Cyclical impact of business cycle comovements on public property market correlations: an empirical exploration

Transcription

1 IRES IRES Working Paper Series Cyclical impact of business cycle comovements on public property market correlations: an empirical exploration Kim Hiang Liow April, 2013

2 Cyclical impact of business cycle co-movements on public property market correlations: an empirical exploration This draft: 20 April 2013 By: Liow Kim Hiang, Department of Real Estate, NUS, paper to be presented at the 2013 Asian Real Estate Society Conference, June 28 July 1, Kyoto, Japan; Abstract Globalization and the emergence of China in the world economy have been two major events in the international economic literature over the last two decades. The key research question addressed in this paper is whether and to what extent the co-movement of business-cycle fluctuations affects public property market correlations and integration within the three GC economies and with their five regional and global partners over the period 1995 Q1 to 2012Q3. Since public property market is a component part of domestic stock market, we control for the influence of stock market cycle correlations. In order to capture dynamic comovement, our main methodology appeals to an alternative measure of business cycle synchronization dynamic correlation proposed by Croux et al. (2001). We find the cyclical co-movement and interdependence dynamics between public property market correlations, business-cycle correlations and stock market correlations have been quite different at three time horizons (long-run frequencies, traditional business-cycle frequencies and short-run frequencies) although the pro-cyclical links between the respective correlation-types are generally statistically significant at the aggregate frequency level. We also found evidence that the detected pro-cyclical co-movement dynamics are different within and across the GC regions. Finally, our results using the GC dataset should be regarded as indicative and implies that the cyclical variations in quarterly public property market cycle correlations could be highly dependent on the business-cycle co-movement dynamics at different time horizons and corresponding stock market cycle correlations, and is hence more complex than what is normally expected. Keywords: co-movement, dynamic correlation, business-cycles, public property market cycles, stock market cycles, Greater China 1. Introduction An important general measure of a country s economic states that is widely analyzed in the economic literature is the business cycle. Motivated by Ragunathan et al. (1999) who argue that the extent to which financial and economic integration are related indicate that business cycles require careful study as a potential driver for shift in the degree of capital market integration, we study the cyclical link between business cycle co-movements and public property market correlations within three Greater China (GC) economies (Mainland China, Hong Kong and Taiwan), as well across between each of the GC 1

3 economies and their selected international and regional partners over the period 1995Q1-2012Q3. This underlying motivation of the research is further supported by the move towards globalization of financial markets which could result in cross-market linkages being affected by the co-movements of business cycles, and could either happen internationally or in a regional setting. However, Kizys and Pierdzioch (2006) have noted that whether the co-movement of business cycles is a factor that explain changes in international stock market correlations has remained as an unsettled question in the international macroeconomic and finance literature. Our research hopes to alternative evidence. Specifically, as an increasingly influential emerging country, China is aspiring to become one of the largest economies in the world. China's impressive economic success so far is attracting the attention of academics, practitioners, and policy makers worldwide. The capital market is set to play a crucial role in China's development since the efficient allocation of financial resources is a key determinant of economic growth. Moreover, an informal economic region that embraces Mainland China, Hong Kong and Taiwan is rapidly emerging as a new epicenter for industrial, commerce and Finance. Increasing globalization of financial and economic activities is expected to impact real estate market which is a significant asset component in the three GC economies. As real estate companies invest in real properties and are themselves real estate vehicles for other investors in international capital markets, an in-depth study of cross-market correlations and how they are altered by business cycle changes will provide fresh insights into the nature and extent of interdependence between business-cycle correlations and public property market integration or segmentation within and across the GC context. On this issue of business cycle and public property cycle integration, we note that whilst authors such as in and Solnik (1995) as well as Erb, Campbell and Viskanta 1994) have indicated that business cycle fluctuations tend to influence international stock market correlations; others like Karolyi and Sultz (1996); King, Sentana and Wadhwani (1994); Ammer and Mai (1996) as well as Kizys and Pierdzioch (20006) have documented a negligible relationship or found no relationship between international stock market correlations and business cycle co-movement. Since real estate securities 2

