This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

THE PRICING OF MULTIPLE-EXPIRY EXOTICS

HYONG-CHOL O, MUN-CHOL KIM

Faculty of Mathematics, Kim Il Sung University, Pyongyang, D. P. R. Korea
HYONG-CHOL O
Faculty of Mathematics, Kim Il Sung University, Pyongyang, D. P. R. Korea
ohyongchol@yahoo.com MUN-CHOL KIM
Faculty of Mathematics, Kim Il Sung University, Pyongyang, D. P. R. Korea
ohyongchol@163.com
Abstract.

In this paper we extend Buchen’s method to develop a new technique for pricing of some exotic options with several expiry dates (more than 3 expiry dates) using a concept of higher order binary option. At first we introduce the concept of higher order binary option and then provide the pricing formulae of nn-th order binaries using PDE method. After that, we apply them to get the pricing of some multiple-expiry exotic options such as Bermudan option, multi time extendable option, multiple shout option and etc. Here, when calculating the price of concrete multiple-expiry exotic options, we do not try to get the formal solution to corresponding initial-boundary problem of the Black-Scholes equation, but explain how to express the expiry payoffs of the exotic options as a combination of the payoffs of some class of higher order binary options. Once the expiry payoffs are expressed as a linear combination of the payoffs of some class of higher order binary options, in order to avoid arbitrage, the exotic option prices are obtained by static replication with respect to this family of higher order binaries.

Key words and phrases:
Multiple-expiry, Exotic option, Bermudan option, Extendable option, Shout option, Higher order Binary Option
2010 Mathematics Subject Classification:
35C15, 91G80
Last revised Aug 7, 2013, published in Elextronic Journal of Mathematical Analysis and Applications, Vol.1(2) July 2013, pp 247-259

1. Introduction

European and American options are referred to as vanilla options. Vanilla options have a single future expiry payoff that corresponds to buying or selling the underlying asset for a fixed amount called the strike price. European options can only be exercised at the expiry date, whereas American options may be exercised at any time before and at the expiry date. The various needs of risk management in financial markets give rise to many exotic (not ordinary) options with various payoff structures, and thus a lot of exotic options continue to be popular in the over-the counter market. Among them, there exists a class of exotic options whose payoff structure involves several fixed future dates, here we write them by T0<T1<<TNT_{0}<T_{1}<\cdots<T_{N}. Usually, at the first expiry date T0T_{0} , the option holder receives a contract related with dates T1,,TNT_{1},\cdots,T_{N}. Options belonging to this class are called multiple-expiry exotic options. Bermudan options or several times extendable options are good examples.
In this paper we extend Buchen’s method [2] to develop a new technique for pricing of multiple expiry exotic options in terms of a portfolio of higher order binary options. An up (or down) binary option is the option that on its expiry date delivers an agreed payoff if the price of the underlying asset is above (or bellow) a fixed exercise price and zero otherwise. Binary options whose agreed payoff is an asset are referred as first order asset binaries and binary options whose agreed payoff is cash are referred as first order bond binaries [2]. In this paper, the binary options whose agreed payoff is a (n1)(n-1)-th order binary are called nn-th order binaries by induction.
The basic idea of expressing the payoffs of complex options in terms of binary options can be seen in previous publications. Rubinstein and Reiner [7] considered the relationship of barrier options and binaries. Ingersoll [3] extended the idea by expressing complex derivatives in terms of ”event-driven” binaries. Event ε\varepsilon-driven binary option pays one unit of underlying asset if and only if the event ε\varepsilon occurs, otherwise it pays nothing.
Buchen [2] developed a theoretical framework for pricing dual-expiry options in terms of a portfolio of elementary binary options. He introduced the concepts of first and second order binary options and provided the pricing formula for them using expectation method. And then he applied them to pricing some dual expiry exotic options including compound option, chooser option, one time extendable option, one time shout option, American call option with one time dividend and partial barrier options.
The purpose of this paper is to extend the Buchen’s procedure to the case of multiple-expiry options (more than 3 expiry dates). To do that, here, at first we introduced the concept of higher order binary options and then provided the pricing formula of nn-th order binaries using PDE method. After that, we applied them to pricing of some multiple-expiry exotic options such as Bermudan options, multi time extendable options, multiple shout options and etc.
As Buchen [2] mentioned, historically, dual expiry exotics as well as multiple-expiry option prices have been derived individually by various authors. This paper has demonstrated that multiple-expiry options can be priced in a unified framework by expressing each as a portfolio of higher order binaries.
The focus of this paper is on explaining the basic pricing method in terms of static replication with higher order binaries, and so issues such as the financial motivation for trading multiple-expiry exotics, computation and simulation will not be considered here in detail and the explanation about another authors’ results on concrete multiple-expiry exotics such as Bermudan option, extendable option and shout option will refer to [6, 9, 10].

2. Higher Order Binary Options

Consider an underlying asset (for example, a stock) whose price xx satisfies Ito stochastic differential equation. Let r,qr,q and σ\sigma be respectively risk free rate, dividend rate and volatility. Then to avoid arbitrage, the price of V(x,t)V(x,t) of any derivative on the stock with expiry date TT and expiry payoff f(x)f(x) must satisfy the following (1)-(2).

Vt+σ22x22Vt2+(rq)xVxrV=0,0t<T,0<x<\frac{\partial V}{\partial t}+\frac{\sigma^{2}}{2}x_{2}\frac{\partial^{2}V}{\partial t^{2}}+(r-q)x\frac{\partial V}{\partial x}-rV=0,~0\leq t<T,~0<x<\infty (1)
V(x,T)=f(x)V(x,~T)=f(x) (2)

This price V(x,t)V(x,t) is called a standard option with expiry payoff f(x)f(x) [2].

Proposition 1[5] Assume that there exist non negative constants MM and α\alpha such that |f(x)|Mxαlnx,x>0|f(x)|\leq Mx^{\alpha\ln x},~x>0. Then the price of standard option, that is, the solution of (1) and (2) is provided as follows:

V(x,t;T)=er(Tt)01σ2π(Tt)1ze[lnxz+(rqσ22)(Tt)]22σ2(Tt)f(z)𝑑zV(x,t;T)=e^{-r(T-t)}\int_{0}^{\infty}\frac{1}{\sigma\sqrt{2\pi(T-t)}}\frac{1}{z}e^{-\frac{[\ln\frac{x}{z}+(r-q-\frac{\sigma^{2}}{2})(T-t)]^{2}}{2\sigma^{2}(T-t)}}f(z)dz
=xeq(Tt)01σ2π(Tt)1z2e[lnxz+(rq+σ22)(Tt)]22σ2(Tt)f(z)𝑑z.=xe^{-q(T-t)}\int_{0}^{\infty}\frac{1}{\sigma\sqrt{2\pi(T-t)}}\frac{1}{z^{2}}e^{-\frac{[\ln\frac{x}{z}+(r-q+\frac{\sigma^{2}}{2})(T-t)]^{2}}{2\sigma^{2}(T-t)}}f(z)dz~. (3)

Remark. It is well known that the change of variable y=lnxy=\ln x transforms (1) to a parabolic PDE with constant coefficients which can be easily transformed into a heat equation. (For example, see [4].) From the theory of heat equations we can know that the singular integral on the left side of (3) and its tt and xx derivatives always exist under the above condition on ff, which can be easily seen in references such as [8] on PDE or equations of mathematical physics, and we can easily check that (3) satisfies (1) and (2).

