The changing diversification benefits of direct real estate within Mixed-Asset investment portfolios.

(1)

The Changing Diversification Benefits of Direct Real Estate within

Mixed-Asset Investment Portfolios

Master Thesis Document

Business Economics: Real Estate Finance Faculty of Economics and Business

Sander Willems 5733693

Thesis supervisor 1: M.A.J. Theebe Thesis supervisor 2: P. van Gool Amsterdam, March 2014

(2)

ACKNOWLEDGEMENTS

I would like to take this opportunity to express my gratitude to my supervisor dr. M.A.J. Theebe for his contribution to the completion of this research and for his patient guidance as the process was laborious from time to time. Also I wish to state my appreciation for prof. dr. P. van Gool, the second reviewer of my thesis. Many thanks to all the lecturers from the Real Estate Finance department for organizing an academic program with an excelling balance between science and disciplines that I can use in practice.

Besides my lecturers at the University of Amsterdam, I want to thank my new colleagues at BNP Paribas Real Estate for their support and patience, as the completion of this work took some more time than expected. Also many thanks to my friends of SoWeBuild, for the ability to develop a promising business plan during my studies and work. I am looking forward to receive our first assignment in the upcoming months.

Thanks to my family and friends who saw me only sporadically in the last few months. A special word of thanks to my parents, who have given me the opportunity to study and encouraged me throughout this time. Finally thanks to my girlfriend, for her care and support during the days I only took very little sleep.

(3)

1. Introduction

1.1 Abstract

In the aftermath of several years of highly volatile capital markets it is even more interesting to analyze the influence of direct real estate as a portfolio diversifier. A commonly held perception on the investments in this asset class is that it will diversify the existing portfolio and therefore lower the overall portfolio risk. Although it was the general perception that an investment in physical properties would work as such, recent movements have shown that in times when the investor needs the diversification benefits the most; they appear to work the least. Why should the mixed-asset portfolio investor therefore continue to invest in the property sector? Previous research has revealed that the benefits of property investments lie in the fact that they enhance portfolio returns and diversify the overall risk of an existing portfolio. The same studies indicate, however, that these benefits are often subject to change over time. The continuous influx and efflux of capital and the significant fluctuations in the allocation in real estate over time underline the changing benefits of this asset class. This research is a quantitative empirical research that looks specifically at the theoretical diversification benefits of real estate over time. The benefits are measured by methods assumed by Lee (2010), Lee & Stevenson (2006), Hoesli et all. (2004) and Liang & McIntosh (1999). Comparable to their research the diversification and return benefits are decomposed from the overall benefits of an investment in real estate to an existing portfolio. With the performance of simple regressions, there is looked for a statistical significance of the change in the diversification benefit over time. The results show that investors operating on a short-time investment horizon are subject to continuously changing diversification benefits, thereby decreasing the envisioned benefits. As a result, this method exposes the investor to capital devastation in times of economic crises.

1.2 Research

A direct real estate investment is the acquisition of a physical property, which can span either the acquisition of the entire property or just part of it. The investor retrieves direct returns through the rent of the tenant(s) of the property, and indirect gains (or losses) from the sale of the property. Where real estate investors have dedicated themselves to this particular investment branch, mixed-asset capital market investors also allocate a significant part of their wealth in bricks and mortar. It is in the interest of investors to diversify their portfolio in order to reduce risk of specific asset classes. It is the primary reason that investors balance their investment portfolio by the determination of a broadly diversified investment strategy. A vigorous allocation of resources in an internationally mixed-asset portfolio might secure the investor from a disastrous ruination of their wealth during uncertain or volatile markets. The impact of an exogenous or endogenous disturbance on the assets

(5)

within a particular class generally differs from the impact of that same disturbance on the assets in any other given class. Fluctuations in returns in separate classes are to a greater or lesser extent related to each other. This is known as the correlation between different asset classes within a portfolio. The diversification potential is the result of the degree of correlation of an investment portfolio.

Especially since the beginning of the recent crisis, the impact of portfolio diversification on risk became a subject of interest to investors, as many investments previously thought to be safe became subject to unexpected portfolio risk. The impact of the crisis on the returns of a diversified portfolio proved to be vast. This disturbance is reflected in the volatility of the total returns of the assets within a portfolio around the mean return and is measured in variance and standard deviation. The ability to counterbalance extreme movements in the returns for an asset class is reflected in the correlation of this asset class with other asset classes in a portfolio. The different assets in a diversified portfolio are related to macro-economic developments over time. Due to this common market dependence, a common strategy is to diversify risk so that the assets within a portfolio are correlated to a lesser extent so risk on returns is diminished. However, there are diversification opportunities that will lower the overall risk of a portfolio because the impact of a market change will affect some assets to a lesser degree than others.

The ability to counterbalance negative returns of highly volatile assets in times of distressed markets is in the interest of the investor. Direct real estate within a traditional investment portfolio is considered as an asset class that bears the ability to further diversify the investment portfolio. However, in times of a distressed market this ability to counterbalance might be significantly lower than is generally presumed in the periods before major financial crises. The recent global financial crisis again spawned the perception that diversification is not working anymore. However, no significant quantitative research has been conducted to study the diversification benefits of direct real estate in a mixed-asset portfolio in the context of global financial crises. This research therefore aims to answer the question whether the benefits of incorporating direct real estate in a mixed-asset portfolio are applicable or not in volatile markets.

The focus of this thesis lies in 1) the diversification opportunities of direct real estate in an efficient mixed-asset portfolio and 2) the diversification opportunities of direct real estate in a typical market portfolio. It is in the interest of this research whether the high volatility of the financial markets had a significant impact on the diversification potential of direct real estate. The use of correlations as the foundation for the process of strategic portfolio allocation conceivably needs to be exercised with

(6)

care, as they are a statistical measure that can vary in different circumstances over time. In addition, statistical measures are subject to estimation errors. Although the investment target for average investors has a long-term nature, on the medium-term these investors are averse to large losses as well. This analysis focuses on this medium-term tendency of correlations between asset classes within a traditional portfolio that has incorporated real estate. What is the tendency of the correlation between the different asset classes and therefore the diversification potential in times of economically distressed markets? Will investors still focus on the diversification potential of direct real estate if this asset class appears to be worthless as a portfolio diversifier when the investor most needed it? The main question will be:

What are the changing direct real estate diversification opportunities within mixed-asset investment portfolios in the period between 1991 and 2012?

