Tests of common real estate risk

premia in a time-varying expected

return framework

Tien Foo Sing and Leiting Deng

Department of Real Estate, National University of Singapore, Singapore, and

Hong Wang

School of Economics and Management, Tsinghua University, China

Abstract

Purpose – This paper aims to test the predictability of the three asset classes, namely direct

property, bond and property stocks in Singapore.

Design/methodology/approach – Using the generalized method of moment (GMM) estimation

methodology, the authors first estimate the excess returns of assets on five instrumental variables and

a constant term. Next the common risk factors are tested in three parts involving different portfolio of

sample assets.

Findings – The empirical results shows that there are at most three common risk factors that can be

used to predict the excess returns of six asset classes, that include four direct property assets, bonds

and property stocks. The results also indicate that there are separate common risk premia that are

priced in property stock and direct property markets, which indirectly reject the hypothesis that the

two property markets are integrated.

Practical implications – The empirical results that reject the market integration between property

and property stock markets imply that there are significant diversification benefits for holding both

assets in investors’ portfolios. The two property assets capture different risk premia in the markets.

Research limitations/implications – The GMM specifications that include five instrumental

variables may not fully capture all risk information. Omission of other variables is, however, traded-off

against the parsimony of the model specification. More independent variables could be included in the

future studies, and more asset classes could also be added to the tests.

Originality/value – The study provides alternative evidence to the test of market integration

between property and property stocks in Singapore. It also verifies the earlier study in the USA that

property and stock market effects could be separately priced by the market.

Keywords Real estate, Risk management, Return on investment, Singapore

Paper type Research paper

1. Introduction

In view of the trends in real estate securitization, investigating the integration between

real estate market and other capital markets has become increasingly important (Liu

et al., 1990). If there is no integration between the two markets, risk diversification can

be achieved through an efficient portfolio consisting of the two assets. Empirical tests

on market integration between real estate and other financial asset markets have been

conducted in the USA (see Liu and Mei, 1992, Mei and Lee, 1994, Ling and Naranjo,

The current issue and full text archive of this journal is available at

www.emeraldinsight.com/1463-578X.htm

The authors wish to thank the participants at the European Real Estate Society Conference in

Milan, Italy, 2004, for their comments.

Common real

estate risk

premia

359

Received December 2006

Accepted April 2007

Journal of Property Investment &

Finance

Vol. 25 No. 4, 2007

pp. 359-369

q Emerald Group Publishing Limited

1463-578X

DOI 10.1108/14635780710762508

1999), the UK (Lizieri and Satchell, 1997), Australia (Wilson et al., 1996) and Hong Kong

(Fu and Ng, 2001). The results were, however, mixed. Evidence of cross-country

integration of direct real estate markets and capital markets was found in studies by

Mei and Hu (2000), Liu and Mei (1998, 1999) and Bond et al. (2003).

Real estate investment trust (REIT) market in Singapore is at its infancy stage of

development. Publicly listed property stocks have been regarded as a close proxy for

securitized claims on real estate assets. The performance of securitized real estate is

expected to track closely the market values of the real estate assets held by the listed

companies, if the two markets are integrated. If this hypothesis is not rejected, it

implies that investors can efficiently price risk premiums embedded in direct real

estate market based on price information of property stocks. However, the results of

the empirical tests using Singapore data have been mixed (Ong, 1994, 1995; Liow, 1998,

2001; Sing and Sng, 2003, Sing, 2004).

This study aims to empirically re-examine the market integration between

securitized and direct real estate markets in Singapore using the multifactor

latent-variable model proposed by Mei and Lee (1994). It tests the existence of common

risk factors across different asset markets, which include property stock, government

bond, and four direct property sub-markets (industry, office, residential and retail). Our

results show that there are at most three common risk factors that explain the price

variations in direct property, property stocks and bond markets in Singapore. The

results imply that the two real estate markets are integrated.

The remainder of the paper is organized into five sections. Section 2 reviews the

literature on the market integration. Section 3 discusses the multi-factor latent variable

empirical methodology. Empirical data used and analyses of empirical results are

covered in Sections 4 and 5. Section 6 concludes the findings of the paper.