An up binary option of exercise price ξ\xi on the standard option with expiry payoff f(x)f(x) is a contract with expiry payoff f(x)f(x) if x>ξx>\xi and zero otherwise. A down binary option pays f(x)f(x) if x<ξx<\xi and zero otherwise. Let ss be the sign ”++” or ”-”. In what follows we use the sign ”++” and ”-” as sign indicators for up and down binaries, respectively. Then the expiry payoff functions for up and down binaries can be written in the form

Vξs(x,T)=f(x)1(sx>sξ).V_{\xi}^{s}(x,T)=f(x)\cdot 1(sx>s\xi).

From this there holds the following parity relation between the standard option price and the corresponding up and down binaries:

Vξ+(x,t)+Vξ(x,t)=V(x,t),t<T.V_{\xi}^{+}(x,t)+V_{\xi}^{-}(x,t)=V(x,t),~~t<T.

If f(x)=xf(x)=x, the standard option simply pays one unit of the asset at expiry date TT, and the price of the standard option is A(x,t;T)=xeq(Tt)A(x,t;T)=xe^{-q(T-t)} for all t<Tt<T. And the corresponding binaries are the very asset-or-nothing binaries, here their prices are denoted by Aξs(x,t;T)A_{\xi}^{s}(x,t;T). Then from the above mentioned parity relation, we have

Aξ+(x,t;T)+Aξ(x,t;T)=A(x,t;T)=xeq(Tt),t<T.A_{\xi}^{+}(x,t;T)+A_{\xi}^{-}(x,t;T)=A(x,t;T)=x\cdot e^{-q(T-t)},~~t<T.

If f(x)=1f(x)=1, the standard option simply pays one unit of cash at expiry date TT, and the price of the standard option is B(x,t;T)=er(Tt)B(x,t;T)=e^{-r(T-t)} for all t<Tt<T. And the corresponding binaries are the very cash-or-nothing binaries (or bond binaries), here their prices are denoted by Bξs(x,t;T)B_{\xi}^{s}(x,t;T). From the parity relation, we have

Bξ+(x,t;T)+Bξ(x,t;T)=B(x,t;T)=er(Tt),t<T.B_{\xi}^{+}(x,t;T)+B_{\xi}^{-}(x,t;T)=B(x,t;T)=e^{-r(T-t)},~~t<T.

The concept of Q-option plays a very useful role in the pricing of dual expiry options and, in particular, simplifying the notation of the price formula. (See [2].) Consider a standard contract that pays f(x)=s(xK)f(x)=s(x-K) at expiry date TT; this is a kind of forward contract, where the holder must buy if s=+s=+ (or sell if s=s=-) one unit of underlying asset for KK units of cash. Since f(x)=s[A(x,T;T)KB(x,T;T)]f(x)=s[A(x,T;T)-K\cdot B(x,T;T)], the price Q(x,t;T,K)Q(x,t;T,K) of this contract at time t<Tt<T is given by

Q(x,t;T,K)=s[A(x,t;T)KB(x,t;T)]=s[xeq(Tt)Ker(Tt)].Q(x,t;T,K)=s[A(x,t;T)-K\cdot B(x,t;T)]=s[x\cdot e^{-q(T-t)}-K\cdot e^{-r(T-t)}].

The binary option that pays f(x)=s(xK)1(sx>sξ)f(x)=s(x-K)\cdot 1(sx>s\xi) at expiry date TT is called a first order Q-option, their prices denoted by Qξs(x,t;T,K)Q_{\xi}^{s}(x,t;T,K). If ξ=K\xi=K, then QKs(x,t;T,K)Q_{K}^{s}(x,t;T,K) is is the very ordinary European (call if s=+s=+, or put if s=s=-) option. So Qξs(x,t;T,K)Q_{\xi}^{s}(x,t;T,K) is called a generalized European option. These options are more general in the sense that their exercise price ξ\xi is different from their strike price KK. The price of first order Q-options are given as follows:

Qξs(x,t;T,K)=s[Aξs(x,t;T)KBξs(x,t;T)].Q_{\xi}^{s}(x,t;T,K)=s[A_{\xi}^{s}(x,t;T)-K\cdot B_{\xi}^{s}(x,t;T)]. (4)

The asset or nothing binary, bond binary and the first order Q-option are called the first order binaries.[2]

Proposition 2[2, 4]. The prices of asset and bond binary options are provided as follows:

Aξs(x,t;T)=xeq(Tt)N(sd),Bξs(x,t;T)=er(Tt)N(sd).A_{\xi}^{s}(x,t;T)=x\cdot e^{-q(T-t)}N(sd),~~~~~B_{\xi}^{s}(x,t;T)=e^{-r(T-t)}N(sd^{\prime}). (5)

Here N(x)N(x) is the accumulated normal distribution function

N(x)=(2π)1xey22𝑑y,N(x)=(\sqrt{2\pi})^{-1}\int_{-\infty}^{x}e^{-\frac{y^{2}}{2}}dy,

and dd, dd^{\prime} are respectively given as follows:

d=lnxξ+(rq+σ22)(Tt)σTt,d=dσTt.d=\frac{\ln\frac{x}{\xi}+(r-q+\frac{\sigma^{2}}{2})(T-t)}{\sigma\sqrt{T-t}},~~~~d^{\prime}=d-\sigma\sqrt{T-t}.

Definition 1. An nn-th order binary option is a binary contract with expiry date T0T_{0} on an underlying (n1)(n-1)-th order binary option. Specifically, the payoff at time T0T_{0} has the following form

V(x,T0)=Fξ1ξn1s1sn1(x,T0;T1,,Tn1)1(s0x>s0ξ0).V(x,T_{0})=F_{\xi_{1}\cdots\xi_{n-1}}^{s_{1}\cdots s_{n-1}}(x,T_{0};T_{1},\cdots,T_{n-1})\cdot 1(s_{0}x>s_{0}\xi_{0}). (6)

Here Fξ1ξn1s1sn1(x,T0;T1,,Tn1)F_{\xi_{1}\cdots\xi_{n-1}}^{s_{1}\cdots s_{n-1}}(x,T_{0};T_{1},\cdots,T_{n-1}) is the price of the underlying (n1)(n-1)-th order binary option with expiry time T1,,Tn1T_{1},\cdots,T_{n-1} at the time T0T_{0} and either F=AF=A if the underlying binary is asset binary, F=BF=B for the underlying bond binary or F=QF=Q for the underlying QQ-option; and s0,,sn1s_{0},\cdots,s_{n-1} are up-down indicators (++ or -) at times T0,,Tn1T_{0},\cdots,T_{n-1} respectively. ξ0,,ξn1\xi_{0},\cdots,\xi_{n-1} are their corresponding exercise prices.
The prices of these nn-th order binary options at time t<T0t<T_{0} are denoted by Fξ0ξ1ξn1s0s1sn1(x,t;T0,T1,,Tn1)F_{\xi_{0}\xi_{1}\cdots\xi_{n-1}}^{s_{0}s_{1}\cdots s_{n-1}}(x,t;T_{0},T_{1},\cdots,T_{n-1}).
Then from the definition we have

Qξ0ξ1ξn1s0s1sn1(x,t;T0,T1,,Tn1,K)~Q_{\xi_{0}\xi_{1}\cdots\xi_{n-1}}^{s_{0}s_{1}\cdots s_{n-1}}(x,t;T_{0},T_{1},\cdots,T_{n-1},K)=
=sn1[Aξ0ξ1ξn1s0s1sn1(x,t;T0,T1,,Tn1)=s_{n-1}[A_{\xi_{0}\xi_{1}\cdots\xi_{n-1}}^{s_{0}s_{1}\cdots s_{n-1}}(x,t;T_{0},T_{1},\cdots,T_{n-1})-

