Amsterdam Business School
Master Thesis
Name: Fan Zhang (10605606)
Track: MSc Business Economics, Finance & Real Estate Finance track
Time: August 2014
Supervisor: Dr. J.E. (Jeroen) Ligterink
Working Title:
The relationship between property prices and bank loans to real estate development
companies: the Chinese case
2
Table of Contents
Acknowledgements... 3
Abstract... 4
1. Introduction... 5
2. Research background... 6
2.1 Usual Capital sources for real estate development ... 6
2.2 Capital sources for real estate development in China ... 7
3. Literature review
... 8
3.1 Dynamics of housing price ... 8
3.2 Real estate markets and bank stability ... 9
4. Data and stylized facts
... 12
4.1 Data definition and Source ... 12
4.2 Description of the data ... 13
5. Econometric models
... 16
5.1 Augmented Dickey Fuller---Unit root test ... 17
5.2 Construction of VAR model ... 18
6. Empirical results
... 19
6.1 Result of the unit root test ... 19
6.2 Result of VAR model ... 20
6.3 Result of Granger causality test... 21
7. Conclusion
... 22
Appendix: ... 23
Bibliography: ... 25
3
Acknowledgements
I wish to express my deep gratitude to Dr. J.E. (Jeroen) Ligterink for his valuable and constructive suggestions on this paper. Different from my peers, I changed my topic once, which took my professor more time and effort on supervision. His patient guidance, enthusiastic beneficial critiques encouraged me to keep my progress on schedule. My great thanks are also extended to Dr. M.K. (Marc) Francke, who provided me with very valuable suggestions on methodology. Finally, I wish to thank my parents and Tianliang for their support and encouragement throughout my study in the Netherlands.
At last, any error of this paper is the sole responsibility of the author.
4
Abstract
This paper investigates the relationship and interaction between property prices and bank loans to real estate development companies in China. Research on this topic is interesting for the following reasons. First, property prices have been escalating rapidly in China. Second, the main capital resource for real estate development in China is bank loans. In addition, previous scholars have already found unidirectional or bidirectional causality between property prices and bank loans in other countries as well as in China but with divergence. Some scholars including Gerlach (2004) argue that the increasing property prices lead to the expansion of bank loans while others such as Liang (2006) believe that bank loans support the property prices to grow. In order to investigate the potential relationship between property prices and bank loans in China, this paper selects the growth rates of consumer price and GDP as control variables with the sample period covering from 2001Q1 to 2012Q4.As this study shows, the growth rate of property prices is negatively correlated with its lags and the growth rate of bank loans is positively correlated with its lags. However, significant causality is not found between property prices and bank loans which might be contributed to existence of other financial channels for real estate development in China, short sample period and omitted variable problem.
5 1. Introduction
From 2001 to 2012, the property prices of China increase from 2,472 RMB per square meter to 5,836 RMB per square meter. According to the calculation of E-house1, national price-to-income ratio is 7.4 in 2011 and even
higher than 10 in the first-tier cities such as Beijing and Shanghai. Such high house price pose pressure on average Chinese buyers. Several causes may explain the high prices of property in China. For example, Guo and Huang (2009) proposed that “hot money” contributed to the fluctuations of real estate due to its enormous amount and short-term characteristics of investing. On the other hand, according to CSMAR Industry Research database, bank loans granted to listed real estate development companies in China increased from 3.79 billion RMB to 93.2 billion RMB. As Herring and Watcher (1998) pointed out that when property prices increase, real estate development companies will profit more from the development and have incentives to obtain more loans from banks; banks might also authorize incremental lending to real estate development companies as property had more economic value as collateral. Therefore, a bidirectional positive relationship is expected to exist between property prices and bank loans.
According to Hott (2011), as real estate development companies get more loans from the bank, they built more houses, increasing the supply of houses in the market. As the new built houses’ prices should cover the cost to get loans as well as opportunity cost of capital, increasing supply does not necessarily decrease housing price but led to higher development costs (Selden, 1959). The other way around, banks offered both loans to households and real estate development companies, who bought houses and built houses, separately. Banks indeed had an incentive to keep a comparably similar growth rate for these two kinds of loans in order to decrease their risk exposure to default or prepayment due to fluctuation of housing price or interest rate (De Jong and Driessen, 2008). Thus, rising loans to real estate development companies were correlated with increasing household loans, which caused more demand for houses followed by higher housing price.
Although previous scholars have described the causality between property prices and bank loans, few identified the relationship from a quantitative or econometric perspective. Gerlach and Peng (2004) applied vector error correction model to investigate the relationship between residential property prices and bank lending in Hong Kong. It turned out that causality went from property prices to bank credit while contemporaneous correlation between lending and property prices is large. For the Chinese case, Liang and Cao (2006) implemented a high dimensional distributed lag framework to examine the relationship between
6 property prices and bank lending. They found that the causality ran from bank lending to property prices. The data period covered in that study from the first quarter of 1999 to second quarter of 2006.
While the relationship between property prices and bank loans varies according to the area and time period, property prices and bank loans have been escalating during recent twelve years in China. Consequently, their relationship in China might have changed after 2006. This paper, extending the time period to 2001Q1 to 2012Q4, contributes to this field by investigating the long run relationship of property prices with bank loans in China. Furthermore, Liang and Cao (2006) used total bank lending instead of property related lending as they believed that evidence showing other lending to the corporate and household sectors working for property-related investments indeed. However, as previous evidence had no solid material to support, this paper takes use of the loans to property-related listed companies and results turn out to be different from previous studies. For the part of methodology, the paper applies unit root test, co-integration test, and vector autoregressive model (VAR) to test the relationship of property prices and bank loans based on time series data.
Background of usual capital sources for real estate development and the case in China will be reviewed in Section 2. It provides evidence that real estate development in China mainly relies on bank loans. Thus investigating the relationship between property prices and bank loans in China becomes more interesting as it can shed lights on whether property prices would be influenced by the methods of obtaining capital for real estate development. If it were the case, diversification of capital sources for real estate development would be a measure to smooth property prices in China, which would decrease the risk exposure of real estate development companies in China as well as Chinese banks’ risk to fluctuating property prices.
The rest of this paper is organized as follows. In section 3, previous studies on property prices, bank loans to real estate sectors, and their relationships will be discussed. Section 4 presents resources and definition of the variables used for regression in the next section and some facts about the data in sample period. Econometric models and empirical results are provided in Section 5 and 6 followed by conclusion.
