The Pricing of Idiosyncratic Risk in European REITs An Empirical study Henok Nuguse Student number: 5995906 University of Amsterdam Thesis Supervisor Dr. M. Petrova MSc Business Economics
Acknowledgment
I would like to thank my thesis supervisor Dr. M. Petrova for her invaluable advice and feedback during the course of this master thesis. Her guidance and recommendations for this research have aided the successful completion of this paper.
Abstract
In this study, total volatility of European REIT returns is decomposed into systematic and idiosyncratic components. The study of the composition of total volatility shows that idiosyncratic risk accounts for a large proportion of total risk with around 80% or more between 2004 and 2014.
Contrary to conventional asset pricing theory, I find that the idiosyncratic risk component has significantly priced in REIT returns. On average, a 1 percentage point increase in idiosyncratic volatility leads to an excess risk premium of 17 basis points. Furthermore, this effect remains when controlling for institutional ownership concentration in addition to size, value and momentum factors.
The Pricing of Idiosyncratic Risk in European REITs An Empirical study
Modern portfolio theory advocates that, due to diversification of securities, firm-specific risk can be eliminated at the portfolio level. Sharpe (1964) builds on this principle and advocates that in equilibrium, investors hold the market portfolio to achieve optimal diversification. Consequently, as the volatility of security returns is composed of a market-related (systematic) and a firm-specific (idiosyncratic) component, traditionally, only systematic volatility has been considered relevant for the pricing of securities. However, research has shown that for various reasons, many investors do not hold well diversified portfolios in reality. Thus, in these circumstances of under-diversification, idiosyncratic volatility still matters to investors. As a result, in recent years the study of idiosyncratic volatility has received more attention. Though, the literature is still inconclusive regarding its significance in the pricing of securities.
This research studies the pricing of idiosyncratic volatility for European real estate investment trusts (REITs) and in particular describes the change in pattern and composition of volatility over the years. Idiosyncratic volatility has mainly been studied in relation to the US equity market. Several studies suggest that firm-specific risk is priced in securities. These findings contradict the proposition of classical asset pricing models that only market risk should be priced. Furthermore, whereas numerous studies on REITs are directed to the study of
risk/return behavior in relation to stock markets and direct real estate, less research is focused on the pricing of risk for REITs.
I focus my study on the Western European market where over the last decade REITs have enjoyed considerable growth. As a result of the region’s economic, financial and monetary
integration, the market has been characterized by increasingly more synchronized equity as well as real estate markets. For instance, Yang et al. (2005) describe the dynamic linkages established among the larger European public real estate markets. Since the European REIT market is a significant component of the global securitized real estate market, it is important to understand the dynamics of its typical REIT volatility and its components; market volatility and
idiosyncratic volatility as well as their relative influences on expected REIT returns. Although this is being studied increasingly more in the US, due to its relative novelty less is known about the characteristics of its European equivalent.1 Hence, a closer investigation is warranted as idiosyncratic volatility is considered important in various applications such as in portfolio context and for the pricing of derivatives. Furthermore, idiosyncratic is shown to make up a sizable portion of total volatility. Dennis and Strickland (2004) show that idiosyncratic
volatility has increased between 1962 and 1997 in the US. Additionally, they found idiosyncratic volatility to be positively related to institutional ownership. In addition to studying the pricing of idiosyncratic risk, this paper will focus on its relation with institutional ownership.
The main contribution of this paper is the study of the composition and temporal variations in European REIT volatility in relation to expected European REIT return. In particular, this study complements the existing literature by decomposing volatility into systematic and idiosyncratic components and studying the relationship between institutional ownership and the pricing of idiosyncratic risk. So far, limited research has focused on the effects of institutional ownership on the pricing of idiosyncratic risk for European REITs. This study thus aims to answer the following questions:
1 Liow & Addae- Dapaah (2010) for instance, find that both market and idiosyncratic variance are
time-varying for the US REIT. Moreover, they conclude that idiosyncratic variance represents a dominant component of a REIT firm’s total variance.
Research question 1:
To what extent is idiosyncratic volatility priced in the return of European REITs
Research question 2:
To what extent does institutional ownership explain the pricing of idiosyncratic volatility
This paper is structured as follows. Section 2 provides a literature review. The data and methodology used for this research are provided in Section 3. In Section 4 I report and interpret the test results to develop the main findings. Finally, section 5 concludes the study.
