An Empirical Study on the Relationship of Interest Rate, Money. Supply and Real Estate Development Investment in China

Dept of Real Estate and Construction Management Master of Science Thesis no. 29 Div of Building and Real Estate Economics An Empirical Study on the Relationship of Interest Rate, Money Supply and Real Estate Development Investment in China Author: Main Supervisor: Co-supervisor: Yannu Zheng Hans Lind Mats Wilhelmsson Stockholm 2010 I

Master of Science Thesis Title: Author: An Empirical Study on the Relationship of Interest Rate, Money Supply and Real Estate Development Investment in China Yannu Zheng Department: Department of Real Estate and Construction Management Division of Building and Real Estate Economics Master Thesis number: 29 Main Supervisor: Hans Lind Co-supervisor: Keywords: Mats Wilhelmsson Interest rate; Money supply; Real estate development investment; China; Error-correction model Abstract The thesis applies error-correction models to examine the causal and quantitative relationship of interest rate, money supply and real estate development investment in China. First, it compares the different impact of specific interest rate (nominal annual rate and effective annual rate) and money supply (nominal and real M1, M2) on aggregated real estate development investment (nominal and real) based on monthly data from January, 1999 to December 2009. Second, it tests the different impact of specific interest rate and money supply on development investment in residential, office and commercial buildings (nominal and real) based on monthly data from January, 2003 to December 2009. The empirical results indicate that: (1) none of the interest rate is the Granger cause for any kind of real estate development investment in the short or long-run, and money supply has much more apparent impact on real estate development investment than interest rate. (2) However, II

different kinds of money supply have different impact on different kinds of real estate development investment. The causal relationship among money supply and investment in different types of real estate is much more complicated than that of money supply and aggregated real estate development investment. For instances, nominal M1 is the Granger cause for both nominal and real aggregated real estate development investment in short and long-run. While to investment in different types of real estate, although nominal M1 is still the Granger cause for nominal and real investment in office buildings and real investment in residential buildings in both short and long-run, but it is the Granger cause for nominal investment in residential and commercial buildings only in the short-run but not in the long-run. Furthermore, it is not the Granger cause for real investment in commercial buildings neither in the short nor long-run. (3) Although interest rate is not the Granger cause for any kind of real estate development investment, it could be cointegrated with some specific money supply and real estate development investment, where interest rate has negative but not significant impact on real estate development investment. However, for different kinds of money supply, either it is alone or combined with specific interest rate, it has significantly positive but different impact on different kinds of real estate development investment. For example, real money supply or it is with effective annual rate have more apparent impact on real investment in all kinds of real estate than they are on the nominal ones. But if it is nominal money supply or it with nominal annual rate, to aggregated real estate development investment, they have more apparent impact on the nominal one than on the real one. However, it is the reservation for investment in residential, office and commercial buildings as the real development investment in these three types could be impacted more by them than the nominal ones. But it should exclude nominal and real investment in office buildings, which are impacted almost the same by nominal M1 with nominal annual rate. III

Acknowledgments The author would like to acknowledge a lot of people for their selfless help to me during my master period, especially when I am doing this thesis. My first appreciation goes to my main superior Hans Lind, who dedicates himself to my research and gives me enlightening guidance and support to my thesis. I also appreciate my co-supervisor Mats Wilhelmsson a lot for his patient guidance of the research methodology. Besides, I want to express my appreciation to P Åke Gunnelin, Ruth-Aida Nahum and all the other teachers, coordinators and teacher assistants at the Department of Real Estate and Construction Management for the knowledge they passed on me and their help to me. I am grateful to KTH as it gives me the chance to study and experience life in Sweden. My eternal gratitude goes to my parents, elder sister and my boyfriend for their encouragement and unconditioned support to me. I also gratefully acknowledge all my former and present friends for their friendliness to me. Stockholm, Sweden. May 2010 Yannu Zheng IV

Abbreviations and Acronyms IR Interest Rate NAR Nominal Annual Rate EAR Effective Annual Rate MS Money Supply N_M1 Nominal M1 N_M2 Nominal M2 R_M1 Real M1 R_M2 Real M2 N_M1_03 Nominal M1 since year of 2003 N_M2_03 Nominal M2 since year of 2003 R_M1_03 Real M1 since year of 2003 R_M2_03 Real M2 since year of 2003 REDI Real Estate Development Investment AREDI Aggregated Real Estate Development Investment N_AREDI Nominal Aggregated Real Estate Development Investment R_AREDI Real Aggregated Real Estate Development Investment RDI Development Investment in Residential Buildings N_RDI Nominal Development Investment in Residential Buildings R_RDI Real Development Investment in Residential Buildings ODI Development Investment in Office Buildings N_ODI Nominal Development Investment in Office Buildings R_ODI Real Development Investment in Office Buildings CDI Development Investment in Commercial Buildings N_CDI Nominal Development Investment in Commercial Buildings R_CDI Real Development Investment in Commercial Buildings AFAI Aggregated Fixed Asset Investment GDP Gross Domestic Product CPI Consumer Price Index Inv. Investment SDRR Statutory Deposit Reserve Ratio No. Number Dev. Development Cos. Companies CHN China ECM Error-correction Model ECT Error Correction Term EG Engle-Granger DF Dickey-Fuller DW Durbin-Watson V