2. Research background
2.1 Usual Capital sources for real estate development
Geltner, Miller and Clayton (2013 P.738) outlined the various capital sources for real estate development in their book “Commercial Real Estate Analysis and Investments.” They indicated that financial markets provide various types of capital during the process of real estate development project. Typically, three main types of capital support the overall development based on different phases of the project. Preliminary Phase is normally initiated by entrepreneurial developer’s seed equity, small in absolute magnitude compared to following sources of capital. In order to finance the purchase of land in second phase, the next investment
7 capital is generally external equity. While the magnitude required for external equity is rather great, it always comes from more than one type of source. In other words, funders of external equity are mixed including entrepreneurial developer itself, financial institutions (like pension funds, life-insurance companies, or endowment funds), foreign investors, private equity vehicles, or sometimes REITs. Finally, “construction capital” is typically provided by commercial banks or syndication of such banks, covering all remaining capital requirement. Payback seniority coincides with the magnitude of capital invested, which means that banks are most senior followed by external equity provider and finally the entrepreneurial developer.
2.2 Capital sources for real estate development in China
There are mainly three types of capital sources providing capital for real estate development in China. The capital come from real estate developers’ fund; domestic loans and other capital constituted of advanced payments and payables. According to the official website of National Bureau of Statistics of China, the capital provided by these three sources amounts to about 90% of total capital for real estate development with real estate development companies themselves’ contribution to about 40% of total capital while contribution portions for domestic loans and other capital are both around 20%. Total capital mentioned here indicates the amount of fund put into practical use for development of real estate.
Nevertheless, channels for real estate development companies to raise capital are rather limited and
restricted. Statistics published by China Securities Regulatory Commission (CSRC)2 present that capital raised
by public real estate companies through capital market only contributed 0.77% of the total capital in 2009. In the “China private real estate investment fund annual report in 2011” published by Qingke Research Center, it was stated that private equity fund set since 2011 raised about 5.86 billion dollar in 2011, contributing about 0.43% of the total capital. Only until 2010, insurance funds were allowed to invest in real estate under restricted laws3 and pension funds are still prohibited from investing in real estate in China. Other
investment vehicles like REITs have been in the process of trial implementation and not raised any capital yet.
As a result, real estate development in China could only be funded by real estate developers’ own capital, domestic loans, and other capital. Moreover, the asset-liability for Chinese real estate development companies remains stable around 75% since 2006, which means capital provided by the real estate developers also comes from loan market mainly constituted of banks in China. On the other hand, capital generated from other resources such as advanced payments rely on bank loans as well as property buyers
2 http://www.gov.cn/jrzg/2010-04/24/content_1591123.htm
8 always pay down payment with the rest paid by mortgage loans provided by banks. In all, Chinese banks’ loan to real estate development companies and individuals housing has a proportion of 20% of total loans. Therefore, Chinese real estate development sector relies on the bank loans as other channel to raise capital is limited in China and two main resources of capital except for domestic loans generate capital from the bank system as well.
3. Literature review
As this paper investigates the relationship between property prices and bank loans in China, previous studies on the dynamics of housing price and interaction between banks and real estate markets will be reviewed. At the end of this section, some similar studies to this paper concentrating on property prices and bank loans are present with their results.
3.1 Dynamics of housing price
Almost 90 percent of sales area of real estate in China is housing and nearly 85 percent of sales of real estate is contributed by housing annually as well. Therefore, the main type of property in China is housing. Consequently, understanding of the dynamics of housing price increases the comprehension of property prices. Pasquale and Wheaton (1992) studied the housing market dynamics under the background that owner-occupied housing became one of the largest components of wealth in the U.S. Facing with an older age distribution of demography, some people forecasted negative prospects of household and expected decreasing home price. However, Pasquale and Wheaton (1992) clarified though demographic perspective would have influences on house price appreciation; long-term downturn shock was not expected in housing market. Instead of momentary clearing of the housing market, the market for single-family housing adjusted their price gradually in the process.
From an international empirical perspective but still studying the dynamics of housing price, Englund and Loannides (1997) compared the dynamics of house prices across fifteen OECD countries. Previous researches, however, almost all focused on a specific country or metropolitan housing markets or only investigated the relationship between house price and savings instead of the dynamics of the house price. Surprisingly, it found similarity across different countries for the first-differenced real annual house price at a remarkable degree. By implementing first-order autocorrelation regression, a highly significant coefficient was obtained and evidence showed negative autocorrelation for lags up to five years. It revealed that house price oscillated around a trend across fifteen OECD countries. Except for own lagged values of house price, contemporaneous GDP growth rate and the change rate of real interest rate were significant with strong predictive power of their lags.
9 Other studies specifically investigate the real estate market in China and analyzed factors that affected property prices in China. Shen and Liu (2004) investigated economic fundamentals and house prices of China from 1995 to 2002 based on city-level panel data. It clarified that past and current data about economic fundamentals could partially explain and predict house prices. They also implemented Granger causality test for real estate development investment and GDP in China and showed that GDP strongly influenced the development of real estate industry. Wang and He (2005) found that a strong relationship existed between expected rate of real estate market return and inflation expectation and suggested that prices of property should be added to the residence price index. CUI (2009) used panel data to analyze the prices of property of 31 provinces in China from 1996 to 2005. It showed that the prices of property were significantly influenced by interest rates, inflation rate and income, which determined the property prices in the intermediate and long run.
3.2 Real estate markets and bank stability
Some researchers have been focusing on the relationship between real estate markets and bank stability. Hott (2011) established an economic model to investigate the relationship between lending behavior and real estate price. It suggested that a cyclical relationship between real estate price, default rates, bank profits, and mortgage lending was the cause of problems in the banking sector in countries such as Switzerland, the UK and the US in the 1980s. With rapid rising real estate prices in these countries, housing bubble burst around 1990. Consequently, banks faced with serious problems as loan losses and default rates increased. Generally, the author clarified that a positive bidirectional feedback effect existed between real estate prices and bank loans. The author applied three irrational expectations approaches to set up a benchmark model. The model suggested that banks provided loans to real estate purchase depending on real estate price, which had impacts on the creditworthiness of borrowers. On the other hand, demand for homes affected by the supply of mortgages influenced real estate prices. Consequently, the economic model showed a simultaneous relationship between bank loans and real estate price. Moreover, markets behaved under the effect of both rational and irrational expectations that defined as mood-swings, momentum forecasts, and disaster myopia, leading to the fluctuations of the markets.
