Stochastic Calculus for Finance II

-
some Solutions to Chapter VII
Matthias Thul

January 4, 2012
Exercise 7.1 (Black-Scholes-Merton Equation for the up-and-out
Call)
(i) We have

p
m(, s) =

t

_
1

_
ln s +
_
r
1
2

2
_

__
=
_

ln s
2

+
r
1
2

2
2

_
=
1
2

_
ln
_
1
s
_
+
_
r
1
2

2
_

_
=
1
2

_
,
1
s
_
(q.e.d.).
(ii) We rst compute

p
m(, s) =

s
_
1

_
ln s +
_
r
1
2

2
_

__
=
1
s

.
Consequently for s = x/c, we get

The author can be contacted via <<firstname>>.<<lastname>>@googlemail.com and

http://www.matthiasthul.com.
1

_
,
x
c
_
=

s

(, s)
s
x
=
1
x

and for s = x/c we have

_
,
x
c
_
=

s

(, s)
s
x
=
1
x

(q.e.d.).
(iii) Since
N

(, s)) =
1

2
exp
_

(, s)
2
_
,
we have
N

(
+
(, s))
N

(, s))
= exp
_
1
2
_

(, s)
2
+
(, s)
_
_
= exp
_
1
2
(

(, s)
+
(, s)) (

(, s) +
+
(, s))
_
.
Here,

(, s)
+
(, s) =

and

(, s) +
+
(, s) =
2 ln s + 2r

.
Thus,
N

(
+
(, s))
N

(, s))
= exp {(ln s + r)}
=
e
r
s
(q.e.d.)
2
and
e
r
N

(d

(, s)) = sN

(d
+
(, s)) (q.e.d.).
(iv) This result is immediately obvious from the denition of

(, s) and has been used

in (iv) already. We have

+
(, s)

(, s) =

2

(q.e.d.).
(v) Again, this result follows immediately from the denition of

(, s). We have

(, s)

_
, s
1
_
=
ln s ln s
1

=
2 ln s

(q.e.d.).
(vi) We have
N

(y) =

y
1

2
exp
_

y
2
2
_
=
y

2
exp
_

y
2
2
_
= yN

(y) (q.e.d.).
(vii) To be continued...
Exercise 7.3 (Markov Property for Geometric Brownian Motion
and its Maximum to Date)
The crucial steps for the solution to this problem have been derived in Section 7.4.4. First,
remember that
S(t) = S(0) exp
_

W(t)
_
,
3
where

W(t) is a drifted Brownian motion. The maximum-to-date process is given by
Y (t) = S(0) exp
_

M(t)
_
,
where

M(t) = max
0ut

W(u).
First notice that
S(T) = S(t)
S(T)
S(t)
= S(t) exp
_

_

W(T)

W(t)
__
,
where S(t) is F(t)-measurable and the increment

W(T)

W(t) is independent of the
-algebra F(t) since by Denition 3.3.1 Brownian motion has independent increments.
Note that adding a drift to the Brownian motion does not change this property. Similarly,
Y (T) = Y (t)
Y (T)
Y (t)
= Y (t) exp
_

_

M(T)

M(t)
__
= Y (t) exp
_

_
max
tuT

W(u)

M(t)
_
+
_
= Y (t) exp
_

_
max
tuT
_

W(u)

W(t)
_

_

M(t)

W(t)
_
_
+
_
= Y (t) exp
_
_
max
tuT

_

W(u)

W(t)
_
ln
_
Y (t)
S(t)
__
+
_
.
Again, ln (Y (t)/S(t)) is F(t)-measurable while
B(t, T) = max
tuT

_

W(u)

W(t)
_
is independent of the ltration F(t). Thus,
E[ f(S(T), Y (T))| F(t)] = E
_
h
_
S(t), Y (t),
S(T)
S(t)
, B(t, T)
_

