An Application Of The Stochastic Mcshane’S

Equations In Financial Modelling

G. Constantin

West Univesity of Timisoara,

300 223 TIMISOARA, Romania
(e-mail: const@math.uvt.ro)

Abstract. We prove a result for the solvability of linear forward-backward stochas-

tic differential equations of McShane type. The motivation for the study is a similar
Black-Scholes type model in mathematical finance.
Keywords: forward-backward stochastic differential equations, McShane type in-
tegral, Black-Scholes type model.

1 Introduction
One of the most remarkable applications of stocahstic analysis is in math-
ematical finance. In particular, the Black-Scholes model enjoys great popu-
larity (see, for example, [Musiela and Rutkowski, 1997]). Recently (see [Ma
and Yong, 1999]), this model was derived by means of the theory of forward-
backward stochastic differential equations of Itô type a setting appropriate
for the case in which the filtration of the undelying probability space is given
by Brownian motion. It appears that for a filtration induced by a finite varia-
tion process with a.s. continous sample paths, the Itô type stochastic integral
is no longer appropriate. In this case one can use a McShane type integral
(introduced by McShane in [McShane, 1969], [McShane, 1974] and further
developed by Srinivasan in [Srinivasan, 1978] and, from a different point of
wiev, by Protter in [Protter, 1992]; so a stochastic calculus could be called
”unified calculus” since it includes ordinary calculus as a special case and
also Itô Calculus) to construct a suitable model. This leads us to forward-
backward stochastic differential equations of McShane type. Despite many
investigation related to McShane type stochastic differential equations (see
[Angulo Ibanez and Gutierrez Jaimez, 1988], [Constantin, 1998], [Ladde and
Seikkala, 1986], [McShane, 1974] for theoretical approaches and [Srinivasan,
1978], [Srinivasan, 1984], [Hangii, 1980] for applications of McShane stochas-
tic calculus to problems in physics) a study of forward-backward stochastic
differential equations of McShane type has not been undertaken, to the best
of our knowledge.
Stochastic calculus appears to be one of the natural tools for the study
of models of those phenomena having some non-deterministic elements. For
example, in the description of brownian motion the stochastic nature is ad-
equately described by a linear differential equation with a random forcing
918 Constantin

term which is identified as a white noise process or has a formal derivative of

the Wiener process.
However, when the results of the stochastic calculus were applied to other
types phenomena, certain difficulties arose in the process of interpretation of
stochastic differentials and approximation process. In many models, white
noise process is explicitly introduced and the basic physical process in ques-
tion is visualised as an approximation. Hence it is reasonable to expect some
kind of a stability in the sense that the solutions that are obtained by approx-
imating the white noise process should themselves approximate the process
in question.
Ito stochastic calculus failed to satisfy this requirement of stability (see
[McShane, 1974]). Moreover, in choosing the type of stochastic processes
that we shall use us models of the noises we meet a dilema. On the one
hand, there is no physical bases for considering an example considering any
simple functions Wj (t) exccept those of a rather simple structure. In fact, the
noise input Wj (t) − Wj (s) is measured be some sort of indicator and if this is
mechanical it cannot move faster than the velocity of light, if it is electrical, it
cannot suport more than some limited current or voltage difference without
destruction and also some similiraties are in the financial modeling case.
In McShane’s Calculus, the standard equations
Zt r Z
X
t
i i i
(I) X (t, ω) = X (0, ω)+ f (s, X(s, ω))ds+ gji (s, X(s, ω))dWj (s, ω)
0 j=1 0

are replaced by what he calls a canonical extension (or canonical form or

canonical system) of equation (I):
Zt
i i
(II) X (t, ω) = X (0, ω) + f i (s, X(s, ω))ds+
0

r Z t r Zt
X 1 X
gji (s, X(s, ω))dWj (s, ω) + i
gj,k (s, X(s, ω))dWj (s, ω)dWk (s, ω)
j=1 0
2
j,k=1 0

in which
n
X
i
gj,k (t, x, ω) = [∂gji (t, x, ω)/∂xm ]gkm (t, x, ω)
m=1

i = 1, 2, ..., n; j, k = 1, 2, ..., r; t ∈ [0, a]; x ∈ Rn .