Herring and Watcher (1998) illustrated real estate booms often led to banking busts though real estate and banking crisis cycles were not necessarily accompanying each other. Increasing real estate prices always brought optimism to banks when extending loans to real estate. Furthermore, moral hazard was also an important cause in busts as bank shareholders and depositors took the protection from regulators and supervisors for granted and deemed that they had little to lose. Poor information and inadequate appraisal also contributed to the cycles of real estate and banks. Koha and Marianoa (2004) investigated the market optimism and asset bubbles in bank lending and real estate in Asia. It was found that the tendency of
10 undervaluing real estate loans existed before the Asian real estate bubbles, and accelerated the process of financial crisis.
Regarding the characteristics of real estate price, and bank stability, Koetter and Poghosyan (2009) constructed an empirical model to test two contradicting hypothesis of the influences that real estate prices on bank stability by using data on regional real estate prices and individual banks in Germany from 1995 to 2004. One of the hypothesis suggested that high prices of real estate contributed to the stability of banks as it increased both the value of collateral and wealth of borrowers and therefore a reduced number of possible credit defaults, which increases the soundness of the banks. Nevertheless, real estate asset is treated non-standard as it varies from quality and location. Therefore markets for the real estate transactions lacked of transparency and liquidity. Real estate prices were sluggish owning to construction lags as well. Consequently, the prices of real estate had a tendency to deviate from their real value decided by construction costs or supply and demand. As a result, increasing prices of real estate was regarded as deviation from their fundamentals instead of rising demand. Considering these features of the real estate price, adverse selection might appear when banks chose to expand their loan portfolios facing with soaring real estate price. In this case, soaring prices of real estate in fact fostered the instability of banks. Therefore, instead of assessing just price levels or changes of the real estate, this paper paid attention to the deviations from the fundamental value when estimating affects stability of banks. Furthermore, the study was implemented at the bank level as it took use of regulatory information about distress. Deviations of real estate prices were found across regions in Germany and facilitated the distress and instability of banks. However, as the research was conducted in a relatively simple period in which the real estate markets were not regarded as overheated, further investigation based on longer period with different market fundamentals is necessary to confirm the results.
3.3 Interaction of property prices and bank loans
Before identifying the relationship between property prices and bank loans, let us look at housing price and credit supply first. Adelino (2012) showed empirically that easier access to credit increased house price significantly. It used exogenous changes in the conforming loan limit as instrument variable. Goodhart (2008) studied the links between money, credit, house price, and economic activity and found that a significant multidirectional link existed based on a fixed-effects panel vector auto regression.
Glaeser (2008) argued that low funding costs and easy access to credit attributed to the boom in house prices and the subsequent price decline when credit dried up. Herring and Watcher (1998) claimed that rising housing price increased economic value of property used as loan collateral. Therefore, banks would slack credit standards and provided more loans to real estate sector. This suggests that the causality can go both ways.
11 Ng and Chow (2004) examined how credit affected the residential property prices in Singapore. The findings suggested that the lag of bank credit determined the property prices instead of the mortgage rate. Moreover, the study conducted separate analysis for private property prices as well as public property prices. Nevertheless, the response period of property prices to bank credit was relatively short, which last only about three quarters. This phenomenon was consistent with the supply push rather than demand pull dynamic in Singapore. Park and Bahng (2008) also investigated the relationship between house price and the access to bank lending in Kangnam and other four markets. Facing with the escalating housing price, the Korean government put strict limits on the qualification to get mortgage loans. As Kangnam was the targeted market, treated as “hot,” house price in this area had been influenced little while other four markets experienced disruption.
Concentrating on the relationship between residential property prices and bank lending in Hong Kong, Gerlach (2004) suggested that property prices had impacts on bank credits rather than the other way around though the contemporaneous correlation that existed between these two variables was large. In other words, the results claimed that excessive bank lending was not the cause of the boom and bust cycles of the real estate market and the bank system remained fundamentally sound. The study considered both demand and supply when analyzing property prices and more weight on supply due to the restriction on land supply during the period of 1985 to 1994. It showed that fluctuations of the property prices in Hong Kong were mainly caused by shifts in economic sentiment, stemming from factor external to Hong Kong. For the econometric part, the logarithms of lending, property prices and GDP measured in real terms with CPI as deflator for the former two variables were selected. It took use of the standard Augmented Dickey Fuller (ADF) unit root test and co-integrating model, which we will follow in the econometric part.
Different from Gerlach (2004), Liang (2006) found that a unidirectional causality running from bank lending to property prices in China over the period 2001Q1 to 2012Q4. It also claimed that the causality ran interactively through the error correction model. Nevertheless, only weak impacts of bank lending on property prices had been observed. Similar to Gerlach (2004), the selected quarter time series for all variables including real property prices, real GDP, real total bank lending and the real interest rate. All the variables were transformed to comparable data by dividing nominal values with CPI of 1999Q1 as the base. In addition, the interest rate denoted lending rate in one year. As the authors carried out co-integration analysis for the variables, they employed the autoregressive distributed lag (ARDL) method for the econometric methodology. Moreover, they integrated the error correction model (ECM) as it combines the short-run and long-run relationship without losing long-run facts. Results showed that the property prices, GDP and bank lending were integrated of order one while interest rate was integrated of order zero. As the variables were a mixture of I (0) and I (1) variables, the ARDL model turned out to be appropriate. Except for the unidirectional relationship, the
12 research found bank lending, GDP and interest rate Granger caused property prices in the long-run though interest rate may not be a valid instrument variable.
Liang (2006) and Gerlach (2004) both chose total bank lending instead of property-related lending for the following considerations. Corporate sectors also got bank loans for property development; anecdotal evidence showed that part of the other lending to corporate and household sector were indeed put use for property-related investment; property market would be affected by broader market influenced by overall credit as well. In addition, Gerlach (2004) re-estimated the equations by using property-related lending with similar results. While the two studies got different conclusions about the relationship between property prices and bank lending.
4. Data and stylized facts 4.1 Data definition and Source
In order to examine dynamic relationships between property prices and bank loans to listed real estate development companies, I use the following variables: quarterly property prices, quarterly bank loans to listed real estate development companies, and control variables of quarterly growth rates for consumer price and gross domestic product. The sample period ranges from first quarter of 2001 to last quarter of 2012.
A simultaneously positive relationship between property prices and bank loans to real estate development companies is expected. Besides this relationship, it is also possible that property prices have positive and significant impacts on bank loans to listed real estate development companies. Instead of the causality running from property prices to bank loans, bank loans can affect property prices the other way around as well. The property prices data (PP), consumer price data (CP), and growth rate for gross domestic product data (GDP) used in this paper come from National Bureau of Statistics of China. Data for the bank loans are taken from CSMAR Industry Research database. In this section, I will discuss methods to construct these variables in detail and will discuss the summary statistics of the data.