F(t)
_
,
4
where
h(S(t), Y (t), A(t, T), B(t, T)} = f
_
S(t)A(t, T), Y (t) exp
_
_
B(t, T) ln
_
Y (t)
S(t)
__
+
__
.
By Lemma 2.3.4, there exists a function g(S(t), Y (t)) dened by
g(x, y) = E
_
h
_
x, y,
S(T)
S(t)
, B(t, T)
__
such that
E[ f(S(T), Y (T))| F(t)] = g(S(t), Y (t)) (q.e.d.).
Exercise 7.4 (Cross Variation of Geometric Brownian Motion and
its Maximum to Date)
We have

j=1
(Y (t
j
) Y (t
j1
)) (S (t
j
) S (t
j1
))

j=1
|Y (t
j
) Y (t
j1
)| |S (t
j
) S (t
j1
)|
max
1jm
|S (t
j
) S (t
j1
)|
m

j=1
Y (t
j
) Y (t
j1
)
= max
1jm
|S (t
j
) S (t
j1
)| (Y (T) Y (0)) .
In the limit as maximum step size goes to zero, we get
lim
||||0
max
1jm
|S (t
j
) S (t
j1
)| (Y (T) Y (0)) = 0,
since the stock price S(t) is a continuous function of time. Consequently,
lim
||||0
m

j=1
(Y (t
j
) Y (t
j1
)) (S (t
j
) S (t
j1
)) = 0 (q.e.d.).
Exercise 7.7 (Zero-Strike Asian Call)
(i) We can split the integral in a F(t)-measurable part and a part independent of F(t)
to get
5
E
Q
__
T
0
S(u)du

F(t)
_
= E
Q
__
t
0
S(u)du

F(t)
_
+E
Q
__
T
t
S(u)du

F(t)
_
=
_
t
0
S(u)du +
_
T
t
E
Q
[S(u)] du.
Using the martingale property of the discounted stock price we can replace E
Q
[S(u)]
with e
r(tu)
S(t) which yields
. . . =
_
t
0
S(u)du + S(t)
_
T
t
e
r(tu)
du
=
_
t
0
S(u)du +
S(t)
r
_
e
r(Tt)
1
_
.
It follows that
e
r(Tt)
E
Q
_
1
T
_
T
0
S(u)du

F(t)
_
= e
r(Tt)
1
T
_
t
0
S(u)du +
S(t)
rT
_
1 e
r(Tt)
_
and
v(t, x, y) = y
e
r(Tt)
T
+ x
1 e
r(Tt)
rT
.
(ii) The partial derivatives of v(t, x, y) are given by
v
t
= (ry x)
e
r(Tt)
T
,
v
x
=
1 e
r(Tt)
rT
,
v
y
=
e
r(Tt)
T
.
Substituting these into the PDE in Equation (7.5.8) yields
(ry x)
e
r(Tt)
T
+ x
1 e
r(Tt)
T
+ x
e
r(Tt)
T
= ry
e
r(Tt)
T
+ x
1 e
r(Tt)
T
.
All terms cancel out and the assertion follows. We further have
v(t, 0, y) = y
e
r(Tt)
T
v(T, x, y) =
y
T
6
These are just the boundary conditions in Equations (7.5.9) and (7.5.11) for K = 0.
Note that both expressions are non-negative since y is an integral over the non-
negative random variable S(t).
(iii) By the proof of Theorem (7.5.1) and Remark (7.5.2), the hedge ratio is given by the
rst derivative of the option price w.r.t. to the spot price.
(t) =
1 e
r(Tt)
rT
.
This quantity is non-random, since it only depends on time but not on the current
value of S(t) or its history.
(iv) The dierential of the discounted portfolio value is
d
_
e
rt
X(t)
_
= (t)e
rt
S(t)dW(t).
Using the process for (t) computed above, we get
d
_
e
rt
X(t)
_
=
e
rt
e
rT
rT
S(t)dW(t).
The dierential of the option value is
dv(t, S(t), Y (t)) =
v
t
dt +
v
x
dS(t) +
v
y
dY (t)
= (rY (t) S(t))
e
r(Tt)
T
dt +
1 e
r(Tt)
rT
dS(t) +
e
r(Tt)
T
S(t)dt
= rY (t)
e
r(Tt)
T
dt +
1 e
r(Tt)
rT
dS(t)
= rv(t, S(t), Y (t))dt +
1 e
r(Tt)
rT
S(t)dW
t
.
By the product rule, we obtain the dierential of the discounted option value
d
_
e
rt
v(t, S(t), Y (t))
_
= re
rt
v(t, S(t), Y (t))dt + e
rt
dv(t, S(t), Y (t))
=
e
rt
e
rT
rT
S(t)dW(t).
7
Since the dierential of the discounted portfolio value and the discounted option
value agree, it follows that if we start with a portfolio that has initial value X(0) =
v(0, S(0), Y (0)) and hold (t) shares of the asset at each point in time, then the
terminal payos agree almost surely.
8