Chapter 8

Option Hedging in Continuous Time

Here we review some applications to mathematical nance of the tools
introduced in the previous chapters. We construct a market model with jumps
in which exponential normal martingales are used to model random prices.
We obtain pricing and hedging formulas for contingent claims, extending the
classical Black-Scholes theory to other complete markets with jumps.
8.1 Market Model
Let (M
t
)
tR
+
be a martingale having the chaos representation property of
Denition 2.8.1 and angle bracket given by dM, M)
t
=
2
t
dt. By a modi-
cation of Proposition 2.10.2, (M
t
)
t[0,T]
satises the structure equation
d[M, M]
t
=
2
t
dt +
t
dM
t
.
When (
t
)
t[0,T]
is deterministic, (M
t
)
t[0,T]
is alternatively a Brownian mo-
tion or a compensated Poisson martingale, depending on the vanishing of
(
t
)
t[0,T]
.
Let r : R
+
R and : R
+
(0, ) be deterministic non negative
bounded functions. We assume that 1 +
t

t
> 0, t [0, T]. Let (A
t
)
tR
+
denote the price of the riskless asset, given by
dA
t
A
t
= r
t
dt, A
0
= 1, t R
+
, (8.1.1)
i.e.
A
t
= A
0
exp
__
t
0
r
s
ds
_
, t R
+
.
For t > 0, let (S
x
t,u
)
u[t,T]
be the price process with risk-neutral dynamics
given by
dS
x
t,u
= r
t
S
x
t,u
du +
u
S
x
t,u

dM
u
, u [t, T], S
x
t,t
= x,
N. Privault, Stochastic Analysis in Discrete and Continuous Settings,
Lecture Notes in Mathematics 1982, DOI 10.1007/978-3-642-02380-4 8,
c Springer-Verlag Berlin Heidelberg 2009
281
282 8 Option Hedging in Continuous Time
0
2
4
6
8
10
12
14
16
18
20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 2.0
Fig. 8.1 Sample trajectory of (S
t
)
t[0,T]
cf. Relation 2.13.5. Recall that when (
t
)
tR
+
is deterministic we have
S
x
t,T
= xexp
_
_
T
t

u
i
u
dB
u
+
_
T
t
(r
u

u

u

1
2
i
u

2
u

2
u
)du
_

k=N
T

k=1+N
t
(1 +
T
k

T
k
), (8.1.2)
0 t T, with S
t
= S
1
0,t
, t [0, T]. Figure 8.1 shows a sample path of
(S
t
)
t[0,T]
when the function (i
t
)
t[0,T]
takes values in 0, 1, with S
0
= 10,

t
= 10, and
t
= 1, t [0, T].
Let
t
and
t
be the numbers of units invested at time t, respectively in the as-
sets (S
t
)
tR
+
and (A
t
)
tR
+
. The value of the portfolio V
t
at time t is given by
V
t
=
t
A
t
+
t
S
t
, t R
+
. (8.1.3)
Denition 8.1.1. The portfolio V
t
is said to be self-nancing if
dV
t
=
t
dA
t
+
t
dS
t
. (8.1.4)
The self-nancing condition can be written as
A
t
d
t
+S
t
d
t
= 0, 0 t T
under the approximation dS
t
,
t
) 0.
8.1 Market Model 283
Let also

V
t
= V
t
exp
_

_
t
0
r
s
ds
_
and

S
t
= S
t
exp
_

_
t
0
r
s
ds
_
denote respectively the discounted portfolio price and underlying asset price.
Lemma 8.1.2. The following statements are equivalent:
i) the portfolio V
t
is self-nancing,
ii) we have

