Mortgage Prepayment and Default Behavior with Embedded Forward Contract Risks in Chinas Housing Market

Yongheng Deng and Peng Liu

First Version: October 2005 This Version: March 2007

Keywords: Mortgage, Prepayment, Default, Credit Risk, Forward Contract JEL Classification Number: G12, G14, G21, H31

Contact author: Peng Liu, F605 Faculty Building #1900, Haas School of Business, University of California, Berkeley,
CA 94720, Tel: 510-6438543, Email: peliu@haas.berkeley.edu; Yongheng Deng, University of Southern California, School of Policy, Planning, and Development, 650 Childs Way, RGL 201A,Los Angeles, CA 90089-0626, Tel: 213821-1030, Email ydeng@usc.edu. We thank Mark Carhart, Pierre Collin-Dufresne, Tom Davidoff, Kent Daniel, Robert Edelstein, Dwight Jaffee, John Quigley, Richard Stanton, Adam Szeidl, Sheridan Titman, Nancy Wallace and seminar participants at Cornell University and U.C. Berkeley for helpful comments. All errors are ours.

Mortgage Prepayment and Default Behavior with Embedded Forward Contract Risks in Chinas Housing Market ABSTRACT Forward housing market, which enables the consumer to lock in future housing price, is very common in Asian countries. Most condominiums in China are sold forward on a pre-sale market, where purchasers and developers transact at the date of pre-sale on an underlying property, which is not yet completed. During the pre-sale period there is significant default risk of developers. However home buyers can borrow mortgages from banks so that they can share the forward contract risk with the banks. This explains the phenomena of irregularly high early stage default and prepayment rates observed in residential mortgage lending in China where there is little or no financial incentives for mortgage borrowers to exercise either put or call options. An investor (home buyer) in Chinas forward housing market will choose to default her mortgage contract if the developer defaults the forward housing contract. If forward house is delivered, the investor may choose to prepay the loan depending on her liquidity constraints and returns in alternative investment channels. Mortgages collateralized by forward housing asset are riskier than those with underlying assets traded on the spot market. However, currently the Chinese mortgage banks charge the same rate to all mortgage borrowers. This results a cross subsidization between mortgage borrowers groups in the forward and spot housing markets. Cross-subsidization, where one group pays a relative high price and thus enables another group to pay a relatively low price is pervasive phenomenon in both product market and financial market. This paper uses a rich set of data from a leading mortgage lender in China. The loan history data set contains not only mortgage loan characteristics, but also borrowers and developers information. The unique data set allows us to study the mortgage borrowers prepayment and default behavior with embedded forward contract risks. The finding of this study will provide valuable insight about emerging housing and mortgage markets in China as well as those in other transition economy.

1. INTRODUCTION The residential mortgage market in China has grown rapidly since 1998 with an average annual growth rate of roughly 100%. By the end of first quarter of 2005 the outstanding balance of residential mortgages reached 1.7 trillion RMB Yuan, approximately $207 billion US dollars.1 There are two highly distinctive features of the Chinese mortgage market: first, the Peoples Bank of China (the Central Bank) set a uniform mortgage rate which is applied to all types of borrowers. There is a long history of not using risk-based interest rates in China or other centralplanning economies. It is also a tradition that in those countries a unified product or service is provided to all consumers. Second, beyond the standard spot market for housing transactions, there is an active forward real estate market in the sense that the developer can sell housing unit before its completion, sometimes even before construction begins. During pre-sale period there is significant default risk of developers. However, the home buyer in China is allowed to borrow a standard mortgage from a bank to finance the pre-sale unit. As a result, home buyers in the presale housing market can share the forward contract risk with the banks. This explains the phenomena of irregularly high early stage default and prepayment rates observed in residential mortgage lending in China where there is little or no financial incentives for mortgage borrowers to exercise either put or call options. An investor (home buyer) in Chinas forward housing market will choose to default her mortgage contract if the developer defaults the forward housing contract. If forward housing unit is delivered, the investor may choose to prepay the loan depending on her liquidity constraints and returns in alternative investment channels. Because of the embedded risk of developer default, the mortgages collateralized on pre-sold property are more risky than their counterparts on the spot market. This paper studies the competing risks of mortgage prepayment and default with embedded forward contract risk of developers default. The economic model is based upon Cox proportional hazard model (Cox, 1972, 1975) of mortgage termination. The empirical analysis is based upon a rich set of mortgage lending data from a leading mortgage lender in China. The loan history data set contains not only mortgage loan characteristics, but also borrowers and developers information. The unique data set allows us to study the mortgage borrowers prepayment and default behavior with embedded forward contract risks. The finding of this study will provide valuable insight about emerging housing and mortgage markets in China as well as those in other transitional economy. The remainder of the paper is organized as follows. Section 2 describes the Chinese mortgage market in detail, including both the single mortgage rate
1

As of March 2007 the exchange rate of Chinese Yuan (CNY)is one U.S. dollar = 7.74 CNY.

constraint and the forward market for new units. Section 3 describes the data used in this study. Section 4 lays out the empirical methodology. Section 5 presents a discussion of empirical results. A brief conclusion follows. 2. THE MORTGAGE AND REAL ESTATE MARKETS IN CHINA There is long history of Real Estate market and consumer loan market in China before 1949, when China adapted the central planning system after the Communist Party took over. The earliest real estate transaction can be traced back to 20 BC, Eastern Han Dynasty 2 . Before adapting the central planning system in 1949, active real estate and mortgage markets had already existed in China especially in big cities, such as Shanghai and Beijing. For a long period of time under central planning regime, housing in China had been social welfare product administrated and delivered by state agencies (e.g. state-owned enterprises and housing bureaus) for its people. Under such a welfare-oriented system, a private housing market and housing mortgage system extinguished. The mission of Chinese Banks was to act as a government directed funding source for SOEs. Since the early 1980s, China has gradually restructured its housing system. Market mechanisms, with the objectives to eliminate state housing allocation, promote privatization of public housing and encourage private housing development, were introduced in stages to replace the welfare housing system. Although China's first modern residential mortgage loan was issued in 1986 by China Construction Bank (CCB), the mortgage market did not play an important role in the Chinese residential housing market for another decade. By the end of 1997, the total outstanding mortgage balance in People's Republic of China was only approximately 19 billion RMB Yuan (USD $2.35 billion). In 1998, the People's Bank of China, which functions as the central bank of China, and the State Council of China announced several administrative laws to speed up housing construction and intensify urban housing reform. Among them, the State council announced that it would no longer allow state-owned enterprises (SOEs) to allocate welfare housing to their employees after December 31, 1999. Since then, residential mortgage lending began to accelerate. By the end of 2005 the outstanding balance of residential mortgages exceeded two trillion RMB Yuan (USD $198 billion), an increase of 40% compared to the 2003 balance, and almost 89 times the1997 balance. Residential mortgages play an increasingly more important role in Chinese banks' lending activities. The outstanding mortgage balance constitutes

The Cambridge History of China Volume 1, The Chin and Han Empires P666, Edited by Denis Twitchett and Michael Loewe.

more than 12% of total loans made by financial institutions in the same period in 2005, compared to less than 0.4% in 1997. [Insert fig. 1] Although mortgage lending currently constitutes about 85% of total consumer lending (no credit card loans and very few automobile loans are issued by banks in China), the ratio of mortgage loans to total loans is 7.5% lower than most developed countries (e.g. the ratio of mortgage to total loan is 39% in U.S. and 59% in U.K.), and even lower than other countries in the Asian-Pacific region (e.g. the ratio in Hong Kong, Taiwan, Singapore, Malaysia and Korea was 34%, 30%, 27%, 26% and 23%, respectively) . Therefore there is much room for continuing growth in China' mortgage lending. [Insert fig. 2] [Insert fig. 3] [Insert fig. 4] The Institutional Background of Residential Mortgage Lending in China The current residential mortgage market in China is dominated by four major state-owned banks. They account for more than 90% of the total outstanding mortgage balance. Of these four banks, Industrial and Commercial Bank of China (ICBC), and China Construction Bank (CCB) are the two leading mortgage lenders, representing about 70% of total outstanding mortgage loans. Currently there are four categories of residential mortgages in China: (1) individual account housing loans, (2) individual provident fund housing loans, (3) combined housing loans, and (4) second-hand housing loans. Individual account housing loans is the most common form of individual household mortgage loans. CCB Individual provident fund housing loans are housing loans using provident funds that bank operate on half of institutions that manage the funds. Usually individual provident fund housing loans enjoy a lower mortgage rate than individual account housing loans. The current spread is 44 basis points. Combined housing loans refer to loans granted to individual home buyers, using both bank's funds and provident fund as a source of funding. Second-hand housing loans are loans that facilitate individual purchases in the secondary housing market. Mortgage application requirements

The loan amount shall not exceed 80% of the purchase price or the appraisal value, whichever is lower. That is 20% of the purchase or maintenance price, is reserved for the down payment. The applicant should provide acceptable statements of stable income and payment to income ratio should not exceed 70%. The mortgage term shall not exceed 30 years. Other required documentations include personal identification, purchasing contract, title to the property, appraisal report, housing (pre-)sale permit, banking statement, proof of stock and bond investments, etc. Payment methods If the loan term is one year or shorter, both principal and interest must be repaid as a lump sum at maturity; if the loan term is longer than one year, the loan may be repaid in equal installments of the principal plus interest, or in equal installments of principal. The borrower may choose either method, but there may be only one payment method for each loan, and after the method is specified in the contract, it may not be changed. Mortgage Interest Rates and Adjustment The mortgage rates that banks charge to customers are regulated by Peoples Bank of China (PBC), the central bank of China. All banks follow the same lending rules. Currently the mortgage interest rates is 4.77% for individual account housing loans with a term of five years or below; for loans with a term above five years, the interest rate is 5.04%. The spread between the long term (more than 5 years) and short term (5 years or less) is 27 basis points. The individual provident fund housing loans is 3.6% and 4.04% for term below 5 years and above 5 years respectively. All residential mortgages in China are adjustable rate mortgages (ARMs). During the loan term, the interest rate may change according to what is set by the People's Bank of China (PBC). Once the new mortgage rate is announced, the rate will be applied to all mortgages. Default Risk Management Currently Chinese banks use a five-category system to manage default risks of mortgage portfolio. The five default risk categories are Normal, Alert, Irregular, Distress, and Default. Before approval of a loan, the bank investigates the credit worthiness of each individual applicant. The bank will approve a mortgage to an applicant according some borrower characteristics like family income, occupation, etc. Therefore there are only two pre-lending classes: Normal for better credit loans and Alert for good credit loans. The bank watches the