2. Literature review

The notion of market integration is an important condition underlying many

well-accepted investment theories. If two asset markets are integrated, it would not be

possible for investors to arbitrage in one market by using information in another

market. Efficient diversification of risks can then be achieved by holding a portfolio of

assets across the markets Liu et al. (1990) tested the market integration between equity,

equity REITs and non-farm commercial real estate in the USA, and found that equity

REIT and stock markets were integrated. However, they rejected the hypothesis that

commercial real estate market and stock market are integrated.

If two markets were integrated, there should be no differential price premia

associated systematic risk factors in the markets. Ling and Naranjo (1999) empirically

tested the risk premiums of a set of pre-defined macro-economic factors in stock and

real estate markets. The fundamental variables include growth rate of industrial

production, consumer consumption, yield spread between long term and short term

government bonds, and unexpected inflation rate. They found evidence of market

integration between REIT and stock market, but no integration was found between

direct real estate and stock market. The results were consistent with those found in the

earlier study by Liu et al. (1990).

Liu and Mei (1992), Mei and Lee (1994) applied the generalized method of moments

(GMM) methodology to test common latent factors in asset markets. Liu and Mei (1992)

showed that the real estate return variations can be fully captured by the common risk

JPIF

25,4

360

factors in stock and bond markets. However, Mei and Lee (1994) subsequently found

that real estate market risk is independently priced as the third common factor in the

model. The results suggest that real estate asset should not be under-represented in

well-diversified portfolios of investors. Li and Wang (1995) using the same GMM

methodology, however, found that latent real estate factor was not significant in

predicting excess returns of REITs and other stocks.

The earlier tests of market integration in Singapore have produced mixed results.

Ong (1994) found a contemporaneous long-term relationship between property stock

return, direct real estate return and 3-month Treasury bills interest rate using a

structural vector autoregressive (VAR) model. The cointegration relationship

between securitized real estate and direct real estate was, however, rejected by Ong

(1995) and Liow (1998). Sing (2004) tested common risk premiums in the two real

estate markets, and found no evidence to support integration in the two real estate

markets. However, using Generalized Autoregressive Conditional Heteroscedasticity

in Mean (GARCH-M) model, Sing and Sng (2003) found that incremental information

in conditional volatility flowed uni-directionally from unsecuritized market to the

securitized property market. They then concluded that the two markets are partial

integrated.

3. Latent-variable asset pricing model framework

The latent factor asset pricing framework as applied by Liu and Mei (1992) to test

common risk factors in the US asset markets is summarized below. First, a k-factor

excess return function of asset i, ~ri;t1, is defined as:

~ri;t1 ¼ Et½~ri;t1_ X

K

k¼1

bik

~fk;t1 ~ 1i;t1 ً1ق

where Et½~ri;t1_ is the expected excess return on asset i conditional on information

known at the end of time t, bik is the time-invariant factor loading of the K-th factors,

and ~ 1i;t1 is the idiosyncratic error. The zero-beta return, Et½~ri;t1_, has the following

linear functional form:

Et½~ri;t1_ ¼X

K

k¼1

bik lkt ً2ق

where lkt is the “market price of risk” for the k-th factor at time t. For an information

set with a vector of L forecasting variables, Xnt, [n¼1. . .L], lkt is a linear function of

these variables:

lkt ¼X

L

n¼1

uknXnt ً3ق

and equation (2) can be represented as:

Et½~ri;t1_ ¼X

K

k¼1

bikX

L

n¼1

uknXnt ¼X

L

n¼1

ainXnt ً4ق

Common real

estate risk

premia

361

where ain is the risk premium for forecasting variable Xnt, which is subject to the

restrictions below:

aij ¼X

K

k¼1

bik ukn ً5ق

where bik and ukn are free parameters.

The model is normalized by setting the factor loadings for the first K assets as

follows: bij ¼ 1 (if j ¼ i) and bij ¼ 0 (if j i) for 1 # i # K. The excess return matrix

is partitioned into, [R ¼ ًR1; R2ق], where R1 is a T £ K matrix of excess returns of K

assets, and R2 is a T £ ًN 2 Kق matrix of excess returns for the rest of the assets:

R1 ¼ X Q m1 ً6aق

R2 ¼ X a m2 ً6bق

where X is a T £ L matrix of the forecasting variables, Q is a matrix of uij, and a is a

matrix of aij.

