International Real Estate Returns: A Multifactor, Multicountry Approach

1. Introduction

The global market for publicly-traded real estate companies has grown dramatically over the last decade. According to the European Public Real Estate Association (EPRA), the market capitalization of real estate companies reached almost $300 billion at the end of 2002. Given that the U.S. alone comprises almost $150 billion, or 50 percent of the global market, it is not surprising that the majority of studies that model commercial real estate returns have been conducted on U.S. securities such as REITs or direct commercial real estate holdings, such as NCREIF indices. However, an increasing number of studies are addressing the risk and return behavior of commercial real estate in Europe and Asia.1 The aim of our study is to examine the returns of securitized real estate markets using new data provided by the European Public Real Estate Association (EPRA). This recently released data set has been developed by a consortium of financial institutions in conjunction with NAREIT to promote awareness of the public real estate market and facilitate research. It comprises monthly historical data on national indexes of almost 300 real estate companies from 14 countries in Asia, Europe, Canada and the U.S. It has quickly become a benchmark index for reference in the financial markets in Europe and Asia with updates to the index widely disseminated on electronic data services.

We study the risks and returns of real estate securities in different countries using multifactor models of international asset pricing. These models incorporate not only global market risk factors, but also country-specific market risk factors. In addition, these models allow for extra-market risks, such as value and size risks, in the spirit of Merton’s

1 See, for example, Eichholtz and Koedijk (1996), Eichholtz, Huisman, Koedijk and Shuin (1998), Case, Goetzmann and Rouwenhorst (1999), and Ling and Naranj o (2002).

(1973) Intertemporal Capital Asset Pricing Model (ICAPM). We are inspired to investigate the importance of these additional risk factors in part by recent developments in the international finance literature (Fama and French, 1998

This study makes two important contributions. First, the data reveal substantial variation in mean real estate returns and standard deviations across countries, but there is clear evidence of a common global market risk factor in international real estate returns (as proxied by the Morgan Stanley Capital International world index). We also find evidence of a significant country-specific market risk factor even after controlling for global market risk (Ling and Naranjo, 2002

The remainder of the paper is organized as follows. Section 2 provides an overview of the international asset pricing literature focusing on the specification global multifactor models. The research methodology of the current study is outlined in Section 3. The EPRA multi-country real estate database is described in Section 4 and our main results are presented in Section 5. Conclusions follow.

2. An International Factor Model for Real Estate Security Returns

International asset pricing research is a well-established field of study in financial economics. Since Karolyi and Stulz (2003) provide a comprehensive survey of this literature, we provide only a brief summary. Beginning with the international CAPM of Solnik (1974), research has concentrated on the specification of returns-generating models of market equilibrium that include factors such as currency risks (Adler and Dumas, 1983) and inflation risks (Vassalou, 2000). The research has also documented important sources of time-variation in the factor premiums associated with these risks (Ferson and Harvey, 1993

In a recent study, Fama and French (1998) examine whether a price-to-book factor is important in explaining international stock returns. A justification for the use of a price-to-book factor is a reflection of a shared sensitivity to financial distress (Fama and French, 1993). However, Griffin (2002) questions the importance of a global value and size factor. He finds that domestic size and value factors capture most explanatory power in a factor model, with little explanatory power added by international factors.

The literature on the returns-generating process for real estate securities is well developed at a country level, but multi-country studies are fewer in number and, to our

knowledge, none examine the explanatory power of extra-market sources of risk, such as of the global value factor. An important challenge for international commercial real estate research has been the absence of a consistent, comprehensive and widely-available database. However, despite the limitations of available data, there have been several studies that have examined multi-factor models (Eichholtz, Huisman, Koedijk and Schuin, 1998

3. Methodology

A commonly used returns-generating model takes the form of an international CAPM for real estate returns:

Rit – Rft = ái + âiw [Rwt – Rft] + åit (1)

where Rit is the U.S. dollar-weighted return on country i’s real estate index in time period t, Rft is the U.S. risk free rate in time period t and Rwt is the U.S. dollar-weighted total return on the world market portfolio in time period t. The expected risk premium associated with an investment in the world market portfolio is given by Rwt – Rft. The parameters to be estimated are the intercept, ái and ßiw, the return sensitivity (exposure) of country i’s real estate returns to returns on the world market portfolio. The error term for country i is represented by åit.

