal process is evaluated, but another explanatory variable is set as the base. Equal to the study of Clayton and MacKinnon (2003) this study uses the unsecuritized real estate return index as second base, reason for using large cap stock and unsecuritized real estate as base is that most of the existing literature has looked at the relationships between these two variables comparing them to REIT returns, which automatically makes it possible to compare the outcomes of this study with results from previously related literature, so the changing dynamics between the relationships becomes visible. The estimates obtained when taking unsecuritized real estate as base provides the upper limit for the influence of unsecuritized real estate. Thus, reformulating of equation (2), using unsecuritized real estate as the chosen base, results in equation (8). For the former equations (3) – (6) the same ordering is used, starting from the right-handed variable working towards the left, similar to the study by Clayton and MacKinnon (2003) and Anderson et al. (2005). 𝒕𝒕𝑩𝑩= 𝜸𝜸𝟎𝟎+ 𝜸𝜸𝟏𝟏𝒕𝒕𝑳𝑳𝑳𝑳+ 𝜸𝜸𝟐𝟐𝒕𝒕𝑳𝑳𝑺𝑺+ 𝜸𝜸𝟑𝟑𝒕𝒕𝑳𝑳𝑺𝑺+ 𝜸𝜸𝟒𝟒𝒕𝒕𝑹𝑹𝑹𝑹+ 𝜺𝜺𝑩𝑩 (𝟖𝟖) 𝒕𝒕𝑳𝑳𝑺𝑺= 𝛌𝛌𝟎𝟎+ 𝛌𝛌𝟏𝟏𝒕𝒕𝑳𝑳𝑳𝑳+ 𝛌𝛌𝟐𝟐𝒕𝒕𝑳𝑳𝑺𝑺+ 𝛌𝛌𝟑𝟑𝒕𝒕𝑹𝑹𝑹𝑹+ 𝛌𝛌𝟒𝟒𝜺𝜺�𝑩𝑩+ 𝜺𝜺𝑳𝑳𝑺𝑺 (𝟗𝟗) 𝒕𝒕𝑳𝑳𝑺𝑺= 𝝋𝝋𝟎𝟎+ 𝝋𝝋𝟏𝟏𝒕𝒕𝑳𝑳𝑳𝑳+ 𝝋𝝋𝟐𝟐𝒕𝒕𝑹𝑹𝑹𝑹+ 𝝋𝝋𝟑𝟑𝜺𝜺�𝑩𝑩+ 𝝋𝝋𝟒𝟒𝜺𝜺�𝑳𝑳𝑺𝑺+ 𝜺𝜺𝑳𝑳𝑺𝑺 (𝟏𝟏𝟎𝟎) 𝒕𝒕𝑳𝑳𝑳𝑳= 𝜹𝜹𝟎𝟎+ 𝜹𝜹𝟏𝟏𝒕𝒕𝑹𝑹𝑹𝑹+ 𝜹𝜹𝟐𝟐𝜺𝜺�𝑩𝑩+ 𝜹𝜹𝟑𝟑𝜺𝜺�𝑳𝑳𝑺𝑺+ 𝜹𝜹𝟒𝟒𝜺𝜺�𝑳𝑳𝑺𝑺+ 𝜺𝜺𝑳𝑳𝑳𝑳 (𝟏𝟏𝟏𝟏) 𝒕𝒕𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝑹= 𝜷𝜷𝟎𝟎+ 𝜷𝜷𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝑳𝑳+ 𝜷𝜷𝑳𝑳𝑳𝑳𝜺𝜺�𝑳𝑳𝑳𝑳𝑳𝑳+ 𝜷𝜷𝑳𝑳𝑺𝑺𝜺𝜺�𝑳𝑳𝑺𝑺𝑳𝑳+ 𝜷𝜷𝑳𝑳𝑺𝑺𝜺𝜺�𝑳𝑳𝑺𝑺𝑳𝑳+ 𝜷𝜷𝑩𝑩𝜺𝜺�𝑩𝑩𝑳𝑳+ 𝒕𝒕 (𝟏𝟏𝟐𝟐)
The resulting estimates (7) and (12) provide a lower and upper bound on the proportion of REIT volatility attributes to each asset class. The difference between the end points of
these ranges represent the effect in REIT volatility of state variables that cannot be uniquely assigned to a single asset class. Variables on the right, or factors in equation (7) and (12), are orthogonal to or independent of each other. To obtain the relative contribution of stocks, bonds and unsecuritized real estate to the volatility of REIT returns, equation (7) and (12) are used to get the variance equation (13) and (14).