To be able to explain the main question for this thesis, the following sub questions are formulated:

How can we measure the change of diversification potential through time?

What is the long-term diversification potential of direct real estate in an efficient mixed-asset portfolio?

What is the medium-term diversification potential of direct real estate in an efficient mixed-asset portfolio?

What is the long-term diversification potential of direct real estate in a typical investment portfolio? What is the medium-term diversification potential of direct real estate in a typical investment portfolio?

The diversification benefits are measured in two distinctive methods, which are performed at the same data set. In the first part of this research the theoretical diversification potential of the incorporation of direct real estate as an asset class is compared to a base portfolio excluding direct real estate. This part of the research is performed by making use of efficient portfolios without any constrains to efficient allocations. The reduction in risk is measured by the comparison of the risk in the latter portfolio opposed to the risk of the former portfolio. In the second part of the research the allocations are based on typical market portfolios as described in Lee & Stevenson (2006). The allocations are regarded as constant for the entire research period an all the considered countries. Also in this approach, the risk reducing ability of direct real estate is seen as the reduced total risk in the portfolio with direct real estate compared to the portfolio without direct real estate as an asset class. For long-term investment periods the research is based on the total research period from 1991

(7)

to 2012. The medium-term portfolio returns are based on the average annual total portfolio return data for 5-year rolling time periods. Interesting is the effect of volatile financial markets on the diversification potential of a diversified portfolio of stocks, bonds, indirect real estate and direct real estate, which is reflected in the difference of the standard deviation of the base portfolios with the standard deviation of the portfolio where direct real estate is included. The effects on the diversification potential emerge from differences in the potential. Whether the potential changes statistically significantly over time is measured through simple regressions. As diversification benefits among assets with low long-term correlations appear to weaken in shorter time periods, the simple regressions are performed in distinguishable periods in economic development from 1991 to 2012. Because of the limited availability of total return data for direct real estate, it was not possible to perform the analysis further back in history or perform the analysis on shorter time periods. Based on relevant literature, the diversification potential is expected to decline in times of crisis.

Hypothesis

Although the diversification potential of real estate declines in times of economic crises, an investor will still be able to reduce risk by the inclusion of direct real estate in a mixed-asset portfolio.

In the next chapter an overview of the diversification issues is provided, laying emphasis on direct real estate as a portfolio diversifier. The next chapter describes the data used for this research and provides the explanation and justification for the research that was explored on these data sets. The results are given in the subsequent chapter, in which the differences in the diversification potential will be shown. These results will finally be reflected and discussed and a conclusion of this thesis will be drawn.

(8)

2. Literature review

2.1 Diversification benefits of direct real estate in an investment portfolio

The correlation structure as a basis for diversification benefits of investment portfolios is subject to extensive research. The diversification potential of investment portfolios results from the variation in the investment opportunity set and the correlation structure between the returns of different asset classes (Goetzmann et all., 2001). Correlation is the statistical measurement of the tendency of the return movements of one asset class to the movement of the returns of another asset class. It is the linear relationship of the returns of these asset classes over time. Correlations are based on the performances of the returns of the assets in the past. Uncorrelated assets are expected to show no systematic or a low linear relationship, but past performance is no guarantee for the results in the future (Philips et all., 2012). Bearing in mind the fundamental assumptions of modern finance, diversification improves the mean variance efficiency and therefore it improves the risk and return tradeoff for investors (Conover et all., 2002; Asness et all., 2010). The potential gains from portfolio diversification arise according to Markowitz (1952) from investment returns in different assets that are not perfectly correlated. With his theory of mean-variance analysis, correlation has been used as a basis for investment portfolio construction since the 1950s. The trade-off between the risk and expected return of an investment is the basic presumption for this kind of analysis, where the combination of risky assets in a portfolio is considered to have a lower risk for the concerned desired level of expected return. The combination of uncorrelated assets can partially mitigate the performance of one asset to a second asset through the reduction of the average volatility of a portfolio, which is especially in times of uncertain and distressed markets an advantageous feature (Philips et all., 2012).

The modern portfolio theory, predominantly based on the normal probability distribution theory of Bachelier (1900), still has enormous implications for the practice of finance. The capital allocation of investors has a crucial impact on the performance of the portfolio, which in turn impacts the amounts that has to be put into a fund or the amount that could be distributed to beneficiaries (Hoesli et all., 2004). The correlation between the assets is one of the important components for the construction of portfolios, next to the expected return and expected volatility characteristics. The combination of imperfectly correlated assets is based on the historical relationship between the assets or asset classes. This historical relationship is used to conduct expectations for the performance and correlations of these assets in the future. The main purpose with regard to the relationship between assets or asset classes, is to strategically construct a portfolio with imperfectly

(9)

correlated assets in such a way that the volatility for an established expected portfolio return is significantly reduced (Philips et all., 2012).

The diversification benefit is sustained as long as the correlation structure remains stable over time (Alexander, 2001). In reality the potential diversification benefits appear to be subject to permanent modifications. There is extensive academic and financial research that suggests that correlations change dramatically through time, which implies that the benefit of a mixed-asset investment portfolio is not constant (Philips et all., 2012; Goetzmann et all., 2001; Longin & Solnik, 1995). There are several reasons for the changes in the correlation between assets. Randomness in the return variables is a first explanation for the difference between the produced and the expected outcomes. The differences in the produced outcomes from the long-term basic statistics are especially examined over shorter time periods. These differences do not explicitly signify that the observed ‘normal’ correlation changes. However, since the recent crisis the statistical measurement is subject of discussion while even over longer periods of time these correlations are observed to change. Mandelbrot et all. (2004) developed an alternative theory on the behavior of markets, in which the basics for modern portfolio theory is put into question. In its intrinsic research is stated that the normal probability distribution as a basic component for portfolio theory as described in Markowitz (1952) could be too simplified. The interesting theory of fractal geometry as a new comprehensive mathematical statistic for the return on assets is subject for further research. In this thesis the conventional statistical measurements are not questioned.