borrowers' payment behaviors very carefully and makes frequent adjustments to the borrowers' risk level if they see any warning signal. Specifically, if the number of delinquencies is between 3 and 6, mortgage borrower's risk level becomes Irregular; if the number of delinquencies is above 6 then borrower's risk level becomes Distress. The last category in the five-category risk level is Default. The headquarter of any mortgage lending bank also actively monitors its branches and subsidiaries. An early warning is given to the branches with 1% to 3% delinquency rate in their mortgage loan portfolio; the headquarter requires a reorganization of mortgage operations at branches whose delinquency rate falls between 3% to 5%; the mortgage lending license would be revoked if a branch has more than 5% delinquency rate. Real Estate Developers and Strong Secular Growth Trend in Housing Demand Real estate development was a highly regulated industry in China before 1992. With the privatization of state owned housing and the fast growth of mortgage lending, real estate developers in china begin to flourish. There are currently more than four thousands active real estate developers in Beijing. Before 1992 there were only 37 state-owned real estate developers in Beijing. With an average increase of 50 developers a year, the total number of developers until 1997 was 335. The total number of developers doubled in 1999 and structure of developer became diversified. With the stimulation of mortgage lending activities, the average annual growth rate jump to 96%, the total number was 548 in year 1998, 1023 in year 2000 and 1979 in year 2002 respectively. In addition to the housing reform, China's capital market begin to grow, with the recent reform such as apparition of Chinese currency Yuan and further opening of financial market to foreign banks as required by the WTO. Moreover the most recent financial innovations such as real estate investment trusts (REITs) and mortgage backed securities (MBS) all start from the real estate finance sector, the emerging and fast-growth capital market in China. The on-going process of these reforms has translated into a booming domestic property sector and acceleration in residential housing investments in recent years. As a consequence, the share of residential housing investment as a percentage of GDP has risen to 5.6% in 2003, up from less than 1% before the 1990s. Notwithstanding fast industrialization, the degree of urbanization, as measured by the ratio of urban residents to total population, is exceptionally low in China. Industrialization without urbanization is a unique Chinese phenomenon, owning to decades of government policies that segregated urban and rural labor markets. The pent-up urbanization demand, evident from the acceleration of rural migration, has also coincided with the unleashing of pent-up demand for

improving housing conditions nationwide. The average per capita housing area in square meters for both urban and rural residents has jumped from7-8 sq. m. in 1978 to above 20 sq. m. in recent years. We believe the evolution of consumer demand is heavily influenced by income distribution and the emergence of the middle class, because at a different income levels, different projects become affordable and available. Experiences in other economies have shown that as annual income crosses US$ 3,000 per capita; there is a period of rapid expansion in the size of the middle class and acceleration in consumer demand. Figure 5 shows the decreasing nationwide trend of supply to demand ratio indicating the strong secular growth trend in China's housing demand. [Insert fig. 5] The Forward (Pre-Sale) Housing Market in China The majority of residential housing units in urban China are condominiums, and most condo purchases are transacted on the forward market. A forward sale is also called pre-sale. In 2003, more than 80% of projects initiated a pre-sale according to a survey. Condo developers usually sell their projects well before completion, usually two to three years,3 in order to hedge the housing price and get additional funding. On the transaction date, home buyers (mortgage borrowers) have to pay purchase price in full on a development that is not completed. To share the risk of developer default, home buyers (mortgage borrowers) usually fund the housing purchase by putting a down payment of 10% to 30% and borrow a 20 year mortgage4 from a bank that charges the same adjustable mortgage rate. The mortgage provides the borrower embedded options to default and prepay the loan at any point before loan maturity. Even though banks pool forward and spot contracts in real estate market by charging the same rate, one should realize that the market structure leads to very different option exercise behaviors for the two types of borrowers. The collateralized borrowing based on the presale units introduces additional risk for mortgage default and prepayment. The developer asks for the full purchase price at presale day, the mortgage loan serves partially as consumption insurance. If the developer defaults on the forward contract, the borrower defaults on her loan. (The loan is non-recourse in China). Presale does not only exist in China, it is a popular practice in other markets as well. In most Asian markets: Hong Kong, Taiwan, Singapore, Japan and in some European countries, like Russia, Italy, Span forward housing sales are very active. The presale practice is also active in the
3
4

The presale period used to be around 4 years, but now decreased to one to two years. Sometimes shorter maturity of 15, 10 years mortgages is borrowed with different rate.

U.S. condominium market, especially in large cities. In United States, pre-selling of buildings and resort condominiums (Miami, Hawaii, for example) has become a standard process, and virtually every condominium is pre-sold today.5 The procedures of forward presale contract can vary from country to country, and from developer to developer in the following dimensions: length of presale period, size of down-payment/deposit, payment schedule, refund policy in case of the developer default, and eligibility of mortgage loan and mortgage rate. In Hong Kong, the common practice is that the developer must complete a certain percentage of the development before the government gives the presale consent. In the U.S., China and Singapore for example, the forward sale can take place after developer getting construction permit and before construction starts, thus the name pre-construction. For example in Florida developers need preconstruction sales of 50% to 90% of the units before they can borrow funds to begin construction. While in Singapore and China the entire purchase price is required to be paid at the time of presale contract is signed; a sequential payment schedule is contracted in U.S. For most Florida condominium developments, investors need only deposit 10% of the presale price at contract time, a second 10% after 3-6 months or when developer breaks ground. For most condo presales in U.S., the deposits are held in escrow. The investors can get full refund in the case of developer defaults. Despite the differences in the contractual details of forward presales, there is a common risk that associated with all forward contracts: the counterparty default risk. On the forward housing market, all investors are well informed of the possible failure of delivery on the contracted houses. Because the investors are buying something that does not yet exist, there is a greater potential for unforeseen problems and setbacks to occur before the investor move into her home. The need not be built disclaimer gives the developer protection from suits of nonperformance if the developer does not proceed with the construction. However there may be a significant social loss in the sense of incomplete structures that arise from developer default. When the presale was first used in China around 1998, majority of developers are state-owned and completion of the development was perceived by forward contract purchasers as guaranteed by government. As the housing demand is so high and real estate development is such a profitable industry, more and more private companies came into the market. With uncertain construction cost, the developers then have incentives to exercise their default option. This introduces inefficiency for there would be social waste resulting from uncompleted projects. The Chinese

Sample of pre-sale contract can be found in Miami real estate website. www.miamirealestatetrends.com

government has realized the problem incentive-compatible mechanisms are under consideration beyond reputation and regulation rules. The convoluted relationship among the developer, home buyer and mortgage lenders can help to explain two paradoxes in China mortgage market: why the prepayment rate is high in the adjustable rate mortgages where there is no incentive to refinance? Why the default rate is also high in a period of a booming housing market? We will explore the phenomenon and explanations in more detail in the empirical modeling session. 3. THE DATA Micro Mortgage Loan Data The unique micro mortgage dataset collected by the largest residential mortgage lender in Beijing, China was first used by Deng, Zheng and Ling (2005). In this paper we enriched the dataset in several dimensions. We updated the mortgage loan data to the end of 2003. Furthermore we extended the loan information to include the property information and developer characteristics. This second extension is crucial because those variables enable us to merge borrowers characteristics of loan with the collateral information from developers in Beijing. The loan dataset includes 103,462 individual mortgage loans issued between 1998 and 2003. The mortgages are 5 year, 10 year, 15 years and 20 years adjustable rate loans with level payment. For each loan, the available information include the year and month of origination and termination (if it has been closed), indicators of prepayment or default, original loan amount, down payment rate, initial loan-to-value ratio, maturity, remaining term, repayment method, mortgage contract rate, the purchase price of the property, size of the house, location of property, sale type, appraisal value of property at time of sale, unit price, etc. Other borrower characteristics include borrower name, education, gender, monthly income, occupation and position, number of dependents in the household and income from spouse. The dataset also has geographical information of the property and its developer. This micro loan data is well suited for the study of mortgage default and prepayment, since Beijing is the capital city of China and the real estate sector is steadily increasing without speculative bubble. Among the 103,462 mortgage loans, 1,384 loans were defaulted and 10,055 loans were fully prepaid during the sampling period, and 92,023 loans were censored at end of year 2003. This data is best for survival analysis because the data avoid the truncation and censoring problems. The sample started from 1998; which is the first year the bank issue mortgage loans. There is no left truncation problem as most research has in sample selection

process. The data collection cutoff date is December 31, 2003. Therefore the right censoring is non-informative6. Tables 1-5 provide the descriptive statistics of the mortgage loans. Among 103,462 residential mortgage loans originated in Beijing, more than 74% are based on forward contracts. Among 1,384 defaulted loans, and among 10,055 prepaid loans, more than 90% are from forward market. The borrowers behavior in default and prepayment vary by loan characteristics like LTV, original loan amount, and borrowers income, occupation and other household characteristics. In addition, borrowers behavior in default and prepayment vary by developer characteristics like type, size, quality, history, etc. For example, about 45% defaulted loans are originated on the properties built or to be built by joint developers with partners in Hong Kong, Macau or Taiwan (HMT thereafter); while only 5% from joint developers with partners in foreign countries. Within all prepaid mortgages, 47% are based on properties built or to be built by limited companies; while only 2.5% are by joint venture with HMT. Among the 103,462 mortgages, more than 63% loans are originated on a property which is the first project by its developer. Only 34% are by developers who have built more than one project in the past. The detailed descriptions can be found in table 1-4. [Insert Table 1] [Insert Table 2] [Insert Table 3] [Insert Table 4] [Insert Table 5] Developers Credit Risk Data Another dataset used in this paper is credit rating data of real estate developer in Beijing from Beijing Urban Construction Development Office (BUCD thereafter). BUCD is a government agency regulating real estate industry in Beijing. In 2004 BUCD begins to annually review all real estate developers in Beijing and publish the results on its official website. With growing concern about information disclosure of developer credit record, BUCD also has built a
6

An alternative sampling method, namely stratified sampling mechanism can be used for a large dataset. To correct the possible sample selection bias, a weighing scheme which is assumed to be independent of error distribution can be used in the maximum likelihood estimation. Specifically, the weight addresses the stratified choice-based sampling of mortgage .les across loan status cells. The weight is commonly defined as the inverse of the probability that the loan is being selected from a cell where was sampled.

