Pricing Credit Risk of Asset-Backed Securitization Bonds in Singapore TIEN FOO SING*, SEOW ENG ONG, GANG-ZHI FAN Department of Real Estate, School of Design & Environment, National University of Singapore, 4 Architecture Drive, Singapore 117566 *rststf@nus.edu.sg and KIAN GUAN LIM School of Business, Singapore Management University, Singapore Revised: June 17, 2004 Abstract Asset-backed securitization (ABS) is a creative arrangement to raise funds through the issuance of marketable securities backed by predictable future cash flows from revenueproducing assets. This paper proposes two pricing models: structural model and intensity model, to value credit spreads on Singapore ABS bonds. Sensitivity analyses were conducted on the ABS credit spreads by varying the values of the key input variables within a plausible range. The property price volatility and its correlations with risk-less interest rates have been shown to have positive effects on the ABS credit spreads. However, when the market volatility is extremely high, the credit spreads decrease with an increase in the time to maturity. The positive effects of the property price volatility were significantly reduced when credit enhancements were added to the ABS bonds, and the credit risks associated with the correlation variable were fully eliminated in the credit enhanced ABS bonds. The rate of loss recovery in the event of default also has significant impact on the credit risks of the ABS bonds. ABS bonds backed by physical property will likely to have high recovery rates thus reducing the credit risks vis-à-vis noncollateralized bonds. Keywords: Asset Backed Securitization; credit risk; structural model; intensity model * We would like to thank National University of Singapore for funding the research project. We also wish to thank Brent Ambrose, CF Sirmans, anonymous referees and others for their constructive comments and suggestions on the earlier version of the paper. Errors, if any, remain the responsibility of the authors. 1 Pricing Credit Risk of Asset-Backed Securitization Bonds in Singapore 1. Introduction Asset-backed securitization (ABS) is a creative arrangement to raise funds by issuing marketable securities backed by future cash flows from revenue-producing assets. In the Singapore context, the assets consist of real estate assets with regular cash flows. A special-purpose vehicle (SPV) is set up in the ABS structure, which issues debt securities to finance the purchase of the real estate assets. After the Asian financial crisis in 1997, ABS transactions became popular in Singapore. The first securitization deal involved the sale of Neptune Orient Lines (NOL)1 headquarter office building in 1999. 10-year fixedrate bonds were issued through a special purpose vehicle (SPV) Chenab Investments Limited, to fund the purchase of the 26-storey building valued at S$185 million. Six other asset securitization transactions were subsequently concluded in 1999. The total bonds issued from the seven ABS transactions amounted to S$1.84 billion. Securitization of S$2billion worth of non-real estate assets was also initiated by the Development Bank of Singapore (DBS)2, with the first tranche of S$200 million short-term interest bearing notes launched in June 2000.3 The asset pool comprises loans granted to statutory boards of Singapore and bonds issued by government-linked companies. The short-term notes were the first asset backed securitization instrument that was assigned an “A-1” rating by the Standard and Poor’s, which marks an important milestone in the development of ABS market in Singapore. ABS is a relatively new financial innovation in Singapore. As yet, there has been no academic study on credit risk evaluation on Singapore ABS transactions due largely to shortage of empirical data given the short history of Singapore ABS. Traditional valuation methods estimate credit spreads by analyzing ex-post default data using 1 2 3 Founded in 1968, Neptune Orient Lines (NOL) is the largest shipping company listed on the Singapore Exchange. NOL operates a network of container transportation and logistic services on major international trade routes. The Development Bank of Singapore (DBS) listed on the Singapore Exchange is one of the largest banks in Singapore, in term of market capitalization. Siow, L.S., “DBS offers retail investors short-term interest notes,” Business Times Singapore, June 1, 2000. 2 empirical credit risk models. The modern derivative theory offers a theoretical approach to pricing credit risk of defaultable bonds, which include the ABS bonds in this paper. There are two approaches commonly used to evaluate credit risk of defaultable bonds: a structural-based (firm value) model and an intensity-based model. In the structural or firm value approach, default and recovery rates are determined based on the evolution of the firm value relative to its liability. While in the intensity model, an exogenous default process is directly modeled as a Poisson process. This study attempts to employ the two models to separately analyze and value credit risk in the ABS transactions. The organization of the paper is as follows: Section 2 reviews the relevant literature on pricing defaultable debt instruments. Section 3 discusses the key security features relating to the selection and applicability of credit risk quantitative valuation techniques. Section 4 specifies the two comparable approaches proposed to price the credit spreads of ABS bonds. Sections 5 and 6 evaluate the numerical properties of the credit spreads based on the proposed models. Section 7 summarizes the results of the investigation and draws some conclusions. 