INTERDEPENDENCE AND CAUSAL LINKAGES OF GLOBAL STOCK AND MAJOR REAL ESTATE MARKETS Kim Hiang LIOW, Department of Real Estate, National University of Singapore Associate Professor (Dr) Kim Hiang LIOW Department of Real Estate National University of Singapore 4 Architecture Drive Singapore 117566 Tel: (65)65163420 Fax: (65)67748684 Email: rstlkh@nus.edu.sg 17 April 2006 1 INTERDEPENDENCE AND CAUSAL LINKAGES OF GLOBAL STOCK AND MAJOR REAL ESTATE MARKETS Abstract This study assesses the interdependence and causal linkages across six major public real estate markets and global stock market since 1990. Using weekly returns of FTSE EPRA/NAREIT series, a multivariate EGARCH model with constant correlation coefficients and causality-in- variance tests, we find international real estate markets are generally correlated in returns and volatilities contemporaneously and with lags. Causality in mean and variance are detected in some real estate markets and the MSCI global stock market, with dynamic adjustments take place on average between one and two weeks Using the vector autoregression technique, we find that different markets’ volatility response to shocks from different sources seems to be diverse. Whilst the conditional volatility of the MSCI global stock market appears to affect the volatilities of the real estate markets significantly, the negligible cross-impact of some real estate markets support the notion that international real estate markets are still fairly segmented. Finally, by implementing the various statistical tests with two shorter sub-sample periods (pre- and postAsian financial crisis periods), we find that the causal interactions among the major real estate /global stock markets under crisis circumstances differ from those in stable times. In particular, significant change in the patterns and significances of the market interdependence over the two sub-periods have been detected as both the stock and real estate markets experienced periods of stability and volatility in the 90’s. This knowledge would help fund managers in managing their exposure in international real estate markets, after controlling for the global stock market influence, and constructing better asset allocation models. 1. INTRODUCTION This study examines the interrelationship among the public real estate markets of Australia, Japan, Hong Kong and Singapore as well as the relationship between each of these four markets and the larger and more developed markets of the US and UK and global stock market since 1990s. Prior literature has studied real estate correlation and diversification issues extensively in international stock markets and some causality and interaction issues related to real estate data to a lesser extent. This paper distinguishes itself from previous real estate studies in at least three aspects. First this paper is unique in that it investigates simultaneously the interdependence of six major real estate markets and the global stock market. The focus on the six real estate markets is of significant interest to the world investors and policy markets. Furthermore, as these sample countries have leading capital markets in the Asian and world economy, the inclusion of the global stock market allows the relationship between pairs of real estate variable to be more fully specified when a whole system of seven series is studied as a dynamic system. Second, in contrast to many real estate studies that examine interdependence in returns only, we also focus on the relationship in return volatilities. The knowledge of how volatility between two real estate markets is important in that the second moment or variance 2 is directly linked to information flow as well as providing insights concerning the characteristics and dynamics of real estate asset prices. Third, rather than estimating separate univariate GARCH models, we employ multivariate GARCH methodology to model the volatility of the seven markets simultaneously and ensure that they are estimated over a common range. Through the use of a MGARCH model instead of estimating seven separate univariate GARCH models, we are able to capture the co-movement of volatility across the seven markets as well as evaluating the causal relationships in returns and return volatilities between any pairs of market throughout the study. The first part looks at correlations and causal relationships in returns and return volatilities. Specifically, causality tests and results provide investors with additional insights into how and when information is impacted on different real estate markets and the global stock market and design more objective pricing models with the appropriate lag structure. Regulators may also use the causality results to identify the foreign market(s) that drive(s) movements in the domestic real estate market and take appropriate corrective measures. In this paper we employ the causality-in-variance test developed by Cheung and Ng (1996) to uncover causal relations in returns and return volatilities with regard to the direction of causality as well as the number of leads/lags involved. So far, no research has examined the causality in variance issue in international real estate investments.1 The second part examines the dynamic responses of each market to shocks in volatilities. As correlation and causality analyzes do not reveal anything about the transmission dynamics in the presence of unexpected shocks, a vector autoregressive (VAR) analysis is employed to quantify the temporal precedence in real estate and stock market return volatilities. Finally, we are also interested in finding out if inter-market relationships vary significantly as the markets go into and out of financial crisis (e.g. 1997-98 Asian financial crisis). By implementing the various tests with two shorter sub-sample periods (pre- and post-crisis periods), we examine if the pattern and significance of the market interdependence has changed over the two sub-periods as both the stock and real estate markets experienced periods of stability and volatility.2 Our study is organized as follows. Section 2 contains a selective literature review. This is followed by an explanation of the research data and methodology in Sections 3 and 4 respectively. 3 Section 5 discusses the empirical results and implications. The last section concludes the study with a summary of the main results. 2. RELATED LITERATURE Of late, there has been increasing interest in the causation in conditional variance across stock market price movements. Such studies are considered in the wider context of the dynamic linkages between national stock markets.3 The study of causation pattern in variance provides useful insights into the relationship between information flow and volatility and the characteristics and dynamics of financial asset prices (Cheung and Ng, 1996). Together with evidence regarding the causality in mean, the overall results will yield a more complete picture regarding the dynamics of interactions among the financial markets involved. The policy implication regarding causality would clearly be important, as the evidence would suggest that financial market policies of one country should not be implemented without taking into account the impacts on other markets (say, in Asian or European regional real estate markets), and vice versa. From the institutional investors’ perspective, there would be little diversification benefits in the short-term if two markets are causally linked in both returns and volatilities. Cheung and Ng (1996) develop a test for causality-in-variance (CIV) that allow researchers to assess how a market evaluates and assimilates new information, and examine the temporal dynamics of return volatilities across national stock markets. The CCF testing procedure does not require modeling of the dynamics of the interaction of the series involved. Instead it is based on the residual cross-correlation function (CCF) and is robust to distributional assumptions. The null hypothesis of no causality in variance is tested through asymptotic normal t statistic. Mathematically, given two stationary time series X t and Yt , the first task is to impose a conditional mean and variance specification for the time series. Then the squared standardized residuals ( ε t and ξ t ) are estimated 2 2 for series X t and Yt , namely: U t = {( X t − μ x ,i ) 2 / hx ,t } = ε t 2 4 Vt = {(Yt − μ y ,t ) 2 / h y ,t ) = ξ t Where 2 μ and h are the conditional mean and variance of the time series respectively. Assume further rUV (k ) and rεξ (k ) are the sample cross-correlations of the squared standardized residuals and standardized residuals at lag k respectively. The CCF testing procedure is employed to test the hypothesis of no causality in variance against the alternative hypothesis of causality at lag k. Given that T is the number of time series observations, the CCF-statistic is give by: CCF − statistic = T * rUV (k ) Similarly, to test the null hypothesis of no causality in mean against the alternative hypothesis of causality at lag k, the CCF statistic is given by: CCF − statistic = T * rεξ (k ) To-date, the CCF testing methodology has been applied in some studies on stock markets, foreign currency markets and interest rate markets. Hu et al (1997) examine the CIV and volatility spillover effects among two developed markets (US, Japan) and four emerging markets in the South China Growth Triangular (Hong Kong, Taiwan, Shanghai and Shenzhen). Employing the CCF methodology, they find significant relationships among the variances of the markets and that there is a feedback relationship between the Hong Kong and the US stock markets. In a study that covers five Pacific-Rim stock markets and the US stock market returns with the CCF testing procedure plus a TARCH model, a MGARCH specification and VAR analyzes, Tay and Zhu (2000) find that in most instances there is bi-directional causality (both) in mean and variance among the markets. Furthermore, the dynamic adjustment of the market return volatilities can take as much longer time than expected. In a foreign exchange study, Kanas and Kouretas (2002) examine the issue of mean and variance causality across four Latin American markets over 1976 -1993. Using an EGARCH-M model and the CIV methodology, they find substantial evidence of causality in both mean and variance with the causality in mean largely driven by the causality in variance. Alaganar and Bhar (2003) employ the 5 CCF testing procedure to examine the linkages between the financial sector returns and interest rates of the G7 countries. They demonstrate that the CIV procedure is able to model the direction and lags in information flow between two time series. They are able to find feedback effects both at the mean and the volatility level. Caporale et al (2002) provide some empirical evidence on the casual relationship between stock prices and exchange rates volatility in four East Asian countries using daily data from January 1987 to March 2000 by appealing to the MGARCH methodology with BEKK representation. They also analyze the effects of the 1997 East Asia crisis by splitting the sample in two, before and after the onset of the crisis. Finally, Fujii (2005) reveals significant causal linkages in mean and variance both within the Asia and Latin America regions and across the two regions. His rolling CCF results further indicate that the significance of the causality varies considerably over time and causal linkages tend to strengthen particularly at the time of major financial crisis. As for real estate studies, key findings of previous work are that diversification into international real estate provides a way of achieving an efficient portfolio by a reduction in the variance of returns and an enhancement of portfolio performance. Additionally, diversification across different markets and regions can result in more efficient portfolios. For example, Eichholtz (1995) compares the diversification benefits derived from international real estate investments to those of international stock and bond investments and concludes that property shares are less strongly correlated than common stock and bond returns across national boundaries. Similarly, several studies on emerging market securities also conclude that international real estate securities offer diversification benefits for a real estate portfolio. Using quarterly and monthly data separately in two studies, Lu and Mei (1999) and Hu and Mei (1999) examine the return-generating process of property indices in ten emerging markets (Argentina, China, Hong Kong, Indonesia, Malaysia, Philippines, Peru, Singapore, Thailand and Turkey), and find that property indices are more volatile than their respective market indices and US NAREIT index. Additional diversification benefits to invest in emerging market property securities beyond that associated with international stocks are found, but correlations are higher during times of high volatility. On the contrary, research studies on real estate market volatility and its transmission across international real estate markets are relatively limited and 6 none of them has examined the issue of causality-in-variance since 1990.. The lack of research in this area is a peculiar oversight, given the increase in indirect global property investment over the last decade and it is these market interdependence and causal linkage issues that this study can make a meaningful contribution.. Garvey et al (1991) examine the linkages between securitized real estate markets of Australia, Japan, Singapore and Hong Kong on both short and long-term basis. The short term analysis covers linkages in both mean and variance of the return series using Granger causality and GARCH models that incorporate volatility spillovers. They find little evidence of co-movements or influences between the markets on a bivariate basis. Stevenson (2002) investigates the influences on volatility in the US REITs. He finds that the volatility in equity REITs has significant influence on other sub-sectors of the REIT market. More recently, Liow et al (2005) examine the long-term and short-run relationships in mean and variance spillover effects across four Asian property stock markets (Japan, Hong Kong, Singapore and Malaysia) and four European real estate stock markets (the UK, France, Germany and Italy). Their evidence of minimal cointegration, weak mean transmission and little cross-volatility spillovers among the Asian and European real estate stock markets imply that diversification opportunities exists among the regional public real estate markets. However, they do not examine the issue of causality in mean and variance across the real estate markets. Additionally, Liow (2006) focuses on the dynamics of volatility persistence and systematic risk in Asian-Pacific real estate markets. Again, the issues of interdependence in term of the causal links among the returns and variances among the real estate markets were not investigated. Recently, Michayluk et al. (2006) construct synchronously priced indices of securitized property listed on NYSE and LSE and then examine dynamic information flows between the two markets. They show that the real estate markets in these two countries experience significant interaction on a daily basis, and the positive and negative news impact the markets differently. 