𝑺𝑺𝑽𝑽𝑹𝑹[𝒕𝒕𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝑹] = 𝝈𝝈𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝟐𝟐 = 𝜷𝜷𝑳𝑳𝑳𝑳𝟐𝟐 𝝈𝝈𝒕𝒕𝑳𝑳𝑳𝑳𝟐𝟐 + 𝜷𝜷𝑳𝑳𝑺𝑺𝟐𝟐 𝝈𝝈𝜺𝜺𝑳𝑳𝑺𝑺𝟐𝟐 + 𝜷𝜷𝟐𝟐𝑳𝑳𝑺𝑺𝝈𝝈𝜺𝜺𝑳𝑳𝑺𝑺𝟐𝟐 + 𝜷𝜷𝑩𝑩𝟐𝟐𝝈𝝈𝜺𝜺𝑩𝑩𝟐𝟐 + 𝜷𝜷𝑹𝑹𝑹𝑹𝟐𝟐 𝝈𝝈𝜺𝜺𝑹𝑹𝑹𝑹𝟐𝟐 + 𝝈𝝈𝒕𝒕𝟐𝟐 (𝟏𝟏𝟑𝟑)
𝑺𝑺𝑽𝑽𝑹𝑹[𝒕𝒕𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝑹] = 𝝈𝝈𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝟐𝟐 = 𝜷𝜷𝑹𝑹𝑹𝑹𝟐𝟐 𝝈𝝈𝒕𝒕𝑹𝑹𝑹𝑹𝟐𝟐 + 𝜷𝜷𝑳𝑳𝑺𝑺𝟐𝟐 𝝈𝝈𝜺𝜺𝑳𝑳𝑺𝑺𝟐𝟐 + 𝜷𝜷𝟐𝟐𝑳𝑳𝑺𝑺𝝈𝝈𝜺𝜺𝑳𝑳𝑺𝑺𝟐𝟐 + 𝜷𝜷𝑩𝑩𝟐𝟐𝝈𝝈𝜺𝜺𝑩𝑩𝟐𝟐 + 𝜷𝜷𝑳𝑳𝑳𝑳𝟐𝟐 𝝈𝝈𝜺𝜺𝑳𝑳𝑳𝑳𝟐𝟐 + 𝝈𝝈𝒕𝒕𝟐𝟐 (𝟏𝟏𝟒𝟒)
Finally, using equation (13) and (14) we can estimate the relative contribution of each factor to total equity REIT return variability by calculating the proportion of the variance of REIT returns due to each factor. This can be done as follows:
𝑳𝑳𝑳𝑳𝒕𝒕𝑳𝑳𝒕𝒕 𝒄𝒄𝑳𝑳𝒓𝒓 𝒔𝒔𝒕𝒕𝒓𝒓𝒄𝒄𝒔𝒔 =𝜷𝜷𝑳𝑳𝑳𝑳𝟐𝟐 𝝈𝝈𝒕𝒕𝑳𝑳𝑳𝑳𝟐𝟐 𝝈𝝈𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝟐𝟐 𝑳𝑳𝑺𝑺𝑳𝑳𝑺𝑺𝑺𝑺 𝒄𝒄𝑳𝑳𝒓𝒓 𝒔𝒔𝒕𝒕𝒓𝒓𝒄𝒄𝒔𝒔 = 𝜷𝜷𝑳𝑳𝑺𝑺𝟐𝟐 𝝈𝝈𝜺𝜺𝑳𝑳𝑺𝑺𝟐𝟐 𝝈𝝈𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝟐𝟐 𝑳𝑳𝑺𝑺𝑳𝑳𝑺𝑺𝑺𝑺 𝒄𝒄𝑳𝑳𝒓𝒓 𝒗𝒗𝑳𝑳𝑺𝑺𝒕𝒕𝒕𝒕 = 𝜷𝜷𝑳𝑳𝑺𝑺𝟐𝟐 𝝈𝝈𝜺𝜺𝑳𝑳𝑺𝑺𝟐𝟐 𝝈𝝈𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝟐𝟐 𝑩𝑩𝒓𝒓𝑩𝑩𝒓𝒓 = 𝜷𝜷𝑩𝑩𝟐𝟐𝝈𝝈𝜺𝜺𝑩𝑩𝟐𝟐 𝝈𝝈𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝟐𝟐 (𝟏𝟏𝟓𝟓) 𝑼𝑼𝑩𝑩𝒔𝒔𝒕𝒕𝒄𝒄𝒕𝒕𝒕𝒕𝑳𝑳𝒕𝒕𝑳𝑳𝑼𝑼𝒕𝒕𝒓𝒓 𝒕𝒕𝒕𝒕𝑳𝑳𝑺𝑺 𝒕𝒕𝒕𝒕𝑳𝑳𝒕𝒕𝒕𝒕 =𝜷𝜷𝑹𝑹𝑹𝑹𝟐𝟐 𝝈𝝈𝜺𝜺𝑹𝑹𝑹𝑹𝟐𝟐 𝝈𝝈𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝟐𝟐 𝑹𝑹𝒓𝒓𝑳𝑳𝒓𝒓𝒔𝒔𝑰𝑰𝑩𝑩𝒄𝒄𝒕𝒕𝑳𝑳𝒕𝒕𝑳𝑳𝒄𝒄 = 𝝈𝝈𝒕𝒕𝟐𝟐 𝝈𝝈𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝑹𝟐𝟐
The sum of the five components: large cap stock, small cap growth stock, small cap value stock, bond and real estate is equal to the coefficient of determiner R-squared, and represents the complete part of the model which has been explained. The idiosyncratic part in equation (15) is equal to one minus the R-squared. The output from the variance decomposition is given in Paragraph 4.2.
3.3. Rolling Regression
In order to investigate time-variation in the relative contributions of the stocks, bonds and unsecuritized real estate factors in more detail the rolling regression is processed. This is