2. Valuation of Defaultable Debts Based on the classical option pricing theory first developed by Black and Scholes (1973) and Merton (1973), Merton (1974) prices the default risk of defaultable bond as a European put option on the total value of firm assets. The firm value is assumed to evolve following a stochastic diffusion process. The defaultable zero-coupon bond is then modeled as an arbitrage-free portfolio consisting of a long position in an equivalent Treasury bond and a short position in a put option on the firm value with an exercise price equal to the bond’s face value. Default occurs when the firm value drops below the face value of its bond at maturity. Merton’s (1974) model suffers several drawbacks. Firstly, default risk is independent of interest rate risk and different seniority levels. Secondly, the possibility of default before debt maturity was excluded. Jones et al (1984) and Franks and Torous (1989) showed that the empirical credit spreads on corporate 3 bonds were too high to be consistent with those generated from the Merton’s model with non-stochastic interest rates. Shimko et al. (1993) make a significant extension to Merton’s model by allowing stochastic interest rates in the model. The interest rates are assumed to follow the Vasicek (1977) process, which fits many observed term structures of interest rates. In the Vasicek process, it is possible to generate negative interest rates. They derived closed-form solutions for defaultable zero-coupon bonds, and showed that the correlation between the returns of underlying asset and the interest rate movements plays an important role in determining the credit spread on bonds. However, their model does not allow default before debt maturity. Kim et al. (1993) developed their corporate bond pricing model by defining default as the time when the firm value first reaches a pre-specified constant default threshold. They incorporated stochastic interest rates in the model, which are assumed to evolve following a square-root CIR process (Cox, Ingersoll, and Ross, 1985). When the default threshold is crossed, default occurs and the firm is assumed to go bankrupt immediately. Bondholders will then exogenously receive a given amount of riskless bonds upon default. They showed that the proposed model is able to generate corporate spreads that match with those observed in practice (also see Nielsen et al., 1993; Longstaff and Schwartz, 1995; and Briys and de Varenne, 1997). This structural valuation framework of Kim et al. (1993) will be adopted as the first credit risk model for ABS bonds. Structural models have, however, some practical limitations. Firstly, it is difficult to estimate the required input parameters related to the value of the firm. The firm’s assets are generally not frequently traded in financial markets. The firm value is, therefore, not directly observed. In our case, the assets are commercial properties. If transactions of comparable properties are not available in the market, the price movements of these properties can still be indirectly inferred through relevant property indices published by relevant authorities. Secondly, structural models do not explicitly account for credit rating information that reflects the credit quality changes of defaultable corporate bonds. 4 It is important, however, in practice, to incorporate the relevant credit-rating information into the proposed valuation models. The intensity model offers an alternative approach, in which rating agencies’ data or other relevant financial markets series can be incorporated into the valuation process. In the intensity model, default or bankruptcy is directly treated as an unpredictable event (i.e. a Poisson process). The probability of a jump from no-default to default in a given time interval is measured by a hazard rate, which is also called an intensity rate. This group of credit risk models have been examined in financial literature, which includes Ramaswamy and Sundaresan (1986), Madan and Unal (1994, 1999), Jarrow and Turnbull (1995), Duffie and Huang (1996), Schönbucher (1997), Duffie and Singleton (1999) and others. In Ramaswamy and Sundaresan (1986) intensity model, they assumed that the instantaneous default risk premium evolves according to a stochastic CIR process (Cox et al. 1985). In a risk-adjusted valuation procedure, Duffie and Singleton (1999) show that risky corporate bonds can be directly priced by discounting promised payments using risk-adjusted interest rates. In Jarrow and Turnbull (1995) intensity model, default is governed by a constant-parameter Poisson process that is independent of default-free interest rates. The payoff is known before default. More recently, Madan and Unal (1999) proposed a two-factor hazard-rate model. The stochastic hazard rate is specified as a function of the firm’s asset value and the default-free interest rate, which are driven by two independent stochastic processes. They showed that the proposed hazard-rate model generates a rich array of credit-spread shapes that are consistent with those observed in practice. 3. Problem Specifications In a typical ABS deal, an owner of an income-generating property asset sells the asset to a special purpose vehicle (SPV), which is created specifically to hold the asset in the securitization process. This process distinguishes a securitization arrangement from the traditional mortgage-backed or collateralized bond issues (Sing, Ong and Sirmans, 2003). 