We are now able to describe the method by which we shall construct
stochastic models of physical systems which in the physically realizable case
of lipschitzian noises are known to satisfy the integral equation (I).
i
If gj,k (t, x, ω) are functions defined for t ∈ [0, a] and x ∈ Rn and bounded
on bounded sets of (t, x), then the solution X i (t, ω) of (I) is also a solution
of (II) since the last integral vanishes for lipschitzian noises.
 An Appl. of the Stoch. McShane’s Eq. in Financial Modelling 919

The McShane Calculus is better suited modeling dynamical phenomena

described typically by McShane systems where Wj (t, ω) are noises processes.
McShane stochastic integral systems enjoy the following three important
properties:
(i) The property of inclusiveness: the model must apply to systems in
which the permitted noises are processes belonging to some family large
enough to include processes with sample paths having lipschitzian property,
all brownian motion processes, and such modifications as have proved conve-
nient in applications;
(ii) The property of consistency: for lipschitzian noises, the solutions of
the equations should coincide with the solutions of the equations that are
normally believed to be applicable to physical systems;
(iii) The property of stability: the model must be such that if the noise
process Wj (t, ω) is replaced by another permissible process Wj0 (t, ω) close
to it, then the corresponding solutions X i (t, ω), X0i (t, ω) are also close to
each other (in the sense that an extreme degree of closeness corresponds
to practical imposibility of distinguishing the process by means of available
experimental procedures).
In section 2 we pursue the study of the solvability of a class of linear
forward-backward stochastic differential equations of McShane type and we
point out some drastic differences from the case of Itô type stochastic equa-
tions.
The approach developed in Section 2 is applied in Section 3 to a similar
Black-Scholes type model in mathematical finance.

2 The main result

Consider the following forward-backward stochastic differential equations
on [0, T ],

 dX(t) = [a(t)X(t) + b(t)]dt + [c(t)X(t) + d(t)]dW (t)
dY (t) = [f (t)X(t) + g(t)Y (t) + h(t)Z(t) + k(t)]dt + Z(t)dW (t) (1)
X(0) = x0 , Y (T ) = α(X(T ))


where T > 0, x0 ∈ R and a, b, c, d, f, g, h, k : [0, T ] → R are continuous func-

tions, while α : R → R is a function of class C 1 . In (1), (X(t), Y (t), Z(t))
is a triplet of adapted stochastic processes on a complete filtered probabil-
ity space (Ω, F , {Ft }t∈[0,T ] , P) such that {F t }t∈[0,T ] is the natural filtration
of a given stochastic process {W (t)}t∈[0,T ] , augmented with all P-null sets.
Throughout this paper, the process {W (t)}t∈[0,T ] inducing the filtration is a
finite variation process with continuous paths i.e. for almost all ω ∈ Ω the
sample path t → W (t, ω) is continuous and of finite variation on [0, T ] as a
particular noise of McShane type. For example, a process satisfying a.s. a
Lipschitz condition
W (t, ω) − W (s, ω)| ≤ L|t − s|, 0 ≤ s ≤ t ≤ T
920 Constantin

for some constant L > 0, is admissible. Let CF [0, T ] be the set of all
{Ft }t∈[0,T ] -progresively measurable continuous processes X : [0, T ] × Ω → R
(that is, for almost all ω ∈ Ω the sample paths t → X(t, ω) is continuous on
[0, T ]), such that E sup |X(t)|2 < ∞. Observe that the space
t∈[0,T ]

MF [0, T ] = CF [0, T ] × CF [0, T ] × CF [0, T ]

is a Banach space under the norm

1
k(X, Y, Z)k = {E sup |X(t)|2 + E sup |Y (t)|2 + E sup |Z(t)|2 } 2 .
t∈[0,T ] t∈[0,T ] t∈[0,T ]

Given a, b, c, d, f, g, h, k ∈ C([0, T ], R), α ∈ C 1 (R, R), x0 ∈ R, and the

finite variation continuous process {W (t)}t∈[0,T ] inducing the filtration on
the probability space, a process (X, Y, Z) ∈ MF [0, T ] is called an adapted
solution of (1) if the following holds for any t ∈ [0, T ], almost surely:

Rt t
X(t) = x0 + [a(s)X(s) + b(s)]ds + ınt[c(s)X(s) + d(s)]dW (s),



0

0



 RT
Y (t) = α(X(T )) − [f (s)X(s) + g(s)Y (s) + h(s)Z(s) + k(s)]ds− (2)

 t
RT





 − Z(s)dW (s),
t

where the stochastic integrals are McShane type integrals (see [Protter, 1992]
for an approach to this integral close in spirit to the original one by McShane
[McShane, 1969]). In Section 3 we will give an example in mathematical fi-
nance that motivates the study of (1). Let us now prove the solvability of (1).

Theorem. The system (1) admits an adapted solution (X, Y, Z) ∈

MF [0, T ].

Proof. To show the existence of a solution we introduce a direct

method, similar to the scheme developed in [Ma et al., 1994] for Itô type
forward-backward stochastic differential equations. We will prove that the
following three-step scheme is realizable:

(A) let θ : [0, T ] × R → R be the C 1 -solution of the following first-order

linear partial differential equation

 θt + ([a(t) − c(t)h(t)]x + b(t) − d(t)h(t))θx =
= g(t)θ + f (t)x + k(t), t ∈ [0, T ], x ∈ R, (3)
θ(T, x) = α(x), x ∈ R;

 An Appl. of the Stoch. McShane’s Eq. in Financial Modelling 921

(B) let X ∈ CF [0, T ] be the solution of the following forward stochastic

differential equation of McShane type:

 dX(t) = [a(t)X(t) + b(t)]dt+
+[c(t)X(t) + d(t)]dW (t), t ∈ [0, T ], (4)
X(0) = x0 ;


(C) then X together with

Y (t) = θ(t, X(t)), Z(t) = θx (t, X(t))(c(t)X(t) + d(t)), t ∈ [0, T ] (5)

is an adapted solution to (1).

The special relations (5) among the components of the adapted solution
(X, Y, Z) ∈ MF [0, T ] to (1) are suggested by the change of variables formula
for McShane type stochastic integrals (see [McShane, 1974], p.146): if X ∈
CF [0, T ] solves (4), then

dθ(t, X(t)) = [θt (t, X(t)) + θx (t, X(t))(a(t)X(t) + b(t))]dt+

+θx (t, X(t))(c(t)X(t) + d(t))dW (t)

and a comparison with the backward stochastic equation (for Y ) in (2) con-
firms that the problem (3) for θ is precisely what is needed for the effectiveness
of the solution scheme. Therefore, the existence part is proved if we show
that steps (A) and (B) can be performed. Both problems can be explicitely
solved. Indeed, the solution of (4) is (see [McShane, 1974], p.129-130)

X(t) = [x0 + Ψ (t)]eΦ(t) , t ∈ [0, T ], (6)

with
Zt Zt
Φ(t) = a(s)ds + c(s)dW (s), t ∈ [0, T ],
0 0

and
Zt Zt
−φ(s)
Ψ= e b(s)ds + e−φ(s) dsdW (s), t ∈ [0, T ].
0 0

On the other hand, the method of characteristics enables us [John, 1962]

to write down the explicit C 1 -solution of the Cauchy problem (3). However,
taking into account the intricacy of the resulting formula, we refrain from
further details- we shall do the full details of the solution in Section 3 for
the choice of coefficients in (1) dictated by a model in mathematical finance.
The proof of the theorem is completed.
The statement of the theorem leaves open the question of uniqueness. Our
solution scheme was constructed in analogy with the four step scheme (see
[Ma et al., 1994]) for the Itô type problem (2)- case in which {W (t)}t∈[0,T ] is
922 Constantin