II. Literature review
A. Asset Pricing Theory
According to conventional asset pricing theory, CAPM, rational investors should be able to eliminate all risk contained in individual risky assets other than that part that coincides with movements in overall economic activity. This can be done by holding risky assets in a diversified portfolio. Sharpe (1964) reasons that combinations of risky assets that are efficient will be
perfectly correlated.2 Furthermore, since only efficient combinations will be perfectly correlated, Sharpe (1964) concludes that this can reasonably be attributed to a common dependence on overall level of economic activity. As a result, in equilibrium a simple linear relationship exists between the expected return and standard deviation of efficient combinations of risky assets. This linear risk/return relationship is named the capital market line (CML), with the slope
referred to as β, as it evidently depicts all attainable risk/return combinations available to rational investors holding optimally diversified portfolios. This β, is thus the single important factor considered relevant for the pricing of securities under the assumption that rational investors optimally diversify. Although the CAPM model has been widely popular, it has its
shortcomings. Criticisms on the model largely stem from the model’s assumption that investors choose their portfolio according to the mean-variance criterion of Markowitz (1952). For instance, the model makes simplifying assumptions regarding the way investor expectations are formed as well as the period over which expectations are formed and utility maximization occurs. These issues have been addressed by ensuing research. Merton (1973), for instance addresses the single-period nature of the CAPM model by developing an inter-temporal model.
2 Where efficiency in this context means the maximization of expected return given the constraint of a
Williams (1977) develops the CAPM under heterogeneous investor expectations about means in the context of continuous trading and continuous processing of available information hence, addressing the temporal critique. Still, these adapted models revolve around the traditional asset pricing model and thus imply that the cross-section of expected returns is explained by a single dominant factor β. Conversely, Fama and French (1992) contend that if assets are priced
rationally, then based on their results security risks are multidimensional. With one dimension of risk proxied by the size of a firm in terms of market value of equity and the other dimension proxied by book-to-market equity ratio. Despite the relevance of these idiosyncratic risk dimensions, often referred to as size and value factors, their model is still consistent with traditional asset pricing theory since they impose a linear factor structure that is consistent with the model produced by Merton (1973). What is more, these risk factors are found to be priced in other research as well (see for example Fama and French (1993) and Carhart (1997)).
B. Deviations from Asset Pricing Theory
As discussed in the last section, based on conventional asset pricing models, traditionally only systematic risk has been considered relevant for the pricing of securities. As such, these models have dominated empirical research focus. Fama and French (1992) have provided an important extension by including idiosyncratic factors in their analysis. However, in preexisting literature other theories have been developed that consider the relevance of idiosyncratic factors in security pricing. Merton (1987) for instance, is one of the first researchers to consider the relevance of idiosyncratic risk in predicting stock returns. Based on his theory, he expects that idiosyncratic risk positively relates to expected returns. Moreover, his theory is based on the assumption of investor under-diversification. Malkiel and Xu (2002) empirically support the
theory of Merton (1987). They argue that under circumstances of investor under-diversification, the pricing of solely systematic risk, as advocated by traditional CAPM theory, may not hold. More importantly, they contend that if for some exogenous reason one group of investors fails to optimally diversify, the remaining investors will also be unable to hold the market portfolio and thus idiosyncratic risk could be priced to compensate rational investors. In their study, they use a proxy for the market wide undiversified idiosyncratic risk factor. Using both the Fama and MacBeth (1973) and Fama and French (1992) frameworks for the period 1935 to 2000 for US stocks and 1975 to 2000 for Japanese stocks, they find that a 1% increase in this proxy results in a 0,57% increase in returns. Malkiel and Xu (2003) further relate their results to institutional ownership behavior, stating that institutional investors often deliberately structure their portfolios to accept considerable idiosyncratic risk in an attempt to obtain extraordinary returns. In
addition, other explanations for investor under-diversification are institutional constraints, taxes and liquidity needs among others. Goyal and Santa-Clara (2003) also establish a link between idiosyncratic risk and returns although they use lagged average stock variance as a predictor. They find a willingness to pay a fee of at least 60 basis points annually by investors to invest in a strategy that forecasts the market based on average stock variance for the period between 1963 to 1999 using US securities.
C. European public REITs and REIT pricing
Since the mid 2000’s the European securitized real estate market has grown substantially. The growth of the public REIT market is not surprising given that the structure enables investors to obtain exposure to real estate assets while requiring substantially less capital and increased liquidity. The publicly traded REIT structure has therefore largely benefitted the small individual
investor as well as small institutional investors (Pagliari et al., 2003). REITs by definition invest their funds primarily in real estate assets by owning and in most cases operating
income-producing commercial property. Most of these incomes in the form of rental income and leases are subsequently paid out as dividends. Accordingly, REITs can provide relatively high income and capital appreciation but are generally also relatively volatile compared to direct real estate investments, especially due to their public trading (Niskanen & Falkenbach, 2012).
