A CLASS OF HEATH-JARROW-MORTON TERM STRUCTURE MODELS WITH STOCHASTIC VOLATILITY

CARL CHIARELLA AND OH KANG KWON School of Finance and Economics University of Technology Sydney PO Box 123 Broadway NSW 2007 Australia carl.chiarella@uts.edu.au ohkang.kwon@uts.edu.au A BSTRACT. This paper considers a class of Heath-Jarrow-Morton term structure models with stochastic volatility. These models admit transformations to Markovian systems, and consequently lend themselves to well-established solution techniques for the bond and bond option prices. Solutions for certain special cases are obtained, and compared against their non-stochastic counterparts. K EY W ORDS : stochastic volatility, interest rate modeling, Heath-Jarrow-Morton, bond option

1. I NTRODUCTION The majority of the term structure models prior to Heath, Jarrow and Morton (1992) were nite dimensional Markovian systems in which the interest rate economy was determined by the spot rate and perhaps one or two additional state variables. This enabled the use of standard arbitrage arguments, along the lines of Black and Scholes (1973) and Merton (1973), to derive the PDE for the bond and bond option prices which, in turn, enabled the application of well-developed techniques from the theory of PDEs to obtain analytic solutions, and numerical solutions in cases where this was not possible. The progenitors of this approach could be regarded as Vasicek (1977) and Brennan and Schwartz (1979). Although these early models were useful from the viewpoint that analytic solutions were often available, the calibration of model parameters to observed market data was a highly non-trivial task. In particular, many models could not be calibrated consistently to the initial yield curve, and the relationship between the model parameters and the market observed variables were not always clear. Furthermore, it was not always possible to incorporate observed market features, such as the humped volatility curve, into these models. By contrast, the Heath, Jarrow and Morton (1992) approach provides a very general interest rate framework, capable of incorporating most, if not all, of the market observed features. The HJM models are automatically calibrated to the initial yield curve, and the
Date: First version April 10, 1998. Current version November 1, 1999.
1

connection between the model parameters and the market variables often emerge from the theory. The main drawback of the HJM framework is that it results in models that are nonMarkovian in general, and consequently the techniques from the theory of PDEs no longer apply. For the general HJM model, Monte Carlo simulation, which can often be time consuming, is the only method of solution. To overcome these problems, many authors, including Carverhill (1994), Ritchken and Sankarasubramanian (1995), Bhar and Chiarella (1997), Chiarella and Kwon (1998a), Chiarella and Kwon (1998b), and Bhar, Chiarella, El-Hassan and Zheng (1999), have considered ways of transforming the HJM models to Markovian systems. In these transformed systems, the desirable properties of the earlier Markovian models and the HJM framework coexist, and provide useful settings under which to study interest rate derivatives. In the standard HJM framework, the uncertainty in the interest rate market is represented by Wiener processes that drive the forward rate process. All other processes in the interest rate market, including the forward rate volatilities, are thus also driven by the same Wiener processes. Consequently, the standard HJM model does not incorporate additional independent sources of stochastic volatility, as considered in Hull and White (1987), Heston (1993), and Scott (1997). The main contribution of this paper is the specication of volatility processes that are driven by additional Wiener processes that are independent of those that drive the forward rate process in the standard HJM framework. Such models will be referred to as stochastic volatility HJM models. The stochastic HJM models considered in this paper transform to Markovian systems, and hence enjoy the benets enjoyed by such models. For certain special cases, explicit bond price formulae, in the spirit of Ritchken and Sankarasubramanian (1995), Inui and Kijima (1998), and Chiarella and Kwon (1998a) are given, along with numerical examples highlighting the effect of the additional Wiener processes. The class of models constructed in this paper can, in some sense, be considered as analogues of the Hull and White (1987) and Heston (1993) stochastic volatility models within the HJM framework, and provides one way of incorporating stochastic volatility into the HJM framework, as alluded to in Jarrow (1997). The remainder of this paper is organised as follows. In 2 the HJM framework is briey outlined. The stochastic volatility model is then introduced in a simplied -dimensional setting in 3, and the general stochastic model is described in 4. Numerical examples illustrating the effect of stochastic volatility are given in 5, and the paper concludes with 6. 2. H EATH -JARROW-M ORTON F RAMEWORK In this section, an overview of the general Heath-Jarrow-Morton framework is given. For further details, the reader is referred to Heath, Jarrow and Morton (1992), Brace and Musiela (1994), or Musiela and Rutkowski (1997). 2.1. The Risk-Neutral Framework. Let , and assume given a ltered prob satisfying the usual conditions, with ltration ability space generated by an -dimensional standard -Wiener process .
2

