The dynamic relationship among REITs, stock and bond returns around the European debt crisis

(1)

University of Amsterdam Amsterdam Business School

MSc Business Economics, Real Estate Finance track Master Thesis

The dynamic relationship among REITs, stock and

bond returns around the European debt crisis

Kim, Yuna

August, 2016

(2)

2

Abstract

This thesis aims to analyze the dynamic relationship between financial assets, which include stocks, fixed-income securities, and real estate investment trusts (REITs) before and after the European debt crisis. A total seven countries have been investigated: The United States (US), Europe (France and the United Kingdom), and Asia-Pacific (Japan, Hong Kong, Singapore, and Australia) from January 2007 to March 2016. A total six variables are investigated which are the return of REITs, of stock indices, of fixed-income securities, VIX index, and CPI.

I found that all level variables are non-stationary after applying three unit root tests which are Hadri LM, LLC and IPS. Therefore, a cointegration test has been conducted as suggested by Westerlund (2007) and a long-run equilibrium among variables has been detected. Since there are long term relationships between variables, a panel error correction model has been performed.

The results show that there are long-term equilibriums between variables, and for all regions, stock markets do not offer diversification effects when they combine with the REIT markets. However, fixed-income securities can have diversification effects with REITs, especially for European REIT markets. Asia pacific REITs, on the other hand, are more stable even when markets are very volatile and are able to keep positive returns.

(3)

3

Statement of originality

This document is written by Yuna Kim who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(4)

4

The purpose of this paper is to analyze the linkage between financial assets, which include stocks and fixed-income securities, and real estate investment trusts (REITs) before and after the European debt crisis for the United States (US), Europe (France and the United Kingdom), and Asia-Pacific (Japan, Hong Kong, Singapore, and Australia) markets. Many researchers have investigated the time dependent behavior of the US REITs and/or Asia-Pacific REITs (also see the literature review) but only a few studies analyze the European REITs markets even though they have been growing fast for the last decades.

Furthermore, this paper would be timely and meaningful since it focuses on the European debt crisis while most of the previous research has covered the impact of the 2007 financial crisis. The 2007 financial crisis indeed triggered the European debt crisis, but the real cause of the European crisis is more complicated than the financial crisis. According to Lane (2012), the European debt crisis is because of financial and external imbalances starting from 2003, the credit boom between 2003 and 2007, and a failure to manage fiscal policy. Therefore, the aim of this thesis will be to study the dynamic relationship between REITs, stock, and bond returns for the US, European, and Asia-Pacific markets before and after the European debt crisis.

The impact of the euro debt crisis was enormous throughout and outside Europe. Acharya et al. (2015) analyze the real effects resulted from the European debt crisis for five European countries (Greece, Ireland, Italy, Portugal, and Spain) and conclude that negative spillovers transferred to the real economy in the euro zone after the debt crisis

(6)

6

(2010-2012). On the other hand, Stracca (2013) exploits the external spillover and contagion from 2010 to 2013 for 25 advanced and emerging countries and provides an evidence of sizeable effects of the euro debt crisis outside Europe. Stracca (2013) also shows notable downturn equity returns, especially in the financial sector even for countries with a relatively strong economy such as Germany and the US. Therefore, the time frame of this thesis will be between 2006 and 2015, such that the results before (2006-2009) and after (2010-2015) the European debt crisis can be efficiently compared.

To the best of the author’s knowledge, there is no research that studies the time behavior of European REITs comparing to that of other regions’ REITs. Although Niskanen and Falkenbach analyzed the correlations of European REITs and other assets in 2010, their study is limited to only for the European market. Thus, this thesis will give meaningful messages for investors to choose which region’s REITs are more suitable for their portfolios. To investigate the relationship between REITs, stocks and fixed-income securities, each country’s REIT index, major stock index and 10 year sovereign bond yields are employed. To capture macroeconomic variables, Chicago Board Option Exchange Volatility Index and the consumer price index for each country are also used. As a methodology, the panel error correction model is employed after conducting unit root tests and cointegration tests. I found that there are long-term equilibriums between variables, and for all regions, stock markets do not offer diversification effects with the REIT markets. However, fixed-income securities can have diversification effects with REITs sometimes, especially for European REIT

(7)

7

markets. Asia pacific REITs, on the other hand, are more stable even when markets are very volatile and are able to keep positive returns.

The rest of the thesis will be organized as follows. In Section Ⅱ, related literature is briefly discussed. In Section Ⅲ the data used are described and in section Ⅳ the methodology applied for this thesis is outlined. Finally, sectionⅤsummarizes the empirical results and concludes.

II. Literature review A. Background of REITs

REITs have started to introduce a new way of investing in income-generating real estate assets, as President Dwight D. Eisenhower confirmed the REIT Act title in 19601_{. According to Semer (2009), there were two financial purpose. The first one is} that to allow investors with a low budget to invest in real estate assets so that they can create diversified portfolios with lower risks. The second reasons is to offer a new approach for real estate developers to attract more investors. After 50 years since REITs has been found, now it has known to have diversification effects, inflation protection, better performances than S&P 500. According to a research conducted by EY in 2016, U.S. REITs contributed around 1.8 million full-time jobs and USD$107.5 billion of labor income to U.S. economy in 2014. In addition, about USD$44 billion and USD$81.6 billion

1_{REIT.com, History of REITs. [online] Available at:}

(8)

8

are distributed as interest income and dividend income respectively, while USD$55.9 billion has been invested by REITs in both new construction and existing real estates.