credit database of real estate developer for public inquiry. This dataset include 3,088 developers and 3,938 real estate development projects in total. Among all developers, about 87 percent of developers only have one property in Beijing; only 3% of the developers have 4 properties or more. The earliest developer in our sample, a state owned enterprises (SOE), started its real estate business in 1966. There was no big growth in real estate sector since then. In 1992 the total number of developers was only 90, the number Quadrupled in 1997 to 335. Since then real estate development industry has experienced a significant growth. By the end of 2003 the total number of developers was almost ten times the number in 1997. For each developer in the sample, we have information of credit ratings from real estate administration office and commercial banks, type of developer, equity value of firm at registration, location of properties and business address of developer, name of CEO and legal person, total number of employees and detailed number of certified professionals in the management. We also have the issuance and expiration date of license, starting date in real estate business. 4. THE EMPIRICAL MODEL AND ESTIMATION METHODOLOGY 4.1 AN ECONOMIC MODEL OF MORTGAGE PREPAYMENT AND DEFAULT WITH EMBEDDED FORWARD CONTRACT RISKS A Simple Model for Forward Housing Market The presale practice, adapted from Hong Kong and other Asian markets created a forward market in Chinese residential housing market. Similar to financial instruments, investors can buy real estate either on a spot market or on a forward market. On the spot market, investors and developers transact on existing housing units - housing stock; while on the forward real estate market, the investors and developers agree on the price at the date of sale but the underlying property, which is not completed yet, is transferred to the buyer only at a certain period later, usually at the date of completion. Because the unit is sold well before completion and occupation, the forward contract is often called a pre-sale or pre-construction. [Insert fig. 6] Figure 6 uses a simple timetable to illustrate the interrelationship between the home buyer (mortgage borrower), the developer and the bank on the forward housing market.

10

Following Liu (2007), we set up a three period model, the agents only make decisions at discrete time at T = 0, 1, 2, 3. T=1 is completion / delivery date of housing units. Home buyer (mortgage borrower) can enter the housing market either at time T=0 (spot market) or at T=1 (forward market). Therefore the transaction date can be either at T=0 or at T=1. At the transaction date, the home buyer has to pay the developer the full purchase price on the housing unit and the home buyer is eligible to borrower an adjustable rate mortgage from the Bank. The mortgage loan is usually less than 80% of the housing value, which means that the purchaser only put down payment of 20%. The mortgage matures at T=3, but the borrower can terminate the loan at any time. For simplicity, we assume the borrower only terminate the loan at T=1, 2, 3. The developer and the bank have an agreement so that the developer deposits the proceeds into the bank at transaction date. Between time 0 and time 1, there is a pre-sale period, in which the developer sells housing units forward before completion of the project. At T=1 of the forward market, there is a probability of developers failure of delivery of the unit, which results in default of the mortgage. In the case of default, the bank takes the collateral residuals, which typically have very low recovery rate currently in China. If the housing unit is delivered in good quality, the home buyer (mortgage borrower) starts to consume the housing consumption. T=2 is the option exercise date for a home buyer (mortgage borrower) who is consuming a housing service and pays a monthly interest and principal of mortgage. She would default if the value of the house is lower than the value of the debt borrower from the Bank. She would prepay if the financing cost, i.e. the mortgage interest cost, exceeds the alternative investment return. The borrower may not optimally prepay the mortgage if she is financially constrained, either due to liquidity constrain (the borrower does not have enough cash to pay off the loan) or she does not participate the stock market investment. Therefore if the home buyer (mortgage borrower) does not terminate the mortgage, at T=3 she makes the last payment and obtained the title of the house. There are at least two main advantages for home-buyers to lock into a pre-sale contract before the construction of the property is completed. First the housing price in the pre-sale market is typically lower than that on the spot market. Second there may be richer menu of choices: location, size, view, etc. For developers, there are two benefits associated with pre sale contract. First, by locking in the selling price, the developer can hedge the project pipe line risk and share the risk with buyers. The uncertainty about future demand and inventory risk is significantly reduced. Secondly the developer can access additional zero cost financing in addition to construction loan which is typically very difficult for private developer to apply from the banks. For new developers, presale funding may be the only source of capital for construction.

11

In China and other emerging markets, home buyers typically lose all the down-payments in the case of developer defaults. However currently Chinese home buyers can borrow a certain amount of mortgage loans from banks, and furthermore they benefit the same mortgage rate as spot market investors. A condo purchaser can default the mortgage loan, if developer fails to deliver the unit according to the presale contract. Because of the popularity of the forward housing market and the practice of the risk sharing mechanism adopted by the home buyers (mortgage borrowers) in China, the risks in the forward real estate market have been embedded into the mortgage contract default risks, which makes the valuation of the mortgage default and prepayment option value becomes more complicated than those in other countries. Default Option with Embedded Forward Contract Risks The option theory (Black and Scholes, 1973 and Merton, 1973) has been widely applied to estimate the mortgage default and prepayment probability. Early examples include Findley and Capozza (1977), Buser and Hendershott (1984), Brennan and Schwartz (1985) among others. Recent applications include more sophisticated modeling techniques (see, for example, Schwartz and Torous, 1989, Stanton, 1995, Deng, Quigley and Van Order, 2000) and more realistic assumptions (see, for example, Quigley, 1987, Quigley and Van Order, 1995, Archer, Ling and McGill, 1996, Calhoun and Deng, 2002). However this method was developed and first used in developed mortgage market in which spot housing contracts are predominant. Understanding of forward housing market and its option structures are necessary in order to apply the option theory to the mortgage pricing in which the forward selling is active. Under the forward contract when buyers borrow mortgages from bank and start to pay back amortized loan amount, they have not seen their house yet. With astonishing growth in real estate market, many private developers enter the residential housing market. There is increasing default risk among developers that fail to deliver the house. Many Chinese borrowers use the mortgage as a consumption insurance instrument to share the credit risk of developer with banks. In case the borrower did not get the desired home, i.e. the developer defaults or fails to satisfy the home buyer on the date of delivery, the borrower defaults. On the other hand, if the borrower gets a house of good quality, she will not default, and will consider prepaying according to borrower's liquidity constraint and other investment opportunities. For example, if borrower can earn a higher return from other investment portfolio (from stock market or bond market), even though the borrower has cash on hand - not liquidity constrained, she rather not prepay the mortgage. However if other investment opportunities give her lower return then mortgage rate and If the investor is not constrained, she will pay off the loan. Although in reality investor has to evaluate the option value of waiting, the

12

prepayment option basically reflects the trade-off between financing cost (mortgage rate) and alternative return, as well as liquidity constraint of the borrower. This forward-selling practice is quite common in other developing countries or transitional economies as well, where there is high demand in residential housing and the governments encourage consumer lending to stimulate real estate construction. The increased leverage via pre-sale then increases credit risk of developer which will be eventually transferred to house buyers. It is rational for the buyer to borrow mortgage loan from a bank, thus sharing the housing-consumption risk. On the demand side, since the commercial housing market is relatively new, there may simply not enough desirable housing on the secondary housing market. Besides, recognizing that the banks charge the same mortgage rate, pre-sale mortgage borrowers implicitly get a cross subsidy from spot market borrowers. The Default and Prepayment Option Analysis on the Forward Market Due to the pre-sale feature, the mortgage options are different from the ones in the spot market. Because all the current mortgage contracts in China are Adjustable Rate Mortgages (ARMs)7 and the mortgage rates are not indexed to any rate, but imposed by the central bank. We found that pre-sale practice in China's housing market plays a crucial role in mortgage defaults and prepayments. An important reason why people want to borrow money from a bank is to facilitate a housing consumption. Since they face tremendous risks when developers fail to deliver properties on the contracted delivery date or fail to meet the quality requirements specified in the pre-sale contracts. Here house refers to all kinds of housing units in China's residential housing market, most of them are condominium, and some are not. In our sample period, every year there are more than five hundred of new properties in Beijing. (Note: most housing units are condos in which one property could have tens of thousands units). The sophisticated borrowers use the mortgage as a consumption insurance instrument to share the credit risk with banks. If the borrower had a bad draw, i.e. the developer defaults, the borrower also defaults. On the other hand, if the borrower has a house of good quality, she will not default, and will consider prepaying according to borrower's liquidity constraint and other investment opportunities. There are two important institutional issues that permit such a risk sharing behavior. Firstly residential mortgages in China are non-recourse; and secondly, so far there is no personal credit system in China, and banks dont share mortgage borrowers information. These features are especially common in transitional economies or developing countries.
7

See Stanton and Wallace (1999) for a discussion of application of option theory to the adjustable rate mortgages.

13

Before the discussion on how to calculate the value of default and prepayment options, further understanding the details of those options and how they differ from their U.S. counterparts are needed. In China, a housing unit can be purchased on either the spot market, where the house has been built, or on forward market (pre-sale), where the building has not been completed yet. Although fundamentally different, (a specific housing unit cannot be simultaneously on both the spot market and on forward market at the same time), the purchase and financing procedures are exactly the same. Consumers have to pay in full using either own equity or mortgage or a combination of both on the transaction date instead of forward delivery date. Mortgage lenders charge the same rate to the borrowers. For the pre-sale house, the borrower can default on a mortgage loan at delivery day. It is essentially a European put option. On delivery day, there is a probability of developer default, hence failure of delivery of a contracted house. Mortgage default can be of two reasons during pre-sale period. The first is from the credit risks of developer. These risks include market risk (via a channel of construction cost) and the idiosyncratic risk of the individual developer. The second risk is from the borrower. An individual borrower can have a negative income shock or other unplanned event (e.g. a big expenditure on medical care). Since during the pre-sale period, the borrower cannot borrow from other places and cannot re-sale the house (due to its adverse selection problem), they will lose the house and payments made earlier plus the down payment if they default before the maturity. Knowing this, a rational borrower will keep enough funds for the mortgage payment beyond the down-payment during pre-sale period. For simplicity, we assume the borrower only exercises the put option on the delivery date. Therefore the mortgage partially serves as consumption insurance instrument. A borrower wants the protection from the credit risk of developer or qualifies of house. At the delivery date, the mortgage options for pre-sale and non- presale becomes identical. After the pre-sale period, the borrower can default or prepay on the mortgage. Mortgage default option can be regarded as a financial put option. If the current market value of the house, which serves as collateral on the mortgage debt, drops below the current value of the remaining mortgage balance, a borrower has an incentive to default. In a boom market, the default option is usually out of the money. Borrowers would be better off by selling the house, rather than default, if they have to move. On the other hand, the prepayment option is a real option (call) with strike price updating every year (depending on bank ARM rate). This call option is does not depend on interest rate, but is closely related to the borrower's alternative investment set. One can think of mortgage debt as a consumption smoothing instrument. In the current Chinese capital market, there are very few investment opportunities, and very few borrowing vehicles. For example there are no credit card loans and other consumer loans are very limited. The mortgage market is a

