Abstract

This paper examines the characteristics of housing price cycles in East Asia and Greater China for the period from 2001:Q1 to 2010:Q1. We find that housing price cycles in East Asia (China, Hong Kong, Japan, Korea, Singapore, and Taiwan) are accounted for mainly by region-specific and country-specific factors. East Asia's regional housing price cycles co-move strongly with the world housing price cycle in the long run, but relatively weak co-movement is found in the short run. Housing cycles in Greater China (China, Hong Kong, and Taiwan) and Singapore co-move with Northeast Asia's regional housing price cycle in the long run, but this tendency fails to show up in the short run. Both domestic monetary and business-cycle effects are important in accounting for housing price cycles in China, Hong Kong, Japan, and Taiwan, while credit supply is crucial for Korea. Fiscal policy does not play a significant role. We find empirical evidence of a China effect in housing price cycles in Hong Kong, Taiwan, and Singapore.

1.  Introduction

Growing attention has been given to the co-movement of housing prices and real economic activity in the past two decades, with the housing market becoming a key source of business cycle fluctuations, adding to recession and expansion phases. Both the Asian currency crisis and the global financial crisis have corroborated the crucial impacts of such boom-and-bust cycles on emerging economies in East Asia and the regional nature of housing price cycles has been highlighted by the housing price cycles and their international co-movement pattern across East Asian economies.

Empirical literature has emphasized the interaction between business cycles and house price boom-and-bust cycles (Bordo and Jeanne 2002; Claessens Kose, and Terrones 2009; Beltratti and Morana 2010; Igan et al. 2011, among others). Despite the importance of housing price cycles and the evidence of their international co-movement, few studies examine these issues and their macroeconomic determinants for East Asian economies. Recently, Cho, Kim, and Renaud (2012) provided a comparative analysis on real estate sectors of six East Asian economies in terms of the housing price boom-and-bust episodes and housing price volatility.

In many OECD countries, there was a synchronized housing price boom in the early to mid 2000s, although divergent housing price paths are observed across countries after the onset of the global financial crisis. In East Asia (shown in Figure 1), China, Hong Kong, Singapore, and Taiwan experienced substantial increases in real house prices but Japan and Korea exhibited weaker performance in the mid to late 2000s. These observations raise several questions about the characteristics of housing price cycles in East Asia: (i) the international co-movement pattern of housing price cycles; (ii) the major driving forces of housing price fluctuations; and (iii) housing price cycle dynamics in the Greater China Region.

Figure 1.

Housing price movements in East Asia

Figure 1.

Housing price movements in East Asia

This study analyzes the characteristics of international housing price cycle co-movement for six East Asian economies for the period from 2001:Q1 to 2010:Q1. The six countries are China, Hong Kong, Japan, Korea, Singapore, and Taiwan. We investigate the international co-movement of housing price cycles and the extent to which major macroeconomic driving forces are significant in accounting for housing price fluctuations across the six economies.1 A widely used approach to measure co-movement of economic variables is to calculate bivariate constant correlations of the variables of interest (e.g., Gong and Kim 2013; Jung et al. 2013). Taking a multilateral perspective, we use a dynamic factor model to extract the latent factors for real housing price cycles and their macroeconomic drivers. Then, instead of calculating constant correlations between individual country pairs, the degree of housing price cycle co-movement is evaluated through dynamic correlation on a frequency band and a time-varying dynamic conditional correlation.

This paper seeks to extend the literature by dealing with the following four issues. First, we decompose housing price fluctuations by estimating a dynamic factor model: e.g. by extracting world, region-specific, and country-specific factors. Second, the co-movement pattern of East Asia's housing price cycles is characterized in the frequency band and in the time domain, respectively. Third, we examine key macroeconomic driving forces of the housing price cycles in East Asia. Finally, we examine the impact of China on housing price cycles in Hong Kong, Taiwan, and Singapore.

The remainder of the paper is organized as follows. Section 2 introduces the empirical methodology and the data used in this study. Section 3 examines East Asia's housing price cycle co-movements, and Section 4 discusses key macroeconomic driving forces of housing price cycles. Section 5 presents the empirical evidence of a China effect in Hong Kong, Taiwan, and Singapore. Section 6 concludes.

2.  Empirical methodology and data

2.1  Dynamic latent factor model

A dynamic latent factor model estimates the unobserved latent factors based on the spectral density matrix of random variables, whereas a static factor model uses the variance-covariance matrix. We use the growth cycles of housing prices, which are seasonally adjusted and detrended, of 14 major economies2 to decompose housing price cycles into a world factor, region-specific factors, and country-specific factors. To this end, a dynamic latent factor model, given by equation (1), is estimated:
formula
1
where
formula
2
formula
3
formula
4
is a stationary time series at time for economy . The factor is the common factor that is shared by all economies considered in the estimation, and thus it is referred to as the world factor. In our context, when is the demeaned growth cycle of real housing prices for economy , the common factor is understood as the world housing price cycle across the 14 major economies. In addition, the factor stands for the region-specific factor that is commonly shared by only the economies in a geographical region, such as North America, Europe, Oceania, and East Asia. is the country factor, including its idiosyncratic component. Both and are factor loadings that represent the quantitative impact of the associated factors. These factors follow autoregressive processes to reflect the dynamic relationship of time series. The length of two is used for the order of autoregressive polynomials , and . The shocks in equations (2)–(4) are cross-sectionally uncorrelated at all leads and lags, and thus the estimated factors are orthogonal.
Estimation of the latent factor model uses the procedure of Kose, Otrok, and Whiteman (2003). Bayesian estimation of the model characterizes the joint posterior of the factors and the parameters by using numerical methods to simulate from the posterior. The prior on the factor loadings is given by:
formula
For the innovation variances of the observable equations, the prior is given by an inverse gamma . The prior on the parameters of autoregressive polynomials is . To implement the procedure, a Markov-Chain Monte Carlo procedure is used to estimate the dynamic factor model. The mean values of the posterior distribution, drawn from 10,000 replications, are used in the empirical analysis.