5 The SPV is usually a shell company with no assets. It will raise funds through the issuance of debt securities to finance the purchase of the securitized assets from the originator. The debt securities contain the feature of an interest-only balloon-payment loan that requires no amortization of principal during the bond periods. The debt securities will be fully redeemed at maturity. The bonds are usually fixed-rate bonds with a term to maturity ranging from 7 to 10 years. The ABS bond issuer relies only on the cash flows produced from the securitized incomegenerating property for the periodic coupon payments. The securitized property also serves as collateral for the ABS bonds. The credit quality of the bonds depends to large extent upon the fluctuation of the underlying property value. When default occurs, the model assumes that the securitized property will be foreclosed and sold immediately. The sale proceeds will then be used to make outstanding coupon payments and also redeem the bonds at par value. With the non-recourse structure of the bonds, the ABS bondholders will bear the risks associated with the shortfalls between property values and ABS issuance bond values. We employ the analytic framework of defaultable bonds to price credit risks of ABS bonds. The model assumes that the ABS bondholders possess a risk-free debt security and simultaneously write to the borrower (SPV) a put option on the underlying property asset. The bondholders receive a coupon spread over and above the Treasury-bill rate as the premium for the put option. If default occurs, the SPV could put the securitized property asset to the bondholders at a strike price that is not more than the present value of the remaining coupon obligations plus the face value of bonds at maturity. For corporate bonds, one of the main problems faced in pricing credit risks using the option pricing techniques is the difficulty in determining the value of miscellaneous assets owned by firms that issues the bonds (Corcoran and Kao, 1997). Like non-recourse loans, the ABS bondholders can only have recourse to the commercial property of the SPV in the event of default. This simplifies the problem in the valuation of ABS credit risk. Compared with other corporate bonds, the structure of ABS consisting of only one 6 real estate asset is more straightforward. Therefore, the ABS is a good case for the application of option pricing approach to value credit risk. In addition to the securitized claims on cash flows generated from the underlying assets, ABS bonds are credit enhanced through both external and internal means like guarantee of minimum rental cash flows by the head-lessee, a senior/subordinated structure, and embedded sale-backed options (Sing, Ong and Sirmans, 2003). The credit enhancements reduce uncertainty in the ABS transactions, and at the same time, increase the credit rating of the bond tranches. The ABS bonds have also a higher credit rating than the corporate rating of the originator because of the bankruptcy-remote feature built-in to the ABS deals. In the structural model, the default risk of ABS bonds can be priced based on the relative value of SPV property to its issued bond/debt value. The model will not be flexible enough to incorporate various credit enhancements for the bonds, such as corporate guarantees included to assure ABS investors of timely payments of bond interests and principals. The corporate guarantees will enhance the credit rating of ABS bonds, but they are independent of whether the underlying property assets are able to produce sufficient incomes to meet the bond payment obligations. The credit enhancements may be an exogenous factor that will shift the optimal triggering boundary of the default options. In other words, the property value may have to fall substantially below its outstanding loan value before a default option will be exercised. This may imply that the structural model may underestimate the credit risks of ABS bonds, when credit enhancements of various forms are incorporated to the bonds. In many credit risk pricing models, prices of underlying assets are assumed to evolve following an exogenous stochastic diffusion process. The prices will only move incrementally subject to predefined parametric constraints, when new information flows in continuously. The price process precludes any sudden falls in prices, and therefore, default never occurs unexpectedly. In the real world, however, shocks and bad news may occur and cause abrupt downward adjustments of property price, which may in turn 7 trigger a sudden default by the bond originator. Therefore, when default event is unpredictable, the reduced form approach may be more appropriate for valuing ABS credit risk. In the model, the probability of default will be treated as an exogenous variable linked to the rating-based information. 