Chapter 4 Measure of Housing Bubbles
4.3 Status of Housing Bubbles
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
4.3 Status of Housing Bubbles
As the literature review implies, the definition of housing bubbles can be in a number of ways. From the point of fundamental value, this study defines housing bubbles as the discrepancy between housing fundamental values and market prices.
The formula of housing bubbles is as follows:
Bubble = (𝑀𝑃𝐹𝑉 − 1) × 100% (4.1)
In this equation, MP represents the market price of property and FV represents the housing fundamental value. The results are reported in Figure 4-7 and Figure 4-8.
However, the fundamental values in this study do not consider the tax shield. It would make the calculated housing bubbles become bigger than the actuals but do not affect the trends.
Figure 4- 7 The Housing Bubbles in Beijing, Shanghai and Guangzhou (50% LTV)
-50.00%
0.00%
50.00%
100.00%
150.00%
200.00%
250.00%
2007/1 2007/4 2007/7 2007/10 2008/1 2008/4 2008/7 2008/10 2009/1 2009/4 2009/7 2009/10 2010/1 2010/4 2010/7 2010/10 2011/1 2011/4 2011/7 2011/10 2012/1 2012/4 2012/7 2012/10
Beijing Shanghai Guangzhou
‧
Figure 4- 8 The Housing Bubbles in Beijing, Shanghai and Guangzhou (70% LTV)
As shown in Figure 4-7, from 2007 to 2009, the housing bubbles in Guangzhou were in the first place and the ones in Beijing come second, while the bubbles in Shanghai were smallest. In this period, although the housing prices in Beijing and Shanghai were higher than the ones in Guangzhou, the housing bubbles in Guangzhou were most serious. It might relate to the seriously low housing rents in Guangzhou. In 2008, the subprime mortgage crisis burst out. The housing bubbles in three cities all reduced. By early 2009, the housing bubbles in three cities reached their lowest levels and even the ones in Shanghai were negative briefly. After that, the real estate markets in three cities shook off the influence of the subprime mortgage crisis and began to recover gradually. The housing bubbles were going up again. However, in April 2010, a series of house purchase restrictions were announced. At that time, the housing bubbles in Beijing and Shanghai suddenly dropped and then ballooned. When the restrictions were officially implemented, the trends of the bubbles in Beijing and Shanghai merely changed a little. It is thus clear that the house purchase restrictions can curb over-heating property markets in a short time. At the same time, the restrictions also gave confidence to the people who would buy houses to live in. They purchased a large number of properties and made the real estate markets take a good turn. On the other hand, the housing bubbles in Guangzhou edged up after 2009. It is possible that the home buyers in Guangzhou took a wait-and-see attitude towards the
0.00%
2007/1 2007/4 2007/7 2007/10 2008/1 2008/4 2008/7 2008/10 2009/1 2009/4 2009/7 2009/10 2010/1 2010/4 2010/7 2010/10 2011/1 2011/4 2011/7 2011/10 2012/1 2012/4 2012/7 2012/10
Beijing Shanghai Guangzhou
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
real estate market.
In Figure 4-8, we can see that the housing bubbles which are calculated under the circumstances of 70% loan-to-value ratio are higher than the ones under the circumstances of 50% loan-to-value ratio. It is more consistent with the practice.
From January 2007 to November 2012, there are 21 months in which the housing bubbles in Beijing were more than 100%. And the highest point (135.8%) arose in February 2011. In Shanghai, the housing bubbles exceeded 100% for 27 times and their peak (144.44%) appeared in January 2012. When it comes to Guangzhou, there are 21 months that the housing bubbles were higher than 100%. And the bubbles in Guangzhou peaked at 214.40% in October 2007. Thus it can be seen that the housing market prices deviated from the fundamental values seriously in Beijing, Shanghai and Guangzhou. The problem of high housing bubbles cannot be ignored any more.
‧
This chapter mainly presents the relationship between housing bubbles and the selected variables. First, the existence of structural change and the unit root are checked. Then the results of cointegration test are interpreted. Finally, the results of Granger causality tests provide suggestions for housing polices.