Each country’s alpha, ái, is a measure of the investor’s excess return on a risk¬adjusted basis. Jensen’s alpha assumes there is one source of risk: the systematic risk in

that particular economy. We follow previous research on multi-factor models applied to international commercial real estate (Naranjo and Ling, 2002) and employ a two-factor model that captures both the global systematic risk variable for the home market and a second variable that attempts to capture additional country-related market risk factors. Estimating this model requires a two-step procedure. For each country, we estimate the following OLS regression:

Rc it – Rft = ái + âiw [Rwt – Rft] + åc it (2)

where Rc it is the U.S. dollar-weighted return on country i’s stock index in time period t and åc it are the unexplained residuals. The residuals represent the portion of each country’s stock returns not explained by movements in the world market portfolio. To estimate a multifactor specification of the unconditional real estate asset-pricing model, these unexplained residuals as an orthogonal country-specific market risk factor is added to the return-generating process:

Rit – Rft = ái + âiw [Rwt – Rft] + âiR [ec it] + åit (3)

where ßiR is the estimated exposure of country i’s real estate return index to returns on the orthogonalized country specific risk factor. Because important factor-influencing events in real estate markets are often local, we expect that ßiR will be significant for a number of markets that we study. But, because of the growing integration of global markets in general through increased cross-border capital flows and government-led market liberalization events, we also expect ßiw will be significant (Karolyi and Stulz, 2003). The most important result to focus on will be the relative magnitudes of the two sensitivity coefficients across the different national real estate markets.

The orthogonalization of the return-generating process is a useful approach to capture the incremental explanatory power of additional variables that impact real estate returns, but it is not without its drawbacks. The residuals used in the orthogonalization are simply the unexplained variation in the country’s share market index. Such a variable provides little information on the nature of the risk which it proxies. Such an approach may be acceptable if the sole purpose of the model is performance evaluation but this is rarely the case.

As a consequence, we attempt to try to identify additional variables that can be used to explain the variation of the dependent variables, the country’s commercial real estate index. One approach is to utilize economic risk factors, such as inflation, interest rate, bond term and default risks, as in the model employed by Karolyi and Sanders (1998). Another approach is to utilize fundamental factor risks, such as book-to-market value and firm size, as in the approach of Fama and French (1992, 1993). They specify a domestic three-factor model in which:

Rit – Rft = ai + bi MRFt + ci SMBt + di HMLt+ eit (4)

where MRFt is the excess return on the country’s stock market index, SMBt is the spread between the return on small stocks and the return on large stocks, HMLt is the spread between the returns on stocks with high book-to-market ratios and the returns on stocks with low book-to-market ratios, and bi, ci, and di are the associated sensitivity coefficients.

In their international model, Fama and French (1998) extend their specification above, but use only a global market risk factor, MRFw t, and a global value risk factor, HMLw t. They also construct country-specific market risk and value risk factors for the

same markets that we study in this paper. We employ their factors and consider a variety of different specifications of a global factor model in an attempt to explain the variation of commercial real estate returns in Europe, Asia, Canada and the United States. For example, in order to examine whether there is additional explanatory power for real estate returns from a country-specific value risk factor beyond a global market and orthogonalized country-specific market risk factor, we test:

Rit – Rft = ái + âiw [Rwt – Rft] + âiR [ec it] + âiH-LB/M HMLc t + åit (5)

where ßiH-LB/M is the sensitivity coefficient to the country-specific value factor, HMLc t. Finally, we test an extension of this model in which we allow U.S.-based market risk, size and value factors beyond the global and orthogonalized country-specific market risk factor. Our objective is to determine whether the additional explanatory power of the world market portfolio stems in part from the U.S. market and specifically due to U.S.-based value or size risk. The model is specified as:

Rit–Rft = ái+âiw [R wt–Rft]+âiR [ åc it]+âUS iR [ åUS t]+âUS iSMB SMB US t+ âUS iHML HML US t+åit (6 )

We are interested in seeing how much of the sensitivity to world market risk, ßiw, is attenuated by the addition of the U. S. based risk factors.