Next to the issue of randomness of asset returns, several researches observed a relationship between macro-economic changes and increasing or decreasing correlations in asset returns over time. In this issue research is focused on the correlation between asset classes as well the (international) correlation within subcomponents of the same asset class (Philips et all., 2012). Increasing correlations between stocks and bonds appear in times of high inflation and significant macro-economic changes. Evidence suggests that this is especially the case with an internationally diversified portfolio of the traditional asset classes stocks and bonds. When the equity markets suffers from high volatility, the correlation between stocks and bonds generally is lowest in the quest for secure investments (Ilmanen, 2003; Conover et all. 2002). Longin and Solnik (2001) found evidence of positive correlation shifts when financial markets experience negative shocks. International market correlations within subcomponents of the same asset class tend to go up when global markets experience negative shocks, which implies that diversification is weak when it is most needed (Asness et all., 2010). Equity markets in particular experience this contribution of the increasing interdependence among countries to an increasing correlation between international assets within a particular asset class. The primary reasons for this increasing interdependence, are

(10)

the increasing (technological) efficiency of markets and the ability to easily invest abroad (Solnik, 2002).

While correlations are suffering from significant volatility, an originally adequately conceived diversified portfolio can lose its diversification potential and efficiency due to significant modifications in the correlation between asset classes. The probability that the correlation between assets is different is relatively high when the investor uses a short window of observation. The increased systematic risk that is experienced in the recent crisis is a natural result of the investors’ search for safe investments. The relatively risky assets tend to correlate with the highly volatile asset class of equities, while asset specific risk is overwhelmed by the systematic risk resulting from changes in market sentiment. While there are minor differences between the risks of bonds and real estate in typical markets, the differences can be significant during times of a highly volatile market. The specific advantages of the riskless allocation in bonds become clearly evident in times when markets are outbalanced. This is expressed in the difference of risk diversification through the allocation in bonds versus return diversification by investing in risky assets, such as real estate (Philips et all., 2012). In the composition of a mixed-asset portfolio, it is important to use relevant historical information while performing this strategy. In practice this means that the investor also has to refer to the correlation figures in a historical time period that equals the intended investment period in the future.

The ability to diversify by investing in assets from different classes affects the mean variance optimization. Especially the investor’s interest in listed real estate increased substantially in this matter. Although the development of the underlying real estate or real estate-related companies should be reflected in the performance of the returns of the public real estate assets, there are considerable variations in the performance of indirect and direct real estate markets. The trading mechanism for indirect real estate is vastly different due to the low-cost, trading-market dimension that is not persistent in direct real estate investments (Brounen et all., 2007). The exchange-traded market price for indirect real estate investments does therefore not necessarily present the market value of the underlying assets. These differences between the two asset classes justify the distribution of recourses among these real estate-related asset classes and will offer some diversification benefits to the mixed-asset portfolio investor (Georgiev et all., 2003).

The advantages offered by direct real estate investments are diverse and thoroughly discussed in real estate literature. Direct real estate is a tangible asset with low volatility that generates an attractive income stream and capital appreciation on the long-term. Most relevant for this research is that direct real estate offers strong diversification benefits to stocks and bonds (Lee, 2010), without a

(11)

ruination of the expected return (Brounen et all., 2007). In general, long-term real estate returns are lowly correlated to the returns of the traditional assets stocks and bonds. Traditionally investors used bonds to diversify the portfolio allocation to equity because of the low correlation of the returns between these asset classes. The disadvantage of the use of bonds is a detriment of the returns of a portfolio. With the improved and low-cost access to other higher-risk portfolio diversifiers as real estate, the advantages of a portfolio diversifier are combined with higher returns than the traditional stock-bond allocated portfolio (Philips et all., 2012). A significant fundamental character of direct real estate is that the asset-class as a diversifier generates returns when benefits are most needed (Chun et all., 2004). However, there are more diversification benefits for real estate. Firstly, real estate returns appear to be predictable. The predictability of stock returns is comparable to the predictability of direct real estate. Furthermore real estate is performing well in an asset-liability framework and concerns that real estate often produces high losses is ungrounded. The proportion of real estate in a mixed-asset portfolio is considered to be between 6 and 12 percent, which enables the investor to mitigate diversifiable or unique risk (Chun et all., 2004).

The weaknesses of direct real estate investments are the illiquidity of the assets, the lack of transparency compared to other asset classes, the need of expert management, and the requirement of a significant amount of capital to develop and structure a diversified portfolio of real estate assets. Only institutional investors with hundreds of millions of euros to invest will be able to build a diversified real estate portfolio and thereby eliminate idiosyncratic risk. The acquisition and transfer of interest in direct real estate means in practice the purchase and sale of actual properties, mostly carried out by fund managers or by third party advisors as brokerage firms. These parties can ask considerable search costs for matching sellers with potential buyers. Especially in markets with low investment activity in direct real estate or no active secondary market facilitations, these costs are substantial (Fuerst & Matysiak, 2013). In addition to transaction cost, a considerable amount is funded to the management of the property (Georgiev et all., 2003). For the analysis of return distributions the investor depends for a large extent on appraisal-based series, causing a high degree of autocorrelation as a result of the appraisal based construction and smoothing of direct market indexes (Georgiev et all., 2003). The lack of transparency in real estate is caused by the low compliance of real estate investors of the guidelines for direct real estate information distribution. The existence of potential asymmetric information is characteristic for this asset class, which is a source of relatively high risk adjusted returns for investors who are able to collect reliable information by alternative sources (Georgiev et all., 2003). However, the commitment to distribute more fund information is increasing over the last years, which underlines a trend towards higher transparency (Fuerst & Matysiak, 2013).

(12)

The opportunity to invest abroad has expanded the ability to diversify an investment portfolio dramatically (Driessen & Laeven, 2007). Under the condition that the returns from investments in different national markets are not perfectly correlated, there is a potential for international portfolio diversification (Markowitz, 1952). The diversification potential of direct real estate is relatively high compared to the potential for the traditional asset classes - stocks and bonds (Eichholtz, 1996). Real estate returns are to a large extent dependent and influenced by local market drivers as growth rates and country specific risks. This results in low or even negative correlations between real estate returns of different countries, and therefore provide diversification benefits that are not provided by the international diversification in stocks, bonds and indirect real estate markets (Hastings & Nordby, 2007).

The explanation for the failure of portfolio diversification in improving the mean variance portfolio efficiency is often found in the integration of the global financial markets. The continuing globalization driven by technological evolutions caused two conflicting developments. The ability to invest in an exceptional set of global assets, vastly improved the ability to maximize the mean variance efficiency of a diversified portfolio. Simultaneously, this resulted in increasingly connected markets that react equally on positive and negative disturbances in the market. The international diversification potential is currently very low compared to the earlier potential in capital market history (Goetzmann et all., 2001).

