BSDEs and Feynman-Kac Formula for L evy Processes with Applications in Finance

David Nualart Universitat de Barcelona Gran Via de les Corts Catalanes, 585 E-08007 Barcelona Spain Email: nualart@mat.ub.es Wim Schoutens K.U.Leuven Eurandom Celestijnenlaan 200 B B-3001 Leuven Belgium Wim.Schoutens@wis.kuleuven.ac.be June 15, 2001

Email:

Running head: BSDEs for L evy Processes

Abstract: In this paper we show the existence and uniqueness of a solution for backward stochastic dierential equations driven by a L evy process with moments of all orders. The results are important both from a mathematical point of view as in nance: An application to Clark-Ocone and Feynman-Kac formulas for L evy processes is presented. Moreover, the Feynman-Kac formula and the related Partial Dierential Integral Equation (PDIE) provide us an analogue of the famous Black-Scholes partial dierential equation and is used for the purpose of option pricing in a L evy market.

AMS Subject Classification: 60J30, 60H05 Keywords: Backward Stochastic Dierential Equations, BSDE, L evy Processes, Orthogonal Polynomials, Option Pricing

Introduction

The rst paper concerned with Backward Stochastic Dierential Equations (BSDEs) is the paper Bismut (1973), where he introduced a non-linear Ricatti BSDE and showed existence and uniqueness of bounded solutions. Pardoux and Peng (1990) considered general BSDEs and this paper was the starting point for the development of the study of these equations. On the other hand, BSDEs have important applications in the theory of mathematical nance, especially, they play a major role in hedging and non-linear pricing theory for imperfect markets (see El Karoui and Quenez (1997)). One can consider a BSDE driven by a Brownian motion as a nonlinear generalization of the integral representation theorem for square integrable martingales. Then it is natural to extend these kind of equations to the case of L evy processes, that is, processes with independent and stationary increments. We recall that a L evy process consists of three stochastically independent parts: a purely deterministic linear part, a Brownian motion and a pure-jump process. In Situ (1997) BSDEs driven by a Brownian motion and a Poisson point process are studied. Ouknine (1998) considers BSDEs driven by a Poisson random measure. In both papers the main ingredient is the integral representation of square integrable random variables in terms of a Poisson random measure (see Jacod (1979)). In Nualart and Schoutens (2000) a martingale representation theorem for L evy processes satisfying some exponential moment condition was proved. The purpose of this paper is to use this martingale representation result to establish the existence and uniqueness of solutions for BSDEs driven by a L evy process of the kind considered in Nualart and Schoutens (2000). The results are important both from a mathematical point of view as in nance. This is illustratated in the applications. The resulting Clark-Ocone and the Feynman-Kac formulas are fundamental ingredients in the build up of an Malliavin calculus for L evy process. Moreover, the Feynman-Kac formula and the related Partial Dierential Integral Equation (PDIE) also have an important application in nance: they provide us an analogue of the famous Black-Scholes partial dierential equation and is used for the purpose of option pricing in a L evy market. The paper is organized as follows. Section 2 contains some preliminaries on L evy processes. Section 3 contains the main result on BSDEs driven by L evy processes. In Section 4 we have included some applications of BSDEs driven by L evy processes to the Clark-Ocone, the Feynman-Kac formulas, and option pricing in a L evy market. Finally, in the appendix one can nd detailed proofs of the main results. 3

Preliminaries

Let X = {Xt , t 0} be a L evy process dened on a complete probability space (, F , P ). That is, X is a real-valued process starting from 0 with stationary and independent increments and with c` adl` ag trajectories. It is known that Xt has a characteristic function of the form 1 E eiXt = exp iat 2 2 t + t 2

eix 1 ix1{|x|<1} (dx) ,

where a R, > 0, and is a measure on R with R (1 x2 ) (dx) < . We will assume that the L evy measure satises for some > 0 e|x| (dx) < ,

( , )c

for every > 0. This implies that the random variables Xt have moments of all orders. Moreover, it will assure us the existence of the below mentioned Predictable Representation, which we will use in our proofs. We refer to Sato (2000) or Bertoin (1996) for a detailed account on L evy processes. For t 0, let Ft denote the -algebra generated by the family of random variables {Xs , 0 s t} augmented with the P -null sets of F . Fix a 2 time interval [0, T ] and set L2 T = L (, FT , P ). We will denote by P the predictable sub- -eld of FT B[0,T ] . First we introduce some notation:
2 denote the space of square integrable and F progressively Let HT t measurable processes = {t , t [0, T ]} such that

||||2 = E

T 0

|t |2 dt < .