Table of Contents Master of Science Thesis... II Abstract... II Acknowledgments... IV Abbreviations and Acronyms... V Table of Contents... VI List of Figures... VIII List of Tables... IX Chapter 1 Introduction... 1 1.1 Background... 1 1.2 Problem Description... 2 1.3 Purpose of the Study... 3 1.4 Methodology... 3 1.5 Limitation... 4 1.6 Organization of the Study... 4 Chapter 2 Literature Review... 6 Chapter 3 History Development... 8 3.1 History Adjustments of IR and MS... 8 3.1.1 History Adjustments of IR... 8 3.1.2 History Adjustments of MS... 10 3.2 History Development of REDI in China... 12 3.3 History Development of Funds on REDI in China... 14 3.4 History Development of REDI in different types in China... 16 3.5 Conclusion... 17 Chapter 4 Empirical Studies on the Relationship of IR, MS and AREDI in China19 4.1 Methodology... 19 4.1.1 Stationary and cointegration test... 19 4.1.2 Granger causality test and Error-correction models... 20 4.2 Data... 21 4.3 Empirical results... 24 4.3.1 Stationarity and cointegration... 24 4.3.2 Granger causality test... 29 4.3.3 Error-correction Models... 31 4.4 Conclusion for the Relationship of IR, MS and AREDI... 37 Chapter 5 Empirical Studies on the Relationship of IR, MS and REDI in different types in China 39 5.1 Data... 39 5.2 Empirical results... 44 VI

5.2.1 Stationary and cointegration test... 44 5.2.3 Error-correction Models... 54 5.3 Conclusion for the Relationship of IR, MS and REDI in different types... 70 Chapter 6 Conclusion, Contribution and Policy Implications... 72 6.1 Conclusion... 72 6.2 Contribution... 74 6.3 Policy Implications... 75 References... 77 Appendix... 80 A.1 ECMs for Chapter 4... 80 A.2 ECMs for Chapter 5... 83 VII

List of Figures Figure 1: History Adjustments of Nominal Annual Rate... 9 Figure 2: A Comparison of Nominal Annual Rate and Effective Annual Rate and Their Amplitude Adjustment... 10 Figure 3: A Comparison of History Changes of Nominal and Real M1 and M2 and Their Growth Rates... 11 Figure 4: Historical Adjustments of Reserve Requirement Ratio... 12 Figure 5: The Average Sales Price of Commercial Housing in China and Its Growth Rate... 13 Figure 6: The Number of Real Estate Development Companies and Its Growth Rate in China... 13 Figure 7: A Comparison of History Growth Rate of Nominal AFAI, N_AREDI and Rate of N_AREDI to Nominal AFAI in China... 14 Figure 8: Loan Rates for Different Sources of Funds of Enterprises for REDI in China... 15 Figure 9: Different Rates of Loans on REDI to Total Loans in China... 16 Figure 10: Growth Rates of Total Loans and Domestic Loans on REDI in China... 16 Figure 11: Development Investment in different types of real estate in China... 17 Figure 12: Rates of REDI in different types to N_AREDI in China... 17 Figure 13: Growth Rates of IR and AREDI in China... 22 Figure 14: Growth Rates of MS and AREDI in China... 23 Figure 15: Growth Rates of IR and REDI in different types in China... 40 Figure 16: Growth Rates of MS and REDI in different types in China... 41 VIII

List of Tables Table 1: Correlations among nominal and real IR, MS and AREDI... 24 Table 2: Results of stationarity by DF tests on IR, MS and AREDI... 25 Table 3: Results of Cointegration by EG test on IR, MS and AREDI... 26 Table 4: Results of Cointegration when IR and MS are combined together as independent variables and AREDI as dependent variables by the EG test... 27 Table 5: Results of normalized cointegrating equations by EG test on IR, MS and AREDI... 28 Table 6: DF test of ECT for IR, MS and AREDI... 29 Table 7: Short-run Granger causality (Prob>F) for IR, MS and AREDI... 30 Table 8: Long-run Granger causality (t-values concerning estimate of ECT) for IR, MS and AREDI... 31 Table 9: Quantitative comparison of the different impact of specific IR, MS on AREDI by Adj R 2... 37 Table 10: Correlations of IR, MS and REDI in different types... 43 Table 11: Results of stationarity by DF tests on IR, MS and REDI in different types... 44 Table 12: Results of Cointegration by EG Test on IR, MS and REDI in different types... 47 Table 13: Results of Cointegration when combined IR and MS are as independent variables and REDI in different types as dependent variables by EG test... 48 Table 14: Short-run Granger causality (Prob>F) for IR, MS and REDI in different types... 50 Table 15: Long-run Granger causality (t-values concerning estimate of ECT)... 53 Table 16: Quantitative comparison of the different impact of specific IR, MS on REDI in different types by Adj R 2... 69 IX

List of Tables in Appendix Table A. 1: The impact of N_M1 on N_AREDI and R_AREDI... 80 Table A. 2: The impact of combined NAR and N_M1 on N_AREDI and R_AREDI... 80 Table A. 3: The impact of R_M1 on N_AREDI and R_AREDI... 81 Table A. 4: The impact of R_M2 or combined EAR and R_M2 on R_AREDI... 81 Table A. 5: The impact of combined EAR and R_M1 on N_AREDI and R_AREDI... 82 Table A. 6: The impact of N_M1_03 on N_RDI, N_ODI and N_CDI... 83 Table A. 7: The impact of N_M1_03 on R_RDI, R_ODI... 84 Table A. 8: The impact of N_M2_03 on N_RDI, N_ODI... 84 Table A. 9: The impact of N_M2_03 on R_RDI, R_ODI... 85 Table A. 10: The impact of R_M1_03 on N_RDI, N_ODI and N_CDI... 86 Table A. 11: The impact of R_M1_03 on R_RDI, R_ODI and R_CDI... 86 Table A. 12: The impact of R_M2_03 on N_RDI, N_ODI... 87 Table A. 13: The impact of R_M2_03 on R_RDI, R_ODI... 88 Table A. 14: The impact of combined NAR_03 and N_M1_03 on N_RDI, N_ODI and N_CDI... 88 Table A. 15: The impact of combined NAR_03 and N_M1_03 on R_RDI, R_ODI... 89 Table A. 16: The impact of combined NAR_03 and N_M2_03 on N_RDI, N_ODI and R_RDI... 90 Table A. 17: The impact of combined EAR_03 and R_M1_03 on N_RDI, N_ODI and N_CDI... 90 Table A. 18: The impact of combined EAR_03 and R_M1_03 on R_RDI, R_ODI and R_CDI... 91 Table A. 19: The impact of combined EAR_03 and R_M2_03 on N_RDI, N_ODI... 92 Table A. 20: The impact of combined EAR_03 and R_M2_03 on R_RDI, R_ODI... 93 X

