DELINQUENCY & DEFAULT IN ARMs: THE EFFECTS OF PROTECTED EQUITY AND LOSS AVERSION Seow Eng ONG * Tien Foo SING Alan Hwee Loon TEO First draft: December 27, 2005 Current draft: March 31, 2006 Abstract This paper extends the extant literature in understanding the effects of equity and debt on delinquency and default by focusing on a variant of borrower equity where part of equity is “protected”. The CPF scheme in Singapore stipulates that the refund of borrower’s retirement funds utilized for property purchase prior to September 2002 takes priority over loan obligations. A decision to utilize CPF for property purchase actually increases ex post delinquency and default risk as it effectively reduces cash equity commitment. In particular, any erosion in house value that places protected equity at risk translates into potential wealth reduction or financial liability for the borrower. While loss aversion is evident for non-distressed sellers, the effect of equity losses for distressed borrowers is not as clear. Our research suggests that averting losses in committed equity may be a secondary consideration for borrower subject to income shocks, recognizing that delinquency and default are precursors to foreclosure. Interestingly, we find that the borrowers are strongly averse to incurring protected equity-induced wealth loss or financial liability. This study suggests that the first-lien “anomaly” associated with CPF refund may reduce delinquency and default risks for mortgage backed securities. * Contact author. Department of Real Estate, National University of Singapore, 4 Architecture Drive, Singapore 117566. Email: seong@nus.edu.sg We wish to thank Brent Ambrose, Tyler Yang and AREUEA participants for many useful comments and suggestions on an earlier draft of this paper, Chun How TAN of Standard Chartered Bank and Fiona TAN of United Overseas Bank for institutional details. We are also grateful to NUS for providing research funding (R-297-000-073-112). 1 DELINQUENCY AND DEFAULT IN ARMs: THE EFFECTS OF PROTECTED EQUITY AND LOSS AVERSION 1. Introduction The recent surge in house prices in the US has prompted many households to utilize their Roth IRAs to purchase homes. Although the intent of the Roth IRA is for retirement savings, the scheme is flexible enough to allow contributors to “use it to save for a home” (Fairmark.com) and first-time home buyers can withdraw 100% of contributions tax- and penalty-free. In addition, a traditional IRA penalty-free withdrawal up to USD10,000 can be made for a first home (Kiplinger.com). While the use of IRA funds seem a good idea to improve housing affordability, it is not entirely clear how the use of IRA / retirement funds would affect subsequent risk of delinquency and default should the real estate market decline. In the event of foreclosure or sale, how should losses in retirement funds be accounted for? If losses are recognized as investment losses, then retirement funds are not substantially different from cash equity. Or if the intent is to preserve savings for retirement subsistence, then should retirement funds be “protected” against such losses? The use of retirement funds for homeownership has been allowed in Singapore since the 1970s. The Central Provident Fund may be used to for down payment on a property purchase as well as regular monthly debt service. The CPF Residential Property scheme provides that when a house, purchased before September 2002, is subsequently sold either for foreclosure or relocation, the sale proceeds must be used to repay the CPF savings first before being utilized to repay the mortgage balance. In other words, the portion of borrower equity financed by CPF funds is “protected” until the borrower reaches retirement age of 55 years. However, the borrower is still liable for any negative equity that may result from the difference in sale proceeds and the sum of CPF refund and outstanding mortgage balance. The shortfall has to paid for by way of new equity (i.e. draw down on borrower wealth/savings), or it may be converted into an unsecured loan or recovered from bankruptcy proceedings against the borrower. To the best of our knowledge, the protection of CPF funds is an unusual arrangement (anomaly). In 2 addition, the policy raises efficiency questions as the priority accorded to CPF refund is a trade off between an immediate loss realization for future retirement benefits. The institutional framework in Singapore allows us to answer the question of how “protected equity” influences the probability of delinquency and default. This study potentially contributes to our understanding of the extent of the borrower equity in affecting the risk of default and delinquency (Kau, et al., 1993 and 1994, and Lambrecht et al., 1997). Clearly the results would be pertinent to mortgage securitization efforts in Singapore, as well as policy and securitization implications for other countries where house purchases are financed in part by retirement funds. In addition, while the use of retirement funds at the point of purchase may seem rational, any protection of such funds may be onerous when the borrower is faced with negative equity subsequently, especially if he/she is still liable for the shortfall, assuming the borrower is subject to non-trivial bankruptcy costs (i.e. has little other assets such that defaulting incurs no addition liability). This phenomenon could potentially provide further insights on our understanding of the issue of time inconsistency and inter-temporal choice (Loewenstein and Thaler, 1989). Our research attempts to address also another related issue – the effect of loss aversion on borrower’s delinquency and default decision. Loss aversion states that the higher amount of potential loss, the more reluctant people are towards realizing this loss. Genesove and Mayer (2001) postulate that owners who are loss-averse have an incentive to attenuate a loss by choosing an asking price that exceeds the level they would set in the absence of a loss. Our research on loss aversion contributes to the literature in two ways. Firstly, Neo, Ong and Somerville (2005) argue that loss aversion implies that owners will seek to avoid loss realization if the expected loss is too high. Extending this concept to borrower behavior, loss-averse borrowers have the incentive to avoid realizing losses to their borrower equity when property value declines by avoiding defaults which may result in foreclosure sales. We investigate the presence of loss aversion in mortgage borrower behaviors at two levels – the potential losses mortgage borrowers suffer when the property prices fall, and the liability that borrowers would incur as a direct consequence of protected equity. If borrowers merely may view delinquency as a temporary delay in payment, any unrealized loss in equity may not be sufficiently strong 3 motivation to deter delinquency compared to say income shocks. In contrast, any shortfall incurred as a result of protected equity translates to financial loss (decrease in wealth) or liability with a high attendant cost (unsecured loan or bankruptcy). The second contribution arises from the fact that non-recourse loans (dominant in the US) effectively mean that borrowers have put options that protect their downside. If so, the S-shape value-wealth function is likely to be left truncated, such that the maximum loss in wealth is the equity committed in the property. This asymmetry could introduce noise in empirical studies on loss aversion behavior on the part of borrowers. Mortgages in Singapore do not provide that downside protection. Even though equity from CPF funds is protected, but the borrower is still liable to the bank for any shortfall, notwithstanding a low bankruptcy cost. We postulate that the impact on default/ delinquency risks of protected equity and conversion into unsecured loan or liability is different from conventional measures of borrower equity financed by cash savings. This provides a natural experiment for us to investigate the role of government policy and the role of borrower equity in affecting mortgage risks. In addition, we postulate that the relationship between borrower equity and decisions to default/delinquent is influenced by loss aversion behavior among borrowers in that loss-averse borrowers have the incentive to avoid realizing losses to their borrower equity when property value declines. In addition, the loss aversion behavior differs between committed cash equity and new equity. Our research finds that the use of protected borrower equity to help finance house purchase actually increases the risk of delinquency when the quantum at origination increases. Put differently, a higher utilization of retirement funds for purchase essentially reduces cash equity commitment. For any given level of loan-to-value, a higher CPF commitment (lower cash equity) translates into a higher affinity to become delinquent, although the affinity to default is not significant. Our empirical results do provide some evidence that the risk of delinquency and default arising from the utilization of CPF funds is mitigated as protected equity is accumulated over time. Interestingly, we find that equity losses induce delinquency, but not when the losses are large enough such that protected equity becomes eroded. The cost of wealth reduction or financial liability as a result of protected equity is sufficiently onerous to 4 induce loss aversion in borrower behavior. In particular, we find evidence that the magnitude of wealth loss matters in delinquency and default decisions – borrowers are less likely to default or become delinquent when the amount of protected equity at risk increases. We find weak evidence that younger borrowers are more likely to default for any given amount of protected equity and that younger borrowers are more loss averse when protected equity is at risk. This can be attributed to the deferment of wealth realization. We conclude by examining some policy implications. 