KBξ0ξ1ξn1s0s1sn1(x,t;T0,T1,,Tn1)].-K\cdot B_{\xi_{0}\xi_{1}\cdots\xi_{n-1}}^{s_{0}s_{1}\cdots s_{n-1}}(x,t;T_{0},T_{1},\cdots,T_{n-1})]. (7)

Note that the strike price KK in the higher order QQ-binary is effective only at last time Tn1T_{n-1}. From the parity relation, we have

Fξ0ξ1ξn1+s1sn1(x,t;T0,T1,,Tn1)+Fξ0ξ1ξn1s1sn1(x,t;T0,T1,,Tn1)=F_{\xi_{0}\xi_{1}\cdots\xi_{n-1}}^{+s_{1}\cdots s_{n-1}}(x,t;T_{0},T_{1},\cdots,T_{n-1})+F_{\xi_{0}\xi_{1}\cdots\xi_{n-1}}^{-s_{1}\cdots s_{n-1}}(x,t;T_{0},T_{1},\cdots,T_{n-1})=
=Fξ1ξn1s1sn1(x,t;T1,,Tn1).=F_{\xi_{1}\cdots\xi_{n-1}}^{s_{1}\cdots s_{n-1}}(x,t;T_{1},\cdots,T_{n-1}).

If the pricing formulae of the higher order asset binary Aξ0ξ1ξn1s0s1sn1(x,t;A_{\xi_{0}\xi_{1}\cdots\xi_{n-1}}^{s_{0}s_{1}\cdots s_{n-1}}(x,t; T0,T1,,Tn1)T_{0},T_{1},\cdots,T_{n-1}) and the higher order bond binary Bξ0ξ1ξn1s0s1sn1(x,t;T0,B_{\xi_{0}\xi_{1}\cdots\xi_{n-1}}^{s_{0}s_{1}\cdots s_{n-1}}(x,t;T_{0}, T1,,Tn1)T_{1},\cdots,T_{n-1}) are provided, then the price of the higher order QQ-binary is easily provided by (7).
Black-Scholes expressions for higher order binary options involve multidimensional normal distribution function N(x0,x1,,xn1;P)N(x_{0},x_{1},\cdots,x_{n-1};P) with zero mean vector and correlation matrix P1P^{-1}:

N(x0,x1,,xn1;P)=x0xn11(2π)ndetPe12y𝖳Py𝑑y.N(x_{0},x_{1},\cdots,x_{n-1};P)=\int_{-\infty}^{x_{0}}\cdots\int_{-\infty}^{x_{n-1}}\frac{1}{(\sqrt{2\pi})^{n}}\sqrt{\det{P}}~e^{-\frac{1}{2}y^{\mathsf{T}}Py}dy. (8)

Here y𝖳=(y0,y1,,yn1).y^{\mathsf{T}}=(y_{0},y_{1},\cdots,y_{n-1}).
Let define the matrix A=(aij)i,j=0,1,,n1A=(a_{ij})_{i,j=0,1,\cdots,n-1} related to the expiry dates T0,T1,,T_{0},T_{1},\cdots, Tn1T_{n-1} as follows:

a00=(T1t)/(T1T0)a_{00}=(T_{1}-t)/(T_{1}-T_{0}),

an1,n1=(Tn1t)/(Tn1Tn2)a_{n-1,n-1}=(T_{n-1}-t)/(T_{n-1}-T_{n-2}),

aii=(Tit)/(TiTi1)+(Tit)/(Ti+1Ti),1in2a_{ii}=(T_{i}-t)/(T_{i}-T_{i-1})+(T_{i}-t)/(T_{i+1}-T_{i}),~~1\leq i\leq n-2,

ai,i+1=ai+1,i=(Tit)(Ti+1t)/(Ti+1Ti),0in2a_{i,i+1}=a_{i+1,i}=-\sqrt{(T_{i}-t)(T_{i+1}-t)}/(T_{i+1}-T_{i}),~~0\leq i\leq n-2,

and another elements are all zero. Then we have

A1=(rij);rij=(Tit)/(Tjt),rji=rij,ij,A^{-1}=(r_{ij});r_{ij}=\sqrt{(T_{i}-t)/(T_{j}-t)},~r_{ji}=r_{ij},i\leq j,
detA=T1tT1T0T2tT2T1Tn1tTn1Tn2.\det{A}=\frac{T_{1}-t}{T_{1}-T_{0}}\cdot\frac{T_{2}-t}{T_{2}-T_{1}}\cdot\cdots\cdot\frac{T_{n-1}-t}{T_{n-1}-T_{n-2}}.

Let si=+s_{i}=+ or - (that is, sign indicators and i=0,1,,n1i=0,1,\cdots,n-1) and define a new matrix by

As0s1sn1=(sisjaij)i,j=0n1.A_{s_{0}s_{1}\cdots s_{n-1}}=(s_{i}\cdot s_{j}\cdot a_{ij})_{i,j=0}^{n-1}.

Then we have

As0s1sn11=(sisjrij),detAs0s1sn1=detA.A_{s_{0}s_{1}\cdots s_{n-1}}^{-1}=(s_{i}s_{j}r_{ij}),\quad\det{A_{s_{0}s_{1}\cdots s_{n-1}}}=\det{A}.

Theorem 1 The price of nn-th order asset binary and bond binary are provided as follows :

Aξ0ξ1ξn1s0s1sn1(x,t;T0,T1,,Tn1)=A_{\xi_{0}\xi_{1}\cdots\xi_{n-1}}^{s_{0}s_{1}\cdots s_{n-1}}(x,t;T_{0},T_{1},\cdots,T_{n-1})=

=xeq(Tn1t)N(s0d0,s1d1,,sn1dn1;As0s1sn1),=x\cdot e^{-q(T_{n-1}-t)}N(s_{0}d_{0},s_{1}d_{1},\cdots,s_{n-1}d_{n-1};A_{s_{0}s_{1}\cdots s_{n-1}}), (9)

Bξ0ξ1ξn1s0s1sn1(x,t;T0,T1,,Tn1)=B_{\xi_{0}\xi_{1}\cdots\xi_{n-1}}^{s_{0}s_{1}\cdots s_{n-1}}(x,t;T_{0},T_{1},\cdots,T_{n-1})=

=er(Tn1t)N(s0d0,s1d1,,sn1dn1;As0s1sn1).=e^{-r(T_{n-1}-t)}N(s_{0}d^{\prime}_{0},s_{1}d^{\prime}_{1},\cdots,s_{n-1}d^{\prime}_{n-1};A_{s_{0}s_{1}\cdots s_{n-1}}). (10)

Here

  di=lnxξi+(rq+σ22)(Tit)σTit,d_{i}=\frac{\ln\frac{x}{\xi_{i}}+(r-q+\frac{\sigma^{2}}{2})(T_{i}-t)}{\sigma\sqrt{T_{i}-t}},~~~~

di=diσTit,i=0,1,,n1.d^{\prime}_{i}=d_{i}-\sigma\sqrt{T_{i}-t},~i=0,1,\cdots,n-1.

