The effect of monetary policy on the housing
price across different regions in China
Name: He Huang
Student Number: 10628053
Business Economics: Finance & Real Estate Finance Amsterdam Business School
August 15, 2014
Table of Contents
1 Introduction ... 3 2 Literature Review ... 4 3 Background ... 12 4 Data ... 17 5 Methodology... 286 Results and Interpretation ... 29
7 Conclusions... 40
8 References ... 45
9 Appendixes ... 53
I.
Introduction
China is the second biggest economy in the world and will be replacing the USA as the biggest economy entity in the coming decades. The real estate sector is said to be the pillar of Chinese economy, which means it also poses a threat. The real estate sector itself is a very controversial one where there are lots of interesting debates around: Is it dangerous that our economy is largely based on real estate? The housing price/disposable income ratio is ridiculously high comparing to the developed countries. Is there a real estate bubble? The inspiring is basically from the recent financial crisis where the crisis origins from US housing mortgage loans. In the US, the highly expansionary monetary policy was cited as one of the key reasons leading to the real estate price bubble that burst during the recent world-wide financial crisis.1 If the easy loans caused the real estate bubble in the US, what the impact
would be for Chinese real estate after the Chinese Central Bank gave out the huge monetary stimulation into the economy? Considering the complexity of different regional economic profiles in China, what impact would a national monetary policy give on all those different regions? When an expansionary monetary policy comes out, will the property market in less-developed regions response the same as the one in the well-developed regions?
The research question for this thesis can therefore be stated as follows: “Firstly, What is the impact of monetary policy on price in regional property markets? Secondly, given that monetary policy indeed has an impact on housing market, how can we explain the different responses from the property markets across the regions?”
Previous research has demonstrated monetary policy would have an impact of the housing price (Taylor 2007, etc.). This thesis carefully either adapts or simplifies their theoretical framework into our scenario. Our empirical evidence shows that the monetary policy indeed also affects the Chinese housing market in a similar way comparing to the original empirical evidence in the previous literature. Secondly, there are very few studies that have analyzed monetary policy on the regional level of the property market in China. The literatures 1 The Fuel That Fed The Subprime Meltdown, by Ryan Barnes, Feb. 26, 2009, Investopedia
3
concerning Chinese housing market and monetary policy only study the average housing price on the national level (Xu and Chen, 2011; So and Lan, 2011)2. The complexity of regional
differences in China is simply ignored by those literatures. In our model, we use “house price of a city” as one dependent variable. And we have around 70 cities, thus 70 study objects. This gives us to a possibility study the regional effects of monetary policy on housing price.
The organization of the paper goes like this: Section 1 is introduction where we raise up an issue here. Section 2 presents literature reviews which gives evidence on the relationship between housing market and monetary policy effects from other studies. In Section 3, we present the background of monetary policy in China, housing markets across different regions and the market behavior on the microeconomic level. In Section 4, we fully describe the dataset and link those data to the real world. In Section 5, we introduce the methodology and design a simple model to test the economic reasoning. Then we test the hypothesis by running regressions, and will highlight the results. In Section 6, we will discuss the results and interpret the economic meaning. In section 7, we summarize all the analyses and give a conclusion of this paper
II. Literature review
2.1. The effect of monetary policy on the asset price
Here we raise our first question: do we have enough evidence to say that the monetary policy affects housing market as well and through which channel? Many studies give evidence on the correlation between the inflation rate, asset price and the rate of money growth. McCandless and Webert, (1995) confirms the inter-linkage between money growth, inflation rate and asset price. They suggest that at least in the short-run, the money growth will boost the asset price and real output. In Lastrapes (2002), he identified the monetary policy effect on housing market
2
Xu, and Chen. January, 2011. The Effect of Monetary Policy on Real Estate Price Growth in China; Jacky So., Olivia Lan., 2011. Monetary Policy and Housing Price Bubbles, the China Experience
4
by VAR model. He reveals that money supply shock has an impact on the housing market. Given that money policy does indeed affect the housing market, here we raise a second question: Should monetary policy seek a stabilized asset prices by implementing different monetary policy tools? Michael Mussa (2002) gives a fundamental analysis of real estate market, suggesting that real estate price jumps around a lot. The mechanism for the property price movement is still not very well understood. Suppression of housing price volatility might not be so feasible. This gap cannot be closed because in financial market, they have options and futures to adjust the fundamental prices while in real estate, there is not such real-time market. In, Taylor (2007), Taylor judges the interest rate standard by applying Taylor rule. He finds out that during the 2003 and 2004 when the short-term interest rates are exceptionally low, and this monetary shock may have substantially contributed to the boom in housing market. The boom brings a higher housing price, a falling delinquency and more favorable credit ratings and higher demand for housing. On the contrary, the following quotes explain everything: “as the short term interest rate returned to normal levels, housing demand rapidly fell bringing down both construction and housing price inflation. Delinquency and foreclosure rates then rose sharply, ultimately leading to the meltdown in the subprime market and on all securities that were derivative from the subprime.”3
Jarocinski (2006) examines the role of housing investment and house prices in U.S. business cycles since the second half of the 1980s using an identified Bayesian VAR. They suggest that there is evidence that monetary policy has significant effects on housing investment and house prices. The easy monetary policy that designed to counter the deflation in 2002-2004 has contributed to the boom in the housing market in 2004 and 2005. Bernanke (2010) studied the housing market in the U.S. he also finds out that the most rapid house price increases occurred when short-term interest rates were at their lowest levels. However the magnitude of house price gains seems too large to be fully explainable by the monetary shocks alone. Above all, the direct link between monetary policy and house price should be weak because monetary policy always works with a lag. Following this doubt, he finds out that the housing finance 3 Taylor, Housing and Monetary Policy, December 2007, NBER Working Paper No. w13682
5
development is the one of the key elements that also contributes to the house price changes. As the mortgage system develops and easy mortgage becomes more available, housing market will be stimulated and house price will increase.
For the empirical evidence, Maclenna, Muelbauer and Stephens (2001) study the 15 member states of the EU. The paper argues that housing and financial markets will response to the output, inflation and short-term interest rates. Given the differences in housing financial system, tenure patterns in different countries, the reactions to interest rate changes are also various. This paper gives a strong hint that property markets across regions with different characteristics will react different to a convergent monetary policy. The scenario is similar to our study object-Chinese housing market, with various regions and one single convergent monetary policy.4
2.2. Basic concept: Valuation, Mortgage rate, and discount rate
The interest rate is the tool that would affect any type of investment, including real estate sector. The mortgage rate is the cost of buying a house, and the mortgage is dependent on the interest rate. The interest rate also has an impact on capital flows, where it often involves the required return of investment. If the interest drops, this means that the cost of funds is reduced and funds would flow into the real estate sector; vice versa, when interest rate goes up, the cost of capital goes up as well. These changes in capital flows have an impact on the supply and demand for property. Same goes for the concept of discount rate. The discount or capitalization rate is seen as an investor’s required dividend rate, which is risk premium plus risk free interest (which is interest rate). When interest rate goes up, the discount rate goes up. The property price will go down because there are not enough favorable property can meet up with such a high discount rate; vice versa, when interest rate goes down, the discount rate goes down as well. The property price will increase because there are quite a lot of properties can easily satisfied the low discount rate requirement and people are buying those properties.5
4
Maclennan, Muelbauer, Stephens, Asymetries in Housing and Financial Market Institutions and EMU, July 2000
5
Based on Rober Stammer, How interest rates affect property values, Sep. 28, 2013 & Barry Nielsen, How interest rates affect the housing market, June 17, 2013, Investorpedia
6
2.3 Regional effects of monetary policy
There are quite a few studies have identified the heterogeneous performance of regional markets as a result of monetary policy. Arnold and Vrugt (2001) analyze the regional effects of monetary policy in Europe. They conclude the activity level will determine the level of sensitivity.6 In this sense, we would tell that in big cities, the house price would react more
actively comparing to that in small cities, due to possible market speculation.