14

major and steady growing sector in Chinese debt market; while the stock market is the major investment sector. Therefore borrowers will make prepayment decisions based on cost of capital and stock market return. They will prepay when the cost of capital (mortgage rate), exceeds investment return. Figures 7-8 show the mortgage rate dynamics and stock index return in China during period of 1998 2005. The optimal stopping time of prepayment depends on borrower's income and her judgment about stock market return and interest rate in the future. If borrowers are the same, i.e. they have same income flow, same information, same perception of macro economy, and same risk aversion, they will prepay at the same time. [Insert fig. 7] [Insert fig. 8] In most developed economies, researchers model mortgage contract in a contingent claim framework. The borrower's option to prepay the mortgage is an embedded call option at a strike price of par while the default option is a put option at a strike price equal to the market value of the collateral property. The prepayment option gives a mortgage borrower the right to repay the mortgage balance when its market value equals or exceeds par, i.e. the market rate drops below the mortgage coupon rate. Exercise of the call option results from two primary motivations: (1) to refinance the existing debt at a lower rate of interest; and (2) to terminate the debt through sale of the underlying asset house, due to relocation for example. However China's mortgage prepayment behavior contradicts to the conventional wisdom in the existing literature which considers the financial call option value of prepayment as an important factor. Because all the current mortgage contracts are ARMs (see Cunningham and Capone, 1990, and Stanton and Wallace, 1999, for a discussion of modeling the termination risks of ARMs). Furthermore, the mortgage rates are not indexed to any rate, but extraneously imposed by the central bank. An analog to prepayment, in U.S., mortgage default is regarded as a financial put option. If the current market value of the house, which serves as collateral of the mortgage debt, drops below the current value of the remaining mortgage balance, a borrower has an incentive to default. Therefore the default may be viewed as a put option that gives the borrower the right to sell the house to the lender at a price equal to the value of the mortgage. In the absence of transaction costs, a rational borrower can maximize her welfare by exercising the options when they are in the money. From a quick glance at China's housing prices, one can easily conclude that the housing market is booming and the financial put option is far out of the money. Then what makes the Chinese borrowers default or prepay their mortgage loans is puzzling.

15

In this paper, we rationalize borrowers' behavior by realizing that the motivation of mortgage borrowing is partly because of risk-sharing and partly due to liquidity constraints. We found that the prepayment and default are closely related to the quality of the underlying property or the credit risk of the developer. Since in our study all the defaulted loans involved a developer with only one property for sale on the market; therefore, determining the default risk of mortgage borrower is equivalent to predicting the credit risk of the developer company. We also find that pre-sale practice in China's housing market play a crucial role in mortgage defaults and prepayments. An important reason why people want to borrow money from a bank to facilitate a housing consumption is that they are facing tremendous risks in the case that developers fail to deliver the properties on the contracted delivery date or fail to meet the quality requirements specified in the pre-sale contracts. However the house buyers have to take the risk, since there are simply not enough existing houses on the market. In our sample period, every year there are about 555 new development offered in Beijing. (Note: most housing units are condos in which one property could have tens to thousands of units) on the market. There are two important institutional issues that permit such a risk sharing behavior. The residential mortgages in China are non-recourse; so far there is no personal credit system in China, and banks don't share mortgage borrowers' information. 4.2 THE COMPETING RISKS HAZARD MODEL OF PREPAYMENT AND DEFAULT The default and prepayment are options of borrowers embedded in any mortgage loans. Recent research on mortgage markets indicates that exercise behaviors of those two options are distinct but not independent. For example, exercising the default option, the borrower gives up the option to prepay the loan in the future. We apply the Cox proportional hazard model to assess the competing risks of mortgage termination by simultaneously modeling the prepayment risk and default risk. Since the early work of Dunn and McConnell (1981), Green and Shoven (1986), researchers have modeled mortgage contracts in a contingent claims framework. The borrowers option to prepay the mortgage is an embedded call option at a strike price of par while the default option is a put option at a strike price equal to the market value of the collateral property. The prepayment option gives a mortgage borrower the right to repay the mortgage balance when its market value equals or exceeds par, i.e. the market rate drops below the mortgage coupon rate. Exercise of the call option results from two primary motivations: (1) to refinance the existing debt at a lower rate of interest; and (2) to terminate the debt through sale of the underlying asset, the

16

house, due to relocation reason for example. In analogy to prepayment, mortgage default is a put option. If the current market value of the house, which serves as collateral of the mortgage debt, drops below the current value of the remaining mortgage balance, a borrower has an incentive to default. Therefore the default option gives the borrower the right to sell the house to the lender at a price equal to the value of the mortgage. In the absence of transaction costs, a rational borrower can maximize her welfare by exercising the options when they are in the money. As discussed in the section above, the special features of Chinas mortgage contract have to be taken into account when calculating the option values. Specifically the collateral of pre-sale house has to be carefully modeled for default option; while prepayment option is different in an economy where there is essentially no interest risk. The prepayment option is at work, even though there is no interest risk, because mortgage borrowers the will decide when to exercise based on the trade-off between borrowers financing cost and alternative investment returns. Furthermore, these two options compete against each other. Kau et al (1992, 1995) have discussed the theoretical relationships among the options. Schwartz and Torous (1993) is among the first to demonstrate the empirical importance. The desired model outputs are explanatory variable effects on conditional probabilities of mortgage default or prepayment. We estimate a proportional hazards model (PHM) which estimates the effects of theses variables on the time to default. In particular, we estimate the probability that a mortgage with certain characteristics will terminate in a given period where there has been no default or prepayment experience up until that period. Following Deng, Quigley, and Van Order (2000), this paper simultaneously estimates the competing risks of mortgage defaults and prepayment. The joint survival function is given in the following form:

S (t d , t p | H , R, X C , X B , Z , d , p ) = Pr{Td > t d , T p > t p | H , R, X C , X B , Z , d , p } = exp{ hdk h pk }

k =1 k =1 td tp

(1)

where hdk = h0 dk (t ) exp{ '1d g dk ( H , R, X C , X B ) + ' 2 d Z + d } h pk = h0 pk (t ) exp{ '1 p g pk ( H , R, X C , X B ) + ' 2 p Z + p }

The log integrated hazard function can be also written in the following form as:

17

lh j = 0 j + 1' j g j ( H , R, X C , X B ) + 2 j Z + 3 j I j + 4 j I j g j ( H , R, X C , X B ) + j (2)
In this formulation, S is cumulative survival probability conditional on housing price H , alternative investment return R , vector of macro-economic variables Z and vector of micro variables g ( H , R, X C , X B ) . Both Z and g are time-varying, where Z measures the systematic risk of the market and loan level information g measures the intrinsic values of the default and prepayment options. Among g, X B represents borrower characteristic and X c stands for collateral information. Ij is a dummy variable equal to unity if the loan is to finance a forward house, and zero otherwise. Td , T p are discrete random variables representing the loan life of a mortgage prior to default and prepayment respectively. hik ; i = d; p are hazard functions; Unobserved error terms associated with the hazard functions for default and prepayment are denoted by d , p respectively to represent heterogeneities. The Kaplan-Meier approach was used to fit the empirical hazard rates of prepayment and default based on the entire sample. However equation (2) assumes that the market effects of Z are the same for both forward mortgage and spot market mortgage. It also assumes that the error structures are the same for all mortgage loans. To accommodate this restriction, a more complete separation regression for forward market and for spot market can be estimated similar to Blinder (1973) and Oaxaca (1973). For Spot market without pre-sale

lh S = 0 j + 1' j g S ( H , R, X C , X B ) + 2 j Z + S j j j
= jG S + S j j
For Forward market with pre-sale

(3)

lh F = 0 j + 1' j g F ( H , R, X C , X B ) + 2 j Z + F j j j
= jG F + F j j

(4)

where G ij = [ g 'j ( H , R, X C , X B ) Z ]' is a column vector of repressors; j = d; p and i = S; F ; = 0 j 1' j 2 j , = 0 j 1' j 2 j are regression coefficients. The log Hazard Ratio, denoted by LHR between two groups of F and S can be decomposed into two parts.

18

LHR j = G F G S = (G F G S ) + G F ( )

(5)

where the upper bar associated to a variable indicates the mean and estimated coefficients. The first part of risk differential is due to differences in the typical loan characteristics of forward versus spot (G jF G jF ) , assuming they are valued at the same spot market. The second part of the risk differential reflects the portion of the higher risk on forward market loans that arises because they are priced differently than otherwise identical mortgages for existing house financing. The default options in pre-sale period versus post-sale period are total different. In the post-sale period the default option is related intensively to the housing price. However in the presale period, the default option is mainly depend on the probability of default of developer. Following Deng, Quigley and Van Order (2000), the proxy values of exercising the put option for mortgage borrower in post-sale period is measured by the probability of negative equity. Typically one cannot estimate directly to which the default option is in the money without knowing the entire path of individual house values. However, we can use the initial loan-to-value ratio and the diffusion process of house prices to estimate the critical value for borrower exercise of put option, known as the probability of negative equity. Specifically, the variable put_proxy is defined as:

Put_Proxy = N (
Where

logVr logVH )

(6)

Vr =
i =1

T k

M +k (1 + r + k ) i

(7)

VH = P (

H +k ) H

(8)

Vr is the market value of mortgage debt; M is monthly payment of mortgage principle and interest, r + k is adjustable mortgage rate for loan originated at time and after the seasoning period of k. T is the mortgage term. VH is the market value of the house which is purchased at

with price P and valued at +k ; H is housing price index at time in the area. The term in
parentheses of equation (8) follows a log normal distribution. N () is the cumulative standard

19

normal distribution function, is housing index volatility 8 . The calculation of Vr is a little complicated. It involves firstly calculate annually re-setting monthly payment of principle and interest. The intrinsic value of call option is defined as:

Call_Proxy =

Vr VR Vr

(9)

where Vr defined previously is the market value of mortgage, the cost of financing a house purchase, and VR is the value of a hypothetical income, the return from alternative investment. Since stock market is a major investment alternative, the Shanghai stock price index is used in return calculation. Specifically, VR is defined as:

VR =
i =1

T k

M + k (1 + R + K ) i (1 + r f ) i

(10)

In addition to intrinsic value of options, three sets of non-option related variables are included in the regression. Those variables include time varying and time invariant determinants of mortgage performance motivated by recent research. The first set of variables, denoted by XB, describe the borrowers characteristics, including gender, age, education, log value of monthly income from borrowers and their spouses, occupation, marital status, number of dependents in borrowers household, etc. The second set of variables, denoted by XC, describes the collateral information. The developer related information includes credit ratings, type of developer, age of the developer, equity value of firm at registration, size of developer, ratio of certified professionals to total employees, and number of projects, etc. Other loan related information includes original loan amount, down payment rate, loan-to-value ratio at origination, year and month, maturity, size of the house, location of property, sale type, appraisal value of property at time of sale, unit price, etc. 5. EMPIRICAL RESULTS The risks analyses are based on the full sample from the combined dataset of bank loans and real estate developer credit database. The competing risks of default and prepayment are

The housing price indices and their volatilities are estimated according to the three stage procedure originally created by Case and Shiller (1989) and modified by Quigley and Van Order (1995).