In the standard HJM interest rate framework, the time instantaneous rate of return on an investment contracted at time , denoted , is assumed to satisfy the stochastic integral equation (2.1)

where , , and represents the dependence of the forward rate process on the Wiener path . For nite dimensional Markovian specialisations of the HJM model, the path ( ) dependence simplies to dependence on a nite number of state variables such as the spot rate . Relatively mild regularity assumptions are imposed on so that the integrals are well dened, and required manipulations are valid. The spot rate process, , is obtained by setting in (2.1), so that (2.2)

The money market account , representing the time made at time , is given by the equation (2.3)

value of unit investment

Finally, the time price of a maturity zero coupon bond , denoted as (2.4)

, is dened

The differential forms of (2.1) and (2.2) are easily obtained and have the form (2.5)

(2.6)

By using the stochastic Fubinis theorem and It os lemma, HJM showed that satises (2.7) where

.
3

2.2. Derivative Prices as Conditional Expectations. It follows from It os lemma and (2.7) that the discounted bond price process satises (2.8)

and is consequently an Thus (2.9)

-martingale under the equivalent martingale measure .

Substituting for and , and imposing the condition

yields the important identity (2.10)

which gives the bond price as the expected value of the discounted payoff. More generally, if is a price process for a -expiry option on , , with payoff , then (2.11)

2.3. Feynman-Kac Theorem and Partial Differential Equations. If the interest rate economy is Markovian, then an application of the Feynman-Kac Thoerem to (2.10) results in the partial differential equation (2.12)

for the bond price, where is the innitesimal generator associated with the Markovian system resulting from (2.5) and (2.6). There are well-developed theoretical and numerical techniques for solving such equations, and consequently many authors, including Ritchken and Sankarasubramanian (1995), Inui and Kijima (1998), Chiarella and Kwon (1998a), Chiarella and Kwon (1998b), and Bhar, Chiarella, El-Hassan and Zheng (1999), have studied conditions under which the HJM model transforms to Markovian systems. 3. O NE D IMENSIONAL HJM M ODELS
WITH

S TOCHASTIC VOLATILITY

Being path dependent, the volatility processes in the standard HJM framework are technically stochastic. However, in the literature (see Hull and White (1987), Heston (1993), and Scott (1997)), the term stochastic volatility appears to be reserved only for those volatility processes which are driven by Wiener processes linearly independent of the Wiener processes that drive the underlying asset price process, or, as in this case, the forward rate process. It is in this sense that the standard HJM models fail to be stochastic volatility models. A good discussion of stochastic volatility models in a non-stochastic interest rate environment is Heston (1993) and Scott (1997) in a generalised setting. In order to give a transparent exposition of the main ideas of this paper, the special case of -dimensional stochastic volatility HJM models are considered in this section. The -dimensional generalisation is considered in 4.
4

3.1. Embedding Stochastic Volatility in the HJM Framework. In the case of the -dimensional HJM model, recall from (2.5) that the forward rate process evolves according to the stochastic differential equation (3.1)