Monthly data or quarterly data usually exhibits some form of seasonality (Enders, 1995). The seasonal variation of time series may cover up the actual trends of economic processes. So we should deseasonalize the time series before the econometric analysis. The variables used in this study, except interest rates, have obvious seasonal patterns. Accordingly, they are seasonally adjusted by using Census X-12 method3.
5.1 Structural Change
Before performing unit root test, it is necessary to confirm that there are no structural changes during the research period. If there are structural breaks, the results of the following econometric analysis would be biased. In order to avoid the deviated estimation, we would identify whether the time series of housing bubbles has structural changes or not. In this study, we apply the Cumulative Sum of the recursive residuals test (CUSUM test).
The results of CUSUM test are shown respectively in Figure 5-1, Figure 5-2 and Figure 5-3. Apparently, from Figure 5-1 to Figure 5-3, the housing bubbles in three cities do not exceed the critical value at the 5% significance level (dashed line). In other words, there are no structural changes in these series of housing bubbles during the research period. Therefore, these time series are appropriate for the following tests, and the results would be authentic.
3 Census X-12 method is a seasonal adjusted procedure established by Bureau of Census. This procedure is revised for many times and becomes one of the most usual seasonal adjusted methods. Much statistical software can perform the Census X-12 procedure.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Figure 5- 1 Result of CUSUM Test for BJ
Figure 5- 2 Result of CUSUM Test for SH
-30 -20 -10 0 10 20 30
II III IV I II III IV I II III IV I II III IV I II III IV I II III IV
2007 2008 2009 2010 2011 2012
CUSUM 5% Significance
-30 -20 -10 0 10 20 30
II III IV I II III IV I II III IV I II III IV I II III IV I II III IV
2007 2008 2009 2010 2011 2012
CUSUM 5% Significance
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Figure 5- 3 Result of CUSUM Test for GZ
-30 -20 -10 0 10 20 30
II III IV I II III IV I II III IV I II III IV I II III IV I II III IV
2007 2008 2009 2010 2011 2012
CUSUM 5% Significance
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
5.2 Unit Root Test
In the time series data analysis, it is essential to confirm whether the series are stationary or not. With the non-stationary variables, the estimates of the classic regression model might be spurious regression4 results. Thus, we employ Augmented Dickey-Fuller (ADF) test and Phillips-Perron (PP) test to examine all the variables used in this study.
The result of unit root tests is reported in Table 5-1. We can learn that housing bubbles in the three cities (BJ, SH and GZ), per-capita disposable incomes in these cities (INCOME_BJ, INCOME_SH and INCOME_GZ), total loans of financial institutions (LOAN), and interest rates of mortgage (INT) accept the null hypothesis at the level order, i.e., these time series are non-stationary. However, they reject the null hypothesis at the first difference order, becoming stationary series. Hence, all the variables used in this study are I(1) series.
Table 5- 1 Result of Unit Root Test
ADF test PP test
Level 1st difference Level 1st difference
BJ -2.544 -8.952 *** -2.593 * -8.970 ***
SH -1.796 -9.567 *** -1.813 -9.472 ***
GZ -1.080 -7.028 *** -1.546 -7.236 ***
INCOME_BJ 0.966 -6.284 *** 1.707 -6.652 ***
INCOME_SH 1.546 -4.160 *** 1.719 -11.027 ***
INCOME_GZ 0.189 -9.427 *** 1.646 -11.985 ***
LOAN 3.043 -6.207 *** 2.228 -6.529 ***
INT -2.086 -2.875 * -1.448 -3.018 **
Note: 1. Null hypothesis: The series has a unit root.
2. *, ** and *** denote significance at the 10%, 5% and 1% level respectively.
‧
In order to investigate the long-run equilibrium relationship between the housing bubbles and other variables, we need to detect whether cointegrations exist in the variables first. Since all the variables are I(1) series, Johansen cointegration test is available in this study. If cointegrations exist, it means that there are long-run relationships in the variables. So we can directly employ the standard values to estimate VECM. However, if there is no cointegration in the variables, we would apply the VAR model to examine the short-run equilibrium relationship between the variables in the first difference.