The significance of common forecasting variable, X, in predicting excess returns of

assets is a test of the rank restriction, [H0: a¼QB], where B is a matrix of bij elements.

The unrestricted conditional excess returns and the restricted conditional excess

returns are estimated using Hansen’s (1982) generalized method of moments (GMM)

methodology.

4. Empirical data

4.1. Data source and definition

Quarterly excess returns of property stock, long-term government bond, industry,

retail, office and residential properties in Singapore are computed for sample periods

from 1988Q2 to 2006Q2.

The Singapore Exchange (SGX) Property Sub-sector Index (PTYS) is used to

represent the performance of securitized real estate. The quarterly bond yield (rq;t) is a

compounded rate calculated from the annualized yield (ra;t ) of five-year government

bond (BOND) as: [rq;t ¼ ً1 ra;tق1=4 2 1]. The five-year government bond (BOND)

data are obtained from the Monetary Authority of Singapore (MAS) database.

Quarterly excess returns of industry (PPII), office (PPIO), residential (PPIR) and retail

(PPIS) properties are computed from the transaction price indices published by the

Urban Redevelopment Authority (URA) of Singapore.

The quarterly excess return of asset x, [ERx ], is computed by subtracting price

returns of indices, [Rx ], by the risk-free rate, [Rf ], which is denoted as [ERx¼Rx 2 Rf ].

The three-month Treasury bills rate published by the MAS is the proxy of the risk free

rate. The annualized three-month Treasury yield (Rf,a) is converted into the quarterly

risk-free rate (Rf,q) based on the same equation used to compute the quarterly five-year

bond yields.

Four instrumental variables are used in the models to control for systematic risks,

which include credit spread (CRDSD), excess equity market return (ESGXR), term

structure of government bonds (TERMS), unexpected inflation rate (UNINF), and yield

spread (YSPRD). CRDSD is defined as the difference between prime lending rate and

three-month inter-bank rate. ESGXR is the return of SGX-all share index over the risk

JPIF

25,4

362

free return. TERM is the term spread of annualized rates between five-year long term

government bond yield and two-year short term government bond yield. YSPRD is the

yield spread between three-month commercial bill and three-month T-bill. The

unexpected inflation (UNINF) is computed based on the interest rate model of Fama

and Gibbons (1984). Data on government bond yields are obtained from the MAS

database, and other data on economic variables are obtained from the Time Series

Retrieval and Dissemination (TREND) database by the Department of Statistics,

Singapore.

4.2. Descriptive statistics

Descriptive statistics of excess returns of six asset classes and four forecasting

variables are reported in Table I. During the sample period from 1988Q2 to 2006Q2,

property stock was the best performer among the asset classes in terms of excess

return with a mean excess quarterly return of 0.36 per cent. Property stock was also the

most volatile asset during the sample period with a standard deviation of 20.8 per cent.

Among the direct property assets, residential property has the highest average excess

return of 1.1 per cent. Shop has a negative return of 20.2 per cent over the sample

period, but it price changes were the least volatile among the direct property assets

with a standard deviation of 4.3 per cent. Bond was the safest investment with the

lowest volatility of 0.2 per cent.

Table II shows the correlation matrices between asset classes and fundamental

variables. Correlations between excess returns of two pairs of direct properties:

office and shop, and office and industry were reported at 0.77 and 0.73 respectively.

Property stock was more closely correlated with residential property (0.43) than

with other direct property assets. Pair-wise correlations between government bonds

and property stock and other direct properties were relatively low, which imply that

diversification benefits can be achieved by mixing bond and other asset classes in

portfolios.

Descriptive statistics and data source Symbol Data source Observation Mean

Std

dev.