2 will denote the subspace of H 2 formed by predictable processes. MT T 2 (l2 ) and M 2 (l2 ) are the corresponding spaces of l2 valued processes HT T equipped with the norm T 0 i=1 (i)

||||2 = E
2 = H 2 M 2 (l2 ). Set HT T T

|t |2 dt .

Following Nualart and Schoutens (2000) we dene for every i = 1, 2, ... (i) the so-called power-jump processes {Xt , t 0} and their compensated (i) (i) (i) version {Yt = Xt E [Xt ], t 0}, also called the Teugels martingales, as follows Xt
(1)

= Xt and Xt = =
(i) Xt

(i)

(Xs )i for i = 2, 3, ...

0<st

(i) Yt

(i) E [Xt ]

= Xt t E [X1 ] for i 1.

(i)

(i)

An orthonormalization procedure can be applied to the martingales Y (i) in order to obtain a set of pairwise strongly orthonormal martingales {H (i) }i =1 ( ) i ) ( j such that each H is a linear combination of the Y , j = 1, ..., i: H (i) = ci,i Y (i) + ci,i1 Y (i1) + ... + ci,1 Y (1) . It was shown in Nualart and Schoutens (2000) that the coecients ci,k correspond to the orthonormalization of the polynomials 1, x, x2 , ... with respect to the measure (dx) = x2 (dx) + 2 0 (dx): qi1 (x) = ci,i xi1 + ci,i1 xi2 + ... + ci,1 . Set pi (x) = xqi1 (x) = ci,i xi + ci,i1 xi1 + ... + ci,1 x Then Ht
(i)

p i (x) = x(qi1 (x) qi1 (0)) = ci,i xi + ci,i1 xi1 + ... + ci,2 x2 . =
0<st

ci,i (Xs )i + ... + ci,2 (Xs )2 + ci,1 Xt tE ci,i X1 + ... + ci,2 X1

(i) (2)

= qi1 (0)Xt +

0<st (i)

p i (Xs ) tE

tci,1 E [X1 ]
0<s1

p i (Xs ) tqi1 (0)E [X1 ] .

As a consequence, Ht i = 1 we obtain where

= pi (Xt ) for each i 1. In the particular case

(1)

Ht

= c1,1 (Xt tE [X1 ]),

c1,1 =

y (dy ) + 5

1/2

and E [X1 ] = a +
{|z |1}

z (dz ).

In the case R |z |v (dz ) < , assuming a = {|z |<1} z (dz ), we obtain E [X1 ] = R z (dz ) . The main results in Nualart and Schoutens (2000) is the Predictable Representation Property (PRP): Every square integrable random variable F L2 T has a representation of the form F = E [F ] +
i=1 0 T (i) (i) Zs , dHs

2 (l2 ). where Zt is a predictable process in the space MT Remark: If = 0, we are in the classical Brownian case and all non-zero (i) degree polynomials qi (x) will vanish, giving Ht = 0, i = 2, 3, . . . . If (i) has only mass in 1, we are in the Poisson case; and also here Ht = 0, i = 2, 3, . . . . Both case are degenerate cases in this L evy framework. From these observations, it is not so hard to see that the PRP property shows that nancial markets based on a non-Brownian or non-Poissonian L evy process, i.e. with a stock price behaviour St = exp(Xt ), are so called incomplete, meaning that perfectly replicating or hedging strategies do not exists for all relevant contingent claims.