4. Data

The European real estate securities data used in this study was obtained from the European Public Real Estate Association (EPRA). This database includes country-level commercial real estate indices comprised of 238 constituent companies totaling over $288 billion in market capitalization as of April 2002. The indices are market¬capitalization weighted on a free-float-adjusted basis. Table 1 presents summary data on

the constituent companies. Fourteen countries are represented from Asia, Europe and North America. The U.S. with its 106 companies and $150 billion in market cap is the largest single country component and is followed by the U.K. (34 companies, $30 billion), Australia (26 companies, $27 billion) and Hong Kong (13 companies, $28 billion).

There are a number of requirements for a company to be included in the EPRA indices related to size, ownership, trading, reporting and to the development of income¬producing real estate. Equally importantly, there are a number of exclusionary features that are not considered relevant real estate activities, such as the construction of residential homes for sale, the provision of construction management, general contracting services and property management services, holding companies and those related to gaming, theme park or other entertainment businesses. Many of the constituent company names are well-known real-estate companies including Starwood Hotels (U.S.), United Dominion Realty Trust (U.S.), Lend Lease (Australia), New World Development (Hong Kong), Unibail (France), IVG Holding (Germany) and Canary Wharf Group (U.K.). An appendix detailing index construction and composition is available from the authors upon request.

The descriptive statistics for the monthly real estate returns data are found in Table 2. All statistics are reported for the full period, February 1990 to December 2001. Of the real estate indices, the United States has the highest mean return (1.205%) with Sweden and Japan recording the lowest mean (-1.162% and –0.694, respectively). The negative mean returns for Sweden arose from the large falls in share prices of the constituent property companies in the early 1990s recession. Singapore records the

highest standard deviation (12.828%) of the group. Only one country shows evidence of skewness (Italy), which has a positive skewness coefficient. Several countries recorded significant excess kurtosis (Germany, Italy, Singapore and Sweden). The finding of skewness in securitized property returns has been discussed by Bond and Patel (2002), however, at the country level, the finding of skewness is less common and accordingly no attempt is made in this study to allow for skewness in the specification of the models.

In addition to the EPRA indices, we include the Morgan Stanley Country indices (MSCI) for each country used in the study. The summary statistics for the MSCI world index are reported in Table 2 with the EPRA indices. We also employ the Fama and French (1993) U.S. book-to-market spread returns, HML, and market capitalization spread returns, SMB, as well as the local book-to-market spread factors, HMLc, for each country (Fama and French, 1998), all of which was obtained from the authors.

5. Results

Using the EPRA data set described above, results for estimation of the models outlined in Section 3 are presented below. The next subsection reports on the estimation of the single-factor International CAPM and is followed by the presentation of various multivariate extensions.

5.1 International Capital Asset Pricing Model

We estimate the single-factor model of equation (1) using ordinary least squares (OLS) for each country’s real estate index excess returns. The returns on the international market portfolio are proxied by the MSCI Global Index. While there are a number of limitations with using this index as the global market portfolio, in particular, in terms of

omitted asset groups (including real estate), it is commonly used in international asset pricing studies (Fama and French, 1998). Note that the model is estimated without restricting the intercept terms to be zero. This provides one means of allowing for the suitability of the model to be assessed. The resulting estimates are shown in Table 3.

A striking feature of the results is that almost all country indexes, with the exception of Australia, Hong Kong and the U. S., show potential underperformance. The Jensen’s ái estimates are lowest for Sweden (-1.77% per month) and Japan (-1.33% per month). However, only in the case of Sweden and the U. S. is the intercept significantly different from zero. The ßiw coefficients for the global market portfolio are positive and significantly different from zero in all but one instance (Germany). However, the coefficients are typically low, positive values. In the table, we compute t-statistics for those coefficients relative to the null hypothesis that they are equal to one and in nine out of the 14 countries, we are able to reject the null.

