Examination of a housing bubble in the Chinese market after the financial crisis in 2008

(1)

Msc Finance, Real Estate Finance track Master Thesis

Examination of a Housing Bubble

in the Chinese Market after the Financial Crisis in 2008

Author: Huijun Luo Studentnumber:11402962

University of Amsterdam, Amsterdam Business School Supervisor: Martijn Dröes

15th_{August 2017}

Abstract

After the financial crisis in 2008, the housing price in China has been increased to an unprecedented level. Even though the Chinese government implemented some polices in order to cool down the market, the housing price still has been increased for years after the transient decrease. The continued price surge is argued that it cannot be accurately explained by the fundamentals anymore. Therefore, it is important to investigate whether a housing bubble exists in the Chinese market after the financial crisis. This thesis constructs error correction models which include the fundamentals in order to examine the housing price in China. 31 Chinese cities are examined in this thesis, and the examined period is between 2008 and 2015. The empirical results suggest that the Chinese housing market is indeed overheated, no matter the model conditions on land price or not. It is also suggested that the land price is out of equilibrium. In addition, the results are in consistent with the assumption that the increasing disposable income, decreasing interest rate and the expensive land price significantly contribute to the upwards of the housing price. Besides the fundamental forces, we point out that the speculation of investors, lacking of private rental sector, and the relation between government revenue and land sales can contribute to the housing bubble.

(2)

2

Statement of Originality

This document is written by Student Huijun Luo who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

The Chinese real estate market has been developed since three decades ago, which is suggested resulting from the rapid economic development in China. In the past three decades, because of the increasing GDP, the increasing income level for citizens, and the expanding urbanization, housing market is booming in China. In 2008, the housing price surged accompanied with more policies were introduced. Both of the stimulation policies and the tightening policies were implemented during the crisis and in the post-crisis period, which are argued to have significant impacts on the housing market. In 2008, in order to deal with the crisis, credits were expanded by Chinese central bank. In this case, commercial banks were encouraged to lend, meanwhile most of the investments inflowed into real estate market and raised the housing prices up. On the other hand, after two years, the Chinese government released policies in order to cool down the continuing high-up housing prices in 2010, but the housing price has been continued to increase after a transient fall. Therefore, it is interesting to know whether the housing price is still in line after the implement of these cool-down policies and it is important to investigate the possibility of the housing bubble in current housing market. The aim of this thesis is to examine the housing prices in the Chinese cities after the financial crisis. The research question therefore is:

Whether a housing bubble exists in China after the financial crisis in 2008?

Some housing indicators present that the home ownership is becoming expensive in China. The price/ income ratio in the first-tier cities in China even achieved 9.2 at the end of 2015 (Balding, 2016). Another evidence of the heated housing prices is that the outflows of populations in cities with high housing prices. The first-tier cities attract graduates from higher education institutions for many years, because of the higher income, and more career opportunities. However, the graduates in cities such as Beijing and Shanghai even consider to migrate to other cities due to the unaffordability of housing purchases.

As the housing prices surged, it becomes a heat topic, which is discussed by several parties in China. From the household’s perspective, the housing bubble will decrease

(6)

6

their affordability of housing, having the impacts on the life quality. Moreover, the real estate investors would tolerate high risks if there is existing a housing bubble. For the government and the policy makers, real estate industry is a significant industry to stimulate the economic development in China, and the previous experience in the financial crisis suggests that the policy makers and the government should pay attention to the possible existence of the housing bubble.

Since the private housing market in China has been developed less than three decades, and it has been blooming in a shorter period in China, the previous literatures mainly focus on mature private housing markets in developed countries. Therefore, the analysis which focus on the Chinese housing market is still lacking, especially the ones focus on the post-crisis period. Before the financial crisis in 2008, the governmental interventions to cool down the real estate market were rare and the magnitude on the real estate market is small. However, 2008 as a split point for the development of real estate market in China, more governmental interactions were introduced in the housing market. Therefore, the results in the previous literatures which focus on the Chinese housing market would be invalid to explain the market after the crisis.

Answering the research question would add the understanding in depth of the Chinese housing market. Answering the research question therefore would be beneficial for citizens for the purchasing decision. In addition, it also beneficial for the investors who are willing to invest in emerging real estate market. It is also important for policy makers to understand the current real estate market in order to release proper policies, avoiding the possible burst of bubble.

In order to observe the possible existence of bubble, this thesis will use error correction models to explain the housing prices. The approach will present the relation between the fundamentals and the housing price index by an econometric regression using panel data. The determinants involve the disposable income, lending interest rate, population and land prices. In this thesis, 31 medium and large cities will be examined. The selection of these cities is based on the released announcement of Chinese government: the 31 cities are important cities which are divided into first-tier cities, second-tier cities and third-tier cities in China. Most of these cities are the capital cities of each province,

(7)

7

excluding Beijing, Shanghai, Shenzhen, Tianjin, and Chongqing, which are mega cities but do not belong to any provinces. There are several sources for the data collection. The data of GDP, disposable income and population are collected from the published Annual Economic and Social Development Report for each selected city. Land price is

collected fromMinistry of Land and Resources. The interest rate is collected from the

historical data from Bank of China. The housing price index for each city is published by National Bureau of Statistics.

In Section 2, the previous literatures, which are relevant to this topic will be provided. The literature review will be divided into four parts. The first sub part will discuss the classical fundamentals which properly explain the housing price. Next, it will answer how can we define a housing bubble. As mentioned above, the literatures focus on analyzing the Chinese housing bubble after the financial crisis, are still lacking, therefore, it will introduce the existing literatures which focus on the housing bubble in other countries. In the fourth sub section, it will introduce the arguments about the Chinese housing market from previous literatures. In Section 3, the comprehensive introduction of the Chinese housing market will be provided. In this section, it will introduce the development of the Chinese housing market. In addition, the roles of the government, state owned enterprises and banks will be introduced as well. In Section 4, it will introduce the data and provide the descriptive statistics. In Section 5, the methodology which in order to observe the possible existence of bubble will be introduced. In this section, we will run the unit root test and cointegration test for the variables first, and then we will use the error correction models to determine whether the price can be explained by the fundamentals. In section 6, the results of the regressions will be presented and be interpreted. In Section 7, it will link the results to the reality and discuss about the possible elements contributed to the overvaluation of the housing prices. In Section 8, limitation of this thesis and the further researches will be discussed, and the main findings will be concluded in Section 9.