As the securitization of real estate has become increasingly more prevalent there has been a growing base of literature on determinants for REIT pricing. For instance, the literature is still inconclusive on whether REIT risk and return characteristics better resemble that of equity stocks or their underlying assets, direct real estate (Hoesli & Oikarinen, 2012). These
characteristics are important from the perspective of determining how they influence the benefits of diversification in a broad investment portfolio. It is generally accepted however, that there are benefits of real estate allocation for the performance of a mixed asset portfolio. Hoesli et al. (2004) for example, show that significant benefits of diversification can be achieved in a mixed-asset portfolio by investing in direct real estate. By making a cross-country comparison they find that a weighting of direct real estate in the 5% to 15% range reduces portfolio risk by 5% to 10% for domestic investments and 10% to 20% for international investments. Furthermore, Hoesli and Oikarinen (2012) focus on the so-called duality of real estate asset markets (i.e. trading in both public and private markets) by researching the extent to which a common “real estate factor” affects both direct and indirect real estate assets over the long horizon. Their findings show that REITs generally do indeed follow a common trend with direct real estate. For 5 of the 7 markets they investigated, long run cointegration is found between REIT and direct market total return indices. This indicates a strong dual market performance relationship.
Niskanen and Falkenbach (2012) make a distinction between two types of listed real estate companies. These are real estate operating companies (REOCs) and real estate investment trusts (REITs). The difference in structure between the two mainly concerns the distribution requirements of REITs that allow them to avoid corporate-level taxation. As described by Niskanen and Falkenbach (2012), REITs therefor effectively resemble mutual funds by passing-through most of their earnings as dividends. Although there are differences with respect to national REIT legislation, according to Niskanen and (2012) the majority of the European REIT structures exhibit similar features. Their study focuses on liquidity of European real estate equities and mainly the difference between REITs and REOCs. They find that REITs have higher correlation with other equities than REOCs and that they are 30% more liquid, measured by daily turnover ratio. Moreover, they find that REITs’ liquidity follow a somewhat similar walk as the stock market. They suggest that the superior liquidity of REITs can be explained by the
limitations regarding REIT ownership structures.
Bond et al. (2003) used an international CAPM model to investigate risk and return characteristics for publicly traded real estate companies. They found that a country-specific risk factor was highly significant even after controlling for the effects of global market risk.
Moreover, they find that a country-specific value factor provides additional explanatory power. Based on their findings they conclude that investors diversifying across real estate markets should consider different dimensions of the real estate market that relate to fundamentals such as size and book-to-market value. Karolyi and Sanders (1998) employ a multiple-beta asset pricing model to examine predictable components in REIT returns and find that REITs have comparable return predictability to stock portfolios. These results were based on a pricing model for return predictability that incorporates time-varying economic risk premiums. Moreover, they conclude
that these macroeconomic economic risk factors such as term spread, risk spread and stock capitalization represent a large fraction of the return predictability of REITs. Furthermore, Liow and Addae-Dapaah (2010) also find time varying pricing factors of REITs. Specifically, they link idiosyncratic risk to the pricing of REITs. In a study for US REITs from 1988 till 2008 they find that idiosyncratic variance is time-varying and represents a significant proportion of around 80% of REIT total variance. Also, this idiosyncratic variance has been declining steadily over time while average US REIT correlation has increased. This decline in firm specific risk is attributed to the US REIT market evolution according to Liow and Dapaah (2010). Hence, increased REIT industry development is related with the extent to which the idiosyncratic variance component makes up of total REIT return variance. Moreover, they find that there is a positive relationship between idiosyncratic risk and expected returns. Thus a risk premium is present although according to standard asset-pricing theories such as CAPM idiosyncratic risk should not be priced in expected return. The main argument for idiosyncratic risk pricing according to Liow and Dapaah (2010) is investor under-diversification so that investors require compensation for firm-specific risks. This is in line with the findings from Pagliari et al. (2003) that investors in REITs tend to be small institutional investors or individuals who do not hold diversified portfolios. Thus REITs resemble small capitalization stocks.