20

estimated jointly. Table 6 presents three variations of the competing risks model of loan termination. All specifications incorporate both market variables and a rich set of loan characteristics including borrowers characteristics and collateral information. The market variables include time-varying proxies for local economic conditions, such as inflation, unemployment rate, the slope of term structure of interest rate, stock market returns, etc. Each of the three models contains separate flexible baseline functions for default and prepayment that follow Han and Hausman (1990). Model 1 tests both default rate and prepayment rate equations following equation (2) where the presale dummy is estimated in hazard regression framework and the interaction term 4 is constrained to be zero. Further, the model does not control directly for the intrinsic value of call and put options in the estimation. Model 1 provides a benchmark for the competing risks specifications discussed below. Model 2 extends model 1 by including the contemporaneous values of options in both default and prepayment equations. Model 3 further extends model 2 by including interaction terms of presale with call option and put option in both risk equations. Overall, the competing risks models are well-specified and control for approximately 40 different characteristics of the loan, borrower and collateral characteristics and market information. [Insert Table 6] As evidenced in model 1, estimation results indicate that economic conditions of the market have important impacts to default and prepayment behaviors. As a proxy for overall economy, the Shanghai Stock Index is negatively associated with default and prepayment. The increase in local unemployment rates positively affects the exercise of the default option and prepayment option. This result is highly significant across model specifications. Unemployment rate is a macro variable indicating the strength of the macro economic environment. It also reflects Chinese borrowers confidence about the job-stability and financial soundness. For most Chinese households, housing is a basic consumption. When facing uncertainty about future wealth, they will choose to invest in housing, rather than risky stocks and bonds. This is quite different from US, where prepayment risk is negatively associated with unemployment rate. The slope of term structure of interest rate, defined as the difference between the five year CD rate and spot rate disclose the expectation of future interest rate. When the yield curve becomes fatter, borrowers in China may prefer to prepay the mortgage debt. As show in model 1 while the relationship between default behavior and slope of term structure is statistically insignificant, the prepayment rate is negatively associated with the slope of term structure and statistically significant.

21

The estimates from model 1 suggest that the loan to value ratio at origination is positively associated with default risk and negatively associated with prepayment risk. Higher LTV reflects that the purchaser is more liquidity constrained and has less equity invested in the house. Model1 also shows that the likelihoods of default and prepayments vary positively with housing-expense burdens proxied by mortgage expense to income ratio. Model 1 also estimates the affects of borrowers characteristics, including borrowers age, sex, marital status, education background, occupation and job positions, log monthly household income and number of dependents in the household. Among these categorical or continuous variables, household income is an important factor in determining default and prepayment risk. High income borrowers are less likely to default and more likely to prepay. The log income is highly significant across 3 models. Another continuous covariate that is statistically significant is number of dependent in a household. The more the dependents in the household, the higher burden of the mortgage borrower, thus the higher the default probability and the lower prepayment risk. Among the categorical covariates, the sex is insignificant; age is significant among these 3 models. Older borrowers have relatively lower default risk and higher prepayment risk. As younger people in China are more liquidity constrained and also prefer to smooth consumption by borrowing, reflecting a difference in lifestyle preference in China. Marital status is another significant factor in determining default rate and prepayment rate. Married borrowers have both lower default rate and prepayment rate reflecting that a family is a more stable and responsible social unit than singles. Occupations and job positions are significantly associated with prepayment risk. Business and Trade industry shows higher prepayment risk due to their relatively volatile income streams, while research and education sector has lower risk of prepayment because of the nature of relatively stable income streams. Managers exhibit relatively higher prepayment risk than technicians and self-employee. On the other hand, none of them is statistically significantly associated with defaults. Borrower with college degree has lower default risk but higher prepayment risk. In addition to borrowers characteristics, model 1 also finds significant impact of collateral information to mortgage termination in China. Specifically, mortgages with underlying property built by bigger or older developer exhibit lower default risk. There are two measures of developer size, one is currency adjusted value of registered equity, and the other is total number of employee in the firm. These two size variables are all statistically significantly associated with default risk, while the association with prepayment risks is insignificant. New developers tend to associated with higher default risk and higher prepayment risk of mortgage loans. In the sample

22

period of the study, Chinese housing sector is experiencing tremendous development which triggers more company enter into real estate business. While it is hard for the investors to monitor and judge the quality of the property they are developing, established developers have operational advantage in this market. Not only the developer quality affect property price, but also have impact to financial market via the channel of mortgage termination. The developers who involved in residential real estate development for more than 20 years are associated with lower default risk and higher prepayment risk. The ratio of certified professionals to total employee of the developer is also significantly associated with mortgage default and prepayment rates. The higher the professional ratio is associated with lower default risk and higher prepayment risk. A strong developer tends to have more certified professionals and have passed higher industry standard (for example ISO9000) and also prefer to take long-term commitment. Since a developer with good quality tends to have better cost management, quality control and reputation concern, its presale house will have higher probability of failure of delivery. Therefore the mortgages associated with good quality developers will have lower default risk and higher prepayment rates. The type of developers also associates with different default risk and prepayment risk. The foreign joint venture developer exhibits lower default risk and higher prepayment risk, while the joint venture with Hong Kong, Macao and Taiwan developer are associated with higher default risk. Comparing with State-Owned developers, the private developers and limited company have both higher default risk and higher prepayment risks. The location of property is also associated with different default and prepayment risks. In central city area, due to competition and tighter regulation or monitoring, mortgages with underlying property in central city area tend to have lower default risk and higher prepayment risk. Model 2 extends model 1 through the introduction of the option-related time-varying covariates into both the default and prepayment equations.9 The estimates confirm that the put option, measured as the probability of negative equity, is positive and highly significant in the exercise of the default option; the call option value is also positively and highly significant in the exercise of the prepayment option. In other words, increase in the probability of negative equity, the incidence of default increases dramatically. On the other hand, declines in returns from alternative investment that bring the call option "in-the-money" will lead to a high prepayment activities. Stock market in China is a fast growing investment vehicle beyond traditional deposit. Investment in stock market has been attached more and more importance in Chinese peoples financial decision. The intrinsic value the prepayment option basically reflects investors portfolio
9

Log Likelihood and Schwarz-Bayesian Criterion (SBC) reported at the bottom of the table provide comparison of the goodness of .t among alternative models. Models with lower values are considered preferable.

23

choice. Model 2 further indicates that a value of call option is negatively associated with default risk. In a bull stock market, household may relocate their asset from housing to stock market by stopping mortgage payment. Model 3 further include the interaction term between presale and option values to test whether the intrinsic value of options in forward market have different influence than on the spot market. The default behavior is less sensitive to the put option and call option for mortgage loans with pre-sale properties. For prepayment behavior, the call option and put option are less sensitive to pre-sale mortgage loan borrowers. This means that the pre-sale in apparently an important factor that highly correlated with default risk and prepayment risk. If the loan is associated with forward housing purchase, it is such a bad news for default and prepayment risks that option value (both call and put) do not add as much to the risks as in spot market. The significance in the product terms indicate that the default risk and prepayment risks exhibit quite differently in forward market for the pre-sale houses than on spot market. In a similar fashion, product terms between presale and other borrower characteristics and collateral information can be tested. Since mortgages on presale houses exhibit strikingly different default and prepayment behaviors, a more complete separation of hazard regressions are performed towards forward market and spot market respectively. Table 7 displays the regression results for the two markets. On forward market, while intrinsic value of call option is significant for exercises of default option and prepayment option, probability of negative equity is associated with neither default option nor prepayment option in statistical sense. The value of mortgage default option has totally different underlying among pre-sale houses, versus completed projects. On the forward market, housing price index has little to do with default risk, since the collateral for presale houses is on paper. Therefore the probability of negative equity is more closely associated with houses on spot market, while houses on forward market depend more on probability of house delivery. Another remarkable feature on the separate regressions is that borrowers characteristics and developer information play different roles on spot market versus forward market. The borrower characteristics (age, marital status, education, occupation and job position) are only statistically significant on spot market, not significant on forward market. On the other hand the developer characteristics (number of projects, size, type and history of the developer) are only statistically significantly associated with mortgage termination for presale loans, none of them is significant on spot market. For example, younger borrowers, singles, elementary school education group and clerk, social service as well as self-employment group are associated with higher default risk, while older borrowers, singles, college education and

24

managers are associated with higher prepayment risk. However none of these borrower characteristics is statistically associated with mortgage termination risks on forward market. On the other hand, good credit quality of developer, e.g. completed more projects in the past, bigger and experienced developers, higher ratio of certified professional to total employee, foreign joint ventures, is associated with lower default risk and higher prepayment risk. None of these developer characteristics is statistically significant in the spot market regression. Although the sensitivities are different, some loan variables, location variable and market condition share the same sign across the spot market and forward market. For instance, on both markets, the more dependents in a household, the higher the default risk, the lower the prepayment risk due to households liquidity constraints. Central city location enjoys lower default risk and higher prepayment risk. [Insert Table 7] To demonstrate the mortgage prepayment and default risks with and without embedded forward contract risks, we simulate 300 paths of house price appreciation rates and stock returns according to a joint stochastic mean-reverting process. These 300 randomly sampled paths are applied to the prepayment and default functions reported by Model 3 in table 6 to compute the monthly prepayment and default risks associated with hypothetical 0.5 million RMB mortgage. We compare two hypothetical mortgage pools characterized by two risk structures between spot versus forward market mortgage borrowers. The present values of the cash flows from the two groups allow us to calculate risk premium on the forward presale housing market. 10 Table 8 reports a simulated result on forward risk premium which is about 250 basis points per year. In other worlds, there is a 2.5% risk premium on forward market relative to the spot market in China. [Insert Table 8] 6. CONCLUSION This paper is the first to realize the mortgage risks on forward presale housing market. This finding indicates that borrower characteristics and collateral information are both important to determine the mortgage termination risks. Currently most Chinese banks price the consumer loan based on borrower characteristics only. The results of this paper indicate a potential improvement in banks mortgage risk modeling. In addition, the mortgage borrowers prepayment and default behavior in the forward housing market is significantly different from those in the spot housing market. The study found that the cross-subsidy between the forward and
10

This is similar to Deng and Gabirel (2006).