The property prices (PP) data are used to measure price of commercial property in China including commercial residential buildings and business buildings. The cumulative sales and cumulative sales areas of commercial property are available monthly. In order to obtain the property prices quarterly, I generate both monthly sales and sales areas of property by taking the first difference of cumulative data. Accordingly, I get the average property prices monthly. To match the periodicity of this paper, I turn the monthly data to quarterly data by taking algorithm mean of the data in one quarter.
Turning next to the data of quarterly growth rates for GDP and consumer price (CP), quarterly growth rate for GDP is denoted as the growth rate of GDP in quarter t compared to GDP in quarter t-4. The original data,
13 monthly consumer price index, is the index that set same month of last year to be 100. Consequently, by taking difference of the index with 100, monthly consumer price growth rate is obtained; quarterly consumer price growth rate is consequently the average of monthly consumer price growth rate.
Before description of variable bank loans (BL) to listed real estate companies, I firstly define the real estate companies. Real estate companies discussed in this paper are defined by China Securities Regulatory Commission4, including construction industry and realty industry with industry code (E) and (K) separately.
CSMAR Industry Research database provides information such as currency, money, term, and purpose of every bank loan to listed real estate companies. The loans denoted by currency except for Renminbi occupy only a tiny fraction of all loans and their main purposes are not for construction of real estate. Therefore, in this paper, I only include the loans denoted by RMB for the real estate development purposes. As I have details of every loan including its amount, term, commence time, a dummy variable is generated to identify whether one loan is in use. If the loan is in use, then dummy variable will be one; otherwise, it will be zero. For example, Bank of China granted 249 million to Vanke in August, 2011 with a three year term. Consequently, this loan is in use for three years since the third quarter of 2011 and the dummy variable is hence 1 from 2011Q3 to 2012Q4 and 0 in other time periods. Finally, I calculate the amount of all loans whose dummy variables are one. Consequently, the sum represents the amount of loans in use, which is also the variable (BL) mentioned in this paper.
In order to make all variables comparable and regression efficient, property prices and bank loans are transformed into ratio by taking the ratio of property price in time t by property price in time t-1, which is denoted as 𝑃𝑃𝑡
𝑃𝑃𝑡−1. Also, it’s the same case for bank loans, of which the ratio is
𝐵𝐿𝑡
𝐵𝐿𝑡−1. In addition, as the ratios of
property prices and bank loans affect each other exponentially, natural logarithms of these ratios are taken in the regression.
4.2 Description of the data
Some description and analysis about the data will be presented before introduction of the econometric models. Instead of examine the econometrical relationship directly, let us first consider the correlation among these variables including natural logarithms of property prices ratio and bank loans ratio, 𝑙𝑛(𝑃𝑃𝑃𝑃𝑡
𝑡−1) and 𝑙𝑛(
𝐵𝐿𝑡
𝐵𝐿𝑡−1)
as well as growth rates of consumer price (CP) and GDP.
14 Table 1 Coefficients of Correlation
Dependent variable 𝐼𝑛( 𝑃𝑃𝑡 𝑃𝑃𝑡−1 ) 𝐼𝑛( 𝐵𝐿𝑡 𝐵𝐿𝑡−1 )
GDP growth rate CP growth rate 𝐼𝑛( 𝑃𝑃𝑡 𝑃𝑃𝑡−1 ) 1 𝐼𝑛( 𝐵𝐿𝑡 𝐵𝐿𝑡−1 ) -0.0802 1 GDP growth rate 0.0486 -0.1495 1 CP growth rate -0.0831 -0.0593 0.4039 1
Note: Table 1 describes the coefficients of correlation among variables𝑙𝑛(𝑃𝑃𝑃𝑃𝑡
𝑡−1) and𝑙𝑛(
𝐵𝐿𝑡
𝐵𝐿𝑡−1), quarterly
growth rates of gross domestic product (GDP) and consumer price (CP). As the table shows, except for the coefficient for correlation between GDP growth rate and𝐼𝑛( 𝑃𝑃𝑡
𝑃𝑃𝑡−1) and the coefficient between GDP growth
rate and CP growth rate, coefficients for other variables’ correlations are all negative. While values for the coefficients are relatively small, I further implement a regression to investigate their relationship.
In addition, Table 2 presents the overall description of all variables by showing their mean, median, and standard deviation. Furthermore, Chart 1, Chart 2, and Chart 3 depict the trends of these variables.
Table 2 Description of the variables
Dependent variable Mean Median Standard deviation Unit
Property Prices 3777 3659 1227 RMB/sq.m
Bank Loans 2367 1457 2181 RMB (Ten Millions)
GDP growth rate 10.25 10.10 1.94 Percentage
CP growth rate 2.47 2.13 2.40 Percentage
As both property price’ mean and bank loans’ mean are higher than the median and these two variables are with great standard deviations, it can be concluded that property prices and bank loans have been increasing rapidly while fluctuating greatly in the sample period. On the other hand, GDP also grew rapidly and steadily as standard deviation for GDP growth rate is low. In addition, consumer price increased a little in sample period with a mean of 2.47% for the growth rate of consumer price. After describing absolute values for these variables, Chart 1, 2 and 3 presents relative trends as following:
As the variables are with different units, different y coordinates are applied for Chart 1 and 2. In order to clearly illustrate the changes of all the variables, I put property prices (PP), bank loans (BL) and the growth rate of GDP in one chart while property prices (PP) with the growth rates of both GDP and consumer price (CP) in another chart to show their relationship.
15 Chart 1
Shown in Chart 1, property prices of China increase steadily and to almost triple times at the end of fourth quarter of 2012. During the same period, growth rate GDP also increased steadily except for the year 2008 due to the effect of global financial crisis. In addition, bank loans surged during the time period and the value of the last quarter was almost 7 times of the value in 2001Q1.