Traditionally, risk and return characteristics of real estate and real estate related assets have been heavily based on market appraisals (Peterson and Hsieh, 1997). However the drawback of appraisal-based data is that they are not transaction driven. Peterson and Hsieh (1997) provide a study of US public real estate assets by using a factor model similar to that of Fama and French (1993). Furthermore they make a distinction between equity REITs and mortgage REITs. They find that there is a relation between book-to-market factor and REIT
returns when analyzing data in the period of 1976 to 1992. Ali et al. (2003) find in their study of return drivers that return premiums for high book-to-market REITs are linked to the amount of idiosyncratic risk present in the total REIT variance. Moreover, they also link the book-to-market effect with lower investor sophistication and higher transaction costs. Their results are consistent with the market mispricing explanation for the book-to-market anomaly posed by Shleifer and Vishny (1997). According Shleifer and Vishny (1997) the anomaly is not arbitraged away as a result of risk aversion by arbitrageurs, who are mainly undiversified investors. If the book-to-market effect arises due to mispricing, then this effect would be higher for REITs with higher idiosyncratic risk. Therefore due to under-diversification of arbitrageurs this mispricing becomes persistent as arbitrageurs are concerned with idiosyncratic risk. Ooi, Wang and Webb (2009) study the pricing of idiosyncratic risk for US REITs. By decomposing risk into a market and firm-specific component, they find that idiosyncratic risk dominates the total volatility of REIT returns in the period between 1990 and 2005. They employ both a CAPM and Fama and French (1992) framework to find similar results with both models. They document an increase of
0,103% with every 1% increase in idiosyncratic risk. Sun and Yung (2009) also study the pricing of risk in REITs using similar factor models. Similarly, they find a significant positive relation between idiosyncratic risk and REIT returns. They document excess returns of between 0,41% to 0,45% per month for different idiosyncratic risk strategies over 12-, 24- and 36-month holding periods.
III. Data and Methodology
This section first presents the data used for the study of the composition and temporal variation in European REIT return volatility. Next, the estimation procedure, variable
construction and control variables used are provided. I conclude this section by presenting descriptive statistics of the sample studied.
A. Data
For this paper, the constituents of the SNL Europe REIT index are studied as these best reflect the universe of publicly traded REITs in Western Europe. Turkish companies are
excluded from the analysis so as to maintain consistency with the control variables used in a later stage of this research.3 The data was gathered from the first of June 2004, the base month for the SNL Europe Index, until the first of June 2014. For some of the individual countries and
constituents represented in the index, REIT return data could have been obtained for earlier periods. However, in this study I limit the research to start from the base date of the SNL Europe REIT index since this ensures a dataset comprehensive enough for an empirical study of risk and returns relations. Both monthly and daily return data are obtained from the SNL financial
database as well as data on institutional holdings. Thomson Reuters DataStream is used to obtain information on number of shares outstanding, market value, debt-to-asset ratio and share
turnover. In addition, I use Fama and French (1993) size (SMB) and book-to-market (HML)
3 Fama-French factors will be used in a later stage for the multivariate analysis. These calculations are
based on the developed European countries, Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, and the United Kingdom.
factors as well as a momentum factor (WML) which are downloaded from Kenneth French's website.4
The MSCI Europe index was chosen as a benchmark so as to best represent the European REIT market. The benchmark represents 15 developed markets countries in Europe and covers approximately 85% of the free float-adjusted market capitalization across the European
developed market equity universe. As can be seen from Figure 1, the SNL Europe REIT index total return series as of June 2004 moves synchronous to the total return series of the MSCI Europe index.
Figure 1: Total return series for SNL Europe REIT index and the benchmark over 2004 - 2014
B. Volatility Components and Measurement
The dynamics of idiosyncratic volatility can be studied at either the aggregate or at the firm level. In the following subsection I provide a decomposition approach of volatility into its idiosyncratic component at the firm-level. For this I use the within-month daily return data to compute monthly idiosyncratic volatility for each REIT where monthly idiosyncratic volatility is computed only for those months for which at least 15 days of return data is available. In the section thereafter, idiosyncratic volatility is related to REIT returns and firm-specific characteristics using a cross-sectional regression.
1. Idiosyncratic risk measurement
Similar to the procedure of Sun and Yung (2009), I measure idiosyncratic volatility as the standard deviation of the residuals of the capital asset pricing model. The following regression is run for each REIT to obtain the residuals: 5
𝑅𝑖𝑡− 𝑅𝑓𝑡 = 𝛼𝑖𝑡 + 𝛽𝑖𝑡(𝑅𝑚𝑡− 𝑅𝑓𝑡) + 𝜀𝑖𝑡 (1)
where 𝑡 is the daily time interval and individual firms are denoted by an 𝑖 subscript. Subsequently monthly idiosyncratic volatility (IV) is then computed for each REIT as:
𝐼𝑉 = (1 𝑁 ∑ 𝜀𝑖𝑡 2 𝑁 𝑡=1 ) 1/2 (2)
5 For the multivariate regression discussed in the next section, I use both CAPM and Fama-French factors
for Developed Europe as computed by Kenneth French. Unfortunately Kenneth French does not yet provide these factors on a daily basis, therefore for the within-month computations only the CAPM is used. This naturally will affect the results of this study to the extent that Fama-French factors affect daily returns. Following research could thus further expand upon the model used to assess the effect of a different model specification.
where, 𝑁 is the number of within-month days used in regression (1) and 𝜀𝑖𝑡 the daily residual for REIT 𝑖. Finally For comparability and use in the monthly regression, the idiosyncratic volatility computed from the within-month regression is multiplied by the square root of the average number of return observations per month.