Separable volatility processes of the form (3.2)

where is a deterministic function, have been considered by numerous authors (Ritchken and Sankarasubramanian (1995), Inui and Kijima (1998), Bhar, Chiarella, El-Hassan and Zheng (1999)), and have been shown to generate a useful class of Markovian interest models in which a closed form formula for the bond price is available. It must be noted that an additional assumption such as must be made in order to obtain a Markovian model, although the bond price formula remains valid in the absence of such assumptions. For example, in the Vasicek type model the additional assumption is made, while for the CIR type model, the , where is a constant, assumption made is . Stochastic volatility may be introduced into the standard HJM model in several ways, but the method adopted in this paper is to assume a volatility process of the form

is a vector of a nite set of xed where tenor forward rates with , and are deterministic functions, and is a stochastic process satisfying (3.4) where is a constant, , , and are deterministic functions, and is a
(3.3)

standard Wiener process such that

Appropriate choice of the parameter allows the volatility or variance to be modeled as

(3.5) a stochastic process. An example of such a volatility specication is

with (3.6)

satisfying

where , , , and are constants. To apply the techniques of Heath, Jarrow and Morton (1992), it is convenient to replace the correlated Wiener processes and with uncorrelated processes, and it is easily seen that and , dened by (3.7) (3.8)

have this property. Inversion of the pair of equations above yields (3.9) (3.10)

and substituting (3.10) into (3.1) gives

Now dene (3.11) (3.12) (3.13)

Then (3.1) can be rewritten as

which is the starting point in Heath, Jarrow and Morton (1992), for a -dimensional model with volatility processes . Assuming the existence of market prices of , associated with the two sources of uncertainty, and following risk ,
(3.14)

the arbitrage arguments of Heath, Jarrow and Morton (1992), the forward rate process is seen to satisfy

(3.15)

where , and are standard Wiener processes under the equivalent martingale measure . The general framework developed in Heath, Jarrow and Morton (1992) applies verbatim to the stochastic volatility model introduced in this section, and in particular, expressions such as (2.2), (2.6), and (2.10) remain valid for the stochastic volatility model. 3.2. Markovian System. The volatility process of the form (3.3) satises the Markovian condition given in Inui and Kijima (1998) and Chiarella and Kwon (1998a), and, consequently, the corresponding HJM model transforms to a Markovian system. Since the transformation in the case of stochastic volatility has not been considered in previous literature, the argument is briey outlined below. Dene state variables and by (3.16) (3.17)

and for (3.18) (3.19) Note that (3.20) (3.21) where (3.22) (3.23)

dene

Lemma 3.1.

and . The variables and satisfy the equations

Proof. See Appendix A. Since the volatility process (3.3) is a function of forward rates , , the equations governing the evolution of the forward rates must also be computed. For this, note that, from (3.15), the stochastic integral equation for is

is

Lemma 3.2. The stochastic differential equation for the forward rate (3.24) where

Proof. See Appendix B. Proposition 3.3. Let the market price of risk be a function of . Then the set and , where forms an -dimensional Markovian system. Proof. From Appendix C, the state variables

, , ,

and from (3.4), satises the equation (3.25)

and satisfy the equations

Recall

that

.
7

Now, if is a function of , , , and , then together with (3.2), these equations determine a Markovian system having , , , and as the state variables. Note the dependence of the drift term in (3.25) on the market price of risk . This is analogous to the corresponding situation in the stochastic volatility models of Hull and White (1987), Heston (1993), and Scott (1997), and arises from market incompleteness, which is in turn a result of the absence of a traded asset related to volatility. 3.3. State Variables and as Functions of Forward Rates. The economic signicance of the state variables and is not at all clear from (3.16) and (3.17). In this subsection, it is shown that they can, in fact, be expressed as linear combinations of forward rates, and hence a useful interpretation of the state variables in terms of economically meaningful variables is established. Setting

in (2.1), and using Lemma 3.1, the equation for the forward rate can be written (3.26)
For notational convenience, let (3.27) (3.28)

and

Then (3.26) can be written

An important feature of (3.28) is that and are deterministic functions, and so, for any , (3.28) gives the value of , and hence the value of , as linear combination of the state variables and . Another useful feature of (3.28) is that it can be used to express the state variables and as linear combinations of a nite set of forward rates. For this, x two tenors . Then setting in (3.28), for , gives rise to the system (3.29) Inverting this system of equations, the state variables and can be written as linear combinations of forward rates and in the form (3.30) (3.31)