In Asia, Japan was the first country to start the REIT market in 2000 followed by other Asian countries including South Korea, Singapore, Taiwan, Thailand, Malaysia and Hong Kong. 186 Asian REITs has been listed in 2014 which is total value of USD$192.32 billion (Loo at el, 2015) and showed fast growth from early 2000s. In 2014, Asian REITs exhibited the total market capitalization of USD$192.32 billion. The Australian REIT market, on the other hand, has started much earlier than the Asian REIT market. In 1971, the first REIT was listed on the Australian Stock Exchange and, in 2015, Australian REITs stood up around USD$100 billion with 43 listed REITs and accounted 8.28% of the global REIT market. In case of the European REIT market, in the mid-2000s the legislation became active and more than 14 REITs have been listed which is accounting 22% of the global REIT market till 2009 (Niskanen and Falkenbach, 2010). Belgium was the first country that adopted the special legislation in 1995, and France and the UK are the largest players in the market. Finland, Spain, and Lithuania are recently jumped into the market.

B. The diversification effect of REITs

According to NAREIT (2002), REITs have a low correlation with stock markets and thus could offer diversification effects to portfolios in the late 1990s. Glascokc et al. (2000) also claim that even though there was a cointegration relationship between REITs and fixed income securities before 1992, it had disappeared afterwards, which

(9)

9

suggests possible diversification benefits exist as well. In addition, Lee and Stevensson (2005) assert that REITs play an important role to diversify risks in a mixed-asset portfolio, after analyzing the efficient frontiers that were made of twenty portfolios. Specifically, they find evidence implying that REITs do provide more diversification benefits with a longer investment horizon and that REITs help to construct low risk and high return portfolios.

However, these diversification benefits provided by REITs are still controversial. Niskanen and Falkenbach (2010) study the sensitivity of European REIT returns to returns in other financial assets. They find a significant positive relationship between REITs and stocks and a negative relationship with fixed income securities, which implies that investors can enjoy diversification effects by including REITs and fixed income securities in one portfolio. Clayton and Mackinnon (2001) also examine the time-varying sensitivity of equity REIT returns on other assets such as real estate, stocks and bonds. The results suggest that the relationship with REIT returns and returns to other assets change over time. Laopodis (2009) adopts both the vector autoregressive methodology and cointegration analysis and finds out similar patterns between the equity and the mortgage REIT categories for the 1971-2007 period. Laopodis also claims that both the mortgage and equity REIT categories show similar interactions with the stock market as well as the movements of industrial production.

REITs from Asia-Pacific region have been also analyzed. Liu et al. (2012) investigate how REIT markets in USA, and the four Asia-Pacific countries including Australia, Hong Kong, Japan, and Singapore are related. They use a DCC-GARH model to estimate correlations and assert that a significant time variation exists in the REIT

(10)

10

series. This correlation was higher with the increasing interactions of national inflation rates and with higher volatility of global equity markets. This implies that investors should take credit spread, the volatility of equity markets, and inflation rates into account when they make a portfolio including REITs.

Some researchers focus on how the relationship has been changed during the financial crisis. Chang et al. (2012) investigate the time-varying relationship between REITs and three market variables (stock market variable, interest rate variable, and the economic growth variable) in four countries (Australia, Japan, Taiwan, and the USA) and suggest that REIT markets have a positive relationship with the stock markets before and after the sub-prime crisis. This linkage is stronger after the financial crisis, meaning less diversification benefits for investors. Similarly, Chiang et al. (2013) focus on four Asian markets such as Taiwan, Hong Kong, Singapore and Japan to study the relationship between REITs and the stock markets. Using the Multivariate GARCH-vech model and the extreme value theory, they conclude that the conditional risks in both REITs and stock markets rapidly increased during the financial crisis. Additionally, they claim that the REIT markets and stock markets have a positive relationship and this correlation becomes even higher after the crisis.

C. European debt crisis

When the global financial crisis started in August 2007, European sovereign debt markets seem relatively stable from 2008 and 2009 (Lane, 2012). The main focus in Europe after the financial crisis was mainly on how European Central Bank would solve

(11)

11

the global financial shock. Of course, major European banks had experienced huge losses due to their high exposure to US asset-backed securities (Shin, 2012). For example, the Irish government offered a two-year liability guarantee to its banks in late 2008 since they were suffered from losses on international short-term funding. However, as the new government of Greece forecasted a budget deficit of 12.7 percent of GDP, which is more than twice of the previous estimate, the European debt crisis finally faced a new phase.

In May 2010, Greece was the first country to receive a financial assistance package from the International Monetary Fund (IMF) and other European governments. Sovereign debt spreads for Ireland and Portugal rapidly increased afterwards, and these two countries were shut out of the bond market in November 2010, and in April 2011 respectively. According to Belkin et al. (2012), the European debt crisis was mainly due to common problems faced by the European “periphery” countries, which refers to a group of mostly southern European countries including Greece, Ireland, Italy, Portugal, and Spain. Starting from 2003, sovereign bond spreads for the periphery countries had decreased swiftly as they tried to transit from their national currencies to the euro. Both the public and private sectors of these countries were able to enjoy new credit at lower prices and their debt amounts started growing.