2.2 De-smoothing and lagging of direct real estate returns

The development of total real estate returns is measured through property indices. Property indices inform investors about the performance of property investments. Information on real estate is in contrast to information about other asset markets difficult to obtain and relatively expensive. This is a result of property specific characteristics and it is usually acknowledged that using real estate data is not straightforward (Hoesli et all., 2004). Property specific risk could be diversified by the inclusion of more property returns that will result in index risk that will be closer to the market risk. The most important features that cause problems for the construction of a property market index is that there is no central trading market for real estate, so it is difficult for property investors to obtain reliable information. The second feature is the infrequent trade of properties. These features make it impossible to regularly compare a sample of properties.

Indexes are typically based on appraised values of properties and are therefore highly dependent on the expertise and the integrity of appraisers. The appraised market value of a property remains an approximation of the real value and is highly dependent on the particular moment of the appraisal

(13)

and the availability of data. With the use of direct real estate data the issue that is called the smoothing bias comes in front. The data suffer, due to the fact that it is appraisal based, on the appraisal-smoothing problem. There is a significant serial correlation of the returns as the volatility of appraised properties is reduced when the value is approximated in a valuation of an appraiser (Pagliari et all., 2005; Geltner et all., 2003; Geltner & Miller, 2001; Brown & Matysiak, 2000). The appraisal problem is exposed to two different problems in the process: random error and a backward looking nature of the appraisal process (Pagliari et all., 2005). The random error term is especially applicable to the valuation at the single property level while giving appraisals is a subjective process. To improve the volatility of random error of single asset appraisals, there is the possibility to expand the number of properties used in the dataset for the research. The problem of random error is canceled out in the realized returns of the Index, because the index is constructed out of the information of a range of appraisers and given a sample that is large enough it will have not affect the overall value (Booth & Marcato, 2003). While in this research index based data series are used, whereby many properties are included, the random error is not applicable for the data. The problem within the appraisal process, which arises from the backward looking nature of the appraisal process, is known as the temporal lag bias. This is attributed to the contemporary use of the weighting averages of real estate information with historical appraisals. Among others, Diaz & Wolverton (1998) and Clayton et all. (2001) found evidence that appraisers judge the value of a particular property insufficiently from previous appraisals or anchor onto the previous appraisals for the valuation of the same property. Within the appraised value there is a specific adjustment factor known as the confidence factor (Pagliari et all., 2005). Periodic, appraisal-based appreciation returns could reflect a value change in the asset that is partial-adjusted to the real value change of the asset. This is called smoothing in the appraisal process and ads first-order autocorrelation (Pagliari et all., 2005). There are possibilities to partially cancel these problems out. As for this research appraisal-based data series are used, the research is appraisal-based on unsmoothed data in order to lose the significant level of serial correlation.

The temporal lag bias is especially in the appraisal of properties with only a few comparable property sales in the market, a rational process of ‘a declining moving average of previous appraisals’ (Pagliari et all., 2005). There several widely used techniques to estimate the unobservable true market value out of an appraisal. For this research the technique of Geltner (1993) is applied, in which true returns are computed by using a reverse filter. De-smoothed direct real estate return data are more comparable with the market based assets (Lee & Stevenson, 2006). The de-smoothed return rates for direct real estate investments are computed as follows:

(14)

In this formula the κt* represents the true return in year t, which is undoubtedly unobservable out

of an appraisal. The appraisal-based return for year t is given by κt and α is called the smoothing

parameter. For this research the mentioned unsmoothing process is performed for all the direct real estate total return data. In this approach the real estate market is not assumed to be efficient and therefore corrects not for all the serial correlation. Instead, the common assumption that has been made relates to the standard deviation of real estate by making use of an arbitrarily estimated smoothing parameter, which for this research is fixed at 40%. While previous research has shown that the smoothing parameter is time-dependent (Booth & Marcato, 2003), for this research the α is considered as fixed for the entire research period.









*



t



(

1 

)

t1 t

(15)

3. Data & Methodology

This chapter is divided into two main sections. The first section is committed to the data and also contains the justification for the data sources as a summary for the descriptive statistics of all countries included in this research. The descriptive statistics are detached for the total research period from 1991-2012 as well as the rolling 5-year periods. The aim of this part is to describe intuitively the mean-variance performance of direct real estate to the performance of the assets in a traditional investment portfolio. In the second section the performed methodology is further explained and divided into two distinguishable parts. In the first methodology part the base statistics that are used to perform portfolio optimizations are explained. In the second part the focus lies on the application of these base statistics in order to derive the diversification potential of direct real estate in both methods used for this research.

3.1 Data

3.1.1 Data Sources

In this research the annual total return indices of stocks, bonds, indirect real estate, direct real estate and cash have been used. Total returns consist of both the capital return as the income return component based on local currencies for the U.S., Canada, France, the Netherlands and the U.K. The study covers a total research period of 23 years from 1990 to 2012. However, because of the de-smoothing optimization process, the first observation is lost. Research is conducted for both the entire research period, compatible to the investor with a long-time investment horizon, as for 5-year rolling window periods, the period, which can be classified as a medium-term investment. With the exception of the direct real estate returns, the data is extracted from the Datastream database (appendix A). The yearly total return data from the Morgan Stanley Capital International (MSCI) are used as the benchmark index for stocks in all the markets. The indices cover at least 60 percent of the market capitalization according to MSCI. For bonds the World Government Bond Index from the Citigroup Global Fixed Income indices is used for the total return data of 5-7 year bonds. The index is a measurement for fixed-rate investment grade sovereign bonds. The FTSE EPRA/NAREIT Global Real Estate Index series is an index that covers trends in real estate equities. For all countries the total return index is used for indirect real estate. For the cash data the three-month T-bills are taken.

The total return data from the Investment Property Databank (IPD) and the National Council of Real Estate investment Fiduciaries (NCREIF) are used as the benchmark for direct real estate. Typically, total return figures in the IPD index are published annually and represent the investments in real estate of the particular country. The property indices exclude the impact of debt, fund management

(16)

fees, taxation and cash on property portfolios in the calculations. The NCREIF represents the investments in U.S. and Canadian real estate. The valuations are at least updated once a year. The cash flows and the estimated changes in value of the aggregated U.S. Funds Portfolio and the aggregated Canadian Funds Portfolio are included in the index. The real estate indices from as well the IPD as NCREIF reflect the investment policy of institutional investors in each of the included countries. The total return data are therefore based on a different set of property types and reflect the investment universe in each of the countries as is presented in table 1.