Generally, the proportion of variation in individual country excess returns on real estate explained by the global market portfolio is low (Ling and Naranjo, 2002), although the adjusted R2 statistics reported in Table 3 are lower on average than even those reported in previous research. Surprisingly, both in this study and in Ling and Naranjo, the single factor model does not explain any fraction of the excess returns for the German data.

We perform two specification tests across the fourteen countries. The first is a Wald test that evaluates whether the global beta risk of securitized real estate is equal across all fourteen international markets. This statistic (?2 with 13 degrees of freedom) is not reported in the table, but the null hypothesis of equality is easily rejected. In addition,

we perform a Wald test (also, ÷2 with 13 degrees of freedom) of the hypothesis that the intercept terms are jointly equal to zero. The resulting test statistic is 27.579 (with a p¬value of 0.01). We infer from these tests that there is potentially important mis¬specification error in the simple single-factor model. There are interesting patterns that suggest regional differences may play an important role in this mis-specification. For example, the largest positive ßiw and largest negative ái are associated with Asian real estate markets, such as Hong Kong, Japan and Singapore. We will continue to explore these regional patterns below.

As a measure to further assess the model, the ability of an international capital asset pricing model to explain the cross section of international real estate returns is examined. A key outcome of the CAPM is that there is a linear relationship between risk (beta) and expected returns. This relationship between systematic risk and expected return can be estimated using the cross-sectional regression approach of Fama and MacBeth (1973). The approach of Fama and MacBeth requires that the following regression model be estimated for each month t over the sample period:

Rit – Rft = ë0t + ëmt âiw (7)

where ßiw is the beta risk coefficient from the international CAPM (equation 1) and Rit, Rft are the country-specific real estate and risk free returns, respectively, as defined in the Section 3. Obtaining the time series of estimated coefficients (?0t,?mt) allows for two implications of the CAPM to be assessed

As the true betas are not observed the estimated betas obtained from an OLS regression of equation are used. The power of our tests is impeded further by two

limitations. First, we estimate the betas only once for the full period and apply them to the cross-sectional regressions each month in the time series. Second, we employ only fourteen national real estate index series in the cross-sectional tests limiting the potential dispersion of the ßiw estimates and the ability of the procedure to detect meaningful variation in the global market risk premium.2 Nevertheless, on testing the above restrictions it was found that neither implication was consistent with the data. Specifically, we computed that E[ë0t] averaged –0.032 with a standard error of 0.026 and thus insignificantly different from zero, consistent with model implications. The expected global market risk premium, E[ëmt], however, is significant and negative with a value of –

0.341 with a standard error of 0.002.

5.2 Multifactor International Asset Pricing Models

As a starting point to our multifactor analysis, we evaluate the results from implementing the model of Ling and Naranjo (2002) in equation (3) for our EPRA data. The findings are presented in Table 4. As stated in Section 2, the method of Ling and Naranjo (2002) creates orthgonalized country indexes that can proxy for a range of priced risks in real estate returns. The results reported are estimated with seemingly unrelated regression (SUR) allowing for potential contemporaneous correlation among the error terms of the model for the various national real estate returns series. The pooled cross¬sectional/time series regression approach is also useful for evaluating cross-market restrictions such as the joint equality of the intercept and slope coefficients using Wald

statistics (distributed ÷2 with N-1 degrees of freedom, where N is the number of countries

2 Roll and Ross (1994) provide a further critique that the Fama-MacBeth testing procedure is sensitive to departures of the market proxy from the true global market portfolio. See Campbell, Lo and MacKinlay (1997) for detailed discussion of the econometric issues relating to the Fama-MacBeth approach.

in the cross-section. At the bottom of the table, we provide Wald tests of the joint equality of the ái, ßiw, and ßiR coefficients, respectively and together, and, in addition, of their joint equality to zero. SUR estimation and the Wald tests are implemented using the GAUSS econometric package.3

