Suggested Solutions for Finance II

Fall 2004
Irina Slinko
(From the Solutions Manual of Raquel M. Gaspar)
1
Contents
4 Stochastic Integrals 3
5 Dierential Equations 7
7 Arbitrage Pricing 9
8 Completeness and Hedging 15
9 Parity Relations and Delta Hedging 17
13 Several Underlying Assets 21
16 Incomplete Markets 24
17 Dividends 25
18 Currency Derivatives 27
20 Bonds and Interest Rates 30
21 Short Rate Models 33
22 Martingale Models for the Short Rate 35
23 Forward Rate Models 39
24 Change of Numeraire 42
2
4 Stochastic Integrals
Exercise 4.1
(a) Since Z(t) is determinist, we have
dZ(t) = e
t
dt
= Z(t)dt.
(b) By denition of a stochastic dierential
dZ(t) = g(t)dW(t)
(c) Using Itos formula
dZ(t) =

2
2
e
W(t)
dt +e
W(t)
dW(t)
=

2
2
Z(t)dt +Z(t)dW
(d) Using Itos formula and considering the dynamics of X(t) we have
dZ(t) = e
x
dX(t) +

2
2
e
x
(dX(t))
2
= Z(t)
_
+
1
2

2
_
dt +Z(t)dW(t).
(e) Using Itos formula and considering the dynamics of X(t) we have
dZ(t) = 2X(t)dX(t) + (d(X(t))
2
= Z(t)
_
2 +
2
_
dt + 2ZdW(t).
Exercise 4.3 By denition we have that the dynamics of X(t) are given
by dX(t) = (t)dW(t).
3
Consider Z(t) = e
iuX(t)
. Then using the Itos formula we have that the
dynamic of Z(t) can be described by
dZ(t) =
_

u
2
2

2
(t)
_
Z(t)dt + [iu(t)] Z(t)dW(t)
From Z(0) = 1 we get,
Z(t) = 1
u
2
2
_
t
0

2
(s)Z(s)ds +iu
_
t
0
(s)Z(s)dW(s).
Taking expectations we have,
E[Z(t)] = 1
u
2
2
E
__
t
0

2
(s)Z(s)ds
_
+iuE
__
t
0
(s)Z(s)dW(s)
_
= 1
u
2
2
__
t
0

2
(s)E [Z(s)] ds
_
+ 0
By setting E [Z(t)] = m(t) and dierentiating with respect to t we nd an
ordinary dierential equation,
m(t)
t
=
u
2
2
m(t)
2
(t)
with the initial condition m(0) = 1 and whose solution is
m(t) = exp
_

u
2
2
_
t
0
2(s)ds
_
= E[Z(t)]
= E
_
e
iuX(t)
_
So, X(t) is normally distributed. By the properties of the normal distribu-
tion the following relation
E
_
e
iuX(t)
_
= e
iuE[X(t)]
u
2
2
V [X(t)]
where V [X(t)] is the variance of X(t), so it must be that E [X(t)] = 0 and
V [X(t)] =
_
t
0

2
(s)ds.
4
Exercise 4.5 We have a sub martingale if E [ X(t)| F
s
] X(s), t s.
From the dynamics of X we can write
X(t) = X(s) +
_
t
s
(z)dz +
_
t
s
(z)dW(z).
By taking expectation, conditioned at time s, from both sides we get
E [ X(t)| F
s
] = E [ X(s)| F
s
] +E
__
t
s
(z)dz

F
s
_
= X(s) +E
s
_

_
_
t
s
(z)dz
. .
0

F
s
_

_
X(s)
so X is a sub martingale.
Exercise 4.6 Set X(t) = h(W
1
(t), , W
n
(t)).
We have by Ito that
dX(t) =
n

i=1
h
x
i
dW
i
(t) +
1
2
n

i,j=1

2
h
x
i
x
j
dW
i
(t)dW
j
(t)
where
h
x
i
denotes the rst derivative with respect to the i-th variable,

2
h
x
i
x
j
denotes the second order cross-derivative between the i-th and j-th variable
and all derivatives should be evaluated at (W
1
(s), , W
n
(s)).
Since we are dealing with independent Wiener processes we know
u : dW
i
(u)dW
j
(u) = 0 for i = j and dW
i
(u)dW
j
(u) = du for i = j,
so, integrating we get
X(t) =
_
t
0
n

i=1
h
x
i
dW
i
(u) +
1
2
_
t
0
n

i,j=1

2
h
x
i
x
j
dW
i
(u)dW
j
(u)
=
_
t
0
n

i=1
h
x
i
dW
i
(u) +
1
2
_
t
0
n

i=1

2
h
x
i
x
j
[dW
i
(u)]
2
=
_
t
0
n

i=1
h
x
i
dW
i
(u) +
1
2
_
t
0
n

i,j=1

2
h
x
i
x
j
du.
5
Taking expectations
E [ X(t)| F
s
] = E
_
_
t
0
n

i=1
h
x
i
dW
i
(u)