2. Literature Review

(8)

8

are considered to promote the housing market, will be given in the first sub section. Definition of a housing price bubble will be discussed next. In the third sub section, literatures relative to the housing bubble in other markets will be discussed. Finally, previous literatures regarding to the Chinese housing market will be discussed.

2.1 The Fundamentals Behind the Housing Prices

According to Geltner, the fundamental drivers to stimulate the housing prices in emerging market usually are strong economic growth, increasing population, increasing income and urbanization. Urbanization can stimulate the housing demand, because more properties are necessary to house migrants and workers. In addition, the citizens who having higher income, will desire more space in the form of better housing. (Geltner, 2007). Therefore, many of those who hold this theory believe that the heated housing market in China is in line with the development, which due to the strong economic development in past decades (World Bank, 2010).

2.2 Definition of a Housing Bubble

A test for the existence of a bubble in real estate market is a tough work. In real estate market, a frequent method to measure a bubble is to compare the price and the fundamentals. (Hui &Yue, 2006) Fundamentals can be categorized into market fundamentals and fundamental value. For the fundamental value, there are three determinants to determine it. However, it is difficult to identify the intrinsic value of the property which is determined by the three determinants. (Flood &Hodrick, 1990) Therefore, the concept of property’s fundamental value is not suitable for empirical studies. In this case, market fundamentals can be used in empirical research in order to test the existence of a bubble (Hui &Yue, 2006). It is to say, if the housing price cannot be properly explained by the exogeneous macroeconomic fundamentals, there exists a housing price bubble (Hui &Yue,2006). In addition, we also can define that a bubble is the increase of price which is not proportional to the raising real demand. House owners purchase houses as investments with the expectations of further rises of housing price— speculators do not interest in the usage of the house but keen on the profits earned from transactions. This results in the rise of price excesses the rational and optimistic anticipation, leading to a sharply and dramatically decrease of the price afterwards, which is called the bubble burst. (Kindleberger,1987)

(9)

9

2.3 Housing Bubbles in the Previous Global Market

Housing bubble is not rare, many literatures have been analyzed the housing bubble in many mature real estate markets such as America, Hong Kong, and Japan. Therefore, learning about the reasons for the bubble will be helpful to understand the Chinese market.

America is considered to be one of the most important and the largest real estate markets in the world, which has been faced the housing price surges several times. In recent decades, there are three important periods for the upwards of the housing prices, before the ultimate housing crisis in 2008 (Levitin & Wachter, 2012). It is believed that the housing price in America began to appreciate from 1997. However, the housing price did not excess the historical peak level at that time. Within the period of 1997-2000, the increase of the housing price was considered to be regular, as the growth rates of housing prices and rental prices remained basically identical. However, from 2000 onwards, the housing prices increased much faster than in the previous periods, and it was considered to be the result of historical low interest rates (Levitin & Wachter, 2012). After 2003, the housing prices continued increasing, leading to a signal of bubble in early 2004. The annual growth rate of housing price was dramatically over 12%, which excessed the highest historical level at that time (Levitin & Wachter, 2012). At that time, the fundamentals cannot explain the increase of the housing prices anymore, as the implemented increasing interest rate would have been reduced the attractiveness of ownership. Meanwhile, mortgage originations increased significantly during this period with massive issuance of subprime mortgages (Levitin & Wachter, 2012). As a result, the housing price collapsed in America after the price surge, and leading to the global crisis in 2008.

Hong Kong, is a densely populated city in Asia, with only a small portion of suitable land for property constructions. Therefore, housing supply is in shortage due to the land scarcity. In this case, supply of housing is considered to be profitable in Hong Kong, making the housing market blooms in several decades. In the recent decades, Hong Kong suffered from several housing bubbles. In 1991, the housing prices increased by 40% (Chan Lee and Woo, 2000). In 1997, more than 45% bank’s loan tied to real estate market, and more than $500 billion were mortgage loan (Hong Kong Government,

(10)

10

1998a). However, the housing bubble burst in 1998, the housing price dramatically declined 50% in one year, making the housing market collapsed in a short time. Previous literatures explained that the misleading prices were triggered by the extensive interactions of government (Chan, 2000; Yiu, 2013).

Because land auctions significantly contributed to the governmental revenue in Hong Kong, the Hong Kong government did not seriously restrain the unregular increase of housing price. In this case, it sent a signal to the public that the housing price would continue increasing without the possibility to decrease (Chan, 2000). This allows the speculators to stimulate the housing price furthermore.

Japan, experienced the strong economic development during 1980s, so did the real estate market. During that period, urbanization prevailed in Japan, cities aimed to expand the urban areas and attracted more skilled labors. Due to the expansions of cities, the real estate investments became highly profitable, and much capital followed into real estate market, leading to the bloom at that period. Behind the bloom, several driving forces are considered to high up the prices and trigger the ultimate housing bubble. First, the increasing land price is the crucial element contributing to the upward of prices (Oizumi,1994 and Shimizu 2007). Since the mid-1980s, land price began to rise in the central Tokyo and then spread to the residential areas outskirt. Later on, this trend spread to other big cities in Japan and finally land prices rose in the urban area throughout the whole country. The rising land price therefore added to the costs for the constructors and it stimulated the price for prospective buyers. Second, the expansions of the property financing also contributed to the oversupply of housing. During 1980s, Japanese yen appreciated against US dollar and the Bank of Japan largely increased the money supply. Meanwhile, the banks in Japan continued lower the interest rates five times, making the interest rate reached the historical lowest level. These polices made the enterprises remarkably easily access to loans. On the other hand, these loans were used for the land speculations by the enterprises (Oizumi,1994). In addition, the Japanese urban planning system even made the problem worse. Japan issued City Planning Act in 1968, but the zoning plan and the regulation were not strict. The data showed that around 40% shops did not locate in retail areas while locate in the residential areas. Office buildings also confound with other types of buildings. Under

(11)

11

the unstrict zoning plan, the price of each piece of land cannot fluctuate separately according to the profitability of each land-use type (Oizumi, 1994). Therefore, the construction enterprises acquit the land without concerning the land-use designation. As a result, the price of residential used land rose independently of land-use. The housing bubble finally burst when the Bank of Japan applied the tight monetary policy in the end of 1989. As interest rate continued shifting upward, the land speculation collapsed and finally resulted in the market depression (Oizumi, 1994).