2. Literature Review It is necessary to note that most literature on mortgage risks was originated from the US, where Fixed Rate Mortgages (FRMs) are prevalent (Ong, 2000). Conversely, all mortgages originated in Singapore are ARMs (Khor and Ong, 1998). The exogeneous and endogeneous factors affecting both forms of mortgages may thus diverge. For instance, the prepayment risk for Singapore mortgages is very low (Ong, Maxam and Thang, 2002) while the prepayment risk for ARMs in US may be higher resulting from potential switches to FRMs to take advantage of interest rate movements (Ambrose and LaCour-Little, 2001). However, Campbell, et al. (1983) found that most determinants that affect default decisions influence delinquency in the same way. Therefore, the methods and factors used in the literature to rationalize mortgage risks in FRMs serve as a platform for our analysis. Practitioners in the US and Singapore1 differentiate delinquency and default by the number of days of missed installments. Delinquency is defined as the nonpayment of a mortgage payment due (e.g. Ambrose and Buttimer, 2000; and Holmes, 2003). Default occurs when a borrower has missed 90 days’ installment and the fourth payment is due (Ambrose and Capone, 2000; and Chen and Deng, 2005). This is also sometimes termed serious delinquency. Therefore, delinquency is a necessary precursor to default. During delinquency, the lender usually sends reminders to the borrower to make up the missed payments. Although the lender has the right to foreclose the property as missing an 1 This definition is supported by Section 25 of the Conveyancing and Law of Property Act in Singapore, which states that "A mortgagee shall not exercise the power of sale conferred by this Act unless - notice requiring payment of the mortgage money has been served on the mortgagor or one of several mortgagors, and default has been made in payment of the mortgage money or part thereof for 3 months after the service…” 5 installment is tantamount to a breach of contract, he would usually refrain until default. Thus, the borrower has the option to repay the missed installments and reinstate the mortgage. Once the loan transited to default, the lender will issue a formal legal letter to the borrower indicating the lender’s right to proceed with foreclosure proceedings any time from then on. The commencement of foreclosure proceedings is significant because borrowers are generally not able to reinstate their delinquent loans once the foreclosure sale occurs (except for some states). Thus, the borrower faces real danger of losing his home with the transition to default. Using this terminology, there are two unambiguous decision points i.e. 1) whether to delinquent, and 2) once in delinquency, whether to default. We adopt this set of definitions in our paper. Determinants of Mortgage Delinquency Ambrose and Capone (1996, 1998) and Waller (1988) described the aim of delinquency is either to put the funds, originally intended to pay the installments, to other uses due to financial difficulties, or to exercise the implicit put option to abandon the property. A third cause of delinquency noted by Waller (1988) is the economic incentive borrowers can gain from living in the house rent-free before foreclosure takes place. Von Furstenberg, et al. (1974) found that the equity-value ratio possesses a significant negative relationship with delinquency while the age of mortgages has a positive relationship. In addition, mortgages of existing houses are more prone to delinquency than those taken on new houses. Herzog and Earley (1970) and Morton (1975) also found income, occupation and the number of children to be influential determinants. Zorn, et al. (1989) argued that delinquency can be regarded as a form of borrowing from the lender at the mortgage contract rate. Therefore, when interest rate increases, delinquency rate will correspondingly rise as people “borrow” at the relatively cheaper source of fund to finance other uses. Canner et al. (1991) found that the receipt of government assistance, headed by a minority, and martial status have positive influences. On a more somber note, Canner et al. (1991) pointed out that delinquency prediction consists of a large unexplained random component as credit problems can arise 6 from events that are difficult to foresee. Thus, the use of ex-ante data has the ability to capture components that systematically affect delinquency and are observable to the lender at loan origination but ignores the more unpredictable ex-post components. Determinants of Mortgage Default Literature on mortgage loan specific characteristics traditionally focuses on the equity position of the borrower. Several proxies are used including the loan-to-value ratio at origination (Campbell, et al., 1983), current loan-to-value ratio (Campbell, et al., 1983; Cunningham and Capone, 1990), value-to-total debt ratio (Waller, 1988; Zorn, et al., 1989) and book value (Giliberto and Houston, 1989; Hendershott and Schultz, 1993). Other mortgage loan specific factors used include the age of the mortgage (Waller, 1989 and Schwartz and Torous, 1993), mortgage term (Bervokec et al., 1994) and mortgage rate (Zorn, et al., 1989 and Ambrose and Capone, 1996 and 2000). Property related factors examined include the price volatility of the property (Schwartz and Torous, 1993; Capozza et al., 1998; Ambrose and Capone, 2000), age (Canner et al., 1991) and neighborhood quality (Vandell and Thibodeau, 1985). Other significant factors consist of the returns from property capital appreciation (Schwartz and Torous, 1993; Kau et al., 1994) and the returns from rental yield (Capozza et al., 1997 and 1998). With regards to borrower related characteristics, the payment-to-income ratio is a popular ability-to-pay measure but yields inconsistent results. Vandell (1978) and Campbell and Dietrich (1983) found a positive relationship while other studies found a negative relationship (Springer and Waller, 1993; and Cunningham and Capone, 1990). Other studies focus on the wealth of the individuals and household income (Canner et al., 1991; and Bervokec et al., 1994), age (Capozza et al., 1997), and the number of years of job tenure (Cunningham and Capone, 1990; Hakim and Haddad, 1999). Exogenous factors include demographic or macroeconomic factors. Unemployment is the more popular factor used by a number of studies that include Campbell and Dietrich (1983), Lea and Zorn (1986), and Capozza et al. (1997). Loss Aversion Studies 7 Loss aversion studies have their theoretical foundation in the prospect theory. Prospect Theory was originated from Kahneman and Tversky (1979) for modeling the decision made for risky gambles of two non-zero outcomes. Tversky and Kahneman (1992) further expanded the design to include gambles of more than two outcomes. With the foundations from Markowitz (1952), the prospect theory defined utility over changes in wealth rather than the level of wealth. The three essential components for prospect theory are that people focus on gains and losses relative to a reference point, value function is steeper for loss than for equivalent amount of gain, and marginal value of both gains and losses diminishes with the size of the gain and loss respectively (Tversky and Kahneman, 1991). Prospect theory also assumes loss aversion. Due to the three components of the prospect theory, Shefrin and Statman (1985) and Odean (1998) extends upon the theory to predict that loss aversion will cause investors to hold on to their loser investments longer than their winner investments, even when the former is expected to have a lower subsequent gain. Genesove and Mayer (2001) applied the concept of loss aversion to explain the significant positive correlation between house prices and sales volume. It postulated that owners who are loss-averse have an incentive to attenuate a loss by choosing an asking price that exceeds the level they would set in the absence of a loss. This results in shrinking sales volume during the downturn of the property market. Neo, Ong and Somerville (2005) argued that this hypothesis implies that owners will seek to avoid loss realization if the expected loss is too high. This is deemed the disposition effect under behavioral arguments. 3. Central Provident Fund The Central Provident Fund (CPF) is a comprehensive social security savings plan. Employees and their employers make monthly contributions to the CPF that may be withdrawn upon retirement (at the age of 55). In addition, these contributions can be used to buy a home, investment and education (http://www.cpf.gov.sg/). However, the CPF Residential Property Scheme provides that “when the property is sold, the sale proceeds will first be used to repay the CPF savings used for payment of stamp duty, legal costs and survey fees, and CPF principal sum … before repayment of the … outstanding 8 housing loan”. This policy has been amended for properties purchased after September 1, 2002, in which case, “when the property is sold, the sale proceeds shall be applied to repay the financier and the Board in the following order of priority: (a) repayment of the outstanding housing loan; (b) repayment of CPF principal sum … plus CPF saving used to pay the legal costs, stamp duty and survey fees.” In Singapore, the use of CPF savings2 to pay for the initial housing down payment and the repayment of the mortgage loan is prevalent. As of September 2005, 1.26 million (215,000) contributors have withdrawn S$73.46 billion (S$41.63 billion) for public (private) housing purchases (see Graph 1). In percentage terms, CPF funds about 20% of the aggregate value of private property transactions from 1995 through 2004 (see Graph 2).3 The Monetary Authority of Singapore (MAS) regularly monitors negative equity in housing mortgages where negative housing equity is defined as “comparison of the property value against the outstanding loan” for mortgages originated after September 1, 2002. Before this date, negative housing equity is the difference between the property value and the “sum of the outstanding loan and the amount of CPF savings used for purchase of the property” (MAS Financial Stability Review, December 2005). As of September 2003, 13.7% of contributors who utilized their CPF to purchase private residential properties are in negative equity, of which 5% are in delinquency (more than 90 days overdue). The data for this study is entirely from mortgages originated before September 2002, so for the rest of the analysis, we will focus on negative housing equity defined as the difference between the property value and the sum of the outstanding loan and the amount of CPF savings to be refunded (protected equity). 2 This is referred to as deferred borrower equity because CPF savings cannot be withdrawn and used freely until retirement age. This is opposed to conventional borrower equity financed by cash savings which can be utilized immediately in other areas. 3 The percentage utilization for private properties is higher, but due to lack of aggregate data on secondary HDB transactions, we are unable to provide point estimates. 