II. Data

The dataset for this paper consists of seven total return indices for REITs in the US, France, the UK, Japan, Hong Kong, Singapore and Australia from January 2007 to

(12)

12

March 2016. According to a survey conducted by European Public Real estate Association (EPRA) in 2009, the US, Asia, Australia, and Europe account for 51%, 14%, 12%, and 22% of the global REIT market respectively. Among European REIT markets, France and the UK, have a market share of 32%, and 19%, each. Japan, Singapore, and Hong Kong have the biggest REIT market capitalizations in Asia (CBRE, 2012). Therefore, a total seven countries mentioned above will be covered in this thesis

To reflect both stock markets and bond markets, major stock index returns for each country will be used, for instance S&P 500 of the US and Hong Kong Hang Seng Index (HIS) of Hong Kong, and yields on each country’s 10 year sovereign bond as well. Table 1 illustrates which stock indices and REIT indices are employed. To capture macroeconomic factors, Chicago Board Option Exchange Volatility Index (VIX Index) will be used since VIX index is a major tool to measure near-term volatility implied in S&P 500 stock index option prices. However, VIX Index itself cannot reflect each country effects, therefore interest rates and the consumer price index (CPI) for each country for the same time period will be employed. The data is mainly collected from Datastream.

Table 1

Country Stock Index REIT Index

Japan NIKKEI 225 Stock Index TSE Reit Index

Hong Kong HANG SENG Index FTSE HONG KONG Reit Index

Singpaore STRAITS TIMES Index FTSE W SINGAPORE Reit Index

Australia S&P/ASX 200 Index FTSE AUSTRALIA Reit Index

France FRANCE CAC 40 Index FTSE FRANCE Reit Index

UK FTSE 100 Index FTSE UK Reit Index

(13)

13

Six variables will be analyzed in this thesis, which are rates of REITs index return (ReitReturn), rates of stock index return (StockReturn), 10 year government bond for each country (BondYield), interest rate or each country (InterestRate), the monthly change of VIX index (VIX), and the month over month CPI for each country (CPI). All the units of variables are percentage and for all variables, monthly data are used. Figure 1 illustrates the scatter plot among variables while Table 2 shows the correlation matrix. A positive relationship between BondYield and InterestRate is clearly depicted in Figure 1, as they have a correlation of 0.8879 from Table 2.

Table 3 depicts the descriptive statistics of the data. The mean of REITs return for the pooled dataset is -0.09% but the median is 0.65%. This is potentially due to the relatively high return of REITs for Asia Pacific region (0.0042%) considering that this is the only positive REITs return for the whole period among three region groups. This average is remarkably higher than the European REITs return average (-0.27%) and United States’ REITs return average (-0.08%). In addition, the standard deviation of REITs return for Asia Pacific region is the lowest for the whole period (6.77%), which implies that Asian Pacific REITs are more stable and less affected by both crises – the 2007 US sup-prime financial crisis and the European debt crisis.

The mean of European REITs return is the lowest for the whole period (-0.27%) along with the lowest stock return. The noticeable fact is that before the European debt crisis, which is from January 2007 to April 2010, the rate of REITs return for Euro zone (-1.76%) is even lower than that for US (-1.64%), and this also applies for the rate of stock returns. This implies that the 2007 US sub-prime financial crisis attacked European REITs/ stock markets even stronger than it did the US market. While the

(14)

14

standard deviation of the EU REITs return is always higher than that of EU stock returns regardless of sub-periods, the standard deviation of the EU REITs return is almost twice higher (10.8132%) before May 2010 than after (5.6451%). To sum up, European REITs/stock markets have been negatively affected during the 2007 financial crisis more than during the EU debt crisis, and have been more volatile than US REITs/Stock markets as well as Asia Pacific markets.

Figure 1

(15)

15

Table 2 Correlation matrix

ReitReturn StockReturn BondYield Interest Rate VIX CPI

ReitReturn 1 StockReturn 0.1107 1 BondYield -0.381 -0.0941 1 Interest Rate -0.4812 -0.2279 0.8879 1 VIX -0.0734 -0.0079 0.7043 0.4209 1 CPI 0.1448 0.0589 0.6768 0.4811 0.5942 1 Table 3 Descriptive Statistics

Pael A : Pooled data

Variable Mean Median Std. dev. Minimum Maximum N ReitReturn -0.0872 0.6469 7.2109 -44.3367 31.7289 770 StockReturn -0.0033 0.6373 5.9064 -27.2907 22.6010 770 BondYield 2.7054 2.5025 1.3536 -0.0560 6.5868 777 InterestRate 1.5631 0.6100 1.8911 -0.3550 7.8900 777 Vix 0.4603 -3.6067 26.2602 -67.9150 78.1407 770 CPI 0.1650 0.1515 0.3538 -1.7710 1.9250 770

Panel B : Asia Pacific region

Variable Mean Median Std. dev. Minimum Maximum N ReitReturn 0.0042 0.6115 6.7684 -44.3367 23.6354 440 StockReturn -0.0268 0.4295 6.2557 -24.6369 22.6010 440 BondYield 2.5158 2.3125 1.4972 -0.0560 6.5868 444 InterestRate 1.6070 0.5625 1.9297 -0.3550 7.8900 444 Vix 0.4603 -3.6067 26.2730 -67.9150 78.1407 440 CPI 0.1817 0.1555 0.4127 -1.2570 1.9250 440

Panel B (a) : Asia Pacific region before EU crisis

Variable Mean Median Std. dev. Minimum Maximum N ReitReturn -1.3929 -0.7411 9.5221 -44.3367 23.6354 152 StockReturn -0.4053 0.3656 8.0858 -24.6369 22.6010 152 BondYield 3.2247 2.7890 1.6187 1.2580 6.5868 156 InterestRate 2.5022 1.4472 2.2812 0.0989 7.8900 156 Vix 1.5076 -5.3999 27.3531 -37.6798 78.1407 152 CPI 0.1733 0.1985 0.4590 -1.2270 1.6740 156

Panel B (b) : Asia Pacific region after EU crisis

(16)