According to the previous bubbles occurred in the American, Hong Kong and Japanese markets, we can find that there are some signals for the housing bubble and some common forces contributed to it. Before the bubble, the rise of housing price should be explained by the fundamentals initially. Larger population, and city expansion allow the demand and the supply for housing increase. However, as the market continues heating, speculations prevail in the market which are usually supported by easing financial conditions. According to the analysis for American, Hong Kong and Japanese markets above, the central bank plays an important role, as the financial policies can significantly affect all of these markets. In addition, governmental interactions are common in the real estate market. The interactions such as land control, zoning plan and sales of land have indirect impacts on the increasing housing price. There are also some signals for the housing bubble. For example, housing price rises dramatically compared to the historical level in a short period. Meanwhile, the implemented financial policies cannot soft the price surge, and the economic and demographic fundamentals cannot explain the increase anymore.

2.4 The Chinese Housing Market

There are many different arguments regarding to the Chinese housing market. Some conclude that the housing price is in line with the Chinese strong economic development (World Bank, 2010). While according to Hui’s opinion, there is a bubble in some segments but it is not nationwide (Hui &Yue, 2006). The ones suggest that there is a housing bubble provide the following reasons.

(12)

12

stimulus packages and dramatically expanded the credits. Therefore, it encouraged state-owned banks to lend, and centrally-controlled state-owned enterprises to invest. In addition, these major investments flowed into real estate market, leading to highly leveraged purchases of real estate (Deng & Morck, 2011). Another explanation for the overheated price is the savings glut shocks, which alone generate the wrong housing price-to-rent ratios (Gete, 2014). In addition, some suggested that a house is an important investment method for the Chinese investors, because of the relative lacking investment opportunities in China (Guo, F., & Huang, Y. S. ,2010). One of them using 35 major cities as the sample to investigate the relationship between housing price and the fundamental elements, during 1998-2009 and found bubble is huge in the cities in the southeastern coastal areas by 2009 (Dreger & Zhang, 2009). However, the examined period in that literature only covers 1 year in the post-crisis period. It therefore did not take the impacts of the cool-down policies into account in Dreger’s literature. Therefore, it still lacks papers to examine the existence of housing bubble in the current Chinese housing market, by taking the impacts of the credit expansion and the tightening housing price regulations into account, after the financial crisis in 2008.

3. Characteristics and the Development in Chinese Housing Market

3.1 Developments of Chinese housing market

Before the reforming of the Chinese real estate market in 1998, the construction of properties was not market-driven, and the Chinese private housing market did not exist before 1998. Before the market reforming, public houses, which are owned by the government, dominated the housing market. Over 80% houses were state-owned in 1990s (Wang, 2012), but the ratio gradually decreased with the private housing market developing, reaching lower than 20% in recent years (Rothman, 2011). Even though the development of the private housing market is in a short period, the Chinese real estate industry has been become one of the most significant industries in China in the recent decade. The investments completed by the enterprises for the residential use real estate development was 208.156 billion in 1998, however, it was 6459.524 billion in 2015, which is 31 times higher than the investments in 1998. The prosperity of real estate

(13)

13

market also can be reflected from other indicators. The number of enterprises was 93426 units in 2015, which is appropriate 4 times compared to that in 1998. Average selling price of residential buildings was 1857yuan/sqm in 1998, and 6473 yuan/sqm in 2015 (China Statistical Yearbook, 2016). These indicators reflect that real estate market, specifically the private housing market in China has been experienced the rapid growth after the market liberalization in 1998. Meanwhile, the surge of the housing price accompanies with the market development, leading to the argument of affordability of housing in different cities.

Cities in China are specified into different tiers according to their size and economic development. According to the division of National Bureau of Statistics, the 31 cities examined in this paper can be divided into three groups. The division is shown as following:

 First-Tier: Beijing, Shanghai, Shenzhen, Guangzhou

 Second-Tier: Tianjin, Chongqing, Shijiazhuang, Taiyuan, Shenyang, Changchun, Haerbin, Nanjing, Hangzhou, Hefei, Fuzhou, Nanchang, Jinan, Zhengzhou, Wuhan, Changsha, Nanning, Chengdu, Kunming, Xian

 Third-Tier: Huhehaote, Haikou, Guiyang, Lanzhou, Xining, Yinchuan, Urumqi Based on the growth rates of housing prices in the 31 sample cities, it can capture the overall change of housing prices in China after the financial crisis. Figure 1 shows the growth rates of housing price index for the sample cities, which excluding Shenzhen, Hangzhou and Haikou, between 2008 and 2015. The reason to exclude the three cities is because they are the outliers having the much higher fluctuations of housing price indexes compared to that in other cities. In this case, Figure 2 shows the growth rates of housing price indexes for Shenzhen, Hangzhou and Haikou solo. In addition, Figure 3 presents the growth rates of housing price index for China as a whole. The growth rates are constructed as the housing price index in year t minus the price index in year t-1 divided by the housing price index in year t-1. It can be seen that the housing prices increased significantly after 2008, while converged between 2010 and 2011 suddenly,

(14)

14

due to the implemental policies.1_{However, the growth rates fluctuate a lot afterwards.} The growth rates in most cities are between 5 and 20 percentage in 2013 and 2015. However, we can see that the growth rates in Shenzhen, Hangzhou and Haikou fluctuated more than in other cities. The housing price in Shenzhen was flat before 2010, and it slightly increased in 2011, finally dramatically increased 47% from 2014 to 2015, which is stunning higher than in other cities. By contrast, the housing price indexes in Hangzhou and Haikou show dramatic increasing trends between 2008 and 2010. However, after the short live of the price surge in Hangzhou and Haikou, the growth rates gradually decreased and reached the lowest points in 2014.

Figure1: Housing price growth rates for 28 cities Data Source: NBSC

When taking the whole 31 cities into account, we can see that the variance of average growth rates for China is much smaller. The average growth rate peaked in 2013 which is around 11%, and the lowest average growth rate is round -4% in 2014. It is also interesting to see that, after the financial crisis, the average growth rate in 2009 is positive, which is opposite to the global market. It is suggested that the Chinese housing market did not significantly correlate to the global market and did not suffer from the external shocks that much. The positive increase of housing price in that period, is however, due to the financial stimulation package implemented by the Chinese

1_{The main implemental housing policies after the financial crisis can be seen at Appendix Table A1} -10%

0% 10% 20%

2008 2009 2010 2011 2012 2013 2014 2015

Housing Price growth rates in sample cities excluding Shenzhen, Hangzhou and Haikou between 2008-2015

Beijing Shanghai Chongqing Tianjin

Shijiazhuang Taiyuan Huhehaote Shenyang

Changchun Haerbin Nanjing Hefei

Fuzhou Nanchang Jinan Zhengzhou

(15)

15

government, leading to much investment inflowed into the housing market. It suggests that the Chinese housing market is relative close and relative sensitive to governmental interventions.