9 Graph 1: CPF contributions and housing withdrawals 20,000.0 18,000.0 16,000.0 $million 14,000.0 12,000.0 10,000.0 8,000.0 6,000.0 4,000.0 2,000.0 20 03 20 00 19 97 19 94 19 91 19 88 19 85 19 82 19 79 19 76 19 73 19 70 0.0 Year Total CPF contribution Total CPF withdrawals housing Source: CPF annual reports Graph 2: CPF utilized for Private Properties 30.00% 25.00% 20.00% 15.00% 10.00% 5.00% 0.00% 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Source: Authors’ compilation from CPF annual reports and REALIS 10 Protected equity and loss aversion theory As previously mentioned, the unique feature of the mortgage market in Singapore is the use of CPF savings to finance the purchase of residential properties. With protection being conferred by regulations to the CPF portion of borrower equity, this provides a natural experiment to investigate the role of government and the role of protected borrower equity in controlling mortgage risk. As mentioned, borrower equity is found to be an important variable of default decisions in past mortgage literature. Furthermore, the effect of protected borrower equity on mortgage risk is likely to be different from the conventional borrower equity, which can be lost if foreclosure occurs. However, this disparity has not been addressed by past mortgage studies. This paper postulates that the relationships between equity and delinquency/ default risk can be explained by the concept of loss aversion among borrowers. When the value of the property falls, the equity component of the mortgage will be reduced first and the current loan-to-value ratio of the mortgage increases. The loss-averse borrowerowner has the incentive to avoid realizing the loss by remaining current in his mortgage. If default occurs, the borrower might be forced to give up his property through foreclosure, and in the process realize the loss in equity. However, there is a kink in the aforementioned situation. As property value continues to decline, there will be a time when the borrower’s entire equity component is wiped out. When this happens, further declines in property value translate to wealth reduction or financial liability should the owner default or sell the property. At the same time, the protected equity still remains in the borrower’s CPF account that may be offset against the liability in future (at the time of retirement). The following equations capture this relationship. Change in wealtht = Pt – Po, where Pt is the estimated property value at time t and Po is the purchase price. In the event that Pt < Po, the loss in excess of cash equity can be characterized as Loss = cash equity + wealth reduction / unsecured liability – present value of protected equity where the offsetting effect from protected equity is less valuable today if the time to retirement is longer. We may assume that the cost of unsecured liability or opportunity 11 cost associated with wealth reduction is higher than the return to protected equity. The cost of unsecured liability is also likely to be higher than the mortgage rate since the mortgage is collateralized. Given a high opportunity cost associated with wealth reduction or high cost of unsecured liability, it is reasonable to expect that a borrower should be more loss averse when the fall in house value exceeds the committed cash equity. This loss could be more acute if the deferred period to realization (retirement) is long. In other words, a young borrower (with a long time to retirement) may be more loss averse in the event of negative equity (Thaler, 1981; Loewenstein and Thaler, 1989). Whether the borrower exhibits different degrees of loss aversion towards bearing unsecured liability and losing cash equity is an empirical question. 4. Model Setup and Explanatory Variables To examine the relationship between the probability of delinquency and probability of default, our empirical model requires an assumption of a delinquency probability function and a default probability function where the ith borrower maximizes a linear indirect utility function Vij* over j outcomes Vij* = α j + β j X i + δ j Wi + φ j Z i + ε ij , i = 1, ….., N; j = 1, ….., J, (1) where Xi is a vector of option-related characteristics, Wi is a vector of individual borrower, property and mortgage characteristics and Zi is a vector of macroeconomic variables other than house price and interest rate to proxy trigger events. Following from (1), where j = 1 only, the independent probit specification is Vi * = x'i βi + ε, yi = 1 if Vi* > 0, 0 otherwise, (2) where Vi * represents the unobserved indirect utility, yi is the observable actual decision to delinquent or default, x'i represents the vector of independent variables, and βi is a vector of model parameters. The probability of delinquency is examined by firstly assigning the dependent binary yi to be either 0, which indicates a non-delinquent (or non-default) loan, or 1, indicating a delinquent (or default) loan. The probability of delinquency (or default) 12 is then modeled using a vector of independent variables, denoted as xi. A general specification is that the probability of observing 1 for yi is: P ( yi = 1 ) F ( β'xi ) = (3) for i = 1, 2, …, N, where F is an appropriate distribution function. We specify a probit distribution: ( ) P (Y = 1 x ) = ∫ φ (t )dt = Φ x ' β x'β −∞ (4) It is well accepted that the probit model can be estimated by maximizing the likelihood function, where n [ ( )] [1 − F (x β )] L = Π F x i' β i =1 yi ' i 1− y i (5) The log-likelihood function is thus n { ( ) [ log L = ∑ y i ln F x i' β + (1 + y i ) ln 1 − F ( x i' β ) ]} (6) i =1 The subsequent estimations are undertaken via the maximum likelihood estimation. The rationale for the use of the probit models is to account for the relatively small sample size used in this study. We also estimated a bivariate probit model where delinquency and default are estimated simultaneously, but the results are qualitatively unchanged. Explanatory Variables Protected Equity Utilized as at Censor Date or Date of Default/ Delinquency Typically, the equity component of a mortgage would be derived from the home purchaser’s cash savings. Thus, for a LVR of, say 0.8, the potential homeowner requires cash equity for 20 per cent down payment. This equity component is usually one of the main loan criteria for the lender to control the default risk of the mortgage. The equity component is used as a buffer against immediate default when the value of the property 13 falls. It is only when property value falls beyond the stipulated margin of, say 20 per cent, which lenders will suffer a loss and face potential defaults. In Singapore, homeowners may use a combination of cash and CPF savings to pay for the equity component of the purchase price upfront so as to reduce the loan quantum. As such, the equity component that is derived from CPF funds is “protected equity” as opposed to the committed cash equity component. As the property value declines, the borrower will suffer a loss in his cash equity component. When the entire cash equity component is wiped out, the borrower does not suffer losses in CPF component of his equity but is still responsible for any difference between the sale proceeds under foreclosure and the mortgage balance. Upon sale or foreclosure, the shortfall has to be paid by way of new equity (i.e. the borrower suffers a reduction in wealth) or will converted into unsecured liability. 4 The conversion of secured loan into unsecured loan may be more onerous for borrowers due to higher interest charges. On the other hand, borrowers who have little to lose by being made bankrupt may not view such conversion to be onerous, in fact, their retirement funds are still protected.5 Depending on whether the borrower is more averse towards holding the unsecured loan, the distribution of the equity between cash and protected equity will have a differential impact on default performance. Thus, we include a control variable for the amount of protected equity divided by the value of the property for each observation, GUARPX. This amount of protected equity includes both the estimated retirement funds utilized to service the mortgage, 6 and also the initial lump sum drawn from the borrower’s CPF account to reduce the loan quantum and meet the down payment requirement. This, however, does not include the interests accrued to the CPF account if the retirement funds have been left in the borrower’s the CPF account.7 4 We have no information on the current wealth of borrowers, but given that the mortgages are delinquent or in default, it is reasonable to infer that such borrowers has limited current wealth. 5 The bank usually lodges a charge against individuals under bankruptcy to access the CPF funds. Some banks deliberately defer bankruptcy proceedings when the borrower is close to retirement age of 55. 6 The amount of monthly CPF funds used to pay for debt service is estimated from prevailing CPF contribution rates, ceilings and the borrower’s monthly income at time of purchase. 7 Refund of interest on CPF funds takes secondary priority to settlement of outstanding loan in the event of a sale or foreclosure (CPF Residential Property scheme). 14 Protected Equity for Initial Down payment & Reducing Loan Quantum CPF funds will normally be first utilized to pay for the property before the borrower takes up a loan, subject to certain stipulated limits. The use of CPF funds reduces the loan quantum and thus the monthly mortgage payments. This increases the affordability of the installments and contributes to the accumulation of financial resources, thus enhancing borrower’s ability to meet any unexpected financial commitments. We include the ratio of the CPF savings used for the initial property down payment to the purchase price of the property (CPFPRICE) as an indication of its influence to delinquency and default risks. Servicing the Mortgage using Retirement Savings Monthly CPF contributions can also be used to pay the monthly installments for the mortgage, and this is a stabilizing source of fund to upkeep the mortgage. If the borrowers’ monthly CPF contributions exceed the monthly mortgage installments, they would be able to service the entire monthly debt service using only CPF contributions. Such borrowers should be less prone to default and delinquency risks because triggerevents other than income disruptions (such as loss of job, retrenchment) that affect their cash flow are less likely to interrupt mortgage repayment. We utilize a dummy variable (DCPF) where value of 0 indicates the mortgage is entirely serviced by retirement funds and the value of 1 indicates otherwise. We expect the latter to have a higher mortgage delinquency risk. If the monthly debt service is larger than the monthly CPF contributions, a combination of both cash and CPF funds is usually employed. As the borrowers’ salaries increase over the years, they will be able to increase the proportion of mortgage installment paid using CPF. For this group of borrowers, we first calculate the ratio of the monthly mortgage installment to the