4 market is a component part of the domestic stock market, we are interested to find out the nature and extent of relationship between property equity correlations and business cycle correlations as there is no a-priori reasons to expect that public property markets should behave similar to general stock markets with regard to the link with business cycle correlations. This is also an increasing important topic for four main practical reasons: (a) Asia-Pacific is the most dynamic region and its regional integration, as well as its global integration (i.e. vis-à-vis rest of world) deserve critical attention from international investors and policy makers; (b) the emerging discussion on convertibility of Renminbi (Chinese Yuan), and an Asian zone implies a need for more research on contagion, market linkages, capital mobility and business cycle synchronization between and among countries of the region; (c) the increasing tradability of the ultimate non-tradable asset real estate, through its complex linkages with the financial world via the development of a global real estate capital market; and (d) as the economic activities among Mainland China, Hong Kong and Taiwan have increased over the years, the significance of common business cycles among these GC economies may have increased. With real estate assets sharing the same trend, the multilateral economic activities could have contributed a great deal to the correlation relationships among the three GC public property markets. Similarly, closer economic integration has raised the question of whether the business cycles in the three GC economies have become more synchronized over time as well as its consequent impact on the degree of capital market (stock and public property) integration. In adding to the existing body of knowledge concerning global financial market integration, this study makes important contribution. It utilizes recent advances in time series econometrics for a relatively long period of time (1995Q1 to 2012Q3) to investigate the link between business cycle co-movement, public property market cycle correlation and stock market cycle correlation from the cyclical perspective among the three GC economies, as well their international relationships. To our knowledge, whilst previous studies have sought to explain business cycle correlations in a cross-country context, less is known regarding the 3

5 link between public property market correlations and business cycle co-movements. This is where our study hopes to contribute. Specifically, our empirical contributions are the following: We estimate a Hodrick-Prescott (HP) cycle for each business cycle, real estate and stock time series. The cyclical co-movements between the HP business cycles, between the HP public property price cycles and between the HP stock market price cycles for the eight markets are then examined for their contemporaneous correlations and semi-correlations. A simple Spearman rank order correlation analysis is conducted to ascertain whether the magnitude of the stock /public property cycle correlations is correlated to the magnitudes of business cycle correlations. We derive an alternative measure of business cycle, property market cycle and stock market cycle co-movements - dynamic correlation. Following Croux et al (2001), dynamic correlations can be interpreted as a decomposition of aggregate correlation into co-movement at particular frequencies. This approach further allow us estimate the degree of co-movement of the correlations at three different time horizons: business cycle frequencies, long-run frequencies and short-run frequencies. We undertake a regression-based analysis in order to formally assess whether the co-movements of business cycle fluctuations are significantly linked to the property market cycle correlations after controlling for the stock market cycle correlation effect at long-run frequencies, business cycle frequencies and short-run frequencies. To increase the power of estimation, we estimate regression models for panel-data using feasible generalized least square (FGLS), random effect GLS and dynamic GMM techniques in order to check the robustness of the results. The plan of this paper is as follows. Section 2 provides a brief update of the relevant literature. Section 3 contains the description of the data used while the explanation of the methodologies that are utilized in the paper is included in Section 4. Section 5 provides a discussion of the results. Section 6 concludes the paper. 4

6 2. Literature review Duarte and Holden (2003) have noted that during the last two decades, there has been an increased interest in the study of business cycles within and between countries. In Europe, the term synchronicity is associated with the concept of symmetry to indicate the convergence and greater harmonization of the business cycles. Gerlach (1988) studied the monthly industrial production index and found evidence of a world business cycle. Other studies on business cycle relationship include Baxter and Stockman (1989), Artis and Zhang (1997), Duarte and Holden (2003), Sato and Zhang (2006), Jarko and Ivana (2008) and Koopman and Azevedo (2008). Second, Barras (1994) has documented the interaction of property market with the four- to fiveyear business cycle, giving rise to an office development cycle of eight to ten years duration. The RICS (1994) has also produced comprehensive reports on the causes, dynamics and implications of commercial real estate cycles. An underlying theme is that an improved understanding of real estate cycles is vital in the property investment process. The RICS report covers the UK real property returns for a period 1962 to 1992, with real estate cycle duration of four to five years. Its main objectives are to prove that property cycles exists, that it is systematically related to movements in the economy, and that most of the relationships can be quantified through statistical analysis. The report defines property cycles as a recurrent but irregular fluctuations in the rate of all-property total returns, which are also apparent in many other indicators of property activity, but with varying leads and lags against the all-property cycle. The report further demonstrates that commercial property cycles are linked to the economic cycle and some internal mechanisms within the property industry. Since stock markets and public property markets are part of the domestic economy, they are expected to be influenced by the business cycle fluctuations. McGough and Tsolacos (1995) use the HP-filter method to detect pro-cyclicality between GDP, manufacturing and business output in the office and industrial sectors in the United Kingdom. A procyclical pattern was also established for demand side variables such as GDP, consumer expenditure and 5