3. RESEARCH DATA As in many previous academic real estate studies, we use returns on real estate stocks to proxy for real estate asset performance. This choice is mainly justified by the availability of longer 7 time series data and higher frequency data (such as monthly and weekly) for real estate stocks. Whilst the adequacy of this proxy has been extensively debated amongst real estate practitioners and researchers, it remains the only substantive “real estate” series appropriate for any rigorous statistical analysis. We include six major public real estate markets and a global stock market for a US-based investor. Apart from the popularly known US and UK real estate markets that have different institutional and market structures from the developing economies, the remaining four are Asia-Pacific markets of Singapore, Hong Kong, Japan and Australia. These six countries are (major) trading partners with one another, and thus that these markets may be (strongly) integrated. Japan is the premier economy in the Asia region, and therefore a potential driver for the regional economy. It has a long history of public real estate. Other countries like Australia, Hong Kong and Singapore have track records of listed real estate companies that play a relatively important role in general stock market indexes. In particular, Australian securitized real estate sector is a leading player in global real estate. The UK real estate market plays a key role in the European property market. Moreover, the six markets represent about 91 percent of the global securitized real estate market and have the world’s most significant listed real estate markets in the respective regions (UBS Investment Bank, 2003). The real estate data are weekly FTSE EPRA/NAREIT total return indexes maintained by the European Public Real Estate Association (EPRA). These global real estate series are designed to track the performance of listed real estate companies and REITs worldwide. The series act as a performance measure of the overall market (www.epra.com). All six real estate index series are extracted from Datastream and the sample period is from 29 December 1989 till 10 February 2006 the longest period for which all weekly data indexes are available.4 Weekly real estate stock returns (R) are obtained by taking the logarithmic difference of the stock index (P) times 100. That is, R t = 100 * (log P t – log P t – 1). All data are expressed in US dollars. Figure 1 displays the real estate index movement over the study period. The world stock market equity index is compiled by MSCI and is also obtained from the Datastream. The MSCI world market index is widely used by international fund managers for asset allocation decisions and performance measurement as well as used by researchers in academic studies. 8 (Figure 1 here) To provide a general understanding of the nature of each market return, Table 1 contains sample characteristics of the weekly real estate stock returns and the global stock market returns over the full sample period from 5 January 1990 – 10 February 2006. The descriptive statistics include mean, standard deviation, coefficient of variation and the measures of skewness, kurtosis and normality. (Table 1 here) Of the real estate index series, the US has the highest average weekly return (0.293 percent) and is followed by Australia (0.258 percent); while Japan has the lowest average mean (0.002 percent) Judging from the sample standard deviations, Singapore, Japan and Hong Kong are characterized by a higher unconditional volatility (range between 5.14 percent and 4.40 percent) compared to the USA 1.82 percent), Australia (2.04 percent) and UK (2.40 percent). Consequently, the US, Australia and UK markets also have the three lowest coefficients of variation (i.e. risk per unit of return). The MSCI world stock index has an average weekly mean return and standard deviation of 0.098 percent and 1.91 percent, respectively. Except for Japan and the UK, all four other markets are negatively skewed (between -0.275 for Australia and -0.430 for the US). These negative skewness coefficients are all statistically significant at the one percent level. The kurtosis measure is more than three in all weekly return series (between 3.506 for Australia and 21.90 for Singapore). The distribution of returns has thus fat tails compared with the normal distribution. Finally, the MSCI world market index is negatively skewed (-0.226), exhibit significant kurtosis (5.032) and fail the test of normality at the one-percent level (Jacque Bera values of 151.81). The presence of inter-temporal dependencies in the returns and squared returns are tested by using Ljung-Box portmanteau test (LB). The LB statistic tests the hypothesis that autocorrelations up to the nth lags are jointly statistically significant. The calculated LB statistics, given by LB-Q (5), LBQ (20), LB-Q2 (5) and LB-Q2 (20) are also reported in Table 1. As can be seen, except for the MSCI global market index, the hypothesis of linear independence in mean return is firmly rejected at the ten percent level or better for Hong Kong, Singapore and Japan and partially rejected for the US, UK and 9 Australia, implying that the conditional mean of the distribution of real estate stock returns is probably a function of either past returns or past residuals. On the contrary, independence of the squared return series is partially accepted at the ten percent level for the UK only. Except for the UK, The LB statistics for the squared returns are several times higher than that of returns themselves in other five real estate markets and the MSCI world index. The implication is that higher-moment dependencies are much more pronounced. Finally, the results for the augmented Dickey Fuller (ADF) and the Phillips and Peron (PP) tests presented in Table 1 depict that a unit root is present in the logarithm of all real estate stock indexes and the MSCI world index. On the contrary, the hypothesis of a unit root is rejected for each of the return series. Consequently, further investigation of the short-term dynamics of the real estate markets and the global stock market requires first differencing. 4. RESEARCH METHODOLOGY Our empirical procedures contain four major steps. They are briefly explained below. First, we conduct a usual preliminary examination of the correlation in raw returns and return volatilities (proxied by mean squared deviation). The correlation structure provides a rough measure of interdependence and is probably the most important feature from the point of view of investors and portfolio mangers. Pearson Correlation analysis has been frequently used in the initial literature on international stock market linkages. Second, GARCH models are used to study the stochastic behavior of return series and, in particular, to explain the behavior of volatility over time (Bollerslev et al. 1992). In addition, Nelson (1991)’s exponential GARCH (E-GARCH) model specifies the conditional second moments of returns to allow for asymmetric effects of market news on the volatility function. These univariate GARCH or EGARCH models can be extended into a multivariate framework (MGARCH or MEGARCH). This extension allows all variables to be considered simultaneously over a common range, i.e. volatility in all markets is analyzed as a whole system. One key advantage of the MGARCH (/MEGARCH) specification is that they permit time-varying conditional covariances as well as variances; thus allow for possible interactions within the conditional mean and variance of two or 10 more financial series. In what follows, we estimate a MEGARCH model suggested by Bollerslev (1990) that assume the correlations in the conditional volatilities are constant. This constant correlation (CC) specification has generally a well-behaved likelihood function as well as limiting the number of estimation of coefficients to a workable level. The model can be estimated by the maximum likelihood method. The AR (p) - MEGARCH (1, 1) model is represented by the following equations: 6 Ri ,t = β i , 0 + ∑ β i , j R j ,t −1 + ξ iσ i2,t + ε i ,t for i, j = 1,2,…7, (1) j =1 ⎧ ⎫ 6 σ i2,t = exp⎨α i , 0 + ∑ α i , j f j ( z j ,t −1 ) + γ i ln(σ i2,t −1 )⎬ j =1 ⎩ f j ( z j ,t −1 ) = z j ,t −1 − E z j ,t −1 + δ j z j ,t −1 ( ( ) ) σ i , j ,t = ρ i , j σ i ,t σ j ,t ⎭ for i, j = 1,2,…7, (2) for i, j = 1,2,…7, (3) for i, j = 1,2,…7 and i ≠ j . (4) Equation (1) describes the returns of the seven markets as an autoregression (AR) each where the conditional mean in each market is a function of pass own returns as well as a time dummy D 97 that takes a value of one for the period July 1997–August 1998 and zero otherwise. The intent is to control for regime swifts in many Asian real estate markets following the eruption of Asian financial crisis in July 1997 (Liow et al. 2005). Finally, while the 1997 Asian financial crisis may have had effects reaching beyond the Asian regions, that issue is beyond the scope of this study. Hence the dummy is not included in the mean equations for UK and USA. Equation (2) explains the EGARCH representation of the variance term ε t . According to the EGARCH representation, the conditional variance of the returns in each market is an exponential function of past own, cross-market standardized innovations and past own conditional variance. The persistence of volatility is measured by γ i . The unconditional variance is finite if γ i < 1 . If γ i = 1 , then the unconditional variance does not exist, and the conditional variance follows an integrated process of order one. The particular function form of f j ( z j ,t −1 ) is given in Equation (3), which captures the ARCH effect, and is asymmetric function of past standardized innovations. The term 11 ( ) z j ,t −1 − E z j ,t −1 measures the magnitude effect. If the magnitude of z j ,t −1 is greater than its ( ) expected value, E z j ,t −1 , the impact of z j ,t −1 on positive. Finally, the term σ i2,t will be greatly positive, providing that α i, j is δ j z j ,t −1 measures the asymmetric volatility transmission mechanism (Koutmos, 1996) Equation (4) provides the conditional covariance specification, which captures the contemporaneous relationship between the returns of the seven markets. This specification implies that the covariance of market i and j to be proportional to the product of their standard deviations. This assumption greatly simplifies estimation of the model and it is a plausible one for many applications (Bollerslev et al., 1992). The coefficient ρ i, j is the cross-market correlation coefficient of the standardized residuals between two markets. Statistically, the significant estimates of ρ i, j indicates that time-varying volatilities across market i and j are correlated over time. The BFGS algorithm provided by RATs is used to produce the maximum likelihood estimates and their corresponding asymptotic standard errors with a Student-t fat tailed distribution. Finally, the Ljung-Box Q statistic is computed to test the null hypothesis the MEGARCH model is correctly specified, or equivalently, that the noise terms are random. Third, once the appropriate AR(p)-MEGARCH (1, 1) model is identified, we employ the causality-in-variance test developed by Cheung and Ng (1996) to detect causal relations and identify patterns of causation in the first and second moments respectively. Here, the sample cross-correlations (CCF) of the resulting standardized residuals and squared residuals standardized at k-lags are determined. The cross-correlation function (CCF) of these standardized residuals and squaredstandardized residuals are used to test the null hypothesis of no causality in mean and variance respectively. Finally, we employ the conventional VAR method5 to conduct a dynamic analysis of the conditional volatilities to shocks originating from various markets through impulse response and variance decomposition analyses. Here, a kth VAR of the conditional volatilities can be written as: 12 ht = c + Φ 1 ht −1 + Φ 2 ht − 2 + .... + Φ k ht − k + vt …………..(5) where c denotes a (7x1) vectors of constant, Φ j is a (7x7) matrix of autoregressive coefficients for j =1,2…..p, and vt ~ i.i.d. N (0, Ω) where Ω is a (7x7) symmetric positive definite matrix, ht is a (7x1) vector of conditional volatilities derived from the MEGARCH estimation. Through impulse response functions, we are able to trace the effect of one standard deviation shocks in the conditional variances of each foreign market has upon the domestic market. The normalized impulse response is the ratio of moving average coefficients and the respective standard errors and will be reported in this study. The normalized coefficients illustrate how long and to what extent each market’ volatility reacts to unanticipated changes in the other markets. We use the generalized impulse response function (Pesaran and Shin, 1998) to decompose the error components. One key advantage of this method is it is not dependent on the ordering of the variables in the VAR system, thus doing away with the need to first conduct Granger causality analysis determine an appropriate causal order of the variables in the VAR system.6 Through variance decomposition, we would be able to examine the composition of the conditional volatilities [h (t) vectors from the MEGARCH estimation] with regard to the extent they are affected by own shocks as well as by shocks from other markets. The results will thus indicate the relative importance of the various markets in causing the volatility fluctuations of that market. 5. RESULTS 5.1 Correlations in Returns and Volatilities The sample correlations of the real estate stock returns and return volatilities are first estimated. They provide a rough measure of market interdependence. The low correlation between returns of international real estate markets has been regarded as an incentive to hold internationally diversified real estate portfolios. In Table 2, the correlation matrices are calculated for the return (first row) and mean squared deviation of the returns (second row) (used to represent return volatilities). 