25

spot housing market in China can be as high as 250 basis points. The finding of this study will provide valuable insight about emerging housing and mortgage markets in China as well as those in other transition economy.

26

Figure 1. Residential Mortgage Market in China

2500

Residential Mortgage Market in China 12.00% ( 98% AAGR since 1997) 11.27% 2,000 12.00%
9.27% 8.59% 1,592 1,330 10.00% 8.00% 6.00% 825 560 4.00% 2.00% 0.00% 1997 1998 1999 2000 2001 Year Total Mortgage Outstanding Ratio of Mortgage to Total Bank Loans 2002 2003 2004 2005

14.00%

2000 Billion RMB Yuan

1500

6.94% 5.07%

1000

500 1.07% 0.39% 19 0 43

2.14% 136

338

Source: Chinese Banking Regulation Committee. Figure 2. International Comparison of Mortgage to Consumer Loan Ratio
International Comparison of Consumer Loans (2003)
80.0% 70.0% Percentage ofTotal Loans 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0%
1.4% 0.0% 7.5% 3.7% 2.0% 9.1% 26.3% 29.5% 3.7% 2.7% 6.7% 3.3% 4.9% 2.6% 16.0% 1.3% 2.5% 34.3% 38.8% 27.3% 22.6% 12.0% 3.9%

8.7% 2.4% 28.4% 59.2%

China

Thailand

Malaysia

Taiwan

HongKong

Singapore

U.S.

Korea

U.K.

mortgage

credit-card loan

other consumer loan

Source: Goldman Sachs

27

Figure 3. International Comparison of Loan to GDP Ratio

International Comparison of Loan-GDP Ratio (2003)
80.0% 70.0% 60.0% Percentage of GDP 50.0%
6.3% 7.1% 4.0% 20.3% 32.6% 11.7% 3.6%

40.0% 30.0% 20.0%

2.1% 4.3% 3.4% 3.7% 2.7%

4.0%

1.5% 3.0% 57.5% 50.0%

36.6% 26.1%

34.7% 25.5%

10.0% 0.0%

3.0% 1.6% 7.2%

0.0% 11.4%

19.2%

Thailand

China

U.S.

Malaysia mortgage

Taiwan

Singapore

Korea

HongKong

U.K.

credit-card loan

other consumer loan

Source: Goldman Sachs Figure 4. International Comparison of Mortgage to Income Ratio

International Comparison of Mortgage-Income Ratio (2002)
200% 180% 160%
Percentage of Disposable Income

174%

140% 120% 100% 80% 60% 40% 20% 0%

India Thailand China Taiwan Malaysia HongKong U.S . Korea

114% 89% 75% 55% 30% 17% 7% 94%

123%

123%

Japan

U.K.

Singapore

Source: Goldman Sachs

28

Figure 5. Supply and Sales in the Residential Housing Market in China

Supply and Sales of Residential Housing in China
400 350 Million Square Meters 1.31 300 250 200 0.8 150 100 50 0 1997 1998 1999 2000 Year Housing Square Meters Completed Supply-Sales Ratio of Residential Housing Housing Square Meters Sold 2001 2002 2003 2004 0.6 0.4 0.2 0 1.58 1.36 1.25 1.22 1.2 1.13 1.03 1.8 1.6 1.4 1.2 1

Source: Goldman Sachs

29

Simple Model for Forward v.s. Spot Housing Market in China

Pre-Sale Period T=0 Transaction Date T=1 House Delivery Date T=3 Mortgage Maturity 1. Payment of Last mortgage payment 2. Consume Housing

1. Forward Housing Market

Time Line

Customer 1.Purchase House on Forward

Market 2. Pay in full to Developer with Downpayment+Mortgage

1. Take Delivery of pre-sale House and Pay Mortgage to Bank 2. Default Mortgage if Developer Fails to Deliver Pre-sale House 3. Consume Housing

T=2 Option Exercise Date 1. Prepay if Financing Cost>Investment Return 2. Default if Value of House<Value of Mortgage 3. Not Exercise Prepayment Option if Borrower is Constrained 4. Consume Housing

Developer 2. Deposit in Bank

3. Develop Pre-Sale House

1. Receive House Price in Full 1. Deliver Pre-Sale House in Good Market 2. Default in Bad Market 3. Sell Completed House on Spot Market

Bank

1. Mortgage Lending Based on Borrower Characteristics 2. Take Deposit from Developer

Figure 6. Market Microstructure of Forward and Spot Housing Market in China

30
T=1 Transaction/Delivery Date 1. Purchase House on Spot Market 2. Pay in full to Developer with Downpayment+Mortgage 3. Consume Housing 1. Receive House Price in Full 2. Deposit in Bank 1. Mortgage Lending Based on Borrower Characteristics 2. Take Deposit from Developer

1. New ARM Rate Announced 2. Prepayment Risk, Receive Payment 1. Receive Last Mortgage Payment 1. Receive Mortgage, if Borrower does not at Par 3. Receive Residual Value of House in Default Case of Borrower Default 2. Receive Residual Value of house if 4. Receive Mortgage Payment if Borrower Defaults Option not Exercised

2. Spot Housing Market

T=2 Option Exercise Date 1. Prepay if Financing Cost>Investment Return 2. Default if Value of House<Value of Mortgage 3. Not Exercise Prepayment Option if Borrower is Constrained 4. Consume Housing T=3 Mortgage Maturity 1. Payment of Last 2. Consume Housing

Time Line

Customer

Developer

Bank

1. New ARM Rate Announced 1. Receive Last Mortgage Payment 2. Prepayment Risk, Receive Payment at Par 3. Receive Residual Value of House in Case of Borrower Default 4. Receive Mortgage Payment if Option not Exercised

Figure 7. Mortgage Rates in China

11 10 9 8 7 6 5 4 3 2 1 0 Jun-98 Nov-98

Mortgage Rates in China

Jun-03

Aug-02

Nov-03

Jan-98

Jul-00

Jan-03

Feb-00

May-01

Oct-01

Apr-99

Mar-02

Sep-99

Dec-00

Sep-04

Feb-05

Apr-04

Jul-05
38169

short

mortgage rate (<5y)

date mortgage rate (>5y)

spread

Source: Chinese Banking Regulation Committee.

Figure 8. Stock Index Return in Shanghai, China

Shanghai Stock Index Return 80% annual return % 60% 40% 20% 0% 35796 35886 35977 36069 36161 36251 36342 36434 36526 36617 36708 36800 36892 -20% -40% date stock index return n 36982 37073 37165 37257 37347 37438 37530 37622 37712 37803 37895 37987 38078 38261

Source: Shanghai Stock Exchange

31

Dec-05

Table 1. Descriptive Statistics for Mortgage Loans Frequency of Loans by Loan Category and by Payoff Types Variable Defaulted Prepaid Other All Loans Origination Year 1998 29 84 200 313 (9.27) (26.84) (63.90) (0.30) 1999 10 905 3511 4426 (0.23) (20.45) (79.33) (4.28) 2000 392 3117 15778 19287 (2.03) (16.16) (81.81) (18.64) 2001 372 3165 21972 25509 (1.46) (12.41) (86.13) (24.66) 2002 529 1644 22132 24305 (2.18) (6.76) (91.06) (23.49) 2003 52 1140 28430 29622 (0.18) (3.85) (95.98) (28.63) Total 1384 10055 92023 103462 (1.34) (9.72) (88.94) (100.00) Loan to Value Ratio LTV<60 309 2608 20232 23149 (1.33) (11.27) (87.40) (22.37) 60<LTV<70 281 2538 22252 25071 (1.12) (10.12) (88.76) (24.23) LTV > 70 794 4909 49539 55242 (1.44) (8.89) (89.68) (53.39) Original Loan Amount 508 3732 31938 36178 OLA < 200,000RMB (1.40) (10.32) (88.28) (34.97) 430 2960 27892 31282 200,000 < OLA < 400,000 (1.37) (9.46) (89.16) (30.24) 446 3363 32193 36002 OLA > 400,000RMB Market of Transaction Spot Market 141 628 26076 26845 (0.53) (2.34) (97.14) (25.95) Forward Market 1385 9427 65805 76617 (1.81) (12.30) (85.89) (74.05)

32

Table 2. Descriptive Statistics for Mortgage Loans Frequency of Loans by Colleteral/Developer Category and by Payoff Types Variable Defaulted Prepaid Other All Loans Rating Group 1 8 543 3536 4087 (0.01) (0.63) (4.13) (4.77) 2 1 1159 13404 14564 (0.01) (7.96) (92.04) (17.01) 3 1 500 4255 4756 (0.02) (10.51) (89.47) (5.55) 4 742 65 578 1385 (53.57) (4.69) (41.73) (1.62) unrated 620 4740 55479 60839 (1.02) (7.79) (91.19) (71.05) Developer Type SOE 123 1310 12462 13895 (0.89) (9.43) (89.69) (16.22) JV_HMT 621 258 2245 3124 (19.88) (8.26) (71.86) (3.65) JV_F 68 404 4175 4647 (1.46) (8.69) (89.84) (5.42) LIMITED 380 4675 54084 59139 (0.64) (7.91) (91.45) (69.03) PRIVATE 192 359 4311 4862 (3.95) (7.38) (88.67) (5.68) History of Developer Less than 5y old 4 2700 24598 27302 (0.01) (9.89) (90.10) (32.33) Longer than 5y old 1334 4132 51672 57138 (2.33) (7.23) (90.43) (67.67) Currency Type of Equity USD 687 6447 71700 6833 (0.87) (8.18) (90.95) (7.98) RMB 1384 7006 77277 78834 (1.62) (8.18) (90.21) (92.02) Valid Period of License Short 1034 1582 16325 18941 (5.46) (8.35) (86.19) (24.30) Long 321 4579 54115 59015 (0.54) (7.76) (91.70) (75.70) (continued)