2. Multivariate regressions
In a second set of regressions, I determine the explanatory power of idiosyncratic volatility for REIT expected returns. I use both the CAPM and a Carhart (1997) four factor model to regress monthly excess return onto idiosyncratic volatility for each REIT.6 In addition to the model factors of CAPM and Carhart (1997), I add to both models control variables that measure the impact of institutional ownership concentration (IO). IO is defined as the percentage of ownership held by the top ten institutional holders and is measured as the percentage held at the start of each year. However, as research has shown institutional ownership to be
endogenously related to firm performance, I use a two-stage least squares estimation procedure to estimate the impact of IO. The regressions run are as follows, with 𝐼𝑂̂ as the estimation from the instrumental variables regression:
6To the extent that SMB, HML and WML factor loadings account for the residual return in equation (1),
the use of the Carhart (1997) four factor model in the multivariate analysis is admittedly not consistent with respect to the computation of within-month idiosyncratic volatility. However, this may still provide valuable insight when compared with the CAPM based regression run at this stage. Further research may improve upon the model by producing consistent daily factors.
Based on CAPM:
𝑅𝑖𝑡− 𝑅𝑓𝑡 = 𝛼𝑖𝑡 + 𝛽𝑀𝐾𝑇,𝑖𝑀𝐾𝑇𝑡 + 𝛽𝐼𝑉,𝑖𝐼𝑉𝑡−1 + 𝛽𝐼𝑂,𝑖 𝐼𝑂̂𝑡 + 𝜀𝑖𝑡 (3)
Based on a four-factor model:
𝑅𝑖𝑡− 𝑅𝑓𝑡 = 𝛼𝑖𝑡 + 𝛽𝑀𝐾𝑇,𝑖𝑀𝐾𝑇𝑡 + 𝛽𝑆𝑀𝐵,𝑖𝑆𝑀𝐵𝑡 + 𝛽𝐻𝑀𝐿,𝑖𝐻𝑀𝐿𝑡 (4) + 𝛽𝐼𝑉,𝑖𝐼𝑉𝑡−1 + 𝛽𝐼𝑂,𝑖 𝐼𝑂̂𝑡=6 + 𝛽𝐴𝐶,𝑖 + 𝜀𝑖𝑡
where, 𝑡 is measured in months for REIT 𝑖. The instruments used when estimating IO are the firms’s Debt-to-asset ratio Log Turnover and Price-to-Book ratio. The first stage regression is then run using the following model:
𝐼𝑂𝑛= 𝑃𝑇𝐵𝑛−1+ 𝐷𝐴𝑅𝑛−1+ 𝐿𝑜𝑔𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟𝑛−1 (5)
C. Descriptive statistics
Table 1 shows general statistics for the composition of the SNL Europe REIT index during the period studied. The number of REITs included in the index has doubled from 36 in June 2004 to 72 in June 2014. Also, the number of countries represented in the index has nearly tripled from 4 to 11 countries. Furthermore, during the time period studied the average market value of the studied REITs showed large fluctuation. This largely coincided with patterns
increased by almost 75 % from 1,05 billion US dollars to 1,76 billion US dollars to reach a low of 0,95 billion US dollars in 2009. This coincides with the large growth and subsequent bust in European capital markets with the outbreak of the debt crisis. Subsequently, REIT market value nearly doubled to reach an average of 1,85 billion US dollars at the start of 2014. Despite the large market fluctuation, for the average REIT, total institutional ownership percentage almost linearly increased from 30% to 38,5% from 2004 to 2014. In contrast, average institutional ownership concentration as measured by top 10 holdings first declined from 23,2% to a low of 19% in 2008 and then increased again to reach an average of 25%. A similar pattern is observed for the average firm Debt-to-Asset ratio which first decreased from 38,7% in 2004 to a low of 32,2% in 2006 and then sharply increased to reach a maximum of 44,4% in 2009 to remain stable at this level. The corresponding figures for the evolution of market value and institutional ownership percentage are contained in the appendix to this paper.
After excluding Turkish firms and firms with insufficient return data a total of 61 REITs are left for study. For the within-month regression in total, 6854 months of data were used covering all 61 firms. The range of trading days per month was 16 to 23 with an average trading day per month of 21.