3.4. Pricing Partial Differential Equation. Since the stochastic volatility model introduced in this section is Markovian, the Feynman-Kac Theorem can be applied to obtain the pricing PDE for interest rate contingent claims. Recall that if the state variables , , for a Markovian system satises (3.32)

where (3.33) where (3.34)

, then the associated innitesimal generator is given by

For the case at hand, the state variables are given by Proposition 3.3, and the equations determining the drift and diffusion coefcients are given in Lemma 3.2 and Proposition 3.3. To clarify the above discussion, the innitesimal generator in the case where and is given below. For this case, the only forward rate contained in the set of state variables is the spot rate, and the relevant equations are

(3.35)

where

The innitesimal generator is then given by

(3.36)

and the pricing PDE for a contingent claim is given by (3.37)

subject to appropriate boundary conditions. 3.5. Bond Price as a Function of the State Variables. The bond price formula for HJM models driven by separable volatility processes was obtained by Ritchken and Sankarasubramanian (1995) for the one dimensional case, and extended to the general case by Inui and Kijima (1998). The formula was then further generalised by Chiarella and Kwon (1998a) to forward rate dependent volatility processes. In this subsection, a brief outline of the derivation of the bond price formula is given. Note that from (2.4) and (3.15), the price of a -maturity bond is given by the equation (3.38)

From Appendix D,

(3.39) where (3.40)

, and consequently

Note that the state variables and can be replaced by forward rates, as shown in

3.3.

3.6. Pricing of European Bond Options. To compute the price of a -maturity European option on a -maturity bond, with payoff , the system of stochastic differential equations (3.35) must be solved numerically for the values of state variables , , , and , and the discount factor , at option maturity . The bond price can then be computed using (3.40), and this, together with . Repeating the payoff function , determines the option price the simulation a suitable number of times, and taking the average, then gives the value of the European option. Results from implementing this procedure are given in 5. 4. G ENERAL HJM M ODELS
WITH

S TOCHASTIC VOLATILITY

In this section, the introduction of stochastic volatility into the -dimensional HJM model, presented in 3, is extended to the general -dimensional HJM framework. 4.1. Embedding Stochastic Volatility in the HJM Framework. Recall from (2.5) that in the -factor risk-neutral HJM framework, the instantaneous forward rate process evolves according to the equation (4.1)

where are independent standard Wiener processes. Let be a positive integer, and x a sequence . Dene a vector of forward rates by (4.2)

As in the -dimensional case, to introduce stochastic volatility into the HJM model, the are assumed to be of the form (4.3) where (4.4)
Note

and

are deterministic functions,

that since

, and satisfy
is redundant.

, the variable

10

where , , and further assumed that (4.5) where dened by (4.6) (4.7)

are deterministic functions, and , for

. It is

, and

if if

for all

. Then the Wiener processes, ,

are independent, and inversion of the the above system yields (4.8) (4.9)

Substituting (4.9) into (4.1) gives

Now dene (4.10) (4.11) (4.12)

Then (4.1) can be written (4.13)

which is the starting point in Heath, Jarrow and Morton (1992) with volatility processes . Assuming the existence of market prices of risk , , and following the arguments of Heath, Jarrow and Morton (1992), one obtains (4.14)

where and are standard Wiener processes under the equivalent martingale measure . The results obtained in Heath, Jarrow and Morton (1992) remain valid for the stochastic volatility model introduced in this section modulo the obvious modications, viz. replacing by and by .
restriction is imposed purely to simplify exposition. The analysis in this section remains valid for any correlation structure.
11
This