Proof: The cases of n=1n=1 and n=2n=2 were proved by Buchen in [2] using probability theory and can be easily proved using the formula (3) too. In the case of n>2n>2 we will give a sketch of the proof by induction.
We assume that theorem 1 holds in the case of n1n-1. From the definition 1, Aξ0ξ1ξn1s0s1sn1(x,t;T0,T1,,Tn1)A_{\xi_{0}\xi_{1}\cdots\xi_{n-1}}^{s_{0}s_{1}\cdots s_{n-1}}(x,t;T_{0},T_{1},\cdots,T_{n-1}) satisfies (1) and

V(x,T0)=Aξ1ξn1s1sn1(x,T0;T1,,Tn1)1(s0x>s0ξ0).V(x,T_{0})=A_{\xi_{1}\cdots\xi_{n-1}}^{s_{1}\cdots s_{n-1}}(x,T_{0};T_{1},\cdots,T_{n-1})\cdot 1(s_{0}x>s_{0}\xi_{0}).

Therefore by the formula (3), If we let

G(z)=1σ2π(T0t)1z2e[lnxz+(rq+σ22)(T0t)]22σ2(T0t),G(z)=\frac{1}{\sigma\sqrt{2\pi(T_{0}-t)}}\frac{1}{z^{2}}e^{-\frac{[\ln\frac{x}{z}+(r-q+\frac{\sigma^{2}}{2})(T_{0}-t)]^{2}}{2\sigma^{2}(T_{0}-t)}},

then Aξ0ξ1ξn1s0s1sn1(x,t;T0,T1,,Tn1)A_{\xi_{0}\xi_{1}\cdots\xi_{n-1}}^{s_{0}s_{1}\cdots s_{n-1}}(x,t;T_{0},T_{1},\cdots,T_{n-1}) is provided as follows:

Aξ0ξ1ξn1s0s1sn1(x,t;T0,T1,,Tn1)=A_{\xi_{0}\xi_{1}\cdots\xi_{n-1}}^{s_{0}s_{1}\cdots s_{n-1}}(x,t;T_{0},T_{1},\cdots,T_{n-1})=

=xeq(T0t)0G(z)Aξ1ξn1s1sn1(z,T0;T1,,Tn1)1(s0z>s0ξ0)𝑑z=x\cdot e^{-q(T_{0}-t)}\int_{0}^{\infty}G(z)A_{\xi_{1}\cdots\xi_{n-1}}^{s_{1}\cdots s_{n-1}}(z,T_{0};T_{1},\cdots,T_{n-1})\cdot 1(s_{0}z>s_{0}\xi_{0})dz.

Here Aξ1ξn1s1sn1(z,T0;T1,,Tn1)A_{\xi_{1}\cdots\xi_{n-1}}^{s_{1}\cdots s_{n-1}}(z,T_{0};T_{1},\cdots,T_{n-1}) is the price of the underlying (n1)(n-1)-th order asset binary option. By induction-assumption, the formula (9) holds for Aξ1ξn1s1sn1(z,T0;T1,,Tn1)A_{\xi_{1}\cdots\xi_{n-1}}^{s_{1}\cdots s_{n-1}}(z,T_{0};T_{1},\cdots,T_{n-1}). Thus we have

Aξ1ξn1s1sn1(z,T0;T1,,Tn1)=A_{\xi_{1}\cdots\xi_{n-1}}^{s_{1}\cdots s_{n-1}}(z,T_{0};T_{1},\cdots,T_{n-1})=

=zeq(Tn1T0)N(s1d1,,sn1dn1;As1sn1).=z\cdot e^{-q(T_{n-1}-T_{0})}N(s_{1}d_{1},\cdots,s_{n-1}d_{n-1};A_{s_{1}\cdots s_{n-1}}).

Substitute this equality into the above singular integral representation and calculate the integral, then we have (9) for the case of n>2n>2. The proof for (10) is similar. (Proof End)

The formulae (7), (9) and (10) give the following price of higher order QQ-option :

Qξ0ξ1ξn1s0s1sn1(x,t;T0,T1,,Tn1,K)=Q_{\xi_{0}\xi_{1}\cdots\xi_{n-1}}^{s_{0}s_{1}\cdots s_{n-1}}(x,t;T_{0},T_{1},\cdots,T_{n-1},K)=

=sn1[xeq(Tn1t)N(s0d0,s1d1,,sn1dn1;As0s1sn1)=s_{n-1}[x\cdot e^{-q(T_{n-1}-t)}N(s_{0}d_{0},s_{1}d_{1},\cdots,s_{n-1}d_{n-1};A_{s_{0}s_{1}\cdots s_{n-1}})-

Ke˙r(Tn1t)N(s0d0,s1d1,,sn1dn1;As0s1sn1)].-K\dot{e}^{-r(T_{n-1}-t)}N(s_{0}d^{\prime}_{0},s_{1}d^{\prime}_{1},\cdots,s_{n-1}d^{\prime}_{n-1};A_{s_{0}s_{1}\cdots s_{n-1}})]. (11)

3. Applications to Multiple-Expiry Exotics

In this section, we applied the results of previous section to the pricing of some multiple expiry exotics.

The Static Replication Theorem If the payoff of an option at expiry time T0T_{0} is a linear combination of prices at time T0T_{0} of higher order binaries, then its price at all time t<T0t<T_{0} is the combination of the corresponding prices at time tt of the higher order binaries.

The static replication theorem can be proved from the uniqueness of solution to the initial value problem of Black-Scholes equation. And we need one lemma that guarantees the monotonousness of the option price on the underlying asset price and provides some estimates about the gradient of the price on the underlying asset price.

Lemma 1 Assume that f(x)f(x) is continuous and piecewise differentiable. If V(x,t)V(x,t) is the solution of (1) and (2), then we have

minxfxeq(Tt)Vx(x,t)maxxfxeq(Tt).\min_{x}\frac{\partial f}{\partial x}\cdot e^{-q(T-t)}\leq\frac{\partial V}{\partial x}(x,t)\leq\max_{x}\frac{\partial f}{\partial x}\cdot e^{-q(T-t)}.

In particular, if E={z:fx(z)<maxxfx},F={z:fx(z)>minxfx},|E|>0E=\{z:\frac{\partial f}{\partial x}(z)<\max_{x}\frac{\partial f}{\partial x}\},~F=\{z:\frac{\partial f}{\partial x}(z)>\min_{x}\frac{\partial f}{\partial x}\},|E|>0 and |F|>0|F|>0, then

minxfxeq(Tt)<Vx(x,t)<maxxfxeq(Tt),t<T.\min_{x}\frac{\partial f}{\partial x}\cdot e^{-q(T-t)}<\frac{\partial V}{\partial x}(x,t)<\max_{x}\frac{\partial f}{\partial x}\cdot e^{-q(T-t)},~t<T.

Using the formula (3) and the assumption of lemma 1, we can easily prove the required results.

3.1. Bermudan Options

The Bermudan option is one type of nonstandard American options and early exercise is restricted to certain dates during the life of the option. For example, let t1<t2<<tn=Tt_{1}<t_{2}<\cdots<t_{n}=T be the dates of early exercise for put option with strike price KK. Let t0=0t_{0}=0. Let denote the option price on the interval (ti1,ti](t_{i-1},t_{i}] by Vi1(x,t)V_{i-1}(x,t), where i=1,,ni=1,\cdots,n. Then [4]

Vi1(x,ti)=max(Vi(x,ti),(Kx)+),i=1,,n1.V_{i-1}(x,t_{i})=\max(V_{i}(x,t_{i}),(K-x)^{+}),~~i=1,\cdots,n-1. (12)

(See figure 1.)

Refer to caption
Figure 1.