Chapter 1 Introduction 1.1 Background In the past decade, China has achieved a dramatic development in the real estate industry. Especially the real estate development investment (REDI), as there has been large housing and real estate financial system reform since 1998. According to the records in China Statistical Yearbook, from 1998 to 2008, the aggregated REDI (AREDI) increased from 361 billion to 3120 billion, which increased more than 275.4 billion every year on average. Its average annual growth rate was up to 22.6%, more than that of aggregated fixed asset investment (AFAI), which was only 19.5% as well as the rate of AREDI to AFAI increased from 12.7% to 18.1%. Here, REDI refers to investment by i.e. real estate development companies on construction of buildings, such as residential buildings, office buildings, commercial buildings, factory buildings, warehouses, hotels, the complementary service facilities and land development projects (National Bureau of Statistics of China, 2009). Among all of constructions, residential, office and commercial buildings are the main three ones, which occupy about 82.8% to 87.1% of the AREDI during 1998 to 2009. However, the development among these three main types is very unbalanced, i.e. during 1998 to 2009, the aggregated development investment in residential buildings (RDI) increased from 208 billion to 2562 billion, whose rate to AREDI increased from 57.6% to 70.2%; however, the aggregated development investment in office buildings (ODI) only increased from 43 billion to 138 billion, while its rate to AREDI decreased from 12.0% to 3.8%; the situation for commercial buildings is a little better than office buildings, which aggregated development investment (CDI) increased from 48 billion to 417 billion with the rate to AREDI just decreasing from 13.2% to 11.5%. But whatever, their average annual absolute increments are huge enough, which are up to 196.1 billion, 7.9 billion and 30.8 billion for residential, office and commercial buildings respectively. Along with the dramatic development of REDI, it becomes increasingly dependent on bank credit. From 1998 to 2009, the domestic loans for AREDI increased from 105 billion to 1129 billion and its average growth rate was up to 24.9%, higher than the average growth rate of AREDI (22.6%). In addition, as estimated, about 50% to 60% of the total investment in real estate development is from the commercial banks, which makes the REDI easily affected by the change of uniform monetary policy by changing its cost. That s because the commercial banks and other related financial institutes, who 1

are the main sources of the funds for REDI, are under the charge of uniform monetary policy. For example, from 2004 to 2008, in order to control the unhealthy development of real estate, especially to protect from a REDI bubble, the Central Bank and government in China adjusted the monetary policy very frequently (see Figure 1 and Figure 4). And a remarkable feature of China's monetary policy at present is the direct control of interest rate (IR) and indirect regulation of money supply (MS), which are so-called "Dual Regulation" (Tang, Li and Xu, 2008). 1.2 Problem Description There were only a few people who studied the relationship of IR, MS and AREDI in China previously, however, their results were different. That s because they based it on different variables, data and methodologies (Zhou, 2007; Cheng, Luo and Ren, 2007; Liu, Li and Liu 2008; Gu, 2008; Lu, Tian and Wu et al, 2008). For example, Zhou (2007), Gu (2008) and Liu, Li and Liu (2008) got the results that both of IR and MS have significant impact on AREDI while MS had more apparent impact than IR by using quarterly data and simple linear regressions; but Lu, Tian and Wu (2008) and Tang, Li and Xu (2008) said that MS had some effect on AREDI while IR was not the effective instrument of controlling AREDI based on monthly data and Granger analysis. Furthermore, when different people did the research on the impact of nominal and real IR and MS on AREDI, some used nominal IR and MS (Nie and Liu, 2005; Qin, 2006; Zhou, 2007; Gu, 2008; Liu Li and Liu; 2008), while some others used real ones (Luo and Jin, 2009) just by their own assumptions but without any strict and convincing demonstration. Additionally, when they researched the impact of MS on AREDI, some preferred to use M1 (Ye, 2008; Cheng, Luo and Ren; 2007) (M1 is the currency in circulation & current deposit) as they believed that M1 was easier to control than M2 and seemed more realistic. However, some others preferred to use M2 (Nie and Liu, 2005; Qin, 2006; Zhou, 2007; Gu, 2008; Liu, Li and Liu, 2008; Yu and Xue, 2008) (M2 is M1 & quasi-money which includes fixed deposit, savings deposit and other deposits like check deposit) as they thought that in real life, M2 should affect AREDI more than M1. Whereas, most of them had not provided forceful proof to demonstrate why they chose the one they used. 2