Figure2: Housing price growth rates for Shenzhen, Hangzhou and Haikou

Data Source: NBSC

Figure3: Average growth rates of housing price index in China between

2008 and 2015 Data Source: NBSC

Even though the growth rates differ from city to city, we can see that the general trend for each city is similar. We can observe that there is a sharply convergence of housing growth rates in 2010 and 2014, which are probably policy-driven. The impacts of the policies, which in order to soft the stress of price surge, only existed a short while, and the price continued increasing after a short convergence. In addition, it is observed that a large variance of growth rates exists between first-tier cities and in other cities. The distribution of the highest growth rates can be seen in Table 1 below.

-0.15 -0.05 0.05 0.15 0.25 0.35 0.45 2008 2009 2010 2011 2012 2013 2014 2015

Housing pricegrowth rates in Shenzhen, Hangzhou and Haikou, bentween 2008 and 2015

Shenzhen Hangzhou Haikou

-0.05 0 0.05 0.1

2008 2009 2010 2011 2012 2013 2014 2015

Average growth rates of housing price index in China between 2008 and 2015

(16)

16

Table 1: Highest growth rate of housing price index in first-tier cities, second-tier cities

and third-tier cities

First-tier cities Second-Tier cities Third-tier cities

Highest growth rate ≥15%

4 4 0

Distribution 100% 25% 0%

Highest growth rate between 10% and 15%

0 15 4

Highest growth rate lower than 10%

0 1 3

It can be seen that the growth rates are correlated to the development and the size of cities. The highest growth rates in the first-tier cities are 100% higher than 15%. However, only 25% second-tier cities have a highest growth rate which is more than 15%. In addition, the highest growth rates in third-tier cities totally below 15%. Therefore, there are eight cities have a highest growth rate which is above 15%, and most cities have a highest growth rate which is between 10% and 15%. The average growth rates in first-tier, second-tier and third-tier cities are10%, 4.2% and 2.86% respectively. Therefore, the dramatic increases of housing price in first-tier cities such as Beijing, Shanghai, and Shenzhen cannot fully represent the changes of housing price in the whole housing market. In order to fully represent the changes of housing price, the sample must include cities in different tiers.

3.2 Governmental Interventions

The Chinese real estate market was reformed in 1998, making the private real estate market liberated since three decades ago. Nevertheless, the current real estate programs in China are still crucially influenced by the governmental interventions. To understand the governmental interventions, it is necessary to explain the importance of the Chinese government, the state-owned real estate enterprises and the commercial banks. First of

(17)

17

all, as it is explained in section 3.1, the real estate enterprises sharply increased when the market was liberated, however, many of them are stated-owned. It other words, the government is the owner of many enterprises in the market. In this case, government can interact the market through these state-owned enterprises. Moreover, the government usually influences the market through the adjustments of interest rate and the land sales. It is important to mention, the central government and the local governments play different roles in the market. The Chinese central government interacts the market mainly through policy releasing. On the other hand, the local governments, they have the right to sell the local land to the property enterprises. The revenue created by land sales can be contributed to the local governmental revenue (Wang, 2012). In this case, when the housing market is blooming, the incentives for the central government and the local government differ.

3.2.1 the Central Government and the Local Government

For the central government, several policies will be released to cool down the overheated market. For example, the interest rate will be set higher in order to soft the increasing housing price. By setting a higher interest rate, the cost for the speculators will be higher, therefore, it will decrease the speculations in the market and will decrease the housing price. Besides, the central government has powerful influences on the real estate market. The central government holds the right to implement policies to influence the country as a whole. The central government has the strong incentive to intervene the market when there is sign of overheating market. Wang et al. (2012) discusses the housing policies applied in the neo-liberal housing market and mentions several political interventions by the central government. For example, when the government observed that the housing price surge in 2009, the interest rate for the extra house purchasing was shifted upward in 2010. Meanwhile, the transaction tax for the land purchase also shifted upward in 2010 in order to lower the land purchases by the enterprises. In order to reduce speculations, the government also restrict the mortgage lending from banks, and raised the minimum down-payment ratio several times to a historical high level. Besides, there are also policies implemented by the central government to directly restrict the housing purchases. According to the policy, each citizen is only allowed to purchase one housing unit. It can be found that these policies

(18)

18

aim to reduce the non-necessary housing demand by increasing the costs to hold extra housing for purchasers. Meanwhile, the implemented policies also aim to reduce the supply by restricting the financial conditions for the enterprises. On the contrast, the local government has the incentive to ask for higher revenues from land sales when the market is blooming, because land sales significantly contribute to the local governmental revenues.

3.2.2 the State-Owned Real Estate Enterprises

As the local government ask for high land price, the costs for the enterprises to bid must be transferred into a high housing price in order to compensate for the high biding costs. When the market is blooming, the state-owned enterprises are willing to bid the land with high price, as they are more easily accessible to financing from banks. On the other hand, the local government realizes that the state-owned real estate enterprises can easily obtain loans from banks, making them be willing to sell the land to these enterprises in higher price. With the engagement of state-owned real estate enterprises, the land price is higher and so does the housing price.

When the market is blooming, there are different incentives for the central government and the local governments. It makes a dilemma for the Chinese housing market when market is blooming. In addition, the active engagement of state-owned enterprises even worsens this problem.

3.2.3 the Commercial Banks

Banks as the financial inter-participants, is also significant in the housing market. The commercial banks are the most significant lending parties for real estate enterprises and mortgage appliers. In addition, the commercial banks in China are closely relative to government’s control. Significant commercial banks were owned by the state in previous decades, and the Chinese government still holds a large portion of shares of these banks currently (Lin & Zhang, 2009). In this case, these banks corporate with the central government to influence the housing market. First of all, the banks are encouraged by the government to provide the priority to state-owned enterprises, allowing relative loose requirements for these enterprises to apply loans. The result is, as mentioned the previous section, indirectly leading to the land sales in high price. In

(19)

19

addition, the Chinese commercial banks are more conventional and risk-aversion. According to Rothman’s argument, the sub-prime mortgages are forbidden to individuals, which is required by the government (Rothman, 2011). Thus, the violation of covenants is lower, and the credibility in the Chinese housing market is higher. 4. Data

In this section, the following fundamental factors are introduced: GDP, disposable income, population, interest rate and land price. They are concerned as the fundamental factors to stimulate the housing price in China. In addition, despite the interest rate, the rest factors will be examined by using logs.