For fixed i=1,,n1i=1,\cdots,n-1, we consider the following equation:

Vi(x,ti)=(Kx)+.V_{i}(x,t_{i})=(K-x)^{+}. (13)

If Vi(x,ti)V_{i}(x,t_{i}) is monotonously decreasing on xx, 0<Vi(x,ti)<K0<V_{i}(x,t_{i})<K and 1<Vi(x,ti)x-1<\frac{\partial V_{i}(x,t_{i})}{\partial x}, then the equation (13) has a unique root aia_{i} such that 0<ai<K0<a_{i}<K. (See figure 2.)

Refer to caption
Figure 2.

Lemma 2 Fix i{1,,n1}i\in\{1,\cdots,n-1\}. Assume that Vi(x,ti)V_{i}(x,t_{i}) is monotonously decreasing on xx, 0<Vi(x,ti)<K0<V_{i}(x,t_{i})<K and 1<Vi(x,ti)x-1<\frac{\partial V_{i}(x,t_{i})}{\partial x}. Let aia_{i} be the unique root of the equation (13). Then for any t[ti1,ti)t\in[t_{i-1},t_{i}), we have

Vi1(x,t)=(Vi)ai+(x,t;ti)+Qai(x,t;ti,K)V_{i-1}(x,t)=(V_{i})_{a_{i}}^{+}(x,t;t_{i})+Q_{a_{i}}^{-}(x,t;t_{i},K)

and 1Vi1x0-1\leq\frac{\partial V_{i-1}}{\partial x}\leq 0. In particular, if ti1t<tit_{i-1}\leq t<t_{i}, then 1<Vi1(x,t)x<0-1<\frac{\partial V_{i-1}(x,t)}{\partial x}<0. Here (Vi)ai+(x,t;ti)(V_{i})_{a_{i}}^{+}(x,t;t_{i}) is the solution of Black-Scholes equation (1) satisfying the condition V(x,ti)=Vi(x,ti)1(x>ai)V(x,t_{i})=V_{i}(x,t_{i})\cdot 1(x>a_{i}).
Proof. From the assumption, we can rewrite (12) as follows. (See figure 2.)

Vi1(x,ti)=Vi(x,ti)1(x>ai)+(Kx)1(x<ai).V_{i-1}(x,t_{i})=V_{i}(x,t_{i})\cdot 1(x>a_{i})+(K-x)\cdot 1(x<a_{i}).

Thus from the definition of the binary option, we have

Vi1(x,t)=(Vi)ai+(x,t;ti)+Qai(x,t;ti,K),t(ti1,ti].V_{i-1}(x,t)=(V_{i})_{a_{i}}^{+}(x,t;t_{i})+Q_{a_{i}}^{-}(x,t;t_{i},K),~t\in(t_{i-1},t_{i}].

From the assumption we have 0<ai<K0<a_{i}<K and thus using lemma 1, we have the conclusion on Vi1x\frac{\partial V_{i-1}}{\partial x}. (Proof End)

Using this lemma, we can easily calculate the price of Bermudan put options. Since the payoff at time tnt_{n} is (Kx)+(K-x)^{+} and we can’t early exercise after the time tn1t_{n-1}, so the option on the interval (tn1,tn](t_{n-1},t_{n}] becomes an ordinary European put option, and thus its price Vn1(x,t)V_{n-1}(x,t) in the interval [tn1,tn)[t_{n-1},t_{n}) is given by

Vn1(x,t)=QK(x,t;tn,K),t[tn1,tn)V_{n-1}(x,t)=Q_{K}^{-}(x,t;t_{n},K),~t\in[t_{n-1},t_{n})

and Vn1(x,tn1)V_{n-1}(x,t_{n-1}) satisfies 0<Vn1(x,tn1)<K0<V_{n-1}(x,t_{n-1})<K and 1<Vn1x(x,tn1)-1<\frac{\partial V_{n-1}}{\partial x}(x,t_{n-1}) <0<0. By the static replication theorem and lemma 2, our option price at time t[tn2,tn1)t\in[t_{n-2},t_{n-1}) is given by

Vn2(x,t)=Qan1K+(x,t;tn1,tn,K)+Qan1(x,t;tn1,K),t[tn2,tn1).V_{n-2}(x,t)=Q_{a_{n-1}K}^{~~~~+~~~~{~~~~-}}(x,t;t_{n-1},t_{n},K)+Q_{a_{n-1}}^{~~~-}(x,t;t_{n-1},K),~t\in[t_{n-2},t_{n-1}).

and 1<Vn2x<0-1<\frac{\partial V_{n-2}}{\partial x}<0. Repeating the similar considerations, we can get the following formulae:

Vi1(x,t)=Qaian1K++(x,t;ti,,tn,K)+V_{i-1}(x,t)=Q_{a_{i}\cdots a_{n-1}K}^{+\cdots{~+}~~~~~~{~~~~~~-}}(x,t;t_{i},\cdots,t_{n},K)+

+Qaian2an1++(x,t;ti,,tn1,K)++Qaiai+1+(x,t;ti,ti+1,K)++Q_{a_{i}\cdots a_{n-2}a_{n-1}}^{+\cdots{~+}~~~~~~{~~~~~~~-}}(x,t;t_{i},\cdots,t_{n-1},K)+\cdots+Q_{a_{i}a_{i+1}}^{+~~~~{~~~~-}}(x,t;t_{i},t_{i+1},K)+

+Qai(x,t;ti,K),t[ti1,ti),i{1,,n}.+Q_{a_{i}}^{-}(x,t;t_{i},K),~t\in[t_{i-1},t_{i}),i\in\{1,\cdots,n\}. (14)

In particular, V0V_{0} gives the price of Bermudan put option at time t=0t=0.

V0(x,t)=Qa1an1K++(x,t;t1,,tn,K)+V_{0}(x,t)=Q_{a_{1}\cdots a_{n-1}K}^{+\cdots{~+}~~~~~~{~~~~~~-}}(x,t;t_{1},\cdots,t_{n},K)+

+Qa1an2an1++(x,t;t1,,tn1,K)++Qa1a2+(x,t;t1,t2,K)++Q_{a_{1}\cdots a_{n-2}a_{n-1}}^{+\cdots{~+}~~~~~~{~~~~~~~-}}(x,t;t_{1},\cdots,t_{n-1},K)+\cdots+Q_{a_{1}a_{2}}^{+~~{~~-}}(x,t;t_{1},t_{2},K)+

+Qa1(x,t;t1,K),t[0,t1).+Q_{a_{1}}^{-}(x,t;t_{1},K),{~}~{~}~{~}~t\in[0,t_{1}).

3.2. Multiple Extendable Options

Extendable options were first studied and analyzed in detail by Longstaff [6]. Here we consider the holder extendable options in his sense. The holder nn-times extendable call (or put) option has the right at some certain dates TiT_{i} to exercise a standard European call (or put) option of strike price KiK_{i} or to extend the expiry date to time Ti+1(>Ti)T_{i+1}(>T_{i}) and change the strike price from KiK_{i} to Ki+1(i=0,1,,n1)K_{i+1}(i=0,1,\cdots,n-1) for a premium CiC_{i}. Here we consider nn-times extendable call option.
Let denote the option price on the interval [Ti1,Ti)[T_{i-1},T_{i}) by Vi(x,t)V_{i}(x,t), where i=0,1,,ni=0,1,\cdots,n and T1=0T_{-1}=0. Then from the definition of extendable option, we have

Vi(x,Ti)=max(Vi+1(x,Ti)Ci,(xKi)+),i=0,,n1.V_{i}(x,T_{i})=\max(V_{i+1}(x,T_{i})-C_{i},(x-K_{i})^{+}),~~i=0,\cdots,n-1. (15)

First, let consider the price Vn(x,t)V_{n}(x,t) in the last interval (Tn1,Tn](T_{n-1},T_{n}]. In this interval we can no longer extend our option contract and thus our option is just an ordinary European call with the expiry date TnT_{n} and strike price KnK_{n}. Therefore

Vn(x,t)=QKn+(x,t;Tn,K),t[Tn1,Tn).V_{n}(x,t)=Q_{K_{n}}^{+}(x,t;T_{n},K),t\in[T_{n-1},T_{n}).