1.3 Purpose of the Study So, in order to make explicit how each specific factor of IR and MS actually impact on AREDI and the differences between them, we separate IR (in this thesis, all IR refers to loan rate) into nominal annual rate (NAR) and effective annual rate (EAR), MS into nominal M1 (N_M1), real M1 (R_M1) and nominal M2 (N_M2), real M2 (R_M2) as well as AREDI into nominal (N_AREDI) and real ones (R_AREDI) to shed light on the causal relationship of these specific factors with monthly data and advanced methodology. 1 Moreover, we will test the quantitative and different impact of specific IR and MS on AREDI. But in China, to different types of REDI, the restrictions to their funds by monetary policies are different. For example, for the RDI, the loan term should not be more than three years (including three years), but for like ODI and CDI, the loan term could be extended to as long as five years, which also makes NAR to different types of real estate different (China Bank of Fuses, 1998). That s because the longer the loan term is, the higher the NAR is. All these lead to different cost of loan for real estate developers, making them have different attitude and interest in investment in different types of real estate. However, all of the limited previous studies only considered about the impact of IR, MS on AREDI, so far there has been no empirical study on the relationship of IR, MS and REDI in different types. So, another purpose of this thesis is to fill the gap in the empirical literature on the relationship of IR, MS and REDI in residential, office and commercial buildings, which are three main types of REDI, by using the same methodology as studying the relationship of IR, MS and AREDI but different period of monthly data. 1.4 Methodology The thesis applies the Error-correction Model (ECM) to examine the causal relationship of IR, MS and REDI as well as compare the different impact of specific IR and MS on 1 If not pointing out specifically, IR in the following thesis represents containing both NAR and EAR; MS represents containing all kinds of MS, including all of nominal and real M1 and M2; M1 points to both nominal and real M1, and same for M2 ; REDI represents including both nominal and real AREDI as well as REDI in different types; AREDI points to both nominal and real AREDI, and same for REDI in different types. 3

AREDI and development investment in residential, office and commercial buildings in China by Adjusted R 2 (Adj R 2 ). In order to make sure every ECM is effective, Durbin-Watson (DW) test or AR(1) test will be applied to test for the autocorrelation problem. However, prior to testing for causal and different impact, the Dickey-Fuller (DF) test and Engle-Granger (EG) test will be used to examine for unit roots and cointegration. All these statistical analysis will be done by statistical software of STATA 10.0. 1.5 Limitation The empirical results could depend on the research approach chosen and the data used. When we test the relationship of IR, MS and REDI in different types, we will just use the monthly data from Jan, 2003, which is the earliest monthly data we can get for REDI in different types. That means we will not use the same monthly data as that testing the relationship of IR, MS and AREDI, which are from Jan, 1999. So, it may have some deviation when we compare the empirical results of specific IR and MS on AREDI and REDI in different types. Another limitation is that when we do the Granger causality test by ECMs, we just test as j=2, that is we haven t looked for the optimal lags by any measure. Moreover, it could also be interesting to search for the time lags of impact of specific IR, MS on REDI, which we are not done in this thesis. 1.6 Organization of the Study The thesis is presented in six chapters. Chapter 1 is the introduction of the thesis, including background, problem description, purpose of the study, methodology and limitation. Chapter 2 reports the literature review and presents the related studies on the relationship of IR, MS and REDI before. In Chapter 3, we outline the background of the history development of IR, MS, REDI, funds on REDI and REDI in different types in China. Following this, Chapter 4 describes empirical studies on the relationship of IR, MS and AREDI in China, containing a discussion of the methodology, data used in our study as well as the empirical findings in this Chapter. Different from Chapter 4 focusing on AREDI in China, in Chapter 5, we consider the development investment in three main types (residential, office and commercial buildings) of real estate to investigate the relationship of specific IR, MS and REDI in these three main types with the same 4

methodology as that in Chapter 4 but different period of data. Based on the findings in Chapter 4 and 5, we draw conclusion, contribution of this thesis and provide policy implications of our empirical study in the final Chapter. 5

Chapter 2 Literature Review There have been only a few studies analyzing the relationship of IR, MS and REDI and almost all of them just focused on the impact of IR and MS on AREDI. However, owning to using different variables, period of data and methodologies, the results concluded by different people were different. Both Zhou (2007) and Gu (2008) used the simple linear statistic methods and got the result that both NAR and N_M2 can have significant impact on N_AREDI based on the quarterly data from the first quarter of 1995 to the first quarter of 2005. It was the same result as in Liu, Li and Liu (2008), which applied the quarterly data from the first quarter of 1998 to the fourth quarter of 2006 in the form of natural logarithm. They also found that NAR and N_M2 had a long-run cointegrated relationship with N_AREDI. Different from the above people, who based on the simple level data, Qin (2006) applied ECM to search for the relationship between N_M2 and N_AREDI. He found that N_M2 was the Granger cause for N_AREDI at 5% significance level but not vice versa. The impact of N_M2 on N_AREDI was significant and when N_M2 changed 1% at present quarter, N_AREDI could change 1.87% as response. According to Nie and Liu s (2005) research results based on quarterly data from the first quarter of 1999 to the first quarter of 2005, N_M2 was the Granger cause for R_AREDI at 3% significance level, and NAR also can Granger cause R_AREDI at 2% significance level. He also got the result that single N_M2 was not cointegrated with R_AREDI, but single NAR or it with N_M2 could be cointegrated with R_AREDI. While when he applied ECM to do the statistical analysis, he got that N_M2 only could be significant at 15% level while NAR even couldn t. In conclusion, except Qin (2006) who didn t test the impact of NAR on AREDI, all the others in above got that both NAR and N_M2 could have impact on AREDI, where NAR had negative and relatively weak impact on AREDI and N_M2 had significantly positive impact. Additionally, NAR and N_M2 had negative relationship with each other. However, different from the above people who preferred to use the variables of NAR and simply verified that NAR also could have some impact on AREDI, some other people like Luo and Jin (2009) put forward that only the change of EAR could have 6

impact on AREDI. Furthermore, Tang, Li and Xu (2008) who based their study on the monthly data since 1998 pointed out that there was no significant relationship between IR and real estate supply and the effect of IR policy on real estate supply was insignificant and limited. Lu et al (2008) supported the previous idea based on the monthly data from 1999 to 2006 and Granger reason-result relationship analysis, indicating that IR was not the effective instrument of controlling the AREDI. She also concluded that in the short run, IR and MS had the same influence on the REDI, but it was not strong. Tang, Li and Xu (2008) also proved that there was cointegration relationship between MS and real estate supply and MS was the Granger cause for real estate supply. Furthermore, Lu et al (2008) proved that in the long run, there was a stable positive cointegrated relationship among IR, MS and the ratio of AREDI and MS policy can impact upon real estate market much more apparent than IR policy. Nevertheless, different from some above persons who preferred to use M2 as they thought in real life, M2 could affect AREDI more than M1 as well as they have proved that M2 could have significant impact on AREDI (Nie and Liu s, 2005; Qin, 2006; Zhou, 2007; Gu, 2008; Liu, Li and Liu, 2008), some others like Ye (2008) got the opposite idea as he thought M1 was easier to control than M2 and seemed more realistic. Cheng, Luo and Ren (2007) also supported this opinion by the Granger causality test and got that M1 was the real reason of REDI, but not M2. 7