There are also studies from a different prospective to analyze the regional effects. Carlino and DeFina (1995) tell a different story. They provide evidence that different regions will react to interest rate quite different in the United States. The analysis found that states with relatively large shares of output accounted for by real estate industry are less sensitive to change in monetary policy than the more industrially diverse states.
Figure 1:industry sensitivity to monetary policy
Share of Total output Attributable To Selected Interest-Sensitive Industries
State Construction Durable Goods Finance, Insurance, & Real Estate Policy Response Delaware 5.0 9.8 19.8 -1.00 Maryland 5.9 7.1 18.1 -0.92 New Jersey 4.4 9.6 18.4 -1.06 New York 3.4 9.4 24.3 -0.72 Pennsylvania 4.5 14.8 15.6 -1.14 U.S. Average 4.7 11.5 14.9 -1.16 7
The higher the concentration on real estate, e.g. New York, the less responsive it would be to 6
Arnold, Page 417, The Regional Effects of Monetary Policy in Europe, Journal of Economic Integration, 16(3), September 2001; 399-420
7
Carlino, DeFina, Page 23, Do States Respond Differently To Changes in Monetary Policy? Business Review July/August 1999, original data can be found in Appendix Picture
7
monetary shock. When there is a monetary shock, the highly concentrated property market in the New York would hold the price while for the city like Pennsylvania, less developed property market would probably suffering a big volatility. The paper also claims that the states with large concentration of small firms tend to be more sensitive to monetary policy shocks, but the correlation is not that significant. This theory focuses only on the composition of industry, instead of the size of the city. However, it suggests that we should check the composition of industry when there are two cities with the same size but different sensitivity on monetary policy shocks.
2.4 M2 Money Supply
Definition of M2 Money supply“M2 money supply is a key economic indicator used to forecast inflation. M2 is the money supply that includes M1 plus savings and small time deposits, overnight repos at commercial banks, and non-institutional money market accounts”.8 M2 is generally considered to be the
intermediate monetary policy tool in China. When we talk about credit supply affects the housing price and investment, we are actually talking about M2 money supply and the housing price & investment. The mechanism of how money supply affects the real estate sector is quite similar to the one with interest rate. Actually, the M2 money supply, interest rate, and reserve requirements ratio are co-related; we still single them out as variables to study our topics.
M2 Money supply, housing price
The amount of money in circulation has a tremendous influence on the performance of the economy, including real estate sector. If there is a growth on M2 supply, the money would flow into real estate and causing housing price increase. This easy-money policy encourages investors to invest more as well as consumers to purchase houses. Vice versa, if central bank targets to lower the inflation, thus implement a tightened money supply, this will cause a drop in real estate activity.9
8
“M2 Definition”. Investorpedia. Retrieved 2008-07-20
9 Dennis J. et al Page 108, Essentials of Real Estate Economics 6th Edition, Cengage Learning
8
2.5 Stock Index & house price
The correlation between the stock index and house price is also an interesting topic. The Chinese house market is considered to be an investment-oriented market. The stock market, another investment alternative, here serves as a convenient comparison group. The attempt to investigate whether the returns are correlated between the stock market and house market would probably bring some interesting results. Sun & Liu (2005) concluded that the returns on residence and stock are not correlated. Those two assets are fundamentally different in nature. Their investment cycles, costs, risks and return patters are quite different. However, there are also empirical studies, e.g. Zhang & Fung, (2006) that prove they do have some weak correlation in terms of return. It is the high level in returns on real estate sector that creates the overheating, which attracts international hot money, private or public savings. And this economic heat would eventually transfer to stock market and boost it. At this moment, the empirical studies on Chinese real estate and stock index indicate that property heat would boost stock index.
2.6 Supply and Demand
We have discussed the different channels that might affect the housing price. We return to the original point where the price is basically reflects the demand and supply. On the demand side, when the interest rate is lower, the mortgage rate will be lower as well. With a lower mortgage cost and lower threshold, more people are accessible for owning a house. In the short term, at least, the house price will go up. But in the long term, things might be more complicated because there will be people who are actually not financially qualified for owning a house while they purchase the house anyway and increase the risk of default to the whole system. On the consumers’ confidence perspective, given that housing is a long term asset and costly, expectations for the future value of their housing are vital in household consumptions decisions. Household expectations reflect consumer confidence and uncertainty of the market. When the interest is low, it gives a signal that the economy will be growing fast in the coming years and consumers starts to buy house. This in return will boost the house price. As for the M2 money supply, it relates to inflation. When the M2 supply is getting higher, people expect a higher
inflation in the near future and they hedge the risk by owning a house.
So far, we have focused only on the effect from demand side. The supply side for housing market also plays an important when responding to different monetary policies. The interest rate plays a critical role in housing supply, interest rate channel that influences construction cost. When the interest rate is low, the construction cost will be lower and the demand will be higher. With a higher demand, thus higher price and a lower construction cost, the supply of housing will go up. On the other hand, when interest rate goes high, the supply side receives the signal of a possible economic overheating from central bank and starts to haul their housing construction.
2.7 Alternative factor: psychological and emotional
Shiller (2007) studied the housing bubble in the late 1990s, and he suggests that this boom is driven largely by extravagant expectations for future price increases. He also examined the housing market boom of 1950 where there are fears of war or terrorism and the boom of the 1970s where there are fears of environmental destruction. Those emotions would have an on the decision whether or not to buy a house. He points out that monetary policy does not come out as central in the case studies examined. The monetary policy is highly effective in the short term, but purchasing a house is a long-term decision. And the opinions about the long-term are hard to quantify. This all leads to people’s expectations. By observing the unpredictable speculative housing market of London in 2005, he suggests that the psychological expectations appear to be a major factor in explaining the extreme momentum of house price surge. Buyers who believe that home prices will continue to grow because they perceive prices as going up in general and this attitude is not easy to change.
2.8 Summary
In conclusion, we believe that monetary policy will indeed affect the housing price development, but only in short-term. On the demand side, the low interest rate provides sufficient easy credits
for the market. This will cause a boom in the housing demand and create a price surge. From the supply side, the price movement boost supply side of housing market and lower interest rate further encourages more supply of housing because of lower construction cost. There are also a lot of empirical studies suggest that there is a link between monetary policy (mostly interest rate) and house price. M2 money supply serves an alternative variable for monetary shock because the amount of M2 money supply is directly the credit availability in the market, although not as important as interest rate. Stock index is not a monetary shock, but we still compare it to house price because we want to see the linkage of financial market and housing market for further reference.
The regional effects of monetary policy on housing price are also given, however, quite limited. The activity level will determine the sensitivity level of house price. In this sense, we would tell that in big cities, the house price would react more actively comparing to that in small cities, due to possible market speculation. There are also studies suggest that the composition of industry in a city would have an impact on the sensitivity of monetary policy shocks. This might give some insight when there are two cities that have the same size, e.g. Beijing and Shanghai, looking at its industry composition might be a good idea to explain the different sensitivities of monetary policy shocks.