13 This is because international real estate markets might be related through their returns, or their return volatilities or both. (Table 2 here) As the numbers indicate, correlations between the returns of all real estate markets are between low and moderate ranges. Here we see that the highest and lowest correlation in returns, respectively, are 0.581 (between Singapore and Hong Kong) and 0.096 (between Japan and the US), and only one other pair’s correlation coefficient is slightly above 0.30 (0.303, between Australia and the UK). The bivariate return correlations between the MSCI world stock index and all real estate markets are higher; but on average still moderate; ranging between 0.336 (MSCI and Australia) and 0.458 (MSCI and the US). Next, the correlation coefficients for the mean squared deviations (return volatilities) are comparatively lower. Here we see that only three pairs of the 21 markets’ pair-wise correlation coefficients are higher in return volatilities than in returns (i.e. Singapore and HK, Japan and the US and UK and the US). Again, Singapore and Hong Kong report the highest correlation coefficient of 0.762. On the contrary, Singapore and the UK are not correlated at all (correlation = 0.022) even though the UK is the largest foreign investor by stock of total foreign direct investment.7 Overall, our findings are generally in agreement with what appears in the literature. We are able to derive a conclusion that similar to stock markets, international real estate markets show small to moderate correlation in returns, but some markets are not independent because they are related through their second moments. 5.2 M-EGARCH Estimation Results Maximum likelihood estimates with conditional t–distribution of the multivariate specification,8 those of the Constant Conditional Correlation AR (p) - EGARCH model for the full sample period are presented in Table 3.9 (Table 3 here) The log-likelihood value of 13506.34 is large enough to suggest that the estimated model is able to capture the temporal dependence of volatility reasonably well. All seven index returns display 14 a significant volatility persistence coefficient (between 0.928 for Australia and 0.995 for the UK). Consistent with expectation, we find a significant and negative Asian financial crisis coefficient (D97) each for Hong Kong, Japan and Singapore real estate markets. Finally, with some minor exceptions, the estimated ARCH values, Ljung-Box statistics for the standardized and squared standardized residuals indicate the AR (p) - MEGARCH (1, 1) model has been adequately specified. The constant conditional correlation coefficients are given by the sample correlation matrix and are set out in Table 3, under the heading CCC. Just as we have found previously in Table 2, although all conditional volatility correlation coefficients are higher than the unconditional ones as well as highly significant, the markets are only moderately correlated in variances with each other. Among the real estate markets, Hong Kong and Singapore remains as the most strongly correlated pair with a correlation of 0.472, and is followed by the UK-Hong Kong pair with a correlation coefficient of 0.305. All other real estate market pair-wise correlations range between 0.102 and 0.301. Additionally, Hong Kong, Singapore, Japan and Australia real estate markets are more correlated with the UK than with the US. However, there is one interesting twist. Unlike the insignificant unconditional correlation found previously (in Table 2) between Singapore and the UK, the conditional correlation in variance between these two markets is now 0.282, which is statistically significant at the one percent level. Finally, the six pairwise correlations between the MSCI world market and real estate markets range between 0.357 and 0.452, which are higher than those of real estate market, except between Hong Kong and Singapore. Table 4 presents the MEGARCH model estimates for the pre-and post-Asian crisis time periods. 10 Focusing our attention on the correlation estimates in volatilities, the conditional contemporaneous correlation relationships are more or less similar to those reported for the full sample period (shown in Table 3), with 20 pairs each of correlation coefficients statistically significant at the one percent level for the two sub-periods. Further comparisons reveal that the post-crisis period witnesses a stronger volatility correlation relationship for 16 of the 21 pairs, and that the numbers of pairs of real estate markets that have a correlation coefficients larger than 0.30 are 1 (pre-crisis) and 4 (post-crisis) respectively. This could have been the results of the contagion effects of the financial 15 crisis. Conversely, the post-crisis volatility correlations between US-Hong Kong, UK-Japan, USJapan, MSCI – Japan and MSCI– UK have weakened (considerably) compared with the pre-crisis period. Among the real estate markets, the strongest volatility correlation for the pre-crisis and postcrisis period occurred between Hong Kong and Singapore (0.381for pre-crisis and 0.531 for postcrisis). (Table 4 here) 5.3 Causality The standardized residuals and squared standardized residuals for each market are extracted from the MGARCH models to implement the CCF tests for cross-market causality in mean and variances. The results are reported in Tables 5(a) and 5(b) for the full sample period. These reported statistics are for causality at a specific lag k. Lags are measured in weeks, which range from -10 to + 10. The test results are organized by market pairs and lag order. For a pair of markets, a significant test statistic with lag k < 0 should be interpreted that the return/variance of the first market causes that of the second market in mean/variance with a k–period lag. 11 Similarly, if the test statistic is significant with k > 0, then the second market’s return / variance is said to cause the first market in mean / variance with a kth lag. A significant test statistic with k = 0 indicates contemporaneous causality. Since these CCF tests for all markets will generate a voluminous amount of information, we provide in Table 6 a combined summary of the causality patterns among the returns and return volatilities. Out of the 420 t-statistics that indicate lead-lag relationship (from Tables 5a and 5b), the results show that there are 44 (10.48%) and 50 (11.9%) significant cases of causality in mean (CIM) and causality in variance (CIV). Although the majority of lead-lag linkages happened at k = 1, statistical significant causal effects can arise from observations as long as 10 weeks ago for some market pairs. Additional evidence reveals that in five cases there are bi-lateral (i.e. feedback) CIM and bi-directional CIV among the returns (AU/HK, AU/MSCI, HK/SG, HK/MSCI and UK/US). In another three cases of CIV where the contemporaneous correlations are weak and not statistically 16 significant (JP/SG, JP/UK and JP/US), the correlations at other lags are statistically significant. For example, the t-statistic for the contemporaneous correlation in variance between the real estate markets for Japan and Singapore is merely 1.102, which is statistically insignificant; but the t-statistics for correlations are, respectively, 2.141 (lag 1), 2.853 (lag3), 2.497 (lag 7) and 5.589 (lag 8) and all are statistically significant at the one percent level. The implication is that one-week and three-weeks lagged variances of the Japan’s real estate returns causes the variances of Singapore’s real estate returns, and longer lags of variances (weeks 7 & 8) of the Singapore’s returns causes the variances of Japanese returns. Interestingly, although Singapore and the USA markets are moderately correlated in their contemporaneous returns and variances, both markets are not correlated at all through lead-lag linkages. Another interesting observation is that our results support significant mean transmission from Hong Kong into the UK, Singapore into Japan and Singapore into the USA despite the UK, Japan and the USA being dominant real estate markets. Hence it is plausible for mean return to be transmitted from the smaller market to the dominant market, a finding which appears to contra most of the previous stock market research (e.g. Janakiramanan and Lamba, 1998). Finally, the MSCI global market has strong degree of interdependence in returns and volatilities with all real estate markets. The most pronounced relationships are between MSCI and Australia and between MSCI and Hong Kong; both pairs reported both feedback CIM and CIV in their returns / volatilities. (Tables 5a, 5b and 6 here) To examine the evolution of cross-market linkages over time and the influence of 1997-98 Asian financial crisis on the relationships, Tables 7(a)-7(b) and 8(a)-8(b) presents the CCF t-statistics for the pre-crisis and post-crisis periods respectively. Some useful comparisons regarding the significances and patterns of CIM and CIV between the pre-crisis and post-crisis periods are presented in Tables 9, 10(a) and 10(b). As the causal structures between all market-pairs are diverse and thus difficult to generalize, we will attempt to make few general observations only. Whilst all markets maintain their significant relationships in the contemporaneous returns in both periods (20 out of 21 pairs), the number of significant contemporaneous volatility pairs increases considerably from 7 pairs in the pre-crisis period to 18 pairs in the post-crisis period. Additionally, 17 Figures 2 and 3 indicate that the contemporaneous correlation coefficients in returns (variances) are significantly higher for 15 pairs (20 pairs) in the post-crisis period. It is thus evident that the contemporaneous linkages in returns and return volatilities across the real estate markets and between each real estate market and the global stock market have strengthened considerably in the post-crisis period due possibly to contagion effects of the financial crisis, a finding which is consistent with those reported in Table 4 and prior stock market studies. [Tables 7(a), 7(b), 8(a), 8(b), 9, 10(a) and 10(b) here] (Figures 2 and 3 here) Another insightful aspect of our results is that the lead-lag linkages among the markets appear to be unstable over the two sub-periods. In Table 9 where three categories of causality are analyzed (i.e. bilateral causality, one-way causality and no causality at all) the post-crisis period witnesses changes in the CIM relationship for 14 pairs (66.7%); the corresponding number is 13 (61.9%) for the CIV relationship. Only four market pairs (JP/US, SG/UK, SG/US and SG/MSCI)’ causal linkages in both returns and return volatilities remain stable throughout the two periods. This finding is similar to some previous stock market studies indicating that the 1997-98 Asian financial crisis has brought changes to the inter-relations of stock markets (and in our case, real estate markets) over time. 5.4 Dynamic responses of the conditional volatilities A VAR moving average system describes the dynamic interdependence of the seven markets included in the model. In the present context, the seven conditional variance series (Figure 4) obtained from the MEGARCH (1, 1) estimation for the full period are used for a VAR model that performs dynamic analyses using impulse response and variance decomposition techniques. The response patterns are simulated by introducing one standard deviation shocks to each of the markets in the system over different time horizons. The decomposition of variance of forecast errors of the conditional volatility of the given market indicates the relative importance of the foreign markets’ shocks in causing the fluctuations in the conditional volatility of the subject market. (Figure 4 here) 18 Table 11 shows the parameter estimates obtained from fitting a VAR model for the full sample period. Based on the maximum log-likelihood and minimum AIC criteria, we determine the lag order of the VAR model to be two.12 (Table 11 here) The variance decomposition of the conditional volatilities for the full sample period for 1week, 4-week, 8-week, 12-week, 16-week, 20-week and 24-week horizons are reported in Table 12. First, the majority of the global stock market volatility is explained by shocks originated in its own market. At the horizon of 24 weeks, the proportion of its own volatility is over 85 percent with the Japanese real estate market explain only about 6.6 percent of the volatility variation in the global stock market. In contrast, the rest of the real estate markets have very weak or negligible influences over the MSCI global stock market volatility. Second, Australian real estate market is not affected by the other real estate markets except for Singapore and to a lesser degree, the MSCI world stock market. Whilst the MSCI world stock market maintains its influence of between 2.3 and 2.92 percent over a 24 week horizon, Singapore real estate market shocks starts to gradually increase the volatility in the Australian market from the eighth week onward. Its impact reaches about 5.5% in 24 weeks. (Table 12 here) In the case of Hong Kong, the MSCI global stock market again has the largest lagged impacts on Hong Kong real estate market volatilities. Its impact starts to explain just over 10 percent of HK return volatility in three weeks and reaches about 11.6% in 24 weeks. The Australian real estate market is the only other market that has a weak influence over time on HK real estate market as well. In contrast, the rest of real estate markets (including Singapore) have very minimal influence on the return volatilities for HK real estate market. Fourth, results for Japan shows that the MSCI global stock market, Hong Kong and Singapore real estate markets have some lagged impacts on the Japanese real estate market volatilities. The impacts occur very gently and in 24 weeks, the proportion of the variation in the Japanese market explained by the global stock market, Hong Kong and 19 Singapore real estate markets are, respectively, about 8.7 percent, 6.5 percent and 6.3 percent; with the remaining 3.2% market volatilities explained by the Australia, UK and US real estate markets. The Singapore real estate market, however, is substantially affected by Hong Kong real estate market and the impact occurs fairly quickly. In the first two weeks, HK market