31 medium and large cities are included into this data set over a period from 2008 to 2015. The cities are the capital cities of each province and the mega cities that are not belong to any provinces. These cities include the first-tier, the second-tier and the third-tier cities, which can highly represent the Chinese housing market. There are several sources for the data collection. Housing price indexes for each city can be accessed from China National Bureau Statistics website online. Other important fundamental drivers, such as GDP, disposable income and population can be accessed from Statistical Yearbook, and the Economic and Social Development Annual Report which are published by the local official statistical departments annually. Land price is the price per square meter for residence-use land, which is published by the Ministry of Land and Resources annually. The interest rate is the mortgage lending rate based on the historical data from Bank of China.

4.1 Housing Price Index

The housing price index for each specific city is collected from the published housing-price-index reports from 2008 to 2015 by China National Bureau of Statistics. On the published reports, it only takes the housing price index in preceding year as 100 and calculates the changes for next year. In this case, we have to convert the data into the one using the indexes in 2008 as the benchmark.

(20)

20

4.2 GDP

GDP is an important variable to measure income. In this thesis, the annual GDP (in 100 million yuan) will be used as an indicator in order to measure the economic development in China. However, this variable can be replaced by the disposable income which is a similar income measure. In this thesis, only GDP or disposable income will be involved in the final model. The selection between GDP and disposable income is based on the result of cointegration test we will discuss in section 6.1.2.

4.3 Disposable Income

Disposable income is highly relative to people’s living standard. It can measure the amount of available income for people to invest and spend. It is a significant indicator used to evaluate the affordability of housing. In the Statistical Yearbook, there are two categories for the disposable income: in the first category, it summarizes the disposable income for residents in city, and in the second category, it summarizes the disposable income for residents in rural areas. Because in this thesis we will focus on the urban area, the disposable income for residents in city will be used in this thesis. The disposable income is in yuan (CNY).

4.4 Population

Population can be seen as the fundamental element for housing demand. Industrialization leads more housing is necessary for workers. The Economic and Social Development Annual Report provides the urban population at the end of year for each sample city. The population is in 10 thousand for each city.

4.5 Land Price

Land prices can be seen as an important driving force for housing prices in Chinese market, because of the local governmental interventions. However, there is no available information about the land price for a specific city. In this case, the price for residential used land in a national level will be used in this thesis. The land price is in yuan/ sqm.

(21)

21

4.6 Interest Rate

Interest rate is an important fiscal tool used by the central government to intervene with the real estate market. The lending mortgage rate used in this thesis is the interest rate for the long-term mortgage (>5 years).

4.7 Descriptive Statistics of Variables

The descriptive statistics of the housing price index and the all the fundamentals for the 31 cities from 2008 to 2015 are shown in Table 2 below.

Table 2: Descriptive statistics of variables from 2008 to 2015

Variable Minimum Maximum Average

Standard Deviation housing price index (2008=100) 100 179.8 118.3 15.13 GDP (in100 million yuan) 422.2 145002 6978 12305 disposable income (in yuan) 2894 52859 25355 8291 population (in 10 thousand) 105.5 2171 659.3 468.8 interest rate 0.0362 0.0470 0.0433 0.00319 land price (in yuan /m2₎ ₃₄₇₉ ₅₂₁₇ ₄₄₁₅ _560.4

The correlation among the selected variables are shown in Table 3. Generally, we expect that housing price index is positively correlated with GDP, disposable income, population and land price. Meanwhile, we expect that there is a negative correlation between housing price index and interest rate. The correlations among the selected variables confirm what we assume. In addition, we can see that the correlation between housing price index and GDP, and the correlation between housing price index and population is much lower. On the contrast, housing price index is strongly correlated

(22)

22

with disposable income, interest rate and land price.

Summarizations of GDP, disposable income and population in the dataset for each city can be seen from Table A2 in the appendix. It can be seen that the growth rates of GDP and disposable income are high for all the cities. For most cities, the growth rates of GDP and disposable income are higher than 10% in each year. The average growth rate of GDP is 12% per year, and the average growth rate of disposable income is 11% per year in this dataset. At the same time, more and more people migrate to most of these cities in our dataset. Only a slight outflow of population in Changchun.

Table 3: Correlation among variables

Housing Price Index GDP Disposable Income Population Interest rate Land Price Housing Price Index 1.000 GDP 0.0905 1.000 Disposable Income 0.5575 0.4011 1.000 Population 0.1461 0.4077 0.4599 1.000 Interest rate -0.1239 -0.0030 -0.2522 -0.0220 1.000 Land Price 0.6819 0.0726 0.6561 0.0760 -0.2667 1.000

Figure 4: Comparison among growth rates of housing price index,

GDP and disposable income Data Source: NBSC

(23)

23

In addition, Figure 4 above compares the growth rates of GDP, disposable income and the housing price index. Taking 2008 as benchmark, we can see that the growth rate of housing price index increased faster than the growth rate of disposable income before 2012, but it became lower after 2012. It may imply that the implement of tightening policies works to soft the increasing prices

. Compared to the growth rate of GDP, the growth rate of housing price is even lower.

5. Methodology

5.1 Drawbacks of OLS

The ordinary least squares (OLS) regression has some issues which make it is not suitable for the bubble examination. In order to run the OLS regression, variables which are in level should be stationary. However, many variables using the panel data usually have a unit root and are non-stationary. If the variables are non-stationary in level, the requirements for running OLS regressions cannot be fulfilled. Due to the non-stationary problem, and the possibility of spurious regressions, it is wise to use a dynamic model to avoid the shortcomings in the OLS regression.

5.2 Error Correction Model

Error correction model is a model which can capture the dynamic adjustments of the independent variables and is widely used in economics. Many of the economists are interested in this model because the error correction mechanism is appropriate to analyze the short-run dynamic adjustments (Salmon, 1982). The key point of error correction model is that the degree of disequilibrium in a period has a correlation with that in other periods (Engel & Granger, 1987). In the error correction model, the included variables can be non-stationary at level, but the variables are integrated at the same order and show a cointegrating relationship (Stock & Waston, 2003). If the variables are cointegrating, it means that a long run relationship exists between the dependent variable and the independent variables. Even this long run relationship can deviate in the short run, the error correction model can estimate whether the dependent

(24)

24

variable is out of equilibrium and how fast will it return to the equilibrium level. In this case, error correction model is one of the most common methods to examine the relationship between housing price and the fundamentals.