16 ReitReturn 0.7416 1.0040 4.5595 -16.7197 22.7420 288 StockReturn 0.1729 0.4295 5.0343 -22.2235 14.8737 288 BondYield 2.1319 2.0255 1.2743 -0.0560 5.7950 288 InterestRate 1.1222 0.3746 1.5053 -0.3550 5.3400 288 Vix -0.0925 -2.9662 25.7160 -67.9150 67.6009 288 CPI 0.1863 0.1420 0.3857 -1.2570 1.9250 284

Panel C: European countries

Variable Mean Median Std. dev. Minimum Maximum N ReitReturn -0.2730 0.4011 7.8759 -23.9778 31.7289 220 StockReturn -0.1137 0.8148 5.5731 -24.3466 9.8513 220 BondYield 2.9852 2.9900 1.1670 0.4400 5.4300 222 InterestRate 1.6257 0.7700 1.9125 -0.2700 6.6500 222 Vix 0.4603 -3.6067 26.3030 -67.9150 78.1407 220 CPI 0.1424 0.1250 0.2047 -0.6020 0.8290 220

Panel C (a) : European countries before EU debt crisis

Variable Mean Median Std. dev. Minimum Maximum N ReitReturn -1.7643 -2.1207 10.8132 -23.9778 31.7289 76 StockReturn -0.6352 0.7636 6.8802 -24.3466 9.8513 76 BondYield 4.2035 4.1650 0.5492 3.2500 5.4300 78 InterestRate 3.5396 4.1725 2.1113 0.4250 6.6500 78 Vix 1.5076 -5.3999 27.4441 -37.6798 78.1407 76 CPI 0.1793 0.1555 0.2408 -0.6020 0.6580 78

Panel C (b) : European countries after EU debt crisis

Variable Mean Median Std. dev. Minimum Maximum N ReitReturn 0.5141 1.3311 5.6451 -17.1052 12.8433 144 StockReturn 0.1614 0.8340 4.7453 -14.0210 9.6117 144 BondYield 2.3253 2.2500 0.8332 0.4400 4.1000 144 InterestRate 0.5890 0.6050 0.4155 -0.2700 1.6300 144 Vix -0.0925 -2.9662 25.7609 -67.9150 67.6009 144 CPI 0.1221 0.1115 0.1796 -0.2600 0.8290 142

Panel D : United States

Variable Mean Median Std. dev. Minimum Maximum N ReitReturn -0.0812 0.8688 7.5898 -37.1017 19.4913 110 StockReturn 0.3117 1.1235 5.0968 -27.2907 11.8888 110 BondYield 2.9041 2.7200 0.9048 1.5300 5.1000 111 InterestRate 1.2623 0.4250 1.6662 0.1950 5.6100 111 Vix 0.4603 -3.6067 26.3632 -67.9150 78.1407 110 CPI 0.1437 0.1735 0.3324 -1.7710 1.0480 110

(17)

17

Panel D (a) : US before EU debt crisis

ReitReturn -1.6412 -0.6584 11.5230 -23.9778 19.4913 38 StockReturn -0.5736 1.2105 7.0144 -27.2907 11.8888 38 BondYield 3.8405 3.7300 0.6592 2.4200 5.1000 39 InterestRate 2.8513 2.8300 1.9976 0.2200 5.6100 39 Vix 1.5076 -5.3999 27.6289 -37.6798 78.1407 38 CPI 0.1752 0.2420 0.4805 -1.7710 1.0480 39

Panel D (b) : US after EU debt crisis

ReitReturn 0.7421 1.2079 4.1536 -23.9778 11.7458 72 StockReturn 0.7790 1.0644 3.6877 -8.8647 9.0947 72 BondYield 2.3968 2.3000 0.5415 1.5300 3.8500 72 InterestRate 0.4017 0.3550 0.1652 0.1950 0.8900 72 Vix -0.0925 -2.9662 25.8515 -67.9150 67.6009 72 CPI 0.1263 0.1600 0.2143 -0.6390 0.5990 71 III. Methodology

To investigate the linkages among REITs, the stock market and bond markets along with macroeconomic factors, 3-step procedure is applied. First, a unit root test has been conducted to test whether variables are stationary. Second, if variables are not stationary, then a cointegration test has been performed to see whether there is a cointegration relationship among variables. Lastly, if they are cointegrated, then a panel error correction model will be employed. Otherwise, a panel auto regression model will be used.

A. Unit root test

There are two generations to test a unit root. The first generation strictly assumes that there is no cross-sectional independence (Breitung, 2000; Hadri, 2000).

(18)

18

The second generation’s assumption is more realistic, that is, there might be a cross-sectional independence between panels. Niu et al (2011) also mention that non-stationary data should be tested by the second generation methods, which include Levin-Lin-Chu (LLC), Im-Pesaran-Shin (IPS), Breitung, adjusted Dikey-Fuller test (ADF), and a test suggested by Phillips and Perron (PP).

In this paper, one of the first generation methods, Hadri LM, and two of the second generation methods, LLC and IPS are used. While LLC and IPS assume that all panels contain unit roots, Hadri LM tests the null hypothesis that all panels are stationary. Hadri (2000) asserts that the traditional null hypothesis of unit root tests needs to be reversed to bring a more powerful test. Even though the null hypothesis of LLC or IPS has been rejected, this only means that at least one panel is stationary, which does not guarantee that all panels are stationary. Therefore, Hadri LM can be used to support or deny the other two results. An autoregressive model is employed to test unit roots in panel data such as:

𝑌𝑖,𝑡 = 𝛼𝑖𝑌𝑖,𝑡−1+ 𝛽𝑖𝑋𝑖,𝑡+ 𝜀𝑖,𝑡 (1)

where 𝑖 = 1, 2, …, N are seven countries and 𝑡 = 1, 2, …, T represent the time period from January 2007 to March 2016. 𝑋𝑖,𝑡 can be either fixed effects or trends of individual panel, where 𝛼_𝑖 are the coefficients of autoregressive model. If this coefficient is less than one, then the variable is stationary. Otherwise, the variable contains a unit root.