BSDE for L evy Processes

Taking into account the results and notation presented in the previous section, it seems natural to consider the BSDE dYt = f (t, Yt , Zt )dt where: Ht is the orthonormalized Teugels martingale of order i associated with the L evy process X .
2 (l2 ) R is a measurable function such that f : [0, T ] RMT 2. f (., 0, 0) HT (i) i=1

Zt dHt ,

(i)

(i)

YT = ,

(1)

 f is uniformly Lipschitz in the rst two components, i.e., there exists C > 0 such that dt dP a.s., for all (y1 , z1 ) and (y2 , z2 ) in Rl2 |f (t, y1 , z1 ) f (t, y2 , z2 )| C (|y1 y2 | + z1 z2 L2 T. If (f, ) satises the above assumptions, the pair (f, ) is said to be standard data for the BSDE. A solution of the BSDE is a pair of processes, 2 M 2 (l2 ) such that the following relation holds {(Yt , Zt ), 0 t T } HT T for all t [0, T ]:
T l2 ) .

Yt = +
t

f (s, Ys , Zs )ds

i=1 t

(i) (i) Zs dHs .

(2)

Note that the progressive measurability of {(Yt , Zt ), 0 t T } implies that (Y0 , Z0 ) is deterministic. A rst key-result concerns the existence and uniqueness of solution of BSDE: Theorem 1 Given standard data (f, ), there exists a unique solution (Y, Z ) which solves the BSDE (2). The proof can be found in the Appendix, as the proof of the continuous dependency of the solution on the nal data and the function f . Theorem 2 Given standard data (f, ) and (f , ), let (Y, Z ) and (Y Z ) be the unique adapted solutions of the BSDE (2) corresponding to (f, ) and (f , ). Then
T

E
0

|Ys Ys | +
T 0

i=1

(i) |Zs Zs(i) |2

ds .

C E [| |2 ] + E

|f (s, Ys , Zs ) f (s, Ys , Zs |2 ds

Applications

Suppose our L evy process Xt has no Brownian part, i.e. Xt = at + Lt , where Lt is pure jump process with L evy measure (dx).

4.1

Clark-Ocone Formula and Feynman-Kac Formula

Let us consider the simple case of a BSDE where f = 0, and the terminal random variable is a function of XT , that is, dYt = or equivalently Yt = g (XT )
i=1 i=1 t T

Zt dHt ;

(i)

(i)

YT = g (XT )

(i) (i) Zs , dHs

(3)

where E (g (XT )2 ) < . Let = (t, x) be the solution of the following PDIE (Partial Dierential Integral Equation) with terminal value g : (t, x) + t ((t, x + y ) (t, x)
{|y |1} y (dy ).

(t, x)y ) (dy ) + a (t, x) = 0, x x (T, x) = g (x), (4) Set (t, x)y. x (5)

where a = a +

(1) (t, x, y ) = (t, x + y ) (t, x)

The following result is a version of the Clark-Ocone formula for functions of a L evy process. Again the proof can be found in the Appendix.
Proposition 3 Suppose that is a C 1,2 function such that x and x 2 are bounded by a polynomial function of x, uniformly in t, then the unique adapted solution of (3) is given by
2

Yt = (t, Xt ) Zt Zt
(i)

= =

R R

(1) (t, Xt , y )pi (y ) (dy ) for i 2, (1) (t, Xt , y )p1 (y ) (dy ) + (t, Xt )( x

(1)

y 2 (dy ))1/2 , is given by

where = (t, x) is the solution of the PDIE (4) and (5).

(1) (t, x, y )

Now by taking expectations we derive that the solution (t, x) to our PDIE (4) equation has the stochastic representation (t, x) = E [g (XT )|Xt = x]. 8

This is an extension of the classical Feynman-Kac Formula. If R |y | (dy ) < , and we take a = {|y|<1} y (dy ), then the equation (4) reduces to (t, x) + t

((t, x + y ) (t, x)) (dy ) = 0, (T, x) = g (x),

and taking into account that p1 (y ) = y ( R y 2 (dy ))1/2 in Proposition 3 we have (1) [(t, Xt + y ) (t, Xt ) ] p1 (y ) (dy ). Zt =

Example: Consider the very special case where we have a compensated Poisson process Xt = Nt t. Then Ht
(1)

Xt 1 (i) = (Nt t) = and Ht = 0 for i = 2, 3, ....

x

Note that p1 (x) = reduces to

and pi (x) = 0, i = 2, 3, .... Moreover the PDIE (4) (t, x) + (t, x) = 0, x t (T, x) = g (x).