The important results in the table arise from the positive and statistically significant ßiR coefficients for the orthogonalized country-specific market factors. These results obtain for each of the markets except Germany. In some countries, the national real estate indices have ßiR as large as 1.42 (Singapore) and 1.28 (Hong Kong). A Wald test of the null that the ßiR coefficients are jointly equal zero is easily rejected with a ÷2 of 1,630.7 (p-value of 0.00). Because these coefficients are constructed to be orthogonal to the global market factor, the ßiw essentially retain similar estimates. The ái coefficients are still negative values, but only in one or two markets are they statistically significant (Japan, Singapore and Sweden). Interestingly, these exceptional cases are enough to lead to a rejection of the joint test that these ái are equal to zero (÷2 of 38.34, p-value of 0.00). Finally, the adjusted R2 are significantly higher than those in Table 3.4 For Australia, Japan and Singapore, the adjusted R2 are as high as 0.60 whereas those in Table 3 rarely exceeded 0.30. These results are consistent with the findings of Ling and Naranjo (2002).

To explicitly consider one extra-market risk variable, the country-specific value factors of Fama and French (1998) are used in addition to the orthgonalized country portfolios. One advantage is that this enables a fundamental source of country-specific

3 Use of OLS in the presence of contemporaneous correlation in the error terms can lead to regression coefficients that are not efficient. This approach is only used in the multifactor models when the explanatory variables differ across the equations.

4 Goodness-of-fit measures for SUR systems have been devised (Greene, Chapter 15, pp. 618-619). These measures are based on the covariance matrix of the residuals using the feasible generalized least squares estimates. We do not compute such a measure, but rather report the individual adjusted R^2 from the OLS regressions.

risk to be assessed rather than through a generalized country-risk term. Another more important advantage is that we are able to judge the extent to which a value factor often associated with distress risk (Fama and French, 1993) is systematically related to real estate returns. The results are shown in Table 5.

For 11 of the 14 countries examined, the value term, ßiH-LB/M, is positive and significant in explaining real estate returns in the sample. The largest values obtained for real estate markets in Hong Kong, Italy, Singapore and the U.K. Overall, the joint test of whether the ßiH-LB/M equal zero is rejected (÷2 of 237.4 with p-value of 0.00). Few, if any, of the inferences about the intercepts or global/country-specific market risk coefficients change with the addition of the value factor. In eight markets, the explanatory power of the model increases compared to the previous model of just including a country risk term. The largest increases in adjusted R2 occurred for Italy, Sweden and the U.K. In a few cases, the explanatory power of the model fell or remained the same (Australia, Belgium and France). Building on the work of Ling and Naranjo (2002), it can be stated that a country distress factor proxied by a value-growth portfolio captures more, or at least a similar amount, of the time-series variation of real estate returns than does a specific country risk proxy for most countries. Interestingly for both models, the intercept terms are now significant for Japan (-2.53%), Singapore (-2.26%), Sweden (-2.37%) and the U.S. (2.30%) suggesting that the parsimonious approach to modeling in this study has likely not captured additional risks priced in these real estate markets.

In results not reported in this study, another of the Fama and French country value portfolios based on the difference between returns on a high dividend yield portfolio and a low dividend yield portfolio was examined. The model was found to explain a lower

proportion of the movements in returns than the use of a portfolio based on the book to market portfolio just reported.

A much discussed risk factor that may also proxy for a distress factor is a measure of capitalization (or size). Fama and French (1993) found, in addition to a value measure, a size factor was important in explaining U.S. stock market returns. As a comparable country based size portfolio to the value portfolio of Fama-French (1998) was not available, alternative ways of capturing a size effect were investigated. In particular, the issue of whether international real estate returns were influenced by U.S. distress factors is tested. This may be valid if international economies are closely integrated to the U. S. market where financial distress factors for the U.S. could be relevant in other international markets. To test this assumption, the Fama-French (1993) size (SMB) and value (HML) portfolios based on U. S. stock returns were used. The results are found in Table 6.