Chapter 3 History Development 3.1 History Adjustments of IR and MS 3.1.1 History Adjustments of IR IR is the price a borrower pays for the use of money they borrow from a lender (Wikipedia, 2009). In this paper, it just refers to the loan rate since it is the real factor that affects the cost for the real estate developers. Keynes has proved that investment demand is determined by the IR and marginal efficiency of capital, if marginal efficiency of capital is constant, the decrease in IR can increase the investment and vice versa, which means there is a negative relationship between IR and investment (Keynes, 1936). As the real estate is a typically capital-intensive industry and loans from banks are the main channels for real estate development companies to get fund, the changes in IR should affect their loan costs, at least in the short run (Ewing & Payne, 2005). So, the process of change in IR impact on REDI could be expressed as below: IR cost for developers expected profit REDI, and vice versa. Moreover, since the demand of the real estate by customers is the main reason whether the developers would like to invest or not, if IR increases, the cost for consumers who consume the real estate depending on housing mortgage would also increase, which could restrain their demand for real estate and then decrease REDI (IR cost for consumers demand expected profit for developers REDI ). So, according to the theoretical analysis above, the adjustments of IR could have impact on REDI. In order to be in line with and control the development of the national economy, the Central Bank and government in China frequently changed the monetary policy especially IR and MS in the past decade. From Figure 1, we can see that from Oct. 23, 1997 to Dec. 23, 2008, NAR has been adjusted 19 times, which could be divided into three periods. The first period was from Oct.1997 to Oct. 2004, the Central Bank adopted Prudent Monetary Policy 2 to encourage the consumption and investment. So, during this period, it lowered the NAR 5 times from 8.46% to 5.13%, for which the cumulative 2 It is a monetary policy according to the economy orientation, if the sign of the economy is in recession, monetary policy would lean to expansion. On contrary, if the sign of the economy is overheated, monetary policy would lean to contraction. The ultimate aim of this monetary policy is to keep the price stable. 8

reduction point was up to 3.33% and the decreasing amplitude was up to 38.5%. The second period was from Oct. 2004 to Sep. 2008, the Central Bank adopted the Tight Monetary Policy 3 to ease and avoid bubble in the economy, especially bubble in real estate industry. So, during this period, NAR was adjusted quite frequently, which was up to 9 times only during four years, especially in 2007, it adjusted NAR 6 times only in one year. IR increased from 5.31% to 7.47%, for which the cumulative growth point was up to 2.16% and the growth amplitude was up to 40.7%. The third period was from Sep. 2008 to present. Affected by the global financial crisis, the Central Bank adopted Moderately Loose Monetary Policy 4 to stimulate the economic development again. During this period, only in 2008, NAR was adjusted 5 times from 7.47% to 5.31%, which was the same rate as the end of the first period. 10% 10% Nominal Annual Rate 8% 6% 4% 2% 0% 1997.10.23 1998.03.25 1998.07.01 1998.12.07 1999.06.10 2002.02.21 2004.10.29 2006.04.28 2006.08.19 2007.03.18 2007.05.19 2007.07.21 2007.08.22 2007.09.15 2007.12.21 2008.09.16 2008.10.09 2008.10.30 2008.11.27 2008.12.23 5% 0% -5% -10% -15% -20% Growth Rate Nominal Annual Rate Growth Rate Adjustment Time for Nominal Annual Rate Figure 1: History Adjustments of Nominal Annual Rate Source: From the People's Bank of China Note: Growth Rate: using preceding period as the base (same for all following figures) Except NAR, the paper would also consider about EAR, which are the real payment the real estate developers should pay for their loans. Here, EAR is the average monthly EAR between the two adjustment times of NAR; Monthly EAR=(1+monthly NAR)/(1+monthly inflation) -1; Monthly inflation=(cpi n, m CPI n-1, m )/CPI n-1, m, and CPI (Consumer Price Index) is the monthly data on national level, where n represents year, m denotes month, and CPI n-1, m is equal to 100. From Figure 2 we can see that EAR was quite different from NAR and its growth rates fluctuated more than those of NAR, 3 It is a course of action undertaken by the Central Bank to constrict spending during times of strong economic growth, or to curb inflation when it is rising too fast by raising short-term IR or increasing the reserves of commercial banks to reduce MS. 4 It refers to carry out low IR, low reserve requirement ratio and loose control of credit scale for commercial banks. 9