There are also possibilities that the monetary policy has nothing to do with housing price development. The main argument is that monetary shocks are only short term while purchasing a house is a long-term decision. There is also enough evidence shows that house price movement is quite unpredictable and largely based on the people’s expectations instead of some monetary policy shocks.
III. Background
3.1 Monetary Policy Tools for Central Bank
10The People’s Bank of China (PBC) is the central bank of China. Similar to the central banks in other countries, the monetary policy in China is to maintain the financial stability and boost economic development. The real estate contributes quite amount to the Chinese economy and has the tendency to grow bigger. Similar to Federal Reserve in the US, there are 3 monetary policy tools that PBC can use to adjust the real estate sector: open market operations, interest rate policy, and reserve requirements.
3.1.1 Open market operations (OMO)
This tool involves the purchase and sale of financial instruments (government bonds) by the central bank. The purchase of government bonds indicates a tight monetary policy while sale of bonds indicates an expansion on money base. We use M2 supply as an indicator for OMO.
3.1.2 Interest rate
Interest rate involves two meaning: First of all, the Central Bank charges a certain interest rate to depository institutions (e.g. commercial banks) on deposits and loans. These commercial banks set their interest rate close to the benchmark; secondly, the floating range around the benchmark is already determined and supervised by the Central Bank. Thus the long-term mortgage rate offered by the commercial banks is often close to the benchmark interest rate by Central Bank. Interest rate by Central Bank affects the real estate investment as well as its price growth. One thing to mention is that the benchmark interest rate set by PBC is not as frequent as that in the US if we check the benchmark rates set by Federal Reserve.
3.1.3 Reserve requirements
This is the proportion of deposits that banks must have as reserves. The reserve requirements tool is used to manipulate the aggregate credit and money supply, also the credit structure 10
Based on Geiger (2006) Monetary Policy in China (1994-2004) Targets, Instruments and their effectiveness & Monetary Policy Tools, The People’s Bank of China,
http://www.pbc.gov.cn/publish/zhengcehuobisi/360/index.html 12
overall. In order to support small and medium enterprises (SMEs), the reserve requirement for regional/rural banks is quite easy, resulting expanded money supply while the requirement tends to be strict for big banks to prevent overflowing of money into the gigantic state-owned enterprises. The current situation is that due to the privileges that state-owned enterprises enjoy, the money still keeps flowing in and later on converts into real estate property in the hands of state-owned companies. Due to the complexity, here again, we only use M2 supply as independent variables.
3.2 Regional differences
3.2.1 GeneralIn legal sense, according to the “Property Laws, People’s Republic of China”, the Chinese government ultimately owns the land and people who buy the house only get the rights of use for 70 years, although the extension after expiration is automatic.11 This is different from
western countries where most of them, the property is often imminent, with required property tax payment every year. This difference in legal prospective would cause some different price movement patterns, and further research might need to be done in this area.
Due to the imbalanced development of regions, the housing price growth differs a lot in different cities. With the national level of interest rate, money supply and any other nation-wide monetary policy, the housing price reacts differently. Our assumption is that there is one single nation-wide monetary policy, however the housing price responds to the monetary policy quite differently. We estimate that the beta of housing price would be quite different according to different regions.
Figure 2: major cities in China
11 Art. 149, The Property Law of the People’s Republic of China
13
12
3.2.2 City Type: Big city, medium city and small city:
Due to China’s large population and area, the administrative divisions of China have been developed. In our case, we apply a more commonly used city tier system which is based on population and GDP weight. The “35 cities” count for 40% of the national GDP while among them, the Tier 1 count 15% of the national GDP. Here I give some examples of how the city tier works: (i) Beijing is a municipality-level city with an equivalent status to a province; It meets up the Tier 1 standard. In total 5 cities belong to Tier 1. (ii) Nanjing is the capital of Jiangsu province and is one of 15 sub provincial cities, which have much greater autonomy and higher status than prefecture-level cities; in total, there are 10 cities belong to Tier 2. (iii) In Tier 3, there cities, e.g. Fuzhou that are capitals of provinces but can’t match up in terms of population and GDP weight and there are also cities that are so called “special economic zone”, e.g. Xiamen. The rest 35 cities that are not on list automatically fall into Tier 3 category.
12 Google Map
14
Figure 3: City tier system
Classification of cities
City %GD P Population( 2009),m %GDP Populatio n(2009), m %G DP Population (2009), mClass-1 Class-2 Class-3
Shanghai 4.38 14.01 Chongqing 1.92 32.8 Jinan 0.98 6.0 Beijing 3.48 12.46 Hangzhou 1.50 6.8 Zhengzhou 0.97 7.3 Guangzhou 2.68 7.95 Qingdao 1.44 7.6 Harbin 0.96 9.9 Shenzhen 2.41 2.46 Wuhan 1.34 8.4 Shijiazhuang 0.91 9.8 Tianjin 2.20 9.80 Chengdu 1.32 11.4 Changchun 0.84 7.6
Dalian 1.30 5.8 Xi’an 0.80 7.8 Shenyang 1.28 7.2 Fuzhou 0.74 6.4 Nanjing 1.24 6.3 Hefei 0.62 4.9 Ningbo 1.24 5.7 Nanchang 0.54 5.0 Changsha 1.10 6.5 Kunming 0.53 5.3 Hohhot 0.48 2.3 Xiamen 0.48 1.8 Taiyuan 0.45 3.7 Nanning 0.44 7.0 Urumqi 0.32 2.4 Lanzhou 0.27 3.2 Guiyang 0.27 3.7 Yinchuan 0.17 1.6 13
3.2.3 Location-Price determination: Empirical evidence from China
Zhang and Tian (2010) study house price in 35 major cities between 1995 and 2006, concluding that Chinese property market is a segmented market. The local economic feature (number of firms, sector, etc.) is an important factor. Deng, Gyourko, and Wu (2010) observe the land auction price and construction cost in 35 major cities. They build a model for land 13
Sources: NBS, own classification, Econnote, Societe Generale, April 2013, original data can be found in appendix Picture
15
supply and applied hedonic model for housing price. They find out that house prices are driven by the land market factor rather than by construction costs. Zheng, Kahn, and Liu (2009) estimate a hedonic house price regression for 35 major cities and find significant location effects that would determine the housing price.
3.2.4 Regional house price dynamics
After a downturn of global economic recession in 2008, the house price starts to soar again. In 2008 house prices in the 35 analyzed cities rose by 17% and by 22% in 2010. On average house price growth in the 35 cities was 8.3% faster in the five years between 2006 and 2010 compared to the previous five year period. The development of the price-rent ratio is of particular interest for investors, as a rising ratio suggests that an increasing share of an investor’s expected return stems from expected capital appreciation, rather than from current rental returns. All the 35 cities are experiencing a continuous growth on price-rent ratio while the Tier 1 cities showing some slowdown.