(19)

19

B. Cointegration test

If the level variables are non-stationary and integrated of order one I(1), a cointegration test should be performed as offered by Westerlund (2007). Westerlund suggests two kinds of group-mean tests (Gt, Ga) and two kinds of panel tests (Pt, Pa) to see whether there is any long-run equilibrium. The alternative hypothesis of group-mean test is that there is a cointegrated relationship in at least one of the cross-section units, and this is similar as other traditional cointegration tests’ null hypotheses. Meanwhile, the alternative hypothesis of panel tests suggests that there is a cointegration for the panel as a whole.

C. Panel error correction model

An autoregressive distributive lag (𝑝, 𝑞₁, 𝑞₂, … , 𝑞_𝑘) can be written as

𝑦_𝑖𝑡 = ∑ 𝜆_𝑖𝑗 𝑝 𝑗=1 𝑦_{𝑖,𝑡−𝑗}+ ∑ 𝛿_𝑖𝑗′ 𝑞 𝑗=0 𝑋_{𝑖,𝑡−𝑗}+ 𝜇_𝑖 + 𝜀_𝑖,𝑡 (2)

where 𝑖 = 1, 2, …, N, 𝑡 = 1, 2, … , 𝑇 and 𝑋_𝑖𝑡 is 𝑘 × 1 explanatory variables and 𝛿_𝑖𝑗 are coefficients. 𝜆𝑖𝑗 and 𝜇𝑖 are scalars and the group-fixed effect respectively. According to Blackburne and Frank (2007), 𝑇 must be long enough so that each panel can be tested separately by the model.

If level variables in equation (2) are non-stationary and integrated of order one I(1), the error term will follow an I(0) process. To investigate dynamic relationships

(20)

20

among various financial assets including REITs, panel error correction model is employed which is can be written as follows:

∆𝑦𝑖,𝑡 = 𝜙𝑖(𝑦𝑖,𝑡−1− 𝜃𝑖′𝑋𝑖,𝑡) + ∑ 𝜆𝑖𝑗∗ 𝑝−1 𝑗=1 Δ𝑦𝑖,𝑡−1+ ∑ 𝛿𝑖𝑗′∗ 𝑞−1 𝑗=0 Δ𝑋𝑖,𝑡−𝑗+ 𝜇𝑖+ 𝜀𝑖,𝑡 (3) where 𝜙_𝑖 = −(1 − ∑𝑝_𝑗=1𝜆_𝑖𝑗), 𝜃_𝑖 = ∑𝑞_𝑗=0𝛿_𝑖𝑗/(1 − ∑ 𝜆_𝑘 _𝑖𝑘), 𝜆_𝑖𝑗∗ _{= − ∑} _𝜆 𝑖𝑚) 𝑝 𝑚=𝑗+1 𝑗 = 1, 2, … , 𝑝 − 1, and 𝛿_𝑖𝑗′∗ _{= − ∑} _𝛿 𝑖𝑚) 𝑞 𝑚=𝑗+1 𝑗 = 1, 2, … , 𝑞 − 1.

The parameter 𝜙 is the error correction term, or so-called the speed of adjustment term. The error correction term should be negative and significant if the variables have a long-run equilibrium. If 𝜙𝑖 equals to zero, then this would be an evidence that variables do not have any long-run relationship. The vector 𝜃_𝑖′_{will show} the specific long-run equilibrium between two variables.

As the equation (3) applies to this thesis, 𝑦 presents the rate of REIT returns and 𝑋 refers to stock returns, bond yields, interest rates, VIX, and CPI. By looking at 𝜃′, we can see the long-run relationship between each variables and REIT returns. On the other hand, 𝜆_𝑖𝑗∗_{and 𝛿}

𝑖𝑗′∗ are short-term parameters of the lagged difference of REIT returns and other rates, respectively. Of particular importance is 𝜃′, since it can tell diversification effects of REITs. For example, if 𝜃′ of stock returns is below (above) zero, then there is (not) diversification effect between stocks and REITs.

(21)

21

IV. Empirical Results A. Unit root test

LLC rejects the null hypothesis for the level values of ReitReturn, StockReturn, VIX, and CPI at the 1% significance level. This is similar with IPS results, under which the level values of BondYield and InterestRate seem to have unit roots at least at one panel. Under the null hypothesis of Hadri LM, which is that all panels contain unit roots, Table 4 shows that all level variables (ReitReturn, StockReturn, BondYied, InterestRate and Vix) contain unit roots. After taking a first-difference, the variables become stationary for all three tests. Therefore, we can conclude that the level variables are non-stationary and integrated of order one I(1).

Table 4 Unit root test

Three unit root tests are performed which are LLC, IPS, and Hadri LM. The null hypotheses of LLC and IPS are that all panels have unit roots, while Hadri LM tests a reversed null hypothesis (H0: All panels are stationary, and Ha: Some panels contain unit roots).

LLC IPS Hadri LM

Level First diff. Level First diff. Level First diff.

