Stochastic Calculus for Finance II

-
some Solutions to Chapter VI
Matthias Thul

January 3, 2012
Exercise 6.1
(i) Let
A(u) =
_
u
t
(v)dW(v) +
_
u
t
_
b(v)
1
2

2
(v)
_
dv
such that Z(u) = exp {A(u)}. For u = t, both integrals evaluate to zero and thus
A(t) = 0 and Z(t) = 1. Let f(u, x) = e
x
with
f
u
= 0,
f
x
= e
x
,

2
f
x
2
= e
x
.
Applying Itos lemma yields for u t
dZ(u) = df(u, A(t))
= Z(u)dA(u) +
1
2
Z(u)dA(u)dA(u)
=
_
b(u)
1
2

2
(u)
_
Z(u)du + (u)Z(u)dW(u) +
1
2

2
(u)Z(u)du
= b(u)Z(u)du + (u)Z(u)dW(u) (q.e.d.).
(ii) By the Ito product rule, we have

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

http://www.matthiasthul.com.
1
dX(u) = d (Y (u)Z(u))
= Y (u)dZ(u) + Z(u)dY (u) + dY (u)dZ(u)
= b(u)X(u)du + (u)X(u)dW(u) + (a(u) (u)(u)) du + (u)dW(u) + (u)(u)du
= (a(u) + b(u)X(u)) du + ((u) + (u)X(u)) dW(u) (q.e.d.).
Finally, using the previous result for Y (t) and Z(t) we see that X(t) = Y (t)Z(t) = x.
Exercise 6.2 (No-Arbitrage Derivation of Bond-Pricing Equation)
(i) The self-nancing portfolio process X(t) is given by
dX(t) =
1
(t)df (t, R(t), T
1
) +
2
(t)df (t, R(t), T
2
)
+R(t) (X(t)
1
(t)f (t, R(t), T
1
)
2
(t)f (t, R(t), T
2
)) dt.
Thus, by the It o product rule,
d(D(t)X(t)) = D(t)dX(t) + X(t)dD(t) + dD(t)dX(t)
=
1
(t)D(t)df (t, R(t), T
1
) +
2
(t)D(t)df (t, R(t), T
2
)
+R(t)D(t) (X(t)
1
(t)f (t, R(t), T
1
)
2
(t)f (t, R(t), T
2
)) dt
R(t)D(t)X(t)dt
=
1
(t)D(t) [R(t)f (t, R(t), T
1
) dt + df (t, R(t), T
1
)]
+
2
(t)D(t) [R(t)f (t, R(t), T
2
) dt + df (t, R(t), T
2
)] .
Here, we used that
dD(t) = R(t)D(t)dt
has zero quadratic variation such that the cross-variation term in the It o dierential
drops out. Assuming that the zero-coupon bond price is smooth enough such that
the Ito formula can be applied, then its dierential is given by
2
df(t, R(t), T) = f
t
(t, R(t), T)dt + f
r
(t, R(t), T)dR(t) +
1
2
f
rr
(t, R(t), T)dR(t)dR(t)
= f
t
(t, R(t), T)dt + (t, R(t))f
r
(t, R(t), T)dt + (t, R(t))f
r
(t, R(t), T)dW(t)
+
1
2

2
(t, R(t))f
rr
(t, R(t), T)dt.
Combining this with the previously obtained dynamics of the discounted portfolio
value yields
d(D(t)X(t)) =
1
(t)D(t) [R(t)f (t, R(t), T
1
) + f
t
(t, R(t), T
1
) + (t, R(t))f
r
(t, R(t), T
1
)
+
1
2

2
(t, R(t))f
rr
(t, R(t), T
1
)
_
dt +
2
D(t) [R(t)f (t, R(t), T
2
)
+f
t
(t, R(t), T
2
) + (t, R(t))f
r
(t, R(t), T
2
) +
1
2

2
(t, R(t))f
rr
(t, R(t), T
2
)
_
dt
+D(t)(t, R(t)) [
1
(t)f
r
(t, R(t), T
1
) +
2
(t)f
r
(t, R(t), T
2
)] dW(t).
This is the rst equality in Equation (6.9.4). To get the second equality, we simply
substitute (t, R(t), T) and obtain
d(D(t)X(t)) =
1
(t)D(t) [(t, R(t)) (t, R(t), T
1
)] f
r
(t, R(t), T
1
) dt
+
2
(t)D(t) [(t, R(t) (t, R(t), T
2
)] f
r
(t, R(t), T
2
) dt
+D(t)(t, R(t)) [
1
(t)f
r
(t, R(t), T
1
) +
2
(t)f
r
(t, R(t), T
2
)] dW(t) (q.e.d.).
(ii) We rst note that by construction of the trading strategy