Excess return variables

Excess return of SGX property stock EPTYS Datastream 73 0.036 0.208

Excess return of five-year government bond EBOND MAS 73 0.004 0.002

Excess return of industry property EPPII URA 73 0.009 0.056

Excess return of office property EPPIO URA 73 0.002 0.060

Excess return of residential property EPPIR URA 73 0.011 0.049

Excess return of shop EPPIS URA 73 20.002 0.043

Forecasting variables

Credit spread between prime lending rate and

three-month inter-bank rate CRDSD Datastream 73 0.030 0.012

Excess return of SGX All-share index ESGXR Datastream 73 0.014 0.123

Yield spread between long and short term

government bonds TERMS Datastream 73 0.025 0.018

Unexpected inflation UNINF Datastream 72 0.009 0.007

Yield spread between three-month T-bill and

commercial bill rates YSPRD Datastream 73 0.008 0.012

Table I.

Summary statistics

Common real

estate risk

premia

363

5. Empirical results

5.1. Regressions for excess returns

Regression models for excess returns of six sample assets on five instrumental variables

and a constant term were estimated by a system of equations using the generalized

method of moments (GMM). The regression results were summarized in Table III. The

model of excess government bond return has the best fit estimation with an adjusted

R-square of 65.5 per cent. Variations in excess property stock returns were the least

predictable by the GMM model as reflected by the adjusted R-square of 1.1 per cent.

Durbin-Watson statistics were not significant for all the models except the excess

property stock return model. Persistence autocorrelations were expectedinthe models[1].

The market betas were negative in all models, but the coefficients were only

significant in the government bond, residential and office models at a 10 per cent

significance level. Credit spread, which measures the costs of external debt, was

negatively related in all direct property excess returns. Property investment and

development activities are in general highly leveraged. Therefore, increases in costs of

debt are expected to have negative effects on the performance in the returns of

properties. The same negative effects of commercial bill rate spread over the

three-month Treasury bill rate were also observed on the office and residential models.

The credit risks and spread in commercial bill rates, on the other hand, increase the

excess returns of government bonds.

In term of term structure of government bond yields, the positive result in the excess

government bond model is observed at a 5 per cent significance level. It is, however,

interesting to also note that excess returns of three of the four property submarkets

that include industry, residential and office, track the term structure positively. The

results also show that shop and residential properties were positive hedges against

unexpected inflation at a 10 per cent significance level.

In the excess return model for property stocks, only yield spread was significant but

negatively related to the variations in property stock excess returns at 5 per cent

significance level.

5.2. Tests of common latent risk factors

Based on the same GMM model specifications presented above in Section 5.1, which

include five instrumental variables and an intercept term, we device our tests of

common latent risk factors into three parts involving different portfolios of sample

Correlation

analysis EPTYS EBOND EPPII EPPIO EPPIR EPPIS CRDSD ESGXR TERMS UNINF

EBOND 20.064

EPPII 0.071 0.029

EPPIO 0.164 20.071 0.731

EPPIR 0.430 0.100 0.435 0.537

EPPIS 0.165 20.057 0.541 0.770 0.462

CRDSD 20.094 0.265 20.350 20.383 20.170 20.268

ESGXR 20.012 0.208 20.015 0.032 0.273 0.145 0.168

TERMS 0.053 0.241 0.483 0.348 0.357 0.236 20.555 0.037

UNINF 0.048 20.394 0.101 0.078 20.161 0.000 20.627 20.194 0.388

YSPRD 20.211 0.589 0.144 20.004 20.102 0.003 20.267 0.036 0.488 0.185

Table II.