monthly CPF contributions of the borrower at origination. Then we multiply this ratio with DCPF to produce an interactive variable (DMORTCPF) to indicate how close borrowers are to servicing the mortgage entirely by CPF. As the salary of the borrowers increased, a greater portion of the mortgage installments can be financed by CPF contributions. Thus, this variable also indicates how 15 quickly borrowers are able to achieve 100% CPF servicing. We expect a lower DMORTCPF to lead to lower risk of default and delinquency. Constituents of Initial Borrower Equity and the Self-selection/ Information Asymmetry Theory The two related concepts of self-selection and information asymmetry is utilized to account for the unobserved heterogeneity of borrowers. Agarwal, Ambrose, Chomsisengphet, and Liu (2005) appealed to the self-selection theory to motivate borrower decisions that reveals the unobserved risk profile of the borrowers. It argued that the risks of first and second mortgage products are different due to different underlying decisions to originate second mortgages as compared to those with only first mortgages. On the other hand, Deng, et al. (2000) argued for the presence of information asymmetry whereby one application is that the borrowers are more aware of their house price volatility. As a result, borrowers can exploit the under-priced options. The paper controlled for this by using the initial Loan-to-Value ratio (LVR) or the Equity-to-Value ratio (EVR) at the time of origination as a proxy. It found that borrowers who have chosen a lower EVR (i.e. a higher LVR) tend to have higher risks. We argue that the decision on the level of initial EVR (and LVR) is a form of self-selection, which allows lenders to identify unobserved heterogeneity among borrowers. To examine the effect of the extent of protected equity within total borrower equity, we include an interactive variable EVRCPF (EVR * CPFPRICE) to attempt to investigate the effects of higher borrower equity caused by higher initial CPF utilized at origination. It indicates the influence of cash equity versus protected equity. This is important because a high EVR may indicate high borrower equity, which should discourage defaults. However, if the high EVR is caused by higher CPFPRICE, it is possible that the protected nature of the CPF funds used may make it easier for borrowers to default instead, depending on the borrowers’ attitude towards financial liability. In our paper, besides initial EVR, we postulate the decisions on mortgage term, number of co-borrowers and previous experiences of delinquency to adhere to the selfselection/ information asymmetry concept. Borrowers are aware of their financial resources. Financially weaker borrowers will self-select to increase the mortgage term of 16 the loan and increase the number of co-borrowers. Borrowers who choose to go into delinquency and are reinstated before transiting to default may reveal themselves to be financially unstable and thus of higher risk. These decisions may reveal the unobserved risk profiles of the borrowers and aid lenders in assessing their risks. These shall be further explained in the subsequent sections. Loss Aversion Variables We first separate the borrowers according to whether they are suffering losses in borrower equity by using a dummy variable (DVL) such that a value of 1 indicate equity losses while a value of 0 otherwise. Borrowers incurring equity losses may be more reluctant to default as the losses will be realized once foreclosure occurs. While we expect a negative relationship in default, we also note that the impact of equity loss may differ for delinquency. If borrowers merely may view delinquency as a temporary delay in payment, any book loss in equity (i.e. unrealized loss) may not be sufficiently strong motivation to deter delinquency compared to say income shocks. To measure the quantitative impact of losses on the probability of default/ delinquency, we measure the nominal quantum of loss suffered (DLOSS). This loss is measured by the difference between current property value and the purchase price, normalized with respect to the purchase price. 8 Similarly, a negative relationship is expected as the higher the quantum of loss, the greater the degree of loss aversion towards realizing the loss. Finally, we measure the effect of the borrower suffering a loss in his committed cash equity versus suffering a financial liability or wealth reduction as a result of deferred protected equity. The “wealth loss” in the latter refers to the shortfall in estimated value and the sum of the outstanding loan and protected equity. We use a dummy variable (CCAP) such that value of 1 refers to borrowers who face wealth loss or conversion of protected equity into debt while value of 0 indicates borrowers who only suffer losses in committed cash equity. We expect the borrower to have different degrees of loss aversion 8 We also compute nominal loss as in Genesove and Mayer (2001) but the results are qualitatively unchanged. 17 when faced with loss of cash equity versus wealth loss. We also quantify the amount of liability arising from protected equity, PEQLOSS that is similar to DLOSS. Time to Wealth Realization As discussed in the previous section, we expect an inter-temporal effect in the use of CPF savings on borrower behavior due to the stipulation that CPF funds can only be withdrawn at retirement age of 55. We investigate the effects of deferred wealth realization of CPF funds on two levels. First, we expect a younger borrower to be more loss averse when protected equity is at risk (i.e. when the estimated house value is less than the sum of outstanding loan and protected equity). If so, the interaction variable between age 9 and wealth loss (AGECCAP = AGE*CCAP) should show a positive relationship with the risks of default and delinquency. This variable allows us to investigate the simultaneously effects of a short time to retirement (wealth realization) and a high level of protected equity at risk. It also helps to isolate the temporal effect from the wealth accumulation effect. Second, on a more general level after controlling for protected equity at risk (or wealth loss), do the amount of CPF funds protected and time to realization affect borrower delinquency and default? When the time to retirement (i.e. the time to wealth realization) is long, the value of CPF funds committed in the house may be less valuable to the borrower than when the time to retirement is shorter. If so, the interactive variable AGEGPX (AGE*GUARPX) should yield a negative coefficient in that older borrowers with a shorter time to realization are less likely to default for any given level of protected equity. Option-related Variable While mortgages in Singapore are not non-recourse loans, to protect against omitted variable biases and to account for low bankruptcy cost borrowers, we compute a proxy for put option value. PROBNEQ indicates the probability of negative equity, which 9 AGE is age of the borrower at time of delinquency / default / censor. Where the property is owned by more than one person, the age of the youngest owner is used. 18 determines the attractiveness of allowing foreclosure. This variable is calculated by the put-option for each loan or the probability of negative equity: ⎛ (log Vi , m − log M i , k ) ⎞ ⎟ Put _ Optioni , k = Φ⎜⎜ ⎟ 2 ω ⎝ ⎠ where Φ (.) is cumulative standard normal distribution function, M i , k value of the property and (8) is the market ω2 is the estimated variance. It is expected that a higher probability of negative equity would induce delinquency and default. Mortgage Loan Specific Characteristics Following Ambrose and Capone (2000), we postulate that the risk profiles of default and delinquency are different between first time delinquents and repeat delinquents who have reinstated from previous delinquencies. This can also be motivated from the self-selection theory point of view, which is explained in the following section. To test this hypothesis, we include a dummy variable REPEAT where the value of 1 indicates loans that have previous delinquency experiences and the value of 0 indicate otherwise. There are two aspects to PREMIUM, which is calculated as the ratio of the difference of purchase price and valuation, over valuation. Firstly, the fact that borrower is willing pay an amount excessive to the fair value implies high preference to the property. As such, his transaction cost of foreclosure will be higher and he will try to keep the mortgage current, i.e. the risk of delinquency and default will be lower. Secondly, paying a premium may stress the borrower’s finances as more savings are spent on the property. This will increase the risk of delinquency when trigger-event occurs. It can be argued that borrowers’ decision on the mortgage term is a congruence of borrowers’ assessment of their financial abilities and financial commitments. Thus, mortgage term (MT) can serve as a useful proxy for self-selection/ information asymmetry for borrowers’ financial abilities, which may be unobserved by lenders. 19 An interesting feature of ARMs in Singapore is that most loans consist of a preferential fixed rate for the first two to three years (Ong, 2002). The below-market rates are to entice new borrowers. On one hand, the preferential fixed rate improves the affordability of borrowers and enables borrowers to accumulate more non-housing wealth for the initial period. On the other hand, such schemes are anticipated to attract higher risk borrowers whom just qualify or can afford for the ‘teaser rates’. The dummy variable PRM is used to for loans with initial preferential rates (allocated the value of 1), otherwise are allocated 0. Property Specific Characteristics The tenure of residential properties in Singapore is essentially categorized into either 99-year leasehold or freehold properties. Holding affordability constant, it is documented that people would prefer freehold properties for continuity. Higher preference generally implies higher transaction costs and leads to lower risk of delinquency. Dummy variables (TENURE) are used to differentiate the effect of the type of lease (99-leasehold properties are allocated the value of 1; and otherwise are allocated 0). The properties can also be classified as either low-rise or high-rise, where people preferring the former. Dummy variables (TYPE) are again used where the latter is allocated the value of 1; and otherwise is allocated 0. An independent variable that is used for low-rise properties is the land area (LAREA). It is expected that with other things held constant, people would prefer larger land area. A determinant for high-rise properties is the floor level the property (FLOOR) is located (Ong and Koh, 2000). People generally prefer to live on higher floors. Delinquency risks are anticipated to be lower for properties that the purchasers fancy. Another variable used is the built-up area of the property (BUAREA). Borrower Specific Characteristics The Payment-to-Income ratio (PINCRATIO) directly indicates affordability or the borrower’s abilities to pay the mortgage installments. A higher ratio reflects lower affordability and the borrower may find it more difficult to keep the mortgage current in 20 the face of negative trigger events. The initial mortgage payment and total household income at the time of origination is used. Some studies suggest a higher number of co-borrowers would lead to lower mortgage risks. This can be motivated as higher total household income (Bervokec, et al., 1994; and Lambrecht, et al., 1997). Neo and Ong (2004) suggest risk sharing as a reason for the negative relationship with foreclosure risks. However, we postulate that the number of co-borrowers (BORROWER) can be an indication of self-selection/ information asymmetry. Borrowers are aware of their wealth and financial commitments. They would include more co-borrowers if their financial circumstances are less favorable, i.e. self-selection. From the lenders’ point of view, they would require more borrowers if the borrowers are perceived as risky. Thus, a higher number of borrowers essentially imply higher risks. Property is usually purchased either for owner-occupation or for investment. Owner-occupiers tend to have emotional attachment to the property as their homes and hence higher transaction cost. They are thus more motivated to continue paying the mortgage when financial difficulties strike. On the other hand, investors are motivated by the profit motive. They may accept higher risks that could impair their inability to pay the mortgages in the event of negative shocks. They may also be dependent on the rents received to pay the mortgage. Dummy variables (PURPOSE) are utilized to categorize purchasers where investors are assigned the value of 1, and 0 if otherwise. The stability of future flows of income is proxied by the number of years the borrower with the highest income has been in his current employment (YRSEMP) (Vandell, et al., 1985; and Cunningham, et al., 1990). With payment-to-income ratio indicating affordability at origination, stability of future flows of income can proxy the probability of continued ability to upkeep the mortgage in the future. Another potential significant factor is occupation of the youngest borrower (OCCUP). Occupations as professionals, executives and managers earn stable income while self-employed persons and sales persons earn unstable income. The stability of future income streams proxy the probability of continued financial affordability. A dummy variable of 1 is allocated to unstable-income occupations, otherwise is allocated 0. 21 Environmental Characteristics Market sentiments can proxy the returns on other investments (Zorn, et al., 1989). When market sentiments are good, funds will be directed away from mortgage payment to other more attractive investments. Conversely, poor sentiments imply a lack of good investments that borrowers can park their money in. Accordingly, funds will be better used in repaying the mortgages to prevent incurring late payment penalties. This is similar to the argument put forth by Ong (2000) and Ong et al. (2002) although the research was on prepayment risk. It is well noted that changes in property prices do reflect changes in fundamentals and sentiments (Ong, 2000 and Ong et al., 2002). However, it is liable to lag the current market sentiments. On the other hand, better market sentiments represent higher returns from the borrowers’ other investments. This will increase their financial wealth, which improves their ability to withstand financial shocks. Market sentiments are proxied by change in the Straits Times Index (CSTI) that is a price-weighted index consisting of 30 major stocks in Singapore. Retrenchments affect borrowers’ abilities to continue with the mortgage payments. The threat of retrenchments and uncertainty of future income affects borrowers can be measured by the change of unemployment rate (CUNEMP) from the origination date to the date of delinquency or if there is no delinquency, the date of censor is used. On the other hand, changes in Gross Domestic Product (CGDP) can proxy the change in income of the borrowers since we do not have the individual income progression of the individual borrowers. Exhibit 1 defines the variables used for this study. 22 Exhibit 1: List of Determinants, Codes and Expected Signs of Influence Variable Code Expected Signs Protected Equity-related Variables Guaranteed Retirement Funds per Property Value Initial CPF-to-Price Ratio Monthly CPF Contributions able cover entire mortgage installment = 0 GUARPX CPFPRICE DCPF +/+/+ Mortgage Installment-to CPF Contribution * DCPF Initial Loan-to-Value Ratio Initial Equity-to-Value Ratio Interactive Variable EVR*CPFPRICE Age of Youngest Borrower at delinquency/default/censor DMORTCPF LVR EVR EVRCPF AGE + + + +/- Interactive Variable AGE*GUARPX Option-related Variables Put Option Value as at Delinquent Date or Censor Date Put Option Value as at Default Date or Censor Date Mortgage Loan Specific Characteristics Previous Delinquency Experience = 1 Price Premium Mortgage Term Teaser Rate Mortgages = 1 Property Specific Characteristics Tenure where Freehold = 0 Type of Property where Low-rise = 0 Land Area Floor Level Built-up Area Borrower Specific Characteristics Payment-to-Income ratio Number of Borrowers Purpose of Purchase where Owner-occupation = 0 Number of years in current employment Occupation where stable income = 0 Environmental Characteristics Change in GDP as at Delinquent / Censor Date Change in GDP as at Default / Censor Date Change in STI as at Delinquent / Censor Date Change in STI as at Default / Censor Date Change in Unemployment Rate as at Delinquent / Censor Date AGEGPX - PROBNEQ1 PROBNEQ2 + + REPEAT PREMIUM MT PRM + + + +/- TENURE TYPE LAREA FLOOR BUAREA + + - PINCRATIO BORROWER PURPOSE YRSEMP OCCUP + +/+ + CGDP1 CGDP2 CSTI1 CSTI2 CUNEMP1 +/+/+ Change in Unemployment Rate as at Default / Censor Date CUNEMP2 + Loss Aversion Variables Borrowers not suffering any losses = 0 as at Delinquent/ Censor Date DVL1 +/- 23 Borrowers not suffering any losses = 0 as at Default / Censor Date DVL2 +/- Quantum of Loss as at Delinquent / Censor Date Quantum of Loss as at Default / Censor Date Borrowers facing conversion of protected equity into debts = 0 as at Delinquent / Censor Date DLOSS1 DLOSS2 CCAP1 +/+/- Borrowers facing conversion of protected equity into debts = 0 as at Default / Censor Date CCAP2 - Quantum of Loss in Protected Equity as at Delinquent / Censor Date PEQLOSS1 - Quantum of Loss in Protected Equity as at Default / Censor Date PEQLOSS2 - Interactive Variable AGE*CCAP AGECCAP + 5. Data and Descriptive Statistics The data used in the empirical estimation is based upon a major insurer in Singapore whose business portfolio includes the issuances of residential mortgages. The dataset provides a rich variety of micro-level borrower, loan and property characteristics and consists of 633 random samples of individual housing mortgages and the observations of delinquency and default are taken monthly, from January 1999 to August 2002. A total of 133 cases have become delinquent at certain times within the period of analysis and a total of 55 observations have been in default. As stated previously, delinquency is defined as nonpayment of mortgage payment due, while default occurs when a payment has been delayed more than 90 days. The summary descriptive statistics delineated by whether the loans are in positive or negative equity (difference between estimated house value and outstanding loan) at the date of censor is shown in Exhibit 2. Over 55% of the mortgages in our sample experienced a drop in property prices, and more than 26% suffered potential wealth losses, i.e. where the house value falls below the sum of outstanding loan and protected equity, but where the value is still higher than the outstanding mortgage balance. In only slightly over 5% of cases do negative equity, as currently defined by MAS, arises (i.e. where house value falls below loan outstanding). 24 The origination dates of the sample range from March 1980 to December 1999.10 Since only 9 cases originated before 1990, a better measure of central tendency would be the median at 1998. The average loan amount is $363,697 with standard deviation of $161,537. The mean value of LVR is 0.56662. The average retirement funds used as at censor date or date of delinquency/ default is $199,167.74, which ranges from zero (for borrowers not using CPF savings for mortgage payments) to $1,131,160.16. The mean value of DMORTCPF is 1.6772 with a standard deviation of 2.6996. The averages of PROBNEQ1 and PROBNEQ2 are 0.1260 and 0.1255 and their standard deviations are 0.1165 and 0.1161 respectively. The average valuation is $670,357 with a range of $447,000 to $3,400,000. While PREMIUM has a wide range, the average value is close to zero. The amount of CPF lump sum used by the borrowers ranges from zero (not utilized) to $631,000. The resultant CPFPRICE ranges from zero to 92.21%. The average CPFPRICE is 17.41%. The average mortgage term (MT) is 24.08 years, which range from 3 to 33 years.11 66.82% of the sample has initial preferential rates (PRM). 73.52% of the sample cases are leasehold properties and the remainders are of freehold tenure. Property type (TYPE) is dominated by high-rise properties. 546 (86.26%) of the mortgages were backed by either condominium housing or apartments. Terraces, semi-detached housing or detached housing, backed the remaining mortgages. The average land area (LAREA) of the low-rise properties is 2,436 sq ft and it ranges from 1,317 sq ft to 8,256 sq ft depending on whether they are terraces, semi-detached or detached housing, in ascending order of the level of land area. The floor levels (FLOOR), associated with high-rise properties, range from 1st to 33rd storey with the average level of 6.8163. The mean built-up area (BUAREA) is 1515 sq ft. Monthly mortgage installments payable has an average of $2,098. The corresponding PINCRATIO ranges from 0.009 to 0.84 with an average of 0.265. The number of borrowers (BORROWER) varies from 1 to 5 with a mean of 2.07. PURPOSE 10 We acknowledge that mortgages may exhibit seasoning effects. However, there are only 9 mortgages that were originated before 1990. The subsequent results remain unchanged when mortgages originated before 1990 were excluded. 