1
(t)f
r
(t, R(t), T
1
) +
2
(t)f
2
(t, R(t), T
2
)
= S(t)f
r
(t, R(t), T
1
) f
r
(r, R(t), T
2
) S(t)f
r
(t, R(t), T
1
) f
r
(r, R(t), T
2
)
= 0.
Thus, the diusion term in the dynamics of D(t)X(t) vanishes and the portfolio is
instantaneously risk-free. Furthermore,
3

1
(t)D(t) [(t, R(t)) (t, R(t), T
1
)] f
r
(t, R(t), T
1
)
+
2
(t)D(t) [(t, R(t) (t, R(t), T
2
)] f
r
(t, R(t), T
2
)
= S(t)D(t) [ (t, R(t), T
2
) (t, R(t), T
1
)] f
r
(t, R(t), T
1
) f
r
(t, R(t), T
2
)
0.
The last inequality follows from the denition of S(t). For no-arbitrage to exist,
the discounted wealth process of a risk-free portfolio has to be a martingale and
thus, the drift has to vanish. This is only the case if (t, R(t), T
2
) = (t, R(t), T
1
).
Since T
1
and T
2
are arbitrary maturities, we conclude that (t, R(t), T) has to be
independent of T.
(iii) The discounted portfolio process immediately follows from the result obtained in (i)
by setting T
1
= T,
1
(t) = (t) and
2
(t) = 0 for all t 0. Then
d(D(t)X(t)) = (t)D(t) [R(t)f(t, R(t), T) + f
t
(t, R(t), T) + (t, R(t))f
r
(t, R(t), T)
+
1
2

2
(t, R(t))f
rr
(t, R(t), T)
_
dt + D(t)(t, R(t))(t)f
r
(t, R(t), T)dW(t).
If f
r
(t, R(t), T) = 0, then the diusion term vanishes. For no-arbitrage to exist, the
change in the discounted portfolio value must be zero as well. Otherwise a risk-free
prot could be made by taking a long or short position in the risk-free portfolio.
Consequently,
R(t)f(t, R(t), T) + f
t
(t, R(t), T) +
1
2

2
(t, R(t))f
rr
(t, R(t), T) = 0 (q.e.d.).
Exercise 6.6 (Moment-Generating Function for Cox-Ingersoll-Ross
Process)
(i) Let f(t, x) = e
1
2
bt
x where
f
t
=
1
2
be
1
2
bt
x,
f
x
= e
1
2
bt
,

2
f
x
2
= 0.
4
When applying It os formula to compute the dierential of f(t, X
j
(t)), then the
geometric drift drops out and we obtain
d
_
e
1
2
bt
X
j
(t)
_
= be
1
2
bt
X
j
(t)
1
2
be
1
2
bt
X
j
(t)dt +
1
2
e
1
2
bt
dW
j
(t)
=
1
2
e
1
2
bt
dW
j
(t).
Integrating over [0, t] yields
e
1
2
bt
X
j
(t) = X
j
(0) +
1
2

_
t
0
e
1
2
bu
dW
j
(u)
and thus
X
j
(t) = e

1
2
bt
_
X
j
(0) +
1
2

_
t
0
e
1
2
bu
dW
j
(u)
_
(q.e.d.).
The expected value is
E[X
j
(t)] = e

1
2
bt
X
j
(0) +
1
2
e

1
2
bt
E
__
t
0
e
1
2
bu
dW
j
(u)
_
= e

1
2
bt
X
j
(0),
where we used that by Theorem 4.3.1, the Ito integral is a martingale starting at
zero. The variance is
Var [X
j
(t)] = Var
_
1
2
e

1
2
bt
_
t
0
e
1
2
bu
dW
j
(u)
_
=
1
4

2
e
bt
Var
__
t
0
e
1
2
bu
dW
j
(u)
_
=
1
4

2
e
bt
E
_
__
t
0
e
1
2
bu
dW
j
(u)
_
2
_
=
1
4

2
e
bt
_
t
0
e
bu
du
=
1
4b

2
e
bt
e
bu

u=t
u=0
=
1
4b

2
_
1 e
bt
_
.
5
In the rst equality, we used that the term e

1
2
bt
X
j
(0) is a constant and thus does
not contribute to the variance of X
j
(t). In the third equality we again used that the
It o integral is a martingale starting at zero and fourth equality is a consequence of
the It o isometry (Theorem 4.3.1). Finally, by Theorem 4.4.9, the Ito integral of a
deterministic integrand is normally distributed and we conclude that
X
j
(t) N
_
e