((t, x + 1) (t, x))

The Clark-Ocone Formula is now given by

T

g (XT ) = E [g (XT )] +
t

(s, Xs + 1) (s, Xs )dXs .

4.2

Nonlinear Clark-Haussman-Ocone Formula and FeynmanKac Formula

i=1

Let us consider the BSDE dYt = f (t, Yt, Zt )dt or equivalently

T

Z t d Ht ;

(i)

(i)

YT = g (XT )

(6)

Yt = g (XT ) +
t

f (s, Ys , Zs )ds 9

i=1 t

(i) (i) . dHs Zs

Suppose that = (t, x) satises the following PDIE: (t, x) + t

(1) (t, x, y ) (dy ) + a

i=1

(t, x) x = 0,

(7)

+f t, (t, x) , (i) (t, x) (T, x) = g (x).

where as in the previous section, we dene (1) (t, x, y ) by (5), (1) (t, x) = and for i 2

(1) (t, x, y )p1 (y ) (dy ) +

(t, x)( x

y 2 (dy ))1/2 ,

(8)

(i) (t, x) =

(1) (t, x, y )pi (y ) (dy ).

2

(9)

and x Proposition 4 Suppose that is a C 1,2 function such that x 2 are bounded by a polynomial function of x, uniformly in t. Then the (unique) adapted solution of (6) is given by

Yt = (t, Xt ) Zt Zt
(i)

= =

R R

(1) (t, Xt , y )pi (y ) (dy ) for i 2, (1) (t, Xt , y )p1 (y ) (dy ) + (t, Xt )( x

(1)

y 2 (dy ))1/2 . is given by

where = (t, x) is the solution of the PDIE (7) and (5). Notice that taking expectations we get
T

(1) (t, x, y )

(t, x) = E [g (XT )|Xt = x]+ f s, (s, Xs ) , (i) (s, Xs )

i=1

E
t

ds|Xt = x .

Example: Consider again the very special case where we have a Poisson process Nt with E [Nt ] = t. Set Xt = Nt t. Then the PDIE (7) reduces to ((t, x + 1) (t, x)) (t, x) + (t, x)+ x t f (t, (t, x) , (t, x + 1) (t, x)) = 0, (T, x) = g (x). 10

(10)

And we derive the nonlinear Feynman-Kac Formula: (t, x) = E [g (XT )|Xt = x]

T

+E
t

f (s, (s, Xs ), (s, Xs + 1) (s, Xs )) ds|Xt = x .

4.3

Option Pricing

Assume a market consisting of one riskless asset (the bond) with price process given by Bt = ert , where r is compound interest rate, and one risky asset (the stock), with price process: St = S0 exp(Xt ), where Xt is a L evy process. Denote by P (dx) the probability measure of X1 . In the last two decades several particular choices for non-Brownian L evy processes where proposed. Madan and Seneta [16] have proposed a L evy process with variance gamma distributed increments. We mention also the Hyperbolic Model proposed by Eberlein and Keller (1995). In the same year Barndor-Nielsen (1995) proposed the normal inverse Gaussian L evy process. Recently the CMGY model was introduced in Carr et al. (2000). Finally, we mention the Meixner model (see Grigelionis (1999) and Schoutens (2001)). All models give a much better t to the data and lead to an improvement with respect to the Black-Scholes model. We recall the density f , the cumulant generating function K , the drift a, and the L evy measure , for the Meixner Process {Mt , t 0}, for which we will illustrated the method.
) 2 ( + i(x (2 cos ) 2) e PMeix (dx) = fMeixner (x; , , , ) = , dx (2 ) + KMeixner (; , , , ) = + 2 log cos log cos , 2 2
(x)