volatility innovation explains about 35 percent of the variations of Singapore real estate market volatility. This influence increases to about 46.7% in 24 weeks, even exceeds the proportion of shocks to its own market (37.1%). At this horizon, the MSCI global stock market and Australian real estate markets contribute about 8.8 percent and 4.3 percent of volatility innovation, respectively, to the volatility of the Singapore real estate market. In contrast, the rest of the real estate markets have very minimal influence on the Singapore real estate market volatility Finally, the MSCI global stock market has the largest lagged impacts on the UK and USA real estate market volatilities. In the case of the UK, the MSCI impact starts with about 5.0 percent in week 1 and gradually increases to 10.0 percent in 24 weeks. At this horizon, about 81.2% of the UK market volatility is explained by shocks originated in its own market, with another 4.7 percent and 2.4 percent, respectively, contributed by the Japanese and Hong Kong real estate markets. In contrast, the remaining three markets (Australia, US and Singapore) explain less than one percent to the UK market volatility. For the USA real estate market, the MSCI global stock market accounts for 13.1 percent of the variation in the USA market in 24 weeks. The Australian real estate market also contributes 12.2 percent of the variation in the USA market as well. To sum up, the variance decompositions have revealed several interesting evidence. First, among the markets included in the VAR system, most of the short-term volatilities for the MSCI global stock market and Australian real estate markets are explained by the shocks to their respective own markets. Hence they are largely exogenous as their own innovations account for most of their error variances. On the contrary, a significant portion (between 19 and 62 percents) of the Hong Kong, Japanese, Singapore, UK and USA return volatilities is explained by the lagged impacts of shocks from other markets. Second, the volatilities of the six major real estate markets are affected by the global stock market. The MSCI global stock market appears to the main “exporter” of shocks to other 20 real estate markets and is only affected by the Australian, UK and USA real estate markets to a much lesser degree. Third, among the six real estate markets, different markets’ response to shocks from different sources appears to be diverse. There are many instances of lead-lag linkages in volatility between any two markets, and is consistent with the causality results reported earlier. Singapore is the most endogenous market with over 60 percent of its forecast error (in 24 weeks) explained by the other markets in the system. This shows how open the Singapore real estate market is and how vulnerable it is to shocks occurring in other leading real estate markets. Geographically and economically close countries like Singapore and Hong Kong show strong volatility linkages in their real estate markets. Whilst Hong Kong affects Singapore considerably with about 46.7 percent of volatility innovations, Singapore real estate market accounts for only 0.23 percent of volatility of Hong Kong real estate market. Furthermore, most of the other foreign markets’ volatility interactions are only below 10 percent. The investment implication is clear. These major real estate markets are still fairly segmented implying that institutional investors would likely to benefit from diversifying public real estate portfolios internationally across Asia and the US/UK markets in the short- and /or medium term. (Table 12 here) Table 13 reports the normalized impulse response coefficients of the markets in the system over 1-week, 4-week, 8-week, 12-week, 16-week, 20-week and 24-week horizons. The numbers demonstrate that shocks originating from all six real estate markets do not influence the behavior of the MSCI global stock market significantly. The MSCI volatility response to one standard deviation shocks from any real estate market in the system is very low on week 1 and is almost negligible on 24 weeks. As an example, the maximum MSCI volatility responses to a shock of one standard deviation on week 1 from Hong Kong, USA, UK, Singapore, Australia and Japan are, respectively, 0.084, 0.084, 0.066, 0.064, 0.050 and 0.043. On the contrary, the results of the dynamic responses also indicate a MSCI response of 0.410 on week 1, 0.218 on week 2, 0.162 on week3, 0.124 on week 4 and so on, to its own one standard deviation shock, a magnitude much larger than that from any other real estate 21 markets. These responses indicate the global stock market behaves largely through its own internal volatility dynamics in the system. The numbers also indicate that the conditional volatilities of the six real estate markets are affected weakly by the global stock market. For Australia, the maximum own-market response to a shock of one standard deviation from other (foreign) markets on week 1 is 0.050 (from MSCI). Similarly, the maximum (foreign market) responses for Japan (0.046), UK (0.065) and USA (0.085) on week 1 are also with respect to one standard deviation shocks from the MSCI global stock market. The volatility responses for Hong Kong and Singapore to one standard deviation shocks on week 1 from MSCI are of comparable magnitudes (0.083 for Hong Kong 0.065 for Singapore), although they are not the highest. The peak coefficient (0.195) of Hong Kong’s response to shocks from Singapore occurs on week 1. The impact of the shocks dies off quickly in about 24th week (0.027). We also see that Singapore’s response to shocks from Hong Kong is the highest with a one-week lag (0.197) and by week 24, the impact of the shocks become very negligible. Finally, various real estate markets’ volatility responses to their own domestic shocks range between 0.404 and 0.411, again a magnitude much larger than any of the cross-volatility response coefficients. To summarize, we find that the generalized impulse response analyses suggest almost the same broad patterns of market interdependence as we find with variance decompositions. The implication is evident. International real estate markets are affected by the global stock market in their second moment. Although there are some volatility interactions among the real estate markets, most of the interactions are quantitatively insignificant. (Table 13 here) Finally, Table 14 compares the variance decomposition results between the pre- and postcrisis periods. In the pre-crisis period, the MSCI global stock market index exerts significant influence on Hong Kong, Japanese, Singapore, UK and US real estate markets but not vice versa. The largest impact is on the US real estate market where its volatility innovation explains up to 36.2 percent of the USA real estate market return volatility in 24 weeks. Among the real estate markets, Australia is largely exogenous as its own volatility innovations are able to account for over 94 percent of its error 22 variances in 24 weeks. Furthermore, there is no major volatility interactions among the real estate markets except those from Hong Kong to Singapore (about 26.2 percent in 24 weeks) and to a lesser degree, from Japan to the USA (6.7 percent in 24 weeks). We witness a different situation in the postcrisis period. Here, we see that the MSCI global stock market is moderately affected by the Japanese real estate market (about 11.4 percent in 24 weeks) but not vice versa. The global stock market also affects Hong Kong (16.4 percent), UK (13.2 percent), USA (11.1 percent) and Singapore (10 percent), respectively, in 24 weeks. Among the real estate markets, Australia is again largely exogenous as only about 5.9 percent of its return volatility is accounted by other markets. There is a bilateral relationship between the Hong Kong and Japanese real estate market. The magnitude of cross-volatility transmission is, respectively, 11.6 percent (from Japan to Hong Kong) and 13.4 percent (from Hong Kong to Japan). Singapore real estate market return volatility is affected by its own shocks and volatility innovations from Japan (26.5 percent in 24 weeks) and Hong Kong (17.5 percent in 24 weeks). Finally, whilst the UK real estate market is only influenced by Australia with approximately 9.7 percent cross-volatility, the USA real estate market is largely unaffected by other real estate markets. (Table 14 here) CONCLUSION In this study, we examine the patterns and significances of dynamic interdependence and casual linkages across six major real estate markets in the USA, UK, Japan, Australia, Hong Kong and Singapore and the MSCI global stock market. We document the general direction and lead/lag structure through mean- and variance-in-causality analyzes and furthermore, how conditional volatilities are related and how important each market’s conditional volatility is to the rest. Finally, by implementing the various statistical tests with two shorter sub-sample periods (pre- and post-Asian financial crisis periods), we investigate whether the patterns and significances of the market interdependence have remained stable over the two sub-periods as both the stock and real estate markets experienced periods of stability and volatility in the 90’s. This knowledge would help fund managers 23 in managing their exposure in the major real estate markets and constructing better asset allocation models. Following Cheung and Ng (1996)’s causality in variance technique, we find that all markets show small to moderate correlations in returns; additionally some markets are not interdependent because they are related through their second moments. There is some evidence of causality-in-mean and /or causality-in-variance among the real estate/global stock markets and the majority of lead-lag linkages happened at one week. The global stock market has strong causal relationships in returns and volatilities with all real estate markets. Additionally, the causal linkages among all markets appear to be unstable across the two sub-periods as only four market-pairs’ causal interactions in mean and variance remain unchanged throughout the two sub-periods. Hence we conclude that the causal linkages among the major real estate /global stock markets under crisis circumstances differ from those in stable times. However, there is no conclusive evidence to suggest that the markets have become more inter-dependent with each other in the post-crisis period. One possible reason is that these markets are the world’s real estate market leaders at the respective continents (i.e. America, Europe and Asia). Volatility shocks on each market can be temporary or highly persistent and that each real estate market’s volatility is subject to the effects of movements in macroeconomic fundamental. As a possible future extension, it would be useful to examine the effects of macroeconomic and financial variables when considering market linkages in volatility during stable and crisis periods. The results on the dynamic volatility response to external shocks analyses indicate that different markets’ responses to shocks from different sources seem to be diverse. The MSCI global stock market appears to be the main “exporter” of shocks to other real estate markets. Geographically and economically close countries such as Hong Kong and Singapore show strong interdependence in their real estate market volatilities. However, most of foreign markets’ shocks only contribute to less than 10 percent of each own market’s volatility. To conclude, our study also underscores the complexity of inter-market relationships in mean and particularly in volatility in at least two aspects: time-dependent market interdependence and 24 temporary/highly persistence in shocks on market linkages. These are some possible future extensions. In the present context, the causality and interdependence evidence provide regulators with useful information to understand own market’s mean and volatility position relative to other foreign markets; i.e. whether it is “endogenous” or “exogenous”; identify relevant mean and volatility “drivers” and take appropriate corrective measures to moderate fluctuations in market movements. Our statistical evidence also provides institutional investors and fund managers with fresh insights into the diversification potential of international real estate investing during stable and crisis periods after controlling for the global stock market effect. This knowledge would further assist fund managers in managing their exposure in international real estate markets and constructing better asset allocation models. REFERENCES Alaganar, V.T. and R. Bhar (2003), An international study of causality in variance: interest rate and financial sector return, Journal of Economics and Finance 27: 39-55 Bollerslev, T. (1990), Modelling the coherence in short-term nominal exchange rates: a multivariate generalized ARCH model, Review of Economics and Statistics 72: 498-505 Bollerslev T., R. Chou and K. Kroner (1992), ARCH modeling in finance: A review of the theory and empirical evidence, Journal of Econometrics, 52, 5-59 Caporale, G.M., N. Pittis and N. Spagnolo (2002), Testing for causality-in-variance: an application to the East-Asian Markets, International Journal of Finance and Economics 7: 235-245 Chung, P.J. and D.J. Liu (1994), Common stochastic trends in Pacific Rim stock markets, The Quarterly Review of Economics and Finance 34(5): 241-259 Cheung, Y. W. and L. K. Ng (1996), A causality-in-variance test and its application to financial market prices, Journal of Econometrics 72: 33-48 Eichholtz, P. M. A. (1995), Does International Diversification Work Better for Real Estate than for Stocks and Bonds? Financial Analysis Journal, 52(1), 56-62. Engle, R.F. and R. Susmei (1993), Common volatility in international equity markets, Journal of Economics and Business Statistics 11(2): 167-176 Eun, G.S. and S. Shum (1989), International transmission of stock market movements, Journal of Financial and Quantitative Analysis 24(2): 241-56 Fujii, E. (2005), Intra and inter-regional causal linkages of emerging stock markets: evidence from Asia and Latin America in and out of the crises, Int. Fin. Markets, Inst. and Money 15: 315-345 Garvey, R., G. Santry and S. Stevenson (2001), The linkages between real estate securities in the AsiaPacific, Pacific Rim Property Research Journal 7(4): 240-258 25 Hamos, Y. D.W. Masulius and V. Ng (1990), Correlations in price changes and volatility across international stock markets, The Review of Financial Studies 3(2): 281-307 Hu, J. and