done by a re-estimation of the orthogonalized regressions and the REIT return regression 25 quarters. In every new period the returns of a new quarter is added and the returns of the last quarter is left out of the sample so that the full sample ranges from 1989 (Q2) until 2012 (Q4). Using the 25 quarters for each estimate, leads to the first estimate of 25 quarters, starting in 1989 (Q2) until 1995 (Q2). The model is calculated 71 times and the parameter estimates in the REIT regressions are used to calculate the variance proportions in each case. The results are given in Paragraph 4.3.
4.
Analysing the Performance
This chapter represents the outcomes from the previously explained regressions and is divided into three parts. It starts with the time variation of factor sensitivity of equity REIT returns on the other asset classes (Paragraph 4.1). This performance analysis is done to answer sub-question (3) Which coefficients of stocks, bonds and unsecuritized real estate factors unfold from the different sub-periods? Paragraph 4.2 presents the estimates of the relative contribution of the other asset class factors to REIT returns variability. These estimates are the direct output used to answer sub-question (4) Which outcomes of stocks, bonds and unsecuritized real estate factors unfold from the different sub-periods doing the variance decomposition? This is followed by showing the rolling regression results of the relative contribution of the factors to the variability of REIT returns over a time path (Paragraph 4.3). Finally, a 100 percent stacked area chart is drawn up (Paragraph 4.4) to get better insight into how the total influence on REIT returns variability has been built up over the years.
4.1. Time Variation in Factor Sensitivity
As described in Paragraph 3.2, the data collected from Datastream was added into the multi-factor regressions of equation (7) and (12). These equations measure the sensitivity of REIT returns as a function of large cap stock, small cap growth stock, small cap value stock, bonds and unsecuritized real estate factors. The results are presented for the full sample period and during the pre-crisis sub-periods: 1989-1993, 1994-2000, 2001-2006, the crisis period 2007-2008 and the post crisis period 2009-2012.