3.2.5 Affordability
From an international prospective, the affordability level in China is definite the lowest comparing to the western countries. The ratio of house prices and disposable income per capita was about 6 for Germany and 8.3 for Spain. For China the figure was as high as 48, for Beijing even 75. The household formation slightly affects the data where the China’s household has 2.9 persons while in Germany it has a bit more than 2. In terms of dynamics, even if disposable income for a Chinese household continues to grow, it would still take 20 years to catch up to the level of Germany.14
3.2. 6 Culture pattern and regulations
There is a popular belief that if a man or his family does not own a property, he will have difficulties finding a bride. There is an article in “the Economist” describe this pheromone in detail.15 In the article, it quotes a recent survey from Horizon China, a Beijing-based
14
DB Research China’s housing markets, April 28, 2011
15 Married to the mortgage, Women and the property market, the Economist, Jul 13th 2013
16
market-research firm, that 3three-quarters of women consider owning a home is very important and after marriage, more women are trying to register their name to the property wealth. This favor for household ownership certainly helps explain the high price in property market.
Due to the limit land resources, the house prices, especially in big cities are climbing high. Worrying about the potential bubbles caused by overheating, the local and central government release several regulations to tighten the house market. However, the government itself belongs to the vested interest group. The tax income obtained by selling lands is tremendous. The regulation from government never meant to depress the real estate, but just a little cool down. From time to time, it would ease the credit and encourage people to buy house. In the Appendix 1, you can find the regulations for tighten and ease the house market, and in Appendix 2, there provides restrictions/bans in terms of purchasing a house in selected cities.
IV. Data
4.1 General
Many studies have been done on investigating relationship between house price and other factors such as rents, income, CPI, etc. Eddie C.M. Hui & Shen Yue (2006) studied the correlation between house price and market fundamentals (CPI and income) to test if the house price is developing in a rational sense in Beijing, Shanghai and Hong Kong during the past decades. Quite a few of the researches mainly used the Vector Autoregressive (VARs), adopting the co-integration, granger causality test. However, for the simplicity and capability, the least square regression method will be applied in my paper.
All the data we need is monthly time series data of all house prices across the nation, interest rate, M2 Money supply, CPI index, and stock price index. The dependent variable is house price and independent variables are interest rate, M2 money supply, and stock price index. In order to test the monetary policy shocks on regional effects, we also need to create city type dummies.
The data are from National Statistics of Bureau of China, People’s Bank of China, and The range of research data runs from Jan. 2011 to Dec. 2013 which is three-year long. The main reason I choose the year 2011 is the data availability. Before 2010, the national-wide data for monthly house price is incomplete. Nevertheless, with monthly house prices in 70 cities, and if each price in a specific city in a specific month counts as one observation, we have actually 70*36 = 2520 observations which is quite decent. The results generated from all those regressions should be reliable. In Section 5.1, we restructure the house price of 70 cites into aggregate level under three city tiers. In Section 5.2, we restructure the data with 32 indices representing 32 major cities and dropping the rest 38 cites that too small in terms of GDP weight and population.
4.2 Variables
House prices (House price index): Monthly residential building selling prices in 70 cities. For
study of regional differences, I convert those observations into 3 tiers: Tier 1 (Big cities), Tier 2 (medium cities) and Tier 3 (small cities) according to the city tier system. What we get is monthly time series data “big city monthly house price”, “medium city monthly house price” and “small city monthly house price data”. Therefore, it is a proxy for the aggregate housing price level. The price index is “fixed based index number” and the year it based on is 2010.
Real Interest rate (interest rate): In China, most household buyers are borrowing money from
commercial banks. The interest rate is determined by the central bank and changes every few months with a period of 3-year long sample size. The biggest doubt is that due to its low frequency in changing and short period of time, the sample size of interest rate is very small. This might bring the results inaccurate when concerning interest rate.
M2 Money supply (M2supply): M2 is definitely an important factor in monetary policy system.
Chinese real estate market is considered to be investment-oriented which always attracts more money flow into it. The unit is in 100 mln RMB. This severs as a good alternative to indicator monetary policy shocks.
Stock price index (stockSH): the real estate is considered to be a good investment and a
better inflation hedge than financial assets. The unique rate of return and risk profile identifies itself from real estate market. We believe those are two independent asset markets. We add stock price as a variable to test the interaction with house price. Monthly Shanghai A-share stock price is used as stock market asset price.
City tier dummy: City tier dummies are created in order to test the monetary policy shocks on
regional property markets. The dummies are “Big city dummy” and “medium city”. In order to avoid dummy trap, the third one for small cities is not needed.
Miscellaneous: Different interaction terms are created. Inter1 (rate*big) means the interest
rate * big city dummy; Inter2 (rate*medium) means the interest rate * medium city dummy; Inter3 (stock*big) means the stock index * big city dummy; Inter4 (stock*medium) means the stock index * medium city dummy.
4.3 House price trend
4.3.1 Aggregate level: The house price index is a fixed index with a base year of 2010. The
data in Figure 4 and Figure 5 shows the country’s house market is starting to recover after the economic downturn during 2008-2010. For the “Big city aggregate house price”, it stayed at a static level with 102 point between 2011 and 2012. Starting from January, 2013, the house price in big cities started to surge up quickly with an average 3% month growth rate. The overall trend for house price in big cities was going up in a stable manner. Overall, the house price in big cities has increased by 20% comparing to level of 2010. The stable trait of price movement curve shows that house price in big city is not very sensitive to external factors. For the house price trend of medium city, the chart gives a much more interesting result. During the year of 2011-2012, the price movement curve was quite similar to those of big city and small city, with a stable movement around 102. However, after 2012, the price began to surged
and fell down frequently. The spikes of price movement show that house prices in medium city were not in a stable level, especially, during year of 2013. Except for the period of September, 2012 and January 2013, the price enjoys a temporary stable level, moving upwards slowly. The amplitude of price is quite enormous comparing to those in big cities and small cities. During the year of 2013, it reaches its price peak 4 times with each time falling back to 110 point afterwards. Overall, the house price has increased by 25% comparing to the price level in 2010. The movement curve spikes show that house price in medium city is quite sensitive to external factors. This might imply that the national monetary policy shock will have a bigger impact on house price in medium city. However later on in Section 6.2.2 where we apply statistically test, we find out that those spikes actually couldn’t tell us anything because most of the betas are significant. Later on, we have to shed light on the regulation which might have a relationship with house price instead of monetary policy shocks.
The trend in small city was even more stable comparing to the other two. The overall price has increased only by 12% during the past 3 years. During the year of 2011 and 2012, the price stables around 105 point level with hardly any change. In 2013, the price went up by a minor level.
Figure 4: house price trend
4.3.2 Trend of individual city
Starting from 2011 the house prices in 70 cities are going upward along with the Chinese 90,00 95,00 100,00 105,00 110,00 115,00 120,00 125,00 130,00 ja n-1 1 m ei -11 se p-1 1 ja n-1 2 m ei -12 se p-1 2 ja n-1 3 m ei -13 se p-1 3 Big city medium city small city 20
economic recovery. For the cities like Hangzhou, Ningbo and Nanjing, their house price were quite volatile and especially during mid-2012, the simultaneous downturn in house price among those specific cities are quite striking, giving a sign of housing price bubbles. Although the house market is a segmented one, the trend curve shows that the house price is positively co-variant and generally moves towards the same direction. When there is a real estate boom occurring in big cities, it house price would follow the booming trend as well. During the year 2011-2013, we can see that the house market is on a boom.