Brownian motion (a process with a.s. continuous sample paths but a.s. the
sample paths are of unbounded variation functions [Protter, 1992]) and all
stochastic integrals in (2) are of Itô type.
For the Itô type problem (2), uniqueness holds (see [Ma and Yong, 1999],
p.82) so that it is not unreasonable to expect uniqueness in the McShane
type problem (2) that we are investigating. However, let us note an essen-
tial difference in the two solution schemes (the four step scheme from [Ma
and Yong, 1999] and our three step scheme) which indicates that the Mc-
Shane type problem is not a perfect replicate to the Itô type problem. In
both problems the forward stochastic differential equation are replaced by a
forward stochastic differential equation coupled with a Cauchy problem for
a partial differential equation: in the Itô type problem we have a parabolic
partial differential equation while in the McShane scheme type problem we
have a linear first order partial differential equation. For parabolic partial
differential equations, time-reversibility is not to be expected whereas for lin-
ear first-order partial differential equations this is not an issue. Here lies an
essential difference between the schemes adapted in the Itô type case, respec-
tively in the McShane type case. An example illustrates that the uniqueness
for the McShane type case isn’t assured.

3 Applications
In this section we analyse a model in mathematical finance that motivates
the study of forward-backward stochastic differential equations of McShane
type.
Consider a market that contains one bond and one stock. Their prices
at time t are denoted by P (t) and X(t), respectively. An investor trades
continuously, the wealth of the investor at time t being denoted by Y (t) and
the amount of money invested into the stock at time t is denoted by π(t),
called portfolio, while the rest of the money at time t, Y (t) − π(t), is put
into the bond. In a stochastic model (model with uncertainly) one assumes
that both prices are stochastic processes, defined on some filtered probability
space (Ω, F , {Ft }t≥0 , P). The fact that both prices can only be determined
by the information up to time t is expressed mathematically by requiring the
processes P (t), X(t) to be both adapted to the filtration {Ft }t≥0 . We assume
that the filtration is generated by a given continuous process {W (t)}t≥0 with
sample paths of bounded variation on compact intervals. If the market is
assumed to be Markovian, that is, the interes rate r(t) of the bond and the
appreciation rate and volatility of the stock b(t), respectively σ, are deter-
ministic (the time-dependence is assumed to be continuous), then the prices
are subject to the following system of stochastic differential equations

 dP (t) = r(t)P (t)dt, (bond)
dX(t) = X(t)b(t)dt + σX(t)dW (t), (stock) (7)
P (0) = 1, X(0) = x0 ,

 An Appl. of the Stoch. McShane’s Eq. in Financial Modelling 923

where x0 > 0 is a constant. The change of wealth dY (t) follows therefore the
dynamics
π(t) Y (t) − π(t)
dY (t) = dX(t) + dP (t). (8)
X(t) P (t)
An option with maturity date T > 0 is an FT -measurable random variable
α(X(T )), where α : R → R is a function of class C 1 . Let us rewrite (7)-(8)
as 
Rt

 r(s)ds
 P (t) = e 0 ,


Rt t
 X(t) = x0 + b(s)X(s)ds + σ ∈ t X(s)dW (s),


 0 0
dY (t) = [π(t)b(t) + r(t)(Y (t) − π(t))]dt + σπ(t)dW (t),


for t ∈ [0, T ]. The interaction between the investor’s wealth/strategy and

the stock price is described by the following forward-backward stochastic
differential equations of McShane type

Rt t
 X(t) = x0 + b(s)X(s)ds + σ ∈ t X(s)dW (s); t ∈ [0, T ],


0

0



 RT
Y (t) = α(X(T )) − [r(s)Y (s) + (b(s) − r(s))π(s)]ds− (9)

 t
RT





 −σ π(s)dW (s), t ∈ [0, T ].
t

The purpose of the investor is to find an adapted solution (X, Y, π) to (9);

this amounts to choosing a strategy π allowing the realization of the option
Y (T ) = α(X(T )).
The problem (9) is of type (2) so that we may apply our three step scheme
developed in Section 2 to find an explicit solution. Relation (6) ensures that
the solution of the equation for X in (9) is precisely
Rt
σ[W (t)−W (0)]+ b(s)ds
X(t) = x0 e 0 , t ∈ [0, T ]. (10)
To find the explicit formula for the wealth Y (t), we have to solve the
problem (3), i.e.