Mei, J. P. (1999), The Return and Risk of Emerging Markets Property Stock Indexes. Emerging Markets Quarterly, 3(1), 10-21. Hu, J.W.S., M.Y. Chen, C.W. Fok and B. N. Huang (1997), Causality in volatility and volatility spillover effects between US, Japan and four equity markets in the South China Growth Triangular, Journal of International Financial Markets, Institutions and Money 7: 351-367 Janakiramanan, S. and A.S. Lamba (1998), An empirical examination of linkages between Pacific-Basin stock markets, Journal of International Financial Markets, Institutions and Money 8: 155-173 Kallburg, J. Liu, C. and P. Pasquariello (2002), Regime shifts in Asian equity and real estate markets, Real Estate Economics 30(2): 263-292 Kanas, A. and G.P Kouretas (2002), Mean and variance causality between official and parallel currency markets: Evidence from four Latin American countries, Financial Review 37: 137-164 Karolyi, G.A. (1995), A multivariate GARCH model of international transmissions of stock return and volatility, Journal of Business and Economic Statistics 13(1): 11-25 Koutmos, G. (1996), Modeling the dynamic interdependence of major European stock markets, Journal of Business Finance and Accounting 23(7): 975-988 Liow, K.H., O.T.L., Ooi and Y. Gong (2005), Cross-market dynamics in property stock markets: some international evidence, Journal of Property Investment and Finance 23(1): 55-75 Liow K.H., H. Zhu, David H.K. Ho and K. Addae-Dapaah (2005), “Regime Changes in International Property Markets” Journal of Real Estate Portfolio Management 11(2), 147-165 (USA) Liow, K.H. (2006), Real estate return volatility and systematic risk: evidence from international markets, working paper, Department of Real Estate, National University of Singapore Lu, K. and Mei, J. P. (1999), The Return Distributions of Property Shares in Emerging Markets, Journal of Real Estate Portfolio Management, 5(2), 145-160 Michayluk D., P. Wilson and R. Zurbruegg (2006), Asymmetric volatility, correlation and returns dynamics between the U.S. and U.K. securitized real estate markets, Real Estate Economics 4 1:109-131 Nelson, D. (1991), Conditional Heteroscedasticity in Asset Returns: A New Approach, Econometrica 59: 323-70 Pesaran. M.H. and Y. Shih (1998), Generalized impulse response analysis in linear multivariate models, Economic Letters 58: 17-29 Sims, C.A. (1980), Macroeconomics and reality, Econometrica 48: 1-48 Stevenson, S. (2002) “An Examination of Volatility Spillovers in REIT Returns” Journal of Real Estate Portfolio Management 8(2): 229-238 Tay, N.S.R and Z. Zhu (2000), Correlations in Returns and Volatilities in Pacific-Rim Stock Markets, Open Economic Review 11: 27-47 UBS Investment Bank (2003), Global Real Estate Market Conditions, Research report 26 Table 1 Summary statistics of weekly real estate market and world stock returns mean(%) std dev(%) Coefficient of variation skewness kurtosis Jarque-Bara (JB) LB-Q(5) LB-Q(20) LB-Q2(5) LB-Q2(20) ADF(I1) PP(I1) MSCI Austraila Hong Kong 0.098 0.258 0.214 1.911 2.038 4.403 19.5 7.89922481 20.57476636 (0.226***) (0.275***) (0.496***) 5.032*** 3.506*** 7.742*** 151.81*** 19.52*** 719.56*** 3.15 9.99* 15.62*** 14.49 22.52 43.84*** 70.88*** 48.89*** 35.76*** 111.13*** 123.31*** 118.73*** -28.88 -30.33 -26.79 -29.91 -30.33 -26.91 Japan 0.002 4.92 2460 0.442*** 4.575*** 114.26*** 9.15* 32.21** 68.21** 133.32** -31.77 -31.76 Singapore UK US 0.074 0.159 0.293 5.144 2.401 1.816 69.5135135 15.1006289 6.19795222 (0.865***) 0.032 (0.430***) 21.90*** 4.684*** 4.349*** 12622*** 99.54*** 419.05 27.33*** 11.05* 11.28** 48.53*** 20.63 25.89 110.34*** 9.08* 58.69*** 164.04*** 26.42 73.21*** -29.34 -26.35 -26.81 -29.37 -26.35 -26.88 Notes: Weekly returns are from January 5, 1990 to Feb 10, 2006 (841 weeks) for MSCI world stock market and public real estate markets of Australia, HK, Japan, Singapore, the UK and the US. LB-Q (5), LB-Q (20), LB-Q2 (5) and LB-Q2 (20) are the Ljung-Box test statistics for serial correlations in the returns and squared returns, respectively. ADF and PP are, respectively, the augmented I (1) statistics for DickeyFuller and Phillips-Perron tests for unit roots. ***, **, * indicates two-tailed significance at the 1, 5 and 10 percent levels respectively. Table 2 Correlations in returns (variances): 05/01/1990 – 06/02/2006 return (variances) MSCI Australia Hong Kong Japan Singapore MSCI 1 1 Australia 0.336*** 0.186*** 1 1 Hong Kong 0.454*** 0.292*** 0.287*** 0.185*** 1 1 Japan 0.424*** 0.171*** 0.171*** 0.122*** 0.221*** 0.170*** 1 1 Singapore 0.421*** 0.209*** 0.204*** 0.107*** 0.581*** 0.762*** 0.246*** 0.106*** 1 UK US UK 0.423*** 0.351*** 0.303*** 0.103*** 0.242*** 0.079** 0.236*** 0.022 0.230*** 0.022 1 1 US 0.458*** 0.431*** 0.226*** 0.123*** 0.220*** 0.162*** 0.096*** 0.116*** 0.251*** 0.093*** 0.256*** 0.280*** 1 1 Notes: Numbers in brackets Pearson correlation coefficients in variance between the respect pairs. ***, ** indicates statistical significance at one and five percent levels respectively. 27 Table 3 Multivariate AR (p) – EGARCH (1, 1) estimates of real estate market /MSCI returns (conditional t distribution) 5/1/1990 - 10/2/2006 (Entire sample period) Constant Condtional Correlation (CCC)- MEGARCH (1,1) Shape = 9.839 (t = 10.19***) Log-likelihood = 13506.3443 MSCI(j=1) AU (j=2) HK(j=3) JP(j=4) SG(j=5) UK(j=6) Mean equation constant D97 variance equation constant arch garch CCC pj1 pj2 pj3 pj4 pj5 pj6 Test on std residuals Arch(10)[prob] Q(10)[prob] Q(20)[prob] Q2(10)[prob] Q2(20)[prob] 0.00187*** 0.00297 0.00367*** -0.00321 0.00388*** -0.01214** 0.00175 -0.00823* 0.00306*** -0.01229* 0.00243*** n.a. US(j=7) 0.07667*** n.a. -0.15627*** -0.66591*** -0.38876*** -0.40295*** -0.28297*** -0.08641*** -0.31777*** 0.09626*** 0.13501*** 0.14009*** 0.19461*** 0.13482*** 0.06731*** 0.12118*** 0.98992*** 0.92768*** 0.95683*** 0.95847*** 0.97230*** 0.99521** 0.97206*** 0.3570*** 1.147[0.32] 13.345[0.21] 19.411[0.50] 10.494[0.40] 17.933[0.59] 0.548[0.87] 12.075[0.28] 24.313[0.23] 4.741[0.91] 15.193[0.77] 0.4394** 0.3006*** 0.4661*** 0.1649*** 0.1771*** 0.4185*** 0.1930*** 0.4723*** 0.2043*** 1.003[0.44] 2.025[0.028] 0.753[0.67] 12.204[0.27] 10.072[0.43] 7.341[0.61] 32.31[0.04] 17.20[0.64] 17.295[0.63] 10.433[0.40] 18.862[0.042] 7.495[0.68] 42.88[0.05] 29.169[0.084] 15.989[0.71] 0.4436*** 0.3054*** 0.2620*** 0.2489*** 0.2819*** 0.4524*** 0.2122*** 0.2102*** 0.1023*** 0.2212*** 0.2485*** 1.828[0.053] 8.492[0.58] 12.952[0.88] 17.464[0.07] 25.449[0.18] 0.634[0.79] 10.108[0.43] 21.770[0.36] 6.935[0.73] 17.098[0.65] ***, ** - indicates statistical significance at the one and five percent levels 28 Table 4 Multivariate AR (p) – EGARCH (1, 1) estimates of real estate market /MSCI returns (conditional t distribution): pre- and post-crisis periods 5/1/1990 - 27/6/1997 (Pre-crisis period) Constant Condtional Correlation (CCC)- MEGARCH (1,1) Log-likelihood = 6596.6634 Shape = 11.068 (t = 8.36***) MSCI(j=1) AU (j=2) HK(j=3) JP(j=4) SG(j=5) UK(j=6) US(j=7) Mean eq constant 0.0018*** variance eq constant -0.5434*** arch 0.0862*** garch 0.9431*** CCC pj1 pj2 pj3 pj4 pj5 pj6 Test on std residuals Arch(10)[prob] 1.68[0.09] Q(10)[prob] 18.01[0.06] Q(20)[prob] 26.06[0.16] Q2(10)[prob] 21.51[0.02] Q2(20)[prob] 42.32[0.01] 0.0032*** 0.0054*** 0.0044** 0.0035*** 0.0041*** -3.0829*** -0.2635*** -1.1423*** -0.5120*** -0.0608*** -0.4691*** -0.1196** -7.9376*** -0.3389 -0.1967 -0.1995* 0.2272* 0.1196*** 0.2867*** 0.0801** 0.0740*** 0.1176*** 0.0862*** -0.0134 0.1457 0.0853 0.1167** 0.6358*** 0.9741*** 0.8533*** 0.9338*** 0.9995*** 0.9553*** 0.9936*** 0.2483 0.9658*** 0.9790*** 0.9836*** -3.3299 0.2158** 0.5964** -9.7904 0.2243*** 0.3207 0.415*** 0.399*** 0.327*** 0.222*** 0.333*** 0.441*** 0.235*** 0.195*** 0.066 0.231*** 0.292*** 1.15(0.32) 7.88[0.64] 12.63[0.89] 12.36[0.26] 22.22[0.33] 0.16(0.99) 9.99(0.44) 15.69[0.74] 1.97[0.99] 6.65[0.99] 0.281*** 0.345*** 0.262*** -0.0001 0.678*** 0.120*** 0.108** 0.0021*** 0.403*** 0.059 0.381*** 0.193*** 0.0019* 1/1/1999 - 10/2/2006 (Post-crisis period) Constant Condtional Correlation (CCC)- MEGARCH (1,1) Log-likelihood = 6096.7621 Shape = 12.962 (t=6.92***) MSCI(j=1) AU (j=2) HK(j=3) JP(j=4) SG(j=5) UK(j=6) US(j=7) 0.493** 0.204*** 0.227*** 0.258*** 0.255*** 0.0035*** 0.0023*** 0.0042*** 0.439*** 0.174*** 0.214*** 0.137*** 0.206*** 0.206*** 0.384*** 1.18[0.31] 1.76[0.07] 1.16[0.32] 0.84[0.59] 0.74[0.68] 0.32[0.98] 0.81[0.62] 1.28[0.18] 11.61[0.31] 7.59[0.67] 22.87[0.01] 5.16[0.88] 4.59[0.91] 12.07[0.28] 11.89[0.29] 5.32[0.87] 23.65[0.26] 28.08[0.11] 28.06[0.11] 10.28[0.96] 11.46[0.93] 26.74[0.14] 22.83[0.30] 10.70[0.95] 14.83[0.14] 15.89[0.11] 11.95[0.29] 7.28[0.70] 7.12[0.71] 3.30[0.97] 9.16[0.52] 27.15[0.02] 30.17[0.07] 21.83[0.35] 19.90[0.46] 19.47[0.49] 18.77[0.54] 8.88[0.98] 14.83[0.79] 42.02[0.12] 0.0031** 0.514*** 0.318*** 2.04[0.03] 16.91[0.08] 26.85[0.24] 32.80[0.01] 27.45[0.30] 0.0047*** 0.287*** 0.171*** 0.228*** 0.425*** 0.292*** 0.531*** 0.225*** 0.51[0.88] 1.23[0.27] 7.33[0.69] 11.12[0.35] 11.57[0.93] 19.46[0.49] 5.94[0.82] 13.06[0.22] 11.40]0.94] 15.77[0.73] ***, ** - indicates two-tailed significance at the one and five percent levels 29 Table 5(a) Causality in mean results: 5/1/90 – 10/2/06 (full-sample period) -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 1 2 3 4 5 6 7 8 9 10 0.625 1.771* -0.226 -1.629 -0.356 1.635 0.579 0.798 0.498 0.139 2.957** 0.069 0.023 1.322 -1.340 -0.454 -1.342 0.917 0.790 0.388 AU/JP 5.072** 1.849* -1.473 0.014 -1.507 0.503 0.691 -0.691 -0.489 -0.819 -0.579 0.793 1.394 1.195 0.573 -0.908 -1.137 -0.608 0.732 -0.966 0.451 AU/SG 5.187** 0.741 -0.729 (2.173**) -0.694 0.975 -1.282 -0.006 2.144** 1.311 2.963** -0.778 0.555 0.709 -0.029 1.169 -0.532 0.804 1.539 0.457 0.356 -1.539 0.084 1.287 -0.165 -0.058 0.700 -0.072 0.541 -0.894 0.663 1.478 -0.506 -0.255 0.865 0.541 -0.723 0.987 -1.577 -1.010 (2.343**) 0.723 0.555 1.001 0.341 0.561 0.995 -0.269 -0.026 -0.547 0.376 -1.140 0.573 -0.680 0.214 0.451 -0.746 -0.822 2.155** 0.414 0.584 0.373 (1.854*) -0.376 -0.338 0.668 0.631 -0.122 Lag 0 AU/HK 8.170** 1.094 AU/UK 9.079** 2.858** AU/US 6.307** 1.736* 2.589** 1.177 AU/WD 10.404** 3.541** 1.496 0.825 (2.592**) -0.255 HK/JP 5.694** 0.602 -0.720 -0.304 0.888 -0.521 -0.179 -1.516 -0.333 0.579 -0.567 1.123 1.131 1.065 (2.031**) 0.023 HK/SG 15.429** 1.823* 0.298 0.529 -0.266 -0.555 -0.955 -0.104 -0.590 0.622 0.373 HK/UK 7.479** 2.054** 0.651 0.376 0.891 -0.139 HK/US 6.515** 0.978 0.868 -0.613 -1.293 0.443 -1.331 -1.134 0.095 -0.312 0.078 -0.257 0.877 -0.012 2.873** 1.325 -0.425 -0.402 -0.046 0.200 0.793 0.758 -0.029 -0.454 -0.182 (2.291**) -0.205 2.199** -0.396 -0.582 0.576 -0.191 0.570 0.570 0.295 -0.474 -0.521 -1.513 0.642 -0.492 -0.127 1.776* -0.706 0.509 0.764 0.174 1.962** 0.908 -0.347 -1.325 0.385 -1.123 1.577 0.822 1.125 HK/WD 12.837** 2.893** 0.694 0.341 -0.654 -0.310 1.238 -0.454 -0.179 0.911 0.547 0.292 1.881* -0.833 -0.862 0.176 0.179 -0.318 -0.674 0.521 JP/SG 6.385** -1.366 -0.784 -0.385 -1.490 0.246 -1.018 -1.423 0.003 1.525 0.587 1.180 1.797* -0.393 0.324 -0.139 0.165 -0.136 1.568 0.477 0.422 JP/UK 7.354** 2.190** 1.438 1.421 1.180 -0.069 1.368 -0.376 1.386 0.801 -0.793 -0.451 -0.671 -0.052 -0.052 0.865 0.793 0.009 0.356 -0.784 -0.446 JP/US 3.136** -1.189 1.125 1.287 0.472 -0.017 0.561 -1.128 1.473 1.458 1.073 0.176 0.156 -0.312 1.528 0.182 0.043 -0.370 -0.671 0.376 0.527 JP/WD 13.314** 1.496 0.529 -0.226 -0.550 1.290 -0.853 0.353 0.729 0.095 (2.902**) 1.007 -0.168 0.353 0.958 -1.206 -0.341 -1.059 -0.162 0.842 -0.133 SG/UK 7.762** 3.165** 1.224 -0.260 -0.437 2.239** 1.180 -1.140 1.727* 2.228** -0.550 0.321 -1.273 -0.634 0.191 0.503 0.041 (1.687*) 0.451 0.772 0.142 SG/US 7.033** 0.023 1.276 -0.515 -0.243 0.231 -1.635 -0.020 1.765* 1.134 0.787 -0.159 1.441 0.631 0.035 1.447 1.247 1.117 -0.509 -0.402 SG/WD 12.012** 3.449** 1.771* -0.226 -0.761 -0.179 0.989 -0.295 1.490 2.653** -0.246 1.701* 0.955 -0.900 -0.987 -0.448 -0.159 -1.201 0.043 1.056 0.509 UK/US 7.808** 2.352** 0.579 0.150 -0.446 0.532 (1.892*) -0.081 1.151 1.956* 0.385 -0.107 1.036 1.322 1.817* 0.518 0.506 0.182 0.691 -1.068 -0.026 UK/WD 13.051** 1.149 -1.293 -0.009 -1.218 0.697 -0.443 0.165 1.383 -0.689 -1.044 1.918* 0.657 0.179 0.347 -0.286 0.625 -0.804 0.923 0.081 0.081 US/WD 13.499** 0.150 1.392 0.165 1.597 0.495 1.493 -0.269 -0.197 -0.599 -0.712 0.367 0.801 0.145 -0.564 -1.189 -1.030 -0.532 0.179 1.123 -0.414 0.984 Notes: Each entry gives the causality in mean t-statistics between the real estate / global stock markets of the countries noted in the first column. The six real estate markets are Australia (AU), Hong Kong (HK), Japan (JP), Singapore (SG), the UK and the USA; the global stock market (WD) is represented by the MSCI World Index. The lag k denotes the number of weeks by which the return of the first market lags or leads that of the second market. A negative (positive) value of k indicates lags (leads). Significant t-statistics with k<0 suggests that the first market causes the second market. Significance with k>0 indicates causality in the opposite direction. ***, ** - indicates statistical significance at one and five percent levels. 