Table 4.1 represents the results from equation (7) and (12) covering the full sample period and sub-periods. This was done in order to determine whether the relative influences of
the financial market factors have changed over time. The results presented use large cap stock as the base orthogonalization (column LSR7 base, Table 4.1) procedure, as well as using unsecuritized real estate as the base (column RER8 base, Table 4.1). The first three periods, the pre-crisis periods, are quite robust in the order of orthogonalization in terms of relative changes in magnitude of the coefficients over time. Only the coefficient on the unsecuritized real estate differs substantially when switching the base from LSR to RER, especially in the periods 1989-1993 and 1994-2000. The impact is higher, suggesting over-purging as discussed in Paragraph 3.2 (Clayton and MacKinnon, 2003). However, during and after the crisis period the coefficients differ much in the order of the orthogonalization, especially in the crisis period. Notable is that the factor sensitivity of all return series during the crisis period are insignificant, probably due to the small sample set of 8 quarters, which could result in insignificant numbers.
The adjusted R-squared of 58 percent during the full sample period shows that the REIT returns are largely driven by the same underlying state variables as the other asset classes. The adjusted R-squared remains consistent in the before- and during- crisis period, it moves between the 42 percent and 55 percent. But in the post-crisis period the adjusted R-squared increases to 88 percent, indicating that the REIT returns become more driven by the same underlying state variables as the five other asset classes.9
Over the full sample period, small cap value and large cap stock returns seems to be the primary drivers of the REIT returns; with respectively 0.85 and 0.75 if LSR is used as the base. The results of the pre-crisis period (Appendix 8.3) support the results of Clayton and MacKinnon (2003) that, in general, state variables that drive small cap stock, large cap
7
LSR, Large Cap Stock used as base in regression
8
RER, Unsecuritized Real Estate used as base in regression
9
In case of small sample size of less than 25 quarters the high R-squared can be misleading. Meaning, the results of the Adjusted R-squared in the during- and post- crisis periods are less accurate.
stock and bond returns are the primary drivers of the REIT returns over the full-sample of the pre-crisis period. However, Clayton and MacKinnon (2003) did not divide small capitals into small cap growth and value stock. As this research does, the results show that the relationship between REIT returns and small cap stock is primarily due to small cap value and not small cap growth stocks. As mentioned before, the sensitivity of small cap value stock is the highest of the five the assets, whereas the small cap growth stock accounts for a negligible 0.05 (LSR base), both the high sensitivity with small cap value stock as the low sensitivity with small cap growth stock corresponds with the results of Anderson et al. (2005).
When looking at the time-path it shows that in the beginning REITs have the greatest sensitivity to large cap stock with 1.06**(LSR base) and 1.14**(RER base) and bonds with 1.22 (LSR base) and 0.92 (RER base). However, this influence decreases when moving to the next time period where both asset classes returns drop - the large cap stock drops to a level of 0.10 (LSR base) and 0.11 (RER base) and in case of bonds they drop to 0.13 (LSR and RER base) during the REIT era (1994-2000). In the last pre-crisis period both large cap stock and bonds emerge, the large cap stock increases to 0.48** (LSR base) or 0.45** (RER base) and the bonds increase to either 0.45 (LSR base) or 0.51 (RER base). When using LSR as the base during the crisis period, both large cap stock (0.82) and bonds (2.62) seem to have an increasing influence on REIT. On the other hand, setting RER as base, the large cap stock and bonds seem to have a negative influence of -1.06 and -1.44 respectively. In the last period, the post crisis period, the large cap stock maintains its high influence of 1.31** (LSR base) and 1.05** (RER base) and both are significant at 5 percent. Opposite are the bond related factors who differs much when switching between LSR base (0.54) and RER base (-0.85).
Similar to the research of Anderson et al. (2005), this research makes a distinction between small cap growth and small cap value stock. The results show that they move in opposite directions. Regarding the small cap growth stock, it starts with negative signs of -0.07 (LSR base) and -0.08 (RER base) and this influence increases from 0.02 (LSR base) and 0.03 (RER base) to 0.51* (LSR base) and 0.61* (RER base) in the last pre-crisis period. During the crisis period both the LSR (0.40) and RER (1.85) show a high positive sensitivity, whereas the post-crisis period results are negative for both the LSR (-0.84) and RER (-0.68). Next, the results for the small cap value stock show that, when compared to the small cap growth stock, they have a higher influence on REITs as it starts at 0.75* (LSR base) and 0.92* (RER base) in 1989-1993. Then it drops slightly to 0.81** (LSR base) and 0.83* (RER base) in 1994-2000, and again to 0.41* (LSR base) and 0.50* (RER base) in 2000-2006. During and after the crisis period it increased to 1.64 and 1.94* respectively using LSR as base. However, with the RER as base it drops to -0.19 during the crisis, but increases to a significant 1.46* after the crisis period.