2.2  Variance decomposition

The variance decomposition of equation (1) allows us to measure the relative contributions of the estimated factors. To quantify the share of the relative contributions, the variance of an observable variable is decomposed into the proportion that is attributable to the relevant factor. With the orthogonal factors, as in Kose et al. (2003), the variance of an observable variable of economy is given by:
formula
Then, for instance, the fraction of the total variability of economy 's observable variable attributable to the world factor is:
formula
which relies on both the value of factor loadings and the relative share of the variations. The fractions of the variability due to the region-specific factor is calculated similarly.

2.3  Measurement of dynamic co-movement

To quantify the degree of housing price cycle co-movement, two measures of dynamic co-movement are used: (i) dynamic correlation on a frequency band, so-called cohesion, and (ii) time-varying dynamic conditional correlation.

A frequency-band dynamic correlation is presented by Croux, Forni, and Reichlin (2001). For two time series and , the dynamic correlation in the frequency domain is defined by:
formula
where is the co-spectrum, and and are the spectral density functions of and at frequency .
Another measure of time-varying dynamic conditional correlation (DCC) is calculated by running a DCC-GARCH model developed by Engle (2002) with:
formula
where is an vector with mean zero and time-varying covariance, is a time-varying conditional correlation matrix, and is a diagonal matrix of conditional standard deviations. The two-step estimation method fits a multivariate GARCH model to estimate , and then models by estimating the parameters by quasi-maximum likelihood estimation on :
formula
where is a sample covariance matrix of as a vector of ones, and stands for the Hadamard product of two matrices.

2.4  Data

We use quarterly data over the period 2001:Q1–2010:Q1. Real housing price indexes are taken from several sources: the Economist data set (for the United States, Canada, the UK, France, Germany, Italy, Austrialia, New Zealand, China, Hong Kong, Japan, and Singapore), the KB Kookmin Bank (for Korea), and the Taiwan Real Estate Research Center (for Taiwan).

Real gross domestic product measures the business cycle and monetary and fiscal policy effects are captured by the real stock of M2 and real government expenditure. The supply of credit, measured by the financial sector's claims on the private sector, is included to assess the effect of mortgage debt on housing price cycles.3 We also consider a measure of trade-openness, proxied by total exports and imports. M2 and total exports and imports are taken from the IMF's International Financial Statistics (IFS) database. Government expenditure is obtained from various sources: the IFS, the Bank of Korea, the Department of Statistics Singapore, and the National Bureau of Statistics China. The supply of credit and foreign capital inflows are available from the IFS except for Taiwan, where the data are sourced from the Central Bank of the Republic of China.

Nominal variables are deflated by the consumer price index, available from the IFS database. The real housing price indexes and macroeconomic variables are seasonally adjusted using the U.S. Census Bureau's X-12 seasonal adjustment program and then processed to generate the demeaned growth cycle series of the respective variable.

3.  International co-movement of housing price cycle

We first estimate a dynamic latent factor model represented by equation (1) for the growth cycles of real housing prices to extract an East Asia region-specific factor of housing price cycles. Because East Asia's regional housing price cycles can be affected by housing price movements in major economies in the world, we consider 14 economies located over four regions: the United States and Canada in North America; the UK, Germany, France, and Italy in Europe; Australia and New Zealand in Oceania; China, Hong Kong, Japan, Korea, Singapore, and Taiwan in East Asia.

Estimating equation (1) allows one to extract a world factor, four region-specific factors, and the country-specific component of housing price cycles. The world factor, which is the factor commonly shared by the 14 major economies, and the East Asia region-specific factor are displayed in Figure 2. These latent factors capture the cyclical feature of the expansion and the contraction phases, with a housing price run-up in the early 2000s following the recovery from the Asian currency crisis, and a sudden, sharp drop during the global financial crisis in mid 2008.

Figure 2.

World and East Asia's region-specific factors of housing price cycles

Figure 2.

World and East Asia's region-specific factors of housing price cycles

It is useful to examine the relative contribution of the world, region-specific, and country factors to the economy's housing price cycle, as shown in Table 1. The world factor accounts for more than half of national housing price cycles in the regions except for in East Asian economies: the United States (62.53 percent), the UK (67.42 percent), France (68.74 percent), Italy (50.23 percent), and New Zealand (61.57 percent).4 On the other hand, the importance of the world factor is less than 20 percent in East Asian economies. In these countries, housing price cycles are explained by the regional and the country factors: China (81.28 percent), Hong Kong (95.62 percent), Japan (89.91 percent), Korea (90.82 percent), Singapore (87.63 percent), and Taiwan (98.72 percent). Given these features of the housing price cycles in East Asia, in the following sections we mainly focus our analysis on regional factors.