And it satisfies 0<Vn(x,Tn1)x<10<\frac{\partial V_{n}(x,T_{n-1})}{\partial x}<1 by lemma 1.
Now consider the price Vn1(x,t)V_{n-1}(x,t) in the interval (Tn2,Tn1](T_{n-2},T_{n-1}]. Then by (15) we have

Vn1(x,Tn1)=max(QKn+(x,Tn1;Tn,Kn)Cn1,(xKn1)+).V_{n-1}(x,T_{n-1})=\max(Q_{K_{n}}^{+}(x,T_{n-1};T_{n},K_{n})-C_{n-1},(x-K_{n-1})^{+}).

From the property of QKn+(x,Tn1;Tn,Kn)Q_{K_{n}}^{+}(x,T_{n-1};T_{n},K_{n}) , the following two equations on xx have unique roots an1a_{n-1} and bn1b_{n-1}, respectively.

QKn+(x,Tn1;Tn,Kn)Cn1=0,{~}~Q_{K_{n}}^{+}(x,T_{n-1};T_{n},K_{n})-C_{n-1}=0,

QKn+(x,Tn1;Tn,Kn)Cn1=(xKn1)+.Q_{K_{n}}^{+}(x,T_{n-1};T_{n},K_{n})-C_{n-1}=(x-K_{n-1})^{+}. (16)

And it is quite natural that we assume that an1<Kn1<bn1a_{n-1}<K_{n-1}<b_{n-1}. (Otherwise the extension of Tn1T_{n-1} to TnT_{n} would not occur as Buchen [2] mentioned. See figure 3. Similarly, in put we assume that an1>Kn1>bn1a_{n-1}>K_{n-1}>b_{n-1}.)

Refer to caption
Figure 3.

Then we can rewrite Vn1(x,Tn1)V_{n-1}(x,T_{n-1}) as follows:

Vn1(x,Tn1)=[QKn+(x,Tn1;Tn,Kn)Cn1]1(an1<x<bn1)V_{n-1}(x,T_{n-1})=[Q_{K_{n}}^{+}(x,T_{n-1};T_{n},K_{n})-C_{n-1}]\cdot 1(a_{n-1}<x<b_{n-1})

+(xKn1)1(x>bn1)=+(x-K_{n-1})\cdot 1(x>b_{n-1})=

=QKn+(x,Tn1;Tn,Kn)1(x>an1)=Q_{K_{n}}^{+}(x,T_{n-1};T_{n},K_{n})\cdot 1(x>a_{n-1})-

QKn+(x,Tn1;Tn,Kn)1(x>bn1)-Q_{K_{n}}^{+}(x,T_{n-1};T_{n},K_{n})\cdot 1(x>b_{n-1})-

Cn11(x>an1)+(xKn1+Cn1)1(x>bn1)-C_{n-1}\cdot 1(x>a_{n-1})+(x-K_{n-1}+C_{n-1})\cdot 1(x>b_{n-1})

Thus using the Static Replication Theorem, for t[Tn2,Tn1)t\in[T_{n-2},T_{n-1}) we have

Vn1(x,t)=[Qan1Kn++(x,t;Tn1,Tn,Kn)Qbn1Kn++(x,t;Tn1,Tn,Kn)V_{n-1}(x,t)=[Q_{a_{n-1}K_{n}}^{~~{+}~~~{~~~~+}}(x,t;T_{n-1},T_{n},K_{n})-Q_{b_{n-1}K_{n}}^{~~+~~{~~~~+}}(x,t;T_{n-1},T_{n},K_{n})

Cn1Ban1+(x,t;Tn1)+Qbn1+(x,t;Tn1,Kn1Cn1)-C_{n-1}\cdot B_{a_{n-1}}^{~~{+}}(x,t;T_{n-1})+Q_{b_{n-1}}^{~~{+}}(x,t;T_{n-1},K_{n-1}-C_{n-1}).

And Vn1(x,Tn1)V_{n-1}(x,T_{n-1}) is a monotone increasing function of xx with a positive inclination less than 1 in x(an1,bn1)x\in(a_{n-1},b_{n-1}). (See figure 3.) By lemma 1, we have

0<Vn1(x,t)x<1,t[Tn2,Tn1).0<\frac{\partial V_{n-1}(x,t)}{\partial x}<1,~t\in[T_{n-2},T_{n-1}).

By similar consideration and induction, we can prove that the formula of price Vi(x,t)V_{i}(x,t) of our option in the time interval [Ti1,Ti)[T_{i-1},T_{i}) is provided by

(jijn1){ai,bi}××{an1,bn1}s(ji)s(jn1)Qjijn1Kn+++(x,t;Ti,,Tn,Kn)+\sum_{(j_{i}\cdots j_{n-1})\in\{a_{i},b_{i}\}\times\cdots\times\{a_{n-1},b_{n-1}\}}s(j_{i})\cdots s(j_{n-1})\cdot Q_{j_{i}\cdots j_{n-1}K_{n}}^{{+}\cdots{~~+}~~{~~+}}(x,t;T_{i},\cdots,T_{n},K_{n})+
k=in1(jijk1){ai,bi}××{ak1,bk1}s(ji)s(jk1)Qjijk1bk+++(x,t;Ti,,Tk,KkCk)\sum_{k=i}^{n-1}\sum_{(j_{i}\cdots j_{k-1})\in\{a_{i},b_{i}\}\times\cdots\times\{a_{k-1},b_{k-1}\}}s(j_{i})\cdots s(j_{k-1})\cdot Q_{j_{i}\cdots j_{k-1}b_{k}}^{{+}\cdots{~~+}~~{~~+}}(x,t;T_{i},\cdots,T_{k},K_{k}-C_{k})
+k=in1Ck(jijk1){ai,bi}××{ak1,bk1}s(ji)s(jk1)Bjijk1ak+++(x,t;Ti,,Tk).+\sum_{k=i}^{n-1}C_{k}\sum_{(j_{i}\cdots j_{k-1})\in\{a_{i},b_{i}\}\times\cdots\times\{a_{k-1},b_{k-1}\}}s(j_{i})\cdots s(j_{k-1})\cdot B_{j_{i}\cdots j_{k-1}a_{k}}^{{+}\cdots{~~+}~~{~~+}}(x,t;T_{i},\cdots,T_{k}). (17)

Here s(j)s(j) is the aa or bb indicator, that is,

s(jm)={1,jm=am,1,jm=bm.s(j_{m})=\begin{cases}1,&j_{m}=a_{m},\\ -1,&j_{m}=b_{m}.\end{cases}

In particular V0V_{0} gives the price of n-times extendable call option in the time interval [0,T0)[0,T_{0}).