4.2. Markovian System. The volatility processes given by (4.11) and (4.12) satisfy the Markovian condition given in Inui and Kijima (1998) and Chiarella and Kwon (1998a), and the corresponding HJM model transforms a nite dimensional Markovian system. Since the method for transforming the model to a Markovian system closely parallels the method employed in 3, the proofs are kept brief. For , dene variables and by (4.15) (4.16) and for (4.17) (4.18) (4.19)

dene and by

Lemma 4.1. The variables (4.20) (4.21) where (4.22) (4.23)

and can be expressed in the form

Proof. See Appendix A. Lemma 4.2. The stochastic differential equation for the forward rate (4.24)

is

where

.
12

Proof. See Appendix B.

Lemma 4.3. The state variables tions (4.25) (4.26)

and satisfy the stochastic differential equa

Proof. See Appendix C.

be a function of , , , and , where , for each . Then the set forms an -dimensional Markovian system.
Proposition 4.4. Let and Proof. From (4.4), the SDE governing the evolution of processes is (4.27)

with respect to Wiener

Since are assumed to be functions of the state variables, the result now follows from Lemma 4.2 and Lemma 4.3. 4.3. State Variables and in Terms of Forward Rates. As in the -dimensional case, the state variables and can be expressed in terms of a nite number of xed tenor forward rates. Setting in (2.1), and using Lemma 4.1, the equation for the forward rate can be written (4.28) Let (4.29)

and

Then (4.28) can be written (4.30)

Once again, since and are deterministic functions, (4.30) gives the value of , and hence the value of , as linear combination of a nite number of state variables and . Now, x tenors . Then setting in (4.30), for , gives rise to the linear system (4.31)

13

where

. . .

. . .

. . .

and can be expressed as .

Inverting this system of equations, the state variables linear combinations of forward rates ,

4.4. Bond Price as a Function of the State Variables. From Appendix D, the price of a -maturity bond, in the stochastic volatility HJM model introduced in this section, is (4.32)

where . Note from 4.3 that the state variables and can be expressed in terms of a nite number of xed tenor forward rates, and so the stochastic volatility model of this paper falls under the exponential afne class of models in the sense of Dufe and Kan (1996). 5. N UMERICAL E XAMPLE In this section, a special case of the general stochastic volatility framework introduced in 4 is considered to illustrate the effect of stochastic volatility on the spot rate, the bond price, and the European call price. 5.1. Model Specication. The model used for numerical simulation in this section is the -dimensional model of 3, with volatility process given by (5.1) where (5.2) where , , and

and

are constants. The dynamics of is restricted to be of the form

are constants.
14

5.2. System of Stochastic Differential Equations. The model specied by (5.1) and (5.2) is a -dimensional Markovian system with state variables , , and . The system of SDEs governing the dynamics are: (5.3) (5.4) (5.5)

5.3. Parameters and Simulation. The set of parameters used for the simulation are as follows: (5.6) (i) time (ii) time

Antithetic variable method was used to compute

distribution of the spot rate and the -year bond price, price of an at-the-money -month call option on the -year bond,

for various correlation coefcients and the volatility of volatility . 5.4. Distribution of Spot Rate and Bond Price. The effect of varying the correlation between the Wiener process driving the forward rate process and that driving the stochastic scaling factor is illustrated in Figure 5.1 for the distribution of the spot rate, and Figure 5.2 for the distribution of the bond price.
0.06 -0.75 -0.50 -0.25 0.00 0.25 0.50 0.75

0.05

0.04

0.03

0.02

0.01

0 0.0492

0.0494

0.0496

0.0498

0.05

0.0502

0.0504

0.0506

0.0508

F IGURE 5.1. Distribution of Spot Rate with

and Varying

The effect of varying the volatility of volatility, , on these distributions is illustrated in Figure 5.3 and Figure 5.4. It is interesting to observe from Figure 5.1 and Figure 5.2 that increasing from negative to positive values tends to skew the spot distribution to the left and the bond distribution to the right. Both distributions tend to peak around zero correlation between the two
15