5.3.1 The Optimal Lag Length
The result of Johansen cointegration test based on the VAR approach is sensitive to the lag length. To ensure the accuracy of the following analysis, it is very important to determine the lag order of an unrestricted VAR. In this study, the optimal lag length is selected by Akaike information criterion (AIC) and Likelihood ratio (LR) test statistic. The results are shown in Table 5-2. As we can see, the third lag is suitable for these three VAR models.
Table 5- 2 The Optimal Lag Order
INCOME_BJ, LOAN, INT and BJ
INCOME_SH, LOAN, INT and SH
Note: * denotes the lag order selected by LR and AIC.
‧
cointegration relationships. The results of cointegration test are reported in Table 5-3.At the 5% significance, there is a cointegration relationship between BJ and other variables. Similarly, under the Trace and Maximum Eigenvalue tests, the results of cointegration between INCOME_SH, LOAN, INT and SH reject the null hypothesis of no cointegration at the 5% significance. Last but not least, the test statistics also show that two cointegration relationships exist between INCOME_GZ, LOAN, INT and GZ. In other words, it is appropriate to employ VECM to discuss the housing bubbles in three cities respectively.
Table 5- 3 Results of Cointegration Test
INCOME_BJ, LOAN, INT and BJ
Note: ** denotes that the null hypothesis can be rejected at the 5% significance.‧
results of VECM analysis are reported in Table 5-4, Table 5-5 and Table 5-6.From Table 5-4, we can learn that the current BJ is affected by the first and second lag of itself. Furthermore, BJ is also influenced by the first lag of INT and the second lag of LOAN. Moreover, BJ is negatively related to itself. With the increase of the housing bubbles in Beijing, the investors predict that the bubbles are in the danger of collapse and prefer to resale the houses. Therefore, the housing bubbles in Beijing can curb themselves when they keep rising. And INT is positively related to BJ at the first lag, which is inconsistent with the general economic theory. This is because the influence of interest rates is reflected more rapidly in fundamental value than in market price. The impact of LOAN on BJ is positive. Once the banks tighten the credit quota which makes the financial market liquidity weakened, the investors’
capital chains are affected. It should be noted that the impact of INCOME_BJ on BJ is not significant. Combined with the influence of LOAN, we can infer that the housing bubbles in Beijing rely more on the financial activities than real economy.
In the adjusting speed, the significant parameter denoted as CointEq1 or CointEqe2 represents the short-run deviations from equilibrium. When the series deviates the long-run equilibrium, they will be adjusted towards the equilibrium by a specific rate. However, as the coefficient of CointEq1 in the VECM analysis on BJ is not significant, it implies the deviation will not be adjusted immediately. Nevertheless, the deviation will be finally adjusted.
As reported in Table 5-5, the current SH is strongly affected by not only the first lag of itself but also the first lag of LOAN. Similar with BJ, SH is negatively related to itself. The rational investors in Shanghai are also afraid that the ballooning housing bubbles are going to collapse. In addition, the impact of LOAN on SH is positive. So it seems that the easy credit policy also encourages overinvestment in Shanghai’s real estate market. The coefficients of INCOME_SH are not significant. Clearly, the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
growth rate of INCOME_SH was far lower than that of housing bubbles. Although the coefficients of INT are also not significant, the directions of INT in Shanghai are the same as the ones in Beijing. In all probability, the interest rates are pretty low for the investors in Shanghai, which leads to a large-scale financing campaign and the inflating housing bubbles. At last, the CointEq1 is also not significant in the VECM analysis on SH. But over time, SH will return to the long-run equilibrium value.
As shown in Table 5-6, the current GZ is strongly affected by the first lag of itself as well as the first lag of INCOME_GZ. Besides, the current GZ is also affected by not only the first and second lag of INT but also the second lag of LOAN. The impact of GZ on itself is negative, which is the same situation as BJ and SH. In Guangzhou’s real estate market, the investors would also intend to exit the market when the housing bubbles keep increasing. And INCOME_GZ is positively related to GZ. It implies that Guangzhou’s income growth can chase the housing bubbles’. People can probably afford the houses in Guangzhou. Specially, the impact of INT on GZ is positive first and then turns negative. When the interest rates increase, the fundamental values drop immediately and the market prices give no response temporarily. Afterwards, the increase of investment cost depresses the speculators from holding houses. LOAN is positively related to GZ, too. Thus, the tight credit policies have a positive effect on curbing the housing bubbles. Moreover, the coefficients of CointEq1 and CointEq2 are significant in the VECM analysis on GZ. And the adjusting speeds are 24% and 0.13% respectively.