Table 1: Description of direct real estate indices

Source: IPD, 2013

3.1.2 Summary statistics yearly data

The various statistics for each asset class for the total research period are summarized in for every country included in this research. The unsmoothed direct real estate data are stated in order to make the return data comparable with the return data for the financial asset classes. With regard to the long-term investment period, the examined data of the assets within a traditional portfolio are congruent with the expected outcomes. The return and risk parameters for stocks are the highest for all countries, while cash exhibits the lowest figures on as well risk and return. Bonds have a higher return and risk than cash, but these figures are lower than those for stocks in all countries. The results are compatible to mean-variance efficiency. In general, return figures are efficient in the situation where the asset classes with a higher expected return are subject to a higher risk.

In both real estate asset classes, the results for the mean-variance statistics show a high variety. The performance of these assets varies strongly from one country to another. Differences in indirect real estate can be explained through the institutional difference across countries in the degree of leverage, the investment in property types and the tax status of indirect real estate funds (Hoesli et all., 2004). From a mean-variance perspective the indirect real estate results are advantageous in some countries, while investors in other countries cannot benefit from investing in this asset class. In the U.S., France and the Netherlands the figures are in line with the mean-variance efficiency. In the U.S. and France both the return and risk figures are higher than those for stocks. In the Netherlands

Country Source Type Properties in index

in 2012 Retail Office Industrial Residential Other

U.S. NCREIF Appraisal-Based nk 23.0% 35.0% 14.0% 25.0% 3.0%

CAN NCREIF Appraisal-Based nk nk

FR IPD Appraisal-Based 6,190 26.9% 51.2% 7.0% 12.4% 2.5%

NL ROZ/IPD

Appraisal-Based/ Repeated

measures regression 4,521 31.6% 16.6% 3.5% 45.7% 2.6%

U.K. IPD Appraisal-Based 21,145 48.2% 26.2% 15.2% 4.4% 6.0%

(17)

as well the return and risk lies between those for stocks and bonds. However, it be pointed out that the risk figure for indirect real estate in the Netherlands is almost as high as for stocks, while the return is only marginally higher than the return for bonds. For Canada and the U.K. the performance of indirect real estate is poor, as the return lies between that for stocks and bonds, while the risk is much higher than that for stocks.

When the mean-variance statistics for direct real estate are considered, the results also vary strongly from one country to another. Only for Canada and the Netherlands the mean-variance efficiency holds while as well the return and risk for direct real estate returns are in between the mean and standard deviation figures for stocks and bonds. In the U.S. and the U.K. the return of direct real estate lies between the return for stocks and bonds, but the risk exceeds the risk for stocks. For France this is even worse, as the returns do not exceed the returns on bonds, while the standard deviation lies between the standard deviation for stocks and bonds. If the mean-variance parameters appear to be not beneficial, the benefits for holding real estate in a mixed-asset portfolio will derive from low correlations between the returns for direct real estate and the other financial assets (Hoesli et all., 2004). In table 2 the summary statistics for the yearly data from 1991 – 2012 for the Netherlands are presented as an example. The statistics for all other countries are collected in a comparable manner.

Table 2 : Summary statistics yearly data 1991-2012 (random example)

3.1.3 Summary statistics rolling 5-year windows

The statistics for rolling windows show higher variations in the mean-variance performance among the different asset classes in the analyzed periods (table 3). There are only a few rolling periods among all countries in which the mean-variance statistics are efficient. The efficient combination is affected if the return of an asset class compared to another asset class is lower than the risk of these compared asset classes. In the U.S. the existence of a rolling window where all asset classes will be included according to the return and risk performance is inexistent. With respect to direct real estate, there appear to be several consecutive periods through time in which the asset class has advantageous statistics. From periods 1992-1996 to 1994-1998 and from 1996-2000 to 2004-2008 the risk and return figures for direct real estate are beneficial opposed to the statistics for financial assets. On average, the return and risk profile for direct real estate appears to be between the return

Stocks Bonds Indirect RE Direct RE Cash Panel IV: The Netherlands

Mean 11,83% 7,42% 8,46% 7,90% 3,75%

Std.Dev 23,22% 5,60% 21,24% 8,49% 2,36%

Max. 45,91% 18,40% 41,06% 23,01% 9,40%

(18)

and risk statistics for stocks and bonds, but for five terms the risk of direct real estate is even lower than the risk on bonds, while the returns are higher than those for bonds. In these cases the generally perceived advantage of including direct real estate within a mixed-asset portfolio is confirmed, as real estate is used as a risk diversifier for riskier financial assets without losing a considerable extent of the returns. This is the case in traditional mixed-asset portfolios where only bonds are used as a portfolio diversifier. In two rolling periods an investment in direct real estate does not provide the desired result to the U.S. investor. In the consecutive periods between 2005-2009 to 2008-2012 the benefit is canceled out completely as direct real estate returns are low or even negative while the standard deviation exceeds the risk parameter for stock returns.

Canada contains two rolling windows in which the mean-variance efficiency for all the assets included in this research is fully applicable. These are the periods 2005-2009 and 2006-2010. In these periods the assets that produce a higher return are subject to higher risk. Concerning the risk and return parameters for direct real estate, the figures are more beneficial than the figures for the U.S. In fourteen out of eighteen rolling windows, the mean-variance performance of direct real estate is efficient. Direct real estate could therefore play a considerable role in efficient portfolio compositions. The performance of direct real estate is significantly better than the performance of this asset class in other countries included in this research. In addition, for nine of the rolling periods, direct real generates higher returns than the asset class of stocks while the risk for this asset is lower than for stocks. In two of the rolling periods an investment in direct real estate is not beneficial for the Canadian investor, which are 1991-1995 and 1992-1996. Solely based on these parameters the allocation to direct real estate should have been as small as possible.

For France the risk and return statistics for direct real estate are appealing. The asset class could play a considerable role in portfolio diversification. For four rolling periods the return data for the entire asset classes are efficient with regard to risk and return: 1996-2000, 1997-2001, 2003-2007 and 2005-2009. However, in five out of eighteen rolling periods a French investor profits only little from the inclusion of the considered asset class, as returns are low or negative or the risk of direct real estate surpasses that for other asset classes with higher returns. The mean-variance profile of direct real estate changes over time in France. The asset class contains very low risk and generates lower returns than bonds in the periods up to period 1994-1998, whereby real estate could be considered as a very safe investment during these years. In the subsequent periods, direct real estate becomes more risky, but inherent to more risk however, is a higher return. Especially opposed to stocks, direct real estate performs better in six periods based on these statistics. For French investors that are less risk averse, allocations to direct real estate count as a substitute for weight in stocks.