Table 1. 
Variance decompositions of housing price growth cycles (%)
World factorRegional factorCountry factor
United States 62.53 13.04 24.43 
Canada 35.92 26.16 37.92 
UK 67.42 5.07 27.51 
France 68.74 8.12 23.14 
Germany 2.10 13.95 83.95 
Italy 50.23 14.77 35.00 
Australia 22.96 21.55 55.48 
New Zealand 61.57 9.86 28.57 
China 18.72 13.15 68.13 
Hong Kong 4.38 18.11 77.51 
Japan 10.09 11.59 78.32 
Korea 9.18 14.53 76.30 
Singapore 12.37 13.80 73.83 
Taiwan 1.28 9.98 88.74 
World factorRegional factorCountry factor
United States 62.53 13.04 24.43 
Canada 35.92 26.16 37.92 
UK 67.42 5.07 27.51 
France 68.74 8.12 23.14 
Germany 2.10 13.95 83.95 
Italy 50.23 14.77 35.00 
Australia 22.96 21.55 55.48 
New Zealand 61.57 9.86 28.57 
China 18.72 13.15 68.13 
Hong Kong 4.38 18.11 77.51 
Japan 10.09 11.59 78.32 
Korea 9.18 14.53 76.30 
Singapore 12.37 13.80 73.83 
Taiwan 1.28 9.98 88.74 

Source: Authors’ calculations.

Figure 3, for example, compares the estimated East Asia region-specific factor to the housing price growth cycle in each of the six countries. Both Japan's and Korea's housing price cycles seem to be divorced from the region-specific factor, whereas those in China, Hong Kong, Singapore, and Taiwan share a close co-movement with regional housing price cycles. The pattern, which is consistent with the house price movements shown in Figure 1, suggests that the four economies share a common relationship.

Figure 3.

East Asia's region-specific factor and national housing price cycles

Figure 3.

East Asia's region-specific factor and national housing price cycles

To further assess the degree of housing price cycle co-movement, Table 2 presents the average values of the dynamic conditional correlations obtained by running a DCC-GARCH(1,1) model. Based upon the results, the East Asian economies can be classified into two groups. Japan and Korea show a somewhat decoupled behavior from the East Asia regional factor (–0.20, –0.08), and China, Hong Kong, Singapore, and Taiwan economies co-move with the regional factor (0.20, 0.37, 0.20, 0.25).

Table 2. 
Dynamic conditional correlations of the national housing price cycle
ChinaHong KongJapanKoreaSingaporeTaiwan
East Asia's regional factor 0.20 0.37 −0.20 −0.08 0.20 0.25 
ChinaHong KongJapanKoreaSingaporeTaiwan
East Asia's regional factor 0.20 0.37 −0.20 −0.08 0.20 0.25 

Source: Authors’ calculations.

Next, housing price cycle co-movement is examined through the dynamic correlations in the frequency domain. The dynamic correlations in Figures 4 and 5 show the degree of housing price cycle co-movement over the frequency range.5 Figure 5 illustrates the co-movement modality between the world factor and the East Asia regional factor. At low frequencies, East Asia's regional housing price cycle co-moves strongly with the world housing price cycle (higher than 0.8), and it shows a somewhat weak co-movement at high frequencies (lower than 0.4). At business cycle frequencies the dynamic correlation is about 0.5. Therefore, we conclude that East Asia's regional housing price cycle co-moves strongly with the world housing price cycle in the long run, but the co-movement weakens in the short run.

Figure 4.

Frequency-band dynamic correlation between world and East Asia's regional factors

Figure 4.

Frequency-band dynamic correlation between world and East Asia's regional factors

Figure 5.

Frequency-band dynamic correlation between East Asia's regional housing price cycle and national housing price fluctuations

Figure 5.

Frequency-band dynamic correlation between East Asia's regional housing price cycle and national housing price fluctuations

Figure 5 illustrates the pattern of co-movement in national cycles with East Asia's regional housing price cycle. At low frequencies, Japan and Korea show weak co-movement or even decoupling from the East Asia regional housing price cycle (i.e., near zero and negative correlations), whereas China, Hong Kong, Singapore, and Taiwan show coupling phenomena with the regional housing price cycle (i.e., positive correlations). At business cycle frequencies, Hong Kong, Singapore, and Taiwan have positive correlations. At high frequencies, none of the six East Asian economies shows close co-movement (i.e., near zero correlations). Therefore, we conclude that the Greater China economies and Singapore have a co-movement tendency in the long run, but not in the short run.

4.  Driving forces of housing price cycles

To characterize major determinants of housing price fluctuations, we seek to identify the driving forces by examining key macroeconomic variables that may contribute to housing price cycles. More specifically, we extract the East Asia regional common factor and the country-specific factor of five macroeconomic drivers by estimating a two-factor model that is similar to that used in the previous section. The five drivers represent monetary policy, fiscal policy, supply of credit, international trade openness, and business cycle fluctuations.

4.1  Regional characteristics of macroeconomic drivers

This section investigates to what extent East Asia's regional housing price cycle is attributable to the region-specific sources of variation in the macroeconomic drivers. To this end, the regional factor of housing price cycles is regressed on the regional factors of five macroeconomic drivers:
formula
5
where is a regional factor shared by the six East Asian economies for . Subscript denotes the housing price cycle, and stand for the macroeconomic drivers monetary aggregate, government expenditure, exports and imports, credit supply, and business cycle fluctuations. Taking the variance operator to both sides of equation (5), we can decompose the variance of the estimated regional housing price cycle into a regional component of each variable and an idiosyncratic component. This is summarized in Table 3, which reveals that both trade openness and credit supply are the major driving forces in the East Asia regional housing price cycle.
Table 3. 
Variance decompositions of East Asia's regional characteristics (%)
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
3.72 0.00 16.30 12.59 0.25 32.86 
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
3.72 0.00 16.30 12.59 0.25 32.86 

Source: Authors’ calculations.