1
2
bt
X
j
(0),
1
4b

2
_
1 e
bt
_
_
(q.e.d.).
(ii) Let f (t, x
1
, x
2
, . . . , x
d
) =

d
j=1
x
2
j
where
f
t
= 0,
f
x
j
= 2x
j
,

2
f
x
2
j
= 2,

2
f
x
j
, x
k
= 0.
An application of the multidimensional Ito formula yields
dR(t) = df (t, X
1
(t), X
2
(t), . . . , X
d
(t))
=
d

j=1
2X
j
(t)dX
j
(t) +
d

j=1
dX
j
(t)dX
j
(t)
=
d

j=1
__
1
4

2
bX
2
j
(t)
_
dt + X
j
(t)dW
j
(t)
_
=
_
1
4

2
d b
d

j=1
X
2
j
(t)
_
dt +
d

j=1
X
j
(t)dW
j
(t)
= (a bR(t))dt +
d

j=1
X
j
(t)dW
j
(t).
Next, we show that the process
B(t) =
d

j=1
_
t
0
X
j
(s)
_
R(s)
dW
j
(s)
is a P-Brownian motion. First observe that by Theorem 4.3.1, each of the Ito
integrals in the denition of B(t) starts at zero, is a martingale and has continuous
sample paths. As a sum over these It o integrals, the process B(t) inherits these
properties. Furthermore,
6
dB(t)dB(t) =
1
R(t)
d

j=1
X
2
j
(t)dt
= dt,
where we used that the Brownian motions W
1
(t), W
2
(t), . . . , W
d
(t) are independent.
By Levys characterization of the Brownian motion (Theorem 4.6.4), it follows that
B(t) is a P-Brownian motion. Consequently,
dR(t) = (a bR(t))dt +
_
R(t)dB(t) (q.e.d.).
(iii) The random variables X
j
(t) and X
k
(t) are independent for j = k since the only
source of randomness in the denition of X
j
(t) is W
j
(t) for any j and the Brownian
motions W
j
(t) and W
k
(t) are independent for j = k. The mean and standard
deviation are obtained by substituting for X
j
(0) in the result obtained in part (i).
We get
X
j
(t) N ((t), v(t)) ,
where
(t) = e

1
2
bt
_
R(0)
d
, v(t) =
1
4b

2
_
1 e
bt
_
(q.e.d.).
(iv) Following the hint and using that X
j
(t) is normally distributed, we get
7
E
_
exp
_
uX
2
j
(t)
_
=
_

exp
_
ux
2
_
1
_
2v(t)
exp
_

(x (t))
2
2v(t)
_
dx
=
_

exp
_
ux
2
_
1
_
2v(t)
exp
_

x
2
2x(t) +
2
(t)
2v(t)
_
dx
=
_

1
_
2v(t)
exp
_

x
2
(1 2uv(t)) 2x(t) +
2
(t)
2v(t)
_
dx
= exp
_

2
(t)
2v(t)
_
_

1
_
2v(t)
exp
_

x
2
(1 2uv(t)) 2x(t)
2
(t)/(1 2uv(t))
2
2v(t)
_
dx
= exp
_

2
(t)(1 1/(1 2uv(t)))
2v(t)
_
_

1
_
2v(t)
exp
_

_
x
_
1 2uv(t) (t)/
_
1 2uv(t)
_
2
2v(t)
_

_
dx.
= exp
_

2
(t)u
1 2uv(t)
_
_

1
_
2v(t)
exp
_

(1 2uv(t)) (x (t)/(1 2uv(t)))

2
2v(t)
_
dx.
=
1
_
1 2uv(t)
exp
_

2
(t)u
1 2uv(t)
_
_

1 2uv(t)
2v(t)
exp
_

(x (t)/(1 2uv(t)))
2
2v(t)/(1 2uv(t))
_
dx.
The integrand is the density of a normal random variable with distribution
N
_
(t)
1 2uv(t)
,
v(t)
1 2uv(t)
_
and thus the integral evaluates to one. We get
E
_
exp
_
uX
2
j
(t)
_
=
1
_
1 2uv(t)
exp
_

2
(t)u
1 2uv(t)
_
(q.e.d.).
This expression is only well dened if
1 2uv(t) > 0 u <
1
2v(t)
.
8
(v) We have
E[exp {uR(t)}] = E
_
exp
_
u
d

i=1
X
j
(t)
__
= E
_
d

j=1
exp {uX
j
(t)}
_
=
d

j=1
E[exp {uX
j
(t)}]
= (E[exp {uX
j
(t)}])
d
=
_
1
1 2uv(t)
_
d/2
exp
_
d
2
(t)u
1 2uv(t)
_
,
where we used that the X
j
(t) are independent and identically distributed in the
third and fourth equality. Alternatively, we could have directly applied Theorem
2.2.7 which states that the joint moment generating function factors for sums of
independent random variables. Now substituting for (t) and d yields
E[exp {uR(t)}] =
_
1
1 2uv(t)
_
2a/
2
exp
_
e