aMeixner (, , , ) = + tan
x

2 2

sinh x x 1 sinh

dx,

e Meixner (dx; , , , ) = dx, x sinh x where > 0, < < , R, and > 0. 11

From the form of the cumulant generating function one easily deduces that the density at any time t can be calculated by multiplying the parameters and by t for both cases. Given our market model, let G(ST ) = F (XT ) denote the payo of the derivative at its time of expiry T . In case of the European call with strike price K , we have G(ST ) = (ST K )+ or equivalently F (XT ) = (S0 exp(XT ) K )+ . According to the fundamental theorem of asset pricing (see Delbaen and Schachermayer 1994) the arbitrage free price Vt of the derivative at time t [0, T ] is given by Vt = EQ [er(T t) G(ST )|Ft ], where the expectation is taken with respect to an equivalent martingale measure Q(dx) and F = {Ft , 0 t T } is the natural ltration of X = {Xt , 0 t T }. An equivalent martingale measure is a probability measure which is equivalent (it has the same null-sets) to the given (historical) probability measure and under which the discounted process {ert St } is a martingale. Unfortunately for most models, in particular the more realistic ones, the class of equivalent measures is rather large and often covers the full no-arbitrage interval. In this perspective the Black-Scholes model, where there is an unique equivalent martingale measure, is very exceptional. Models with more than one equivalent measures are called incomplete. Our L evy model is such an incomplete model. Following Gerber and Shiu (1994) and Gerber and Shiu (1996), we can, by using the so-called Esscher transform, easily nd at least one equivalent martingale measure, which we will use in the sequel for the valuation of derivative securities. The choice of the Esscher measure may be justied by a utility maximizing argument (see Gerber and Shiu (1996)). Let K be the cumulant generating function of X under the measure P (dx), and let be the solution of K ( + 1) K () = r. Then, we dene the risk-neutral measure Q(dx) as the probability measure with the Radon-Nykodym derivative with respect to P (dx) given by Q(dx)/P (dx) = exp(x K ()). For our Meixner-example, the parameters for the Esscher transforms are easily found; explicit values for can be found in Schoutens (2001) or Grigelionis (1999). In the Meixner case one only has to shift to + to get the density under the measure Q(dx). This means that under the riskneutral measure Q(dx) our process Mt is again a Meixner process. In all such cases where the underlying process is a L evy process in the risk-neutral world and the price Vt = V (t, Mt ) at time t of a given derivative satises 12

some regularity conditions (i.e. V (t, x) C (1,2) ), the function V (t, x) can also be obtained by solving a partial dierential integral equation (PDIE) with a boundary condition: a V (t, x) + V (t, x) + x t
+

V (t, x + y ) V (t, x) y

V (t, x) Q (dy ) x = rV (t, x) V (T, x) = F (x) ,

where Q (dy ) is the L evy measure of the risk-neutral distribution Q(dx). This PDIE is the analogue of the famous Black-Scholes partial dierential equation and follows from the above Feynman-Kac formula for L evy Processes. In the Meixner case, it is clear that: Q (dx) = d and a = aMeixner (, + , , ). exp((a + b)x/a) dx. x sinh(x/a)

Appendix: Proofs of the Results

Proof of Theorem 1:
2 into itself such that (Y, Z ) H2 is a We dene a mapping from HT T solution of the BSDE if and only if it is a xed point of . Given (U, V ) 2 , we dene (Y, Z ) = (U, V ) as follows: HT T

Yt = E +
t

f (s, Us , Vs )ds|Ft , 0 t T,

and {Zt , 0 t T } is given by the martingale representation of Nualart and Schoutens (2000) applied to the square integrable random variable
T

+
0

f (s, Us , Vs )ds,

i.e.,
T T

+
0

f (s, Us , Vs )ds = E + +
i=1 0

0 T

f (s, Us , Vs )ds
(i) (i) Zs dHs .