0.06 -0.75 -0.50 -0.25 0.00 0.25 0.50 0.75

0.05

0.04

0.03

0.02

0.01

0 0.87

0.8705

0.871

0.8715

0.872

0.8725

0.873

F IGURE 5.2. Distribution of Bond Price with

0.06

and Varying
0.0 0.2 0.4 0.6 0.8

0.05

0.04

0.03

0.02

0.01

0 0.0492

0.0494

0.0496

0.0498

0.05

0.0502

0.0504

0.0506

0.0508

F IGURE 5.3. Distribution of Spot Rate with

and Varying

Wiener processes. When there is positive correlation between the two Wiener processes, Figure 5.3 and Figure 5.4 show that an increase in volatility of volatility skews the spot rate distribution to the left and the bond price distribution to the right. 5.5. Call Price as Function of and . The effect of varying the correlation and the volatility of volatility on the price of a -month call option on a -year bond is illustrated in Figure 5.5 and Figure 5.6 respectively. Figure 5.5 shows that the call option value increases with increasing correlation , and this is to be expected given the skewing to the right of the bond price distribution in this situation. Similarly, Figure 5.6 shows that the call option value increases with increasing volatility of volatility.
16

0.06 0.0 0.2 0.4 0.6 0.8

0.05

0.04

0.03

0.02

0.01

0 0.87

0.8705

0.871

0.8715

0.872

0.8725

0.873

F IGURE 5.4. Distribution of Bond Price with

0.0002

and Varying

0.00018

0.00016

0.00014

0.00012

0.0001

8e-005

6e-005 -1 -0.5 0 0.5 1

F IGURE 5.5. Call Price against with 6. C ONCLUSION

A fairly broad class of forward rate volatility processes within the HJM framework has been considered in this paper. These forward rates depend on a set of xed tenor forward rates, time dependent quantities, and stochastic quantities driven by Wiener processes independent from those driving the forward rate dynamics. It is shown how the stochastic dynamics of the resulting system can be reduced to a Markovian form, and that many of the subsidiary state variables introduced by the Markovian reduction procedure can be expressed in terms of a set of xed tenor forward rates. The only non-market traded quantities in the stochastic dynamics are the stochastic volatility quantities.
17

0.0002

0.00018

0.00016

0.00014

0.00012

0.0001 0 0.2 0.4 0.6 0.8 1

F IGURE 5.6. Call Price against

with

It is possible to obtain an explicit formula for the bond prices in this framework so that Monte Carlo simulation is required only for the calculation of option prices. This fact makes option pricing feasible with the class of stochastic volatility models presented in this paper. Some numerical results have been given indicating how the level of the volatility of volatility, and the correlation between the noises driving the forward rates and the stochastic volatility, affect the spot rate, bond price, and European call option values. These calculations indicate the computational feasibility of the approach developed in this paper, at least as far as off-line calculations are concerned. No doubt further research on numerical methods tailored to these stochastic volatility models would also make feasible calculations in trading time. It has also been shown how the models developed in this paper may be viewed as a subclass of the exponential afne class of interest rate models considered by Dufe and Kan (1996). A PPENDIX A. P ROOF
OF

L EMMA 3.1

AND

L EMMA 4.1

Consider rstly Lemma 4.1. It follows easily from denition (4.3) that

Using this identity,

18

Next, for (A.1) so that

, let

. Then for

and likewise

Addition of the previous two equations yields the required identity

Lemma 3.1 is obtained by setting .