7 no-food retail sales in relation to the retail property cycle. Employing the same methodology, Brooks, Tsocalos and Lee (2000) fit a long-term trend to the various economic time series in the UK, and to derive short-term cycles of each series. They find that the cycles of consumer expenditure, total consumption per capita, the dividend yield and the long-term bond yield were moderately correlated, and mainly coincident, with the property stock price cycle. The approach in the study is largely based on the long tradition of research on the stylized facts of business cycles. In another study by Witkiewicz (2002), the author uses HP-filter technique to identify cycles in the real estate market and to construct real estate cycle indicators. His findings suggest that the HP-filter offers a means for constructing such indicators and for correctly identifying the turning points that are largely dependent on the real business cycles. With increasing globalization of global financial markets, there are considerable studies investigating the correlation structure across GC and international stock markets (e.g. Cheng and Glascock, 2005; Caporale et al. 2006; Qiao et al, 2008 and Johansson and Ljungwall, 2009), as well as among international public property markets (e.g. Liow et al, 2009; Liow, 2012; Liow and Newell, 2012). In these studies, increased time-varying correlations imply higher stock/ real estate stock market integration. Liow and Newell (2012) investigate simultaneously the effects of volatility spillover and time-varying conditional correlation on the cross-market relationships between the three GC securitized real estate markets, as well as their international links with the US securitized real estate markets. They find that the conditional correlations between the GC securitized real estate markets have outweighed their conditional correlations with the US market, supporting closer integration between the GC markets due to geographical proximity and closer economic links. 3. Data We use data on real GDP to measure business cycle fluctuations. The quarterly data covers from 1995Q1 to 2012Q3 in eight countries, and comes from the International Financial Statistics issued by the IMF. In addition, we include Standard & Poor s quarterly closing public property indexes and stock market indexes for three GC {China (CH), Hong Kong (HK), Taiwan (TW)} and five non-gc economies 6

8 {Australia (AU), Japan (JP), Singapore (SG), the US and the UK} over the same period. The US, UK, JP and AU have well-developed financial markets and open capital accounts. In particular, the US and JP are the main investors and trading partners with the GC economies. SG, due to its geographical proximity and cultural similarity, has had close economic ties with the three GC economies. Exhibit 1 provides the mean, standard deviation, maximum and minimum of the real GDP growth, public property market returns and stock market returns (for local dollars) over the full study period. (Exhibit 1 here) 4. Research methodology We use two complementary approaches in our analysis. First we apply the Hodrick-Prescott (1980) (HP) filter to determine the cyclical component of the GDP, property and stock series. The HP filter is a de-trending technique that is widely used in the economic literature. Second, an alternative measure of business cycle synchronization is dynamic correlation, as was proposed by Croux et al. (2001). We use the methodology to estimate the dynamic correlation measure for the business cycle, property market cycle and stock market cycle series at three different time horizons: business cycle frequencies, long-run frequencies and short-run frequencies. Panel regressions are conducted to determine whether business cycle co-movements significantly affect property market cycle correlations after controlling for the stock market cycle correlations under different time horizons. The main points of each of the methods are explained below 4.1 Hodrick-Prescott (HP) Filter Bilateral correlations in business cycles, property market cycles and stock market cycles are computed on the basis of the cyclical components of quarterly real GDP; real estate and stock market time series indexes, isolated using the bandpass filter introduced by Hodrick and Prescott (1980) (HP) which 7

9 has been used extensively in the literature. 1 The HP methodology consists of determining the cyclical component of the data in analysing the degree of synchronization /correlation between any two series. Further, the HP-filter is a linear filter that decomposes a time series into a cyclical component and a growth component. Its basic assumption is that the trend growth rate is smooth, and the cyclical components sum to zero over time. The Lagrangian equation set up by this technique is expressed as follows, where the square of the cyclical part of a variable is minimised, subject to a smooth growth rate (Hodrick and Prescott, 1980, 1997) Min { T T 2 ct t 1 t 1 [( g t g t 1 ) ( g t 1 g t 2 2 T )] } { g } t 1 t where c t and g t represent the cyclical and trend component of variable Y t, respectively, with Y t c g. For quarterly time series, =1600. t t After the HP filter was used to fit a smooth trend to all time series, the cyclical component of each series could be derived. The cycle of each series is defined as the deviations of the actual values from the HP trend fitted to the series (i.e. cycle = actual series HP trend). Following this procedure, the cyclical co-movements between the HP real GDP cycles, between the HP public property price cycles and between the HP stock market price cycles for the eight economies are examined for their contemporaneous correlations and semi-correlations. 2 Finally, a simple non-parametric Spearman rank order correlation analysis is conducted to determine whether higher magnitudes of the unconditional stock 1 For example, the HP Filter has been applied by the European Commission to estimate output gaps, which is the difference between actual and potential GDP in Europe. 2 For the purpose of this study, we focus on 18 synchronization/ correlation pairs; (a) within the GC areas (between CH-HK, CH-TW, HK-TW); (b) China and international (between CH-JP, CH-SG, CH-AU, CH-US, CH-UK); (c) Hong Kong and international (between HK-JP, HK-SG, HK-AU, HK-US, HK-UK) and (d) Taiwan and international (between TW-JP, TW-SG, TW-AU, TW-US and TW-UK). Groups (b), (c) and (d) together are also labeled as across the GC regions. 8