Figure 5: National average house price trend in 70 cities
As the property market boom continues to boom, the average house price for second-hand residential in Beijing is 40k RMB per m2 in 2013. The average house price for second-hand residential in Shanghai reaches 30k in 2013.16 In Section 3.2.5, we analyzed the alarming
house price-income ratio in China already, and this continuous boom in house market really alerts policy makers’ nerves. Governments make special restrictions and bans, especially for big cities to tighten the unbounded property market. Here are some examples: “in Beijing, one household is only allowed to buy one new apartment; in Shanghai, households, local or from outside the city are allowed to purchase only one new home; in Hangzhou, households – both local residents and non-residents, are only allowed to buy one additional home; in Guangzhou, each local household, permanent resident or non-permanent resident who has resided in the 16 Source: Fangjiawang, a private property market agency & data center, http://bj.fangjia.com/zoushi/
90,0 95,0 100,0 105,0 110,0 115,0 120,0 125,0 130,0 ja n-1 1 m ei -11 se p-1 1 ja n-1 2 m ei -12 se p-1 2 ja n-1 3 m ei -13 se p-1 3
National average housing price
National average housing price
21
city for over a year, can only buy one new home. Non-residents are not allowed to purchase homes in the city.”17
Figure 6: House price trend in selected cities:
4.3.3 Regulation impact
The Chinese government has recognized the possibility of housing bubbles and has reacted quickly implementing regulations to control the overheating. The main measures are including restrictions on purchase from non-locals, setting up a cap limit (usually 2) for the amount of house one person could own, tax on transaction, etc. Besides the central government,18 the
main regulating agents are the municipal and provincial governments. The local governments guided by the national regulation, implementing their own regulations. However, the problem is that local governments and officials have conflicting interests, i.e. they benefit from booming house markets. Thus, the implementation for tightening the house market is often delayed. This makes the whole issue far more complicated. If we look at the regulations during 2007-2011, the trend of the regulation is to tightening the house market, especially for big cities; relaxing and tightening for medium cities.19 Here we would like to examine whether the regulation have
a direct impact on house price or not. Due to its complex nature, it seems to be not easy. In Home
17
Source: HIS Global insight, IMF, details can be found in Appendix 1
18
The departments includes Ministry of Land and Resources, the state Administration of Taxation, and the Ministry of Housing and Urban-rural construction
19
In the Appendix regulation 1, you can find the selected regulations, for the complete version of it, please refer it to http://www.dbresearch.com/PROD/DBR_INTERNET_EN-PROD/PROD0000000000272841.PDF 90 95 100 105 110 115 120 125 130 ja n-1 1 ap r-1 1 ju l-1 1 okt -1 1 ja n-1 2 ap r-1 2 ju l-1 2 okt -1 2 ja n-1 3 ap r-1 3 ju l-1 3 okt -1 3 Beijing Shanghai Hangzhou Fuzhou Zhengzhou Chongqing 22
purchase restrictions/bans for selected cities in detail20, we find an interesting policy restriction that
imposes on 3 cities: Fuzhou, Xiamen, and Ningbo. This regulation is “One household is only allowed to buy one new home during the period from 1 October to 31 December”. The first instinct tells us it is normal restriction with different ways. It also hints that this kind of regulation might have some impact on the micro-behavior of house price trend. With a period of 3 months each year within 3 cities, the sample size is big enough to make an observation where we simple check if there is any price spike during the months where the purchasing is allowed and compare it to other cities. The figure below shows the result. We can see there is a slightly a bit price spike during the months of October till December, though it is not that obvious. Here we can conclude that the regulation where purchase can only occurred in certain month is a restriction in general, but also would have an impact on the micro-behavior of house price.
Figure 7: house price during “purchase month” in Xiamen, Fuzhou, and Ningbo
4.4 Explanatory variables
M2 money supply: Money supply was once an important monetary tool for many countries
over the world. However, currently, most central banks of developed countries have chosen interest rate as the most important instrument over the money supply. It is interesting to mention that Central bank of China still stick to money supply as the most important instrument 20 Appendix regulation 2 90 95 100 105 110 115 120 125 130 ja n-1 1 ap r-1 1 ju l-1 1 okt -1 1 ja n-1 2 ap r-1 2 ju l-1 2 okt -1 2 ja n-1 3 ap r-1 3 ju l-1 3 okt -1 3 Xiamen Ningbo Fuzhou 23
over the interest rate rule. Figure 9 shows the M2 money supply is growing on a monthly base. The rapid growth of the money supply shows that market liquidity is improving and a sign for the government turning towards a loosen monetary policy. Judging from the growth of M2 money supply, the overall liquidity is relatively eased during the year 2011-2013, and this will stimulate the aggregate demand. The real estate property will definitely absorb the majority of credits. This will result a growth on house price. From this figure we can see that a rise in M2 money supply has caused an increase in house price. M2 money supply influences the house price significantly and persistently, which is in line with our previous analysis in Section 2.4. Figure 8: M2 supply & house price comparison
Interest rate: As mentioned in the Section 3.1, the interest rate is considered to be one of the
most important monetary policy tools by the Central Bank. The Figure 10 displays the interest 0,00 200.000,00 400.000,00 600.000,00 800.000,00 1.000.000,00 1.200.000,00 ja n-1 1 ap r-1 1 ju l-1 1 okt -1 1 ja n-1 2 ap r-1 2 ju l-1 2 okt -1 2 ja n-1 3 ap r-1 3 ju l-1 3 okt -1 3
M2 supply (in 100 mln RMB)
0,00 20,00 40,00 60,00 80,00 100,00 120,00 140,00 160,00 180,00 200,00 ja n-1 1 m ei -11 se p-1 1 ja n-1 2 m ei -12 se p-1 2 ja n-1 3 m ei -13 se p-1 3 Big city medium city small city M2 money base Year2010 24rate shock on the house prices. The red curve represents the house price growth index with 2010 base year, while the blue line represents the interest movement trend. The problem with this comparison is that the adjustment for interest rate occurs only a few times per year while the house price changes every month. Due to the lack of longer period data, the result is not significant in statistics. We can hardly conclude anything whether the interest rate is effective on controlling the house price or not.
Figure 9: Interest rate & house price change
Stock index: From Figure 8, we can observe that stock index has negative correlation with
house prices regardless of the city type. During the past 3 years, the trend of stock index shows a turning down while the house price is heating up. This is not in line with our previous study in Section 2.5. Our original academic guess was that correlation between stock market and house market is positive. From the macroeconomic prospective, they go hand in hand where they have positive mutual influence on each other. Figure 8 plots a different story the correlation is negative. We can only explain this phenomenon by treating the two markets as two types of investment goods. And they are not complements, but rather substitutes to each other. When the stock index is not performing well, real estate market would attract more
0 5 10 15 20 25 1 7 13192531 interest rate % Big city change % 0 10 20 30 1 7 13 19 25 31 interest rate % Mediu m city change 0 5 10 15 1 7 13 19 25 31 interest rate % Small city change 25
capital which further deteriorates the stock market. We consider this type of investment behavior irrational. When the stock index is performing badly, it is already a sign of economic bubble. Under a potential economic bubble, the real estate heat must be sort of bubble as well.