30 Table 5(b) Causality in variance results: 5/1/90 – 10/2/10 (full-sample period) -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 1 2 3 AU/HK 4.360** 1.256 -1.279 1.209 -1.175 0.327 -1.435 1.250 0.411 -1.360 -0.067 1.504 -0.845 0.584 1.828* 0.657 0.249 -1.070 Lag 0 4 5 -0.683 2.610** 7 8 1.389 -1.513 9 10 -0.715 (1.655*) -1.267 -0.145 -0.758 -1.027 1.909* 1.510 0.746 0.515 1.412 1.079 0.243 -0.119 -0.333 1.397 -0.599 AU/SG 2.121** 2.951** -0.778 2.627** 0.194 -0.255 -1.293 1.765* 1.820* 0.142 -1.096 0.307 -0.651 0.145 -0.393 0.535 0.159 1.551 -1.444 -0.159 -1.079 AU/UK 4.450** 1.204 0.156 -0.527 -0.249 0.610 -1.146 0.373 -0.576 (1.996**) -1.050 2.555** 2.060** -0.388 0.894 0.240 AU/US 2.902** 1.177 0.029 -0.231 0.038 -0.078 -0.642 -0.278 -0.657 (1.739*) -1.371 AU/WD 5.731** 0.353 0.295 0.663 -1.241 -0.370 -0.715 1.042 -0.347 HK/JP 3.370** 0.385 AU/JP 2.485** -0.084 6 0.865 0.804 0.639 -0.171 -1.042 0.622 0.506 0.382 0.133 1.516 0.744 0.266 2.115** -0.338 -0.101 0.437 0.095 (1.753*) -0.084 0.214 -0.243 0.084 1.426 1.559 2.332** -1.565 1.290 -0.020 -0.515 -0.330 0.107 -0.249 0.061 1.522 -0.078 -0.139 -0.139 -0.527 -0.932 -0.237 0.185 0.822 -0.286 -0.367 -0.784 0.428 2.245** 0.038 -0.906 1.490 0.000 HK/SG 18.215** 0.365 -0.278 -0.778 3.072** -0.639 0.041 -0.396 0.200 -0.761 -0.527 1.389 0.356 HK/UK 3.628** -0.405 0.385 -0.243 -0.260 0.605 -0.793 0.830 -1.510 0.634 1.713* 4.695** -0.781 -0.451 -0.926 1.166 -0.419 0.998 -0.663 -1.331 HK/US 3.668** 1.993** -0.605 1.516 -0.107 0.417 0.547 0.257 -0.110 -1.632 1.215 1.585 -0.856 -0.706 0.677 0.495 -1.206 -0.220 -0.914 0.477 0.573 -1.328 -0.402 0.396 -0.252 1.230 2.720** 1.672* -0.492 2.222** 1.050 1.805* HK/WD 7.675** 2.155** 0.284 -0.564 -0.518 -0.845 1.073 1.672* 0.949 -0.017 1.664* 2.115** 1.580 -0.448 1.285 -0.712 0.639 1.102 2.141** 0.657 2.853** 1.195 -0.599 -0.463 -0.770 0.830 -0.344 0.995 -0.489 0.851 0.009 -0.284 -0.211 -0.984 2.497** 5.589** JP/UK 0.804 0.197 -0.150 -0.474 1.455 -0.853 0.124 1.811* 0.124 -1.125 -0.315 0.744 0.081 -0.281 -0.153 3.232** 0.498 0.017 1.366 1.669 -0.176 JP/US -0.122 0.495 0.405 1.134 -0.466 1.687* 1.189 0.038 0.729 -1.056 -0.336 -0.934 0.130 -0.521 -0.009 4.473** 0.385 0.107 0.026 -0.486 -0.746 JP/WD 4.774** 0.746 -0.906 0.683 1.241 0.515 -0.240 -0.133 1.001 0.457 0.272 1.785* 1.519 -0.315 0.041 0.871 0.868 0.914 0.894 0.654 -0.203 SG/UK 1.768* -0.260 -0.365 0.443 -0.113 -0.553 0.610 -0.961 0.087 -0.833 -0.414 1.826* 3.538** -0.197 -0.336 -0.532 0.139 -0.414 0.489 -0.628 -0.775 SG/US 2.184** 0.608 -0.110 -0.124 1.004 0.093 0.182 0.052 -0.307 -0.492 0.796 1.435 0.393 -0.032 -0.593 -0.272 -0.373 -0.289 -0.758 -0.558 -0.906 SG/WD 5.086** 1.053 0.307 -0.012 0.865 -0.064 -0.150 2.242** 0.162 -0.593 2.031** 1.192 -0.448 -0.023 1.467 -0.524 0.017 0.278 -0.156 -0.712 -0.544 1.125 0.463 -0.324 -0.451 0.113 0.865 0.292 0.029 -0.917 4.354** -0.457 -0.639 -0.729 0.003 -1.267 (1.730*) 0.237 0.208 -0.286 UK/WD 11.428** 3.191** 2.083** 1.065 0.223 -0.362 0.469 1.328 0.451 -0.642 -0.952 2.346** 0.726 -0.880 0.307 -0.393 -0.981 -1.230 -1.042 -1.039 -0.535 US/WD 11.598** 2.936** 1.177 1.177 0.506 0.116 -0.489 -0.443 -0.113 -1.499 3.813** 0.767 0.443 2.236** 0.411 -0.093 -0.356 -0.862 -0.443 -0.660 JP/SG UK/US 10.924** 3.775** 1.337 Notes: Each entry gives the causality in variance t-statistics between the real estate / global stock markets of the countries noted in the first column. The six real estate markets are Australia (AU), Hong Kong (HK), Japan (JP), Singapore (SG), the UK and the USA; the global stock market (WD) is represented by the MSCI World Index. The lag k denotes the number of weeks by which the variance of the first market lags or leads that of the second market. A negative (positive) value of k indicates lags (leads). Significant t-statistics with k<0 suggests that the first market causes the second market. Significance with k>0 indicates causality in the opposite direction. ***, ** - indicates statistical significance at one and five percent levels. 31 Table 6 Summary of causality patterns: full sample period (5/1/90-10/2/06) Australia Australia HongKong Japan Singapore UK US MSCI CIM CIV CIM CIV CIM CIV CIM CIV CIM CIV CIM CIV CIM CIV 0,-1 0, -10 0 0, -3 0, -1 0 0 0,-1,-2 0 0, -7 0,-1,-5 0, -7 Hong Kong 0, -2 0, -5 0,-1 0,-3,-4 0, -1 0, -4 0 0,-1,-2 0,-2,-7 0,-1,-2 0,-1 0, -1 Japan 0, -1 0, -8 0 0, -8 0, -1 -7,-8 0 -5 0 -5 0 0, -1 Singapore 0,-4,-9 0,-1,-3,-7,-8 0,-1 0, -4 0 -1,-3 0, -7 0,-1,-2 0 0 0,-1 0 UK 0, -1 0, -9 0,-1,-7,-9 0 0,-1 -7 0,-1,-5,-8,-9 0 0,-4 0,-1,-7 0,-1 0, -1 US 0,-1,-6 0, -9 0 0, -1 0 -5 0, -9 0 0,-1,-6,-9 0, -1 MSCI 0,-1,-4 0, -10 0,-1 0,-1,-7,-10 0, -10 0 0,-1,-2,-9 0,-7,-10 0 0,-1,-2 0 0, -1 0 0, -1, -4 Notes: This summary is constructed based on the results reported in Table 5(a) (causality in mean – CIM) and Table 5(b) (causality in variance – CIV). 0 indicates contemporaneous correlation between two markets exist. –k indicates the presence of correlation at lag k. For example: 0,-2 in the first cell of the Hong Kong column indicates there is contemporaneous relation between the means of the Australia and Hong Kong real estate markets, and the mean return of the Australia market at lag 2 has an impact on the current mean return of the Hong Kong market 32 Table 7(a) Causality in mean results: 5/1/90 – 27/6/97 (pre-crisis period) -2 -3 -4 -5 -6 -7 -8 -9 -10 1 2 3 4 5 6 7 8 9 10 AU/HK Lag 4.100** -0.057 0 -1 1.525 0.748 -1.544 0.521 0.218 2.672** 0.366 2.467** 0.795 1.664* -0.498 1.064 -1.021 -0.352 -0.726 -1.328 1.064 1.418 -0.193 AU/JP 2.353** 2.227** -1.310 0.201 0.732 -1.013 -0.240 -1.096 -1.411 1.348 -0.606 0.089 2.075** 0.173 0.539 0.033 -0.193 -0.751 0.417 -1.060 -0.429 0.773 0.942 0.437 -1.399 0.053 0.376 0.474 -0.429 2.841** 1.316 1.277 -0.151 0.002 -0.035 0.445 1.422 -1.574 1.434 0.690 -0.340 AU/UK 3.718** 1.798* -1.310 0.053 -0.370 1.253 0.106 1.617 1.521 0.565 0.626 -0.582 -0.476 0.797 0.061 -0.706 -0.810 0.883 -0.374 -1.100 1.289 AU/US 3.470** 0.110 0.734 -1.009 0.338 -1.639 -0.236 -0.061 0.946 0.370 -0.429 0.728 -0.651 -0.610 0.421 -0.661 -0.724 1.566 -0.268 0.720 AU/WD 5.253** 2.221** -1.121 0.818 (2.048**) 0.002 -0.423 -0.090 0.492 0.885 0.212 0.584 0.846 0.667 -0.128 -0.541 0.063 -0.203 1.355 (1.818*) 0.360 HK/JP 2.154** 0.244 0.277 0.797 -0.584 -0.230 0.559 0.645 1.111 2.308** -1.106 (1.957*) 0.130 (2.630**) -1.586 0.177 -1.119 -0.033 -0.134 0.576 -1.174 HK/SG 8.152** 0.063 1.367 -0.338 0.437 0.262 0.120 1.108 -0.069 -0.008 1.497 -0.110 0.612 1.568 0.336 -0.124 -0.942 HK/UK 4.426** 1.708* 1.560 0.470 0.779 -0.329 0.157 -1.096 -0.714 2.506** -0.506 HK/US 4.322** -0.285 0.395 0.787 0.002 -0.663 -0.468 0.710 1.108 0.155 0.866 -0.679 AU/SG 0.911 0.885 -0.441 0.207 0.761 -0.492 -0.041 0.207 -0.136 0.486 0.610 -0.938 -0.773 -0.704 (1.942*) 0.712 -1.576 1.845* -1.237 -0.468 -0.547 -1.027 1.672* -0.100 1.171 (1.739*) -0.502 0.515 -0.462 0.677 -0.710 HK/WD 6.877** 1.098 1.100 -0.285 0.374 0.793 -0.031 2.485** -0.026 -0.834 JP/SG 3.814** 0.919 -0.124 (2.567**) (1.727*) 0.403 -0.551 -0.423 -0.169 1.283 0.852 2.593** JP/UK 5.075** 2.068** 1.194 0.659 -1.147 1.397 -0.443 0.637 1.729* -0.165 0.669 JP/US 2.988** 2.217** 0.515 -1.007 -1.414 2.190** -0.582 1.955* 0.344 -0.470 -0.797 0.081 JP/WD 13.411** 1.710* 0.301 0.177 (1.794*) 1.076 -1.265 0.301 0.911 0.409 (3.814**) 1.613 1.121 SG/UK 4.957** 2.492** 0.850 0.114 -0.643 1.509 0.708 0.307 -0.297 1.674* 0.171 1.385 1.430 SG/US 4.296** 1.391 -0.724 1.363 -0.201 0.159 0.081 0.232 -0.205 0.498 0.087 -0.877 1.662* SG/WD 8.032** 1.649* 1.727* 0.384 -0.753 -0.374 -0.004 1.586 0.445 1.436 UK/US 3.954** 1.123 -0.765 -1.051 0.057 -0.321 -1.499 -0.840 -0.035 -0.586 -0.051 0.724 UK/WD 9.752** 0.454 -1.007 -0.395 -0.230 0.618 -0.927 -0.264 1.316 -0.757 9.039** -0.631 1.346 0.214 0.710 -0.049 0.610 2.003** 0.012 0.818 US/WD 0.124 -1.627 -1.119 0.708 (1.865*) 0.049 -0.502 0.421 0.360 -0.547 0.246 0.671 0.275 -1.265 0.321 0.970 1.784* -0.480 -0.462 1.710* 0.163 -0.753 1.210 0.944 0.822 -0.079 0.478 -0.159 0.171 0.919 -0.090 1.206 0.653 -0.370 (2.040**) -0.157 0.964 1.279 (2.801**) -0.620 0.246 0.140 -0.614 (2.188**) 0.104 0.224 -0.456 -0.295 0.513 -0.185 1.058 1.218 1.326 0.151 -1.021 1.826* -1.446 -1.350 -0.081 0.014 -0.773 -0.179 1.371 -1.104 0.529 0.220 1.039 1.072 0.506 1.308 0.317 0.710 -1.133 -1.121 2.252** 0.993 0.675 -0.321 -0.382 1.145 0.655 0.114 0.797 -0.083 -0.474 0.203 0.622 -0.470 0.043 0.096 0.795 0.205 0.214 -1.129 2.638** -0.907 0.787 2.268** -1.106 Notes: Each entry gives the causality in mean t-statistics between the real estate / global stock markets of the countries noted in the first column. The six real estate markets are Australia (AU), Hong Kong (HK), Japan (JP), Singapore (SG), the UK and the USA; the global stock market (WD) is represented by the MSCI World Index. The lag k denotes the number of weeks by which the return of the first market lags or leads that of the second market. A negative (positive) value of k indicates lags (leads). Significant t-statistics with k<0 suggests that the first market causes the second market. Significance with k>0 indicates causality in the opposite direction. ***, ** - indicates statistical significance at one and five percent levels. 33 Table 7(b) Causality in variance results: 5/1/90 – 27/06/97 (pre-crisis period) Lag 0 -1 -2 AU/HK 0.289 -0.179 -0.565 -0.238 (1.822*) 1.117 -3 -4 -5 -6 -7 -8 -9 -10 1 2 3 -0.590 0.488 -0.594 -0.008 0.732 -0.415 -1.279 1.519 4 5 -0.490 2.160** 6 7 8 9 10 0.952 1.351 -0.946 0.026 -1.348 AU/JP 0.016 1.953* 0.285 -0.716 -1.078 -0.067 0.545 -1.060 0.797 0.537 -0.218 0.696 2.020** 0.708 -0.018 0.096 -0.616 1.391 (1.991*) 1.393 -0.504 AU/SG -0.364 1.804* -0.407 -0.407 1.212 -0.179 -1.161 0.014 -0.687 0.964 -0.470 -0.089 -0.604 0.342 -0.449 0.413 -0.270 1.420 -1.102 AU/UK -0.549 0.826 -0.750 0.366 0.057 0.338 (1.749*) 2.172** 0.409 -1.127 -1.161 0.417 0.287 0.293 -0.031 1.318 0.689 -1.076 1.987** -0.879 -1.072 2.347** AU/US 0.508 1.505 0.704 0.035 1.503 -0.954 -0.022 -1.534 -0.624 -1.070 0.130 -1.351 -0.240 -0.047 0.830 -0.102 0.338 -0.006 0.748 -0.321 AU/WD 1.176 0.722 -0.476 -0.130 -0.561 0.014 0.069 -0.055 1.174 1.414 -1.076 -1.477 1.407 0.588 0.488 0.142 1.249 1.361 (1.682*) -0.130 0.671 HK/JP 1.393 -0.610 0.466 -0.647 -0.344 -0.151 -0.388 -0.519 -0.936 -0.122 -1.348 0.277 0.498 0.445 0.954 -0.063 0.476 1.774* 1.058 -0.698 2.913** HK/SG 6.724** 0.570 0.862 -0.502 0.360 -1.210 -0.199 -1.041 -0.193 -0.925 -0.771 0.999 0.342 -0.677 1.076 0.944 -0.155 -0.004 -0.340 HK/UK 0.134 -0.087 -1.031 -0.824 -0.915 -0.667 0.624 -0.559 1.678* -0.860 1.418 -0.065 2.620** -0.787 1.692* -0.942 2.441** -0.218 1.365 0.281 -0.397 2.398** -1.127 0.671 -0.832 2.939** 1.961** HK/US 0.830 0.140 -0.205 1.405 0.035 0.348 -0.291 -0.118 0.346 -0.844 0.035 0.012 HK/WD 3.864** 0.370 0.427 -0.687 0.395 -0.895 0.244 -0.346 -0.871 -0.222 -0.582 -0.578 3.124** 0.134 0.873 0.216 JP/SG 0.673 0.474 -0.059 2.886** 0.490 -0.087 -0.256 2.101** 3.063** -0.094 0.736 -1.037 0.860 -0.128 0.173 JP/UK -0.083 0.938 1.680* -0.781 0.124 0.578 -0.631 -0.020 -0.537 1.070 1.035 0.128 -0.372 JP/US 0.327 1.505 0.431 1.489 0.799 0.364 1.430 -0.260 -0.415 -0.917 0.028 -1.306 0.612 JP/WD 9.163** 0.073 0.334 1.564 SG/UK 0.612 0.311 -0.565 -0.472 1.991** -0.995 0.102 1.971** 3.714** 2.097** -0.746 0.370 2.750** 2.673** -0.763 -0.232 0.464 0.500 -0.146 1.509 2.406** 2.115** 1.967** -0.799 -1.049 1.652* 0.903 -0.004 1.940* 2.311** 0.264 -0.445 0.234 -0.706 0.161 -0.537 -0.352 0.968 0.582 0.342 -0.364 0.472 1.204 1.074 1.239 1.127 1.084 -1.039 0.030 -0.059 -0.923 1.373 0.915 0.628 -0.586 0.435 -0.761 -0.633 -1.377 0.053 0.059 -0.724 -0.616 1.312 0.612 1.058 0.153 1.975* 0.073 0.614 -0.932 0.187 1.216 -0.146 -0.232 -1.214 SG/US 1.694* 1.200 0.832 -0.350 3.734** 0.767 -1.412 0.413 0.429 -0.885 1.580 0.063 1.273 1.645* -0.209 1.051 0.380 SG/WD 3.803** 0.433 0.950 1.121 1.338 -0.372 0.425 1.033 -0.641 0.077 1.666* 0.297 1.202 2.994** 1.591 0.978 1.625 UK/US 0.659 2.388** 1.054 0.301 0.728 0.877 0.513 0.085 0.126 -0.075 0.313 0.612 -0.716 -0.183 0.112 -0.807 -0.537 -1.322 -0.108 0.661 0.012 UK/WD 4.688** 2.315** 0.793 -0.285 0.950 0.081 2.142** 2.327** 1.713* 0.761 0.901 -0.496 1.324 -1.643 1.322 -1.389 -0.138 -0.675 -0.472 -0.506 -0.004 US/WD 5.731** 0.106 0.596 0.744 0.409 1.062 -0.706 3.334** 0.362 0.059 1.869* 1.153 0.352 0.035 0.574 -0.033 -0.146 0.980 -0.061 0.976 0.325 2.654** 2.095** -0.452 0.201 Notes: Each entry gives the causality in variance t-statistics between the real estate / global stock markets of the countries noted in the first column. The six real estate markets are Australia (AU), Hong Kong (HK), Japan (JP), Singapore (SG), the UK and the USA; the global stock market (WD) is represented by the MSCI World Index. The lag k denotes the number of weeks by which the variance of the first market lags or leads that of the second market. A negative (positive) value of k indicates lags (leads). Significant t-statistics with k<0 suggests that the first market causes the second market. Significance with k>0 indicates causality in the opposite direction. ***, ** - indicates statistical significance at one and five percent levels. 34 Table 8(a) Causality in mean results: 4/9/98 – 10/2/06 (post-crisis period) -1 -2 -3 -4 -5 -6 -7 -8 -9 AU/HK 6.736** Lag 0 1.423 0.253 -0.667 -1.105 -0.632 1.882* -1.409 -1.034 -1.273 AU/JP 3.550** 1.248 -1.468 0.441 -1.056 0.642 1.368 AU/SG 6.283** 1.615 -0.742 -0.122 (2.135**) -0.120 1.837* AU/UK 8.433** 1.848* 1.407 -1.489 0.094 0.257 -0.084 -1.215 2.211** 1.211 -1.258 -0.389 -1.101 0.759 0.347 0.693 0.457 0.267 0.275 0.757 1.185 -0.791 0.565 1.462 (1.682**) 1.016 1.342 -0.854 -0.840 -1.175 -1.258 1.387 -0.567 0.167 0.936 -1.652 -0.479 -1.330 AU/US 4.919** 1.220 AU/WD 7.972** 2.980** 1.678* HK/JP 5.176** 2.192** (1.791**) 0.385 2 3 4 5 6 7 8 9 10 -0.006 1.995** 0.298 -0.691 1.650* -0.679 -0.290 0.487 -0.473 0.188 0.408 0.614 (2.459**) 0.024 -0.806 0.595 -0.846 1.069 -0.467 -0.549 -1.403 0.806 0.365 0.137 0.840 -1.454 (2.070**) 1.974** 0.926 1.925* -0.296 0.239 -0.102 0.173 0.430 2.058** -0.322 1.446 0.434 -0.375 0.724 (2.068**) 0.077 2.349** 0.667 -0.039 0.434 -0.540 -0.051 -0.545 -0.141 -1.458 -0.891 -0.184 -0.088 1.164 -0.057 0.349 -1.111 -0.463 -10 1 0.188 -0.545 0.047 0.061 0.320 -0.233 1.941* -0.251 -0.967 -0.014 -0.139 0.822 1.224 -0.357 -0.381 0.557 -1.081 0.232 -0.438 0.010 0.198 -0.744 (1.666**) 1.819* 1.077 3.989** 1.122 0.245 0.135 0.133 0.930 0.850 0.133 -0.742 0.445 HK/UK 6.416** 1.970** -0.746 0.842 1.067 0.706 -0.522 -1.222 0.532 -0.265 0.865 -1.132 0.418 0.518 1.138 -0.752 -1.313 -0.799 0.047 HK/US 4.083** 0.296 0.571 -0.706 -0.194 0.440 (2.037**) 0.783 0.950 1.001 1.583 0.981 0.122 1.028 0.720 0.708 1.364 0.112 (2.121**) -0.271 HK/WD 10.189** 3.306** -0.700 0.453 -0.389 1.067 1.842* -0.912 -0.944 0.210 1.503 0.932 0.557 -0.377 -0.512 0.879 -0.806 (1.715*) -0.084 -0.390 0.355 -0.336 (1.748**) 0.700 1.360 0.604 1.007 -0.410 -0.436 0.445 0.237 0.852 0.175 -1.479 0.838 0.135 0.310 1.071 -0.320 -0.169 -0.826 -0.024 -0.757 -0.108 -0.151 0.918 0.963 (1.854*) -1.943 0.712 -1.052 HK/SG 11.116** 1.946* 0.610 -0.151 -0.785 0.748 1.466 JP/SG 5.184** -0.506 -1.128 0.055 JP/UK 4.317** 1.097 0.498 1.674* 2.111** 0.300 JP/US 1.230 -0.671 -0.147 0.545 0.708 -0.483 0.037 0.757 1.191 1.472 -0.702 0.716 -1.373 1.578 0.494 -0.355 -0.273 (1.780*) -0.120 0.553 JP/WD 5.637** 0.481 -0.818 -0.671 0.818 1.111 0.312 0.481 0.593 -0.467 -0.365 0.283 (1.929**) 0.082 1.128 -0.734 0.799 -0.257 (2.500**) 0.573 2.402** -0.167 SG/UK 6.665** 1.264 0.404 -0.404 0.188 1.880* 0.228 -0.948 0.983 0.997 -0.241 0.306 (2.296**) -1.222 0.408 0.330 0.665 -0.610 -0.646 1.791* 0.712 SG/US 0.410 0.347 0.455 -1.032 -0.006 0.628 -1.160 0.184 1.550 1.126 1.444 -1.395 2.149** 0.801 0.190 1.042 0.257 -0.481 -1.366 -0.047 SG/WD 8.608** 3.210** 0.430 -0.940 -0.775 0.665 1.570 0.477 1.285 1.099 -0.398 1.452 -1.040 -0.192 0.075 -0.563 0.506 -0.595 -0.808 1.167 2.164** UK/US 6.550** 2.337** 1.140 0.775 -0.748 0.956 -0.822 1.114 1.432 2.614** 0.636 -0.334 0.137 1.315 2.262** -0.304 -0.328 -0.632 0.210 (2.667**) -0.039 UK/WD 8.455** 1.758* -0.810 -0.077 -0.161 1.264 0.889 1.073 0.135 -0.885 -0.204 0.714 -0.881 -0.518 0.901 -0.634 -1.209 (1.803*) -0.118 -1.277 0.147 US/WD 9.002** 0.561 0.418 1.393 0.483 1.315 -1.352 -0.808 -1.631 -0.801 -0.589 0.080 -0.653 -1.181 -1.629 -1.307 1.370 0.351 4.899** 0.877 -0.836 -0.116 Notes: Each entry gives the causality in mean t-statistics between the real estate / global stock markets of the countries noted in the first column. The six real estate markets are Australia (AU), Hong Kong (HK), Japan (JP), Singapore (SG), the UK and the USA; the global stock market (WD) is represented by the MSCI World Index. The lag k denotes the number of weeks by which the return of the first market lags or leads that of the second market. A negative (positive) value of k indicates lags (leads). Significant t-statistics with k<0 suggests that the first market causes the second market. Significance with k>0 indicates causality in the opposite direction. ***, ** - indicates statistical significance at one and five percent levels. 