10 /property stock correlations/semi-correlations are associated with higher magnitudes of business cycle correlations / semi-correlation; or vice-versa. 4.2 Dynamic correlation Although the classical correlation and semi-correlation analyses are two popular approaches for describing co-movements between economic variables, they are basically static analyses that fail to capture any dynamics in the co-movements of business cycles/capital market cycles. We use an alternative measure of synchronization, dynamic correlation, which can be decomposed by frequency and frequency band in order to study business cycle synchronization, property market cycle correlation and stock market cycle correlation at different time horizons. Following Croux et al (2001), if S ( ) and S y ( ) are the spectral density estimates of x and y and Cxy ( ) is the co-spectrum, which are defined for all frequencies, then the dynamic correlation is x ( ) xy C x xy ( ) S ( ) S y ( ).(1) The dynamic correlation coefficient is defined to be between -1 and +1. In addition, the average value of dynamic correlation over all frequencies is approximately equal to the static correlation. It can be interpreted as a decomposition of the aggregate correlation into co-movements at particular frequencies. Croux et al. (2001) further point out that the interpretation of coherence (or squared coherency) is quite similar to dynamic correlation. To motivate the dynamic correlation measure, we perform co-spectral analyses to obtain the three component estimates in (1) for the 18 economy pairs GDP, property and stock market cycles at each frequency. Co-spectral analysis is a means if revealing variance and covariance, but in the frequency domain. On the basis of NBER studies, we look at three components of the aggregate dynamic correlations. First, the long-run movements (over 8 years) correspond to the low frequency band below /16. Second, the traditional business cycles (lasting between 1.5 and 8 years) belong to the medium part of the figure and are between /16 and /3. Finally, the short-run 9

11 movements are defined by frequencies over /3. We examine the average dynamic correlation of the GDP/public property/stock cycles at these three time horizons and search for preliminary evidence regarding the co-movements between the dynamic correlations of business cycles, public property market cycles and stock market cycles in our data set using Spearman rank correlation analysis.. Finally, we formally test the cyclical co-movements between property market correlations (RE), business cycle correlations (GDP) and stock market correlation (ST) for the 18 economy pairs using the dynamic correlation estimates. In this way, we are interested to find out whether and to what extent business cycle co-movements could affect property market cycle correlation after controlling for the stock market cycle correlation at the long-run, business cycle and short-run frequencies. By undertaking an orthogonalization in Model II, we hope to extract the co-movements in the public property market in excess of the general stock market correlations, so that any residual relationship between the detected market correlation and business cycle co-movement can be reasonably attributed to the public property markets per se. We run the following two models: RE f ( GDP ) frequency Model I freqwency RE f GDP, ST ) Model II frequency ( frequency frequency The two models are specified for the 18 economy-pairs. The regression series are estimated in three different ways. First, all 18 equations are estimated as a pooled cross-sectional model to minimize unobserved panel-level effects, thereby constraining all regression coefficients, except the intercepts, to be identical across all 18 equations. As the Hausman test indicates the GLS random effect estimator is better than a within estimator for fixed effect models, we estimate the two models with a GLS estimator for random effects model. Second, we estimate the two models using feasible generalized least square (FGLS) to account for possible cross-sectional correlation dependence and heteroskedasticity across panels (i.e. the variance for each of the panels differs). Finally, we use a consistent generalized method of 10

12 moment (GMM) estimator (Arellano and Bond, 1991) to re-estimate the two models. This dynamic panel GMM model is implemented to render the unobserved panel effects orthogonal to the one-quarter lagged correlation variable, as well serves a robustness check on the pooled sample specification. 5. Results Applying the HP filter on the time series, Exhibit 2 shows average business-cycle correlation, public property market cycle correlation and stock market cycle correlation for the 18 economy-pairs. We observe that the highest business-cycle (GDP cycle) correlation is found between Hong Kong and Singapore (0.7985). This highest business cycle correlation corresponds with the highest property cycle correlation (0.8750), as well as the stock market cycle correlation (0.9099) between Hong Kong and Singapore. Additional Spearman rank correlation estimates for the full sample are (average rank coefficient between GDP cycles and stock market cycle correlations), (average rank coefficient between GDP cycles and property market cycle correlation) and (average rank coefficient between stock market cycles and property market cycle correlation), indicating for the entire sample, higher business-cycle correlations are moderately associated with higher real estate market cycle correlations and higher stock market cycle correlations over the entire study period. (Exhibit 2 here) In order to test whether the link between the property market cycle correlations, stock market cycle correlations and the co-movements of business cycle fluctuations exists, we define phases of economic booms and recessions. Using the quarterly real GDP growth rate as a measure of economic performance, we define common-up when a pair of economies experienced a boom (positive real GDP growth) at the same quarter. Similarly, we define common-down when a pair of economies experienced a recession (negative real GDP growth) at the same quarter. Finally, we define a mixed market when the business cycles of a pair of economies were out-of-phase. The semi-correlation results are reported in Exhibit