which was due to different inflations in different periods. During the first period, although the Central Bank and government wanted to lower the NAR, the EAR was still very high and even much higher than the NAR. But during the second period, it was opposite: the EAR decreased fast period by period and was much lower than the NAR. During the third period, although mainly the EAR was still lower than the corresponding NAR, but it increased fast especially from Dec, 2007 to Sep, 2008, the growth rate was suddenly up to 1000%. This went on the opposite way as the Central Bank and government wished. When the economy was in contraction with decreasing NAR and low inflation, the borrowers relatively need to pay more than the nominal cost. But when the economy was in expansion with increasing NAR and high inflation, the borrowers relatively paid less than the nominal cost. Annual Rate 10% 9% 8% 7% 6% 5% 4% 3% 2% 1% 0% 1997.10.23 1998.03.25 1998.07.01 1998.12.07 1999.06.10 2002.02.21 2004.10.29 2006.04.28 2006.08.19 2007.03.18 2007.05.19 2007.07.21 2007.08.22 2007.09.15 2007.12.21 2008.09.16 2008.10.30 2008.11.27 2008.12.23 Adjustment Time for Nominal Annual Rate 1100% 1000% 900% 800% 700% 600% 500% 400% 300% 200% 100% 0% -100% Amplitute Adjustment Nominal Annual Rate Effective Annual Rate Growth Rate of Nominal Annual Rate Growth Rate of Effective Annual Rate Figure 2: A Comparison of Nominal Annual Rate and Effective Annual Rate and Their Amplitude Adjustment Source: From the People's Bank of China and Wind Database Note: Since there were two adjustments for the NAR in Oct. 2008, one was on 09 th, Oct (NAR=6.93%) and the other was on 30 th Oct. (NAR=6.66%), here we just use the later one because when we calculate the EAR, it depends on monthly CPI. 3.1.2 History Adjustments of MS In economics, MS is the total amount of money available in an economy at a particular point in time. There are several ways to define "money", but standard measures usually include currency in circulation and demand deposits (Wikipedia, 2009). Usually the Central Bank would change the MS by tools of open-market operations, rediscount rate 10

and reserve requirement, which could change the commercial banks deposit reserve and deposit to affect on the banks lending ability. By changing the banks lending ability, the availability of loans for both real estate consumers and investors could be affected. So, the ways to affect REDI by changing MS could be expressed as follows: if the Central Bank adopts the Tighten Monetary Policy, then MS lending ability real estate loans (or by IR real estate demand ) REDI ; On the contrary, if the Central Bank adopts the Expansionary Monetary Policy, then MS lending ability real estate loans ((or by IR real estate demand ) REDI. Because of the fast development of the national economy and significant increase in foreign exchange reserve (according to the exchange rate mechanism in China, when foreign exchange reserve increases, the supply of RMB (Currency in China) should increase with the same value based on the exchange rate at that time) (Eichengreen, 2006; Liu, 2007; Frankel and Wei, 2007), the MS in China increases year by year. From Figure 3, we can see that the absolute nominal M1 and M2 were much larger than the corresponding real ones and the growth rates of nominal M1 and M2 were also higher than those of the real ones especially from 2003. Additionally, the growth rates of nominal and real M2 were higher than those of correponding M1 in most years. MS (Unit: 100 Million Yuan) 8000000 7000000 6000000 5000000 4000000 3000000 2000000 1000000 0 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Year 35% 30% 25% 20% 15% 10% 5% 0% Growth Rate of MS Nominal M1 Nominal M2 Real M1 Real M2 Growth Rate of Nominal M1 Growth Rate of Nominal M2 Growth Rate of Real M1 Growth Rate of Real M2 Figure 3: A Comparison of History Changes of Nominal and Real M1 and M2 and Their Growth Rates Source: From People's Bank of China and China Statistical Yearbook Note: Real MS=Nominal MS/P, P is the Price level, here uses GDP Deflator as P; GDP Deflator=Nominal GDP/ Real GDP *100; GDP Base Year=1999. The high increase of MS supplies sufficient funds and build a solid foundation for the real estate development, which even lead to risk for its healthy development. In order to control the unhealthy development of the economy, especially the real estate industry, the 11

Central Bank in China in the last years tried very hard to control the MS. For example, in the last dozen years, the Central Bank changed the reserve requirement ratio for 25 times (see Figure 4). From the figure below, we can see it could be divided into three periods which was very similar to the changes of IR. And according to statistics in 2006, if the reserve requirement ratio changed 1%, the absolute change in MS could be up to 300 billion (Li, 2006). Reserve Requirement Ratio 20.0% 18.0% 16.0% 14.0% 12.0% 10.0% 8.0% 6.0% 4.0% 2.0% 0.0% 1.5% 1.0% 0.5% 0.0% -0.5% -1.0% -1.5% -2.0% -2.5% Amplitude Adjustment Reserve Requirement Ratios 1998.03.21 1999.11.21 2003.09.21 2004.04.25 2006.07.05 2006.08.15 2006.11.15 2007.01.15 2007.02.25 2007.04.16 2007.05.15 2007.06.05 2007.08.15 2007.09.25 2007.10.25 2007.11.26 2007.12.25 2008.01.25 2008.03.18 2008.04.25 2008.05.20 2008.06.07 2008.09.25 2008.10.15 2008.12.05 2008.12.25 Amplitude Adjustment Adjustment Time Figure 4: Historical Adjustments of Reserve Requirement Ratio Source: From the People's Bank of China Note: Here in 2008.12.05 and 2008.12.25, only use the Reserve Requirement Ratios for major financial institutions, the Reserve Requirement Ratios for small and medium sized financial institutions are 14.0% and 13.5% respectively. 3.2 History Development of REDI in China Since 1998, there has been large housing and real estate financial system reform in China. Typically when the residential buildings are commercialized and encouraged to trade freely in the market, the high potential demand of the real estate is developed, which drives the average sales price of commercial housing in China growing dramatically (see Figure 5) as well as effectively promotes the development of REDI. From Figure 6, we can see that, except from 2004 to 2005, the number of real estate development companies keeps growing in China, which means the development trend for real estate development industry is quite good and it attracts more people investing on it. 12