ReitReturn Statistic -17.37*** -31.28 *** -11.46*** -18.63 *** 1.38* -1.98 p-value 0.00 0.00 0.00 0.00 0.08 0.98 StockReturn Statistic -15.76*** -30.03*** -11.77*** -20.24 *** 3.43*** -2.08 p-value 0.00 0.00 0.00 0.00 0.00 0.98 BondYield Statistic -0.27 -15.85*** -1.60 -10.60*** 18.19*** -0.89 p-value 0.39 0.00 0.41 0.00 0.00 0.81 InterestRate Statistic -0.36 -13.74*** -1.88 -10.81*** 13.98*** -0.10 p-value 0.36 0.00 0.16 0.00 0.00 0.54 Vix Statistic -33.81*** -43.41*** -3.48*** -20.21*** 2.73*** -1.52 p-value 0.00 0.00 0.00 0.00 0.00 0.94 CPI Statistic -12.66*** -27.55 *** -10.75*** -19.32*** 5.14 -2.77 p-value 0.00 0.00 0.00 0.00 0.000 0.1

Note: *, ** and *** denote rejection of null hypothesis of unit root based on their P-value at the 0.1, 0.05 and 0.01 significance levels respectively

(22)

22

B. Cointegration test

From both group-mean tests and panel tests, a null hypothesis of no integration has been rejected at the 1% significance level and confirmed that variables are cointegrated and therefore have a long-run equilibrium (Table 5). Since level variables of this thesis are non-stationary and integrated of order one I(1), the error term will follow an I(0) process and a panel error correction model can be used.

Table 5 Cointegration test

Westerlund ECM panel cointeration test under a null hypothesis of no cointegration

Statistic Value Z-value P-value Gt -3.1160 -1.30* 0.0000 Ga -25.7220 -3.45*** 0.0000 Pt -8.3700 -1.94** 0.0000 Pa -25.8870 -4.63*** 0.0000

Note: *, ** and *** denote rejection of null hypothesis of no cointegration based on their P-value at the 0.1, 0.05 and 0.01 significance levels respectively

C. Panel error correction model

Table 6 shows the coefficients and z-values of the panel error correction of the pooled data, Asia pacific region, and European countries. The dependent variable is the return of REITs and each region has been tested divided into two sub-periods separately, which is between January 2007 and Aril 2010 (Panel B) and between May 2010 and March 2016 (Panel C). Each panel shows long run and short run effects of each variable. Long run coefficients and the error correction term are of particular interest of this thesis. As mentioned above, the coefficient of the error correction term

(23)

23

should be negative and significant lying between -1 and 0 to prove that a factor will converge to the equilibrium in the long term. ΔREIT t-2 is used as an instrumental variable.

In panel A, while all coefficients for stock return are positively significant, the coefficient for European stock return is the highest (1.20%) meaning that when the stock return increases by 1% the REITs return increases by 1.20%. Asian pacific stock return has the lowest coefficients (0.80%) implying the return of REITs is less sensitive to the stock markets than that of US or European REITs. On the other hand, the coefficient for BondYield is negative and significant at 1% level only for European markets which gives investors an opportunity to enjoy diversification effects if they put bonds and European REITs in a same portfolio. The error correction terms are significantly negative for all regions and the absolute value is the highest in Asian pacific region. It means this region goes back to the long-run equilibrium at the fastest speed and this result is in line with the finding from the descriptive statistics that is Asian pacific REITs are most stable one.

Panel B reflects the impacts from the 2007 financial crisis and the long run coefficient for stock returns is the highest for the European stock market (0.97%). Additionally, the coefficients of the European stock returns are the highest regardless of the sub-periods, implying stock returns of this region is most sensitive to REIT returns which causes higher risks. The coefficient of StockReturn for Asian pacific region is not significant whereas the one for the pooled data is significant at 1% level (0.95%). With regard to bond returns, the pooled data and Asia pacific region also show the negatively significant coefficient, -6.01% and -5.43% respectively, but the one for European region is the lowest again (-8.10%). This can be very interesting for investors because even

(24)

24

though there was a severe crisis, European REITs still had the greatest diversification effects when it combined with fixed income securities. REITs can be a wise choice to hedge against stock markets during crisis. The error correction term for each panel is all significantly negative again and is the lowest for Asia pacific region (-0.6015). This speed of adjustment of Asia pacific region is the lowest (highest absolute value) among all error correction terms, which means Asia pacific region could overcome the 2007 financial crisis fastest.

Panel 3 shows only the European debt crisis period. The noticeable fact is that all StockReturn coefficients for this period are higher than those for the other two periods. This suggests that returns of REITs for all regions are more elastic, or volatile, on stock indices after European debt crisis. For example, if FRANCE CAC 40 Index rebounds by 1%,

FTSE FRANCE REITs Index rallies by 1.77%. The differences between stock return

coefficients for this period are notably larger than those for other periods, for instance, they do not exceed 0.34%points for Panel A and Panel B. However, in Panel C, the StockReturn coefficient for Europe is 0.84%points higher than that for the pooled data. This might be because the European debt crisis may have affected the European economy stronger than it did with regard to other regions’ economies. The speed of adjustment is the lowest for the pooled data.

(25)

25

Table 6

Panel error correction model

The dependent variable is the first-differenced ReitReturn and independent factors are stock return, bond yield, interest rate, VIX and CPI. This test has been applied for three periods, which are the whole period (Jan 2007-March 2016), and two sup-periods (Jan 2007 – April 2010 and May 2010-March 2016).