(19)

In the Netherlands, two rolling periods are considered as efficient for all asset classes concerning their mean-variance performance: 2003-2007 and 2005-2009. In six periods the inclusion of direct real estate is not at all beneficial to the Dutch investor, as returns are lower than the returns for bonds while the risk lies between that for bonds and stocks. Nevertheless, investments in direct real estate are beneficial in the remaining 12 periods. In five rolling periods the mean-variance profile lies between that for stocks and bonds. In 1995-1999 and 1996-2000 direct real estate investments proved to be extremely safe as the returns are between that for stocks and bonds, while the risk is lower than that for bonds. In the consecutive periods from 1998-2002 to 2002-2006 and in rolling period 2004-2008 the risk lies between that for stocks and bonds, while the returns of direct real estate outperform those for stocks. Direct real estate is something less a safe investment in the Netherlands as in France and could therefore be an interesting investment for investors that are less risk averse.

The least efficient combinations are found in the U.K. In ten rolling periods direct real estate is not beneficial according to their mean-variance statistics. Only in the consecutive periods from 1996-2000 to 2002-2006 investment in direct real estate perform well compared to the financial assets. Direct real estate could be considered as a highly risky asset class in the U.K. ,as the risk is higher than for stocks in twelve periods and the risk is between that for stocks and bonds in six periods. Noticeable is that as well indirect as direct real estate performs very poorly from 2004-2008 to the last rolling period 2008-2012.

The mean-variance performance of the asset classes for both the total research period and the rolling periods among the countries that are subject of this research, particularly affect the allocations, and therefore the diversification potential, of direct real estate in the first approach in which efficient portfolios are considered. As allocations in the second approach are held constant to the typical market allocation, the positive or negative impact of the inclusion of direct real estate to a base portfolio without this asset class exclusively represents the potential to diversify the portfolio as a result of the interrelationship between the asset classes.

(20)

Table 3 : Summary statistics yearly data rolling 5-year windows (random example)

3.2 Methodology

3.2.1 Basic statistics

The research on the diversification opportunities of direct real estate is based on the total return of the assets within the portfolio and therefore total returns are the fundamental basis to research the diversification potential of direct real estate investments in a mixed-asset portfolio and the diversification potential within a portfolio of typical market asset allocations. The total return out of investments in the asset classes is divisible over a direct return and an indirect component. In contrast to the relatively constant cash flows that are known as direct returns, the indirect returns arrive from price changes in the underlying asset and are generated when these assets are sold. The direct component of the stock returns are dividends, for bonds this is the coupon payment and for real estate this is net rent. Indirect returns are known as capital gain. The realized returns are computed by taking the sum of (direct) income and (indirect) capital gain over a particular time period. Investors in a mixed asset portfolio regard the total return as the indivisible sum between both components. The average return of an asset class ( ̅) the average of the realized returns ( ) of a selected period (appendix B, 1-4). The total return series are, as described in previous chapter, extracted from the databases that are used for this research.

Mean Std.Dev Max. Min. Mean Std.Dev Max. Min.

Rolling periods U.S.: 1991-1995 Rolling periods Canada: 1995-1999

Stocks 17,79% 15,95% 38,19% 2,00% Stocks 22,22% 16,71% 45,88% 1,22% Bonds 10,09% 9,02% 18,94% -3,90% Bonds 9,72% 7,49% 19,98% -0,78% Indirect RE 26,95% 18,66% 58,44% 8,40% Indirect RE 9,11% 30,26% 60,56% -12,67% Direct RE 2,66% 12,74% 13,90% -17,42% Direct RE 14,04% 12,66% 35,73% 2,55%

Cash 4,33% 1,16% 5,53% 2,99% Cash 4,81% 1,40% 7,05% 3,18%

Mean Std.Dev Max. Min. Mean Std.Dev Max. Min.

Rolling periods France: 1999-2003 Rolling periods the Netherlands: 2003-2007

Stocks 4,20% 32,83% 51,93% -32,83% Stocks 14,56% 11,00% 32,56% 5,95%

Bonds 4,88% 4,63% 10,29% -2,34% Bonds 3,74% 2,81% 7,60% -0,21%

Indirect RE 12,16% 7,00% 19,24% 2,37% Indirect RE 21,35% 21,48% 41,06% -12,27% Direct RE 11,58% 9,00% 25,51% 2,93% Direct RE 10,54% 4,48% 15,98% 4,69%

Cash 3,37% 0,88% 4,32% 2,30% Cash 2,80% 0,93% 4,30% 2,10%

Mean Std.Dev Max. Min. Rolling periods U.K.: 2007-2011

Stocks 3,24% 20,72% 27,66% -28,46%

Bonds 8,38% 4,57% 14,29% 2,00%

Indirect RE -14,09% 26,27% 14,95% -46,12% Direct RE -2,90% 40,52% 41,92% -50,12%

(21)

The portfolio return is the weighted average of the total returns of the assets within the portfolio. The portfolio weight wi is the proportion of the value of asset i to the total value of the portfolio.

Combining portfolio weight and the returns for the individual assets for particular time series gives the return for the portfolio (Rp) (appendix B, 5):

The variability or risk of the total asset class returns are measured by the standard deviation ( ) or volatility. The standard deviation is computed by taking the square root of the variance ( ) of the total return data. Variance and standard deviation are common measures for risk in modern finance. The variability of the returns of different asset classes reveals large differences that are expressed through the spread in the average return ( ̅). From the variability of these returns it is possible to compute the standard deviation, which quantifies the different distribution of the realized returns within an asset class or the portfolio. In the measurement of this variability, the average realized return is the best estimate for the mean and replaces therefore the mean in the variance computation (appendix B, 6-7):

√ ∑( ̅)

Portfolio risk requires some more explanation, as also the co-movement of the return data is an extensive component. The co-movement of the returns is measured by covariance (Cov) and correlation ( ). The covariance between the return of asset i (Ri) and asset j (Rj) is computed by the

multiplication of the excess returns ( ̅ ) for every period in the dataset. The result of this multiplication is the product of these differences, which are summarized and divided by the number of years (T) included in the calculation:

( ) ∑( ̅)( ̅ )

A high covariance is an indication for a substantial increase in the return of one asset if the other asset increases significantly. This implies a high degree of dependency from one asset to another. If the covariance appears to be low the assets are independent and therefore react barely in the same direction or react in the opposite direction. Covariance is the average value of the product of deviations across different assets, but provides no clear insight into the degree of interdependence or the direction as covariance measures variables have different units of measurement. These components are determined by correlation (Chen et. all., 2002).