3.3. Multiple Shout Options

Shout options (Thomas [10]) are exotic options that allow the holder to lock in a payoff at times prior to the final expiry date. If the shout times would be selected randomly according to the holder’s mind, the pricing of shout option will be quite challenging, but if the shout times are pre-determined, then it would be simpler ([2]). In a fixed time multiple shout call option with expiry date TT and strike price KK, their payoff can be locked in at some predetermined times T0<<Tn1(<T)T_{0}<\cdots<T_{n-1}(<T). So at the final expiry date T=TnT=T_{n}, its payoff is given as follows:

V(x,T)=max(x0K,,xn1K,xK,0),V(x,T)=\max(x_{0}-K,\cdots,x_{n-1}-K,x-K,0), (18)

where xi(i=0,,n1)x_{i}~(i=0,\cdots,n-1) are the underlying asset price x(Ti)x(T_{i}) at the shout time TiT_{i}.
For simplicity, here we consider the case n=2n=2. Then the expiry payoff (18) is given by

V(x,T2)=max(x0K,x1K,xK,0),V(x,T_{2})=\max(x_{0}-K,x_{1}-K,x-K,0), (19)

Note that in the last interval (T1,T2](T_{1},T_{2}] both of the underlying asset prices x0=x(T0)x_{0}=x(T_{0}) and x1=x(T1)x_{1}=x(T_{1}) are known constants and in the interval (T0,T1](T_{0},T_{1}] the underlying asset price x0=x(T0)x_{0}=x(T_{0}) is a known constant.
The case of x0<Kx_{0}<K. We can rewrite (19) as

V(x,T2)=max(x1K,xK,0)V(x,T_{2})=\max(x_{1}-K,x-K,0)

and this is the terminal payoff of one-shout option in the time interval (T0,T2](T_{0},T_{2}] and thus by the method of [2], for T0t<T1T_{0}\leq t<T_{1} we have

V(x,t)=QKK+(x,t;T1,T2,K)+er(T2T1)QK+(x,t;T1,K)+V(x,t)=Q_{KK}^{{-}+}(x,t;T_{1},T_{2},K)+e^{-r(T_{2}-T_{1})}Q_{K}^{+}(x,t;T_{1},K)+
+g(T1,T2)AK+(x,t;T1),T0t<T1.+g(T_{1},T_{2})A_{K}^{+}(x,t;T_{1}),~{~}~{~}T_{0}\leq t<T_{1}.

Here

g(T1,T2)=eq(T2T1)N(d+(T1,T2))er(T2T1)N(d(T1,T2)),g(T_{1},T_{2})=e^{-q(T_{2}-T_{1})}N(d^{+}(T_{1},T_{2}))-e^{-r(T_{2}-T_{1})}N(d^{-}(T_{1},T_{2})),
d±(T1,T2)=[(rq)/σ±σ/2]T2T1.d^{\pm}(T_{1},T_{2})=[(r-q)/\sigma\pm\sigma/2]\sqrt{T_{2}-T_{1}}. (20)

In particular x0=xx_{0}=x at the time t=T0t=T_{0}, we have

V(x,T0)=QKK+(x,T0;T1,T2,K)+er(T2T1)QK+(x,T0;T1,K)+V(x,T_{0})=Q_{KK}^{{-}+}(x,T_{0};T_{1},T_{2},K)+e^{-r(T_{2}-T_{1})}Q_{K}^{+}(x,T_{0};T_{1},K)+
+g(T1,T2)AK+(x,T0;T1),x<K.+g(T_{1},T_{2})A_{K}^{+}(x,T_{0};T_{1}),~~~~~~~x<K. (21)

The case of x0>Kx_{0}>K. If x0x1x_{0}\leq x_{1}, then we can rewrite (19) as follows:

V(x,T2)=max(x1K,xK,0)=(x1K)+(xx1)+.V(x,T_{2})=\max(x_{1}-K,x-K,0)=(x_{1}-K)+(x-x_{1})^{+}.

Thus in the time interval [T1,T2)[T_{1},T_{2}) we have

V(x,t)=(x1K)er(T2t)+Qx1+(x,t;T2,x1).V(x,t)=(x_{1}-K)\cdot e^{-r(T_{2}-t)}+Q_{x_{1}}^{+}(x,t;T_{2},x_{1}).

In particular, x1=xx_{1}=x at the time t=T1t=T_{1} and

V(x,T1)=(xK)er(T2T1)+Qx+(x,T1;T2,x),xx0.V(x,T_{1})=(x-K)\cdot e^{-r(T_{2}-T_{1})}+Q_{x}^{+}(x,T_{1};T_{2},x),~~x\geq x_{0}.

Here by (4) and (5) we have

Qx+(x,T1;T2,x)=xg(T1,T2),Q_{x}^{+}(x,T_{1};T_{2},x)=x\cdot g(T_{1},T_{2}),

where g(T1,T2)g(T_{1},T_{2}) is as in (20). Thus

V(x,T1)=x[er(T2T1)+g(T1,T2)]Ker(T2T1),xx0.V(x,T_{1})=x\cdot[e^{-r(T_{2}-T_{1})}+g(T_{1},T_{2})]-K\cdot e^{-r(T_{2}-T_{1})},~~x\geq x_{0}. (22)

If x0>x1x_{0}>x_{1}, then we can rewrite (19) as follows:

V(x,T2)=max(x0K,xK,0)=(x0K)+(xx0)+.V(x,T_{2})=\max(x_{0}-K,x-K,0)=(x_{0}-K)+(x-x_{0})^{+}.

Thus in the time interval [T1,T2)[T_{1},T_{2}) we have

V(x,t)=(x0K)er(T2t)+Qx0+(x,t;T2,x0).V(x,t)=(x_{0}-K)\cdot e^{-r(T_{2}-t)}+Q_{x_{0}}^{+}(x,t;T_{2},x_{0}).

In particular, at the time t=T1t=T_{1}

V(x,T1)=(x0K)er(T2T1)+Qx0+(x,T1;T2,x0),x<x0.V(x,T_{1})=(x_{0}-K)\cdot e^{-r(T_{2}-T_{1})}+Q_{x_{0}}^{+}(x,T_{1};T_{2},x_{0}),~~x<x_{0}. (23)

Combining (22) with (23) to get

V(x,T1)=[er(T2T1)+g(T1,T2)]x1(x>x0)er(T2T1)K1(x>x0)V(x,T_{1})=[e^{-r(T_{2}-T_{1})}+g(T_{1},T_{2})]\cdot x\cdot 1(x>x_{0})-e^{-r(T_{2}-T_{1})}\cdot K\cdot 1(x>x_{0})
+(x0K)er(T2T1)1(x<x0)+Qx0+(x,T1;T2,x0)1(x<x0).+(x_{0}-K)\cdot e^{-r(T_{2}-T_{1})}\cdot 1(x<x_{0})+Q_{x_{0}}^{+}(x,T_{1};T_{2},x_{0})\cdot 1(x<x_{0}).

Note that in the interval (T0,T1](T_{0},T_{1}] the underlying asset price x0=x(T0)x_{0}=x(T_{0}) is a known constant. Then using the Static Replication Theorem, for t[T0,T1)t\in[T_{0},T_{1}), we have

V(x,t)=[er(T2T1)+g(T1,T2)]Ax0+(x,t;T1)Ker(T2T1)Bx0+(x,t;T1)+V(x,t)=[e^{-r(T_{2}-T_{1})}+g(T_{1},T_{2})]\cdot A_{x_{0}}^{+}(x,t;T_{1})-K\cdot e^{-r(T_{2}-T_{1})}\cdot B_{x_{0}}^{+}(x,t;T_{1})+

+(x0K)er(T2T1)Bx0(x,t;T1)+Qx0x0+(x,t;T1,T2,x0)+(x_{0}-K)\cdot e^{-r(T_{2}-T_{1})}\cdot B_{x_{0}}^{-}(x,t;T_{1})+Q_{x_{0}x_{0}}^{{-}+}(x,t;T_{1},T_{2},x_{0})

=[er(T2T1)+g(T1,T2)]Ax0+(x,t;T1)Ker(T2t)+=[e^{-r(T_{2}-T_{1})}+g(T_{1},T_{2})]\cdot A_{x_{0}}^{+}(x,t;T_{1})-K\cdot e^{-r(T_{2}-t)}+

+x0er(T2T1)Bx0(x,t;T1)+Qx0x0+(x,t;T1,T2,x0)+x_{0}\cdot e^{-r(T_{2}-T_{1})}\cdot B_{x_{0}}^{-}(x,t;T_{1})+Q_{x_{0}x_{0}}^{{-}+}(x,t;T_{1},T_{2},x_{0}).