V
t
=

V
0
+
_
t
0

u

S
u
dM
u
, t R
+
, (8.1.5)
iii) we have
V
t
= V
0
exp
__
t
0
r
u
du
_
+
_
t
0

u
exp
__
t
u
r
u
du
_
S
u
dM
u
, (8.1.6)
t R
+
.
Proof. First, note that (8.1.5) is clearly equivalent to (8.1.6). Next, the self-
nancing condition (8.1.4) shows that
dV
t
=
t
dA
t
+
t
dS
t
=
t
A
t
r
t
dt +
t
r
t
S
t
dt +
t

t
S
t
dM
t
= r
t
V
t
dt +
t

t
S
t
dM
t
,
t R
+
, hence
d

V
t
= d
_
exp
_

_
t
0
r
s
ds
_
V
t
_
= r
t
exp
_

_
t
0
r
s
ds
_
V
t
dt + exp
_

_
t
0
r
s
ds
_
dV
t
= exp
_

_
t
0
r
s
ds
_

t
S
t
dM
t
, t R
+
,
i.e. (8.1.5) holds. Conversely, if (8.1.5) is satised we have
dV
t
= d(A
t

V
t
)
=

V
t
dA
t
+A
t
d

V
t
=

V
t
A
t
r
t
dt +
t

t
S
t
dM
t
= V
t
r
t
dt +
t

t
S
t
dM
t
=
t
A
t
r
t
dt +
t
S
t
r
t
dt +
t

t
S
t
dM
t
=
t
dA
t
+
t
dS
t
,
hence the portfolio is self-nancing.
284 8 Option Hedging in Continuous Time
8.2 Hedging by the Clark Formula
In the next proposition we compute a self-nancing hedging strategy leading
to an arbitrary square-integrable random variable F, using the Clark formula
Proposition 4.2.3.
Proposition 8.2.1. Given F L
2
(), let

t
=
exp
_

_
T
t
r
s
ds
_

t
S
t
IE[D
t
F[T
t
], (8.2.1)

t
=
exp
_

_
T
t
r
u
du
_
IE[F[T
t
]
t
S
t
A
t
, t [0, T]. (8.2.2)
Then the portfolio (
t
,
t
)
t[0,T]
is self-nancing and yields a hedging strategy
leading to F, i.e. letting
V
t
=
t
A
t
+
t
S
t
, 0 t T,
we have
V
t
= exp
_

_
T
t
r
u
du
_
IE[F[T
t
], (8.2.3)
0 t T. In particular we have V
T
= F and
V
0
= exp
_

_
T
0
r
u
du
_
IE[F].
Proof. Applying (8.2.2) at t = 0 we get
IE[F] exp
_

_
T
0
r
u
du
_
= V
0
,
hence from (8.2.2), the denition (8.2.1) of
t
and the Clark formula we obtain
V
t
=
t
A
t
+
t
S
t
= exp
_

_
T
t
r
u
du
_
IE[F[T
t
]
= exp
_

_
T
t
r
u
du
_
_
IE[F] +
_
t
0
IE[D
u
F[T
u
]dM
u
_
= V
0
exp
__
t
0
r
u
du
_
+ exp
_

_
T
t
r
u
du
_
_
t
0
IE[D
u
F[T
u
]dM
u
= V
0
exp
__
t
0
r
u
du
_
+
_
t
0

u
S
u
exp
__
t
u
r
s
ds
_
dM
u
, 0 t T,
8.2 Hedging by the Clark Formula 285
and from Lemma 8.1.2 this also implies that the portfolio (
t
,
t
)
t[0,T]
is
self-nancing.
The above proposition shows that there always exists a hedging strategy
starting from
V
0
= IE[F] exp
_

_
T
0
r
u
du
_
.
Conversely, since there exists a hedging strategy leading to

V
T
= F exp
_

_
T
0
r
u
du
_
,
then by (8.1.5), (

V
t
)
t[0,T]
is necessarily a martingale with initial value

V
0
= IE[

V
T
] = IE[F] exp
_

_
T
0
r
u
du
_
.
We now consider the hedging of European call option with payo F =
(S
T
K)
+
using the Clark formula in the setting of deterministic struc-
ture equations. In this case the next proposition allows us to compute the
hedging strategy appearing in (8.2.1).
Proposition 8.2.2. Assume that
t
0, t [0, T]. Then for 0 t T we
have
IE[D
t
(S
T
K)
+
[T
t
] = IE
_
i
t