35 Table 8(b) Causality in variance results: 4/9/98 – 10/2/06 (post-crisis period) -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 1 2 3 4 5 6 7 8 9 10 AU/HK 4.038** Lag 0 1.032 0.208 -0.491 -1.599 -0.320 -0.318 0.767 -0.475 -1.617 -0.245 1.530 1.442 -0.595 -0.506 0.534 -0.608 -0.078 -0.673 0.573 -0.128 AU/JP -0.116 0.905 -0.206 -0.445 -0.226 -0.441 0.292 1.175 -0.049 0.141 0.067 0.610 -0.024 -1.311 0.049 -0.377 -1.040 -0.343 0.263 0.300 AU/SG 3.767** 1.354 0.239 -0.804 -0.198 -1.346 -0.290 1.903* -0.165 -0.496 -1.056 0.092 0.840 -0.993 -0.145 1.289 0.432 1.136 -1.342 0.363 -0.381 AU/UK 6.762** 0.616 0.920 -1.077 -0.801 0.569 0.208 -0.147 -0.494 -1.326 0.165 1.432 2.353** -1.140 1.044 0.098 -1.271 -0.232 -0.691 2.537** -0.371 AU/US 3.618** 0.418 0.377 -0.551 -1.046 0.224 -0.510 0.579 -0.165 -1.177 (1.646*) 0.314 0.485 -0.583 0.620 1.183 -1.132 2.245** -0.463 -0.775 0.192 AU/WD 5.904** 0.267 0.922 -0.106 -0.836 -1.056 -0.106 1.585 -1.585 0.067 1.615 -0.959 HK/JP 1.595 -1.458 0.857 0.642 -0.806 -0.200 0.718 -0.120 0.084 -1.807 5.429** 5.386** 1.146 0.371 1.138 0.365 -0.145 -0.547 -0.275 0.198 0.390 1.954* -0.573 -0.487 -0.775 0.567 -0.457 -0.538 -0.700 0.188 0.253 HK/SG 10.406** 3.118** 0.441 -1.340 -0.771 -0.318 -0.481 -0.649 -0.031 -0.477 -0.744 7.105** 0.891 -0.799 0.122 1.252 1.075 -0.616 -0.369 -1.201 -1.034 HK/UK 5.822** 1.990** -0.357 0.055 -0.789 -0.540 -0.408 0.031 0.626 -0.390 -0.422 2.151** 2.688** -1.340 HK/US 6.663** 3.999** -0.263 -0.067 -0.977 -0.145 -0.824 -0.092 0.135 -1.101 0.251 HK/WD 8.465** 3.569** -1.258 -0.373 -0.359 -0.644 0.006 -0.434 -0.379 2.980** 2.386** 0.194 -0.212 -0.924 0.438 -1.215 -0.383 -0.526 -0.593 -0.447 -0.498 -0.871 0.237 -0.789 -0.200 -1.107 -0.999 0.602 1.099 1.748* 0.306 -0.290 -0.697 0.004 0.230 1.542 1.638 0.549 -1.154 -0.408 -0.077 0.261 0.400 0.041 0.108 3.924** JP/SG 3.083** 0.567 -0.871 0.469 -1.338 -0.039 -0.069 -0.810 -0.300 -0.387 0.137 6.540** 2.588** JP/UK 1.458 0.196 -1.120 0.024 0.237 -1.024 -0.110 1.156 0.579 -0.693 -0.396 0.018 -0.290 -0.186 -0.175 0.351 -1.275 -0.247 1.815* -0.975 JP/US 1.742* 0.334 0.226 -0.424 -1.116 -0.153 0.371 -0.557 0.494 -1.079 -0.914 0.192 0.483 0.027 -0.267 4.499** -0.296 0.043 -0.661 -0.459 -0.651 JP/WD 0.767 2.374** -0.812 0.078 -0.549 -0.871 -0.130 0.004 -0.216 -0.108 -0.479 2.590** 2.773** -0.485 -0.061 0.251 -0.267 1.995** -0.736 SG/UK 4.874** SG/US 0.616 4.776** 2.804** 1.519 1.397 0.549 -0.485 0.441 -0.591 -0.500 1.024 -0.491 -0.763 -0.316 -0.834 2.831** 0.169 -0.630 -0.063 -0.659 0.524 -0.826 0.029 0.318 -0.824 0.893 -0.131 -0.059 0.383 1.083 -0.361 -0.777 0.343 -0.163 3.789** 0.781 0.022 -0.014 -0.675 -1.067 -0.928 -0.536 -0.512 -0.879 SG/WD 8.298** 1.752* 2.384** 0.220 0.100 0.577 0.049 0.020 -0.171 -0.310 -0.045 3.349** -0.141 0.177 -0.073 -0.451 0.328 -0.334 -0.534 -0.750 -0.897 UK/US 8.641** 1.381 -0.230 -0.538 -0.728 -0.294 0.757 0.740 -0.349 -0.983 3.024** -0.365 -0.198 -0.549 0.508 -0.983 -1.470 0.318 -0.165 -0.298 UK/WD 9.232** 0.905 1.262 0.926 -0.663 -0.308 -1.071 0.332 -0.206 -0.781 -0.842 2.716** -0.279 0.257 -0.775 -0.169 -1.309 -0.952 -0.889 -0.491 -0.249 US/WD 8.826** 1.809* 1.676* 0.512 0.624 0.249 -0.648 -1.270 -0.041 -0.818 -0.918 2.975** -0.336 0.285 -0.726 -1.179 -0.781 -0.934 -0.606 -0.632 -0.306 0.469 Notes: Each entry gives the causality in variance t-statistics between the real estate / global stock markets of the countries noted in the first column. The six real estate markets are Australia (AU), Hong Kong (HK), Japan (JP), Singapore (SG), the UK and the USA; the global stock market (WD) is represented by the MSCI World Index. The lag k denotes the number of weeks by which the variance of the first market lags or leads that of the second market. A negative (positive) value of k indicates lags (leads). Significant t-statistics with k<0 suggests that the first market causes the second market. Significance with k>0 indicates causality in the opposite direction. ***, ** - indicates statistical significance at one and five percent levels. 36 Table 9 Direction of causality in mean and causality in variance: pre- and post-crisis periods Market-pair Australia / Hong Kong Australia / Japan Australia / Singapore Australia / UK Australia / USA Australia / MSCI Hong Kong / Japan Hong Kong / Singapore Hong Kong / UK Hong Kong / US Hong Kong / MSCI Japan / Singapore Japan/ UK Japan / USA Japan / MSCI Singapore / UK Singapore / USA Singapore /MSCI UK / USA UK / MSCI US / MSCI Causality-in-mean PRE-CRISIS POST-CRISIS Bilateral Bilateral Bilateral Uni-directional Uni-directional Bilateral Uni-directional Bilateral No Uni-directional Bilateral Uni-directional Bilateral Uni-directional No Bilateral Bilateral Uni-directional Uni-directional Bilateral Bilateral Bilateral Bilateral Uni-directional Bilateral Bilateral Uni-directional Uni-directional Bilateral Uni-directional Bilateral Bilateral Uni-directional Uni-directional Bilateral Bilateral No Bilateral Uni-directional Bilateral Uni-directional No Causality-in-variance PRE-CRISIS POST-CRISIS Bilateral No Bilateral No Uni-directional Uni-directional Bilateral Uni-directional No Bilateral Uni-directional No Uni-directional Bilateral Uni-directional Bilateral Bilateral Bilateral Uni-directional Bilateral Uni-directional Bilateral Bilateral Uni-directional Bilateral Uni-directional Uni-directional Uni-directional Uni-directional Bilateral Uni-directional Uni-directional Bilateral Bilateral Bilateral Bilateral Uni-directional Uni-directional Uni-directional Uni-directional Uni-directional Bilateral Notes: Based on the results reported in Tables 7(a) – 8(b), the direction of causality (in mean and variance) for all 21 market pairs is classified into three groups, and the results for both pre- and post-crisis periods are compared. “Bilateral” causality means there are lead-lag interactions between the two markets (i.e. from lagged A to B and from lagged B to A); “Uni-directional” causality means either A causes B or B causes (and not both): “No” means that there is no lead-lag interaction between the two markets. 37 Table 10 (a) Causality in mean patterns: pre- and post-crisis periods Australia Australia HongKong Japan Singapore UK US MSCI PrePostPrePostPrePostPrePostPrePostPrePostPrePost- 0, -1 0,-1,-4 0, -2 0 NIL 0,-1,-7 0 0,-2,-4 0 0 0, -9 0 Hong Kong 0,-7,-9 0, -6 0, -1, -3 0 0 0,-1 0, -9 0 0,-2,-7,-10 0, -9 0, -3 0, -8 Japan 0, -1 0, -8 0, -9 0,-1,-2,-10 0,-2,-9 0 0,-5,-8 0, -8 0 -8 0,-6,-10 0,-2,-8,-10 Singapore -9 0,-4,-6,-8,-9 0 0,-1,-8,-9 0,-3,-4 0, -7 0, -7 0, -2,-9 0, -2 0, -3 0,-2 0, -10 UK 0, -1 0, -1 0,-1,-9 0, -1 0,-1,-9 0, -3,-4 0,-1,-9 0,-5 US 0 0, -2 0 0, -7 0,-1,-5,-7 NIL 0 0 0 0, -1,-9 0 0,-4,-9 0, -1 0, -7 MSCI 0,-1,-4 0,-1,-2,-4 0, -9 0,-1,-6 0,-1,-4 0 0,-1,-2,-9 0, -1 0 0, -1 0, -1 0 0 0 Notes: This summary is constructed based on the results reported in Tables 7(a) and 8 (a). 0 indicates contemporaneous correlation between two markets exist. NIL indicates no relationship at all between the two markets. –k indicates the presence of correlation at lag k. For example: 0,-7, -9 in the first cell of the Hong Kong column (pre-crisis) indicates there is contemporaneous relation between the means of the Australia and Hong Kong real estate markets, and the mean returns of the Australia market at lags 7- and 9week have an impact on the current mean return of the Hong Kong market in the pre-crisis period. Table 10 (b) Causality in variance patterns: pre- and post-crisis periods Australia Australia HongKong Japan Singapore UK US MSCI PrePostPrePostPrePostPrePostPrePostPrePostPrePost- -5 0 -2, -8 NIL NIL 0 -7,-10 0,-2,-9 NIL 0, -10 -8 0 Hong Kong -4 0 -7, -10 0, -1 0,-2-3 0, -1 -2,-4,-6 0,-1,-2 -2,-3,-7 0,-1,-2 0,-2,-6,-7,-8 0,-1 Japan -1 NIL NIL 0, -1 -7 0,-1,-2,-5 -7,-8 -9 NIL 0,-5 0 -1,-2,-9 Singapore -1 0, -7 0 0, -1 -3,-7,-8 0 -6 0,-1 0,-3,-7,-8 0,-1 0,-3,-7,-8 0,-1 UK -6, -7 0 -8 0, -1 -2, -4 NIL NIL 0 NIL 0,-1 0 0,-1 US NIL 0, -10 NIL 0, -1 -3 0 0, -4 0,-1 -1 0 MSCI NIL 0 0 0, -1 0,-4,-5 -1 0, -10 0,-1-2 0,-1,-6,-7,-8 0 0 -1,-2 0,-1,-4 -1 Notes: This summary is constructed based on the results reported in Tables 7(b) and 8 (b). 0 indicates contemporaneous correlation between two markets exist. NIL indicates no relationship at all between the two markets. –k indicates the presence of correlation at lag k. For example: -4 in the first cell of the Hong Kong column (pre-crisis) indicates there is no contemporaneous relation between the variances of the Australia and Hong Kong real estate markets, and the return variance of the Australia market at lag 4-week has an impact on the current return variance of the Hong Kong market in the pre-crisis period. 38 Table 11 Vector Autoregressive Estimates: 5/1/90 – 10/2/06 MSCI 0.939019 [ 24.3382]* MSCI(-2) 0.04487 [ 1.17108] AU(-1) -0.003224 [-0.13579] AU(-2) 0.007828 [ 0.32912] HK(-1) 0.0095 [ 1.68495]* HK(-2) -0.007759 [-1.38146] JP(-1) 0.003919 [ 1.56843] JP(-2) -0.001063 [-0.42672] SG(-1) -0.002653 [-0.85458] SG(-2) 0.000376 [ 0.12094] UK(-1) 0.007956 [ 0.24004] UK(-2) -0.026472 [-0.79518] US(-1) 0.177891 [ 5.85865]* US(-2) -0.171584 [-5.67478]* C 7.53E-06 [ 1.54565] R-squared 0.980109 Adj. R-squared 0.979769 Sum sq. resids 5.29E-07 S.E. equation 2.54E-05 F-statistic 2886.014 Log likelihood 7657.505 Akaike AIC -18.3054 Schwarz SC -18.22048 Mean dependent 0.000366 S.D. dependent 0.000179 Determinant resid covariance Log likelihood Akaike information criterion Schwarz criterion MSCI(-1) Australia -0.055085 [-0.94805] 0.048385 [ 0.83855] 0.936922 [ 26.2085]* -0.011951 [-0.33364] 0.000839 [ 0.09887] -0.003681 [-0.43524] 0.001174 [ 0.31190] 4.49E-05 [ 0.01197] 0.007018 [ 1.50122] -0.003948 [-0.84392] 0.046319 [ 0.92797] -0.046856 [-0.93458] 0.046898 [ 1.02560] -0.054826 [-1.20404] 3.30E-05 [ 4.50107]* 0.885058 0.883096 1.20E-06 3.83E-05 451.0038 7315.629 -17.48654 -17.40161 0.000427 0.000112 1.74E-58 47232.23 -112.8796 -112.2851 HongKong 0.09257 [ 0.30126] -0.074562 [-0.24434] 0.274388 [ 1.45134] -0.214902 [-1.13445] 1.011039 [ 22.5158]* -0.058074 [-1.29827] 0.020789 [ 1.04457] -0.013724 [-0.69166] -0.047565 [-1.92393]* 0.04454 [ 1.80050]* 0.201405 [ 0.76297] -0.218147 [-0.82276] 0.789645 [ 3.26533]* -0.827158 [-3.43490]* 5.89E-05 [ 1.52007] 0.918253 0.916857 3.36E-05 0.000202 657.9219 5924.907 -14.15547 -14.07054 0.001732 0.000702 Japan 0.197849 [ 0.36367] -0.104445 [-0.19332] -0.405881 [-1.21259] 0.481385 [ 1.43532] 0.013281 [ 0.16705] -0.034573 [-0.43655] 1.010169 [ 28.6694]* -0.10778 [-3.06801]* 0.068704 [ 1.56961] -0.025759 [-0.58815] -0.018208 [-0.03896] 0.108376 [ 0.23087] 0.944668 [ 2.20640]* -0.963882 [-2.26079]* 7.20E-05 [ 1.04848] 0.900566 0.898868 0.000105 0.000358 530.4757 5447.915 -13.01297 -12.92805 0.002417 0.001126 Singapore -0.212607 [-0.39159] 0.28807 [ 0.53428] 0.186872 [ 0.55942] -0.075385 [-0.22523] 0.159888 [ 2.01524]* -0.099342 [-1.25693] 0.001806 [ 0.05137] 0.008785 [ 0.25057] 0.92231 [ 21.1138]* 0.015524 [ 0.35518] -0.17314 [-0.37121] 0.004109 [ 0.00877] 1.224346 [ 2.86543]* -1.227306 [-2.88449]* 1.79E-05 [ 0.26154] 0.950072 0.94922 0.000105 0.000358 1114.555 5449.607 -13.01702 -12.9321 0.002011 0.001587 UK 0.060304 [ 1.44762] -0.054691 [-1.32204] 0.029813 [ 1.16318] -0.026943 [-1.04914] -0.002303 [-0.37836] -0.00121 [-0.19961] 0.000436 [ 0.16151] 0.001739 [ 0.64662] 0.00138 [ 0.41167] -0.001322 [-0.39411] 0.966742 [ 27.0141]* 0.021917 [ 0.60976] 0.085444 [ 2.60628]* -0.086559 [-2.65144]* 4.52E-06 [ 0.86016] 0.978781 0.978419 6.17E-07 2.74E-05 2701.748 7593.472 -18.15203 -18.06711 0.000597 0.000187 US 0.07479 [ 1.63997] -0.067342 [-1.48696] -0.023014 [-0.82020] 0.064341 [ 2.28853]* 0.003477 [ 0.52170] -0.004651 [-0.70065] 7.38E-05 [ 0.02499] 0.001314 [ 0.44622] 0.001313 [ 0.35786] -0.002892 [-0.78765] 0.091267 [ 2.32958]* -0.102044 [-2.59321]* 1.101469 [ 30.6898]* -0.150368 [-4.20734]* 3.39E-06 [ 0.58949] 0.948411 0.94753 7.39E-07 3.00E-05 1076.772 7517.88 -17.97097 -17.88605 0.000314 0.000131 Notes: The number of the lags in the VAR is 2. T-statistics are indicated in the brackets. The 10% critical value for a two-tailed test is 1.645. Statistical significance is denoted by * 39 Table 12 Variance Decomposition of forecast error of conditional volatilities (full-sample period) Market MSCI Australia HongKong Japan Singapore UK USA Horizon (wks) 1 4 8 12 16 20 24 1 4 8 12 16 20 24 1 4 8 12 16 20 24 1 4 8 12 16 20 24 1 4 8 12 16 20 24 1 4 8 12 16 20 24 1 4 8 12 16 20 24 MSCI 100.00 95.48 92.91 90.61 88.54 86.72 85.12 2.92 2.90 2.77 2.63 2.52 2.44 2.38 8.08 10.87 11.55 11.70 11.69 11.64 11.58 2.19 3.98 5.39 6.49 7.39 8.13 8.74 4.75 6.32 7.41 8.05 8.45 8.67 8.79 5.03 8.53 9.27 9.53 9.69 9.85 10.03 8.06 12.22 13.33 13.58 13.52 13.35 13.14 Australia 0.00 0.02 0.20 0.47 0.75 1.00 1.20 97.08 96.24 95.33 94.12 92.73 91.30 89.95 1.54 2.82 3.72 4.45 4.96 5.28 5.46 0.19 0.09 0.41 0.97 1.56 2.08 2.51 0.48 1.03 1.85 2.66 3.35 3.89 4.28 0.24 0.62 0.74 0.82 0.87 0.90 0.92 0.01 0.24 2.12 4.91 7.76 10.24 12.22 Hongkong 0.00 0.31 0.25 0.19 0.15 0.12 0.11 0.00 0.19 0.30 0.44 0.64 0.92 1.26 90.38 84.67 82.87 81.94 81.32 80.88 80.57 1.20 2.54 3.49 4.32 5.09 5.80 6.45 31.99 37.80 41.08 43.23 44.75 45.87 46.71 0.01 0.04 0.35 0.85 1.40 1.94 2.43 0.27 0.40 0.24 0.29 0.46 0.66 0.83 Japan 0.00 0.71 2.30 3.86 5.11 6.01 6.62 0.00 0.05 0.19 0.34 0.45 0.52 0.56 0.00 0.21 0.53 0.82 1.05 1.21 1.32 96.41 92.22 88.18 84.31 80.81 77.83 75.38 0.00 0.05 0.24 0.48 0.68 0.82 0.91 0.01 0.21 1.10 2.16 3.14 3.97 4.65 0.03 0.14 0.65 1.23 1.73 2.09 2.32 Singapore 0.00 0.18 0.62 1.03 1.34 1.57 1.74 0.00 0.44 1.27 2.33 3.46 4.53 5.45 0.00 0.29 0.29 0.26 0.23 0.22 0.23 0.00 0.56 1.77 3.17 4.46 5.50 6.29 62.77 53.95 48.40 44.43 41.39 39.00 37.11 0.01 0.01 0.02 0.05 0.10 0.19 0.29 0.00 0.01 0.07 0.12 0.13 0.12 0.11 UK 0.00 0.02 0.10 0.43 0.96 1.67 2.53 0.00 0.09 0.08 0.07 0.07 0.08 0.11 0.00 0.12 0.11 0.09 0.08 0.09 0.10 0.00 0.02 0.06 0.08 0.08 0.08 0.08 0.00 0.01 0.11 0.35 0.70 1.14 1.64 94.71 89.83 87.73 85.91 84.20 82.63 81.21 0.40 1.36 1.13 0.90 0.78 0.79 0.91 US 0.00 3.27 3.62 3.42 3.15 2.90 2.69 0.00 0.09 0.07 0.08 0.14 0.21 0.29 0.00 1.02 0.92 0.74 0.66 0.68 0.75 0.00 0.59 0.70 0.67 0.62 0.59 0.56 0.00 0.84 0.90 0.79 0.68 0.60 0.56 0.00 0.76 0.79 0.70 0.60 0.53 0.47 91.23 85.62 82.46 78.97 75.63 72.76 70.46 Notes: The entries indicate the percentage of conditional volatility of returns of a particular market displayed at the first column explained by conditional volatilities of returns of all markets. 