‧
Table 5- 4 VECM Analysis on INCOME_BJ, LOAN, INT and BJ
Variables D(BJ)
Note: *, ** and *** denote significance at the 10%, 5% and 1% level respectively.
Table 5- 5 VECM Analysis on INCOME_SH, LOAN, INT and SH
Variables D(SH)
Note: *, ** and *** denote significance at the 10%, 5% and 1% level respectively.
‧
Table 5- 6 VECM Analysis on INCOME_GZ, LOAN, INT and GZ
Variables D(GZ)
D(INCOME_GZ(-1)) 0.001204 ** 2.50782
D(INCOME_GZ(-2)) 0.000263 0.71503
Note: *, ** and *** denote significance at the 10%, 5% and 1% level respectively.
As has been shown, the three factors above contribute to the housing bubbles. It is noteworthy that the low interest rates and high total loans encourage many speculators to invest real estate in Beijing, Shanghai and Guangzhou. Besides, the economic disparities between different cities also lead to the present situation of real estate market in China. numbers. From 2007 to 2012, the population in Beijing had increased by 810,000. In Shanghai and Guangzhou, the population had both expanded by 480,000. The employed population in Beijing increased from 7,981,000 to 10,744,000, expanding by 34.61%. In Shanghai, the employed people rose from 6,578,000 to 9,486,000, an increase of 44.21%. The employed population has expanded by 46.12%, which increased from 2,236,900 to 3,268,500. Since these three large cities attract plenty of extraneous people, we can reasonably infer that there are huge house demands for
‧
non-investment in Beijing, Shanghai and Guangzhou. In addition, the house purchase restrictions make many potential buyers take a wait-and-see attitude and postpone buying houses. So the house demands for non-investment are accumulated constantly, which also stimulates greater investment demands. But the housing supply which is limited by the urban land cannot meet the vast house demand. In a short time, housing prices would not decline and so do housing bubbles. It is essential to achieve a better distribution of resources and the balanced regional economic developments, rather thanconcentrate resource in the big cities.
Figure 5- 4 The population in Beijing, Shanghai and Guangzhou Source: National Bureau of Statistics of China
Figure 5- 5 The employed population in Beijing, Shanghai and Guangzhou Source: National Bureau of Statistics of China
0.00
‧
considers the error correction term to determine the lead/lag relationship between China’s housing bubbles and the variables. The results of Granger causality tests are shown from Table 5-7 to Table 5-9.In Table 5-7, it shows that LOAN Granger causes BJ while BJ do not Granger cause LOAN. This implies that the total loans of financial institutions lead the housing bubbles in Beijing. Generally speaking, the speculators tend to invest in real estate when they can get mortgages easily, which might lead to housing bubbles.
Since the loan scale is regarded as a significant tool for credit policy, we can infer that the housing bubbles in Beijing are affected by the financial institutions’ credit policy.
However, INCOME_BJ and BJ do not Granger cause each other. Likewise, there is no lead/lag relationship between INT and BJ.
From Table 5-8, we can learn that INT and SH do not Granger cause each other.
And there is also no lead/lag relationship between LOAN and SH. It suggests that the housing bubbles in Shanghai are not controlled by the credit policy. It is possible that the investors in Shanghai’s real estate market have adequate free money. As a result, they are able to purchase houses without bank loans. In addition, SH Granger causes INCOME_SH but INCOME_SH do not Granger causes SH. The development of real estate market has promoted Shanghai’s economy, which can create new jobs and increase people’s income.