Next, we perform variance decompositions for the national housing price cycle in East Asia. Equation (6) is estimated to decompose the variance of national housing price cycles:
formula
6
where is the growth cycle of the real housing price index for economy , and is the East Asia regional factor for .

Table 4 tabulates the variance decompositions for the East Asia region-specific factors of the drivers. On average, the East Asia regional components of macroeconomic drivers account for a significant share of the variations (28.22 percent). In particular, 46.87 percent of Singapore's housing price cycles are explained by the regional components of the macroeconomic drivers, followed by 39.06 percent, 38.82 percent, and 25.17 percent for Japan, Hong Kong, and China, respectively. In addition, the international driver of trade openness is the most important regional determinant for Hong Kong and Singapore (37.71 percent and 32.60 percent, respectively). On the other hand, the contributions of regional factors of macroeconomic drivers are relatively small in Korea, which suggests that the country-specific factors are important in accounting for its housing price cycle.

Table 4. 
Variance decompositions with respect to regional factors of macroeconomic drivers (%)
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
China 1.74 2.68 12.81 6.78 1.16 25.17 
Hong Kong 0.04 0.58 37.71 0.30 0.19 38.82 
Japan 8.88 0.51 7.50 10.25 11.92 39.06 
Korea 2.14 1.23 1.80 1.83 2.35 9.35 
Singapore 4.05 0.05 32.60 6.38 3.79 46.87 
Taiwan 0.30 0.59 0.43 6.91 1.85 10.08 
Average 2.86 0.94 15.48 5.41 3.54 28.22 
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
China 1.74 2.68 12.81 6.78 1.16 25.17 
Hong Kong 0.04 0.58 37.71 0.30 0.19 38.82 
Japan 8.88 0.51 7.50 10.25 11.92 39.06 
Korea 2.14 1.23 1.80 1.83 2.35 9.35 
Singapore 4.05 0.05 32.60 6.38 3.79 46.87 
Taiwan 0.30 0.59 0.43 6.91 1.85 10.08 
Average 2.86 0.94 15.48 5.41 3.54 28.22 

Source: Authors’ calculations.

4.2  National characteristics of macroeconomic drivers

We now turn to the explanatory power of country-specific components of the drivers. Equation (7) is estimated to determine the contribution of country-specific macroeconomic factors to national house price cycles, and the results are presented in Tables 5 through 10.
formula
7
where is the growth cycle of real housing prices for economy , and is a country-specific factor of the driver variables .
Table 5. 
Variance decompositions with respect to China's country factor of macroeconomic drivers (%)
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
China 6.78 0.36 0.84 2.90 7.49 18.37 
Hong Kong 1.08 0.41 1.13 0.24 2.80 5.66 
Japan 0.18 0.11 9.02 0.03 0.00 9.34 
Korea 0.14 0.45 6.50 2.42 1.35 10.86 
Singapore 4.72 7.57 2.62 0.32 5.04 20.27 
Taiwan 0.30 2.84 0.11 1.84 0.04 5.13 
Average 2.20 1.96 3.37 1.29 2.79 11.60 
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
China 6.78 0.36 0.84 2.90 7.49 18.37 
Hong Kong 1.08 0.41 1.13 0.24 2.80 5.66 
Japan 0.18 0.11 9.02 0.03 0.00 9.34 
Korea 0.14 0.45 6.50 2.42 1.35 10.86 
Singapore 4.72 7.57 2.62 0.32 5.04 20.27 
Taiwan 0.30 2.84 0.11 1.84 0.04 5.13 
Average 2.20 1.96 3.37 1.29 2.79 11.60 

Source: Authors’ calculations.

The overall picture that emerges from Tables 5 through 10 is that Japan's macroeconomic drivers account for the largest portion of housing price variations (21.24 percent in total), followed by Hong Kong (19.98 percent), Korea (17.04 percent), Taiwan (14.70 percent), China (11.60 percent), and Singapore (10.63 percent). In Table 5, 20.27 percent of Singapore's housing price variation is explained by China's macroeconomic drivers. Table 6 shows that 42.68 percent of Singapore's housing price growth cycle is explained by Hong Kong's macroeconomic drivers and 21.26 percent by China’s. Table 7 shows that Japan's macroeconomic drivers account for around 20 percent of most East Asian economies’ housing price variations: Singapore (26.13 percent), Taiwan (25.28 percent), Hong Kong (23.37 percent), and Korea (19.50 percent). The results in Table 8 indicate that 24 percent of China's and Hong Kong's housing price variations are explained by Korea's macroeconomic drivers. In Table 9, Singapore's macroeconomic drivers account for 19.28 percent of China's housing price variations and 16.39 percent of Korea's housing price fluctuations. In Table 10, 25.38 percent of China's and 22.31 percent of Singapore's housing price cycles are explained by Taiwan's macroeconomic drivers.

Table 6. 
Variance decompositions with respect to Hong Kong's country factor of macroeconomic drivers (%)
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
China 1.45 0.96 2.55 7.94 8.37 21.26 
Hong Kong 6.60 0.07 1.49 2.65 20.36 31.18 
Japan 0.68 0.99 0.02 1.39 0.03 3.11 
Korea 0.39 8.19 0.03 0.14 0.06 8.81 
Singapore 2.63 0.23 3.53 0.00 36.30 42.68 
Taiwan 0.61 2.34 0.62 0.76 8.52 12.85 
Average 2.06 2.13 1.37 2.15 12.27 19.98 
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
China 1.45 0.96 2.55 7.94 8.37 21.26 
Hong Kong 6.60 0.07 1.49 2.65 20.36 31.18 
Japan 0.68 0.99 0.02 1.39 0.03 3.11 
Korea 0.39 8.19 0.03 0.14 0.06 8.81 
Singapore 2.63 0.23 3.53 0.00 36.30 42.68 
Taiwan 0.61 2.34 0.62 0.76 8.52 12.85 
Average 2.06 2.13 1.37 2.15 12.27 19.98 

Source: Authors’ calculations.