Pooled data Asia pacific region Europe

Coefficient Z-value Coefficient Z-value Coefficient Z-value

Panel A: Jan 2007 - March 2016

Long Run StockReturnt 0.8638*** 12.21 0.8085** 2.39 1.2003*** 3.12 BondYieldt -1.1475 -1.53 -1.8714 -1.51 -0.4496*** -2.62 Interest Ratet 0.9332 1.27 1.5415 1.25 0.4979 1.54 VIXt -0.5678*** -4.71 -0.5309*** -2.63 -0.7425*** -8.95 CPIt 3.8804 1.13 2.9371 0.86 1.8246 0.15 Short Run Error Correctiont-1 -0.5079*** -12.51 -0.5502*** -10.22 -0.4850*** -6.51 ΔStockReturn t-1 10.2796 1.05 0.3969** 2.43 0.5534*** 5.68 ΔBondYield t-1 -0.6065 -1.58 -0.9926 -1.59 -0.2053*** -4.13 ΔInterest Rate t-1 0.4945 1.32 0.8171 1.3 0.2174* 1.82 ΔVIX t-1 -0.2932*** -4.47 -0.3009*** -2.7 -0.3539*** -23.48 ΔCPI t-1 1.3879 0.87 1.2984 0.83 -0.0023 0 ΔREIT t-1 -0.1893*** -3.88 -0.1783* -1.88 -0.1722 -1.64 ΔREIT t-2 -0.1379** -2.68 -0.2628** -1.99 -0.1417*** -6.39 Constant 0.7166* 1.93 1.1116* 1.87 0.3304*** 8.11

Panel B: Jan 2007 - April 2010

Long Run StockReturn t 0.9503*** 10.13 0.7538 1.6 0.9675*** 3.96 BondYield t -6.0123*** -3.87 -5.4349** -2.11 -8.0968*** -4.02 Interest Rate t 2.766** 1.97 3.6079 1.47 2.2256** 2.17 VIX t -0.7224*** -4.69 -0.7698*** -2.86 -0.7396*** -3.99 CPI t 12.3723* 1.76 12.1191 1.17 5.5431 0.36 Short Run Error Correction t-1 -0.5502*** -10.54 -0.6015*** -8.85 -0.5547*** -15.23 ΔStockReturn t-1 13.177 1.03 0.3578 1.61 0.5277*** 5.26 ΔBondYield t-1 -3.43*** -3.65 -3.4446** -2.19 -4.4176*** -5.37 ΔInterest Rate t-1 1.6933* 1.87 2.3245 1.48 1.1971** 2.45

(26)

26 ΔVIX t-1 -0.4299*** -3.79 -0.5086*** -2.68 -0.4035*** -5.31 ΔCPI t-1 5.1646* 1.66 5.5115 1.25 2.5065 0.3 ΔREIT t-1 -0.2690** -2.23 -0.3423* -1.67 -0.2645** -2.51 ΔREIT t-2 -0.2422*** -5.22 -0.2847*** -4.45 -0.1231*** -5.58 Constant 7.5766*** 3.17 4.6668 1.64 14.5979*** 9.45

Panel C: May 2010 - March 2016

Long Run StockReturn t 0.9308*** 6.7 1.0541*** 5.36 1.7700*** 3.47 BondYield t -0.2621 -0.75 -0.737 -1.55 0.4634 1.38 Interest Rate t -4.4388 -1.32 -6.0518 -1.02 -0.1658 -0.2 VIX t -0.2255 -1.4 -0.0678 -0.48 -0.755*** -13.95 CPI t -3.0635 -1.3 -3.7685 -1.42 -0.0468 -0.01 Short Run Error Correction t-1 -0.4283*** -11.94 -0.4194*** -11.86 -0.3771*** -4.44 ΔStockReturn t-1 8.8115 1.06 0.4426*** 4.5 0.6242*** 14.91 ΔBondYield t-1 -0.102 -0.85 -0.278* -1.75 0.1462* 1.67 ΔInterest Rate t-1 -2.0908 -1.28 -2.7385 -0.96 0.0089 0.03 ΔVIX t-1 -0.0723 -1.12 -0.0152 -0.27 -0.2801*** -6.41 ΔCPI t-1 -1.5831* -1.69 -1.5474 -1.41 -0.6689 -0.23 ΔREIT t-1 0.1459* 1.7 0.1022 1.17 -0.0113 -0.18 ΔREIT t-2 0.0688* 1.85 0.0379 0.86 -0.0044 -0.07 Constant 2.906** 2.09 1.2733** 1.99 -0.2210*** -2.89

Note: *, ** and *** denote rejection of null hypothesis of unit root based on their P-value at the 0.1, 0.05 and 0.01 significance levels respectively

V. Conclusion

This thesis has explored the dynamic relationship between REITs return, stock returns, and bond yields. A total of seven countries have been investigated, divided into three regions (pooled, Asia pacific, and Europe) for the period between January 2007 and March 2016. From the descriptive statistics analysis, I found that Asia pacific area is the most stable market during the sample period and the 2007 sup-prime crisis affected

(27)

27

European market stronger than European debt crisis did. These findings are in line with results from the panel error correction model. After analyzing three regions for the whole sample period and two sub-periods, the coefficients of stock return for the Asia Pacific market are usually the lowest among three regions which implies that Asia Pacific REITs returns are more stable in the long-term.

With regard to diversification effect, all stock return coefficients, regardless of regions, are significantly positive suggesting investors cannot expect diversification effects from a portfolio consisting of stocks and REITs. However, the coefficients of fixed income securities for European market is negative and significant at the 1% level for the whole sample period, even for during the 2007 financial crisis period. Therefore, European REITs and fixed income securities can be a good hedging strategy for investors. The BondYield coefficient for Asia Pacific market is significant at the 5% level during the 2007 financial crisis period as well, but it is not significant for the whole sample period of 2007-2016. The scale of diversification effects are bigger for European REITs as well since the coefficients for BondYield are always larger than that of other regions. This suggests that European REITs offer more diversification effects in general, but during the severe crisis, Asia pacific REITs can also offer some of them.