33

Table 3. Descriptive Statistics for Mortgage Loans Frequency of Loans by Colleteral/Developer Category and by Payoff Types Variable Location Outer City Inner City Quality Control Standard Non Pass Positive Review by R.E Non Good Dev and Project in same District Different District Same District ne or two Missing Internal Monitoring Strong Weak Quality Control Standard Non Pass Positive Review by R.E Non Good Professional Ratio Low High Defaulted 244 (2.04) 1140 (1.25) 1375 (1.40) 9 (0.17) 1376 (1.33) 8 (0.01) 1246 (1.68) 18 (0.06) 120 (0.12) 778 (1.62) 606 (1.61) 1375 (1.40) 9 (0.17) 1376 (1.33) 8 (0.01) 757 (3.86) 616 (0.93) Prepaid 1702 (14.20) 8353 (9.13) 9445 (9.61) 610 (11.83) 9252 (8.94) 803 (0.78) 7752 (10.45) 2178 (7.75) 79 (0.08) 3633 (7.57) 3379 (8.96) 9445 (9.61) 610 (11.83) 9252 (8.94) 803 (0.78) 1690 (8.61) 5306 (8.05) Other 10044 (83.77) 81979 (89.62) 87484 (88.99) 4539 (88.00) 84888 (82.05) 7135 (6.90) 65214 (87.88) 25915 (92.19) 701 (0.68) 43576 (90.81) 33726 (89.43) 87484 (88.99) 4539 (88.00) 84888 (82.05) 7135 (6.90) 17177 (87.53) 59998 (91.02) All Loans 11990 (11.59) 91472 (88.41) 98304 (95.01) 5158 (4.99) 95516 (92.32) 7946 (7.68) 74212 (72.53) 28111 (27.47) 900 (0.87) 47987 (56.00) 37711 (44.00) 98304 (95.01) 5158 (4.99) 95516 (92.32) 7946 (7.68) 19624 (22.94) 65920 (77.06) (continued)

34

Table 4. Descriptive Statistics for Mortgage Loans Frequency of Loans by Category and by Payoff Types(Continued)

Variable Developer Size Employee Employee < 30 Employee >= 30 Registered Equity (Currency Exchange adj. ) Equity < 30 Million RMB Equity >=31 Million RMB Number of Projects Single Property Multiple Properties

Defaulted

Prepaid

Other

All Loans

1318 (3.07) 55 (0.13)

3333 (7.76) 3665 (8.60)

38314 (89.17) 38875 (91.27)

42965 (50.22) 42595 (49.78)

1268 (3.46) 57 (0.12) 1373 (2.10) 11 (0.03)

2872 (7.84) 4140 (8.45) 6592 (10.07) 3417 (9.05)

32497 (88.70) 44805 (91.44) 57495 (87.83) 34335 (90.92)

36637 (42.78) 49002 (57.22) 65460 (63.42) 37763 (36.58)

35

Table 5. Descriptive Statistics for Mortgage Loans Frequency of Loans by Borrower Category and by Payoff Types

Variable Marital Status Married Single Education College Secondary Sch. Primary Sch. Age Cohort Age < 40 Age > 40 Income Groups Monthly Income < 8888RMB Monthly Income >8888RMB Occupation Business/Trade Social Service Self-Employment R&D Others Job Position Manager Others Technician Clerk

Defaulted 562 (1.21) 822 (1.44) 843 (1.30) 482 (1.38) 59 (1.60) 1036 (1.29) 348 (1.49) 1006 (1.31) 378 (1.43) 248 (1.11) 130 (1.33) 602 (1.42) 80 (0.81) 324 (1.69) 768 (1.22) 214 (10.77) 168 (1.21) 234 (1.14)

Prepaid 4277 (9.24) 5778 (10.10) 6432 (9.92) 3254 (9.32) 369 (9.99) 6574 (8.21) 3481 (14.86) 6623 (8.60) 3432 (12.99) 1854 (8.27) 849 (8.72) 4960 (11.72) 773 (7.82) 1619 (8.47) 6184 (9.86) 607 (16.76) 1256 (9.03) 2008 (9.77)

Other 41440 (89.54) 50583 (88.46) 57563 (88.78) 31195 (89.30) 3265 (88.41) 72420 (90.49) 19603 (83.66) 69403 (90.10) 22620 (85.58) 20303 (90.62) 8759 (89.95) 36746 (86.85) 9038 (91.38) 17177 (89.84) 55766 (88.92) 5457 (172.48) 12482 (89.76) 18318 (89.10)

All Loans 46279 (44.73) 57183 (55.27) 64838 (62.67) 34931 (33.76) 3693 (3.57) 80030 (77.35) 23432 (22.65) 77032 (74.45) 26430 (25.55) 22405 (21.66) 9738 (9.41) 42308 (40.89) 9891 (9.56) 19120 (18.48) 62718 (60.62) 6278 (6.07) 13906 (13.44) 20560 (19.87)

36

Table 6 Maximum Likelihood Estimates for Competing Risks of Chinese Adjustable Rate Mortgage Prepayment and Default
Model 1 Default Fraction of Contract Value (Call Option) Probability of Negative Equity (Put Option) Presale (dummy) Interaction of Presale and Call Option Interaction of Presale and Put Option age > 40 (dummy) female (dummy) married (dummy) occupation: business and trade (dummy) occupation: social service (dummy) occupation: others (dummy) occupation: R&D (dummy) education: secondary school (dummy) education: college (dummy) job position: manager (dummy) job position: self-employment (dummy) job position: technician (dummy) number of dependents log monthly income -0.283 (4.65) -0.044 (0.82) -0.017 (1.29) -0.077 (0.84) 0.070 (0.70) -0.141 (1.78) 0.017 (0.15) -0.102 (0.78) -0.149 (2.63) -0.021 (0.24) 0.994 (1.29) 0.013 (0.12) 0.229 (20.63) -0.100 (2.60) 0.105 (4.73) 0.009 (0.44) -0.057 (2.85) 0.051 (1.96) 0.013 (2.33) 0.060 (2.15) -0.033 (2.84) -0.083 (1.55) 0.119 (5.37) 0.045 (2.71) -0.042 (1.92) -0.063 (1.98) -0.568 (50.02) 0.130 (6.62) -0.280 (4.54) 0.010 (0.18) -0.027 (1.42) -0.095 (1.03) 0.943 (0.44) -0.180 (0.23) 0.040 (0.34) -0.110 (0.83) -0.158 (2.78) -0.019 (0.22) 0.943 (0.85) 0.031 (0.29) 0.224 (20.09) -0.103 (1.63) 0.111 (4.96) 0.020 (1.02) -0.064 (3.15) 0.050 (2.33) 0.018 (1.95) 0.047 (2.62) -0.024 (2.17) -0.046 (0.86) 0.108 (4.86) 0.036 (2.18) -0.041 (2.82) -0.049 (1.95) -0.550 (48.49) 0.123 (6.07) 1.058 (2.56) 1.118 (28.10) Prepay Default -2.616 (11.93) 3.967 (5.69) 1.013 (1.96) Model 2 Prepay 2.722 (30.49) 2.105 (13.13) 1.097 (27.56) Default -3.017 (5.80) 3.533 (9.55) 1.090 (2.88) -1.937 (7.42) -3.103 (7.54) -0.245 (3.93) 0.007 (0.13) -0.058 (1.97) -0.022 (0.24) 0.054 (0.53) -0.106 (1.32) -0.136 (1.15) -0.056 (0.42) -0.127 (2.23) -0.025 (0.29) 0.836 (1.58) 0.082 (1.76) 0.213 (18.67) -0.152 (2.40) Model 3 Prepay 4.618 (21.45) 5.377 (10.04) 1.020 (12.74) -2.029 (8.86) -3.462 (6.45) 0.103 (4.61) 0.022 (1.12) -0.067 (3.28) 0.045 (1.94) 0.013 (0.33) 0.045 (1.61) -0.022 2.06 0.044 (0.82) 0.106 (4.79) 0.036 (2.36) -0.040 (1.88) -0.051 (1.45) -0.549 (48.40) 0.122 (6.03)

37

multiple project (dummy) size (in 10mill. currency adj.) number of employee license longer than 20y (dummy) new developer (dummy) developer type: joint venture with foreign developer (dummy) developer type: joint venture with HK Macau, TW (dummy) developer type: private developer (dummy) developer type limited co.(dummy) district central city location (dummy) ratio of certified professionals Loan To Value Ratio Housing Exp. To Income log loan amount stock market slope term structure unemployment

-1.829 (10.68) -0.104 (6.31) -0.063 (15.82) -0.458 (3.86) 2.431 (9.80) -0.936 (7.16) 0.834 (8.59) 0.196 (1.14) 0.504 (5.03) 0.399 (6.24) -1.306 (9.55) -4.501 (9.55) 0.068 (1.70) 0.578 (9.95) 0.129 (0.99) 0.002 (1.79) -0.494 (2.16) 0.059 (7.69) 165089 165545

0.077 (3.35) 0.000 (0.79) 0.000 (1.58) 0.152 (6.89) 0.123 (5.13) 0.056 (1.14) -0.059 (0.87) 0.158 (2.68) 0.094 (2.36) 0.113 5.05 0.086 (3.38) 0.256 (5.38) -0.043 (3.25) -0.363 (12.68) -0.117 (1.87) -0.005 (39.45) -1.549 (48.74) 0.081 (48.20)

-1.748 (10.08) -0.099 (5.96) -0.060 (15.25) -0.426 (3.57) 2.499 (9.83) -0.987 (7.37) 0.844 (8.63) 0.295 (1.72) 0.451 (4.48) 0.352 (5.45) -1.224 (8.93) -4.402 (18.38) 0.046 (3.08) 0.537 (9.08) 0.047 (0.72) 0.002 (0.50) -0.762 (3.11) 0.046 (5.44) 164254 164747

0.119 (5.14) 0.000 (0.31) 0.000 (1.43) 0.105 (4.72) 0.117 (4.89) 0.130 (2.68) -0.049 (0.72) 0.157 (2.66) 0.101 2.56 0.108 (4.83) 0.084 (3.30) 0.269 (5.67) -0.073 (4.11) -0.352 (12.43) -0.190 (1.36) -0.005 (39.50) -1.512 (47.30) 0.077 (46.21)