13 (Exhibit 3.1 here) It is unclear from the numbers that the magnitudes of HP property cycles /HP stock market cycles semi-correlations are systematically associated with the magnitudes of the HP business cycles semicorrelations. For the full sample, additional Spearman rank correlation analysis (Exhibit 3.2) reveals that whilst the rank correlation coefficient between the business-cycle correlations and property market cycle correlations are higher in common-down market condition (0.627) than in common up market condition (0.539), the reverse is true between the business-cycle correlations and stock market cycle correlations during common-down (0.455) and during common-up (0.549) market periods. It is therefore interesting to examine more formally by means of dynamic correlation analysis any conclusive link between property market cycle correlations, stock market cycle correlations and the co-movements of business cycle fluctuations. (Exhibits 3.2 here) Exhibit 4 compares average dynamic correlations at the business cycle, short-run and long-run frequencies for the 18 country-pairs and 4 regional-pairs. It must be cautioned that the long-run results should be interpreted with great caution due to the small number of observations available for each economy pairs 3. Within the three GC economies, the business cycles display low levels of dynamic correlation (between and 0.405), especially for the short-run frequencies. Similarly, there are between low and moderate correlations of property cycles between CH, HK and TW, implying that three are diversification opportunities in the GC public property markets, especially for the short-run frequencies. At the business-cycle frequencies, the three GC economies (as a group) display the highest average GDP cycle correlation (0.555), property market cycle correlation (0.457) and stock market cycle 3 For each economy pair, the co-spectral results have 3 observations for the long run frequencies, 12 observations for the business cycle frequencies and 24 observations for the short run frequencies. 12

14 correlation (0.755). Across the GC areas, the numbers indicate the average dynamic correlation of the GDP, property and stock market cycles are the highest for the business cycle frequencies for Chinainternational (0.428, and 0.622), Hong-Kong-international (0.570, and 0.765), as well as Taiwan-international (0.520, and 0.658). These results imply that the interdependences between the economic and capital market (stock /public property markets) development in the three GC economies and in developed economies are becoming similar for the business cycle development and more permanent shocks should influence the business cycle frequencies. Next whilst all dynamic correlation estimates are positive for the business frequencies, there are several instances of negative correlations for the long-run and short-run frequencies. For example, the correlations of business cycles are negative between CH and the US, between CH and the UK, between HK and the US, as well between HK and AU for the long run frequencies. In addition, CH s stock market cycle is negatively correlated with those of US and UK. For the short-run frequencies, negative correlations of business cycle are found between HK and the UK, HK and AU, as well as between TW and US. In addition, the property market cycle of TW is negatively correlated with those of the US and AU, albeit at the very low negative levels ( and respectively). Overall, the correlations of business cycle, property cycle and stock market cycle between each of the three GC economies and two Asian partners (JP and SG) are stronger than those of non-asian partners (The US, the UK and AU) probably because of geographical proximity and some similarities in their economic development, especially with SG. Finally, additional Spearman rank correlation analyses for the full sample reveal that the co-movement between the business cycle correlations and property market cycle correlation (0.521), as well as between the business cycle correlations and stock market cycle correlations (0.684) are the highest and statistically significant at least at the 5% level for the long-run frequencies compared to those for the business cycle frequencies and short-run frequencies. (Exhibit 4 here) 13