Chart 2
Chart 2 presents bank loans and growth rates of both GDP and consumer price (CP). From the chart, a rough positive relationship between growth rate of GDP and property prices can be detected. Furthermore, growth
0% 2% 4% 6% 8% 10% 12% 14% 16% 0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000 10,000 20 01 Q 1 2001Q 3 2002Q 1 2002Q 3 2003Q 1 2003Q 3 2004Q 1 2004Q 3 2005Q 1 2005Q 3 2006Q 1 2006Q 3 2007Q 1 2007Q 3 2008Q 1 2008Q 3 2009Q 1 2009Q 3 2010Q 1 2010Q 3 2011Q 1 20 11 Q 3 2012Q 1 2012Q 3
Chart 1 Property prices, bank loans and GDP growth rate
Bank loans Property prices GDP growthrate
-4% -2% 0% 2% 4% 6% 8% 10% 12% 14% 16% 0 1000 2000 3000 4000 5000 6000 7000 2001Q 1 2001Q 3 2002Q 1 2002Q 3 2003Q 1 2003Q 3 2004Q 1 20 04 Q 3 2005Q 1 2005Q 3 2006Q 1 2006Q 3 2007Q 1 2007Q 3 2008Q 1 2008Q 3 2009Q 1 2009Q 3 2010Q 1 2010Q 3 2011Q 1 2011Q 3 2012Q 1 2012Q 3
Chart 2 Property prices, GDP growth rate and CP growth rate
16 rates for GDP and consumer price (CP) presents some kind of simultaneity. Nevertheless, as property prices and bank loans were not stable in sample period, natural logarithms of their ratios based on the values of last quarter are used in the regression and graphed in Chart 3.
Chart 2
Shown by Chart3, the natural logarithms of both bank loans ratio and property prices ratio are positive of the most time as both bank loans and property prices increased in our sample period. In addition, while property prices fluctuated more in the sample period, natural logarithms of property prices ratio also deviate more form zero.
5. Econometric models
Liang (2006) applied a high dimensional autoregressive distributed lag (ARDL) model and took gross domestic product (GDP) and interest rate into consideration. In order to adopt the model, the study employs the autoregressive distributed lag bonds testing approach that has numerous advantages: better small sample properties; avoiding possible threats of variables being nonstationary; integrating short-run dynamics with the long-run dynamics through linear transformation to error correction model (ECM). Moreover, the study declared that the interest rate, especially in short-term, was not an effective explanatory variable for changes in property prices in China over the sample period. Gerlach (2004) based the study on the multivariate approach to integration tests and employed VAR model to analyze the long-run relationship between bank lending, real GDP and real property prices.
Several other scholars have also implemented the vector autoregressive model to research on property prices and economic fundamentals. Seow Eng (1994) use the vector autoregressive model to investigate the link
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 2001Q 2 2001Q 4 2002Q 2 2002Q 4 2003Q 2 2003Q 4 20 04 Q 2 2004Q 4 2005Q 2 2005Q 4 2006Q 2 2006Q 4 2007Q 2 2007Q 4 2008Q 2 2008Q 4 2009Q 2 2009Q 4 2010Q 2 2010Q 4 2011Q 2 20 11 Q 4 2012Q 2 2012Q 4
Chart 3 Natural logrithms of bank loans ratio and property prices ratio
17 between real estate and property prices. The VAR model provided interactive analysis between property stock and prices of property and offered insights of how variables influenced one another. Luintel and Khan (1999) also used multivariate time series framework to analyze the long-run causality between financial development and economic growth. The model filled the gap of a conspicuous lack of multivariate time series studies and eliminated single equation bias, which might exist in previous research.
Following previous study, the thesis applied multivariate vector autoregressive (VAR) model to exhibit dynamic relationship between property prices and bank loans to real estate development companies as it could analyze the interactive causality between these variables and long-term relationship.
Moreover, interest rate is not regarded as control variable in this paper as for the following reasons. Liang (2006) has already clarified that interest rate was not an effective control variable for the property prices in China. What’s more, the interest rate is determined by the authority instead of the market and kept stable in sample period. Interest rate discussed here refers to the yearly RMB benchmark lending rate of financial institutions provided by the People’s Bank of China. In addition, the adjustment of yearly RMB benchmark lending rate of financial institutions over sample period is presented in the appendix. Except the year 2007 and 2008, all terms of interest rate were adjusted about every two years during the sample period. At the same time, property prices and bank loans increased at an annual rate of 7.42% and 30.59%, respectively, which means that the interest rates of lending stayed same while bank loans and property prices increased rapidly. Therefore, interest rate could hardly explain fluctuation of property prices nor bank loans in sample period. 5
5.1 Augmented Dickey Fuller---Unit root test
To ensure that the model constructed will be stationary, I firstly conduct a unit root test in the raw time series data. If the time series data has a unit root, then the first order difference is taken to make the time series stationary. It also makes the variables in the model more comparable. Aiming to control for plagued by serial correlation, I adopt the Augmented Dickey–Fuller (ADF) test procedure. The ADF test is performed on the quarterly property prices, quarterly bank loans to listed real estate development companies, quarterly GDP, and quarterly consumer price, respectively. Since these four variables’ time series exhibit different means and trends, I use separate regression equations to conduct ADF tests for these variables.
∆𝑦𝑡 = 𝛼 + 𝛽𝑦𝑡−1+ (𝛿𝑡) + 𝜑1∆𝑦𝑡−1+ 𝜑2∆𝑦𝑡−2+ ⋯ + 𝜑𝑘∆𝑦𝑡−𝑘+ 𝜖𝑡
18 In the equation above, ∆𝑦𝑡 and k stands for first order difference of time series 𝑦𝑡 and number of lags for∆𝑦, respectively. The null hypothesis transforms from variable containing unit root to𝛽 = 0. Since time series for GDP and consumer price do not show strong increasing or decreasing trends, time trend 𝛿𝑡 is excluded from the regression. On the other hand, property prices and bank loans increase rapidly especially after the second season of 2008, thus time trend item is included and the null hypothesis does not restrict 𝛼 as the regression is conducted with or without drift.
As results show that property prices and bank loans are not co-integrated in China during sample period. This paper implements VAR (vector autoregressive model) that analyzes the linear interdependences among property prices and bank loans time series.