As reported in Table 5-9, GZ is Granger caused by INCOME_GZ, INT and LOAN while GZ Granger causes INT and LOAN. The per-capita disposable incomes lead the housing bubbles in Guangzhou, which accords with the normal development of real estate market. As people accumulate more wealth, they tend to invest their money in property. Moreover, INT and GZ have the feedback relationship. The feedback relationship also exists between LOAN and GZ. Similar to the case of Beijing, we can easily learn that the housing bubbles in Guangzhou are under the impact of credit policy.
According to the results of Granger causality tests above, we can find inspiration
‧
for housing policy. In Beijing and Guangzhou, the financial institutions’ credit policy could be the principal mean for addressing the over-heating real estate market.
Because the housing bubbles in Beijing do not concern per-capita incomes, Beijing’s credit policy should be stricter. In Shanghai, since the credit policy has no effect on real estate market, it’s better to focus on establishing some related tax and administrative policies.
Table 5- 7 Granger Causality Test on INCOME_BJ, LOAN, INT and BJ
Dependent Variable Independent Variable
D(BJ) D(INCOME_BJ) D(INT) D(LOAN)
D(BJ) — 4.77433 5.313607 6.780424*
D(INCOME_BJ) 2.333389 — 6.999474* 8.635452**
D(INT) 1.066239 1.34843
—
0.537529D(LOAN) 1.588207 0.799552 3.581872
—
Note: 1. This study conducted VEC Granger Causality/Block Exogeneity Tests.
2. Figures in the table are Chi-square statistic.
3. *, ** and *** denote significance at the 10%, 5% and 1% level respectively.
Table 5- 8 Granger Causality Test on INCOME_SH, LOAN, INT and SH
Dependent Variable Independent Variable
D(SH) D(INCOME_SH) D(INT) D(LOAN)
D(SH) — 1.301459 3.327519 5.457886
D(INCOME_SH) 9.702212** — 0.244825 3.53238
D(INT) 3.846077 0.424296
—
1.226377D(LOAN) 3.126434 3.840389 5.331905
—
Note: 1. This study conducted VEC Granger Causality/Block Exogeneity Tests.
2. Figures in the table are Chi-square statistic.
3. *, ** and *** denote significance at the 10%, 5% and 1% level respectively.
Table 5- 9 Granger Causality Test on INCOME_GZ, LOAN, INT and GZ
Dependent Variable Independent Variable
D(GZ) D(INCOME_GZ) D(INT) D(LOAN)
D(GZ) — 6.425987* 41.68709*** 8.23999**
D(INCOME_GZ) 1.995909 — 4.243666 3.051406
D(INT) 10.42858** 0.54275
—
0.056404D(LOAN) 6.697625* 3.040764 5.531587
—
Note: 1. This study conducted VEC Granger Causality/Block Exogeneity Tests.
2. Figures in the table are Chi-square statistic.
3. *, ** and *** denote significance at the 10%, 5% and 1% level respectively.
‧
In recent twenty years, China’s economy develops rapidly. Many cities attract a large number of capital and people. At the same time, the demand and supply of real estate in these cities increase. It pushes up the housing prices and even forms housing bubbles. But the over-heating real estate investments do harm to the development of China’s economy. In order to control the housing bubbles, we must find out the driving factors. Therefore, this study intends to measure China’s housing bubbles and explore the factors that may contribute to China’s housing bubbles.
First, this study evaluates the housing bubbles in Beijing, Shanghai and Guangzhou from 2007 to 2012 by comparing fundamental values with market prices.
The fundamental value of real estate is calculated by annual rents and WACC. The results show that the housing bubbles in Beijing, Shanghai and Guangzhou have risen rapidly before the financial crisis in 2008. After the subprime mortgage crisis burst out, foreign investment in the China’s real estate market was withdrawn. It led to a tide of selling houses and made the housing bubbles decreased sharply. Due to the excellent investment environment in Beijing, Shanghai and Guangzhou, the real estate
The fundamental value of real estate is calculated by annual rents and WACC. The results show that the housing bubbles in Beijing, Shanghai and Guangzhou have risen rapidly before the financial crisis in 2008. After the subprime mortgage crisis burst out, foreign investment in the China’s real estate market was withdrawn. It led to a tide of selling houses and made the housing bubbles decreased sharply. Due to the excellent investment environment in Beijing, Shanghai and Guangzhou, the real estate