11 The maximum mortgage term is 30 years, but the maximum term of 33 years is due to negotiations between the delinquent borrower and the lender after loan origination to extend the period over which the loan shall be paid. 25 is dominated by that of owner-occupation at 95.6% or 605 cases while the reminder is purchased for investment. The average YRSEMP is 9.3 years with standard deviation of 7.69. In the period under study, unemployment rates have been increasing due to the economic crisis in the region. The stock market exhibited relatively large standard deviations over the period from the origination dates of the loans to the delinquency and default dates or the censor dates. Exhibit 2: Descriptive Statistics Loan Amount (S) LVR Retirement Funds used ($) DMORTCPF PROBNEQ1 PROBNEQ2 PREMIUM CPFPRICE DVL1 DVL2 DLOSS1 DLOSS2 CCAP1 CCAP2 PEQLOSS1 PEQLOSS2 MT FRM TENURE TYPE LAREA FLOOR BUAREA PINCRATIO BORROWER AGE PURPOSE OCCUP YRSEMP CUNEMP1 CUNEMP2 All Loans* Mean Std.Dev. 364,682 16,1589 0.56662 0.169857 19,9168 1.67717 0.125974 0.125535 0.00133478 0.174087 0.551343 0.551343 0.0643 0.6027 0.262243 0.262243 0.0363 0.0357 24.0774 0.668246 0.729858 0.862559 334.849 5.89021 1515.85 0.265437 2.06951 36.3924 0.0442338 0.229068 9.12317 0.522937 0.514662 144,334 2.69959 0.116484 0.116056 0.0529033 0.148288 0.49775 0.49775 0.1152 0.1189 0.440202 0.440202 0.0769 0.0761 6.29587 0.471215 0.444384 0.344584 914.556 5.61821 577.372 0.11331 0.543869 7.09307 0.205777 0.420565 7.68984 0.583481 0.582245 Loans with Positive Equity# Mean Std.Dev. 361,637 162,796 0.557116 0.168416 202,748 1.68467 0.110265 0.10983 0.00113918 0.180076 0.53333 0.53333 0.06179 0.06128 0.21833 0.21833 0.02828 0.02775 23.9183 0.693333 0.72 0.855 353.266 5.86417 1528.35 0.260666 2.07 36.5643 0.0466667 0.228333 9.23411 0.491607 0.484109 146,672 2.76697 0.0949849 0.0944695 0.0537147 0.149153 0.49930 0.49930 0.10032 0.09941 0.41346 0.41346 0.06648 0.06536 6.35287 0.461495 0.449374 0.352395 935.936 5.71224 587.125 0.111416 0.555831 7.11967 0.2111 0.420109 7.7146 0.57405 0.573364 Loans with Negative Equity@ Mean Std.Dev. 420,038 127,521 0.739408 0.080006 134,068 1.53571 0.410659 0.410125 0.00489122 0.0651947 0.87879 0.87879 0.04183 0.04183 0.87889 0.87889 0.18245 0.18169 26.9697 0.212121 0.909091 1 0 6.36364 1288.64 0.352173 2.06061 33.2668 0 0.242424 7.10606 1.09257 1.07018 64,323.2 0.581734 0.103049 0.102675 0.0353986 0.0703916 0.33143 0.33143 0.07105 0.07105 0.33143 0.33143 0.10400 0.10580 4.27555 0.415149 0.291937 0 0 3.5162 273.894 0.11411 0.242306 5.84633 0 0.435194 7.0298 0.449903 0.453836 26 CGDP1 0.178057 0.21474 0.177477 0.21986 0.188597 0.076049 CGDP2 0.180145 0.214834 0.179455 0.219989 0.192685 0.0739465 CSTI1 0.0260575 0.276199 0.0354261 0.27918 -0.144282 0.126909 CSTI2 0.0335119 0.277631 0.0431259 0.280655 -0.141288 0.121038 *Total sample size is 633 observations, where 133 are delinquent loans and 55 are defaulted loans. # Sample size of loans with positive equity is 600, where 123 are delinquent loans and 49 are defaulted loans. @ Sample size of loans with negative equity (estimated house value < outstanding loan) is 33, where 10 are delinquent loans and 6 are defaulted loans. 6. Empirical Analysis We estimated separate probit models for default and delinquency. This is to examine the importance and divergence of the two measures of mortgage risks. Protected Equity As motivated earlier, we examine three measures of protected borrower equity to evaluate their effect of delinquency and default performances. The first measure CPFPRICE represents the effect of protected equity committed at loan origination on delinquency/default risk. The results in Exhibit 3 show that CPFPRICE has a positive and significant relationship with delinquency (regression 1). As a higher CPFPRICE represents smaller cash equity for a given LVR, this result is not surprising. Interestingly, the amount of CPF lump sum utilized has a positive but insignificant relationship with default, suggesting that while borrowers are more to go into delinquency when a higher proportion of equity is protected, this protection is not a significant factor in default. We defer a more in-depth discussion on this issue to a latter section. As an aside, it is noted that higher loan to value (LVR) mortgages are more likely to go into delinquency as expected, but less likely to lead to default. This is likely to be due to the lack of non-recourse protection. The put option variable is consistently insignificant in the three regressions, which is not entirely surprising given that the fact that mortgages in Singapore are not non-recourse loans. The second measure of protected equity is GUARPX, which represents the impact of protected equity over the life of the mortgage as the loan is paid down and equity is 27 built up as at the date of delinquency/ default if delinquency/ default occur or at the date of censor if otherwise. This includes both the initial amount used at origination and the estimated amount accumulated with the monthly servicing of the mortgage. GUARPX is not significant factor in delinquency and default (regression 2). We re-run our regressions with GUARPX replaced with an absolute measure of the amount of protected equity as at the date of censor or the date of delinquency/ default. The effect is positive, although insignificant. We attribute the sensitivity of results on GUARPX to estimation error in computing the amount of accumulated CPF money through monthly contributions. Alternatively, monthly CPF contributions are already controlled for by way of DCPF and DMORTCPF variables. The third measure of protected equity provides clearer evidence. The equity-tovalue ratio (EVR) directly measures the effect of borrower equity in a mortgage loan while the interactive variable EVRCPF measures the extent of protected equity within total borrower equity. In other words, we are differentiating the influence of cash equity from that of protected equity. We expect a higher initial EVR to discourage delinquency/default but if the high EVR is due primarily to a high CPFPRICE, then it is possible that the protected nature of the retirement funds will encourage delinquency/default instead. The initial EVR is negatively and significantly correlated to delinquency risks although the point estimate for EVR for default is positive but insignificant12 (Exhibit 3 regression 3). This apparent inconsistency between delinquency and default seems awkward and maybe due to the effects of varying proportion of cash versus protected equity within this borrower equity. Therefore, we look to EVRCPF for a more coherent explanation for this. EVRCPF yields consistently positive signs in accordance to our theoretical predictions. Overall, the evidence suggests that a high EVR due to high proportion of CPFPRICE tends to lead to higher delinquency likelihood. These results suggest that our conventional understanding of the effect of borrower equity on delinquency / default has to be modified for situations where part of the equity is from retirement funds. Hence, borrowers with high EVR can no longer be treated as having lower risks. The self-selection argument may more appropriately be 12 This result is consistent with the negative coefficient observed for LVR on default in regression 1. 28 directed towards EVRCPF. Specifically, riskier borrowers will choose their EVR together with the CPFPRICE simultaneously. Thus, this group of borrowers will be characterized by higher EVRCPF. Monthly CPF contributions DCPF and DMORTCPF provide insights on the extent to which having the retirement funds to make up for any cash insufficiencies in times of crisis are able to moderate mortgage risks. The results on delinquency and default across the 3 regressions show that borrowers who are not able to pay the entire monthly mortgage payment using CPF funds tend to have higher default rates in accordance to our predictions. The unexpected finding is that borrowers who do not service repayments entirely using CPF contributions have a lower delinquency risk. Among the borrowers who are not paying the entire mortgage installment via CPF, greater ease of achieving 100% CPF servicing as their salary increased over time after origination (DMORTCPF) is shown to be useful in averting default (positive albeit insignificant effect). This is in accordance to our hypothesis. Overall, our results here suggest that the policy in allowing retirement funds to service mortgage debt payments reduces default risk. The findings on CPF utilization and mortgage service are unchanged even after we allow for dependence between delinquency and default in a bivariate probit model. In addition, we estimate a conditional default model (based on delinquent loans) and find that CPF utilization at origination leads to higher conditional default risk while adequate cover through CPF contributions reduces conditional default probability.13 The effects of other variables are mainly inline with expectations. For more information and elaboration, please refer to Teo (2004; 2005). The analysis in Exhibit 3 focuses only on the amount of protected equity that is committed at the time of loan origination. Do potential wealth losses or liability arising from equity that is protected affect the probability of delinquency and default? The next section addresses this question. 13 Results available upon request. 29 Exhibit 3: Effect of Protected Equity on Delinquency and Default Determinant Regression 1 Delinq Default Coeff Std Err Coeff Std Err ONE 1.00 1.99 0.94 4.40 Protected Equity-related Variables LVR 1.77 * 1.04 -3.23 * 1.99 CPFPRICE 6.46 *** 1.76 2.06 3.38 GUARPX EVR EVRCPF DCPF -0.54 * 0.33 1.22 ** 0.58 DMORTCPF 0.00 ** 0.00 0.00 0.00 AGEGPX -0.23 *** 0.04 -0.09 0.07 Option-related Variable PROBNEQ1 -0.00 1.13 PROBNEQ2 3.31 2.07 Mortgage Loan Specific Variables REPEAT 4.18 17.35 0.82 ** 0.34 PREMIUM -1.60 2.13 -1.54 2.80 MT -0.00 0.03 0.00 0.03 PRM -3.31 *** 0.35 -2.23 *** 0.52 Borrower Specific Variables AGE 0.05 ** 0.02 0.02 0.03 PINCRATIO -1.34 1.15 -1.61 1.54 BORROWER -0.00 0.21 0.07 0.30 PURPOSE 0.50 0.42 0.68 0.69 OCCUP 0.11 0.24 0.93 ** 0.38 YRSEMP -0.00 0.01 0.01 0.02 p 0.99 0.07 ln L -221.60 McFadden R2 0.55 Regression 2 Delinq Default Coeff Std Err Coeff Std Err 2.90 1.77 0.44 3.87 0.14 0.75 -3.49 * 1.84 -2.95 2.27 -1.18 4.21 -0.53 * 0.00 -0.03 0.31 0.00 0.07 -0.00 0.36 1.27 ** 0.00 -0.02 0.59 0.00 0.11 3.07 1.96 Regression 3 Delinq Default Coeff Std Err Coeff Std Err 2.07 1.65 -2.91 4.10 -1.88 * 8.01 *** -0.43 0.00 ** -0.18 *** 1.00 2.13 0.32 0.00 0.03 -0.00 1.08 1.91 7.40 1.15 ** 0.00 -0.13 ** 2.29 5.86 0.57 0.00 0.06 3.01 2.08 3.77 -0.82 0.00 -3.48 *** 12.50 2.34 0.02 0.31 0.84 ** -1.47 0.01 -2.21 *** 0.35 2.64 0.03 0.50 4.19 -1.64 0.00 -3.22 *** 19.83 2.22 0.02 0.34 0.84 ** -2.12 0.01 -2.02 *** 0.34 2.82 0.04 0.51 -0.02 -1.01 0.00 0.36 0.20 -0.00 0.99 -231.56 0.58 0.02 1.11 0.19 0.39 0.23 0.01 0.13 0.01 -1.40 0.11 0.61 1.06 0.01 0.03 1.53 0.32 0.68 0.38 0.02 0.04 ** -0.89 0.00 0.56 0.12 -0.00 0.99 -223.73 0.59 0.02 1.07 0.21 0.41 0.25 0.01 0.20 0.03 -1.04 0.05 0.76 0.94 ** 0.01 0.06 1.61 0.32 0.73 0.39 0.02 *** significant at 1%; ** significant at 5%; * significant at 10% The probit models include property specific variables (TENURE, TYPE, LAREA, FLOOR, BUAREA) and environmental variables (CGDP, CSTI CUNEMP). 