1
2
bt
R(0)u
1 2uv(t)
_
(q.e.d.).
Exercise 6.8 (Kolmogorov Backward Equation)
Let h(y) be a Borel measurable function and dene
g(t, x) = E
t,x
[ h(X(T))| F(s)] =
_

0
h(y)p(t, T, x, y)dy.
The lower bound at zero is due to the assumption of X(t) being strictly positive. By
Lemma 6.4.2, g(t, X(t)) is a martingale. Its partial derivatives are
g
t
(t, x) =
_

0
h(y)p
t
(t, T, x, y)dy
g
x
(t, x) =
_

0
h(y)p
x
(t, T, x, y)dy
g
xx
(t, x) =
_

0
h(y)p
xx
(t, T, x, y)dy
The dierential of g(t, x) is
9
dg(t, X(t)) = g
t
(t, X(t))dt + g
x
(t, X(t))dX(t) +
1
2
g
xx
(t, X(t))(dX(t))
2
= g
t
(t, X(t))dt + (t, X(t))g
x
(t, X(t))dt + (t, X(t))g
x
(t, X(t))dW(t)
+
1
2

2
(t, X(t))g
xx
(t, X(t))dt
Since g(t, X(t)) is a martingale, the drift term has to be equal to zero
g
t
(t, X(t)) + (t, X(t))g
x
(t, X(t)) +
1
2

2
(t, X(t))g
xx
(t, X(t)) = 0
Substituting the derivatives yields
0 =
_

0
h(y)
_
p
t
(t, T, x, y) + (t, X(t))p
x
(t, T, x, y) +
1
2

2
(t, X(t))p
xx
(t, T, x, y)
_
dy
In order for this equation to hold for all values of h(y) > 0, we require
p
t
(t, T, x, y) = (t, X(t))p
x
(t, T, x, y) +
1
2

2
(t, X(t))p
xx
(t, T, x, y) (q.e.d.)
Exercise 6.10 (Implying the Volatility Surface)
(i) We get

_

K
(y K)

y
(ry p(0, T, x, y)) dy
= (y K)ry p(0, T, x, y)|
y=
y=K
+
_

K
ry p(0, T, x, y)dy
=
_

K
ry p(0, T, x, y)dy (q.e.d.).
Here, we used the assumption given in Equation (6.9.55) in the last step.
(ii) We get
10
1
2
_

K
(y K)

2
y
2
_

2
(T, y)y
2
p(0, T, x, y)
_
dy
=
1
2
(y K)

y
_

2
(T, y)y
2
p(0, T, x, y)
_

y=
y=K

1
2
_

K

y
_

2
(T, y)y
2
p(0, T, x, y)
_
dy
=
1
2
_

K

y
_

2
(T, y)y
2
p(0, T, x, y)
_
dy
=
1
2

2
(T, y)y
2
p(0, T, x, y)

y=
y=K
=
1
2

2
(T, K)K
2
p(0, T, x, K) (q.e.d.).
Here, we used the assumption given in Equation (6.9.57) in the second equality and
the assumption given in Equation (6.9.60) in the fourth equality.
(iii) Starting from Equation (6.9.53), we get
c
T
(0, T, x, K)
= rc(0, T, x, K) + e
rT
_

K
(y K) p
T
(0, T, x, y)dy
= re
rT
_

K
(y K) p(0, T, x, y)dy
+e
rT
_

K
(y K)
_

y
_
ry p(0, T, x, y) +
1
2

2
y
2

2
(T, y)y
2
p(0, T, x, y)
__
dy
= re
rT
_

K
(y K) p(0, T, x, y)dy + re
rT
_

K
y p(0, T, x, y)dy
+
1
2
e
rT

2
(T, K)K
2
p(0, T, x, K)
= rKe
rT
_

K
p(0, T, x, y)dy +
1
2
e
rT

2
(T, K)K
2
p(0, T, x, K).
This is the rst equality in Equation (6.9.59). Here, we used the Kolmogorov forward
equation from Equation (6.9.51) in the second equality and the results from (i) and
(ii) in the third. Next, we use the result from Exercise 5.9, namely that the risk-
neutral distribution can be represented as
p(0, T, x, K) = e
rT
c
KK
(0, T, x, K).
Thus
11
c
T
(0, T, x, K)
= rK
_

K
c
KK
(0, T, x, y)dy +
1
2

2
(T, K)K
2
c
KK
(0, T, x, K)
= rKc
K
(0, T, x, y)|
y=
y=K
+
1
2

2
(T, K)K
2
c
KK
(0, T, x, K)
= rKc
K
(0, T, x, K) +
1
2

2
(T, K)K
2
c
KK
(0, T, x, K) (q.e.d.),
where we used that
lim
K
c(0, T, x, K) = 0
and thus
lim
K
c
KK
(0, T, x, K) = 0.
12