5000 25% Average Sales Price of Commercial Housing (Unit: Yuan/Sqr) 4000 3000 2000 1000 0 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 20% 15% 10% 5% 0% -5% Growth Rate of Average Sales Price of Commercial Housing The Average Sales Price of Commercial Housing in CHN Growth Rate of the Average Sales Price of Commercial Housing in CHN Year Figure 5: The Average Sales Price of Commercial Housing in China and Its Growth Rate Source: From Infobank: www.bjinfobank.com and China Statistical Yearbook No.of Dev. Cos. (Unit) 100000 80000 60000 40000 20000 0 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 70% 60% 50% 40% 30% 20% 10% 0% -10% Increased Rate of No. of Dev. Cos. No. of Total Dev. Cos. in CHN Growth Rate of No.of Total Dev. Cos. in CHN Year Figure 6: The Number of Real Estate Development Companies and Its Growth Rate in China Source: From China Statistical Yearbook In addition, in line with the high increase of nominal AFAI, N_AREDI also keeps high increase and even grows faster than the growth of nominal AFAI for most years as the real estate boom greatly arouses the investors enthusiasm of investing. The rates of N_AREDI to nominal AFAI also increased from 13% to 19% during 1997 to 2008. This implies that there are more and more funds invested in real estate development industry, which becomes more and more attractive to investors than the other aspects of the fixed assets because of its high return and it drives REDI becoming more and more important for the development of the national economy. At the same time, the high growth rates of REDI also increase the risk of the real estate industry itself and the national economy, which attract great attentions from the Central Bank and government who want to control its too fast development by monetary policy, especially by IR and MS. 13

35% 30% 25% 20% 15% 10% 5% 0% 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 Year Growth Rate of Nominal AFAI in CHN Growth Rate of N_AREDI in CHN Rate of N_AREDI to Nominal AFAI in CHN Figure 7: A Comparison of History Growth Rate of Nominal AFAI, N_AREDI and Rate of N_AREDI to Nominal AFAI in China Source: From Wind Database and China Statistical Yearbook From all of the figures above, we can get that the year 2004 seems like a turning point as from then the Central Bank and Chinese government began to fully implement the Tighten Monetary Policy by increasing IR and reserve requirement ratio (see Figure 1 and Figure 4). Simultaneously, we can see that the trend of growth rate of N_AREDI in China in Figure 7 basically follows the trends of growth rates of MS. Before 2004, the trends of most figures were increasing, but after 2004, they were mainly decreasing or just increasing with small rates until 2007 or 2008. 3.3 History Development of Funds on REDI in China In China, the sources of funds of enterprises for REDI are mainly four ways: domestic loans, self-raised funds, foreign investment and others, which usually contain 60% to 80% of the funds from deposit and advance payments. Since the source of foreign investment includes few domestic loans and usually is little affected by the changes of IR and MS, here it is not taken into the consideration. As known for all, credit funds from the commercial banks are the most important source of funds for REDI in China. According to Figure 8, we can get that, from 1997 to 2009, the domestic loans occupied about 18% to 24%, the rates of self-raised funds occupied about 25% to 39% and the rates of other funds occupied about 38% to 50% to total funds on REDI respectively. However, among the self-raised funds, most of them came from the sales of commercial housing, which funds were mainly from the bank credits. As estimated, 70% of the self-raised funds and 30% of the funds of others were 14

from the bank credits (Beijing WEFore Investment Consultant Ltd, 2008). According to the above estimation, we can calculate that about 18% to 27% of the self-raised funds and 11% to 15% of the funds of others to total funds on REDI were from bank loans. If to sum the total loans from the bank credits together, it denotes that about 53% to 58% of the funds on REDI in China were from loans. This provides the probability for the Central Bank and government to use monetary policy by changing IR and MS to affect REDI. Funds on REDI (Unit: 100 Million Yuan) 60000 50000 40000 30000 20000 10000 0 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Year 70% 60% 50% 40% 30% 20% 10% 0% Total Funds on REDI Domestic Loans self-raised Funds Others Rate of Domestic Loans to Total Funds Estimated Rate of Loans in selfraised (70%) Estimated Rate of Loans in Others (30%) Estimated Total Rate of Loans Figure 8: Loan Rates for Different Sources of Funds of Enterprises for REDI in China Source: From Wind Database Owning to the fast development of the real estate industry, there are more and more funds invested in the real estate development, which leads to higher rates of domestic loans as well as estimated rate of total loans on REDI to the total loans in China (see Figure 9). This could make REDI much easier to be affected by change of monetary policy of IR and MS. Figure 10 indicates that from year of 2001 to 2008, the growth rates of domestic loans on REDI in China were much higher than that of total loans in China except year of 2004 and 2008, which means that there were more loans flowing into REDI rather than other areas. Moreover, the figure also shows that the growth rates of domestic loans on REDI in China fluctuated more than that of total loans in China. Especially in year of 2004, following with the high increases in IR and reserve requirement ratio (see Figure 1 and Figure 4), the growth rate of domestic loans on REDI decreased a lot immediately. These imply that the loans on REDI may be more sensitive to the changes of IR and MS than the other areas. 15

Rates of Loans on REDI to Total Loans in China 0.8% 0.7% 0.6% 0.5% 0.4% 0.3% 0.2% 0.1% 0.0% 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Rate of Domestic Loans on REDI to Total Loans in CHN Estimated Rate of Total Loans on REDI to Total Loans in CHN Year Figure 9: Different Rates of Loans on REDI to Total Loans in China Source: From the People's Bank of China and Wind Database 50% Growth Rate of Different Loans 40% 30% 20% 10% Growth Rate of Total Loans in CHN Growth Rate of Domestic Loans on REDI in CHN 0% 2001 2002 2003 2004 2005 2006 2007 2008 Year Figure 10: Growth Rates of Total Loans and Domestic Loans on REDI in China Source: From the People's Bank of China and China Statistical Yearbook 3.4 History Development of REDI in different types in China Among many different types of real estate, the most attractive three to the real estate developer are residential, office and commercial buildings. Besides, some other types of real estate development projects like warehouses, hotels, complementary service facilities and land development projects, which are only small parts of AREDI, will not be considered here. Figure 11 below shows the increases of absolute investment in N_AREDI and different types of real estate, while Figure 12 indicates the rates of nominal investment in different types of real estate to N_AREDI from 1997 to 2009. Among the N_AREDI, the residual buildings contributed most, which aggregated development investment increased from 154 billion to 2562 billion (see Figure 11), occupying almost 48.4% to 70.2% of the total investment in real estate (see Figure 12) and working as the main drive for the high increase of N_AREDI. The second biggest part was the investment in commercial buildings, which developed very stably with aggregated development investment 16