30 Loss Aversion The first measure of loss aversion focuses simply on the effect of equity loss (when estimated value is less than purchase price). An equity loss, whether measured as a dummy variable (DVL) or quantified variable (DLOSS), has a positive and significant effect on both default and delinquency probabilities (see Exhibit 4, regressions 4 & 5). While this may be somewhat unexpected, it could be because distressed borrowers are not as worried about averting losses especially if cash equity is regarded as a sunk cost. We also note that the impact of the quantum of equity loss (DLOSS) on default is not significant. To further unravel the impact of equity losses, we look to the effect of potential wealth loss or protected equity at risk (CCAP). Interestingly, once the fall in house value is greater than the cash equity and impinges into protected equity (when CCAP = 1 and when PEQLOSS increases), the probability of delinquency and default becomes lower. The results from regressions 6 and 7 of Exhibit 4 show that the probabilities of delinquency and default are reduced when the borrower’s protected equity is at risk (and the borrower potentially faces a wealth reduction / conversion of the shortfall into unsecured loan / bankruptcy proceedings). In other words, this finding suggests that borrowers are cognizant of the higher costs associated with shortfall arising from default and when the sale proceeds are insufficient to cover the liability to the bank after deducting CPF funds (protected equity). Consequently, borrowers are averse to incurring this wealth loss. In contrast, they are less averse to incurring losses in committed (sunk) cash equity. We note that while this result may seem contradictory to the earlier finding that a higher use of CPF funds increases the likelihood of delinquency, it is actually not so. This is because CPFPRICE captures the amount of CPF funds used at the point of purchase and PEQLOSS quantifies the potential reduction in wealth or financial liability that the borrower would incur should the proceeds from the defaulted property are insufficient to cover outstanding loan as a result of protected equity. In other words, PEQLOSS is a more pertinent measure of loss, and borrowers are averse to this more immediate loss (as opposed to sunk loss). 31 Exhibit 4: Effect of Loss Aversion on Delinquency and Default Determinant Regression 4 Delinq Default Coeff Std Coeff Std Err Err Protected Equity-related Variables EVR -0.75 0.73 0.71 1.02 EVRCPF 6.18*** 1.74 4.74* 2.52 Regression 5 Delinq Default Coeff Std Coeff Std Err Err Regression 6 Delinq Default Coeff Std Coeff Std Err Err Regression l 7 Delinq Default Coeff Std Coeff Std Err Err -0.96 6.49*** 0.73 1.72 0.97 4.23* 0.93 2.27 -0.42 5.53*** 0.72 1.69 1.11 3.93* 0.93 2.28 -0.50 5.36*** 0.72 1.69 1.06 3.83* 0.93 2.28 DCPF DMORTCPF -0.23 0.00** 0.21 0.00 0.61* 0.00* 0.32 0.00 -0.30 0.00*** 0.21 0.00 0.60** 0.00 0.31 0.00 -0.21 0.00** 0.21 0.00 0.62** 0.00 0.31 0.00 -0.21 0.00** 0.21 0.00 0.62* 0.00 0.31 0.00 AGEGPX -0.00*** 0.00 -0.00*** 0.00 -0.00*** 0.00 -0.00*** 0.00 -0.00*** 0.00 -0.00*** 0.00 -0.00*** 0.00 -0.00*** 0.00 AGE 0.04*** 0.01 0.03* 0.02 0.04*** 0.01 0.03* 0.02 0.04*** 0.01 0.03* 0.02 0.04** 0.01 0.03* 0.02 1.46*** 0.36 3.75*** 0.88 1.29 1.04 -0.63** 0.28 -0.21 0.32 -2.96** 1.34 Loss Aversion Variables DVL1 0.83*** DVL2 DLOSS1 DLOSS2 CCAP1 CCAP2 PEQLOSS1 PEQLOSS2 -193.0779 ln L 0.91 McFadden R2 0.25 -111.0092 0.92 -189.9149 0.90 -119.4022 0.93 -196.0607 0.91 -119.9117 0.93 -196.1057 0.91 -1.15 1.61 -119.8771 0.93 *** significant at 1%; ** significant at 5%; * significant at 10% The probit models include property specific variables (TENURE, TYPE, LAREA, FLOOR, BUAREA), borrower specific characteristics (PINCRATIO, BORROWER, PURPOSE, YRSEMP, OCCUP) and environmental variables (CGDP, CSTI CUNEMP). 32 Time to Wealth Realization Another interesting observation is the negative coefficient on AGEGPX. Recall that the CPF scheme may have an inter-temporal effect on borrower behavior in that any CPF refund upon resale or foreclosure goes back to the borrower’s account and is not accessible until retirement age of 55 is reached. AGEGPX captures the interactive effect of time to wealth change realization and amount of protected equity. The negative and significant coefficients consistently observed through out all regressions (Exhibits 3 & 4) mean that the probabilities of delinquency and default decrease when time to retirement (wealth realization) is shorter and when the amount of protected equity is larger. Interpreted together with the positive coefficient on AGE, the results suggest that the amount of deferred protected equity matters when lock-up period is short. This result is consistent with that for CCAP and PEQLOSS. Interpreted differently, borrowers of a given age are less likely to default / become delinquent when the amount of protected equity increases. This is consistent with the magnitude effect for given time delay (Thaler, 1981; Benzion, et al., 1989). However, do younger borrowers differ in delinquency probability when CPF is used (i.e. is there a time delay effect)? Regression 2 in Exhibit 3 includes GUARPX, AGE and AGEGPX variables, so controlling for given CPF sum utilized, we find that the coefficient on AGEGPX is negative but insignificant. In addition, are younger borrowers are more loss averse when protected equity is at risk? We investigate this question by introducing a new variable AGECCAP (AGE*CCAP) in regression 6 of Exhibit 4, with and without AGEGPX; results not reported. The coefficient on AGECCAP is positive but insignificant. Our findings do not provide strong support for our expectation that younger borrowers should be more loss averse. Future research would seek to further unravel the differential effects of time and loss quantum. Further tests were carried out where interactive variables were created to evaluate the effect of DVL and CCAP on CPFPRICE, GUARPR and EVR to safeguard against biases arising from a relatively small sample. The results are consistent with those reported earlier. 33 6. Conclusion While the extant literature is well developed in understanding the effect of equity (and debt) on delinquency and default, we seek to provide further insights into a different variant of borrower equity where part of equity is protected. The CPF scheme in Singapore stipulated that the refund of borrower’s retirement funds utilized for property purchase takes priority over loan obligations. This legislature was deemed to be necessary to preserve the retirement nest egg. While the ability to utilize CPF retirement funds is widely acknowledged to facilitate home ownership, a rational decision to fully utilize CPF for property purchase actually increases ex post delinquency and default risk as it effectively reduces cash equity commitment. In particular, any erosion in house value that places protected equity at risk translates into potential reduction in wealth or financial liability for the borrower. Our results demonstrate that the borrower is strongly averse to incurring protected equity-induced wealth loss or financial liability. While loss aversion is evident for non-distressed sellers (Genesove and Mayer, 2001), the effect of equity losses for distressed borrowers is not as clear. Our research suggests that averting losses in committed cash equity may be a secondary consideration to say income shocks, recognizing that delinquency and default are precursors to foreclosure. There is also a tradeoff between discount rates over time – between incurring a loss now versus a compensating gain / protection from CPF funds – and the quantum of that financial liability. From a policy perspective, this study shows how government policies and regulations can manipulate the default and delinquency performances of mortgages and their securities. More practically, understanding the effects of protected equity is significant to policy-makers and advocates of pension fund reforms. As pension funds are being liberalized or reformed, the use of such funds to finance the purchase of housing may become a possibility. This will probably be advocated to improve the accessibility and affordability of homeownership. The Singapore ‘anomaly’, where pension funds are allowed to finance home purchases and to service the monthly mortgage installments, provides an opportunity to investigate the potential impacts of such a policy on mortgage risk. 34 Equally important to local context, the direction and strength of the relationship between the use of CPF savings and the risk of default will likely provide an answer to whether the policy of allowing CPF funds to partially furnish the mortgage is a bane for mortgage securitization in Singapore. Despite tax incentives and regulatory liberalizations, Singapore has yet to see its first MBS. It is generally believed that the use of CPF savings is holding up securitization of mortgages. The main reason is the regulation that stipulates the CPF Board as having first lien on the properties; i.e. private lenders have secondary claims on the properties is a deterrent in mortgage securitization. Our research shows that this may only be a misperception. Borrowers in our sample actually exhibit a lower likelihood to go into delinquency and default when house values fall below the sum of outstanding loan and CPF funds committed. In other words, the protection accruing to CPF funds actually deter delinquency and default as any shortfall from such refund is translated into immediate financial liability. As such, by contributing to a better understanding of the effects of the protection conferred to the retirement funds used, our research could underpin further development of Singapore’s securitization market. While the CPF legislation has been amended, all mortgages that originated before September 2002 are still subjected to the CPF priority rule. It will be interesting to evaluate how CPF utilization for mortgages originated after September 2002 affect delinquency and default, but that remains for future research. Since the new regulations provide that lender obligations takes priority, any losses suffered when house values fall below outstanding loans could be absorbed by adjustments in the retirement fund. If the amount of CPF funds used is sufficiently large, any negative equity may be comfortably absorbed by borrowers with long time to retirement. Intuitively, the recent regulation change could increase delinquency and default risk, and possibly prepayment risk. In addition, future research on post September 2002 mortgages may offer further insights on inter-temporal choice and time inconsistency, in particular to examine how borrowers view gains and losses in immediate and deferred equity. 