Table 7. 
Variance decompositions with respect to Japan's country factor of macroeconomic drivers (%)
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
China 0.91 4.56 7.15 0.02 0.46 13.09 
Hong Kong 0.03 0.90 1.65 10.38 10.40 23.37 
Japan 15.82 0.50 0.00 3.04 0.70 20.06 
Korea 2.71 1.97 0.01 5.46 9.34 19.50 
Singapore 2.86 0.22 4.95 2.28 15.82 26.13 
Taiwan 16.98 3.47 0.02 1.81 3.00 25.28 
Average 6.55 1.94 2.30 3.83 6.62 21.24 
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
China 0.91 4.56 7.15 0.02 0.46 13.09 
Hong Kong 0.03 0.90 1.65 10.38 10.40 23.37 
Japan 15.82 0.50 0.00 3.04 0.70 20.06 
Korea 2.71 1.97 0.01 5.46 9.34 19.50 
Singapore 2.86 0.22 4.95 2.28 15.82 26.13 
Taiwan 16.98 3.47 0.02 1.81 3.00 25.28 
Average 6.55 1.94 2.30 3.83 6.62 21.24 

Source: Authors’ calculations.

Table 8. 
Variance decompositions with respect to Korea's country factor of macroeconomic drivers (%)
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
China 0.99 9.25 8.13 1.39 4.87 24.64 
Hong Kong 3.00 1.14 11.88 2.32 6.35 24.68 
Japan 7.08 0.05 0.51 2.83 0.21 10.68 
Korea 0.39 1.12 0.57 23.28 0.30 25.66 
Singapore 1.57 0.27 5.05 0.01 0.56 7.46 
Taiwan 0.01 0.84 0.27 4.09 3.91 9.13 
Average 2.17 2.11 4.40 5.65 2.70 17.04 
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
China 0.99 9.25 8.13 1.39 4.87 24.64 
Hong Kong 3.00 1.14 11.88 2.32 6.35 24.68 
Japan 7.08 0.05 0.51 2.83 0.21 10.68 
Korea 0.39 1.12 0.57 23.28 0.30 25.66 
Singapore 1.57 0.27 5.05 0.01 0.56 7.46 
Taiwan 0.01 0.84 0.27 4.09 3.91 9.13 
Average 2.17 2.11 4.40 5.65 2.70 17.04 

Source: Authors’ calculations.

Table 9. 
Variance decompositions with respect to Singapore's country factor of macroeconomic drivers (%)
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
China 8.42 7.15 1.70 0.85 1.16 19.28 
Hong Kong 0.82 1.76 0.32 3.69 0.04 6.62 
Japan 0.27 0.83 0.32 1.15 3.55 6.13 
Korea 0.18 0.22 2.59 11.71 1.69 16.39 
Singapore 4.80 0.01 0.74 0.50 0.06 6.10 
Taiwan 0.29 0.09 0.75 3.21 4.93 9.28 
Average 2.47 1.68 1.07 3.52 1.90 10.63 
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
China 8.42 7.15 1.70 0.85 1.16 19.28 
Hong Kong 0.82 1.76 0.32 3.69 0.04 6.62 
Japan 0.27 0.83 0.32 1.15 3.55 6.13 
Korea 0.18 0.22 2.59 11.71 1.69 16.39 
Singapore 4.80 0.01 0.74 0.50 0.06 6.10 
Taiwan 0.29 0.09 0.75 3.21 4.93 9.28 
Average 2.47 1.68 1.07 3.52 1.90 10.63 

Source: Authors’ calculations.

Table 10. 
Variance decompositions with respect to Taiwan's country factor of macroeconomic drivers (%)
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
China 0.02 5.31 9.27 7.39 3.39 25.38 
Hong Kong 1.21 0.56 1.26 3.61 0.12 6.76 
Japan 3.86 0.03 0.02 3.90 0.58 8.38 
Korea 2.88 0.02 0.01 2.01 3.08 7.99 
Singapore 9.78 0.76 8.54 2.33 0.89 22.31 
Taiwan 5.22 3.23 0.00 0.92 7.97 17.35 
Average 3.83 1.65 3.18 3.36 2.67 14.70 
Monetary policyFiscal policyTrade opennessCredit supplyBusiness cycleTotal
China 0.02 5.31 9.27 7.39 3.39 25.38 
Hong Kong 1.21 0.56 1.26 3.61 0.12 6.76 
Japan 3.86 0.03 0.02 3.90 0.58 8.38 
Korea 2.88 0.02 0.01 2.01 3.08 7.99 
Singapore 9.78 0.76 8.54 2.33 0.89 22.31 
Taiwan 5.22 3.23 0.00 0.92 7.97 17.35 
Average 3.83 1.65 3.18 3.36 2.67 14.70 

Source: Authors’ calculations.

The fiscal policy driver is generally not found to be significant in accounting for housing price variations across East Asian economies.