13

Taking the conditional expectation with respect to Ft in the last identity yields
t

Yt +
0

f (s, Us , Vs )ds = Y0 +

i=1 0

(i) (i) Zs dHs ,

from which we deduce that

T

Yt = +
t

f (s, Us , Vs )ds

i=1 t

(i) (i) Zs dHs

2 solves our BSDE if and only if it is a and we have shown that (Y, Z ) HT xed point of . 2 equipped with the Next we prove that is a strict contraction on HT norm T s 2 i=1 1/2 (i) 2 |Zs |

(Y, Z )

E
0

|Ys | +

ds

2 and set for a suitable > 0. Let (U, V ) and (U , V ) be two elements of HT (U, V ) = (Y, Z ) and (U , V ) = (Y , Z ) . Denote (U , V ) = (U U , V V ) and (Y , Z ) = (Y Y , Z Z ) . Applying It os formula from s = t to s = T , to es (Ys Ys )2 , it follows that

et Yt Yt

= 2
t

t T t T

es Ys Ys

ds

es Ys Ys d(Ys Ys ) (11)

es d[Y Y , Y Y ]s .

We have d(Yt Yt ) = (f (t, Ut , Vt ) f (t, Ut , Vt ))dt d Y Y ,Y Y H (i) , H (j )

t i=1

Zt Zt
(i)

(i)

(i)

dHt , Zt
(j )

(i)

i=1 j =1 t

Z t Zt

(i)

Zt

(j )

d[H (i) , H (j ) ]t ,

= ij t. 14

Hence, taking expectations in (11), we have

2 i=1 T t T 2

E e

Yt Yt

E
t

(i) es Zs Zs(i) 2

ds

= E
T

es Ys Ys

ds

+2E
t

es Ys Ys

f (s, Us , Vs ) f (s, Us , Vs ) ds .

Using the fact that f is Lipschitz with constant C yields

2 i=1 T T 2

E e

Yt Yt

E
t

(i) es ( Zs Zs(i)

ds

If we now use the fact that for every c > 0 and a, b R we have that 2 2 2 2 2ab ca2 + 1 c b and (a + b) 2a + 2b , we obtain E et Yt Yt (4C 2 )E 1 + E 2
T t 2

+2CE

E
T t

es Ys Ys |Us Us | +

es (Ys Ys )2 ds
i=1 (i) (i)

|Vs Vs |2 ds .

i=1

E
t 2

(i) es Zs Zs(i)

ds

T t s

es Ys Ys
2

ds
i=1

|Us Us | +

|Vs(i) Vs (i) |2

ds .

Taking now = 4C 2 + 1, and noting that et E [(Yt Yt )2 ] 0, we nally derive

T 2 i=1 i=1 T

E
t

es Ys Ys
T

ds +

E
t

(i) es (Zs Zs(i) )2 ds

1 E 2

es

|Us Us |2 + 15

|Vs(i) Vs (i) |2

ds ,

that is, (Y, Z )

2

1 (U, V ) 2

2 equipped with the from which it follows that is a strict contraction on HT 2 norm if = 4C + 1. Then has a unique xed point and the theorem is proved.

Proof of Theorem 2: Applying It os formula from s = t to s = T , to (Ys Ys )2 , it follows that YT YT

2

Yt Yt

= 2
t T

Ys Ys d(Ys Ys ) d[Y Y , Y Y ]s .

+
t

Taking expectations and using the relations d(Yt Yt ) = f (t, Yt , Zt ) f (t, Yt , Zt )dt d[Y Y , Y Y ]t = H (i) , H (j ) we have E [ Yt Yt = E[ +2E
t 2 T 2 t i=1

Zt Zt
(i)

(i)

(i)

dHt

(i)

i=1 j =1

Zt Zt

(i)

Zt

(j )

Zt

(j )

d[H (i) , H (j ) ]t ,

= ij t,

]+

i=1

E
t

(i) Zs Zs(i) ds

] Ys Ys f (s, Ys , Zs ) f (s, Ys , Zs ) ds .

16

Using the Lipschitz property of f , and computations similar to those of the proof of Theorem 1 we obtain 1 2 E [ Yt Yt ] + E 2
T t i=1 (i) Zs(i) |2 ds | Zs T t

E [| |2 ] + (1 + 2C + 2C 2 )E
T

|Ys Ys |2 ds

+E
t

|f (s, Ys, Zs ) f (s, Ys, Zs |2 ds .