Table 1: SNL Europe REIT index composition
Year As of June 1 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Number of Firms 36 37 39 41 41 55 60 67 71 69 72 Number of Countries Represented 4 5 6 8 8 8 9 9 9 8 11
IV. Results and Discussion
In this section the results of both the within-month estimation and multivariate model are presented. I provide full regressions of the models in the Appendix to this paper.
A. Idiosyncratic risk measurement
For the first part of the analysis, the appropriate model for CAPM estimation is first determined. The panel data of daily excess REIT returns is regressed on market excess returns using both a fixed and random effects model. By examining both models, as expected, the fixed effects model shows that the variation in errors is not due to firm specific effects with rho close to zero. Thus the variation in errors is largely idiosyncratic. Also with a large F-statistic, the model is significant with a P-value lower than 1% of alpha (see Regression 1). Also, the random effects model shows that the variation in errors is not due to firm specific effects. The Hausman test for a fixed versus random effects model performs a formal test and shows that the difference in coefficients is insignificant (see Regression 2). Thus a fixed effects model is supported. The average β is estimated to be 0,64 and thus shows lower than average exposure to swings in the broad market for the average REIT. Figure 2 presents the time path of idiosyncratic volatility followed by the average European REIT between 2004 and 20014. The average idiosyncratic volatility on a monthly basis fluctuated largely during the period and shows a significant increase leading up to the financial crisis. As can be seen from the figure, both total and idiosyncratic volatility measures tend to move opposite the of the sector’s performance, with a low in 2009, corresponding to the peak of both volatility measures Moreover, idiosyncratic volatility remained a large fraction of total variation during the entire period.
Figure 2: Volatility measures against European REIT sector return
I also find that idiosyncratic volatility is high and accounts for almost 80% or more on average of total REIT return volatility. This is similar to the findings of Liow and Addae-Dapaah (2010) who also observe a significant proportion of around 80%. However, contrary to their observation of decreasing idiosyncratic variance over time, no such trend seems to exist in this research. Other interesting patterns can be observed from Figure 3 however. Firstly, during the debt crisis years, it seems that idiosyncratic volatility ratio was somewhat lower than before 2008, with exception of the low reached in 2006. An ultimate low was reached in 2010. This suggests that during this period, systematic risk was more important. Secondly, idiosyncratic risk measured as a ratio to total volatility exceeds 1 in certain periods, implying that firm-specific risk
-.3 -.2 -.1 0 .1 .2 .3 Vo la ti lit y Me a su re s vs SN L D e ve lo p e d Eu ro p e I n d e x
Jan 2004 Jan 2006 Jan 2008 Jan 2010 Jan 2012 Jan 2014
Average Idiosyncratic Volatility Average Total Volatility
is more important in these periods than systematic risk. Furthermore, an unusual hike can be observed at the end of 2013, with a ratio exceeding 1.5.
Figure 3: Idiosyncratic volatility as a ratio of total volatility
B. Multivariate analysis
For the multivariate analysis, again a test is performed to determine the appropriate model for the multivariate estimation procedure. Similar to the within-month regressions, a fixed effects model is supported as appropriate for both the CAPM and Carhart models (see
Regression 4 and 6). The coefficients on all variables except institutional ownership
concentration, momentum factor and the constant are highly significant in the regression using the Carhart model. Thus, the results show contradiction with respect to the traditional asset
.8 1 1.2 1.4 1.6
Jan 2004 Jan 2006 Jan 2008 Jan 2010 Jan 2012 Jan 2014 Idiosyncratic Volatility Ratio
pricing model. The regressions were run using both Total institutional ownership percentage as well as institutional ownership concentration. Neither of these provdided significant coefficients nor affected the coefficient on idiosyncratic risk.The coefficient on idiosyncratic volatility is highly significant with a value of 0,17, implying that for a 1% increase in monthly idiosyncratic volatility, a premium in excess returns of 17 basis points is obtained on a monthly basis. The results are similar for both the CAPM and Carhart model, although the CAPM produces a higher beta of 0,89 versus a beta of 0,77 from the Carhart model. Thus, the additional factors
introduced in the Carhart model do not take away from the effect of idiosyncratic risk, rather the coefficient of systematic risk is reduced. With a market factor, β, of 0,77, the market volatility’s effect upon excess REIT returns is more than 4 times higher than the idiosyncratic effect. The size and value factors have a coefficient of 0,63 and 0,43 respectively. The average REIT studied thus resemble small size stocks and value stocks respectively. The empirical results from this study are economically significant and largely in line with the findings from Ooi, Wang and Webb (2009) who find a premium of 10 basis points for US REITs for the studied period
between 1990 and 2005. The higher premium for idiosyncratic risk found here could possibly be explained in the line of Ali et al. (2003) who firstly develop a link between the value factor and idiosyncratic risk and furthermore relate these premiums to low investor sophistication and higher transaction costs. Lower investor sophistication could be one explanation for the higher premium since the institutional ownership is lower than found in the US. However, this would contradict Malkiel and Xu (2003) who argue that higher idiosyncratic risk could be a result of higher institutional ownership. Higher transaction costs seems more of a plausible explanation in this respect since the European REIT market is characterized by cross country real estate
regulations between different local real estate markets. This adds to the opacity that characterizes real estate markets.