A PPENDIX B. P ROOF
OF

L EMMA 3.2

AND

L EMMA 4.2

Consider the more general case of Lemma 4.2, and, for notational convenience, write for each (B.1) (B.2)

can be written

Then the integral equation for the forward rate process (B.3)

19

Assuming volatility processes of the form (3.3), and differentiating w.r.t. (B.4) (B.5) for (B.6) (B.7)

. Similar computations establish

These equations combine to yield the expression

and the result follows from Appendix A. Setting case.

yields Lemma 3.2 as a special

A PPENDIX C. P ROOF The key identity (C.1)

OF

L EMMA 4.3

follows from (4.3). Now, rst consider (4.25).

by (C.1)

Next consider (4.26). From (B.1) and (B.2), (C.2)

and from (B.4)-(B.7),

20

A PPENDIX D. P ROOF Recall the denitions of (D.1) Then

OF

(3.39)

AND

(4.32)

and from Lemma 3.1, and dene

and for

Analogous computation yields

Now, using (A.1) and Fubinis Theorem

for

, and similar computations yield

Addition of the previous two equations implies

21

Summation over

results in the equation

and hence

R EFERENCES
Bhar, R. and Chiarella, C. (1997), Transformation of Heath-Jarrow-Morton Models to Markovian Systems, European Journal of Finance 3, 126. Bhar, R., Chiarella, C., El-Hassan, N. and Zheng, X. (1999), Reduction of Forward Rate Dependent HJM Models to Markovian Form: Pricing European Bond Options, Working paper, School of Finance and Economics, University of Techonology Sydney. Black, F. and Scholes, M. (1973), The Pricing of Options and Corporate Liabilities, Journal of Political Economy 81, 637659. Brace, A. and Musiela, M. (1994), A Multifactor Gauss Markov Implementation of Heath, Jarrow, and Morton, Mathematical Finance 4(3), 259283. Brennan, M. and Schwartz, E. (1979), A Continuous Time Approach to the Pricing of Interest Rates, Journal of Banking and Finance 3, 133155. Carverhill, A. (1994), When is the Short Rate Markovian?, Mathematical Finance 4(4), 305312. Chiarella, C. and Kwon, O. (1998a), Forward Rate Dependent Markovian Transformations of the HeathJarrow-Morton Term Structure Model, Working paper, School of Finance and Economics, University of Techonology Sydney. Chiarella, C. and Kwon, O. (1998b), Square Root Afne Transformations of the Heath-Jarrow-Morton Term Structure Model and Partial Differential Equations, Working paper, School of Finance and Economics, University of Techonology Sydney. Dufe, D. and Kan, R. (1996), A Yield Factor Model of Interest Rates, Mathematical Finance 6(4), 379 406. Heath, D., Jarrow, R. and Morton, A. (1992), Bond Princing and the Term Structure of Interest Rates: A New Methodology for Contingent Claim Valuation, Econometrica 60(1), 77105. Heston, S. (1993), A closed form solution for options with stochastic volatility with application to bond and currency options, Review of Financial Studies 6(2), 327343. Hull, J. and White, A. (1987), The Pricing of Options on Assets with Stochastic Volatilities, Journal of Finance 42(2), 281300. Inui, K. and Kijima, M. (1998), A Markovian Framework in Multi-Factor Heath-Jarrow-Morton Models, Journal of Financial and Quantitative Analysis 33(3), 423440. Jarrow, R. (1997), The HJM Model: Its Past, Present, and Future, Journal of Financial Engineering 6(4), 269279. Keynote Address, IAFE 1997 Conference. Merton, R. (1973), The Theory of Rational Option Pricing, Bell Jounal of Economics and Management Science 4, 141183. Musiela, M. and Rutkowski, M. (1997), Martingale Methods in Financial Modelling, rst edn, SpringerVerlag, New York. Applications of Mathematics 30. Ritchken, P. and Sankarasubramanian, L. (1995), Volatility Structures of Forward Rates and the Dynamics of the Term Structure, Mathematical Finance 5(1), 5572. Scott, L. (1997), Pricing Stock Options in a Jump-Diffusion Model with Stochastic Volatility and Interest Interest Rates: Applications of Fourier Inversion Methods, Mathematical Finance 7(4), 413426. Vasicek, O. (1977), An Equilibrium Characterisation of the Term Structure, Journal of Financial Economics 5, 177188.

22