t
S
x
t,T
1
[K,)
(S
x
t,T
)
+
j
t

t
(
t

t
S
x
t,T
(K S
x
t,T
)
+
)1
[
K
1+
t
,)
(S
x
t,T
)
_
x=S
t
.
Proof. By Lemma 4.6.2, using Denition 4.6.1 and Relation (4.6.4) we have,
for any F o,
D
t
F = D
B
t
F +
j
t

t
(T

t
F F), t [0, T]. (8.2.4)
We have T

t
S
T
= (1 +
t

t
)S
T
, t [0, T], and the chain rule D
B
f(F) =
f

(F)D
B
F, cf. Relation (5.2.1), holds for F o and f (
2
b
(R). Since o
is an algebra for deterministic (
t
)
t[0,T]
, we may approach x (x K)
+
by polynomials on compact intervals and proceed e.g. as in [97], p. 5-13. By
dominated convergence, F = (S
T
K)
+
Dom(D) and (8.2.4) becomes
D
t
(S
T
K)
+
= i
t

t
S
T
1
[K,)
(S
T
) +
j
t

t
((1 +
t

t
)S
T
K)
+
(S
T
K)
+
),
0 t T. The Markov property of (S
t
)
t[0,T]
implies
IE
_
D
B
t
(S
T
K)
+
[T
t

= i
t

t
IE
_
S
x
t,T
1
[K,)
(S
x
t,T
)

x=S
t
,
286 8 Option Hedging in Continuous Time
and
j
t

t
IE[(T

t
S
T
K)
+
(S
T
K)
+
[T
t
]
=
j
t

t
IE
_
((1 +
t

t
)S
x
t,T
K)
+
(S
x
t,T
K)
+

x=S
t
=
j
t

t
IE
_
((1 +
t

t
)S
x
t,T
K)1
[
K
1+
t

t
,)
(S
x
t,T
)
_
x=S
t

j
t

t
IE
_
(S
x
t,T
K)
+
1
[K,)
(S
x
t,T
)

x=S
t
=
j
t

t
IE
_

t
S
x
t,T
1
[
K
1+
t

t
,)
(S
x
t,T
) + (S
x
t,T
K)1
[
K
1+
t

t
,K]
(S
x
t,T
)
_
x=S
t
=
j
t

t
IE
_

t
S
x
t,T
1
[
K
1+
t

t
,)
(S
x
t,T
) (K S
x
t,T
)
+
1
[
K
1+
t

t
,]
(S
x
t,T
)
_
x=S
t
=
j
t

t
IE
_
(
t

t
S
x
t,T
(K S
x
t,T
)
+
)1
[
K
1+
t

t
,)
(S
x
t,T
)
_
x=S
t
.

If (
t
)
t[0,T]
is not constrained to be positive then
IE[D
t
(S
T
K)
+
[T
t
] = i
t

t
IE
_
S
x
t,T
1
[K,)
(S
x
t,T
)

x=S
t
+
j
t

t
IE
_

t
S
x
t,T
1
[
K
1+
t

t
,
)
(S
x
t,T
) + (S
x
t,T
K)1
[
K
1+
t

t
,K
]
(S
x
t,T
)
_
x=S
t
,
with the convention 1
[b,a]
= 1
[a,b]
, 0 a < b T. Proposition 8.2.2 can
also be proved using Lemma 3.7.2 and the It o formula (2.12.4).
In the sequel we assume that (
t
)
tR
+
is deterministic and
dM
t
= i
t
dB
t
+
t
(dN
t

t
dt), t R
+
, M
0
= 0,
as in Relation (2.10.4).
Next we compute
exp
_

_
T
0
r
s
ds
_
IE
_
(S
T
K)
+

in terms of the Black-Scholes function

BS(x, T; r,
2
; K) = e
rT
IE[(xe
rT
2
T/2+W
T
K)
+
],
where W
T
is a centered Gaussian random variable with variance T.
8.2 Hedging by the Clark Formula 287
Proposition 8.2.3. The expectation
exp
_