V. Conclusion
This research is aimed at studying the composition and temporal variation in European REIT volatility. I find that idiosyncratic risk is significantly related to REIT excess returns, contradicting traditional asset pricing theory. On average a 1 percentage point increase in idiosyncratic volatility is associated with an increase in excess return of 17 basis points per month. These results are in line with the research of Ooi, Wang and Webb (2009) who find excess returns of 10 basis points per 1 percentage point increase in idiosyncratic risk.
Furthermore, I find that the effect of idiosyncratic risk does not differ between the CAPM and Carhart (1997) models employed. The results also hold when controlling for institutional ownership concentration, or total institutional ownership percentage. Possibly, the higher premium for idiosyncratic risk found in this research compared to Ooi, Wang and Webb (2009), could be attributed to higher opacity in European REIT markets as a result of different real estate market regulation and possible language barriers. This could affect the pricing of risk since European REITs are characterized by cross-country real estate investments. Thus larger pricing of idiosyncratic volatility when compared to that found by US REITs could be expected. In this research, idiosyncratic risk was measured solely based on CAPM due to the limitation in factor data availability. This naturally, affects the implications of the results obtained. Nonetheless, the results in this research provide a justification for expansion by comparing against idiosyncratic volatility measurement based on Fama and French (1992) factors. However, the result from the multivariate regression showed that additional factors did not reduce the effect of idiosyncratic
risk. This is similar to the findings of Sun and Yung (2009). Finally, the observed increase in European REIT market integration merits a further expansion of the model. Such research is important, since the pricing of risk and its components remain an important part of many applications, such as portfolio diversification and the pricing of derivatives.
VI. Appendix
Regression 1: Fixed effects within-month regression
Note: In the regression of excess REIT returns, denoted FDEXReturn, on market returns, denoted as MDEXReturn, the results show that average beta amounted to 0,64. In addition, the model rho shows that residuals are largely idiosyncratic.
F test that all u_i=0: F(60, 139635) = 0.51 Prob > F = 0.9994 rho .00022169 (fraction of variance due to u_i)
sigma_e 2.0383876 sigma_u .03035356 _cons .0062986 .0054539 1.15 0.248 -.0043909 .0169882 MDEXReturn .6398818 .0035482 180.34 0.000 .6329274 .6468362 FDEXReturn Coef. Std. Err. t P>|t| [95% Conf. Interval] corr(u_i, Xb) = 0.0002 Prob > F = 0.0000 F(1,139635) = 32522.49 overall = 0.1889 max = 2493 between = 0.0598 avg = 2290.1 R-sq: within = 0.1889 Obs per group: min = 264 Group variable: Firmid Number of groups = 61 Fixed-effects (within) regression Number of obs = 139697
Regression 2: Hausman Test within-month regressions
Note: The Hausman test of fixed efects versus random effects model supports the use of a fixed effects model since the coefficients for the models do not significantly differ.
rho 0 (fraction of variance due to u_i)
sigma_e 2.0383876 sigma_u 0 _cons .0062986 .0054533 1.15 0.248 -.0043898 .0169869 MDEXReturn .6398853 .0035478 180.36 0.000 .6329317 .6468388 FDEXReturn Coef. Std. Err. z P>|z| [95% Conf. Interval] corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 Wald chi2(1) = 32530.07 overall = 0.1889 max = 2493 between = 0.0598 avg = 2290.1 R-sq: within = 0.1889 Obs per group: min = 264 Group variable: Firmid Number of groups = 61 Random-effects GLS regression Number of obs = 139697
Note: The market excess variable is denoted MktRF, top institutional ownership concentration is denoted by T10. Price-to-Book; Debt/Asset Ratio and Log Turnover are used as instruments in the first stage regression and are denoted PTB, DAR and logTurnover respectively.
Regression 4: Hausman Test multivariate model based on Carhart (1997)
Note: The Hausman test of fixed efects versus random effects model supports the use of a fixed effects model since the coefficients for the models do not significantly differ.
Regression 5: Fixed effects model based on CAPM
Note: The market excess variable is denoted MktRF, top institutional ownership concentration is denoted by T10. Price-to-Book; Debt/Asset Ratio and Log Turnover are used as instruments in the first stage regression, which are not tabulated here. The results are largely in line with the Carhart (1997) model.