Then by Gronwalls inequality the result follows. Lemma 5 Let h : [0, T ] R R be a random function measurable with respect to P B R such that |h(s, y )| as (y 2 |y |) a.s., (12)
T 0

where {as , 0 s T } is a nonnegative predictable process such that E [ . Then for each t [0, T ] we have h(s, Xs ) =
t<sT i=1 t T

a2 s ds] <

h(s, ), pi

(i) L2 ( ) dHs

+
t t 0

h(s, y ) (dy )ds.

Proof of Lemma 5: Because (12) implies that E [ , we have that

t

2 R |h(s, y )| (dy)ds] <

Mt =
0<st

h(s, Xs )

h(s, y ) (dy )ds.

is a square integrable martingale. By the Predictable Representation Theo2 (l2 ) such that rem, there exists a process in the space MT Mt =
i=1 0 t (i) i) ( s dHs .

Taking into account that H (i) , H (j ) M, H (i)

= tji , we have
t (i) s ds.

=
0

(13)

17

On the other hand, using that Ms Hs = h(s, Xs )pi (Xs ) we obtain M, H (i)
t t

(i)

=
0

h(s, y )pi (y ) (dy )ds.

(14)

Consequently, (13) and (14) imply

i) ( s =

h(s, y )pi (y ) (dy ),

and the result follows. Proof of Proposition 3: Under the hypotheses of Proposition 3 the function (1) (t, x, y ) given by 5) satises the hypotheses in Lemma 5 imposed on h due to the mean value theorem, when we take x = Xt . Apply It os lemma to (s, Xs ) from s = t to s = T :
T

(T, XT ) (t, Xt ) =

(s, Xs )ds + t

+
t<sT

(s, Xs )dXs (15) x t (s, Xs )Xs . (s, Xs ) (s, Xs ) x

x (s, Xs )y ,

If we apply Lemma 5 to h(s, y ) = (s, Xs + y ) (s, Xs ) we obtain (s, Xs ) (s, Xs ) (s, Xs )Xs x

t<sT T i=1 t T

R R

(i) (1) (s, Xs , y )pi (y ) (dy ) dHs

+
t

(1) (s, Xs , y ) (dy )ds.

(16)

Hence, substituting (16) into (15) yields

18

g (XT ) (t, Xt )
T

=
t

(s, Xs )ds + t
T

T t

(s, Xs )dXs x

+ +

i=1 t T t

(i) (1) (s, Xs , y )pi (y ) (dy ) dHs

(1) (s, Xs , y ) (dy )ds.

(17)

Notice that Xt = Yt and E (X1 ) = a +

{|y |1} (1)

+ tE (X1 ) = (

y 2 (dy ))1/2 Ht

(1)

+ tE (X1 ),

y (dy ).

We also have Y0 = E [Y0 ] = E [g (XT )] so we can rewrite (17) as

T

g (XT ) = E [g (XT )] +
t

(s, Xs )( x

i=1 t

(1) y 2 (dy ))1/2 dHs

(i) (1) (s, Xs , y )pi (y ) (dy ) dHs ,

which completes the proof of the Proposition. Proof of Proposition 4: Apply It os lemma to (s, Xs ) from s = t to s = T . By using Lemma 5, we obtain the equality (17). Now, using (7) we get
T

g (XT ) (t, Xt ) =
T

f s, (s, Xs ) ,

(k) (s, Xs )

k=1

ds

+
t

(1) (s, Xs , y )p1 (y ) (dy ) + +

i=2 t T

(s, Xs )( x

(1) y 2 (dy ))1/2 dHs

(i) (1) (s, Xs , y )pi (y ) (dy ) dHs ,

which completes the proof of the Proposition. 19

Acknowledgement: The second named author is a Postdoctoral Fellow of the Fund for Scientic Research - Flanders (Belgium) (F.W.O. - Vlaanderen). The authors also would like to thank the referees for their useful suggestions for improvement of the manuscript.