35 References Ambrose, B.W. and Buttimer, R.J. (2000). Embedded Options in the Mortgage Contract. Journal of Real Estate Finance and Economics, 21(2), 95-111. Ambrose, B.W., Buttimer, R.J. and Capone, C.A. (1997). Pricing Mortgage Default and Foreclosure Delay. Journal of Money, Credit, and Banking, 29(3), 314-325. Ambrose, B.W. and Capone, C.A. (1996). Cost-Benefit Analysis of Single-Family Foreclosure Alternatives. Journal of Real Estate Finance and Economics,13(2), 105-120. Ambrose, B.W. and Capone, C.A. (1998). Modeling the Conditional Probability of Foreclosure in the Context of Single-Family Mortgage Default Resolutions. Real Estate Economics, 26(3), 391429. Ambrose, B.W. and Capone, C. A. (2000). The Hazard Rates of First and Second Defaults. Journal of Real Estate Finance and Economics, 20(3), 275-293. Ambrose, B.W. and LaCour-Little, M. (2001). Prepayment Risk in Adjustable Rate Mortgages Subject to Initial Year Discounts: Some New Evidence, Real Estate Economics, 29(2), 305-327. Benzion, U., Rapoport, A. and Yagil, J. (1989) Discount Rates Inferred from Decisions: An Experimental Study, Management Science, 35, 270 – 284. Campbell, T., and J. Kimball Dietrich (1983). The Determinants of Default on Insured Conventional Residential Loans. Journal of Finance, 38(December), 1569-1581. Canner, G.B., Gabriel S.A. & Woolley J.M. (1991). Race, Default Risk and Mortgage Lending: A Study of the FHA and Conventional Loan Markets. Southern Economic Journal, 58(1), 249-262. Capozza, R., Kazarian, D. and Thomas, A. (1997). Mortgage Defaults in Local Markets. Real Estate Economics, 25(4), 631-655. Capozza, R., Kazarian, D. and Thomas, A. (1998). The Conditional Probability of Default. Real Estate Economics, 26(3), 359-389. Cox, D.R. and Oakes, D. (1984). Analysis of Survival Data, Monographs on Statistics and Applied Probability. London: Chapman and Hall. Clapp, J.M., Deng, Y, and An, X. (2004). Alternative Models for Competing Risks of Mortgage Termination. Working Paper. Cunningham, F. and Capone, A. (1990). The Relative Termination Experience of Adjustable to Fixed-Rate Mortgages. The Journal of Finance, 45(5), 1687-1703. Deng, Y. (1997). Mortgage Termination: An Empirical Hazard Model with Stochastic Term Structure. Journal of Real Estate Finance and Economics, 14, 309-331. Deng, Y., Quigley, J.M. and van Order, R. (2000). Mortgage Terminations, Heterogeneity and the Exercise of Mortgage Options. Econometrica, 68(2), 275-307. 36 Deng, Y. and Gabriel, S. (2002). Enhancing Mortgage Credit Availability Among Underserved and Higher Credit-risk Populations: An Assessment of Default and Prepayment Option Exercise Among FHA-Insured Borrowers. Working Paper, Marshall School of Business, USC. Foster, C. and Van Order, R. (1984). An Option-Based Model of Mortgage Default. Housing Finance Review, 3(4), 351-372. Genesove, D. and C. Mayer (2001). "Loss aversion and seller behavior: Evidence from the housing market. Quarterly Journal of Economics, 116(4), 1233-1260. Green, R.K. and LaCour-Little, M. (1999). Some Truths about Ostriches: Who Doesn’t Prepay Their Mortgages and Why They Don’t. Journal of Housing Economics, 8(3), 233-248. Greene W.H. (2003). Econometric Analysis. New Jersey: Prentice Hall. Hakim, R. and Haddad, M. (1999). Borrower Attributes and the Risk of Default of Conventional Mortgages. Atlantic Economic Journal, 27(2), 210-220. Han, A. and Hausman, J.A. (1990). Flexible Parametric Estimation of Duration and Competing Risk Models. Journal of Applied Econometrics, 5, 1-28. Heckman, J.J. (1979). Sample Selection Bias as a Specification Factor. Econometrica, 47(1), 153162. Herzog, J. and Earley, J. (1970). Home Mortgage Delinquency and Foreclosure. New York: National Bureau of Economic Research. Holmes, C. (2003). The Outcome of Commercial Mortgage Delinquency: Foreclosure or Reinstatement. Working Paper, Sauder School of Business, UBC. Hunt, G.L. (2000). Alternative Nested Logit Model Structures and the Special Case of Partial Degeneracy. Journal of Regional Science, 40(1), 89-113. Kahneman, D. and Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk. Econometrica, 47, 263-91. 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 and Banking, 24(3), 279-299. Kau, J. B., Keenan, D.C. and Epperson, J.F. (1993). Option Theory and Floating-Rate Securities with a Comparison of Adjustable- and Fixed-Rate Mortages. Journal of Business, 66(4), 595-618. Kau, J.B., Keenan, D.C. and Kim, T. (1993). Transaction Costs, Suboptimal Termination and Default Probabilities. Journal of the American Real Estate and Urban Economics Association, 21(3), 247-263. Kau, J.B., Keenan, D.C. and Kim, T. (1994). Default Probabilities for Mortgages. Journal of Urban Economics, 35, 278-296. Khor, A. & Ong, S.E. (1998). Housing loans: finding the best fit. Business Times Property Supplement, 15 July, p.6. 37 Knapp, T.A., White, N.E., and Clark, D.E. (2001). A Nested Logit Approach to Household Mobility. Journal of Regional Science, 41(1), 1-22. Lambrecht, B.M., Perraudin, W.R.M., and Satchell, S.E. (1997). Time to Default in the UK Mortgage Market. Economic Modelling, 14(4), 485-499. Lea, M.J. and Zorn, P.M. (1986). Adjustable-rate Mortgages, Economic Fluctuations, and Lender Portfolio Change. Journal of The American Real Estate and Urban Economics Association, 14(3), 432-447. Loewenstein, G. and Thaler, R. H., (1989). Anomalise: Intertemporal Choice. Journal of Economic Perspectives, 3(4), 181 – 193. Markowitz, H. (1952). The Utility of Wealth. Journal of Political Economy, 60, 151-158. McCall, B.P. (1996). Unemployment Insurance Rules, Joblessness, and Part-time Work. Econometrica, 64, 647-682. McFadden, D. (1973). Conditional Logit Analysis of Qualitative Choice Behavior. Frontiers in Econometrics. New York: Academic Press. Morton, T.G. (1975). A Discriminant Function Analysis of Residential Mortgage Delinquency and Foreclosure. Journal of The American Real Estate and Urban Economics Association, 3(1), 73-90. Neo, P.H. and Ong, S.E. (2004). Risk Sharing in Mortgage Loan Agreements. Review of Pacific Basin Financial Markets and Policies, 7(2), 233 – 258. Neo, P. H., Ong, S. E. and Somerville, T. (2005) Who are Loss Averse and From Which Reference Point? European Real Estate Society conference, Dublin. Odean, T. (1998). Are Investors Reluctant to Realize their Losses? Journal of Finance, 53(5), 1775-1798. Ong, S.E. (2000). Prepayment Risk and Holding Period for Residential Mortgages in Singapore. Journal of Property Investment & Finance, 18(6), 586-601. Ong, S.E., Ooi, J. and Sing, T.F. (2000). Asset Securitization in Singapore: A Tale of Three Vehicles. Real Estate Finance, 17(2), 47-56. Ong, S.E., Maxam, C.L. and Thang, C.L. (2002). Mortgagor Motivations in Prepayments for Adjustable Rate Mortgages. Review of Urban & Regional Development Studies, 14(2), 97-116. Phillips, R.A. and VanderHoff, J.H. (2004). The Conditional Probability of Foreclosure: An Empirical Analysis of Conventional Mortgage Loan Defaults. Real Estate Economics, 32(4), 571587. Quigley, J.M. and Van Order, R. (1990). Efficiency in the Mortgage Market: The Borrower’s Perspective. Journal of The American Real Estate and Urban Economics Association, 18(1), 237252. 38 Quigley, J.M., and Van Order, R. (1991). Defaults on Mortgage Obligations and Capital Requirements for U.S. Savings Institutions: A Policy Perspective. Journal of Public Economics, 44(3), 353-370. Quigley, J.M. and Van Order, R. (1995). Explicit Test of Contingent Claims Models of Mortgage Default. The Journal of Real Estate Finance and Economics, 11(2), 99-117. Quercia, R.G. and Stegman, M.A. (1992). Residential Mortgage Default: A Review of the Literature. Journal of Housing Research, 3(2), 341-379. Schwartz, E.S. and Torous, W.N. (1993). Mortgage Prepayment and Default Decisions: A Poisson Regression Approach. Journal of the American Real Estate and Urban Economics Association, 21 (4), 431 – 449. Shefrin, H. and Statman, M. (1985). The Disposition to Sell Winners too Early and Ride Losers too Long: Theory and Evidence. Journal of Finance, 40(3), 777-790. Sing, T.F. and Ong, S.E. (2004). Residential Mortgage Backed Securitization in Asia: The Singapore Experience. Journal of Real Estate Literature, 12(2), 159-179. Stanton, R. (1995). Rational Prepayment and the Valuation of Mortgage-Backed Securities. The Review of Financial Studies, 8, 677-708. Sueyoshi, G. (1992). Semiparametric Proportional Hazards Estimation of Competing Risks Models with Time-varying Covariates. Journal of Econometrics, 51, 25-58. Teo, A., Time in Delinquency: Implications for Mortgage Lending and MBS. Briefings in Real Estate Finance, 2005, 4(4) 275 - 289. Teo, A., Delinquency Risk in Residential ARMs: A Hazard Function Approach, Journal of Real Estate Portfolio Management, 2004, 10(3), 243 - 258. Thaler, T. H., (1981). Some Empirical Evidence on Dynamic Inconsistency, Economic Letters, 8, 201 – 207. Tversky, A. and Kahneman, D. (1991). Loss Aversion in Riskless Choice: A Reference Dependent Model. Quarterly Journal of Economics, 106(4), 1039-1061. Tversky, A. and Kahneman, D. (1992). Advances in Prospect Theory: Cumulative Representation of Uncertainty. Journal of Risk and Uncertainty, 5, 297-323. Waller, G. (1988). Residential Mortgage Default: A Clarifying Analysis. Housing Finance Review, 7, 321-333. Waller, G. (1989). Residential Mortgage Default: An Empirical Note on Changes in Property Value and Debt Over Time. Housing Finance Review, 8, 155-164. Vandell, K.D. (1993) “Handling Over the Keys: A Perspective on Mortgage Default Research”, Journal of the American Estate and Urban Economics Association, 21(3), 211 – 246. 39 Vandell, K.D. and Thibodeau, T. (1985). Estimation of Mortgage Defaults Using Disaggregate Loan History Data. ARUEA Journal, 13(3), 292-316. Von Furstenberg, G.M. & Green, R.J. (1974). Home Mortgage Delinquencies: A Cohort Analysis. The Journal of Finance, 29(3), 1545-1548. Zorn, M. and Lea, J. (1989). Mortgage Borrower Repayment Behaviour: A Microeconomic Analysis with Canadian Adjustable Rate Mortgage Data. Journal of the American Estate and Urban Economics Association, 17(1), 118-136. 40