-1.868 (2.40) -0.099 (6.44) -0.061 (15.37) -0.502 (4.15) 2.457 (9.57) -0.934 (6.96) 0.800 (8.20) 0.299 (1.78) 0.506 (4.99) 0.402 (6.17) -1.271 (9.18) -4.470 (18.51) -0.003 (0.06) 0.360 (5.89) -0.061 (0.92) 0.002 (1.43) -0.649 (2.65) 0.053 (6.27) 164107 164638

0.116 (5.03) 0.000 (0.51) 0.000 (1.00) 0.103 (4.64) 0.118 (4.94) 0.128 (2.63) -0.039 (0.58) 0.150 (2.55) 0.105 (2.65) 0.110 (4.93) 0.085 (3.35) 0.271 (5.71) -0.074 (4.12) -0.343 (12.13) -0.191 (1.53) -0.005 (39.36) -1.511 (47.28) 0.076 (46.01)

Log Likelihood SBC

38

Table 7 Maximum Likelihood Estimates for Competing Risks of Chinese Adjustable Rate Mortgage Prepayment and Default
Model 4: Spot Market Default Prepay Fraction of Contract Value (Call Option) Probability of Negative Equity (Put Option) age > 40 (dummy) female (dummy) married (dummy) occupation: business and trade (dummy) occupation: social service (dummy) occupation: others (dummy) occupation: R&D (dummy) education: secondary school (dummy) education: college (dummy) job position: manager (dummy) job position: self-employment (dummy) job position: technician (dummy) number of dependents log monthly income multiple project (dummy) size (in 10mill. currency adj.) number of employee -0.278 (1.45) 9.064 (2.35) -0.454 (1.99) -0.076 (1.32) -1.552 (4.08) -0.699 (1.83) 0.932 (2.37) -0.579 (1.60) 0.064 (0.16) -0.879 (2.06) -0.457 (2.82) -0.001 (0.00) 0.819 (1.82) 0.270 (0.54) 0.799 (6.45) -0.200 (2.10) -0.153 (0.51) -0.025 (1.46) -0.004 (0.72) 4.048 (9.98) 2.781 (2.05) 0.127 (2.38) -0.013 (1.16) -0.202 (2.43) 0.142 (1.04) 0.083 (0.51) 0.135 (1.10) -0.118 0.71 -0.135 (0.57) 0.118 (2.25) 0.028 (2.24) -0.134 (1.72) 0.045 (0.29) -0.741 (14.18) 0.235 (2.88) -0.065 (0.68) 0.004 (1.74) 0.000 (0.10) Model 4: Forward Market Default Prepay -2.456 (9.52) 2.762 (0.88) -0.198 (1.95) 0.023 (0.41) -0.137 (1.17) -0.023 (0.24) -0.008 (0.08) -0.137 (1.63) -0.081 (0.64) 0.007 (0.05) 0.039 (0.66) 0.056 (0.62) 0.857 (1.29) 0.071 (0.64) 0.220 (14.67) -0.109 (1.68) -2.702 (8.74) -0.220 (8.32) -0.072 (14.12) 2.669 (28.76) 2.075 (0.56) 0.107 (1.65) 0.025 (1.22) -0.054 (1.60) 0.038 (1.17) 0.002 (0.04) 0.033 (1.14) 0.013 (0.31) -0.040 (0.73) -0.097 (1.24) -0.047 (1.70) -0.054 (1.15) -0.058 (1.60) -0.533 (45.82) 0.120 (5.72) 0.131 (5.49) 0.000 (0.04) 0.000 (2.99)

39

license longer than 20y (dummy) new developer (dummy) developer type: joint venture with foreign developer (dummy) developer type: joint venture with HK Macau, TW (dummy) developer type: private developer (dummy) developer type limited co.(dummy) district central city location (dummy) ratio of certified professionals Loan To Value Ratio Housing Exp. To Income log loan amount stock market slope term structure unemployment

-0.082 (0.27) 0.159 (0.32) 18.082 (0.00) 19.330 (0.00) 18.793 (0.00) 18.078 (0.00) 0.219 (0.52) -0.078 (2.18) 0.003 (0.00) 0.967 (2.94) 1.215 (7.16) 1.530 (5.35) 0.010 (0.17) 3.627 (0.00) 1.829 (0.01) 7463 7803

-0.007 (0.08) 0.034 (0.34) 0.027 (0.14) 0.344 (1.33) 0.026 (0.11) 0.051 (0.31) 0.208 (2.22) 0.057 (2.55) -0.017 (0.09) -0.127 (1.69) -0.790 (5.95) -0.254 (3.13) -0.008 11.84 -1.555 (10.14) 8.824 (2.40)

-0.254 (1.77) 3.636 (6.23) -0.863 (6.40) 0.681 (6.94) 0.324 (1.86) 0.435 (4.08) 0.237 (3.53) -2.071 (8.40) -4.742 (17.84) -0.046 (0.91) -0.006 (0.08) -0.201 (3.01) 0.002 (6.06) -0.566 (2.28) 5.514 (6.46) 151198 151678

0.104 (4.51) 0.117 (4.75) 0.140 (2.78) -0.057 (1.80) 0.162 (2.64) 0.109 (2.66) 0.109 (4.73) 0.085 (3.27) 0.289 (5.90) -0.073 (3.96) -0.330 (11.37) -0.186 (8.97) -0.005 (37.78) -1.494 (45.50) 7.536 (44.23)

Log Likelihood SBC

40

Table 8: Credit Option-Adjusted-Spread of Forward v.s. Spot Mortgage Rete One-Year Seasoned Credit OAS T-Ratio 1.81% (4.5) Two-Year Seasoned 2.52% (32.1) Five-Year Seasoned 2.12% (10.4)

Notes: T-ratios are in parentheses. The simulated market value are computed based on model 4 in table 7, together with joint processes of stock market and housing prices.

41

REFERENCE Archer, W. R., D. C. Ling, and G. A. McGill. (1996). The Effect of Income and Collateral Constraints on Residential Mortgage Terminations, Regional Science and Urban Economics, 26(3-4) 235-261. Black, F. and M. S. Scholes. (1973). The Pricing of Options and Corporate Liabilities, Journal of Political Economy 81, 637-654. Blinder, Alan S. (1973). Wage discrimination: Reduced Form and Structural Estimates, The Journal of Human Resources. Volume 8, Issue 4, pages 436-455, Fall 1973. Brennan, M. J., and E. S. Schwartz. (1985). Determinants of GNMA Mortgage Prices, Journal of the American Real Estate and Urban Economics Association 13, 209-228. Buser, S. A., and P. H. Hendershott (1984). Pricing Default Free Mortgage, Housing Finance Review, 3, 405-429. Calhoun, A. Charles, and Yongheng Deng. (2002). A Dynamic Analysis of Fixed- and Adjustable-Rate Mortgage Terminations, Journal of Real Estate finance and Economics 24(1-2), 9-33. Case, Karl E. and Robert J. Shiller. (1989) The Efficiency of the Market for Single Family Homes, American Economic Review, 79(1) 125-37. Cox, D. R. (1972). Regression Models and Life-Tables, Journal of the Royal Statistical Society B 34(2), 187-220. Cox, D. R. (1975). Partial Likelihood, Biometrika 62, 269-276. Cunningham, D., and C. Capone. (1990). The Relative Termination Experience of Adjustable to Fixed-Rate Mortgage, Journal of Finance 45(5), 1687-1703. Deng, Yongheng and Stuart A. Gabriel. (2006). Risk-Based Pricing and the Enhancement of Mortgage Credit Availability among Underserved and Higher Credit-Risk Populations, Journal of Money, Credit, and Banking 38 (6), 1431-1460. Deng, Yongheng, John M. Quigley, and Robert Van Order. (2000). Mortgage Terminations, Heterogeneity and the Exercise of Mortgage Options, Econometrica 68(2), 275-307. Deng, Yongheng, Della Zheng and Changfeng Ling, (2005) An Early Assessment of Residential Mortgage Performance in China, Journal of Real Estate Finance and Economics, 31 (2), 117136. Dunn, K. B., and J. J. McConnell. (1981). Valuation of Mortgage-Backed Securities, The Journal of Finance 36, 599-617. Findley, M. C., and D. R. Capozza. (1977). The Variable Rate Mortgage: An Option Theory Perspective, Journal of Money, Credit and Banking 9, 356-364. Green, J., and J. B. Shoven. (1986). The Effect of Interest Rates on Mortgage Prepayment, Journal of Money, Credit and Banking 18, 41-50. Han, Aron K. and Hausman, Jerry (1990). Flexible Parametric Estimation of Duration and Competing Risk Models. Journal of Applied Econometrics, 5:1, 1-28. Hayashi, Fumio, Altonji, Joseph and Kotlikoff, Laurence.(1996) Risk-sharing between and within families, Econometrica, Vol.64, No.2, 261-294 Kau, J.B., Keenan D.C., Muller, W.J., and Epperson, J.F.(1992). A Generalized Valuation Model for Fixed-Rate Residential Mortgages, Journal of Money, Credit & Banking, 1992.

42

Kau, J.B. and Keenan D.C.(1995) An Overview of the Otion-Theoretic Pricing of Morgages, Journal of Housing Research 6(2), 217-244. Kocherlakota, R. Narayana.(1996). Implications of Efficient Risk Commitment, The Review of Economic Studies, Vol. 63, NO.4, 595-609. Sharing without

Liu, Peng (2007), Ph.D. Dissertation, Haas School of Business, University of California, Berkeley. Merton, R. C. (1973). Theory of Rational Option Pricing, Bell Journal of Economics and Management Science 4, 141-183. Oaxaca, Ronald (1973), Sex Discrimination in Wages in Ashenfelter, O. and Rees, A. (eds), Discrimination in Labor Markets (Princeton, NJ: Princeton University Press), 1973. Quigley, John M. (1987). Interest Rate Variations, Mortgage Prepayments and Household Mobility, Review of Economics and Statistics, 69, 636-643. Quigley, Jonn, and Robert Van Order. (1995). Efficiency in the Mortgage Market: The Borrowers Perspective, Journal of the American Real Estate and Urban Economics Association 18(3), 237-252. Schwartz, E. S., and W. N. Torous. (1989). Prepayment and the Valuation of Mortgage-Backed Securities, Journal of Finance 44, 375-392. Stanton, Richard (1995). Rational Prepayment and the Valuation of Mortgage-Backed Securities, The Review of Financial Studies 8(3), 677-708. Stanton, Richard and Wallace, Nancy. (1999). Anatomy of an ARM: The Interests-Rate Risk of Adjustable-Rate Mortgages, Journal of Real Estate Finance and Economics 19(1), 49-67.

43