References
[1] Barndor-Nielsen, O.E. (1995) Normal inverse Gaussian distributions and the modeling of stock returns. Research Report No. 300, Department of Theoretical Statistics, Aarhus University. [2] Bertoin, J. (1996) L evy Processes. Cambridge Tracts in Mathematics 121, Cambridge University Press, Cambridge. [3] Black, F. and Scholes, M. (1973) The pricing of options and corporate liabilities. Journal of Political Economy 81, 637654. [4] Bismut, J.M. (1973) Conjugate Convex Functions in Optimal Stochastic Control. J. Math. Anal. Appl. 44, 384404. ole des syst` emes lin eaires quadratiques: ap[5] Bismut, J.M. (1978) Contr plications de lint egrale stochastique. S emin. Probab. XII. Lecture Notes in Mathematics 649, 180264. [6] Carr, P., Geman, H., Madan, D.H. and Yor, M. (2000) The ne structure of asset returns: an empirical investigations. Preprint. [7] Delbaen, F. and Schachermayer, W. (1994) A general version of the fundamental theorem of asset pricing. Math. Ann. 300, 463520. [8] Eberlein, E. and Keller, U. (1995) Hyperbolic distributions in nance. Bernoulli 1, 281299. [9] El Karoui, N. and Quenez, M.C. (1997) Non-linear Pricing Theory and Backward Stochastic Dierential Equations. In: Financial Mathematics. Lecture Notes in Mathematics 1656, 191246. [10] Gerber, H.U. and Shiu, E.S.W (1994) Option pricing by Esschertransforms. Transactions of the Society of Actuaries 46, 99191. [11] Gerber, H.U. and Shiu, E.S.W (1996) Actuarial bridges to dynamic hedging and option pricing. Insur. Math. Econ. 18 (3), 183218.

20

[12] Grigelionis, B. (1999), Processes of Meixner Type. Lith. Math. J. 39 (1), 3341. [13] Grigelionis, B. (2000), Generalized z -Distributions and related Stochastic Processes. Matematikos Ir Informatikos Institutas Preprintas Nr. 2000-22, Vilnius. [14] Jacod, J. (1979) Calcul stochastique et Probl` emes de Martingales. Lecture Notes in Mathematics 714. Springer-Verlag. [15] Ma, J. and Yong, J. (1999) Forward-Backward Stochastic Dierential Equations and Their Applications. Lecture Notes in Mathematics 1702. Springer, Berlin. [16] Madan, D.B. and Seneta, E. (1990) The v.g. model for share market returns. Journal of Business 63, 511524. [17] Nualart, D. and Schoutens W. (2000) Chaotic and predictable representations for L evy processes. Stochastic Processes and their Applications 90, 109122. [18] Oldman, K. and Spanier, J. (1987) An Atlas of Functions. Springer, 1987. [19] Ouknine, Y. (1998) Reected Backward Stochastic Dierential Equations with Jumps. Stochastics and Stochastics Reports 65, 111125. [20] Pardoux, E. (1996) Backward Stochastic Dierential Equations and Viscosity Solutions of Systems of Semilinear Parabolic and Elliptic PDEs of Second Order. In: Stochastic Analysis and Related Topics VI. The Geilo Workshop 1996 (eds. L. Decreusefond et al.) Progress in Probability 42, 79127. [21] Pardoux, E. and Peng, S. (1990) Adapted Solution of a Backward Stochastic Dierential Equation. Systems and Control Letters 14, 55 61. [22] Protter, Ph. (1990) Stochastic Integration and Dierential Equations. Springer-Verlag. Berlin [23] Sato, K. (2000) L evy Processes and Innitely Divisible Distributions. Cambridge Studies in Advanced Mathematics 68. Cambridge University Press, Cambridge.

21

[24] Schoutens, W. (2001) The Meixner Process in Finance. EURANDOMReport 2001-02, Eurandom, Eindhoven. [25] Situ, R. (1997) On solutions of backward stochastic dierential equations with jumps and applications. Stochastic Processes and their Applications 66, 209236.

22