15 Exhibits 5.1 to 5.4 present dynamic correlation profile of business-cycles, property market cycles and stock market cycles within and across the CH-international, HK-international and TW-international groups, separated by the long-run, business cycle and short-run frequencies One key observation from the graphs is that while there are pro-cyclical co-movements across the three dynamic correlation types (GDP, property and stock) for most of the long run and certain business cycle frequencies for some economy-pairs, the co-movements between the correlations of business cycles, property cycles and stock market cycles fluctuate in different patterns especially at various short-run frequencies. The general pattern of co-movement across the three correlation-types over the short-run cycles is not as clear as might have been expected. In general, the cyclical co-movements are weaker (and in some cases negative) although there is ad-hoc evidence of high positive co-movements between the dynamic correlations of business cycles, public property and stock market cycles at intermediate short-run frequencies. (Exhibit 5 here) Within the GC region, Exhibit 6 shows the average co-movement (represented by the Spearman rank coefficient) between the dynamic correlations of business cycles and property cycles is positive (0.333) for long run frequencies; but is negative for business cycle (-0.117) and short-run (-0.009) frequencies. In contrast, the relevant average co-movement is all positive for all three time horizons between each of the three GC economies and their global partners. Of the 18 economy-pairs, there are, respectively, 3 cases and 11 cases of negative cyclical co-movement between the business-cycle correlations and property market cycle correlations for the traditional business cycle and short run frequencies. Similarly, we find another 4 cases (business cycle frequencies) and 9 cases (short-run frequencies) of negative cyclical co-movements between the business cycle correlations and stock market cycle correlations over the same period. 4 (Exhibit 6 here) 4 Due to lack of data for long run frequencies, we were not able to estimate the relevant Spearman correlation coefficients for each economy-pair. 14

16 Finally, Exhibit 7 reports the results of estimating Model 1 and Model II as a pooled crosssectional and time series regression using (a) a generalized least square (GLS) estimator for random effect model; (b) feasible GLS method; and (c) dynamic panel GMM estimation. The dependent variable is the Fisher-transformed dynamic correlation coefficient since the raw estimates are not normally distributed. At an aggregate frequency level (Panel A), we find, except for the China-international group, there are significant pro-cyclical movements between property market cycle correlations and business cycle correlations for other five groups, a finding that is reasonably expected. However, when the effect of the underlying stock market correlation is controlled, the link between the property market cycles and business-cycles becomes somewhat weaker since significant pro-cyclical co-movements between the dynamic correlations of property cycles and business cycles are only detected for four groups. Nevertheless, these findings imply that the co-movement of business cycle fluctuations is a potential driver in influencing the degree of public property integration and is in agreement with the impact of the economic activity level on stock returns. Panel B shows in the long run, the link between the correlations of property cycles and business cycles is significantly positive within the GC region, Hong Kong and international, as well as for the Taiwan and international groups. In contrast, the correlations of property cycles are not linked to the stock market cycles for CH-International, HK-international and TWinternational groups, implying that there is little integration of public property markets with their stock markets between each of the three GC markets and their global partners. It can be argued that these results indicate public property cycles are not aligned with stock market cycles in the long run, a finding which is in agreement with the understanding that the long-run performance of public property is tied to the performance of the underlying direct property assets; but not the general stock market. Finally, it is evident from the results that the long-run co-movement dynamic between the correlations of business cycles and property markets cycles is different within and across the GC groups. (Exhibit 7 here) 15

17 For the business cycle frequencies, results of Panel C reveal that except for the within GC group, the correlations of business cycle fluctuations affect significantly the correlations of property market cycles even after the effect of the stock market cycle correlations are controlled for the full sample, GCinternational, CH-international, HK-international and TW-international groups, Similarly, with one exception (Taiwan-international) there are significant pro-cyclical co-movements between the dynamic correlations of property and stock market cycles for other five groups. Based on these results, we are inclined to conclude that the cyclical variations in quarterly public property correlations are reasonably linked to the co-movements of business cycle fluctuations and stock market correlations at the traditional business frequency horizon. Finally, the short-run link (Panel D) between public property cycle correlations between business-cycle correlations is found to be very weak and negative for most of the panels examined. However, there are strong cyclical links between the public property correlations and stock market correlations for all groups examined. The results are generally robust with the different estimation methods. Based on the regression results generated in our analyses, we are inclined to conclude the cyclical co-movement dynamics between public property market correlations, business-cycle correlations and stock market correlations have been different at the three time horizons. In the long run, public property cycle correlations are significantly pro-cyclical to business-cycle co-movements but are less aligned with the stock market cycle correlations. Between 1.5 years and 8 years (the traditional businesscycle frequencies), the positive links between the co-movements of property cycle correlations and business-cycle correlations, and between property cycle correlations and stock market cycle correlations are generally stronger. In the short run, however while there is generally insignificantly and negative link between the cyclical co-movement of public property market correlations and business-cycle correlations, the cyclical link between the public property market correlations and stock market correlations emerge to be the strongest in many cases. These results are generally in agreement with the requirements of 16