5.2.1 Unit Root Test and Cointegration Test

Cointegration test is a method for detecting cointegration relation among non-stationary variables. Before constructing the error correction model, we should run the cointegration test for the variables first. If an independent variable is not stationary in level, it can still have a stationary residual in the case that the long run relationship exists between the independent variable and the dependent variable. If it is the case, we can say the variables are cointegrated. In this thesis, we will use the Engle-Granger approach, which is a residual based approach that will be applied for the cointegration test. The cointegration test can be run by the following steps:

 First, we should check whether the examined variables have a unit root in level but are stationary after the first differencing.

 Second, run the regression for the dependent variable and the independent variables, which are non-stationary at level, in order to check the unit root for the residual. If there is no unit root, it suggests that the residual is stationary and the involved variables are cointegrated

For the first step, this thesis will apply the unit root test developed by Hardi and Levin- Lin-Chu (LLC). According to Hardi’ s null hypothesis, all the panels are stationary. If the null hypothesis can be rejected, it means that at least one panel is non-stationary (Hardi, 2000). On the other hand, the null hypothesis of LLC test is the opposite. The null hypothesis of LLC test is that panels contain unit roots. If the null hypothesis can be rejected, meaning that panels are stationary (Levin, 2002). In case variables are non-stationary, they will be checked the stationarity after first differencing.

After the unit root test, the variables which are non-stationary at level but are stationary after first differencing will be included in the regression to check the cointegrating relation. For the last step, Hardi LM test and LLC test will be run again to check whether

(25)

25

the residuals are stationary.

5.2.2 Error Correction Model Construction

After the cointegration test, the error correction model can be constructed for the examination of bubble. In the error correction model, all the cointegrating variables will be included in the model. It is assumed that the error correction term should be zero if the dependent variable is in equilibrium. If the dependent variable is disequilibrium, the error correction term should be non-zero, and should capture the degree of the disequilibrium. Because the independent variables and the dependent variable are cointegrated in the error correction model, the deviation from equilibrium will be corrected over time until the equilibrium is reached. In this thesis, based on the result of cointegration test which is shown in section 6.1.22_{, we assume the long-run} relationship between the housing price index and the fundamentals as following:

ln (housing price indexit ) = β0 +β1 ln (disposable incomeit) + β2 ln (populationit)

+ β3 interest rateit + β4 ln (land priceit) + εit (1)

In addition, as mentioned in section 3.2, the bubble could be resulted from the increases of land price, thus it is interesting to see the result when the model does not condition on land price. In this case, we also run the regression which excludes land price and the model is shown in equation (2). Furthermore, in order to have more insights of land price, we also run a regression which uses land price as the dependent variable, and it is shown in equation (3).

ln (housing price indexit ) = β5 + β6 ln (disposable incomeit) + β7 ln (populationit)

+ β8 interest rateit + εit (2)

ln (land priceit ) = β9 +β10 ln (disposable incomeit) + β11 ln (populationit)

+ β12 interest rateit + εit (3)

2

As GDP and disposable income are the similar variables to measure income, only the one which has a better cointegrating relation with the housing price index will be included finally. Based on the result of cointegration test shown in section 6.1.2, only disposable income is included. The detailed discussion of the cointegration test can be seen in section 6.1.2.

(26)

26

The basic form of the error correction model is shown in equation (4): the differenced dependent variable at time t, which has a relation with constant λ0, the differenced

independent variable△ X at time t-1, and the lagged error correction term. The

coefficient λ1 indicates the short run effect of the changes of the independent variable in previous period, on the changes of dependent variable in current period. The

coefficient α of the lagged error correction term captures the long run effect, meaning

the speed that the dependent variable returns to equilibrium.

△Yt= λ0 + λ1△Xt-1 +α(εt-1) +μt (4)

The error correction model captures both of the short run effect due to the changes of the independent variable and the long run effect due to the error correction term. Apply the basic error correction model to our case, we can obtain a specific error correction model shown in equation (5):

△ln(housing price indexit)= λ0 +λ1 △ln(disposable incomei,t-1)+ λ2△ln(populationi,t-1) +λ3△interest ratei,t-1+ λ4△ln(land pricei,t-1) +α1 error correction term(εi, t-1)+ μit (5)

Again, in order to observe whether the possible housing bubble comes from the increases of land price, we will run further error correction models.

First, we exclude land price from the model, and the error correction model without land price is presented in equation (6):

△ln(housing price indexit)= λ5 +λ6 △ln(disposable incomei,t-1)+ λ7△ln(populationi,t-1) +λ8 △interest ratei,t-1+α2 error correction term(εi, t-1)+ ŋit

(6) Secondly, we also run the error correction model by using land price as the dependent

variable. The error correction model by using land price as dependent variable is shown in equation (7):

(27)

27

△ln(land priceit)= λ9 +λ10△ln(disposable incomei,t-1)+ λ11△ln(populationi,t-1)

+λ12△interest ratei,t-1+α3 error correction term(εi, t-1)+ δit (7)

5.2.3 Detecting the Bubble by the Error Correction Term

The long run effect can be observed from the lagged error correction term. This term indirectly indicates the speed that the dependent variable returns to the long run equilibrium. If the coefficient of this term is negative, meaning that the dependent variable is going to return to the equilibrium. In order to observe the speed, the formula will be used: 1/α. The result of the formula would suggest how many years will the housing price index and land price go back the long run equilibrium.

6. Results

6.1 Unit Root Test and Cointegration Test

6.1.1 Unit Root Test

Whenever we estimate the regressions with panel data, we should run the unit root test and the cointegration test in order to avoid the spurious regression. In the first step, we run a unit root test to check whether the selected variables are non-stationary in level and stationary after differencing. Table 4 shows the results of unit root tests based on Hardi LM test and Levin-Lin-Chu (LLC) test.