Figure 6: Hausman Test multivariate model based on CAPM
Note: The Hausman test of fixed efects versus random effects model supports the use of a fixed effects model since the coefficients for the models do not significantly differ.
Mean market value 1000 1200 1400 1600 1800 Me a n Ma rke t Va lu e (U S $ Mi lli o n )
Mean Debt-Asset Ratio and Price-to-Book .4 .6 .8 1 1.2
Jan 2004 Jan 2006 Jan 2008 Jan 2010 Jan 2012 Jan 2014 Mean Debt-to-Asset Ratio Mean Price-to-Book Ratio
Institutional ownership development .2 .25 .3 .35 .4
Jan 2004 Jan 2006 Jan 2008 Jan 2010 Jan 2012 Jan 2014
modate
References
Ali, A., Hwang, L., & Trombley, M. A. (2003). Arbitrage risk and the book-to-market anomaly.
Journal of Financial Economics, 69(2), 355-373.
Bond, S. A., Karolyi, G. A., & Sanders, A. B. (2003). International real estate returns: A multifactor, multicountry approach. Real Estate Economics, 31(3), 481-500.
Carhart, M. M. (1997). On persistence in mutual fund performance. The Journal of Finance,
52(1), 57-82.
Dennis, P., & Strickland, D. (2004). The determinants of idiosyncratic volatility. Unpublished
Working Paper, University of Virginia,
Fama, E. F., & French, K. R. (1992). The cross-section of expected stock returns. Journal of
Finance, , 427-465.
Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and bonds.
Journal of Financial Economics, 33(1), 3-56.
Fama, E. F., & MacBeth, J. D. (1973). Risk, return, and equilibrium: Empirical tests. The Journal
of Political Economy, , 607-636.
Goyal, A., & Santa‐Clara, P. (2003). Idiosyncratic risk matters! The Journal of Finance, 58(3), 975-1008.
Hoesli, M., Lekander, J., & Witkiewicz, W. (2004). International evidence on real estate as a portfolio diversifier. Journal of Real Estate Research, 26(2), 161-206.
Hoesli, M., & Oikarinen, E. (2012). Are REITs real estate? evidence from international sector level data. Journal of International Money and Finance, 31(7), 1823-1850.
Karolyi, G. A., & Sanders, A. B. (1998). The variation of economic risk premiums in real estate returns. The Journal of Real Estate Finance and Economics, 17(3), 245-262.
Liow, K. H., & Addae-Dapaah, K. (2010). Idiosyncratic risk, market risk and correlation
dynamics in the US real estate investment trusts. Journal of Housing Economics, 19(3), 205-218.
Malkiel, B. G., & Xu, Y. (2002). Idiosyncratic risk and security returns. Unpublished Working
Paper, University of Texas at Dallas,
Markowitz, H. (1952). Portfolio selection*. The Journal of Finance, 7(1), 77-91.
Merton, R. C. (1973). An intertemporal capital asset pricing model. Econometrica: Journal of the
Econometric Society, , 867-887.
Merton, R. C. (1987). A simple model of capital market equilibrium with incomplete information. The Journal of Finance, 42(3), 483-510.
Niskanen, J., & Falkenbach, H. (2012). Liquidity of european real estate equities: REITs and REOCs. International Journal of Strategic Property Management, 16(2), 173-187. Ooi, J. T., Wang, J., & Webb, J. R. (2009). Idiosyncratic risk and REIT returns. The Journal of
Real Estate Finance and Economics, 38(4), 420-442.
Pagliari Jr, J. L., Scherer, K. A., & Monopoli, R. T. (2003). Public versus private real estate equities. The Journal of Portfolio Management, 29(5), 101-111.
Peterson, J. D., & Hsieh, C. (1997). Do common risk factors in the returns on stocks and bonds explain returns on REITs? Real Estate Economics, 25(2), 321-345.
Sharpe, W. F. (1964). Capital asset prices: A theory of market equilibrium under conditions of risk. The Journal of Finance, 19(3), 425-442.
Shleifer, A., & Vishny, R. W. (1997). The limits of arbitrage. The Journal of Finance, 52(1), 35-55.
Sun, Q., & Yung, K. (2009). Idiosyncratic risk and expected returns of equity real estate investment trusts. Journal of Real Estate Portfolio Management, 15(1), 45-57.
Williams, J. T. (1977). Capital asset prices with heterogeneous beliefs. Journal of Financial
Economics, 5(2), 219-239.
Yang, J., Kolari*, J. W., & Zhu, G. (2005). European public real estate market integration.