5.2 Construction of VAR model
After the unit root test and co-integration test, a VAR model is constructed to discuss the relationship between bank loans and property prices from 2001Q1 to 2012Q4 in China. As mentioned above, growth rates of GDP and consumer price are two variables that have impacts on both property prices and bank loans. Therefore, in the VAR model, they are regarded as exogenous variables controlling for other factors that determine bank loans and property prices. As the unit root test shows that only first-difference for the growth rates of GDP ad CP are stationary, changes of these two variables are included in the regression instead of absolute values. The following is the model specification for the VAR model
𝑙𝑛 ( 𝑃𝑃𝑡 𝑃𝑃𝑡−1 ) = ∑ 𝛽1,𝑖 𝑝 𝑖=1 𝑙𝑛 ( 𝑃𝑃𝑡−𝑖 𝑃𝑃𝑡−𝑖−1 ) + ∑ 𝛾1,𝑖 𝑝 𝑖=1 𝑙𝑛 ( 𝐵𝐿𝑡−𝑖 𝐵𝐿𝑡−𝑖−1 ) + 𝛿1∆𝐶𝑃𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒𝑡+ 𝜗1∆𝐺𝐷𝑃𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒𝑡+𝜀1,𝑡 𝑙𝑛 ( 𝐵𝐿𝑡 𝐵𝐿𝑡−1 ) = ∑ 𝛽2,𝑖 𝑝 𝑖=1 𝑙𝑛 ( 𝐵𝐿𝑡−𝑖 𝐵𝐿𝑡−𝑖−1 ) + ∑ 𝛾2,𝑖 𝑝 𝑖=1 𝑙𝑛 ( 𝑃𝑃𝑡−𝑖 𝑃𝑃𝑡−𝑖−1 ) + 𝛿2∆𝐶𝑃𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒𝑡+ 𝜗2∆𝐺𝐷𝑃𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒𝑡+ 𝜀2,𝑡
𝑃𝑃𝑡denotes property prices in quarter t, 𝐵𝐿𝑡 represents bank loans to real estate companies in quarter t with 𝐶𝑃𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒𝑡 and 𝐺𝐷𝑃𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒𝑡 denoting growth rates of consumer price and gross domestic product in the same period. 𝜀𝑖 is the residual term and p is the number of lags. While the ratios of property prices and bank loans affect each other exponentially, natural logarithms (ln) of these ratios are taken in the regression.
In order to determine the optimal lag number, I use the approach similar to Ullah et al (2003), which choose the m that minimize the Final Prediction Error (FPE) and it is the same case with n. In the following equation, T represents the sample size with k and SSR denotes number of coefficient and sum of squired residuals, separately. FPE(𝑚1) = (𝑇 + 𝑘) (𝑇 − 𝑘)[ 𝑆𝑆𝑅(𝑚1) 𝑇 ]
19 6. Empirical results
Results of unit root test, co-integration test will firstly be presented. Following these tests and the construction of the VAR model, I will finally exhibit the regression results of VAR model and describe the relationship between property prices and bank loans.
6.1 Result of the unit root test
Table 3 exhibits the result of unit root tests carried out on property prices, bank loans, GDP and consumer price.
Table 3 Augmented Dickey Fuller--- Unit root test
Level Change 𝑙𝑛 ( 𝑃𝑃𝑡 𝑃𝑃𝑡−1) -10.98 (N)** -17.21 (N)** 𝑙𝑛 ( 𝐵𝐿𝑡 𝐵𝐿𝑡−1) -4.874 (N) ** -13.12(N)** GDP growth rate -1.605 (C) -6.097 (C)** CP growth rate -1.798 (C) -3.717 (C)**
Note: Letters T, C and N clarify whether the test regression includes a time trend and a constant (T), or just a constant (C) or neither a trend nor a constant (N). As quaterly growth rates of GDP, consumer price does not show a trend but have a nonzero mean, the regression equation contains constant term but without trend. The natural logrithm of property prices ratio and bank loans ratio, on the other hand fluctuate around zero and show no trend. Thus regeression model for the natural logrithm of propety prices ratio and bank loans ratio include neither constant nor trend. **denotes the statistics is significant at 1% level.
Shown in the table, the hypothesis that time series of 𝑙𝑛 (𝑃𝑃𝑃𝑃𝑡
𝑡−1) and 𝑙𝑛 (
𝐵𝐿𝑡
𝐵𝐿𝑡−1)have unit root can be
rejected at 1% level, which means the variables are stationary at their original level. Variables GDP growth rate and CP growth rate are stationary after taking first-order difference. Therefore, ∆𝐺𝐷𝑃𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒 and ∆𝐶𝑃𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒are used as control variables to keep the whole model stationary.
20 6.2 Result of VAR model
Exploiting the method of minimizing the Final Prediction Error (FPE) to determine the optimal lag, it turns out to be two. After specifying the regression equation, I get the results of VAR model, only coefficients with significant statistics are shown in equations to highlight the relationship between the interested variables. The equation obtained for 𝑙𝑛 ( 𝑃𝑃𝑡
𝑃𝑃𝑡−1)is as the following:
𝑙𝑛 ( 𝑃𝑃𝑡 𝑃𝑃𝑡−1 ) ̂ = 0.038 − 0.549𝑙𝑛 (𝑃𝑃𝑡−1 𝑃𝑃𝑡−2 ) − 0.042𝑙𝑛 (𝑃𝑃𝑡−2 𝑃𝑃𝑡−3 ) − 0.046𝑙𝑛 (𝐵𝐿𝑡−1 𝐵𝐿𝑡−2) − 0.015𝑙𝑛 ( 𝐵𝐿𝑡−2 𝐵𝐿𝑡−3) (0.011)** (0.145) ** (0.147) (0.057) (0.060) +0.999∆𝐺𝐷𝑃𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒𝑡− 0.049∆𝐶𝑃𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒𝑡 (0.901) (0.762) The R square is 30.65%.
The equation obtained for 𝑙𝑛 (𝐵𝐿𝐵𝐿𝑡
𝑡−1) is as the following: 𝑙𝑛 ( 𝐵𝐿𝑡 𝐵𝐿𝑡−1) ̂ = 0.028 + 0.001𝑙𝑛 (𝑃𝑃𝑡−1 𝑃𝑃𝑡−2 ) + 0.222𝑙𝑛 (𝑃𝑃𝑡−2 𝑃𝑃𝑡−3 ) + 0.174𝑙𝑛 (𝐵𝐿𝑡−1 𝐵𝐿𝑡−2) + 0.353𝑙𝑛 ( 𝐵𝐿𝑡−2 𝐵𝐿𝑡−3) (0.027) (0.353) (0.359) (0.139) (0.146) * −2.033∆𝐺𝐷𝑃𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒𝑡+ 1.563∆𝐶𝑃𝑔𝑟𝑜𝑤𝑡ℎ𝑟𝑎𝑡𝑒𝑡 (2.198) (1.890) The R square is 23.85%.
Numbers in the parentheses denote standard deviation of the coefficients; * denotes for 5% level and **denotes 1% level.