Across almost all country real estate returns the model results are generally poorer than those contained in Table 5. The only exceptions are those of the U. S., and surprisingly, Australia, where a marginally higher proportion of real estate returns were explained by the U. S. size and value factors (and international market returns) than explained by country specific value factors. In the U.S. the real estate index has positive and significant coefficients for SMB and HML indicating a small-capitalization and value bias, whereas for Australian real estate markets, the same coefficients have negative values indicating a large-capitalization and growth bias. For the other markets, the U.S.-based (orthogonalized) market, HML and SMB factors are not significant. We perform Wald tests of the joint significance of these coefficients across the 14 real estate markets

and report them at the bottom of the table. As expected, the null that the U. S. market and SMB factors are jointly equal to zero cannot be rejected (with ÷2 of 11.23, p-value of 0.60, and 15.26, p-value of 0.29, respectively). The Wald test for the joint significance of the U.S. HML factor is marginally significant (÷2 of 22.69, p-value of 0.05).

6. Conclusions and Implications

This study has examined the risk and return attributes of securitized international real estate shares. To this end, we take advantage of a comprehensive new database provided by the European Public Real Estate Association (EPRA) that covers 288 real estate companies in 14 countries in Asia, Europe, and North America. For monthly returns on national capitalization-weighted real estate indexes in each country, we examine the usefulness of a range of single-factor and multifactor returns-generating models. Specifically, we estimate an international CAPM model with the MSCI world index as global market proxy as well as multifactor models that capture country-specific and global market risks as well as country-specific and global size and value risks.

We find that there is evidence of a strong global market risk component in the real estate sectors of most countries. However, even after controlling for the effects of global market risk, an orthogonalized country-specific market risk factor is highly significant, especially for real estate indexes in Asia-Pacific markets. We find that a country-specific value risk factor has some explanatory power in addition to the country-specific market factor, but U.S.-based market, value and size risk factors do not provide any additional explanatory power.

These findings are important in a number of ways. First, while our evidence of a significant country-specific market risk factor after controlling for global market risk is similar to that in previous studies (Ling and Naranjo, 2002

That a country-specific value risk factor provides significant explanatory power for real estate securities leads to another key implication of the study. We show that this country-specific value risk factor is unique and not subsumed by global or local market risks or by U.S.-based value risk factors. This is an important finding for the current debate about the relative importance of an international value risk factor in international equity markets (Fama and French, 1998

We must caution readers about several important limitations. Even though the EPRA database is one of the most comprehensive and consistent available, there are still important country-specific variations in the definition of a real estate security. While real estate investment trusts (REITs) in the U.S. are essentially closed-end funds with real property in their underlying portfolios, the composition of foreign real estate securities is quite varied. In particular, a number of property developers are included in the EPRA series. Since it is difficult to ascertain which security is a U.S.-type pass-through and which is more complicated in terms of development and investment, we do not specifically control for it in the study. Future research should focus on the constitution of these indices and perhaps examine disaggregated indexes that account for these distinctions across countries.

We also understand that there are important country-level differences in real estate markets that can influence inferences about risk and returns patterns. Consider, for example, the fact that the market for U. S. REIT and U.K. property companies is much more actively traded than those in markets like Denmark, Singapore and Austria. There are important differences in regulation, disclosure standards and accounting standards that are required for publicly listing securities in different markets. In addition, the governance standards of these companies vary considerably ranging from companies with large, dominant controlling shareholders to others with widely-dispersed shareholder bases. We strongly encourage research to evaluate the sensitivity of our inferences to these factors.

Finally, while we are among the first international real estate studies to evaluate multifactor models with extra-market risks, the scope and variety of the models we

examine is still limited. We look forward to studies that evaluate an even richer variety of factor models for real estate securities that include economic risk factors (such as inflation, exchange rate risks), fundamental risk factors (including value, size, but also dividend yield, turnover rate, earnings growth factors) and technical risk factors (such as momentum, reversals and other forms of predictability).

Bibliography

Adler, M and Dumas, B. 1983, “International Portfolio Choice and Corporate Finance: A Synthesis,” Journal of Finance, 38, 925-984.

Bond, S.A. and Patel, K. 2002, “The Conditional Distribution of Real Estate Returns: Are Higher Moments Time Varying?” Journal of Real Estate Finance and Economics, 26(2&3)..

Campbell, J.Y, Lo, A.W. and MacKinlay, A.C., 1997, The Econometrics of Financial Markets, Princeton University Press, Princeton.