Summary statistics

JPIF

25,4

364

Intercept

Credit

spread

Unexpected

inflation

Excess market

return

Term

structure

Spread of commercial

bills Adjusted

INT CRDSD UNINF ESGXR TERMS YSPRD R 2

Government

bond Coefficient 0.002 0.065 0.001 20.147 0.037 0.128 0.655

(EBOND) t-statistic 1.809 3.802 0.540 24.420 3.103 9.080

Prob. 0.071 0.000 0.589 0.000 0.002 0.000

Property stock Coefficient 0.096 21.840 0.020 20.609 1.899 25.729 0.011

(EPTYS) t-statistic 0.786 20.721 0.117 20.132 1.213 22.883

Prob. 0.432 0.471 0.907 0.895 0.226 0.004

Industry

property Coefficient 0.030 21.206 20.007 22.026 1.663 20.752 0.246

(EPPII) t-statistic 1.015 22.099 20.138 21.456 4.127 21.382

Prob. 0.311 0.036 0.890 0.146 0.000 0.168

Office Coefficient 0.078 22.316 0.034 22.604 1.184 21.390 0.226

(EPPIO) t-statistic 2.181 23.033 0.803 21.827 2.642 22.130

Prob. 0.030 0.003 0.423 0.069 0.009 0.034

Residential Coefficient 0.052 21.232 0.102 23.376 1.649 21.804 0.403

Property

(EPPIR) t-statistic 2.128 22.575 2.412 23.338 4.769 23.550

Prob. 0.034 0.010 0.016 0.001 0.000 0.000

Shop Coefficient 0.053 21.508 0.065 21.841 0.568 20.765 0.146

(EPPIS) t-statistic 1.479 22.029 1.870 21.620 1.391 21.635

Prob. 0.140 0.043 0.062 0.106 0.165 0.103

Table III.

Regression results of

unrestricted excess

returns of sample assets

Common real

estate risk

premia

365

assets. The first set of the tests involves only all direct property excess returns. We

then add government bond and property stocks in the two subsequent tests. The

incremental approach could allow us to separate the common property market effects

from the bond and equity market effects. The results of the tests of existence of

common latent factors were summarized in Table IV.

The chi-square statistics, which were computed from J-statistics in the GMM model,

were used to test the restriction in equation (5). In Panel (A) of Table IV, the regression

results do not reject the null hypothesis, Ho: [K ¼ 1], which implies that there is at most

one latent systematic risk factor that explains variations in excess returns of the four

direct property assets. Next, we add government bond into the sample portfolio in the

second stage tests. The results in Panel (B) shows that when bond is used as the reference

asset, the chi-square statistic of 49.405 rejected the null hypothesis for [K ¼ 1] latent

factor. However, the chi-square test (x2 ¼ 12:576) failed to reject the hypothesis for

[K ¼ 2] when industry property excess return (EPPII) was included as the second

common reference factor. The results implies that there are at most two common latent

risk factors, which include a common bond market factor and a common direct property

market factor, that are sufficient to predict the excess returns of the sample assets.

In stage three (Table IV – Panel C), property stock was added to the five asset

sample portfolio consisting of four direct property assets and government bond. The

chi-square statistics rejected the null hypotheses for [K ¼ 1] and [K ¼ 2] at 10 per cent

significance level when EBOND and EPTYS were used as reference assets

respectively. The null hypothesis for [K ¼ 3] when EPPII was added as the third

reference asset was not rejected at 5 per cent significance level. The results collectively

imply that three common latent factors that include government bond, property stock

and direct property could predict the variations in the excess returns of the six sample

assets. The results were not different from those in Mei and Lee (1994) in term of

number of common risk factors found in the US markets. However, in Mei and Lee

(1994), they showed that bond, stock (equity) and equity REITs latent factors were the

three significant latent risk factors. Whereas in our model, equity market excess return

is used as an instrumental variable to control for systematic market risks. The three

common factors in our model include bond, property stocks (a close proxy of equity

REITs in the US market) and direct property latent variables.

The above results offer further empirical evidence to reject that property stock

(securitized) and direct property markets were integrated in Singapore markets. At

least, the two property market risk factors were priced significantly different in the

multi-factor latent risk models. The finding that rejects the market integration

hypothesis for the two property markets in Singapore was consistent with the earlier

results of Ong (1995), Liow (1998), Sing and Sng (2003) and Sing (2004). The results

suggest that there was no dilution of diversification efficiency if investors include both

direct property and property stocks in the same portfolio. These two assets capture

significantly different risk premiums in the markets, and variations in excess returns

could be incrementally explained by common latent factors in bond, property stocks

and direct property markets.