In particular, x0=xx_{0}=x at the time t=T0t=T_{0} and

V(x,T0)=[er(T2T1)+g(T1,T2)]Ax+(x,T0;T1)Ker(T2T0)V(x,T_{0})=[e^{-r(T_{2}-T_{1})}+g(T_{1},T_{2})]\cdot A_{x}^{+}(x,T_{0};T_{1})-K\cdot e^{-r(T_{2}-T_{0})}
+xer(T2T1)Bx(x,T0;T1)+Qxx+(x,T0;T1,T2,x).+x\cdot e^{-r(T_{2}-T_{1})}\cdot B_{x}^{-}(x,T_{0};T_{1})+Q_{x~x}^{-+}(x,T_{0};T_{1},T_{2},x).

Here by (4) and (5), we have

Ax+(x,T0;T1)=xeq(T1T0)N(d+(T0,T1)),A_{x}^{+}(x,T_{0};T_{1})=xe^{-q(T_{1}-T_{0})}N(d^{+}(T_{0},T_{1})),
Bx0(x,T0;T1)=er(T1T0)N(d(T0,T1)),B_{x_{0}}^{-}(x,T_{0};T_{1})=e^{-r(T_{1}-T_{0})}N(d^{-}(T_{0},T_{1})),
Qxx+(x,T0;T1,T2,x)=g1(T0,T1,T2)x,Q_{x~x}^{-+}(x,T_{0};T_{1},T_{2},x)=g_{1}(T_{0},T_{1},T_{2})\cdot x,

where d±(T0,T1)d^{\pm}(T_{0},T_{1}) is as in (20) and

g1(T0,T1,T2)=eq(T2T0)N2(d+(T0,T1),d+(T0,T2);A+)g_{1}(T_{0},T_{1},T_{2})=e^{-q(T_{2}-T_{0})}\cdot N_{2}(-d^{+}(T_{0},T_{1}),d^{+}(T_{0},T_{2});A_{-+})-
er(T2T0)N2(d(T0,T1),d(T0,T2);A+).-e^{-r(T_{2}-T_{0})}\cdot N_{2}(-d^{-}(T_{0},T_{1}),d^{-}(T_{0},T_{2});A_{-+}). (24)

Thus we have

V(x,T0)=G(T0,T1,T2)xKer(T2T0),x>K.V(x,T_{0})=G(T_{0},T_{1},T_{2})\cdot x-K\cdot e^{-r(T_{2}-T_{0})},~~~x>K. (25)

Here

G(T0,T1,T2)=[er(T2T1)+g(T1,T2)]eq(T1T0)N(d+(T0,T1))+G(T_{0},T_{1},T_{2})=[e^{-r(T_{2}-T_{1})}+g(T_{1},T_{2})]\cdot e^{-q(T_{1}-T_{0})}\cdot N(d^{+}(T_{0},T_{1}))+
+er(T2T0)N(d(T0,T1))+g1(T0,T1,T2).+e^{-r(T_{2}-T_{0})}\cdot N(-d^{-}(T_{0},T_{1}))+g_{1}(T_{0},T_{1},T_{2}). (26)

Putting the two expressions (21) and (25) of V(x,T0)V(x,T_{0}) together, we have

V(x,T0)=QKK+(x,T0;T1,T2,K)1(x<K)+V(x,T_{0})=Q_{KK}^{{-}+}(x,T_{0};T_{1},T_{2},K)1(x<K)+

+er(T2T1)QK+(x,T0;T1,K)1(x<K)++e^{-r(T_{2}-T_{1})}Q_{K}^{+}(x,T_{0};T_{1},K)1(x<K)+

+g(T1,T2)AK+(x,T0;T1)1(x<K)+G(T0,T1,T2)x1(x>K)+g(T_{1},T_{2})A_{K}^{+}(x,T_{0};T_{1})1(x<K)+G(T_{0},T_{1},T_{2})x1(x>K)

Ker(T2T0)1(x>K)-Ke^{-r(T_{2}-T_{0})}1(x>K).

Thus the price of fixed time twice shout call option at t<T0t<T_{0} is given by

V(x,t)=QKKK+(x,t;T0,T1,T2,K)+er(T2T1)QKK+(x,t;T0,T1,K)V(x,t)=Q_{KKK}^{{-}{-}+}(x,t;T_{0},T_{1},T_{2},K)+e^{-r(T_{2}-T_{1})}Q_{KK}^{{-}+}(x,t;T_{0},T_{1},K)

+g(T1,T2)AKK+(x,t;T0,T1)+G(T0,T1,T2)AK+(x,t;T0)+g(T_{1},T_{2})A_{KK}^{{-}+}(x,t;T_{0},T_{1})+G(T_{0},T_{1},T_{2})A_{K}^{+}(x,t;T_{0})

Ker(T2T0)BK+(x,t;T0),t<T0.-Ke^{-r(T_{2}-T_{0})}B_{K}^{+}(x,t;T_{0}),~~~~~~t<T_{0}. (27)

4. Conclusions

In this paper we introduced the concept of higher order binary options and then provide the pricing formulae of nn-th order binaries using solving method of PDE. Then we applied them to pricing of some multiple-expiry exotic options such as Bermudan options, multi time extendable options, fixed time twice shout options and etc. Here when calculating the price of concrete multiple-expiry exotic options, the focus of discussion was on explaining how to express the expiry payoffs of the exotic option as a combination of the payoffs of some class of higher order binary options. Here we assumed that risk free rate, dividend rate and volatility are constant but we could easily extend to the case with time dependent coefficients.

References

  • [1] Broadie, M. and Yamamoto Y., Application of the Fast Gauss Transform to option pricing, Management science, Vol. 49, No. 8, 1071-1088, 2003
  • [2] Buchen, P., The Pricing of dual-expiry exotics, Quantitative Finance, 4, 101-108, 2004
  • [3] Ingersoll, J. E., Digital contract: simple tools for pricing complex derivatives, J. Business, 73, 67-88, 2000
  • [4] Jiang, Li-shang, Mathematical Modeling and Methods of Option Pricing, World Scientific, Singapore, 2005
  • [5] Kwok, Y.K., Mathematical models of Financial Derivatives, Springer-verlag, Berlin, 55-82, 1999
  • [6] Longstaff, F., Pricing options with extendable maturities: analysis and applications, J. Finance, 45, 935-957, 2000.
  • [7] Rubinstein, M. and Reiner, E., Unscrambling the binary code, Risk Mag. 4, 75-83, 1991
  • [8] Rubinstein, I. and L. Rubinstein, Partial Differential Equations in Classical Mathematical Physics, Cambridge Univ. Press, 296-324, 1998
  • [9] Schweizer, M., On Bermudan Options, Advances in Finance and Stochastics, in Essays in Honor of Dieter Sondermann, Springer, 257-269, 2002
  • [10] Thomas, B., Something to shout about, Risk Mag. 6, 56-58, 1994