(22)

The fraction of the volatility due to risk that is common to the securities is called the correlation (appendix B, 8). It is a measurement of how the returns of two assets move in relation to each other (Philips et all., 2012). It is able to analyze as well the strength of the correlation and the linear direction between the variables, as correlation standardizes the measure of interdependence between two variables. The correlation between asset classes is besides the mean-variance performance of assets, the second important parameter for the risk reducing potential of one asset class to another and determinative for the diversification potential of direct real estate in an existing portfolio. The correlation can be calculated by dividing the covariance of the returns by the standard deviation of each return (appendix B, 9). Correlations that are increasingly positive are indicating that the total returns have a stronger relationship. When two variables show negative correlations the relationship is inverse. The strength of the relationship between two assets is provided on a scale from -1, returns always move in opposite direction, to 1, the returns always move together (Philips et all., 2012; Berk and DeMarzo, 2007).

Correlation is especially useful to intuitively explain differences in portfolio correlation. Nevertheless, the useful information that is provided by correlation as one of the essentials of the portfolio allocation strategy process has to be considered to be a statistical measure that is based on a variable return relationship. Assets with a low correlation can be seen as potential risk reducing assets in the portfolio because of increased diversification benefit of these particular assets. As correlations are determinative for the asset allocation in efficient portfolios, the figures are extensively discussed in the results. As the standardization of the dependency of the returns of two assets has no further effect on the calculation of portfolio risk, the covariance matrix is used to be able to compare the risk of a new portfolio that includes direct real estate with a base portfolio in which direct real estate is excluded.

The formula for portfolio variance displays that the variability of a portfolio is a function of the assets weights in the portfolio and the co-movement of the assets as is provided by the covariance (Philips et all., 2012):

∑∑ ( )

The variance between the assets included in the portfolio are collected in a portfolio variance matrix and summarized to derive at portfolio variance. By taking the square root out of this basic variance

(23)

measure, it will be able to present an intuitively meaningful description of the risk of portfolio returns:

√

3.2.2. Diversification potential

The research on the diversification potential of direct real estate in an investment portfolio is based on the diversification opportunities of direct real estate in an efficient mixed-asset portfolio and the diversification opportunities of direct real estate in a typical market portfolio. The first approach aims to describe the optimal allocations of the investors’ capital through time, at different levels of risk. Optimal allocations have a crucial impact on the performance of the portfolio, as better portfolios could yield higher returns to their beneficiaries or could lead to lower payments into an investment funds in the future (Hoesli et. all., 2004). In the second approach the diversification benefit is the natural result of the diversification potential of direct real estate as an asset class. In this approach the allocations are approximated as stable through time, in order to be able to exclude the diversification benefit that results from changes in allocations.

The analysis of the reduction of risk in portfolios that include direct real estate is examined solely on domestic portfolios. In both approaches, the ability to reduce portfolio risk is analyzed for investors with a long-time investment horizon as well as for investors with a medium-term investment plan. For long-term investments the yearly total returns for the entire research period from 1991-2012 have been analyzed. For medium-term investments the statistics are based on the 5-year rolling window average returns. The averages are a result of the geometric mean of yearly total returns from 1991 to 1995, 1992 to 1996, …, 2007 to 2011 and 2008 to 2012. The total research period from 1991-2012 comprises 18 of these rolling windows.

The analysis of the contribution of adding direct real estate as an asset class to a mixed-asset portfolio without real estate is performed with unhedged returns, which means that all the returns are calculated in the currency of the country that is considered. Especially investors who consider mean reversion in exchange rates will be likely to adopt such a strategy (Hoesli et all., 2004). Mean reversion is a mathematical concept that assumes that price changes will tend to move to an average over time so price changes in the exchange rates will be compensated over time. Another motivation for not making use of hedging could be that the cost of this hedge is too precious according to the investor.

(24)

The diversification potential of direct real estate is based on the relative comparison of the level of risk in a base portfolio without direct real estate and a new portfolio that includes the researched asset class. The method is identical in both approaches. The approaches are different in that in the first approach the allocations differ according to the optimization according the efficient portfolio premise, but in the second approach the allocations are held constant.

In the first approach the diversification benefit of real estate in efficient mixed-asset portfolios is determined. The calculation of efficient frontiers is performed by minimizing portfolio risk for given levels of portfolio return, a method provided by the earlier research of Lee & Stevenson (2006). In this minimization process, that is mathematically demonstrated in the efficient portfolio theory of Markowitz (1952), the variance of a portfolio is determined by the covariance between and the allocations to the included assets. The optimal portfolios are constructed by varying the proportion invested in each asset class within a mixed-asset portfolio. The efficient portfolios are generated for as well the entire research period as for the 5-year rolling windows for the U.S., Canada, France, the Netherlands and the U.K. The 5-year rolling windows produce eighteen configurations for each country, containing seven efficient portfolios for a base portfolio without direct real estate and seven efficient portfolios with direct real estate. Each portfolio exemplifies a particular asset allocation and standard deviation for a given expected return. The return is held constant in order to determine the diversification benefit of the portfolio in which direct real estate is included. The seven portfolios result from the return as a result of the minimum variance portfolio (σmin), the maximum return

portfolio for (σmax) and five portfolios that have next lowest fixed returns. These are consecutively

the portfolios up 10%, 30%, 50%, 70% and 90% of the difference between the minimum variance portfolio and the maximum return portfolio. The total research consists of 70 portfolios for the total research period from 1991-2012 and 1260 portfolios for the 5-year rolling periods for all the countries considered in this research

Several assumptions have been made in order to compute optimal portfolio allocations that minimize the variation:

1) the asset weights of the portfolio have to be positive as no short sales are allowed 2) the portfolio weights must sum up to one

3) the expected return of a portfolio ( ) is equal to the defined return of the portfolio ( ) a. except for the determined minimum variance portfolio (σmin)

(25)

The determination of the efficient portfolios at the lowest level of risk for a given expected return is determined by the solvation of the following calculation:

) ∑ ∑ ( ) ) ( ) ∑ ( ) ) ∑ ∑ ( ) ( ) ( ) ∑

It has to be mentioned that for this research no assumptions are made for the investment portfolios, which leads to corner solutions as the weights of different assets in the of efficient portfolios could be zero according to the equations (Black & Litterman, 1992; Lee & Stevenson, 2006). In order to be able to estimate the impact of the inclusion of direct real estate in portfolios that are more consistent with real portfolios, the second approach considers market portfolios as fixed.