Finally, looking at the influence of unsecuritized real estate setting with LSR as base, it starts negatively in the first two pre-crisis periods; -0.51 in the 1st period and -.0.29 in the 2nd period, and then increases dramatically to 1.54* in the 3rd pre-crisis period, then during the crisis it increases to 2.41 and decreases slightly to 2.31 in the post-crisis period. When RER is used as the base then there is a difference in the influence of unsecuritized real estate. Opposite to the result when using LSR as base, the 1st period starts positive at 0.25 and increases to 0.42 in the 2nd and 1.35 in the 3rd pre-crisis period. During the crisis it drops to a level of 0.74 and afterwards it shows a dramatic increase to 5.23**, with a 5 percent significance level.
4.2. Relative contribution to the volatility of REIT returns
To empirically quantify the impact of the various asset class drivers in REIT returns variability, the REIT returns variance is divided into components associated with each asset class factor using the parameter estimates of Table 4.1 to estimate the contributions of the orthogonalized asset class returns to REIT return volatility. Table 4.2 presents the results, with formulas where the chosen base is either large cap stock orthogonalization (column LSR base, Table 4.2) or unsecuritized real estate orthogonalization (column LSR base, Table 4.2). Results are again recorded over the full period and the three sub-periods 1989-1993, 1994-2000, 2001-2006, the crisis period 2007-2008 and the period after the crises 2009-2012.
Consistent with the results from the Adjusted R-squared in Table 4.1 the idiosyncratic risk appears to be at its lowest the 25 quarter period ending after the crisis period. This indicates that the REIT returns became more driven by the same underlying state variables as other asset classes. Meaning the diversification benefits of REITs diminishes; this phenomenon is further described in chapter 6.
For the full sample period with S&P 500 returns as base in the first regression, large cap stock factors and small cap value stock indicate 35.62 percent (LSR base) and 20.48 percent (LSR base) of REIT volatility, respectively. Regarding small cap growth stock, using either LSR or RER as base, bonds and unsecuritized real estate appear to play a negligible role. As expected, after studying the results from related literature, the small cap growth stocks and bonds factors have a negligible influence which does not change during any of the sub-periods (Anderson et al., 2005).
When studying the pre-crisis sub-periods, the results show a decrease in the explanatory power of large cap stock from 43.43 percent (LSR base) and 44.22 percent (RER base) to almost zero in the REIT-era, regardless of the base used. In the 3rd pre-crisis period, 2001-2006, it indicates 33.19 percent (LSR base) and 27.27 percent (RER base), and continues to grow during and after the crisis period, when using LSR as base, to 41.46 percent (LSR base) and 68.23 percent (LSR base) respectively. However, the contribution is less when using the RER as base, where the influence during the crisis is 7.98 percent (RER base) and is less than the 41.46 percent (LSR base). A similar pattern emerges during the post crisis period, where the difference is approximately 34 percent (34.96 RER base).
The proportion of the variance due to small cap value stock, shows an overall fit of 20.48 percent (LSR base) and 17.80 percent (RER base). This is mainly the result of the exceptionally high influence during the ‘REIT era’ in the 1993-2000 period with 50.33 percent (LSR base) and 49.31 percent using RER as base. The coefficients differ less than 5 percent in all periods, regardless of the orthogonalization, except during the financial crisis period, in which the influence is 22.30 percent (LSR base) or 0.12 percent (RER base).
Interesting to note are the results of the relative contribution of unsecuritized real estate to the REIT return volatility. This factor strongly depends on which base is taken during and after the crisis. The influence of unsecuritized real estate is, in almost all periods, insignificant when LSR is used as the base, with one exception during the crisis period. Where it accounts for 15.03 percent (LSR base). Switching to a RER base the results in the pre-crisis period are identical, but is highly over-purged when looking at the crisis period (62.11) and post-crisis period (43.53).