4. Valuation Frameworks Two credit risk pricing models are proposed to evaluate the default risks of the ABS bonds in this paper: structural model and intensity model. a) Structural Model In the structural model, credit risk of ABS bonds is driven by two state variables: the underlying property value and the instantaneous risk-less interest rate. Credit enhancements are not considered in this model. The default is triggered by the joint and simultaneous occurrence of two events: when the underlying property value falls below the market value of outstanding principal and interest payments of the bonds, and when the net operating income from the property is less than the scheduled coupon payments at any payment dates. We extend the credit risk pricing framework of Shimko et al. (1993) and Kim et al. (1993), which is used to price zero-coupon bonds, by incorporating discrete coupon payments in the ABS bonds. Like the earlier models, we assume that the evolution of the value of underlying property (V) follows a lognormal diffusion process, which is defined in a specific form as follows, dV = (a − b )Vdt + σ V VdZV , (1) where a is the instantaneous expected rate of return of the underlying property, b is the continuous payout rate for the property, σ V is the instantaneous standard deviation of property return, and Z V is the standardized Wiener process. bV (t ) , in this model, defines the net operating income from the securitized property. We also assume that the instantaneous risk-less interest rate(r) evolves in a meanreverting diffusion process (see Cox, Ingersoll, and Ross, 1985). 8 dr = κ (µ − r )dt + σ r r dZ r (2) where κ is the speed at which the interest rate r reverts to the long-term mean rate, µ is the long-term mean value of r, σ r is the interest rate volatility, and Z r is the standardized Wiener process. A risk-free portfolio consisting of a short position in the ABS bonds and a long position in the underlying asset is constructed under the arbitrage-free conditions. The risk-free portfolio yields an instantaneous risk-free rate of return on the portfolio over a short time interval, dt, in a market where arbitrage is strictly eliminated. Using Itô’s Lemma on the two stochastic variables, the ABS bond pricing equation can be defined by the following partial differential equation (PDE), 1 2 2 ∂2S ∂2S 1 2 ∂2S ∂S ∂S σVV σ σ ρ + + σ r r 2 + (r − b )V + (κ (µ − r ) − λr ) r V V r 2 2 ∂V ∂r ∂V∂r 2 ∂V ∂r ∂S + − rS = 0 ∂t (3) where λ denotes the market price of interest rate risk, and ρ is the instantaneous correlation coefficient between dZ V and dZ r . On the assumption that the ABS bonds consist only of one single fixed-rate tranche with coupons payments due semi-annually, Equation (3) is expanded to include the discrete fixed-rate coupon payments as follows, ∂S ∂S 1 2 2 ∂2S ∂2S 1 2 ∂2S + (κ (µ − r ) − λr ) σVV + σ V σ r rVρ + σ r r 2 + (r − b )V 2 ∂V ∂r 2 ∂V∂r 2 ∂r ∂V ∂S + − rS + ∑ C i δ (t − t i ) = 0 ∂t i (4) 9 where C i is ith constant coupon payment during the bond periods4, δ (⋅) is the Dirac delta function and t i is the time of the ith coupon payment (see, e.g., Cossin and Pirotte, 2001). If ruthless default is not restricted, the bond originator can still choose not to fulfill the redemption obligations at maturity (i.e., t=T), if and only if the following default condition holds, V (T ) < F + C (5) where F represents the outstanding bond values at maturity, C is the constant coupon payment, and V(T) is the value of the underlying property at maturity. Therefore, the value of ABS bonds, S(V, r, T) at maturity is a corner solution, which can be defined as, S(V,r,T) = Min {F + C , V (T )} . (6) At time t prior to bond maturity, the general solution for the ABS bond value with default options can be derived by solving PDE (4) subject to two default conditions5: (a) the net operating income, bV (t ) , is less than or equal to the scheduled coupon payment, C; and (b) the value of the underlying property, V(t), falls to or below the value of ABS bonds, S(V, r, t): bV (t ) ≤ C (7) V (t ) ≤ S (V , r , t ) (8) The two default conditions are jointly binding, which means that default will occur at a coupon payment date only if both conditions are satisfied.6 Therefore, at time t prior to bond maturity, [t < T], the equations can be written simply as follows, 4 5 6 As there are no earlier redemptions of the bonds in the ABS, the fixed-rate coupon payment, C i , is assumed to be constant throughout the bond periods. Leland (1994) suggests that default triggered by the condition that asset value falls to or below the debt value will be irrelevant when the asset value becomes large. He includes an endogenous triggering mechanism that is dependent on firm ability to raise capital to meet the debt obligations, which is consistent with the condition in our equation (7), [ C ≤ bV (t ) ]. A similar default definition can be found in Westhoff et al. (2000), who developed a simulation-based approach of estimating CMBS defaults. 10 V (t ) = Min [ S (V , r , t ), C / b] (9) Like in Titman and Torous (1989) and Childs et. al. (1996), the default may occur in the absence of default cost, at a critical property value, which is way before the lower triggering value is reached, V * (r , t ) = S (V * , r , t ) (10) where V * (r , t ) is determined endogenously.7 If either one of the conditions were not met, default will not occur. In other words, even if the ABS originator was not able to meet the scheduled coupon payments with the net operating income from the property, the default will still not be triggered as long as the property value is still higher than the bond redemption value. Similarly, even if the underlying property value falls below the value of ABS bonds, the originator will still continue to keep the bond alive as long as the net operating income is higher than the scheduled coupon payments. After determining the value of defaultable ABS bonds, we could back-up the yield-tomaturity from the following pricing equation, T S t = ∑ C t e − yt + Fe − y (T −t ) (11) t =1 where S t is the value of ABS bonds at time t, and F is the face value of ABS bond.8 The credit spread, which reflects the default premium over the remaining life of the security, τ where [τ= T – t], can then be determined as the difference between the yieldto-maturity of ABS bond, y (τ ) , and the comparable yield on a Treasury bond with the same maturity, r (τ ) , cs = y (τ ) − r (τ ) 7 8 (12) The critical V (r , t ) is determined endogenously using a spline methodology as in Childs et al. (1996). In order to calculate the yield-to-maturity from equation (10), we have developed a bisection algorithm to approach the real solution of this equation. * 11 b) Intensity Model The intensity-based model is an alternative approach to pricing the credit risk of ABS bonds, where default is modeled as an exogenous Poisson process. The underlying property value and the instantaneous risk-less interest rate are the two state variables in the model, which are assumed to follow the same stochastic processes defined in the previous structural model. Following Wilmott’s (1998) approach in pricing the ABS bonds, S, a perfectly hedged portfolio of a long position in underlying asset and a short position in risk-less zerocoupon bond with price Z (r , t ) is constructed, which is represented as follows, Π = S (V , r , t ) − ∆Z (r , t ) . (13) where ∆ is the hedge ratio for the risk-less bonds. Let the instantaneous default risk be p at time t, the default will be triggered at the first jump of a Poisson process. For a simple case with only binary outcomes, the risky bonds will either be or not be defaulted by the originator. The probability of default in a short time interval from times t to t+dt can be represented as [pdt]. Conversely, [1-pdt] indicates the probability of non-default of the ABS bonds. The value of the risk-less hedge portfolio after the short time interval of dt, if default does not occur, can then be derived as follows, 1 ∂2S ∂2S 1 2 ∂2S ∂S ∂S dΠ = σ V2V 2 r V dV + dr + + σ r r 2 + C dt + σ σ ρ V r 2 ∂V∂r 2 ∂V ∂r ∂V ∂r 2 ∂Z 1 2 ∂ 2 Z ∂Z dr ⋅ − ∆ + σ r r 2 dt + ∂r ∂r ∂t 2 (14) On the other hand, if default occurs with a probability of pdt, the change in value of the portfolio can then be given as, dΠ = − S + Q + O(dt1 2 ) (15) 12 ( where Q is the recovery amount of the ABS bond in default, and O dt1 / 2 ) is an infinitesimal value relative to the loss of the ABS bond. To solve the PDE (12), an optimal ∆ is selected such that the dr term in the equation can be eliminated. By taking the expectation of the change in portfolio value, the basic pricing equation for the ABS bonds with no interim coupon payments can be derived as, ∂2S 1 2 ∂2S ∂S ∂S 1 2 2 ∂2S + + σ r r 2 + (r − b )V + (κ (µ − r ) − λr ) r V σ VV σ σ ρ V r 2 ∂V∂r 2 ∂V ∂r 2 ∂V ∂r ∂S + − (r + p )S + pQ = 0 ∂t (16) When the discrete coupon payments are taken into account, Equation (14) can be further expanded as follows, 1 2 2 ∂2S ∂2S 1 2 ∂2S ∂S ∂S σ VV σ σ ρ + + σ r r 2 + (r − b )V + (κ (µ − r ) − λr ) r V V r 2 2 ∂V∂r 2 ∂V ∂r ∂V ∂r ∂S + − (r + p )S + pQ + ∑ C i δ (t − t i ) = 0 ∂t i (17) which is subject to the following boundary condition, S (V , r , T ) = F + C (18) For simplicity reasons, we assume that the instantaneous risk of default and the recovery rate are constant. In other words, the instantaneous default risk and the recovery rate are not correlated with the two specified stochastic variables.9 The credit spreads in a continuous time intensity-based framework are also estimated by taking the difference between the yield-to-maturity of ABS bond and the comparable yield of the Treasury bond with the same maturity period. 9 The assumptions can be relaxed for more generalized cases. However, this study does not intend to elaborate on the generalized cases, instead it focuses on the analysis of the effects on of these variables on credit spreads. 13 5. Estimation of ABS Credit Spread In this section, we numerically analyze the credit spreads of ABS bonds using the two proposed models. We consider a 10-year, 6% fixed rate coupon ABS bond backed by the rental incomes from a securitized office property. The risk-less rate of return is represented by the coupon rate of 3% for a Treasury bond proxy with a comparable maturity term. In the ABS structure, the securitized property asset is transferred off-thebalance sheet of the original owner of the real estate asset. It insulates the bond investors, on one hand, against any bankruptcy liability of the original owner. On the other hands, bondholders will also have no recourse to the original firm’s asset in the case of default on the bond obligations by the SPV. In other words, the ABS bondholders will only have direct recourse on the income and sale proceeds from the securitized property, if it is foreclosed. The credit spreads of the ABS bonds can be estimated by solving equations (4) (structural model) and (17) (intensity model). There are no analytical solutions to the two equations, explicit finite-difference methods will be used to numerically compute the bond prices and credit spreads from the two equations. For the numerical analyses, we make necessary assumptions for the input parameters in the two stochastic processes, which include: σ V , σ r , ρ , κ , µ , λ and b in equation (4), and σ V , σ r , ρ , κ , µ , λ , b, p and Q in equation (17). We use the secondary data sources10: the Urban Redevelopment Authority (URA)11 of Singapore data on commercial property prices, and the Monetary