_
T
0
r
s
ds
_
IE
_
(S
T
K)
+

can be computed as
exp
_

_
T
0
r
s
ds
_
IE
_
(S
T
K)
+

= exp (
0
(T))

k=0
1
k!
_
T
0

_
T
0
BS
_
S
0
exp
_

_
T
0

s
ds
_
i=k

i=1
(1 +
t
i

t
i
) , T; R
T
,

0
(T)
T
; K
_

t
1

t
k
dt
1
dt
k
.
Proof. Similarly to Proposition 3.7.3 we have
IE
_
e
TR
T
(S
T
K)
+

k=0
IE
_
e
TR
T
(S
T
K)
+
[N
T
= k

P(N
T
= k),
with
P(N
T
= k) = exp(
0
(T))
(
0
(T))
k
k!
, k N.
Conditionally to N
T
= k, the jump times (T
1
, . . . , T
k
) have the law
k!
(
0
(T))
k
1
{0<t
1
<<t
k
<T}

t
1

t
k
dt
1
dt
k
,
since the process (N

1
0
(T)
t
)
tR
+
is a standard Poisson process. Hence, con-
ditionally to
N(
1
0
(
0
(T))) = k = N
T
= k,
its jump times (
0
(T
1
), . . . ,
0
(T
k
)) have a uniform law on [0,
0
(T)]
k
. We
then use the fact that (

B
t
)
tR
+
and (N
t
)
tR
+
are also independent under P
since (r
t
)
tR
+
is deterministic, and the identity in law
S
T
law
= S
0
X
T
exp
_

_
T
0

s
ds
_
k=N
T

k=1
(1 +
T
k

T
k
) ,
288 8 Option Hedging in Continuous Time
where
X
T
= exp
_
TR
T

0
(T)/2 +
_

0
(T)
T
_
1/2
W
T
_
,
and W
T
is independent of N.
8.3 Black-Scholes PDE
As in the standard Black-Scholes model, it is possible to determine the
hedging strategy in terms of the Delta of the price in the case (r
t
)
tR
+
is
deterministic.
Let the function C(t, x) be dened by
C(t, S
t
) = V
t
= exp
_

_
T
t
r
u
du
_
IE[(S
T
K)
+
[ T
t
]
= exp
_

_
T
t
r
u
du
_
IE[(S
T
K)
+
[ S
t
], t R
+
.
cf. (8.2.3). An application of the It o formula leads to
dC(t, S
t
) =
_
C
t
+r
t
S
t
C
x
+
1
2

2
C
x
2
i
t

2
t
S
2
t

2
t
+
t
C
_
(t, S
t
)dt
+S
t

t
C
x
(t, S
t
)dM
t
+(C(t, S
t
(1 +
t

t
))C(t, S
t
)) (dN
t

t
dt)
(8.3.1)
where
C(t, S
t
) = C(t, S
t
(1 +
t

t
)) C(t, S
t
)
C
x
(t, S
t
)S
t

t
.
The process

C
t
:= C(t, S
t
) exp
_

_
t
0
r
s
ds
_
= exp
_

_
T
0
r
u
du
_
IE[(S
T
K)
+
[ T
t
]
=

V
t
8.3 Black-Scholes PDE 289
is a martingale from Propositions 2.5.8 and 8.2.1-(ii), with
d