40 Table 13 Generalized Impulse Response of all markets to a one standard deviation shock in each of the markets Market MSCI Horizon (wks) 1 4 8 12 16 20 24 Australia 1 4 8 12 16 20 24 HongKong 1 4 8 12 16 20 24 Japan 1 4 8 12 16 20 24 Singapore 1 4 8 12 16 20 24 UK 1 4 8 12 16 20 24 US 1 4 8 12 16 20 24 MSCI 0.410 0.218 0.162 0.124 0.098 0.079 0.064 0.050 0.031 0.021 0.013 0.009 0.006 0.004 0.083 0.068 0.054 0.039 0.029 0.022 0.016 0.044 0.044 0.043 0.037 0.032 0.028 0.025 0.065 0.055 0.047 0.039 0.031 0.024 0.020 0.065 0.062 0.052 0.043 0.034 0.030 0.027 0.085 0.072 0.061 0.043 0.031 0.024 0.017 Australia 0.050 0.034 0.028 0.025 0.022 0.020 0.018 0.407 0.183 0.095 0.063 0.045 0.036 0.027 0.049 0.042 0.030 0.024 0.020 0.017 0.014 0.020 0.013 0.018 0.019 0.019 0.019 0.019 0.031 0.030 0.026 0.024 0.021 0.019 0.017 0.025 0.025 0.018 0.014 0.012 0.011 0.010 0.017 0.026 0.037 0.037 0.037 0.035 0.033 HongKong 0.084 0.060 0.040 0.027 0.021 0.017 0.015 0.050 0.039 0.024 0.019 0.016 0.016 0.016 0.404 0.192 0.115 0.076 0.058 0.045 0.038 0.045 0.044 0.035 0.029 0.027 0.026 0.026 0.197 0.131 0.093 0.068 0.055 0.048 0.042 0.018 0.012 0.000 -0.004 -0.006 -0.007 -0.007 0.039 0.030 0.014 0.006 0.002 0.000 0.000 Japan 0.043 0.046 0.044 0.039 0.035 0.032 0.029 0.020 0.017 0.015 0.012 0.011 0.010 0.009 0.044 0.039 0.028 0.022 0.020 0.018 0.016 0.407 0.167 0.084 0.056 0.042 0.035 0.029 0.031 0.026 0.023 0.020 0.019 0.017 0.016 0.012 0.021 0.024 0.024 0.023 0.022 0.022 0.020 0.021 0.021 0.020 0.018 0.017 0.015 Singapore 0.064 0.038 0.018 0.011 0.007 0.005 0.004 0.031 0.038 0.037 0.033 0.030 0.028 0.027 0.195 0.104 0.071 0.051 0.038 0.031 0.027 0.031 0.048 0.049 0.044 0.040 0.036 0.034 0.411 0.204 0.132 0.091 0.068 0.054 0.044 0.012 0.011 0.005 0.002 0.002 0.002 0.002 0.029 0.019 0.007 0.002 0.001 0.001 0.002 UK 0.066 0.043 0.027 0.012 0.003 0.004 0.008 0.026 0.022 0.014 0.007 0.002 0.001 0.002 0.018 0.024 0.017 0.010 0.005 0.003 0.001 0.013 0.016 0.017 0.014 0.010 0.007 0.004 0.012 0.008 0.000 -0.006 -0.009 -0.012 -0.013 0.409 0.225 0.171 0.134 0.110 0.088 0.074 0.038 0.040 0.027 0.014 0.005 -0.001 -0.005 USA 0.084 0.087 0.058 0.040 0.030 0.024 0.020 0.017 0.016 0.005 0.000 -0.002 -0.003 -0.004 0.038 0.045 0.025 0.012 0.005 0.002 -0.001 0.020 0.032 0.021 0.014 0.011 0.008 0.007 0.029 0.039 0.025 0.016 0.011 0.007 0.004 0.036 0.043 0.030 0.020 0.015 0.013 0.011 0.411 0.190 0.116 0.073 0.052 0.039 0.031 Notes: The entries contain the normalized impulse response coefficients. The response patterns are simulated by introducing one s.d. shocks in the system and tracing out the normalized responses of the markets in the system over different time horizons. 41 Table 14 Variance Decomposition of forecast error of conditional volatilities (pre- and post-crisis periods) Pre-crisis period: 5/1/90 - 27/6/97 Post-crisis period: 1/1/99 - 10/2/06 Lag(wks) MSCI Australia HongKong Japan Singapore UK USA MSCI Australia HongKong Japan Singapore (a) Percentage of conditional volatility of MSCI World market returns explained by conditional volatilities of real estate market returns of 1 100.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 0.00 0.00 0.00 0.00 4 95.21 0.21 0.00 0.16 0.06 0.20 4.16 94.77 0.44 0.66 0.87 0.44 8 94.62 0.21 0.00 0.68 0.37 0.27 3.85 93.05 0.50 0.41 2.92 0.32 12 93.92 0.26 0.00 1.22 0.75 0.38 3.46 90.99 0.41 0.43 5.30 0.24 16 93.26 0.30 0.00 1.62 1.09 0.55 3.16 88.66 0.33 0.67 7.60 0.26 20 92.64 0.33 0.00 1.91 1.37 0.80 2.94 86.38 0.28 0.98 9.65 0.37 24 92.05 0.36 0.00 2.13 1.57 1.11 2.77 84.29 0.24 1.31 11.39 0.54 (b) Percentage of conditional volatility of Australian real estate market returns explained by conditional volatilities of market returns of 1 0.17 99.83 0.00 0.00 0.00 0.00 0.00 3.29 96.71 0.00 0.00 0.00 4 0.31 96.08 0.05 0.59 1.57 0.69 0.71 3.35 95.86 0.07 0.32 0.19 8 0.30 95.07 0.07 0.58 2.20 1.11 0.69 3.66 94.91 0.21 0.76 0.19 12 0.30 94.60 0.07 0.58 2.38 1.38 0.68 3.80 94.50 0.24 0.98 0.19 16 0.30 94.36 0.08 0.58 2.43 1.57 0.68 3.88 94.30 0.24 1.10 0.19 20 0.31 94.18 0.08 0.58 2.44 1.72 0.69 3.92 94.18 0.24 1.16 0.19 24 0.33 94.02 0.09 0.58 2.45 1.85 0.70 3.95 94.11 0.24 1.20 0.19 (c) Percentage of conditional volatility of Hong Kong real estate market returns explained by conditional volatilities of market returns of 1 3.77 0.00 96.23 0.00 0.00 0.00 0.00 13.43 1.96 84.61 0.00 0.00 4 4.86 0.11 94.44 0.20 0.21 0.07 0.11 15.58 5.10 74.66 3.22 0.01 8 4.85 0.14 94.15 0.27 0.29 0.11 0.18 15.63 5.82 71.60 5.21 0.02 12 4.65 0.18 93.84 0.29 0.64 0.20 0.21 15.74 5.71 69.60 7.03 0.04 16 4.42 0.22 93.49 0.27 1.02 0.36 0.22 15.94 5.57 67.76 8.71 0.06 20 4.21 0.27 93.10 0.26 1.36 0.60 0.22 16.17 5.43 66.08 10.23 0.08 24 4.03 0.31 92.65 0.24 1.63 0.93 0.22 16.42 5.30 64.58 11.57 0.10 (d) Percentage of conditional volatility of Japaneses real estate market returns explained by conditional volatilities of market returns of 1 18.67 0.08 0.00 81.25 0.00 0.00 0.00 0.04 0.71 8.35 90.90 0.00 2 17.33 0.20 0.08 81.17 0.11 0.01 1.11 0.63 0.66 10.68 87.86 0.03 8 24.13 0.98 0.10 73.35 0.36 0.16 0.91 1.84 0.90 13.14 82.93 0.05 12 27.22 1.19 0.16 69.83 0.51 0.23 0.87 2.55 0.86 13.48 81.68 0.05 16 29.29 1.22 0.24 67.48 0.64 0.27 0.85 3.36 0.85 13.59 80.67 0.06 20 30.77 1.21 0.32 65.81 0.76 0.28 0.84 4.21 0.85 13.54 79.83 0.06 24 31.88 1.20 0.40 64.56 0.85 0.29 0.83 5.06 0.84 13.37 79.12 0.07 UK USA 0.00 0.74 0.50 0.40 0.35 0.31 0.29 0.00 2.07 2.30 2.24 2.13 2.02 1.92 0.00 0.06 0.10 0.11 0.12 0.12 0.12 0.00 0.16 0.17 0.18 0.18 0.18 0.18 0.00 0.13 0.31 0.44 0.52 0.58 0.61 0.00 1.29 1.41 1.43 1.44 1.43 1.43 0.00 0.10 1.05 1.25 1.31 1.32 1.31 0.00 0.05 0.10 0.13 0.17 0.20 0.22 42 Table 14 (Continued) Pre-crisis period: 5/1/90 - 27/6/97 Post-crisis period: 1/1/99 - 10/2/06 Lag(wks) MSCI Australia HongKong Japan Singapore UK USA MSCI Australia HongKong Japan Singapore (e) Percentage of conditional volatility of Singapore real estate market returns explained by conditional volatilities of market returns of 1 6.72 0.02 7.56 0.07 85.63 0.00 0.00 7.99 1.12 12.95 0.69 77.25 4 9.51 0.14 13.40 1.29 74.79 0.07 0.79 8.53 1.04 19.14 6.04 63.81 8 14.63 0.38 18.13 2.19 61.97 0.62 2.08 8.22 0.92 19.36 11.13 57.74 12 18.42 0.47 21.34 2.82 52.40 1.57 2.98 8.38 0.79 19.02 15.75 52.96 16 20.86 0.48 23.53 3.23 45.91 2.59 3.41 8.79 0.73 18.56 19.88 48.74 20 22.31 0.47 25.08 3.48 41.58 3.55 3.53 9.33 0.69 18.06 23.46 45.08 24 23.09 0.46 26.24 3.62 38.66 4.42 3.51 9.95 0.66 17.54 26.51 41.94 (f) Percentage of conditional volatility of UK real estate market returns explained by conditional volatilities of market returns of 1 3.83 0.17 0.15 0.90 0.10 94.85 0.00 10.66 2.97 0.47 0.14 0.19 4 6.62 1.26 0.38 0.52 0.07 90.66 0.48 13.26 9.58 0.92 0.24 1.12 8 9.09 1.36 0.62 0.30 0.04 87.55 1.04 13.17 9.79 1.48 0.54 1.14 12 11.38 1.24 0.88 0.27 0.04 84.64 1.55 13.20 9.76 1.88 0.70 1.13 16 13.62 1.10 1.15 0.25 0.07 81.82 1.99 13.20 9.75 2.05 0.79 1.13 20 15.82 0.98 1.41 0.22 0.11 79.09 2.36 13.20 9.75 2.13 0.83 1.12 24 17.99 0.88 1.66 0.19 0.17 76.44 2.67 13.20 9.74 2.16 0.85 1.12 (g) Percentage of conditional volatility of US real estate market returns explained by conditional volatilities of market returns of 1 7.35 0.14 0.05 1.05 0.04 0.34 91.04 11.50 0.25 1.46 0.00 0.24 4 15.30 0.12 0.34 4.20 0.06 0.24 79.75 11.01 0.49 2.27 0.31 0.74 8 21.14 0.15 0.69 5.50 0.08 0.26 72.18 11.06 0.50 2.33 0.37 0.82 12 26.03 0.18 1.09 6.16 0.11 0.30 66.13 11.08 0.50 2.35 0.39 0.88 16 30.14 0.20 1.47 6.50 0.16 0.35 61.18 11.09 0.50 2.35 0.39 0.91 20 33.52 0.21 1.81 6.66 0.20 0.42 57.18 11.10 0.50 2.36 0.39 0.93 24 36.24 0.21 2.11 6.72 0.24 0.50 53.98 11.11 0.50 2.36 0.39 0.94 UK USA 0.00 0.53 1.67 2.14 2.35 2.43 2.45 0.00 0.90 0.96 0.96 0.95 0.95 0.94 85.56 74.47 73.45 72.89 72.64 72.53 72.48 0.00 0.42 0.43 0.44 0.44 0.44 0.44 4.08 4.74 4.74 4.74 4.73 4.73 4.73 82.48 80.45 80.17 80.07 80.02 79.99 79.96 Notes: The number of the lags in the VAR is 2 43 AUSTRALIA HONG KONG JAPAN SINGAPORE UK 2005-12-29 2005-6-29 2004-12-29 2004-6-29 2003-12-29 2003-6-29 2002-12-29 2002-6-29 2001-12-29 2001-6-29 2000-12-29 2000-6-29 1999-12-29 1999-6-29 1998-12-29 1998-6-29 1997-12-29 1997-6-29 1996-12-29 1996-6-29 1995-12-29 1995-6-29 1994-12-29 1994-6-29 1993-12-29 1993-6-29 1992-12-29 1992-6-29 1991-12-29 1991-6-29 1990-12-29 1990-6-29 1989-12-29 INDEX LEVEL Figure 1 EPRA FTSE/NAREIT INDEXES 3500 3000 2500 2000 1500 1000 500 0 WEEK USA 44 Figure 2: Comparison of t-statistics for contemporaneous conditional correlation in mean: pre-and post-crisis periods 14.000 12.000 10.000 t-statistic 8.000 6.000 4.000 2.000 0.000 AU/HK AU/JP AU/SG AU/UK AU/US AU/WD HK/JP HK/SG HK/UK HK/US HK/WD JP/SG JP/UK JP/US JP/WD SG/UK SG/US SG/WD UK/US UK/WD US/WD Market pair Pre-crisis Post-crisis Figure 3: Comparison of t-statistics for contemporanoues conditional correlation in variance: pre- and post-crisis periods 12.000 10.000 8.000 t-statistic 6.000 4.000 2.000 0.000 -2.000 AU/HK AU/JP AU/SG AU/UK AU/US AU/WD HK/JP HK/SG HK/UK HK/US HK/WD JP/SG JP/UK JP/US JP/WD SG/UK SG/US SG/WD UK/US UK/WD US/WD Market pair Pre-crisis Post-crisis 45 Figure 4 Conditional Volatility Graphs from the MEGARCH (1, 1) estimation A us tralia real es tate market Ho ng K o ng re al es tate market .0010 Ja pa n rea l es tate market .006 .009 .008 .0009 .005 .007 .0008 .004 .0007 .006 .005 .0006 .003 .004 .0005 .002 .003 .0004 .002 .001 .0003 .001 .0002 .000 100 200 300 400 500 600 700 800 .000 100 S ing ap ore rea l es ta te ma rke t 200 300 400 500 600 700 800 100 UK re al es ta te market .0014 .0012 .010 .0012 .0010 .008 .0010 .0008 .006 .0008 .0006 .004 .0006 .0004 .002 .0004 .0002 .0002 100 200 300 400 500 600 700 800 300 400 500 600 700 800 700 800 US re al e s tate ma rke t .012 .000 200 .0000 100 200 300 400 500 600 700 800 100 200 300 400 500 600 MS CI glo ba l s to c k market .0010 .0009 .0008 .0007 .0006 .0005 .0004 .0003 .0002 .0001 100 200 300 400 500 600 700 800 46 Endnotes 1 Whilst the examination of causation in mean using Granger causality test has commonly appeared in studies relating to financial market movements; the issue of variance causality has received less attention in international finance. 2 In this study, the pre- and post-Asian financial crisis periods are arbitrary defined as 5/1/90-27/6/97 and 4/9/98- 10/2/06 respectively. In the second sub-period (i.e. post-crisis), in addition to the Asian financial crisis, two other common shocks are the August 1998 Russian crisis and the January 1999 Brazilian exchange rate crisis. Since our sample mainly comprises Asia real estate markets, we thus focus our attention on the impact of Asian financial crisis on the cross-market linkages. Kallberg et al. (2002) note that the 1997-98 Asian financial crisis has led to a reduction in real estate returns and an increase in real estate market volatility and correlations following the crisis. 3 Many stock studies were conducted to investigate the benefits of international diversification and to understand the transmission mechanisms across national /regional boundaries. Examples of the some most recent studies include Engle and Susmei (1993), Hamos et al. (1993), Karolyi (1995) and Koutmos (1996), among others. 4 Weekly data are employed in this study as the monthly sampling interval is considered too long a measurement interval to capture the dynamics of volatility and the asymmetry of returns. Furthermore, weekly returns represent a compromise between the many daily observations within a given calendar time interval and the less severe measurement errors in monthly returns. For example, problems arising from non-synchronized trading, rounding errors and bid-ask spreads should not as important in weekly return data as in daily data. 5 The VAR methodology was developed by Sims (1980) with the purpose of estimating unrestricted reduced-form equations that have uniform sets of lagged dependent variables as regressors. The VAR model thus estimates a dynamic simultaneous equation system, free of a priori restrictions on the structure of relationships. 6 Specifically, to obtain the decomposition, the innovations in the VAR system are usually orthogonolized by Cholesky decomposition according to a causal ordering of the variables. However, Pesaran and Shin (1998) propose an orthogonal set of innovations that does not depend on the VAR ordering for obtaining the generalized impulse response function. 7 According to a Strait Times report (18/3/06), the UK is now the biggest investor in Singapore to the tune of S$45.7 billion as at end of 2004. Further, the UK is Singapore second largest European trading partner and a popular investment destination for Singapore companies. 8 The coefficient for the t-shape distribution is 9.839 (standard error is 0.965), which is statistically significant at the one percent level. We estimate the model with BFGS and robust standard error calculations written in RATs program. The estimates converge in 147 iterations. 9 The appropriate lag length (p) chosen for each market is based on the highest negative Akaike information criterion (AIC). The resulting p values are 1 for MSCI, Australia, Japan, UK and US; 2 for Hong Kong and 4 for Singapore. 10 The coefficients for the t-shape distribution are, respectively 11.068 (t = 8.36) and 12.962 (t=6.92), for the pre- and post-crisis models, which are statistically significant at the one percent level. Using RATs estimation, the two models achieved convergence in 206 iterations (pre-crisis) and 134 iterations (postcrisis). Finally, the test statistics (ARCH and LB tests) indicate that the two models are adequately specified. 47 11 A statistical significant level of 10% is adopted in this study. 12 The log-likelihood values are, respectively, 47228.83, 47232.23, 47212.64 and 47185.28 at lag orders 1, 2, 3 and 4. The AIC values are -112.853 (lag 1), -112.880 (lag 2), -112.851 (lag 3) and -112.804 (lag 4). 48