6. Conclusion

In this paper, the multi-factor latent variable model was used to test the predictability

of the five asset markets and the significance of common risk premia in bond, property

JPIF

25,4

366

Number of latent

factor (K)

Reference

asset

Excess

return

Property stock

(EPTYS)

Industry

property (EPPII)

Office property

(EPPIO)

Residential

property (EPPIR)

Shop

(EPPIS) Chi-square

Degree of

freedom Probability

Panel (A): All direct property asset (EPPII, EPPIO, EPPIR, EPPIS)

K¼1 EPPII bi 1.043 1.130 0.542 18.607 15 0.232

t-statistic 6.942 6.191 4.239

Prob. 0.000 0.000 0.000

Panel (B): All direct property assets (EPPII, EPPIO, EPPIR, EPPIS) Government Bond (EBOND)

K¼1 EBOND bi 11.583 13.947 14.866 7.472 49.405 20 0.000

t-statistic 3.720 3.580 4.507 2.859

Prob. 0.000 0.000 0.000 0.005

K¼2 EBOND bi 23.005 20.216 20.732 12.576 8 0.127

t-statistic 22.925 20.141 20.877

Prob. 0.004 0.888 0.381

EPPII bi 1.254 1.420 0.635

t-statistic 6.297 5.186 3.652

Prob. 0.000 0.000 0.000

Panel (C): All direct property assets (EPPII, EPPIO, EPPIR, EPPIS) Government Bond (EBOND) Property Stock (EPTYS)

K¼1 EBOND bi 24.298 9.520 11.203 13.458 5.066 44.315 25 0.010

t-statistic 2.514 4.202 3.727 4.795 2.436

Prob. 0.012 0.000 0.000 0.000 0.015

K¼2 EBOND bi 4.210 2.212 4.528 1.863 25.241 16 0.066

t-statistic 1.916 0.882 1.847 1.330

Prob. 0.056 0.378 0.065 0.184

EPTYS Coefficient 0.561 0.704 0.721 0.311

t-statistic 2.222 2.384 2.381 2.363

Prob. 0.027 0.018 0.018 0.019

K¼3 EBOND bi 21.858 0.628 20.593 12.730 9 0.175

t-statistic 21.810 0.470 20.720

Prob. 0.071 0.639 0.472

EPTYS bi 0.113 0.053 20.014

t-statistic 1.875 0.626 20.240

Prob. 0.062 0.532 0.810

EPPII bi 0.915 1.194 0.725

t-statistic 5.116 5.141 4.116

Prob. 0.000 0.000 0.000

Notes: Et ½~ri;t1_ ¼X

K

k¼1

bikX

L

n¼1

uknXnt ¼X

L

n¼1

ainXnt (4)

aij ¼X

K

k¼1

bik ukn (5)

Empty cells indicate that either the excess returns of the assets were not included in the sample or they were used as reference assets in the latent variable tests

Table IV.

Estimation results for

multi-factor latent

variable models

Common real

estate risk

premia

367

stock and four property markets in Singapore. Compared with the results in Mei and

Lee (1994) study, who found bond, stock and equity REIT as three significant common

latent risk factors in the US asset markets, our results show that securitized property

stock (a close proxy of equity REITs) and direct property captured different risk

premia, in addition to common bond risk premium in the model. The results imply that

risk premium in securitized property market could not be replicated by holding direct

property and vice versa. In other words, the two property markets were not perfectly

integrated. They were not perfect substitutes from the portfolio diversification

perspective.

In our GMM framework, five instrumental variables that include credit risk spread,

excess stock market return, unexpected inflation, terms structure between two

government bonds of different terms, and the spread between commercial and

Treasury bill rates. The explanatory variations in the significance of the estimated

coefficients of instrumental variables also pointed to the evidence there was no single

common risk factor that can fully capture the risk premiums in bond, property stocks

and direct property markets. One technical caveat in the empirical specification is that

it is difficult to fully capture all risk information as represented by the GMM model

with five instrumental variables. The omission of other variables is, however,

traded-off against the parsimony of the model specification in this study.

Note

1. Ferson (1990, footnote 18) explained that residual autocorrelations in the GMM estimation

could be caused by the omission of “market” information that is not captured by the

instruments. The results do not imply mis-specification of the conditional means in the

model, and therefore, the model structure does not affect our tests of significance of latent

risk factors for the sample assets in the next section.