Authority of Singapore (MAS) data on Treasury bonds over the past fifteen years to provide a close approximation to the input parameters. In this paper, 5-year Treasury bond yield data were used, because longer time series data were available compared to the 7-year and 10year Treasury bonds, which have only a relatively short history from the late 1990s.12 10 11 12 The secondary sources of information may include the Urban Redevelopment Authority (URA) of Singapore published office property price and rental indexes and also the Treasury bond yields published by the Monetary Authority of Singapore (MAS), the de-facto central bank of Singapore. URA is the national planning authority of Singapore entrusted with the role to plan the long-term land use and physical development of the country. The 5-year Treasury bond yields offer a closer available proxy for the risk-less interest rate for the ABS bond valuation, but we recognize that the Treasury bond yields with comparable maturity term of 10 years, if available, should be a better substitute (Titman and Torous, 1989). 14 6. Results of Numerical Analyses This section evaluates the sensitivity of the credit spread estimates to the changes of key input parameters. The objective of these sensitivity analyses is to identify the important factors and how these factors will have significant influences, in terms of both direction and magnitude, on the credit spreads of the ABS bonds. The bond value is normalized to one unit dollar, and the property value is set in a range from 0 to 2.5*F. The effects of the changes in selected key variables, which include property value volatility, ( σ V ), the correlation coefficient parameter, (ρ), and the recovery rate, (Q), on he ABS credit spreads can then be illustrated. For a base case scenario, we specify the parameter values for the relevant parameters based on the historical data of Singapore office property market and the 5-year government Treasury bond yields. The property payout rate, (b), is assumed to be 7% based on the rental data, and the volatility of the office price is estimated at σ V =13% based on the quarterly office rental return of the URA. The correlation coefficient of [ ρ = 0.1 ] is assumed to reflect the historical correlations of the risk-less interest rate changes and property returns over the last fifteen years, which was less than 10 percent. The market price of interest rate risk is set to be [ λ = 0 ]. For the stochastic risk-less interest rates, we use the quarterly yield data for the 5-year Treasury bonds to estimate the reversion speed, [ κ =0.5], the long-term mean of risk-less interest rates, [ µ =0.037], and the volatility of riskless interest rates, [ σ r =0.05]. For the intensity model, two more parameters were assumed for the risk intensity, [p=0.01], and the recovery rate, Q=0.9]. a) Effects of Property Price Volatility By varying the property value volatility parameter from 0% to 45%, we evaluate how the value surfaces of the ABS credit spreads will be affected in both the structural model (Figure 1) and the intensity model (Figure 2). The maturity periods of the ABS bonds are also added to show the time dimension of the volatility-credit spread relationships. 15 In the structural model, where no credit enhancement is included, the ABS credit spreads increase with increases in volatility of the property value over time (Figure 1). However, the rate of changes in the credit spread decrease as the volatility increases. In fact, the positive relationships are no longer observed when the property value volatility exceeds approximately 40%. When the property price volatility is above this level, the reverse effects set in when the maturity of the ABS bonds is shortened. For example, at σ V =45%, the credit spread premiums decrease from 4.62% for a 2-year ABS bonds to 4.41% and 3.56% for a 3-year and a 10-year ABS bonds respectively. We expect the bond originator to keep the “default option” alive when the potential upside associated with the high property price is valuable when the market is highly uncertain. The risk premium for the ABS originator to default will also be higher when the bond maturity is shortened. [Insert Figure 1] The positive credit spread-property price volatility relationships as shown in Figure 2 are more discernible in the intensity model. However, in the relative scale, the credit spread premiums of ABS bonds and also the rate of change of the premium over the volatility changes estimated in the intensity model are much lower than those obtained in the structural model. The effects are, however, independent of the maturity term of the ABS bonds. [Insert Figure 2] The above results imply that ABS bonds without credit enhancement as represented by the structural model are more sensitive to the shocks in the securitized property prices. The rates of credit spread change are also dependent on the maturity term of ABS bonds, if they are not provided with adequate credit enhancements. Clearly, ABS bonds with no credit enhancements are riskier than those with credit enhancements. Therefore, credit enhancements should be used as an effective mean to reduce default risks in the ABS, especially in a high property price volatility environment. 