The error correction term are significantly negative for all regions and for all periods showing that there is long-run equilibriums. Asia pacific region always has lower speed of adjustments that European countries, suggesting the rate of REIT returns can converge to long-run equilibriums faster in Asia Pacific region.

The limitation and possible further research would be adding more macroeconomic variables. As mentioned above, VIX index is based on S&P 500 stock

(28)

28

index option prices so that the information it contains is skewed to the American market. As previous research claims that macroeconomic variables including industrial production, GDP, and money supply can also affect REIT markets (McCue and Kling, 1994;Chang et al., 2011), a more accurate study can be done by adding macroeconomic factors.

(29)

29

REFERENCES

Abrigo, M. R. M., and Love, I., 2015, Estimation of panel vector autoregression in Stata: a Package of Programs, The Quarterly Review of Economics and Finance, 46, 190-210. Acharya, V.V., Eisert, T., Eufinger, C. and Hirsch, C.W., 2015, Real effects of the sovereign debt crisis in Europe: Evidence from syndicated loans, Available at SSRN 2612855. Ada Choi and Leo Chung, 2012, Asian REITs return to acquisition mode but 2012 brings

new challenges, CBRE Research

Belkin, P., Mix, D.E. and Weiss, M.A., 2012, The eurozone crisis: overview and issues for congress, Congress Research Service, 42377.

Blackburne, E. F., & Frank, M. W., 2007, Estimation of nonstationary heterogeneous panels, Stata Journal, 7(2), 197–208.

Breitung, J., 2000, The Local Power of Some Unit Root Tests for Panel Data, Available at http://edoc.hu-berlin.de/series/sfb-373-papers/1999-69/PDF/69.pdf

Canova, F., and Ciccarelli, M., 2013, Panel Vector Autoregressive Models A Survey. ECB Working Paper Series, 15, 1–53.

Chang, C.-Y., Chou, J.-H., and Fung, H.-G., 2012, Time dependent behavior of the Asian and the US REITs around the subprime crisis, Journal of Property Investment &

Finance, 30(3), 282–303.

Chiang, M.-C., Tsai, I.-C., and Sing, T.-F., 2013, Are REITs a good shelter from financial crises? Evidence from the Asian markets, Journal of Property Investment & Finance, 31(3), 237–253.

Clayton, J. and MacKinnon, G., 2001, The time-varying nature of the link between REIT, real estate and financial asset returns, Journal of Real Estate Portfolio Management, Vol. 7 No. 1, 43-54. Economics and Statistics, 69(6), 709-748.

Ernst & Young, 2016, pdf, Economic contribution of REITs in the United States European public real estate association EPRA, pdf, 2009, Global REIT survey

Farrell, S., Global REIT Survey 2012: Australia, Global REIT Survey, Retrieved 2013-02-15

Glascock, J., Lu, C., and So, R., 2000, Further evidence on the integration of REIT, bond, and stock returns, Journal of Real Estate Finance and Economics, 194, 177–194. Hadri, K., 2000, Testing for stationarity in heterogeneous panel data, The Econometrics

(30)

30

Lane, P.R., 2012, The European sovereign debt crisis, Journal of Economic

Perspectives, 26(3), 49-67.

Laopodis, N., 2009, REITs, the stock market and economic activity, Journal of Property

Investment & Finance, 27(6), 563–578.

Lee, S., & Stevensson, S., 2005, The case for REITs in the mixed-asset portfolio in the short and long run. Journal of Real Estate Portfolio Management, 11(1), 55–80. Levin, A., C. F. Lin, and C. S. Chu, 2002, Unit root tests in panel data: asymptotic and finite

sample properties, Journal of Econometrics, 108(1), 1-24.

Liu, J., Loudon, G., and Milunovich, G., 2012, Linkages between international REITs: the role of economic factors, Journal of Property Investment & Finance, 30(5), 473–492. Loo, W.K., Anuar, M.A. and Ramakrishnan, S., 2016, Integration between the Asian REIT

markets and macroeconomic variables. Journal of Property Investment & Finance, 34(1), pp.68-82.

McCue, T. and Kling, J., 2009. Real estate returns and the macroeconomy: some empirical evidence from real estate investment trust data, 1972-1991. Journal of

Real Estate Research.

NAREIT, 2002, Diversification Benefits of REITs—An Analysis by Ibbotson Associates, Available at www.nareit.com.

Niskanen, J., & Falkenbach, H., 2010, REITs and correlations with other asset classes: a European perspective, Journal of Real Estate Portfolio Management, 16, 227–239. Niu, S., Ding, Y., Niu, Y., Li, Y. and Luo, G., 2011, Economic growth, energy conservation

and emissions reduction: a comparative analysis based on panel data for 8 Asian-Pacific countries, Energy Policy, 39(4), 2121–31

Semer, S. L., 2009, A brief history of US REITs, Canadian Tax Journal, 57(4), 960–971. Shin, H.S., 2012. Global banking glut and loan risk premium. IMF Economic Review, 60(2),

55-192.

Stracca, Livio, 2013, The global effects of the euro debt crisis, ECB Working Paper No. 1573. Available at SSRN: http://ssrn.com/abstract=2302515

Westerlund, J., 2007. Testing for error correction in panel data. Oxford Bulletin of Economics and statistics, 69(6), pp.709-748.

The dynamic relationship among REITs, stock and bond returns around the European debt crisis