increasing from 43 billion to 417 billion. However, its rates to N_AREDI decreased a little, which was from 13.4% to 11.5%. The smallest part was the investment in office buildings, which aggregated development investment inly increased from 39 billion to 138 billion, with its rates to N_AREDI decreasing from 12.2% to 3.8%. The unbalanced development of the investment in these three different types of real estate implies that developers have different attitude and interest on them. But what are the reasons resulting in different development for these three types of real estate? Is that somewhat because that they are impacted differently by changes of IR and MS? We would explore these questions in Chapter 5. 40000 Dev. Inv.on Different Types of Real Estate (Unit: 100 Million Yuan) 35000 30000 25000 20000 15000 10000 5000 N_AREDI Dev. Inv. on Residential Buildings Dev. Inv. on Office Buildings Dev. Inv. on Commercial Buildings Dev. Inv. on Other Types 0 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Year Figure 11: Development Investment in different types of real estate in China Source: From China Statistical Yearbook Rates of Nominal Inv. on Different Types of Real Estate to N_AREDI 80% 70% 60% 50% 40% 30% 20% 10% 0% 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Rate of Nominal Inv. on Residential Buildings to N_AREDI in CHN Rate of Nominal Inv. on Office Buildings to N_AREDI in CHN Rate of Nominal Inv. on Commercial Buildings to N_AREDI in CHN Year Figure 12: Rates of REDI in different types to N_AREDI in China Source: from China Statistical Yearbook 3.5 Conclusion 17

The Central Bank and government in China emphasize using the monetary policy of changing IR and MS to control the national economy more than the western countries, as they believe their changes could have impact on the national economy, especially like the REDI whose funds are mainly from banks. As we estimated above, usually there are more than half funds on REDI from banks, and the more the funds on REDI from banks, the more sensitive REDI seems to be to changes of monetary policy by changing cost, expectation and financing environment for the investors. But whether the changes of IR and MS could really have impact on REDI in China? If they really have, how strong they are and what are the differences between each nominal and real factors? We would try to explore these questions with advanced methodology in the following chapters. 18

Chapter 4 Empirical Studies on the Relationship of IR, MS and AREDI in China 4.1 Methodology According to Engle and Granger (1987), a linear combination of two or more non-stationary series (with the same order of integration) may be stationary. If such a stationary linear combination exists, the series are considered to be cointegrated and long run equilibrium relationship exists. Incorporating these cointegrated properties, an error-correction model (ECM) could be constructed to test for Granger causation of the series in at least one direction (Shui & Lam, 2004). Moreover, the ECM is specifically adopted to examine the quantitative impact of changes of IR and MS on REDI. 4.1.1 Stationary and cointegration test As the prerequisite for using the ECM is that the series should be cointegrated with the same order, it is necessary to test the stationarity and cointegration first. A series is considered to be stationary if it is integrated of order zero I(0), otherwise, it s non-stationary. If a non-stationary series has to be differenced d times to become stationary, then it is said to be integrated of order d: i.e. I (d) (Shui & Lam, 2004). In this paper, the Dickey-Fuller (DF) test has been applied to test the unit root and stationarity. The regression model is as follows: Y Y = + δ + t t 1 u α (1) Where is first difference operator,α is the drift, δ = ρ 1, where ρ is the unit root from AR(1) model ( Y Y = α + ρ + t t 1 u t t ), t is the time index, u t is the error term. If the null hypothesis H 0 : δ 0 is true, then Y t is non-stationary; Otherwise, if H 1 : δ < 0 is true, then Y t is stationary. Here we can use the t-values depending on the size of sample directly. If t is smaller than the critical value at given significance levels, then get the series stationary. If both series are integrated of the same order, then we can test for the cointegration. 19

The Engle-Granger (EG) procedure (1987) is used for this purpose. The EG test can proceed in two steps in a simple linear regression. The first step involves the following static OLS regression: Get the equation: ε t Y = α + β t X + (2) t Then save the residuals, ε t, which equals ^ ^ Y t ^ ^ X = α + β (3) t ^ ε ^ t = (4) t t Y Y The second step is to run the DF test on the residual ^ ε t to see whether it is stationary or not. The time series are said to be cointegrated if the t-statistic for testing ^ ε t is smaller than the corresponding critical value, otherwise, it indicates that the series are not cointegrated (Noriega & Santaularia, 2006). 4.1.2 Granger causality test and Error-correction models Granger causality is a term for a specific notion of causality in time-series analysis. Simply it defines as a variable X Granger-causes Y if Y can be better predicted using the histories of both X and Y than it just use the history of Y alone ( POLS571 - Longitudinal Data Analysis, 2001) and vice versa. The existence of cointegration relationship indicates that there is long-run relationship among the variables and thereby Granger causality among them in at least one direction. According to the introduction of Banerjee, Dolado, Juan, and Hendry (1993), Shui and Lam (2004) and Wigren and Wilhelmsson (2007), ECM is used for correcting disequilibrium and testing for short and long-run causality among cointegrated variables. The dynamic regression models are as follows: Y X t + β + 1 ( ) ( ) 0 yut β j Y β j X 1 2 = β + (5) t + t j t j j = 1 j = 1 α 1( j) Y α 2( j X + + + 0 α xut 1 ) t j t j j= 1 j = 1 ε yt = α + ε (6) xt 20