We can see that the null hypothesis of Hardi LM test can be rejected at 1% significant level for all the variables. It means that the examined variables have a unit root at level, and they are non-stationary. However, after the first differencing, the null hypothesis cannot be rejected, meaning that all the panels are stationary for all the differenced variables. In this case, we can conclude that the variables are integrated at order one I (1) based on the result of Hardi LM test. There is the same result based on the LLC test. The null hypothesis of LLC test is that the variable is not stationary. When we test the variables in level, the null hypothesis cannot be rejected, therefore, the variables are not

(28)

28

stationary at level. After the first differencing, the statistics are all significant at 1% level. In this case, we can conclude that the variables are non-stationary at level and all of them are integrated at order one I (1). Therefore, the further regression to examine the cointegrating relation will be run.

Table 4: Results of unit root test, using Hardi LM test and LLC test

Hardi LM test LLC test

Level First difference Level First Difference

ln house price index Statistics: 6.1029*** Statistics: -2.1329 Statistics: 2.0331 Statistics: -9.3747*** ln GDP Statistics: 6.2260*** Statistics: -2.1423 Statistics: 1.0050 Statistics: -17.7181*** ln disposable income Statistics: 7.7983*** Statistics: 0.4158 Statistics: 1.7386 Statistics: -31.1919***

ln land price Statistics:

9.8720*** Statistics: -2.2175 Statistics: -0.1900 Statistics: -19.5564*** ln population Statistics: 5.1481*** Statistics: -1.2932 Statistics: 1.7386 Statistics: -80.6339***

interest rate Statistics:

9.0254*** Statistics: -1.3686 Statistics: 1.5461 Statistics: -13.6714*** *significant at 10% level, **significant at 5%, and *** significant at 1% level

6.1.2 Cointegration Test

Because GDP and disposable income are the similar measure, we will only include the one having better cointegrating relation with the house price index. In addition, we also examine the cointegrating relation between housing price index and the independent variables when land price is excluded. The cointegrating relation between land price and other independent variables are also examined. In this section, we also use Hardi LM test and LLC test to examine whether the residual has a unit root. The results are shown in the Table 5 below.

(29)

29

Table 5: Results of cointegration test, using Hardi LM test and LLC test

Dependent variable Independent variables Hardi LM LLC ln house price index ln GDP, ln land price, lnpopulation, interest rate Statistics: 1.4546* p-value: 0.0729 Statistics: -4.7361*** p-value: 0.0000 ln house price index ln disposable income, ln land price, lnpopulation, interest rate Statistics: 0.6672 p-value: 0.2523 Statistics: -4.7607*** p-value: 0.0000 ln housing price index ln disposable income, lnpopulation, interest rate Statistics: 1.1959 p-value: 0.1159 Statistics: -7.2302*** p-value: 0.0000

ln land price ln disposable

income, lnpopulation, interest rate Statistics: 1.2063 p-value: 0.1139 Statistics: -2.7049*** p-value: 0.0034 *significant at 10% level, **significant at 5%, and *** significant at 1% level

Firstly, we should decide either GDP or disposable income will be included in the further error correction model. The results of cointegration tests can be found in the second row and the third row. We can see that the null hypothesis of Hardi LM test cannot be rejected at 5% and 1% for each regression, which means that the residuals are stationary and thus no unit root. Based on the LLC test, the statistics is significant at 1% level for each regression. Thus, based on the results of LLC test, the residuals are also stationary, which as same as the result of Hardi LM test. Therefore, there is a

(30)

30

cointegrating relation between the housing price index and the included independent variables. However, based on the Hardi LM test, we can see that the regression, the one includes GDP, the statistics value is significant at 10% level. It means that some panels are non-stationary when significant level is 10%. In this case, it means the disposable income having a better cointegrating relation with the housing price index. In this case, we will exclude GDP, and only include disposable income in the further error correction model.

Secondly, we should observe the cointegrating relation when land price is excluded and the cointegrating relation when land price is the dependent variable. In the fourth row, we can see the result of cointegration test when land price is excluded. In the fifth row, it is the result of cointegration test, using land price as the dependent variable. For the two models, the null hypothesis of Hardi LM tests cannot be rejected, on the other hand, the null hypothesis of LLC test can be rejected in the two models. It means that there is a cointegrating relation among land price and the included independent variables, and there is a cointegrating relation between housing price index and the included variables when land price is excluded.

6.2 Error Correction Model

6.2.1 Error Correction Model: When Land Price is Included

After the cointegration test, the error correction model is run in order to observe the possible existence of the housing bubble. In this section, we will show the results of the model which includes land price. The results of the error correction model and the long run relationship are shown in Table 6 below.

In the short run, the coefficients of disposable income, land price, population and interest rate indicate the short run effect (one year) on the housing price index. According to the results shown in Table 6, we can see that interest rate negatively influences the housing price index, which is consistent with our assumption. On the other hand, the result suggests that disposable income and land price stimulate the housing price positively. Even though we can see that the coefficient for population is

(31)

31

positive, it is not significant at all. In addition, the error correction term is significant at 5% level. The result thus suggests that disposable income, land price, interest rate and the error correction term have significant effects on the housing price index. The coefficient of disposable income suggests that an 1% increase in the disposable income leads the housing price index increasing 0.28%. On the contrast, an 100% increase of the interest rate will lead to a 1100 % decrease of the housing price index. Meanwhile, the coefficient of land price is also significant, which suggests that an 1% increase of land price will lead to an 1% increase of housing price index in one year. The coefficient of εt-1 is significant at 5% level, and it is negatively relative to the housing price index. The negative relation means that the disequilibrium of housing price index is going to gradually decline in the future. According to the formula 1/α, we can calculate the speed of the housing price correction. In this case, it is 1/.098=10.2, which means it will take approximate 10 years to correct the existing disequilibrium in the housing price index.

Table 6: Results of error correction model using housing price index as

dependent variable

Short run Long run

△ln disposable incomet-1 .2785219** (.1144252) ln disposable incomet .0902288*** (.0279474) △ln land pricet-1 1.004582*** (.2142165) ln land pricet .530015*** （.062975） △ln populationt-1 .0294338 （.0294338） ln populationt -.0151263 （.0090486） △interest ratet-1 -11.82168*** （1.568084） interest ratet -1.497572 （1.065055） εｔ－１ -.098681** （.0456703） constant -.0515165 (.015749) constant -.5804428 (.3934483) R-square: 0.5293 R-square: 0.2727

*significant at 10% level, **significant at 5%, and *** significant at 1% level

In the long run, we can see that disposable income and land price still have the impacts on the housing price index, however, interest rate and population do not show significant effects on the housing price index. In the long run, an 1% increase of

(32)

32

disposable income will increase the housing price index by 0.09%, and an 1% increase of land price will increase the housing price index by 0.53% percentage. In this case, we can see that the effects of disposable income and land price on housing price index are smaller in long run compared to the effects in short run. Based on the results shown in Table 6, we can suggest that the growth of population in cities do not significantly affect the housing price either in short run or in long run. The increasing income for citizens, on the other hand, can drive the housing price high. Interest rate as a tool of governmental intervention only shows the significant effect in short run, but land price can influence the housing price index in short run and long run.