Figure 10: Stock index & house price
Descriptive data
Table 1: Descriptive statistics of the variables in China, January 2011-December 201321
Variable Observations Mean Std. Dev. Min Max
Big City 36 107.0683 6.36834 102.16 121.26 Medium City 36 108.8711 6.699064 101.23 124.79 Small City 36 105.7094 2.989423 103.23 112.97 M2supplyBillion 36 915104.4 116568.1 733884.8 1106509 Interest Rate 36 6.198333 0.2671169 5.81 6.56 Stock Index 36 2509.846 326.7754 2092.87 3212.22
The aggregateed observations are 36 for each variable. The first row is time where it is a noun variable. The next 3 rows are for house prices in big, medium and small cities respectively. The 21 The original output can be found in appendix Stata-Output
0,00 20,00 40,00 60,00 80,00 100,00 120,00 140,00 ja n-1 1 m ei -11 se p-1 1 ja n-1 2 m ei -12 se p-1 2 ja n-1 3 m ei -13 se p-1 3 Big city medium city small city stock index base 2010 26
average house price index is sort of similar which floats around 107. The standard deviations for house prices in big cities and medium cities are 6.36834 and 6.699064 which seem close. The standard deviation for small city is only 2.98 which is significantly lower comparing to those in big and medium cities. All the 3 house price index has a similar lowest point of 102; for the maximum point of house price in small cities, it reaches only to 112.97 while for medium and big city, they reach on the level of about 122. The house price in big cities and medium cities enjoy a signficant growth while the price in small cities does not; however, the price movement is much more volatile in big and medium cities than that in small cities.
M2 money supply has a mean of 915104.4 in mln RMB during 2011-2013. It reaches its max in the late 2013, with 1106509 in mln RMB. The minimum point is about where the data starts, in early 2011, with the level of 1106509 in mln RMB. The interest rate floats around 6%. The standard deviation is 0.2671169. It means that change occurs every few months with a minor adjustment. The stock index has a mean of 2509.846. The peaking is 3212.22 in early Year 2011 and it goes all the way down to the level of 2092.87 in the middle of Year 2013.
Table 2: Correlation among the variables22
InBig InMedium InSmall InM2 inInterest InStock
InBig 1.00 InMedium 0.4128 1.00 InSmall 0.9815 0.5280 1.00 InM2 0.8763 -0.0197 0.7747 1.00 InInterest -0.2129 -0.1430 -0.2488 -0.1308 1.00 InStock -0.0555 -0.0373 -0.0649 -0.0341 -0.1041 1.00
22 From Stata-Output in the Appendix
27
V. Methodology
5.1 Regression for Aggregate house price level
Aggregate house price level with big, medium and small city: Model (1)
House Price Index (t, i) = c + 𝛽𝛽1* Monetary policy related index + 𝛽𝛽2* Big-city-dummy +𝛽𝛽3 *
medium-city-dummy + 𝛽𝛽4* interaction term (mpri * bvscd) + 𝜀𝜀𝑡𝑡,𝑖𝑖23
The “monetary policy related index” including “bank loan rate”, “change in money M2 supply”, “stock index” and “CPI index”. For the dummy categories, there will be 2 dummies: 1/0 big city + 1/0 medium city. The “small-city” dummy is left out on purpose in order to prevent the dummy trap.𝛽𝛽2 is the beta for “Big-city” dummy and will give a clue about the price movement due to
its large city size under a monetary policy shock. The 𝛽𝛽3 is the beta for “medium-city” dummy
and will explain the price movement due to its medium city size under a monetary policy shock. The interaction term 𝛽𝛽4 will tell about the differences in terms of the effect of monetary policy
on house price index for both types of cities. One result might be that neither of the individual variables (monetary policy and city type dummy) has much effect on the housing price index, but a combination of the two does. This depends on whether the 𝛽𝛽4 would be significant or
not. If there is a significant result on 𝛽𝛽4, this will indicate that there might be a bubble growing
in the larger cities as a result of monetary expansion.
5.2 Regression for House price level of each city (32 selected cities)
The model would contain much more information if we didn’t aggregate the house prices but present them all. The original data has the monthly house price of 70 cities over the past 3 years. The total observation is 2520. However, if we present them all, the table would be very 23
The subscript (i) means that the variables vary per city and subscript (t) means for variation in time, (t, i) stands for variables that vary for both cities and time
28
long and unnecessary because 50% from the city groups are belonging to small city and the GDP weight as well population from those small cities are negligible. The common used city tier system tells us that it consists of 5 big cities, 10 medium cities and 15+ small cities, about 30 cities in total. After compromising those factors, we design a regression with house price indices of 32 selected cities, and the rest 38 cities are dropped because their GDP and population weight is not that significant comparing to the whole. The regression is quite similar to the one with aggregate price level. The only difference is that dummy city type is substitute with 31 dummies each representing a city. And we have more interaction terms because we have more dummies. The total number of observation is 1152. The form goes the following:
House price level with 32 selected city: model (2)
32 selected cities House Price Index (t, i) = c + 𝛽𝛽1* M2 (t) + 𝛽𝛽2* interest rate (t) +𝛽𝛽3 * stockSH (t)
+ 𝛽𝛽4* Beijing + 𝛽𝛽5*Shanghai + 𝛽𝛽6*Nanjing + 𝛽𝛽7*Fuzhou +... + interaction term 1
(interestrate * Beijing) + interaction term 2 (interestrate*Nanjing)...+ 𝜀𝜀𝑡𝑡,𝑖𝑖
VI. Results & interpretation
6.1 Results from House price level with 32 selected city: model (2)
After running the regression from Model (2), we have 1152 observations, the adj. R-Square is 0.7322 and Root MSE is 2.8533 where you can see from the table below. The number of observations is 1152 which consist of the monthly housing price in 32 selected cities over the period of 36 months. The sufficient number of observations makes our housing price trend analysis more accurate. R-square is a statistical measure of how close the data are to the fitted regression line. However, for each variables added in the regression, the R-square value would always increase, thus here we use “adjusted R-Square” where it would adjust the statistic as extra variables are included in the model. The adjusted R-Square in our model (2) is 73%, this indicates that the assumed trend-line fitted is a reasonable approximation to the actual values. In other words, 73% of the variance of housing price is due to its affecting factors which include
monetary policy, city tier, etc. Considering the complexity of the affecting factors, we are very happy that 73% of the variance can be explained by our model (2) and the results are reliable. The Root MSE is a measure of the differences between values predicted by a model or an estimator and the value observed. The value itself might be meaningless unless they are used for comparative purposes between two or more statistical models. In Model (2), Root MSE is 2.8533 while in Model (1) the Root MSE is 3.2718. In Model (3), the Root MSE is 3.0689. By comparison, the Root MSE in Model (2) is significantly smaller than that in either Model (1) or Model (2). It means in Model (2), the differences between values predicted by a model and the value actually observed is relatively smaller, and this interprets as the results from Model (2) are more reliable. And this is what we expected since we apply over 1000 observations in the model (2) while in Model (1) or Model (3), we use the average housing price on 3 city-tier in which the total number of observations is around 100.
Statistical quality estimation from model (2) Observations 1152
Adj. R-Square 0.7322
Root MSE 2.8533
Table 3: Beta for monetary related variables from Model (2)
Variables Coefficient Standard Err. Significance
M2 Money supply 0.00003 3.28e-06 Yes
Interest Rate -0.570224 0.7784265 No
Stock Index Shanghai 0.0040464 0.0010913 Yes
M2 Money supply: The coefficient of M2 supply from Model (2) is 0.00003, positive. It means
the growth of M2 supply would boost the house price in individual city at the level of beta 0.00003. The P value is shows that it is significant that proves M2 money supply is indeed one of factors that would affect house price movement. This result matches our hypothesis. From
several previous studies we know that if there is a growth on M2 supply, the money would flow into real estate and causing housing price increase. This easy-money policy encourages investors to invest more as well as consumers to purchase houses. Vice versa, if central bank targets to lower the inflation, thus implement a tightened money supply, this will cause a drop in credit supply and less credit available in the property market means less activity, eventually the house price would go down due to low activity.