C
t
=
t
d

S
t
(8.3.2)
=
t

t
S
t
dM
t
= i
t

t
S
t
dB
t
+
t

t
S
t
(dN
t

t
dt),
from Lemma 8.1.2. Therefore, by identication of (8.3.1) and (8.3.2),

r
t
C(t, S
t
) =
_
C
t
+ r
t
S
t
C
x
+
1
2
i
t

2
t
S
2
t

2
t

2
C
x
2
+
t
C
_
(t, S
t
),

t
S
t
dM
t
= S
t

t
C
x
(t, S
t
)dM
t
+(C(t, S
t
(1 +
t

t
))C(t, S
t
)) (dN
t

t
dt) .
Therefore, by identication of the Brownian and Poisson parts,

i
t

t
S
t

t
= i
t
S
t

t
C
x
(t, S
t
)
j
t

t
S
t

t
= C(t, S
t
(1 +
t

t
)) C(t, S
t
).
(8.3.3)
The term C(t, S
t
) vanishes on the set
t R
+
:
t
= 0 = t : i(t) = 1.
Therefore, (8.3.3) reduces to

t
=
C
x
(t, S
t
),
i.e. the process (
t
)
tR
+
is equal to the usual Delta (8.3) on t R
+
: i
t
= 1,
and to

t
=
C(t, S
t
(1 +
t

t
)) C(t, S
t
)
S
t

t
on the set t R
+
: i
t
= 0.
Proposition 8.3.1. The Black-Scholes PDE for the price of a European call
option is written as
C
t
(t, x) +r
t
x
C
x
(t, x) +
1
2

2
t
x
2

2
t

2
C
x
2
(t, x) = r
t
C(t, x),
on t :
t
= 0, and as
C
t
(t, x) +r
t
x
C
x
(t, x) +
t
C(t, x) = r
t
C(t, x),
on the set t R
+
:
t
,= 0, under the terminal condition C(T, x) =
(x K)
+
.
290 8 Option Hedging in Continuous Time
8.4 Asian Options and Deterministic Structure
The price at time t of an Asian option is dened as
IE

_
T
t
r
s
ds
_
1
T
_
T
0
S
u
du K
_
+

T
t

.
The next proposition provides a replicating hedging strategy for Asian
options in the case of a deterministic structure equation. Following [74],
page 91, and [13], we dene the auxiliary process
Y
t
=
1
S
t
_
1
T
_
t
0
S
u
du K
_
, t [0, T]. (8.4.1)
Proposition 8.4.1. There exists a measurable function

C on R
+
R such
that

C(t, ) is (
1
for all t R
+
, and
S
t

C(t, Y
t
) = IE

_
1
T
_
T
0
S
u
du K
_
+

T
t

.
Moreover, the replicating portfolio for an Asian option with payo
_
1
T
_
T
0
S
u
du K
_
+
is given by (8.1.3) and

t
=
1

t
e

_
T
t
r
s
ds
_

C(t, Y
t
)
t
(8.4.2)
+(1 +
t

t
)
_
j
t

t
_

C
_
t,
Y
t
1 +
t

t
_

C(t, Y
t
)
_
i
t

t
Y
t

2

C(t, Y
t
)
__
.
Proof. With the above notation, the price at time t of the Asian option
becomes
IE
_
e

_
T
t
r
s
ds
S
T
(Y
T
)
+

T
t
_
.
For 0 s t T, we have
d (S
t
Y
t
) =
1
T
d
__
t
0
S
u
du K
_
=
S
t
T
dt,
8.4 Asian Options and Deterministic Structure 291
hence
S
t
Y
t
S
s
= Y
s
+
1
T
_
t
s
S
u
S
s
du.
Since S
u
/S
t
is independent of S
t
by (8.1.2), we have, for any suciently
integrable payo function H,
IE
_
H (S
T
Y
T
)

T
t
_
= IE
_
H
_
S
t
Y
t
+
1
T
_
T
t
S
u
du
_

T
t
_
= IE
_
H
_
xy +
x
T
_
T
t
S
u
S
t
du
__
y=Y
t
, x=S
t
.
Let C (
2
b
(R
+
R
2
) be dened as
C(t, x, y) = IE
_
H
_
xy +
x
T
_
T
t
S
u
S
t
du
__
,
i.e.
C(t, S
t
, Y
t
) = IE
_
H (S
T
Y
T
)