The country-specific factor is important in China (18.37 percent), Hong Kong (31.18 percent), Japan (20.06 percent), Korea (25.66 percent), and Taiwan (17.35 percent), but relatively less important in Singapore (6.10 percent). In particular, the domestic monetary policy driver is crucial for China (6.78 percent), Hong Kong (6.60 percent), Japan (15.82 percent), Singapore (4.80) and Taiwan (5.22 percent). Domestic business cycle fluctuations are also significant in China (7.49 percent), Hong Kong (20.36 percent), and Taiwan (7.97 percent).

Credit supply is the most important driver for Korea, accounting for 23.28 percent of housing price variations. The finding implies that credit supply measures may be an effective tool to stabilize Korea's housing market.

4.3  Country-specific policies and factors impacting housing markets

This section discusses country-specific factors and policies impacting housing markets. The results in Tables 5, 6, and 10 suggest that the business cycle and monetary policy are a significant influence in the housing market in China, Hong Kong, and Taiwan. Nevertheless, widespread housing price boom-and-bust experiences across major Chinese cities suggest that local government policies may also play a key role. In Taiwan, a sustained housing price boom has been observed since 2006. Our analysis reveals that the boom was attributable to economic growth and expansionary monetary policy. In addition, Taiwanese housing market participants had optimistic expectations about Chinese investments in Taiwan, which led to the signing of a financial supervisory cooperation's memorandum of understanding (MOU) between China and Taiwan.

For the Japanese housing market, Table 7 shows that monetary policy has the largest impact on housing price cycles. Interest rate and credit controls began to be removed in the early 1980s in Japan, but the process was not completed until the mid 1990s. Following the deregulation of financial markets, including the mortgage market, Japan moved toward more expansionary monetary policy with quantitative easing between 2001 and 2006 and more competitive housing finance, which gave housing market participants easier access to mortgage credit due to more diverse funding sources, lender types, and loan products (IMF 2008). Moreover, as emphasized by Seko, Sumita, and Naoi (2011), the real estate investment trusts (REITs) industry played a crucial role in the late 2000s and the U.S. sub-prime mortgage shock increased the demand for Japanese REITs.

The Korean housing market also experienced a housing price boom-and-bust cycle. The rapid housing price appreciation was countered by various policy interventions and counter-speculative policy measures. Table 8 reveals that the credit supply driver was the most important for Korea. As of 2012, mortgage debt outstanding amounts to 35 percent of Korea's GDP. Even though the central bank cannot accomplish housing price stability completely through monetary policy only, the monetary authority tends to be increasingly alert to the risk of short-term housing price behavior and its boom-and-bust cycles. Lending-restrictions on home mortgage lending, such as loan-to-value and debt-to-income ceilings, might have been effective in responding to the housing price boom in Korea.6

The housing price cycles in Singapore and Hong Kong are accounted for by the regional factors of macroeconomic drivers to a great extent. In particular, openness to international trade is the most important regional determinant for Hong Kong and Singapore. These economies are highly exposed to trade liberalization and have pursued an open-door policy toward becoming offshore financial hubs. Furthermore, Cho, Kim, and Renaud (2012) discuss the role of large foreign capital flows into these economies following a decline in world interest rates and abundant global liquidity. Moreover, Singapore has experienced a housing price boom thanks to active job creation, and a rapid influx of foreigners following the change in the country's immigration policy to attract highly skilled workers.

5.  Housing price cycle dynamics in Greater China

In this section, we examine housing price cycle dynamics in the Greater China Region. We also include Singapore in the analysis. Cho, Kim, and Renaud (2012) suggest a plausible China effect in housing price movements in the mid and late 2000s. They mention that, because of expectations of positive spillovers from China's economic performance, housing prices in the Greater China region share similar trends from the mid 2000s and again after the global financial crisis. Building on this conjecture, we examine the dynamic relation of housing price cycles in Greater China and also Singapore.

As a preliminary examination we estimate the dynamic factor model (equation (1)) by separating the Greater China and Singapore economies from the East Asia region. That is, we consider five regions: North America, Europe, Oceania, Northeast Asia (consisting of Japan and Korea), and Greater China and Singapore. Comparing Table 11 (case of five regions) with Table 1 (case of four regions), the share of the regional factor for the Greater China and Singapore economies gets significantly larger (13.80 percent vs. 18.32 percent for Singapore, 9.98 percent vs. 20.41 percent for Taiwan). This suggests that co-movement in housing prices is more pronounced in the Great China region and Singapore than in the six economies in East Asia and raises the question whether China leads housing price cycles in the Greater China region and Singapore.

Table 11. 
Variance decompositions of housing price growth cycles in Greater China and Singapore (%)
World factorRegional factorCountry factor
China 18.68 15.15 66.18 
Hong Kong 4.53 20.93 74.54 
Singapore 12.58 18.32 69.10 
Taiwan 1.22 20.41 78.37 
World factorRegional factorCountry factor
China 18.68 15.15 66.18 
Hong Kong 4.53 20.93 74.54 
Singapore 12.58 18.32 69.10 
Taiwan 1.22 20.41 78.37 

Source: Authors’ calculations.

In the subsequent analysis, we use the country-specific component of the housing price cycle after removing the world and region-specific components from the dynamic factor model (equation (1)) of five regions. We apply the following two conditions to determine if a variable leads another variable : (i) is weakly exogenous, and (ii) Granger-causes . To examine a China effect, the variables and correspond to the country-specific component of the Chinese housing price cycle and the other economy's housing price cycle.