18 economic theory and imply that assessing and understanding the links between business-cycle correlations, public property market cycle correlations and stock market cycle correlations can be more complicated than generally expected. Second, it appears that the above co-movement of the two groups: i.e. within the GC region and GC-international are largely different. For example, while the property market cycle correlations are not linked to the business-cycle correlations within the three GC economies, this link is statistically significant across the GC-international group. This can be probably attributed to the fact that the three GC economies (especially China) and their global developed partners are quite different in terms of their macroeconomic conditions, degree of market openness, informational transparency, legal system, size and maturity level, as well as levels of government intervention on the real estate market. 6. Conclusions Globalization and the emergence of China in the world economy have been two major events in the international economic literature over the last two decades. The key research question addressed in this paper is whether and to what extent business cycle co-movements affect correlations and integration of public property market cycles within the GC and across the GC areas over the study period from 1995Q1 to 2012Q4. In this paper we have used the Hodrick-Prescott filter to decompose real GDP (a common proxy for business cycle), public property market index and stock market index into cyclical and trend components. The resulting series of cyclical components were then examined for across relationships using correlation and semi-correlation measures. Generally, higher (semi-) correlations of business cycles are moderately associated with higher (semi-) correlations of public property market cycles and higher (semi-) correlations of stock market cycles over the entire study period. Further, we derive dynamic correlation, which can be decomposed by frequency and frequency band to study business cycle co-movement, public property market cycle correlations and stock market cycle correlations at different time horizons using descriptive statistics and panel data regressions. In short, based on the results using our data, we find the cyclical co-movement dynamics between public property market 17

19 correlations, business-cycle correlations and stock market correlations are different at the three time horizons, although the pro-cyclical links between the respective correlation-types are generally statistically significant at the aggregate frequency level.. We also find evidence to indicate that the detected pro-cyclical co-movement dynamics are different between the within and across the GC regions Despite our broad findings, some of results remain inconclusive. As such, our results on the GC dataset should be regarded as indicative and implies that the cyclical co-movement variations in quarterly property market cycle s dynamic correlations could be highly dependent on the business-cycle comovement dynamics at different time horizons and corresponding stock market cycle correlations, and is hence more complex than what is normally expected. Notwithstanding this shortcoming the central point of our analysis still holds true: co-movement of business-cycle fluctuations impact upon public property cycle and stock market correlations. In future research the impact of business-cycle changes on capital market integrations warrants deeper analysis using a combination of time and frequency domain approaches to testing for their significant relationship using different international datasets. 18

20 References Artis, M.J. and Zhang, W. (1997), International business cycles and the ERM Is there a European business cycle? International Journal of Finance and Economics 2: 1-16 Barras, R. (1994), Property and the Economic Cycle: Building Cycles revisited, Journal of Property Research 11(3), pp Baxter, M. and Stockman, A.C. (1989), Business cycles: the exchange rate regime Journal of Monetary Economics 23: Brooks, C., S. Tsolacos and S. Lee (2000), The Cyclical Relations between Traded Property Stock Prices and Aggregate Time-series, Journal of Property Investment and Finance 18(6): Cappiello, L., Robert, E. and Sheppard, K. (2006), Asymmetric dynamics in the correlation of global equity and bond returns Journal of Financial Econometrics 4(4), Cheng, H. and Glascock, J. (2005), Dynamic linkages between the Greater China Economic Area stock markets Mainland China, Hong Kong and Taiwan Review of Quantitative Finance and Accounting 24, 343-,357 Croux, C., Forni, M. and Reichlin, L. (2001), A measure of co-movement for economic variables: theory and empirics The Review of Economics and Statistics 83(2): Duarte, A. and Holden, K. (2003) The business cycles in the G-7 economies International Journal of Forecasting 19: Erb, C., Campbell, H. and Viskanta, T. (1994), Forecasting international equity correlations Financial Analysts Journal: 32-45, November/December Gerlach, H.M.S. (1988), World business cycles under fixed and flexible exchanger rates Journal of Money, Credit and Banking 20: Hodrick, R. and E. Prescott (1980), Post-War US Business Cycles: An Empirical Investigation, Carnegie Mellon University. Hodrick, R. and E. Prescott (1997), Post-War US Business Cycles: An Empirical Investigation Journal of Money, Credit and Banking 29: 1-16 Jacko, F. and Ivana, B. (2008), China in the world economy: dynamic correlation analysis of business cycles Research paper, UNU-WIDER, United Nations University (UNU) No: 2008/12. Johansson, A. and Lijunwall, J. (2009), Spillover effects among the Greater China stock markets World Development 37, Karolyi, G.A. and Stulz, R.M. (1996), Why do markets move together? An investigation of US-Japan stock returns movements Journal of Finance 51: King, M., Sentana, E. and Wadhwani, S. (1994), Volatility and links between national stock markets Econometrica 62:

21 Kiyzs, R. and Pierdzioch, C. (2006), Business-cycle fluctuations and international equity correlations Global Finance Journal 17: Koopman, S.J. and Azevedo, J.V.E. (2008), Measuring synchronization and convergence of business cycles for the Euro areas, UK and US Oxford Bulletin of Economics and Statistics 70(1):23-51 Liow, K.H., Ho, K.H.D., Ibrahim, M.F. and Chen, Z. (2009), Correlation and volatility dynamics in international real estate securities markets Journal of Real Estate Finance and Economics 39: Liow, K.H. (2012) Co-movements and correlations across Asian Securitized Real Estate and Stock Markets, Real Estate Economics 40(1): Liow K.H and Newell, G (2012) Investment dynamics of the Greater China securitized real estate markets and international linkages Journal of Real Estate Research 34(3): in, F. and Solnik, B. (1995), Is the correlations in international equity returns constant: ? Journal of International Money and Finance 14: 3-26 McGough, T. and Tsolacos, S. (1995), Property Cycles in the UK: An Empirical Investigation of the Stylized Facts, Journal of Property Finance 6(4), pp Qiao, Z., Chiang, C. and Wong, W. (2008), -run equilibrium, short-term adjustment, and spillover effects across Chinese segmented stock markets and the Hong Kong stock market Int. Fin. Markets., Inst. & Money 18, Ragunathan, V., Faff, R.W. and Brooks R.D. (1999), Correlations, business cycles and integration Journal of International Financial Markets, Institutions and Money 9: Sato, K. and Z. Zhang (2006), Real output co-movements in Ease Asia: any evidence for a monetary union? World Economy 29: Witkiewicz, W. (2002), The Use of the HP-filter in Constructing Real Estate Cycle Indicators, Journal of Real Estate Research 23(1/2):

22 Exhibit 1 Average and standard deviation of quarterly real GDP growth, public property market returns and stock market returns: 1995Q1 to 2012Q3 (Local dollars) Real GDP growth rate Real estate securities market return Stock market return Mean (%) Standard deviation (%) Mean (%) Standard deviation (%) Mean (%) Standard deviation (%) China Hong Kong Taiwan US UK Australia Japan Singapore Source: Derived from IMF statistics, S& P Global BMI, based on local currency. 21

23 Exhibit 2 Unconditional correlations: HP real GDP cycles, HP public property market cycles and HP stock market cycles: 1995Q1 to 2012Q3 Pairs Correlation of real GDP growth Correlations of real GDP cycles Correlations of stock market cycles Correlations of real estate market cycles Within Greater China (GC) CH and HK CH and TW HK and TW China and International CH and US CH and UK CH and AU CH and JP CH and SG Hong Kong and International HK and US HK and UK HK and AU HK and JP HK and SG Taiwan and international TW and US TW and UK TW and AU TW and JP TW and SG

24 Exhibit 3.1 Unconditional semi-correlations: HP GDP cycles, HP real estate securities market (RE) cycles and HP stock market (ST) cycles: 1995Q1 to 2012Q3 Market Pair Common up Markets Correlation Common down markets Correlation Mix markets correlation # of observations for ea GDP cyclere Cycles ST Cycles GDP growt GDP cyclere Cycles ST Cycles GDP growt GDP cyclere Cycles ST Cycles GDP growtup Down Mixed CH HK CH TW na na na na CH US CH UK CH AU na na na na CH JP na na na na CH SG na na na na HK TW HK US HK UK HK AU na na na na HK JP na na na na HK SG na na na na TW US TW UK TW AU na na na na TW JP na na na na TW SG na na na na US UK US AU na na na na US JP na na na na US SG UK AU na na na na UK JP na na na na UK SG na na na na AU JP na na na na AU SG na na na na JP SG na na na na

25 Exhibit 3.2 Spearman rank correlation coefficient: 1995Q1 2012Q3 Common-up markets Common-down markets Mixed markets Between HP GDP cycle correlations and HP real estate cycle correlations Between HP GDP cycle correlations and HP stock market cycle correlations Between HP real estate market cycle correlations and HP stock market cycle correlations Between real GDP growth correlations and HP real estate market cycle correlations Between real GDP growth correlations and HP stock market cycle correlations

26 Exhibit 4 Average dynamic correlations of business cycles, stock market cycles and public property market cycles: 1995Q1-2012Q3 Country Pair ALL frequency Business cycle Property cycle Stock market cycle run Business cycle Short run ALL frequency run Business cycle Short run ALL frequency run Business cycle CH-HK CH-TW HK-TW Within GC CH-US CH-UK CH-AU CH-JP CH-SG CH-INT HK-US HK-UK HK-AU HK-JP HK-SG HK-INT TW-US TW-UK TW-AU TW-JP TW-SG TW-INT GC-INT Short run 25