T
t
_
.
When H(x) = max(x, 0), since for any t [0, T], S
t
is positive and
T
t
-measurable, and S
u
/S
t
is independent of T
t
, u t, we have:
IE
_
H (S
T
Y
T
)

T
t
_
= IE
_
S
T
(Y
T
)
+

T
t
_
= S
t
IE
_
_
Y
T
S
T
S
t
_
+

T
t
_
= S
t
IE

_
Y
t
+
1
T
_
T
t
S
u
S
t
du
_
+

T
t

= S
t
IE

_
y +
1
T
_
T
t
S
u
S
t
du
_
+

y=Y
t
= S
t

C(t, Y
t
),
with

C(t, y) = IE

_
y +
1
T
_
T
t
S
u
S
t
du
_
+

.
We now proceed as in [13], which deals with the sum of a Brownian motion
and a Poisson process. From the expression of 1/S
t
given by (8.1.2) we have
292 8 Option Hedging in Continuous Time
d
_
1
S
t
_
=
1
S
t

__
r
t
+

2
t

2
t
1 +
t

t
_
dt

t
1 +
t

t
dM
t
_
,
hence by (2.12.4), It os formula and the denition (8.4.1) of Y
t
, we have
dY
t
= Y
t
_
r
t
+

2
t

2
t
1 +
t

t
_
dt +
1
T
dt
Y
t

t
1 +
t

t
dM
t
.
Assuming that H (
2
b
(R) and applying Lemma 3.7.2 we get
IE
_
D
t
H (S
T
Y
T
)

T
t
_
= L
t
C(t, S
t
, Y
t
)
= i
t
_

t
S
t

2
C(t, S
t
, Y
t
)
Y
t

t
1 +
t

3
C(t, S
t
, Y
t
)
_
(8.4.3)
+
j
t

t
_
C
_
t, S
t
+
t
S
t
, Y
t

Y
t

t
1 +
t

t
_
C(t, S
t
, Y
t
)
_
,
where L
t
is given by (2.12.5). Next, given a family (H
n
)
nN
of (
2
b
functions,
such that [H
n
(x)[ x
+
and [H

n
(x)[ 2, x R, n N, and converging
pointwise to x x
+
, by dominated convergence (8.4.3) holds for C(t, x, y) =
x

C(t, y) and we obtain:

IE

D
t
_
1
T
_
T
0
S
u
du K
_
+

T
t

= i
t

C(t, Y
t
)
t
S
t
+S
t
_
j
t

t
_

C
_
t,
Y
t
1 +
t

t
_

C(t, Y
t
)
_
i
t

t
Y
t

2

C(t, Y
t
)
_
+S
t

t
_
j
t

t
_

C
_
t,
Y
t
1 +
t

t
_

C(t, Y
t
)
_
i
t

t
Y
t

2

C(t, Y
t
)
_
.

As a particular case we consider the Brownian motion model, i.e.

t
= 0, for
all t [0, T], so i
t
= 1, j
t
= 0 for all t [0, T], and we are in the Brownian
motion model. In this case we have

t
= e

_
T
t
r
s
ds
_
Y
t

2

C(t, Y
t
) +

C(t, Y
t
)
_
= e

_
T
t
r
s
ds
_
S
t

C
_
t,
1
x
_
1
T
_
t
0
S
u
du K
__
|x=S
t
+

C(t, Y
t
)
_
=

x
_
xe

_
T
t
r
s
ds

C
_
t,
1
x
_
1
T
_
t
0
S
u
du K
___
|x=S
t
, t [0, T],
which can be denoted informally as a partial derivative with respect to S
t
.
8.5 Notes and References 293
8.5 Notes and References
See e.g. [74] and [135] for standard references on stochastic nance, and [97]
for a presentation of the Malliavin calculus applied to continuous markets.
The use of normal martingales in nancial modelling has been rst considered
in [35]. The material on Asian options is based on [70] and [12]. Hedging
strategies for Lookback options have been computed in [15] using the Clark-
Ocone formula.