Unit root tests reveal that the four series of housing price cycles in Greater China and Singapore are stationary. Before estimating a vector autoregression model, we perform a block-exogeneity test through a likelihood ratio test to determine the weakly exogenous variables in the vector autoregression:
formula
8
where and are the variance-covariance matrix of the restricted and unrestricted model, is the number of observations, and is the number of coefficient estimates in the unrestricted model. The test statistic (equation (8)) follows a distribution with the degree of freedom equal to the number of restrictions. The null hypothesis is that the coefficients of all other lagged variables are zero. Table 12 indicates that both China's and Singapore's housing price cycles are weakly exogenous, and both Hong Kong's and Taiwan's housing price cycles are incorporated as endogenous variables in a block-exogenous vector autoregression model.
Table 12. 
Block Exogeneity test
Test statisticDegree of freedomp-value
China 3.328 0.767 
Hong Kong 11.554* 0.073 
Singapore 10.343 0.111 
Taiwan 17.409*** 0.008 
Test statisticDegree of freedomp-value
China 3.328 0.767 
Hong Kong 11.554* 0.073 
Singapore 10.343 0.111 
Taiwan 17.409*** 0.008 

Source: Authors’ calculations.

***Statistically significant at the 1% level; *statistically significant at the 10% level.

Because China's and Singapore's housing price cycles turn out to be weakly exogenous, we impose block-exogenous restrictions that the exogenous variables are not affected by the lagged endogenous variables. Thus, a block-exogenous vector autoregression is modeled as follows:
formula
where is a vector of exogenous variables and is a vector of endogenous variables. According to Table 13, we find that China leads Taiwan's housing price cycle and Taiwan leads Hong Kong's housing price cycle in a transitive way. One of the reasons for this indirect China effect on Taiwan could be market participants’ optimistic expectations of Chinese investments in Taiwan in the mid 2000s, which led to the signing of the financial supervisory cooperation's MOU and the Economic Cooperation Framework Agreement. In addition, Singapore leads Taiwan's housing cycle, implying that a China effect is evident for Taiwan in a direct way and for Hong Kong in an indirect way between 2001 and 2010.
Table 13. 
Granger causality test
Null hypothesisTest statisticp-value
China does not G.C. Taiwan 5.816* 0.055 
Hong Kong does not G.C. Taiwan 7.879** 0.019 
Singapore does not G.C. Taiwan 8.097** 0.017 
China does not G.C. Hong Kong 0.627 0.627 
Singapore does not G.C. Hong Kong 1.283 0.527 
Taiwan does not G.C. Hong Kong 7.078** 0.029 
Null hypothesisTest statisticp-value
China does not G.C. Taiwan 5.816* 0.055 
Hong Kong does not G.C. Taiwan 7.879** 0.019 
Singapore does not G.C. Taiwan 8.097** 0.017 
China does not G.C. Hong Kong 0.627 0.627 
Singapore does not G.C. Hong Kong 1.283 0.527 
Taiwan does not G.C. Hong Kong 7.078** 0.029 

Source: Authors’ calculations.

Note: G.C. = Granger-cause.

**Statistically significant at the 5% level; *statistically significant at the 10% level.

6.  Conclusion

The purpose of this paper was to examine the characteristics of housing price cycles in six East Asian countries during the period 2001:Q1 to 2010:Q1. We investigated housing price cycle co-movements and the extent to which macroeconomic driving forces are significant in accounting for housing price fluctuations.

The key findings can be summarized as follows. First, the largest portion of housing price cycles in East Asian economies is explained by region-specific and country-specific factors, rather than a world factor. Second, we find evidence that East Asia's regional housing price cycles co-move strongly with the world housing price cycle in the long run, but relatively weak co-movement is found in the short run. Third, the economies in the Greater China region and Singapore co-move with Northeast Asia's regional housing price cycle in the long run but not in the short run.

As for determining significant region-specific factors of macroeconomic drivers, international trade openess is the major driving force of the regional factor of housing price cycles in East Asian economies. Moreover, the regional factors of macroeconomic drivers are important in accounting for housing price cycles in Singapore, Japan, and Hong Kong. However, their contributions are relatively small in Korea. Regarding country-specific factors of macroeconomic drivers, domestic monetary policy and business cycle fluctuations are important in explaining housing price movements in China, Hong Kong, Japan, and Taiwan, whereas fiscal policy is found to be insignificant. By contrast, credit supply proves to be the most important influence in Korea. Moreover, the existence of a China effect is found in the Greater China region and Singapore, especially for Taiwan and Hong Kong between 2001 and 2010. For Hong Kong, we find an indirect linkage running from China to Hong Kong via Taiwan.7

Because we intended to identify macroeconomic driving forces of international housing price cycles, we did not consider other socioeconomic or microeconomic elements, such as demographic structure, supply-side regulations, and the role of taxation. In fact, all six East Asian economies considered in this study are experiencing rapid population aging and shrinking family size. It would be interesting to investigate the interrelationship of major metropolitan areas in East Asia by considering these socioeconomic determinants. Another future research agenda is related to the effects of global liquidity expansion on housing price cycles, especially after the global financial crisis. This is not examined here because of the short time span currently available after the global financial crisis.