6.2.2 Examination of Land Price

It is a section to show more insights of land price. According to the results shown in Table 6, it is observed that land price is a significant element leading to the disequilibrium of housing price. Therefore, it leaves a question, whether the bubble of housing price is the result of the land price. In order to observe more insights of land price, two regressions are run. In the first regression, land price is excluded from the regression model, and in the second regression, land price is used as the dependent variable. The results can be found respectively in Table 7 and Table 8 below.

Table 7: Results of the error correction model, using housing price index

as the dependent variable and excluding land price

Short run Long run

△ln disposable incomet-1 .3626822*** (.1163526) ln disposable incomet .2640835*** (.0213486) △ln populationt-1 .0414524 (.0804021) ln populationt -.0416696 (.0096187) △interest ratet-1 -6.117446 *** (1.036443) interest ratet -1.166998 (1.207067) εｔ－１ -.1508196*** (.0425776) constant -.0037407 (.0126781) constant 2.293058 *** (.2217542) R-square: 0.2208 R-square: 0.1921

(33)

33

When land price is excluded, the R-squares decrease. In the short run, the results are the same to the regression which includes land price. It can be seen that disposable income, interest rate and the error correction term still have significant effects on housing price index. On the other hand, population does not significantly affect the housing price index either in short run or in long run. An 1% increase of disposable income can lead to a 0.36% increase of housing price index, but an 100% increase of interest rate can decline the housing price index by 610%. What is remarkable is the error correction term, compared to the model including land price, the value of the error correction term is still significant but is smaller, which is -0.15. It means that, if we withdraw land price from the model, the Chinese housing market requires 6.7 years rather than 10 years to return to the equilibrium. It means that the increases of land price indeed results in larger disequilibrium of housing price index.

Table 8: Results of error correction model using ln land price as dependent variable

Short run Long run

△ln disposable incomet-1 -.0319566 (.0586341) ln disposable incomet .3280185 *** (.019136) △ln populationt-1 .065704 (.0406268) ln populationt .0500803*** (.0086218) △interest ratet-1 -1.24204** (.5215901) interest ratet -.6237072 (1.081965) εｔ－１ - .1582442*** (.0275802) constant .0575794*** (.0063677) constant 5.421547*** (.1987713) R-square: 0.1995 R-square: 0.5644

*significant at 10% level, **significant at 5%, and *** significant at 1% level

Furthermore, land price is examined as the dependent variable. Using land price as the dependent variable, in the short run, disposable income cannot significantly explain the changes of land price, which is opposite to the result when housing price index is the dependent variable. Population does not show significant effects on land price neither. On the opposite, the coefficients of interest rate and the error correction term are significant. It suggests that an 100% increase of interest rate decreases the land price by 124%. The error correction term is negative which suggests that land price is out of

(34)

34

equilibrium, and it requires about 6 years to return to the equilibrium. In the long run, disposable income, population can promote the land price, however, interest rate dose not have significant effect on land price. The results suggest that an 1% increase of disposable income leads to a 0.33% increase of land price, and an 1% increase of population can drive land price increasing 0.05%.

6.3 The Error Correction Term

It can be seen from the results of the error correction model that the market will take a

long time returning to the equilibrium. This means thatthere exists a disequilibrium for

the Chinese housing market, and the housing market can only slowly correct the disequilibrium. To obtain more insights, the analysis of the error correction terms for specific cities will be discussed in following. The results will be discussed based on their tiers.

The Figures 5-8 show the lagged error correction term εｔ－１ for the first-tier cities: Beijing, Shanghai, Shenzhen, and Guangzhou. The figures of error correction terms for all the 31 cities can be seen in the appendix. It can be seen that the error correction terms for the four first-tier cities are non-zero, which implies that the housing markets in these cities were not in equilibriums. For Beijing and Shanghai, the housing price indexes were gradually far beyond the equilibrium level. For Shenzhen and Guangzhou, the housing price indexes were approaching to the equilibrium level around 2014, and disequilibrium sizes were smaller compared to that in Beijing and Shanghai. This implies that the housing markets in Shenzhen and Guangzhou were close to the steady level, which is consistent with the result of applying the new financial policies in 2014.

(35)

35

Figure 6: Error correction term for Shanghai

Figure 7: Error correction term for Shenzhen

Figure 8: Error correction term for Guangzhou

Figure 9: Error correction term for all the cities

(36)

36

For the second-tier cities, we can observe that the sizes of disequilibrium were smaller than that in the first-tier cities in general. In addition, most of the disequilibrium were reduced after 2014. This situation was similar in most third-tier cities. In this case, we can conclude that there is a disequilibrium of housing prices in all the selected cities in China, but the degree of the disequilibrium differs city to city. The degree of the housing bubble for a specific city is dependent on the economic development of that city. In general, the Chinese housing market has a long way to the long run steady level. Even though we can see that the housing price are far from equilibrium, the positive signal is that the disequilibrium has been reducing for several second-tier and third-tier cities after 2014.

Figure 9 shows the error correction term for the whole cities by taking the average value of each city in each year. It can be seen that the deviation is smaller than in the first-tier cities by taking all the cities into account. The lagged error correction term is mostly negative between 2011 and 2013. It suggests that even after the implement of cool-down policies, the housing market is still out of long run equilibrium. Nevertheless, the correction of housing market is lagged, and slower than in other liquid asset markets. By the end of 2015, we can see that the error correction term is much closer to zero than in the previous years, which is a good signal for the Chinese housing market.

The error correction term for the whole cities when land price is excluded from the model and the the error correction term when uses land price as dependent variable can be found in appendix. The error correction terms are non-zero in each year. Therefore, even the model does not condition on land price, the result still suggests that the housing market is overheated. The error correction term also confirms that land price is out of equilibrium.

7. Insights of the Empirical Results

According to the results in the error correction models, it can be concluded that the Chinese housing market is indeed out of equilibrium after the financial crisis, meaning that the housing prices have been deviated from the fundamentals. Even though the

Examination of a housing bubble in the Chinese market after the financial crisis in 2008

Contents