16 b) Effects of Correlation Coefficient between property values and risk-less rates The effects of changing correlations between the risk-less interest rates and the securitized ABS property return are analyzed by varying the coefficient, (ρ), from a negative range of -0.4 to a positive range of +0.5. In the structural model (Figure 3), the credit spread curves shifts in an upward direction along the correlation coefficient axis, but the rate of change is marginal. For a case of 10-year ABS bond, the credit spread premium increases from 0.0303 to 0.0322 when the correlation coefficient changes from –0.4 to 0.4. The effects created by the correlations between the two stochastic processes in the model were eliminated when credit enhancements were included in the intensity model (Figure 4). The credit spread curves were the same over the entire range of maturity periods, despite the changes in the coefficients of correlation. [Insert Figures 3 and 4] The difference in the estimates of the credit spread premium between Figure 3 and Figure 4 show the effects of having credit enhancement in ABS bonds on the credit spreads. This suggests the importance of the credit enhancements in mitigating the default risks associated with the correlations between property returns and risk-less interest rate changes in ABS bonds. c) Intensity-based Factors Effects In the intensity model, default is a Poisson process that is dependent on the instantaneous risk of default, p, and the recovery rate in the event of default, Q. The probability of [1- pdt] indirectly measures the level of credit enhancements within which default of ABS will not occur. The credit enhancement effects will be evaluated by changing the recovery rate in a range from 50% to 90% (Figure 5). [Insert Figure 5] The results show that the ABS credit spread is negatively related to the changes in the recovery rates. The rate of decrease of the ABS credit spread as shown by the slope along 17 the recovery rate axis increases when the term to maturity increases. For a 2-year ABS bonds, every 10% increase in the recovery rate, the credit spread premium decreases by 0.1%, whereas the rate of changes was approximately -0.14% for a 10-year fixed rate coupon payment ABS bonds. The results are not unexpected, which imply that ABS credit risk is lower when ABS bondholders are assured that they can recover a substantial portion of the bond values through foreclosure and disposal of the securitized property in the event of default, i.e. a higher recovery rate. Compared to other paper-based non-recourse bonds, the probability and the rate at which the bondholders can recover their investments from the securitized assets are likely to be higher, since property price is less prone to inflationary risks. The underlying property assets securitized indirectly credit enhance ABS bonds. In summary, our results show that in the absence of credit enhancement, the ABS credit spread is more sensitive to the positive shocks in the securitized property prices over time. However, when credit enhancements are included, the effect of volatility on the credit spread is significantly reduced. It is also shown that the credit spread is an increasing function of the maturity term of ABS bonds. The effects of an increasing ABS bond term are less critical when the price volatility level is high, [ σ V =45%], and when credit enhancements are not available. Another significant and positive determinant of ABS credit spread is the correlation between the risk-less interest rates and the property return. The effect of the correlation can be eliminated by including appropriate credit enhancements to the ABS structure. We also find that the credit spread is a decreasing function of the recovery rates. 7. Conclusion This paper extended and applied two pricing models: structural model and intensity model, to value credit spreads for ABS bonds. Numerical analyses were conducted with a set of input assumptions that represent a base case scenario. Unlike the zero-coupon bond models (Shimko et. al., 1993), the model is extended to allow for discrete coupon payments during the life of the ABS bonds. However, the structural model does not 18 explicitly account for credit enhancements in the ABS bonds. In the intensity process, a Poisson default process is included to represent the exogenous effects of credit enhancements on the ABS credit risks. We could model different level of credit enhancements by varying the intensity of default risks, p, and also the recovery rate, Q, in the event of default. We examine the sensitivities of the ABS credit spreads with respects to changes of key parameters such as property price volatility, correlations between property price and riskless interest rate and also the intensity-based recovery factor. By varying the values of the controlled variable within a plausible range, value surfaces of credit spreads estimated in the two different models could be analyzed and compared. Credit spreads are highly sensitive to the property price volatility, and the risk premiums are higher in the absolute scale when credit enhancements are not included in the ABS structure. In the absence of credit enhancement, credit spreads in a highly volatile environment, ( σ V =40%), decrease with the length of the bond maturity term. The credit spread analysis shows that the effects of an increase in the correlations between property price return and risk-less interest rate on the ABS credit spreads were positive, but at a marginal rate of change. These positive effects were eliminated in the intensity model when credit enhancements are included. Based on the above sensitivity analyses, credit enhancements are important features in ABS that will improve the credit