Interest rate: The coefficient of interest rate from Model (2) is -0.570224 and negative. This is
in line with our hypothesis. The higher interest rate means higher cost and will discourage the house market and damp the house price. When interest rate goes up, the house price goes down; vice versa. However the P value is extremely large, this makes the result insignificant. Before checking the hypothesis, this result is flaw to some extent. The dataset range is only 3 years, and this makes the valid number of observations for interest rate below 10 because interest rate changes only every few months. The previous studies are observing data with period longer than 10 years, and the valid number of observation would be much higher and makes the result significant.
Stock Index Shanghai: The coefficient for stock index from model (2) is 0.0040464 and
positive. The P-value shows that it is significant. The positive beta is not what we expected. Nevertheless, the positive beta means the positive change in stock index would result a positive change in house price. We must keep in mind that the beta is too small. This micro movement trait is certainly interesting, but the overall trend curve tells us during the past 3 years, the stock market is performing badly while the house price is still increasing. This implies that the investment behavior in real estate might be irrational. We predict the Chinese house market is facing a bubble. The betas for different regions are indeed different in which the beta for medium city is quite big while the betas for big and small cities are small by comparison.
6.2 City Beta
City dummy comparison: Now we then compare the betas for different city dummies. The
betas for big city group are the following:
6.2.1 Big cities:
We first look at the significance level. The betas for Beijing, Shenzhen and Guangzhou are significant from model (2). We conclude from here in statistics, the relationship between house price and monetary policy exists. The betas for Shanghai and Tianjin are insignificant, thus we can’t really conclude anything from them in statistics. Besides the significance level, we also take a closer look at the beta values. Overall, the beta values for the big cities are negative and quite big. This is in line with our previous conclusion: in big cities, the house price will be sensitive to monetary policy because the house market as well capital market in big cities is generally more active. We expect more speculations in big cities. Shanghai’s beta is 0.5633 which is quite small in absolute value comparing to the betas of others. There are two possible explanations here: the first one is that due to its insignificancy in statistics, the beta value itself hardly has any meaning; the second one can be referred to Carlino, Defina (1995). They suggest that city which has a higher concentration in real estate, finance and insurance has a lower sensitivity on monetary policy response. In their example, the New York has 24.3 share of real estate, finance and insurance output which is the highest while its policy response is only -0.72 which is the lowest. In our case, Shanghai is notably the city with highest share of output in the area of real estate, finance and insurance industry. The Shanghai, resembles the New York in Carlino, Defina (1995), meaning that its sensitivity on policy response would be low. Table 4: Betas for Big cities (5 cities) from Model (2)
Variable Coefficient Standard Err. Significance
Beijing -54.75311 24.20082 Yes Shanghai 0.5633718 24.20082 No Shenzhen -83.05364 24.20082 Yes Guangzhou -81.65714 24.20082 Yes Tianjin -8.589127 24.20082 No Summary (-) *4, (+)*1 60% of them 32
significant
6.2.2 Medium cities:
After observing the betas for 10 medium cities, the significance level from model (2) overall is quite low. Shenyang and Dalian are the only two cities that have betas with statistical significance. Beta values overall various largely in terms of absolute value. There are 6 cities show negative relationship while 4 cites have positive relationship. Unfortunately, we couldn’t really conclude anything due to the fact their hardly any of betas are significant. This might also solve our doubt in Section 4.3.1 Figure 4 and Figure 8 where the price development in medium cities is quite unstable comparing to small and big cities. Nevertheless, if we take a look at Shenyang and Dalian, their betas are 75.25 and 63.29 which are quite similar. This might suggest that to some extent, medium cities do behavior in the same pattern when response to monetary policy shocks and at least for those two cities, they are quite sensitive to monetary policy shocks. The biggest difference between the betas of medium city and big cities is that they go in the opposite direction; the betas for big cities are negative while the betas for medium cities are 50% positive and 50% negative. We could partly conclude that medium cities have the same level of policy sensitivity as the big cities, but the reaction towards policy goes in the opposite way. The different development levels of housing finance and credit systems in different regions can only explain the magnitude of the betas, but not the direction of it. We have to shed light on the regulation where the house market is heavily regulated in big cities, in which most of the regulation types are for tightening the market.
Table 5: Beta for Medium cities (10 cities) from Model (2)
Variable Coefficient Standard Err. Significance
Shenyang 75.25055 24.20082 Yes
Hangzhou 21.66005 24.20082 No
Chongqing -18.58798 24.20082 No
Wuhan -31.11727 24.20082 No
Ningbo -9.310229 24.20082 No Qingdao -29.79238 24.20082 No Nanjing 34.63861 24.20082 No Changsha -37.24311 24.20082 No Chengdu -19.5023 24.20082 No Dalian 63.29955 24.20082 Yes Summary (-)*6, (+)*4 20% of them significant 6.3.3 Small cities:
In general, betas for small cities in model (2) are generally small and negatively related. This result is in line with what we expected. The small beta means less speculation and we expect less speculation in small cities where there is a monetary policy shock. Overall, 43.75% of the betas are significant meaning that in a rough sense, the relationship between monetary policy shocks and house price exists in small city group. There are two cities that need to pay attention: Fuzhou and Hefei. Both of them are statistically significant. We can refer to Appendix Regulation 3: list of cities that abandon/loosen restrictions, Fuzhou is one of the few Tier 3 cities that haven’t loosened its restrictions on purchasing houses. Although Hefei increase its housing funds support in a way to loosen the restrictions, this is far from abandon its restrictions. The regulations still imposes great pressure. It’s worth mentioning that although the big cities loosen its restriction to some extent, this only implies a little relaxing after a very long policy tightening.
Table 6: Beta for small cities (16 cities) from Model (2)
Variable Coefficient Standard Err. Significance
Shijiazhuang -14.21753 24.20082 No
Taiyuan -4.344444 0.6725198 Yes
Huhehaote -2.080556 0.6725198 Yes
Changchun 14.57386 24.20082 No Harbin 7.67579 24.20082 No Hefei 62.4971 24.20082 Yes Fuzhou 75.00358 24.20082 Yes Xiamen 0.5333333 0.6725198 No Nanchang -17.57041 24.20082 No Jinan -20.06541 24.20082 No Zhengzhou -36.70565 24.20082 No Nanning -5.1 0.6725198 Yes Haikou -7.108333 0.6725198 Yes Guiyang -2.166667 0.6725198 Yes Kunming -1.478026 24.20082 No Xian -23.30155 24.20082 No
Summary (-)*11, (+)*5 The city group with Std. Err. 0.67 consists of cities that are very small
43.75% of them significant
6.3.4 Interaction term:
6.3.4.1 Interaction 1: Interest rate * city in model (2)
We use interest rate * each individual city (5 big cities, 10 medium cities and 10 selected small cities) as interaction term. From the results of model (2), we find out that only 6 of the interaction terms are significant. This means that excluding all the other factors, we can say that interest rate as the single factor alone, would affect the house price in the following city: Hangzhou, Dalian, Shenyang, Ningbo, Fuzhou, and Hefei. To sum up, only 24% of the interaction terms tend to be significant. None of the big cities show any significant implying that there is not enough evidence to claim the relationship between interest rate and big cities. And
this is not quite in line with what we have presented in the previous section. We sense that it is a signal showing that either the interest rate in general won’t affect the house price, or due to the natural pitfall of my test design, the result itself is inaccurate.