θt + r(t)xθx = r(t)θ, t ∈ [0, T ], x ∈ R,
(11)
θ(T, x) = α(x) x ∈ R.
In accordance to the study pursued in Section 2, we apply the method
of characteristics to solve (11). The characteristic curves are given by the
system of ordinary differential equations (with parameter s)
dt


 = 1,

 ds
dx

= r(t)x, (12)
 ds
dθ


= r(t)θ,


ds
924 Constantin

and the Cauchy data corresponds to (at s = 0)

t = T, x = ξ, θ = α(ξ). (13)

The solution to (12)-(13) has the parametric representation

Rs Rs
r(τ +T )dτ r(τ +T )dτ
t = s + T, x = ξe 0 , θ = α(ξ)e 0 ,

as it can be easily verified. Eliminating s, ξ we find for the C 1 -solution of

the Cauchy problem (11) the representation
T
R
r(τ )dτ
limitsT
R
θ(t, x) = α(xe t )e− t r(τ )dτ (14)

since
T
R
r(τ )dτ
s = t − T, ξ = xe t .
We can check directly that (14) solves (11).
As a consequence of our theorem, taking into account relations (5), (10)
and (14), we find that a solution of the problem (9) is given by

Rt

 σ[W (t)−W (0)]+ b(s)ds
X(t) = x0 e , t ∈ [0, T ],
 0


 Y (t) = θ(t, X(t)), t ∈ [0, T ],
Z(t) = X(t)θx (t, X(t)), t ∈ [0, T ].


Remark. The model presented above is of Black-Scholes type because if

we consider b(t), r(t) to be positive constants and the process {W (t)}t∈[0,T ]
to be a Brownian motion, interpreting (9) as an Itô type problem, we end up
with a parabolic problem instead of (11): the Black- Scholes partial differen-
tial equation (see [Ma and Yong, 1999], p.227).

References
[Angulo Ibanez and Gutierrez Jaimez, 1988]J.M. Angulo Ibanez and R. Gutier-
rez Jaimez. On the existence and uniqueness of the soltuion processes in
mcshane’s stochastic integral equation systems. Ann.Sci.Univ.Blaise Pascal,
pages 1–9, 1988.
[Constantin, 1998]A. Constantin. On the existence, uniqueness and parametric
dependence on the coefficients of the solution processes in mcshane’s stochastic
integral equations. Publ.Math., pages 11–24, 1998.
[Hangii, 1980]P. Hangii. Langevin description of markovian integro-differential mas-
ter equations. Z.Phyzik B., pages 271–282, 1980.
[John, 1962]F. John. Partial Differential Equations. Springer Verlag, New York,
1962.
 An Appl. of the Stoch. McShane’s Eq. in Financial Modelling 925

[Ladde and Seikkala, 1986]G.S. Ladde and S. Seikkala. Existence, uniqueness and
upper estimates for solutions of mcshane type stochastic differential systems.
Stoch.Anal.Appl., pages 409–430, 1986.
[Ma and Yong, 1999]J. Ma and J. Yong. Forward-Backward Stochastic Differential
Equations and Their Applications. Springer Verlag, Berlin, 1999.
[Ma et al., 1994]J. Ma, P. Protter, and J. Yong. Solving forward-backward stochas-
tic differential equations explicitly – a four step scheme. Prob.Th.Rel.Fields.,
pages 339–359, 1994.
[McShane, 1969]E.J. McShane. Stochastic integral and stochastic functional equa-
tions. SIAM J.Appl.Math., pages 287–306, 1969.
[McShane, 1974]E.J. McShane. Stochastic Calculus and Stochastic Models. Aca-
demic Press, New York, 1974.
[Musiela and Rutkowski, 1997]M. Musiela and M. Rutkowski. Martingale Methods
in Financial Modelling. Springer Verlag, Berlin, 1997.
[Protter, 1992]P. Protter. Stochastic Integration and Differential Equations.
Springer Verlag, Berlin, 1992.
[Srinivasan, 1978]S.K. Srinivasan. Stochastic integrals, solid mechanics archives.
Solid Mechanics Archives, pages 325–379, 1978.
[Srinivasan, 1984]S.K. Srinivasan. Stochastic calculus and models of physical phe-
nomena. J.Math.Phys.Sci., pages 163–168, 1984.