In the second approach, a typical market portfolio is considered. The market portfolio is derived from the analysis of Lee & Stevenson (2006) and is kept constant for all the included countries in this research. In the typical market portfolio, the allocation to real estate is established at 20% of the total portfolio. In the base portfolio the share of capital is entirely allocated to indirect real estate. In the new portfolio the total allocation to real estate is equally divided between indirect real estate and direct real estate. The distribution of the investor’s capital in the assets that are included in this research is explained in the following table 4:

Table 4: Market portfolio allocations

The return and risk figures that result from this research are not necessarily optimal regarding the efficient market premise, but are definitely useful as investors are restrictively capable to adapt their

Allocations excluding direct real estate Allocations including direct real estate

Stocks 45,0% Stocks 45%

Bonds 30,0% Bonds 30%

Indirect RE 20,0% Indirect RE 10%

Cash 5,0% Cash 5%

(26)

investment portfolio to the efficient market portfolio that results out of historic data. As the allocations to the different assets are constant, the diversification potential of direct real estate is the natural result of the diversification potential of direct real estate as an asset class in a typical market portfolio. In the second approach the research is focused on the long-term and the medium-term investors perspective similar to the research in the first approach, for long-term investment perspectives the risk and return profile of a portfolio excludes and includes direct real estate respectively. The figures are collected for the U.S., Canada, France, the Netherlands and the U.K. For medium-term investments, in all 5-year rolling windows the risk of a base portfolio without direct real estate is compared to the risk of a new portfolio in which direct real estate is concluded.

The results are not necessarily optimal according to the efficient market premise. Therefore, in times that direct real estate have disadvantageous parameters; the diversification potential of the inclusion of this asset class is completely negated.

(27)

4. Results

The chapter is divided into four sections. First the correlations for each of the five countries are described for as well the total research period as the 5-year rolling windows. Correlation is next to the mean-variance parameter a determinative determinant of the diversification benefit of a particular asset class within a portfolio. In the second part the results for the efficient portfolios are determined in order to describe the diversification potential of direct real estate in an efficient mixed-asset portfolio. In the third part the results for the diversification potential of direct real estate in a typical investment portfolio are explained. Finally, the changes in the diversification potential of direct real estate are statistically justified.

4.1 Correlations

The correlations are described and summarized for each of the 5 countries that are considered in this research. The general perception on the benefits of the inclusion of direct real estate in a mixed-asset portfolio is the low correlation with stocks. This is a comparable feature as the characteristic of bonds in a traditional mixed-asset portfolio, with the difference that bonds generally have a lower return. Regarding the total research period from 1991 – 2012, direct real estate appears to be positively correlated to stocks. Direct real estate returns for Canada appear to have the lowest correlation with stock returns of 0.18 and the moderate correlation in the U.K. is the highest of all panels with a correlation of 0.58. Direct real estate provides therefore the least diversification benefit in the U.K. The correlation between bonds and direct real estate is negative for all countries, where the U.S. has a low negative correlation of -0.11 and France a relatively strong negative relationship of -0.47.

During the period of analysis the correlation between direct real estate and indirect real estate returns is moderate to high. This is a noticeable result considering the general perception that the financial market is regarded as a poor indicator for the behavior of the direct real estate market results. Especially the correlation of indirect and direct real estate returns in the U.K. is high, with a correlation of 0.75 over the entire period. In the U.S. the correlation between indirect real estate and direct real estate is the lowest with a correlation of 0.20. The correlations between indirect real estate and stocks are moderate for all the countries, with the lowest correlation for Canada and the highest for France (0.41 and 0.55 respectively). This is further evidence for the diversification benefits of direct real estate investments in order to lose risk in a mixed-asset portfolio, as a complementary investment asset to indirect real estate. The correlation between bonds and stocks appear to be as expected, with low negative or very low positive correlations between the returns of these asset

(28)

classes over the entire period. In the next table 5, the correlations for the Netherlands are provided as an example:

Table 5: Correlation Matrix for total research period (1991-2012)

To be able to justify the extensive research on the consecutive 5-year rolling window portfolios, the absolute difference between the correlation of the total research period and the average correlation of the rolling windows is examined. The average asset return correlations for the consecutive 5-year periods can differ strongly from the correlation figures of the total research period. In table 6 the descriptive statistics for the Netherlands are provided as an example. The difference varies between a minimal difference of 0.00 for the correlation between stocks and indirect real estate in the U.K. The maximum difference of 0.35 is found for the correlations between stocks and bonds in as well France as in the Netherlands. On average the differences are 0.07 for the U.S., 0.14 for Canada, 0.18 for France, 0.20 for the Netherlands and 0.11 for the U.K. The existence of this difference indicates that correlations can differ significantly for investors that consider shorter investment horizons. This result is in line with the general perception that diversification benefits are subject to constant modifications and are not stable over time.

The results and conclusions for the comparison of the average correlations for the rolling windows with the correlation for the total research period, give reason to explore the descriptive data for the different panels that are collected in order to observe remarkable correlations among two asset classes within the panels. The focus lies on the absolute values of the correlations between different asset classes. Where the correlation coefficient cannot exceed +1 for perfect positive correlated assets and -1 for perfect negative correlated assets, the greatest possible range between the minimum and maximum correlation coefficient will be 2. Regarding all asset classes for the studied panels the minimum range is 0.834 for the correlations between stocks and direct real estate in the U.K. and the maximum range is 1.908 for the correlation between stocks and bonds that is also found in the U.K. A range lower than 1 is exceptional which means that the statistical relationship between two assets have to differ for the examined investment periods, since this relationship can be positive as well negative for the same asset classes for different investment periods.

Panel IV: The Netherlands

Stocks Bonds Indirect RE Direct RE Cash

Stocks 1,00

Bonds -0,02 1,00

Indirect RE 0,57 -0,10 1,00

Direct RE 0,43 0,11 0,30 1,00

The changing diversification benefits of direct real estate within Mixed-Asset investment portfolios.