quality and mitigate possible losses of ABS investors in the event of default. The effectiveness of the credit enhancements is, however, dependent on the loss recovery rate when default occurs. ABS bonds backed by physical property are likely to have lower credit risk than non-collateralized ABS bonds, because bondholders is assured that they can recover a substantial portion of the bond values through foreclosure and disposal of the securitized property in the event of default, i.e. a higher recovery rate. 19 References Black, F. and M. Scholes, The Pricing of Options and Corporate Liabilities, Journal of Political Economy, 1973, 81, pp.637-654. Briys, E. and F. de Varenne, Valuing Risky Fixed Rate Debt: An Extension, Journal of Financial and Quantitative Analysis, 1997, 32(2), June, pp. 239-249. Childs, P. D., S. H. Ott and T. J. Riddiough, The Pricing of Multiclass Commercial Mortgage-Backed Securities, Journal of Financial and Quantitative Analysis, 1996, 31(4), pp.581-603. Corcoran, P. and D. L. Kao, Assessing Credit Risk of CMBS, in The Handbook of Commercial Mortgage-Backed Securities, Eds. Frank J. Fabozzi and David P. Jacob, 1997, pp. 251-268, Frank J. Fabozzi Associates and Nomura Securities International, Inc., USA. Cossin, D. and H. Pirotte, Advanced Credit Risk Analysis: Financial Approaches and Mathematical Models to Assess, Price, and Manage Credit Risk, 2000, New York: John Wiley & Sons. Cox, J.C., J. Ingersoll and S. Ross, A Theory of the Term Structure of Interest Rates, Econometrica, 1985, 53, pp.385-407. Duffie, D. and M. Huang, Swap Rates and Credit Quality, Journal of Finance,1996, 51(3), July, pp.921-949. Duffie, D. and K. Singleton, Modeling Term Structures of Defaultable Bonds, Review of Financial Studies, Special 1999, 12(4), pp. 687-720. Fama, E. F., Term Premiums and Default Premiums in Money Markets, Journal of Financial Economics, 1986, September, pp.175-196. Franks, J. R. and W. Torous, An Empirical Investigation of US Firms in Reorganization, the Journal of Finance, 1989, 44, pp.747-769. Jarrow, R. and S. Turnbull, Pricing Derivatives on Financial Securities Subject to Credit Risk, Journal of Finance, 1995, 50(1). March, pp.53-85. Jones, E. Philip, S. P. Mason and E. Rosenfeld, Contingent Claims Analysis of Corporate Capital Structures: An Empirical Investigation, Journal of Finance, 1984, 39(3), pp.611627. Kim, I. J., K. Ramaswamy, and S. Sundaresan, Does Default Risk in Coupons Affect the Valuation of Corporate Bonds?: A Contingent Claims Model, Financial Management, Autumn1993, 22(3), pp.117-131. 20 Leland, H.E., Corporate Debt Value, Bond Covenants, and Optimal Capital Structure, Journal of Finance, 1994, 49(4), pp. 1213-1252. Longstaff, F. and E. Schwartz, A Simple Approach to Valuing Risky Fixed and Floating Rate Debt, Journal of Finance, 1995, 50(3), pp.789-819. Madan, D. and H. Unal, Pricing the Risks of Default, Working Paper, College of Business, University of Maryland, September, 1995. Madan, D. and H. Unal, A Two-Factor Hazard-Rate Model for Pricing Risky Debt and the Term Structure of Credit Spreads, Working Paper, College of Business, University of Maryland, November, 1999. Merton, R. C., Theory of Rational Option Pricing, Bell Journal of Economics and Management Science, 1973, 4, pp.141-183. Merton, R. C., On the Pricing of Corporate Debt: The Risk Structure of Interest Rates, Journal of Finance, 1974, 29, pp.449-470. Nielsen, L. T., J. Saá-Requejo, and P. Santa-Clara, Default Risk and Interest Rate Risk: The Term Structure of Default Spreads, Working Paper, INSEAD, France, 1993. Ramaswamy, K., and S. Sundaresan, The Valuation of Floating-Rate Instruments, Journal of Finance, 1986, 17, February, pp.251-272. Saá-Requejo, J. and P. S. Santa-Clara, Bond Pricing with Default Risk, Working Paper, John E. Anderson Graduate School of Management, UCLA, Los Angeles, 1997, 23. Schönbucher, P., The Term Structure of Defaultable Bond Prices, Working Paper, University of Bonn, 1997. Shimko, D., N. Tejima and D. V. Deventer, The Pricing of Risky Debt When Interest Rates are Stochastic, The Journal of Fixed Income,1993, September, pp.58-65. Sing, T. F., S. E. Ong and C. F. Sirmans, Asset-Backed Securitization in Singapore: Value of Embedded Buy-Back Options, Journal of Real Estate Finance and Economics, 2003, 27:2, pp.173-189. Titman, S. and W. Torous, Valuing Commercial Mortgages: An Empirical Investigation of the Contingent-Claims Approach to Pricing Risky Debt, Journal of Finance, 1989, 44:2, pp.345-373. Vasicek, O., An Equilibrium Characterization of the Term Structure, Journal of Financial Economics, 1977, 5, pp.177-188. Westhoff, D., V. S. Srinivasan and M. Feldman, An Empirical Framework for Estimating CMBS Defaults, in The Handbook of Nonagency Mortgage-Backed Securities, Eds. F. J. 21 Fabozzi, C. Ramsey and M. Marz, 2nd, 2000, pp.435-441, Pennsylvania: Frank J. Fabozzi Associates. Wilmott, P., Derivatives: the theory and practice of financial engineering, Chichester: John Wiley, 1998 22 Figure 1 Sensitivity Measures of Credit Spreads vs. Varying Property Value Volatility (Structural Approach) 23 Figure 2 Sensitivity Measures of Credit Spreads vs. Varying Property Value Volatility (Intensity Approach) 24 Figure 3 Sensitivity Measures of Credit Spreads vs. Varying Correlation (Structural Approach) 25 Figure 4 Sensitivity Measures of credit Spreads vs. Varying Correlation (Intensity Approach) 26 Figure 5 Sensitivity Measures of Credit Spreads vs. Varying Recovery Rate (Intensity Approach) 27