I also calculate the eigenvalue for VAR model to test for its stability and the results show that all the eigenvalues lie inside the unit circle. Therefore, VAR satisfies stability condition
According to the regression results, natural logarithm of property prices ratio is negatively correlated with its first and second lags. As the property price ratio indeed represents the quarterly growth rate of property prices, it can be concluded that if the property prices increased rapidly in quarter t-1, then the growth rate would decrease in quarter t. Specifically, as the first lag 𝑙𝑛(𝑃𝑃𝑡−1
𝑃𝑃𝑡−2) is highly significant, it coefficient can be
21 rate for property prices in next quarter. In other words, property prices in our sample period adjusted its growth rate based on last quarter’s growth and performed kind of oscillation. This result coincides with Peter Englund and Ioannides (1997), which found the natural logarithm of the difference of property prices
negatively correlated with fifth lag. Regression equation for 𝑙𝑛 ( 𝐵𝐿𝑡
𝐵𝐿𝑡−1) shows that the growth rate of bank loans are positively correlated with its lags. As data for bank loans is generated as “in use”, which means that loans granted in quarter t-1 or in quarter t-2 with a term longer than one year is also included in quarter. In addition, as the banks granted loans to real estate development companies according to performance of previous loans, it’s reasonable for the banks to grant more loans in the situation of growing property prices, in which there’s little bad debt. Coefficient for 𝑙𝑛 (𝐵𝐿𝑡−2
𝐵𝐿𝑡−3) shows that 1% increase of bank loans growth rate in quarter t-2 in correlated with
0.353% increase of bank loans growth rate in quarter t.
Nevertheless, no significant relationship has been found between natural logarithm of bank loans growth rate and natural logarithm of property prices growth rate. Moreover, the control variables are not significant in neither regression. Several possible explanations are discussed in the following paragraph:
Although real estate development mainly depends on domestic loans in China, other financing institutions such as trusts also provide debts to the real estate development companies. In other words, loans granted to the real estate companies did not only come from banks. In addition, some variables such as demographic characteristics are not included in the regression and the sample period only contains 48 quarters.
Therefore, omitted variables and data scarcity both might cause relationship between banks loans and property prices not significant.
6.3 Result of Granger causality test
Regression equations have already revealed some relationship between property prices and bank loans to real estate companies. In order to get deeper understanding of the causality between these two variables, I carry out the Granger causality test. In terms of “causality,” it means that given the past values of y, if past values of x are useful for predicting y, then variable x Granger-causes variable y. The Granger causality firstly calculates and reports Wald tests that the coefficients on all the lags of an endogenous variable are jointly zero. Consequently, it tests the hypothesis that each of the other endogenous variables does not Granger-cause the dependent variable in the equation.6 In other words, the Granger causality test is a joint test for the
prediction power of endogenous variables on dependent variables. Results of the Granger-causality test are exhibited in Table 4.
22 Table 4 Results of Granger causality test
Equation Excluded chi2 df Prob >chi2 𝑙𝑛 ( 𝑃𝑃𝑡 𝑃𝑃𝑡−1) 𝑙𝑛 ( 𝐵𝐿𝑡 𝐵𝐿𝑡−1) 0.7134 2 0.7 𝑙𝑛 ( 𝑃𝑃𝑡 𝑃𝑃𝑡−1) ALL 0.7134 2 0.7 𝑙𝑛 ( 𝐵𝐿𝑡 𝐵𝐿𝑡−1) 𝑙𝑛 ( 𝑃𝑃𝑡 𝑃𝑃𝑡−1) 0.4327 2 0.805 𝑙𝑛 ( 𝐵𝐿𝑡 𝐵𝐿𝑡−1) ALL 0.4327 2 0.805
Note: The “Equation” column stands for dependent variable with the “Excluded” column meaning endogenous variables.
The test statistics chi2 states that the hypotheses of natural logarithm of property prices ratio (bank loans ratio) do not Granger cause natural logarithm of banks loans ratio (property prices ratio) cannot be rejected. In other words, the lags of natural logarithm of property prices ratio (bank loans ratio) are not jointly useful in predicting natural logarithm of banks loans ratio (property prices ratio). Moreover, fist differences of GDP and consumer price are not included in granger causality test as they are not significantly as control variables in the regression.
7. Conclusion
The study focuses on the relationship between bank loans and property prices in China during the period from 2001Q1 to 2012Q based on VAR model while controlling for the first difference of quarterly growth rates for GDP and consumer price. The model is based on the natural logarithm of property prices ratio and bank loans ratio instead of absolute values, which explains more about the growth rates for both bank loans and property prices. The results show that growth rate of property prices is negatively correlated with its lags while growth rate of bank loans is positively correlated with its lags. Referring to the empirical results, no significant relationship can be drawn between the growth rates of property prices and bank loans. Further research may include variables such as demographic characteristics and apply panel data over longer time period.
23
Appendix:
Renminbi Benchmark Lending Rate of Financial Institutions
Unit: annual interest rate%
Within Six Months Six Months to One Year One Year to Three Years Adjustment Time (Including Six Months) (Including One Year) (Including Three Years)
1999.06.10 5.58 5.85 5.94 2002.02.21 5.04 5.31 5.49 2004.10.29 5.22 5.58 5.76 2006.04.28 5.4 5.85 6.03 2006.08.19 5.58 6.12 6.3 2007.03.18 5.67 6.39 6.57 2007.05.19 5.85 6.57 6.75 2007.07.21 6.03 6.84 7.02 2007.08.22 6.21 7.02 7.2 2007.09.15 6.48 7.29 7.47 2007.12.21 6.57 7.47 7.56 2008.09.16 6.21 7.2 7.29 2008.10.09 6.12 6.93 7.02 2008.10.30 6.03 6.66 6.75 2008.11.27 5.04 5.58 5.67 2008.12.23 4.86 5.31 5.4 2010.10.20 5.1 5.56 5.6 2010.12.26 5.35 5.81 5.85 2011.02.09 5.6 6.06 6.1 2011.04.06 5.85 6.31 6.4 2011.07.07 6.1 6.56 6.65 2012.06.08 5.85 6.31 6.4 2012.07.06 5.6 6 6.15