Case, B., Goetzman, W. and Rouwenhorst, 1999, “Global Real Estate Markets – Cycles and Fundamentals”, mimeo, Yale School of Management.

Chan, K.C., P.H. Hendershott, and A.B. Sanders, 1990, “Risk and Return on Real Estate:Evidence From Equity REITs,” AREUEA Journal 18(4), 431-452.

DeSantis, G. and Gerard, B. 1997, “International Asset Pricing and Portfolio Diversification with Time-Varying Risk,” Journal of Finance, 52, 1881-1912.

Dumas, B. and Solnik, B., 1995, “The World Price of Foreign Exchange Rik,” Journal of Finance, 50, 445-479.

Eichholtz, P.M.A., R. Huisman, K. Koedijk, and L. Schuin, 1998, “Continental Factors in International Real Estate Returns,” Real Estate Economics 26, 493-509.

Eichholtz, P.M.A., and K. Koedijk, 1996, “The Global Real Estate Securities Market,” Real Estate Finance, 13, 76-82.

Fama, E. and J. MacBeth, 1973, “Risk, Return and Equilibrium: Empirical Tests,” Journal of Political Economy, 81, 607-636

Fama, E. and French, K.R., 1992, “The Cross Section of Expected Stock Returns,” Journal of Finance, 47, 427-465.

Fama, E. and French, K.R, 1993, “Common Risk Factors in the Returns on Stocks and Bonds,” Journal of Financial Economics 33, 23-49.

Fama, E. and French, K.R., 1998, “Value versus Growth: The International Evidence,” Journal of Finance, 53, 1975-1999.

Ferson, W. and Harvey, C.R, 1991, “The Variation of Economic Risk Premiums,” Journal of Political Economy 99, 385-415.

Ferson, W and Harvey, C.R., 1993, “The Risk and Predictability of International Equity Returns,” The Review of Financial Studies, 6, 527-566.

Greene, W.H., 2000, Econometric Analysis, 4th edition, Prentice Hall, Saddle River, NJ.

Griffin, J., 2002, “Are the Fama and French Factors Global or Country-Specific?” Review of Financial Studies 15, 783-803.

Karolyi, G.A. and A.B. Sanders, 1998, “The Variation in Economic Risk Premiums in Real Estate Returns,” Journal of Real Estate Finance and Economics 17, 245-262.

Ling, D. C. and A. Naranjo, 1996, “Economic Risk Factors and Commercial Real Estate Returns,” Journal of Real Estate Finance and Economics, 14, 283-307.

Ling, D. C. and A. Naranjo, 1999, “The In of Commercial Real Estate Markets and Stock Markets,” Real Estate Economics, 27(3), 483-516.

Ling, D.C., A. Naranjo, and M. Ryngaert, 2000, “The Predictability of Equity REIT Returns,” Journal of Real Estate Finance and Economics, forthcoming.

Ling, D. C. and A. Naranjo, 2002, “Commercial Real Estate Return Performance: A Cross-Country Analysis,” Journal of Real Estate Finance and Economics, 24, 119-142.

Liu, C.H. and J. Mei, 1996, “The Predictability of International Real Estate Markets, Exchange Risk, and Diversification Consequences,” Real Estate Economics 26, 3-39.

Merton, R.C., 1973, “An Intertemporal Capital Asset Pricing Model,” Econometrica, 41, 867-887.

Roll, R. and Ross, S.A, 1994, “On the Cross-Sectional Relation between Expected Returns and Beta,” Journal of Finance, 49, 101-122.

Sanders, Anthony, 1997. “The Historical Behavior of REIT Returns: A Capital Market Perspective,” The Handbook of Real Estate Investment Trusts, edited by R. Garrigan and J. Parsons (Irwin Publishing Co.).

Solnik, B.H., 1974, “An International Market Model of Security Price Behavior,” Journal of Financial and Quantitative Analysis, 9, 537-554.

Vassalou, M (2000), “Exchange Rate and Foreign Inflation Risk Premiums in Global Equity Returns,” Journal of International Money and Finance, 19, 433-470.