References

Bond, S.A., Karolyi, G.A. and Sanders, A.B. (2003), “International real estate returns: a

multifactor multicountry approach”, Real Estate Economics, Vol. 31, pp. 481-500.

Fama, E.F. and Gibbons, M.R. (1984), “A comparison of inflation forecasts”, Journal of Monetary

Economics, Vol. 13, pp. 327-48.

Ferson, W. (1990), “Are the latent variables in time-varying expected returns compensation for

consumption risk?”, Journal of Finance, Vol. 45 No. 2, pp. 397-429.

Fu, Y.M. and Ng, L.K. (2001), “Market efficiency and return statistics: evidence from real estate

and stock markets using a present-value approach”, Real Estate Economics, Vol. 29,

pp. 227-50.

Hansen, L.P. (1982), “Large sample properties of generalized method of moments estimators”,

Econometrica, Vol. 50, pp. 1029-286.

Li, Y. and Wang, K. (1995), “The predictability of REIT returns and market segmentation”,

Journal of Real Estate Research, Vol. 10, pp. 471-82.

Ling, D.C. and Naranjo, A. (1999), “The integration of commercial real estate markets and stock

markets”, Real Estate Economics, Vol. 27 No. 3, pp. 483-515.

Liow, K.H. (1998), “Singapore commercial real estate and property equity markets: close

relations?”, Real Estate Finance, Vol. 41, pp. 603-16.

Liow, K.H. (2001), “The long-term investment performance of Singapore real estate and property

stocks”, Journal of Property Investment & Finance, Vol. 19 No. 2, pp. 156-74.

JPIF

25,4

368

Liu, C. and Mei, J. (1992), “The predictability of returns on equity REITs and their co-movement

with other assets”, Journal of Real Estate Finance and Economics, Vol. 5, pp. 401-18.

Liu, C. and Mei, J. (1998), “The predictability of international real estate markets, exchange rate

risks and diversification consequences”, Real Estate Economics, Vol. 26, pp. 3-39.

Liu, C. and Mei, J. (1999), “A time-varying risk analysis of equity and real estate markets in the

US and Japan”, Pacific Economic Review, Vol. 4, pp. 321-36.

Liu, C.H., Hartzell, D.J., Greig, W. and Grissom, T.V. (1990), “Integration of the real estate market

and the stock market: some preliminary evidence”, Journal of Real Estate Finance &

Economics, Vol. 3, pp. 261-82.

Lizieri, C. and Satchell, S. (1997), “Interactions between property and equity markets: an

investigation of linkages in the United Kingdom 1972-1992”, Journal of Real Estate

Finance & Economics, Vol. 15 No. 1, pp. 11-26.

Mei, J. and Lee, A. (1994), “Is there a real estate factor premium?”, Journal of Real Estate Finance

and Economics, Vol. 9, pp. 113-26.

Mei, J. and Hu, J. (2000), “Conditional risk premiums of Asian real estate stocks”, Journal of Real

Estate Finance and Economics, Vol. 21, pp. 297-313.

Ong, S.E. (1994), “Structural and vector autoregressive approaches to modelling real estate and

property stock prices in Singapore”, Journal of Property Finance, Vol. 5 No. 4, pp. 4-18.

Ong, S.E. (1995), “Singapore real estate and property stocks – a cointegration test”, Journal of

Property Research, Vol. 12, pp. 29-39.

Sing, T.F. (2004), “Common risk factors and risk premia in direct and securitized real estate

markets”, Journal of Property Research, Vol. 21 No. 3, pp. 189-208.

Sing, T.F. and Sng, S.B.S. (2003), “Conditional variance tests of integration between direct and

indirect real estate markets”, Journal of Property Investment & Finance, Vol. 21 No. 4,

pp. 366-82.

Wilson, P., Okunev, J. and Ta, G. (1996), “Are real estate and securities markets integrated? Some

Australian evidence”, Journal of Property Valuation & Investment, Vol. 14 No. 5, pp. 7-24.

Corresponding author

Tien Foo Sing can be contacted at: rststf@nus.edu.sg

Common real

estate risk

premia

369

To purchase reprints of this article please e-mail: reprints@emeraldinsight.com

Or visit our web site for further details: www.emeraldinsight.com/reprints