Table 7: interaction term 1-Interest rate * City dummy from Model (2) Variable Coefficient Standard Err. Significance
RateBeijing -1.795477 2.201723 * RateShanghai -3.020777 2.201723 * RateTianjin -0.6161981 2.201723 * RateShenzhen -0.9407117 2.201723 * rateGuangzhou -1.831606 2.201723 * RateChongqing -1.134536 2.201723 * RateHangzhou 6.87739 2.201723 Yes RateQingdao 1.56695 2.201723 * RateWuhan 0.1539659 2.201723 * RateChengdu -0.7293837 2.201723 * RateDalian 5.472373 2.201723 Yes RateShenyang 5.003888 2.201723 yes RateNanjing 1.338359 2.201723 * RateNingbo 5.19275 2.201723 yes RateChengsha 0.7640006 2.201723 * RateJinan 0.1421538 2.201723 * RateZhengzhou -0.4596405 2.201723 * RateHarbin -2.061441 2.201723 * RateShijiazhuang -0.2593032 2.201723 * RateChangchun -1.580205 2.201723 * RateXian 0.4417591 2.201723 * RateFuzhou -7.962419 2.201723 yes RateHefei -7.589353 2.201723 yes 36
RateNanchang -0.9940203 2.201723 *
RateKunming 0.2209576 2.201723 *
Summary 24% of them are
significant
6.3.4.2 Interaction 2: M2 money supply * city dummy
The second interaction term we use is the “M2 money supply * city dummy”. In the previous analysis, we points out the pitfall of interest rate in our dataset-sample size too small due to the very short period of time. The M2 money supply here would serve as a good alternative as an indicator for monetary policy. In the results of model (2), the betas of M2*city dummy showed a higher percentage share of being significant. Among them, the betas from 4 out of 5 big cities are significant. This is strongly suggesting that there is indeed a relationship between M2 money supply and house price in big cities. 50% of the medium cities have the betas that are significant. This implies that M2 money supply have some sort of link with house price in medium city, but not strong enough. Only 3 out of 10 selected small cities have betas that are significant. At this step, we can further conclude that the relationship between M2 money supply and house price in small cities is quite weak. In conclusion, we find out this result is in line with our previous argument-the relationship between monetary policy and house price is strong in big cities, modest in medium cities and weak in small cities.
Table 8: Interaction 2, M2 money supply*city dummy from model (2)
Variable Coefficient Standard Err. Significance
M2Beijing 0.0000447 9.27e-06 Yes
M2Shanghai -0.0000262 9.27e-06 Yes
M2Tianjin 3.10e-06 9.27e-06 *
M2Shenzhen 0.0000603 9.27e-06 Yes
M2Guangzhou 0.0000673 9.27e-06 Yes
M2Chongqing 0.0000119 9.27e-06 *
M2Hangzhou -0.0000579 9.27e-06 Yes
M2Qingdao 1.29e-06 9.27e-06 *
M2Wuhan 0.0000181 9.27e-06 *
M2Chengdu 0.0000114 9.27e-06 *
M2Dalian -0.0000616 9.27e-06 Yes
M2Shenyang -0.000064 9.27e-06 Yes
M2Nanjing -0.000065 9.27e-06 Yes
M2Ningbo -0.0000165 9.27e-06 *
M2Chengsha 0.000025 9.27e-06 Yes
M2Jinan 7.26e-06 9.27e-06 *
M2Zhengzhou 0.0000287 9.27e-06 Yes
M2Harbin -0.0000232 9.27e-06 Yes
M2Shijiazhuang 0.0000128 9.27e-06 *
M2Changchun -0.0000237 9.27e-06 Yes
M2Xian 0.0000132 9.27e-06 *
M2Fuzhou -0.0000123 9.27e-06 *
M2Hefei -0.0000174 9.27e-06 *
M2Nanchang 0.0000141 9.27e-06 *
M2Kunming -1.45e-06 9.27e-06 *
Summary 48% of them are
significant
6.3.4.3 Interaction 3-Stock Index * city dummy
We are interested in finding out the relationship between stock index and house price. This is the link between financial market and housing market. The results are below in the table and looks quite similar to the one in M2*city dummy. Overall, 36% of the betas are significant. Among them, the betas from 4 out of 5 big cities are significant. This is strongly suggesting that
there is a strong link between stock index and house price in big cities. 4 out of the 10 medium cities have the betas that are significant. This implies that stock index have some sort of link with house price in medium city, but not strong enough. Only 1 out of 10 selected small cities have betas that are significant. This indicates that stock index has nothing to do with the house price in small cities. In summary, the results are in line with our previous theory that the link is strong in big cities, moderate in medium cities and weak in small cities
Table 9: interaction 3-stock index*city dummy: from model (2)
Variable Significance Standard Err. Significance
StockBeijing 0.0091267 0.0030866 Yes StockShanghai 0.0150055 0.0030866 Yes StockTianjin 0.0024585 0.0030866 * StockShenzhen 0.0130522 0.0030866 Yes StockGuangzhou 0.0127697 0.0030866 Yes StockChongqing 0.0046113 0.0030866 * StockHangzhou -0.0087281 0.0030866 Yes StockQingdao 0.0052834 0.0030866 * StockWuhan 0.0044269 0.0030866 * StockChengdu 0.0039189 0.0030866 * StockDalian -0.0167489 0.0030866 Yes StockShenyang -0.019155 0.0030866 Yes StockNanjing 0.0029205 0.0030866 * StockNingbo -0.0060436 0.0030866 Yes StockChengsha 0.0045481 0.0030866 * StockJinan 0.0035041 0.0030866 * StockZhengzhou 0.0055868 0.0030866 * StockHarbin 0.0086927 0.0030866 Yes StockShijiazhuang 0.0022376 0.0030866 * StockChangchun 0.0054447 0.0030866 * 39
StockXian 0.0024536 0.0030866 *
StockFuzhou -0.0054435 0.0030866 *
StockHefei 0.000079 0.0030866 *
StockNanchang 0.0043403 0.0030866 *
StockKunming -5.94e-06 0.0030866 *
Summary 36% of them are
significant
Constant 74.83367 8.566189 Yes
6.2 Robustness checks
6.4.1 Finding alterationsA robustness check is necessary in order to see if they property market system can still continue to perform despite alterations in market conditions, e.g. what would happen if we use aggregate house price to represent big city, medium city and small city; or find alterations for variables. E.g. what change would it be if we use some other alternative variables such as inter-bank lending rate instead of interest rate, stock index Shenzhen instead of stock index Shanghai, etc.?
6.4.2 Aggregate house price level with 3 city tiers
It is worth to mention that interaction terms are created in order to identify the pure monetary policy shock on the house price. For example, in the results of Model (1), the beta 𝛽𝛽6 for
Inter1(rate*big) means the impact of interest rte in big cities where keeping the effect of the other variables constant. We thus correct for differences in monetary policy and correct for unobserved differences in cities; the observed result can therefore be contributed to the impact of the monetary variable in that particular city-big city. This goes the same for other interaction terms.