References

Baxter
,
Marianne
, and
Robert G.
King
.
1995
.
Measuring Business Cycles: Approximate Band-Pass Filters for Economic Time Series
. NBER Working Paper No. 5022.
Cambridge, MA
:
National Bureau of Economic Research
.
Beltratti
,
Andrea
, and
Claudio
Morana
.
2010
.
International House Prices and Macroeconomic Fluctuations
.
Journal of Banking and Finance
34
(
3
):
533
545
.
Bordo
,
Michael D.
, and
Olivier
Jeanne
.
2002
.
Boom–Busts in Asset Prices, Economic Instability, and Monetary Policy
. NBER Working Paper No. 8966.
Cambridge, MA
:
National Bureau of Economic Research
.
Cho
,
Man
,
Kyung-Hwan
Kim
, and
Bertrand
Renaud
.
2012
.
Real Estate Volatility and Economic Stability: An East Asian Perspective
.
KDI Policy Study 2012. Seoul, Korea
:
Korea Development Institute
.
Claessens
,
Stijn, M.
Ayhan
Kose
, and
Marco E.
Terrones
.
2009
.
What Happens During Recessions, Crunches and Busts
?
Economic Policy
24
:
653
700
.
Croux
,
Christophe
,
Mario
Forni
, and
Lucrezia
Reichlin
.
2001
.
A Measure of Co-movement for Economic Variables: Theory and Empirics
.
Review of Economics and Statistics
83
(
2
):
232
241
.
Engle
,
Robert
.
2002
.
Dynamic Conditional Correlation: A Simple Class of Multivariate Generalized Autoregressive Conditional Heteroskedasticity Models
.
Journal of Business and Statistics
20
(
3
):
339
350
.
Gong
,
Chi
, and
Soyoung
Kim
.
2013
.
Economic Integration and Business Cycle Synchronization in Asia
.
Asian Economic Papers
12
(
1
):
76
99
.
Igan
,
Deniz
,
Alain
Kabundi
,
Francisco
Nadal De Simone
,
Marcelo
Pinheiro
, and
Natalia
Tamirisa
.
2011
.
Housing, Credit, and Real Activity Cycles: Characteristics and Co-movement
.
Journal of Housing Economics
20
(
3
):
210
231
.
Igan
,
Deniz
, and
Heedon
Kang
.
2011
.
Do Loan-to-Value and Debt-to-Income Limits Work? Evidence from Korea. IMF Staff Paper No. WP/11/297
.
Washington, DC
:
International Monetary Fund
.
IMF
.
2008
.
The Changing Housing Cycle and the Implications for Monetary Policy
. In:
World Economic Outlook: Housing and the Business Cycle
, pp.
103
132
.
Washington, DC
:
International Monetary Fund
.
Jung
,
Yongseung
,
Soyoung
Kim
,
Doo
Yong Yang
, and
Tack
Yun
.
2013
.
Are Asian Business Cycles Different
?
Asian Economic Papers
12
(
3
):
94
113
.
Kose
,
M. Ayhan
,
Christopher
Otrok
, and
Charles H.
Whiteman
.
2003
.
International Business Cycles: World, Region, and Country-Specific Factors
.
American Economic Review
93
(
4
):
1216
1239
.
Seko
,
Miki
,
Kazuto
Sumita
, and
Michio
Naoi
.
2011
.
The Recent Financial Crisis and the Housing Market in Japan
. In:
Global Housing Markets: Crises, Policies, and Institutions
, edited by
Ashok
Bardhan
,
Robert
Edelstein
, and
Cynthia
Kroll
, pp.
357
375
.
Kolb Series in Finance. Hoboken, NJ
:
John Wiley & Sons, Inc
.

Notes

*

We would like to thank Bhanupong Nidhiprabha, Anwar Nasution, Deborah Swenson, Wing Thye Woo, and participants at the 2014 Asian Economic Panel Meeting for valuable comments. Kyung-Hwan Kim was at the Korea Research Institute for Human Settlements when the paper was submitted, and he is currently at the Ministry of Land, Infrastructure, and Transport of the Republic of Korea. Young-Joon Park is the corresponding author.

1 

Because the six East Asian economies are constrained by different supply-side conditions in their housing markets, including housing supply elasticity, we restrict our analysis to the demand-side macroeconomic driving forces in accounting for housing price cycles. For example, Singapore operates the most active program among the six East Asian economies to support homeownership through its unique system of integrating housing supply, financing and saving with the aid of the national pension fund. In China, local governments play a key role in housing supply because they are involved in managing land sales which are a major source of revenue. On the other hand, tight land-use regulations are in operation in Korea and Hong Kong (Cho, Kim, and Renaud 2012).

2 

The United States and Canada in North America; the UK, Germany, France, and Italy in Europe; Australia and New Zealand in Oceania; China, Hong Kong, Japan, Korea, Singapore, and Taiwan in East Asia.

3 

Mortgage debt is measured by total domestic credit less net claims on the government sector and claims on other financial institutions.

4 

The German housing market is regarded as an outlier and thus the world factor accounts for a minute portion of German housing price variations. One of the reasons is that German municipal authorities consistently encourage housing supply by releasing land for development on a regular basis.

5 

Throughout this paper, three frequency ranges are used to interpret the dynamic correlation in the frequency domain. The business cycle frequency, which corresponds to 6 to 32 quarters’ periodicity as indicated in Baxter and King (1995), refers to the intermediate frequencies around 1.5 on the horizontal axis in Figure 4. Accordingly, the low and high frequencies are associated with the longer and shorter periodicities than the business cycle frequency.

6 

For instance, Igan and Kang (2011) provide empirical evidence that the debt-to-income restriction was more effective than the loan-to-value regulation.

7 

The fact that there is housing price cycle co-movement among the four economies in Greater China and that Japan is an outlier imply that the contribution of the regional factor to each country's housing price cycle might have been underestimated. It might be that what happens to the China region spills over to Korea, whereas Japan operates on its own.