24 Renminbi Benchmark Lending Rate of Financial Institutions (continued)
Unit: annual interest rate%
Within Six Months Six Months to One Year One Year to Three Years Adjustment Time (Including Six Months) (Including One Year) (Including Three Years)
1999.06.10 5.58 5.85 5.94 2002.02.21 5.04 5.31 5.49 2004.10.29 5.22 5.58 5.76 2006.04.28 5.4 5.85 6.03 2006.08.19 5.58 6.12 6.3 2007.03.18 5.67 6.39 6.57 2007.05.19 5.85 6.57 6.75 2007.07.21 6.03 6.84 7.02 2007.08.22 6.21 7.02 7.2 2007.09.15 6.48 7.29 7.47 2007.12.21 6.57 7.47 7.56 2008.09.16 6.21 7.2 7.29 2008.10.09 6.12 6.93 7.02 2008.10.30 6.03 6.66 6.75 2008.11.27 5.04 5.58 5.67 2008.12.23 4.86 5.31 5.4 2010.10.20 5.1 5.56 5.6 2010.12.26 5.35 5.81 5.85 2011.02.09 5.6 6.06 6.1 2011.04.06 5.85 6.31 6.4 2011.07.07 6.1 6.56 6.65 2012.06.08 5.85 6.31 6.4 2012.07.06 5.6 6 6.15
25
Bibliography:
Manuel Adelino, Antoinette Schoar, Felipe Severino (2012) Credit Supply and House Price: Evidence from mortgage market segmentation, NBER working paper 17832
Guangcan Cui (2009), An empirical study on the interaction between prices of property and macro economy, Economic Theory and Business Management, Jan, 2009
China Insurance Regulatory Commission (2010), Interim Measures of Insurance Fund Investing in Real Estate,” No. 80
Zhongdong Duan, Linghua Zeng and Zexian Huang (2007), An Empirical Study of Price Fluctuation of Real Estate Industry and Banking Credit Increase, Finance Forum, Feb, 2007
Peter Englund, Yannis M. Ioannides (1997), House Price Dynamics: An International Empirical Perspective, Journal of Housing Economics 6, 119–136
Ong Seow Eng (1994), Structural and Vector Autoregressive Approaches to Modelling Real Estate and Property Stock Prices in Singapore, Journal of Property Finance, Volume 5 Iss 4 ,pp. 4 – 18
Trevor Fitzpatrick and Kieran Mcquinn (2007), House prices and mortgage credit: empirical evidence from Ireland, The Manchester School, Volume 75, No. 1, Jan 2007, Pages 82–103
Edward L. Glaeser, Joseph Gyourko, Albert Saiz (2008), Housing supply and housing bubbles, Journal of Urban Economics 64, 198–217
Stefan Gerlach, Wensheng Peng (2004), Bank lending and property prices in Hong Kong, Journal of Banking & Finance, Volume 29, Issue 2, February 2005, Pages 461–481
Mark Gertler, Nobuhiro Kiyotaki and Albert Queralto (2012), Financial crises, bank risk exposure and government financial policy, Journal of Monetary Economics, Volume 59, Supplement, 15 December 2012, Pages S17–S34
David Geltner, Norman G Miller, Clayton (2013), Commercial Real Estate Analysis and Investments, 738-743
Charles Goodhart (2000), Do asset prices help to predict consumer price inflation? The Manchester School Supplement, pages 122-140
Charles Goodhart and Boris Hofmann(2008), House price, money, credit, and the macro economy, Oxford Review of Economic Policy, Volume 24, Number 1, pp. 180–205
26 C.Hott (2011), Lending behavior and prices of property, Journal of Banking& Finance, Volume 35, pp. 2429-2442
Richard J. Herring, Susan Watcher (1998), Real Estate Booms and Banking Buts: an international perspective, presented at the Wharton Conference on Asian Twin Financial Crises
Qi Liang, Hua Cao (2006), Property prices and bank lending in China, Journal of Asian Economics, Volume 18, Issue 1, February 2007, Pages 63–75
Winston T.H. Koha, Roberto S. Marianoa, Andrey Pavlovb, Sock Yong Phanga, Augustine H.H. Tana, Susan M. Wachterc, Bank lending and real estate in Asia: market optimism and asset bubbles, Journal of Asian Economics, Volume 15, Issue 6, January 2005, Pages 1103–1118
Michael Koetter, Tigran Poghosyan (2009), Prices of property and bank stability, Journal of Banking & Finance, Volume 34, 2011, pp. 1129-1138
Koll B. Luintel and Mosahid Khan (1999), A quantitative reassessment of the finance–growth nexus: evidence from a multivariate VAR, Journal of Development Economics, Volume 60, 1999, Pages 381– 405
Edward H K Ng, Rosaline Chow (2004), Availability of bank credit and the residential property prices level: evidence from Singapore
John Okunev and Patrick J. Wilson. (1997). " Using Nonlinear Tests to Examine Integration between Real Estate and Stock Markets,” Real Estate Economics, Volume 25, Issue 3, pages 487–503
Sae Woon Park & Doo Woan Bahng & Yun W. Park (2008), Price Run-up in Housing Markets, Access to Bank Lending and House Price in Korea
Denise Di Pasquale, William C. Wheaton (1991), Housing Market Dynamics and the Future of Housing Price, Journal of Urban Economics, Volume 25, 1994, pp.1-27
Qingke Research Center (2011), China private real estate investment fund annual report in 2011 Richard T. Selden (1959), Cost-Push versus Demand-Pull Inflation, Journal of Political Economy, Vol. 67, No. 1 pp. 1-20
Yue Shen and Hongyu Liu, Relationship between real estate development investment and GDP in China, Journal of Tsinghua University(Science and Technology), Sep, 2004
27 Weian Wang and Cong He (2005), Prices of property and Inflation Expectation, Journal of Finance and Economics, Dec, 2005
Benhua Xiao (2008), Credit Expansion and Prices of property in China, Journal of Shanxi Finance and Economics University, Jan, 2008
Shen Yue and Liu Hongyu (2004), Housing Prices and Economic Fundamentals: A Cross City Analysis of China for 1995—2002, Economic Research Journal, Jun, 2014
Websites:
NATIONAL Bureau of Statistics of China, http://data.stats.gov.cn/workspace/index?m=hgnd , 26th, Jun
2014
Ullah, A. & Z. G. Zhou (2003) Real estate and stock returns: a multivariate VAREC model. Journal of Forecasting, 21(1), 181 ‐192
The Central People's Government of the People's Republic of China, http://www.gov.cn/jrzg/2010-04/24/content_1591123.htm , 17th, Jun 2014
The Central People's Government of the People's Republic of China, http://www.gov.cn/xinwen/2014-06/06/content_2695695.htm , 22nd, Jun 2014
An American Perspective from China, http://chovanec.wordpress.com/2012/07/31/whats-driving-chinas-real-estate-rally-part-3/ , 29th, Jul 2014
The People’s Bank of China,
http://www.pbc.gov.cn/publish/zhengcehuobisi/627/index.html, 13th, Aug 2014
http://tianjin.pbc.gov.cn/publish/fzh_tianjin/2909/2014/20140331145449847589008/201403311454 49847589008_.html, 15th, Nov, 2014