F
s
_
+E
_
_
1
2
_
t
0
n

i,j=1

2
h
x
i
x
j
du

F
s
_
_
=
_
s
0
n

i=1
h
x
i
dW
i
(u) +
1
2
_
s
0
n

i,j=1

2
h
x
i
x
j
du
. .
X(s)
+E
_
_
t
0
n

i=1
h
x
i
dW
i
(u)

F
s
_
. .
0
+E
_
_
1
2
_
t
s
n

i,j=1

2
h
x
i
x
j
du

F
s
_
_
= X(s) +E
_
_
1
2
_
t
s
n

i,j=1

2
h
x
i
x
j
du

F
s
_
_
.
If h is harmonic the last term is zero, since

n
i,j=1

2
h
x
i
x
j
= 0, we have
E [ X(t)| F
s
] = X(s) so X is a martingale.
If h is subharmonic the last term is always nonnegative, since

n
i,j=1

2
h
x
i
x
j

0 we have
E [ X(t)| F
s
] X(s) so X is a submartingale.
Exercise 4.8
(a) Using the Itos formula we nd the dynamics of R(t),
dR(t) = 2X(t)(dX(t)) +2Y (t)(dY (t)) +
1
2
_
2(dX(t))
2
+ 2(dY (t))
2
_
= (2 + 1)
_
X
2
(t) +Y
2
(t)
_
dt
= (2 + 1)R(t)dt
From the dynamics we can see immediately that R(t) is deterministic
(it has no stochastic component!).
6
(b) Integrating the SDE for X(t) and taking expectations we have
X(t) = x
0
+
_
t
0
E [X(s)] ds
Which once more can be solve setting m(t) = E [X(t)],taking the
derivative with respect to t and using ODE methods, to get the answer
E [X(t)] = x
0
e
t
5 Dierential Equations
Exercise 5.1 We have:
dY (t) = e
t
x
0
dt, dZ(t) = e
t
dt, dR(t) = e
t
dW(t).
Itos formula then gives us (the cross term dZ(t) dR(t) vanishes)
dX(t) = dY (t) +Z(t) dR(t) +R(t) dZ(t)
= e
t
x
0
dt +e
t
e
t
dW(t) +
_
t
0
e
s
dW(s) e
t
dt
=
_
e
t
x
0
+
_
t
0
e
(ts)
dW(s)
_
dt +dW(t)
= X(t)dt +dW(t).
Exercise 5.5 Using the dynamics of X(t) and the Ito formula we get
dY (t) =
_
+
1
2
( 1)
2
_
Y (t)dt +Y (t)dW(t)
= Y (t)dt +Y (t)dW(t)
where = +
1
2
( 1)
2
and = so Y is also a GBM.
7
Exercise 5.6 From the Ito formula and using the dynamics of X and Y
dZ(t) =
1
Y (t)
dX(t)
X(t)
Y (t)
2
dY (t)
1
Y (t)
2
dX(t)dY (t) +
X(t)
Y (t)
3
(dY (t))
2
= Z(t)
_
+
2
_
dt +Z(t)dW(t) Z(t)dV (t).
Exercise 5.9 From Feyman-Kac we have We have
F(t, x) = E
t,x
[2 ln[X(T)]] ,
and
dX(s) = X(s)ds +XdW(s),
X(t) = x.
Solving the SDE, we obtain (check the solution of the GBM in th extra
exercises if you do not remmeber)
X(T) = exp
_
lnx + (
1
2

2
)(T t) +[W(T) W(t)]
_
,
and thus
F(t, x) = 2 ln(x) + 2(
1
2

2
)(T t).
Exercise 5.10 Given the dynamics of X(t) any F(t, x) that solves the prob-
lem has the dynamics given by
dF(t, x) =
F
t
dt +
F
x
dX(t) +
1
2

2
F
x
2
(dX(t))
2
=
F
t
dt +
F
x
[(t, x)dt +(t, x)dW(t)] +k(t, x)dt k(t, x)dt
+
1
2

2
F
x
2
_

2
(t, x)dW(t)
_
=
_

_
F
t
+(t, x)
F
x
+
1
2

2
(t, x) +k(t, x)
. .
0
_

_
dt k(t, x)dt
+
F
x
(t, x)dW(t)
= k(t, x)dt +
F
x
(t, x)dW(t)
8
We now write F(T, X(T)) in terms of F(t, x) and the dynamics of F during
the time period t . . . T(recall that we dened X(t) = x)
F(t, X(T)) = F(t, x)
_
T
t
k(s, X(s)ds +
_
T
t
F
x
(s, X(s))dW(s)

F(t, x) = F(T, X(T)) +

_
T
t
k(s, X(s)ds
_
T
t
F
x
(s, X(s))dW(s)
Taking expectations E
t,x
[.] from both sides
F(t, x) = E
t,x
[F(T, X(T))] +E
t,x
_
_
T
t
k(s, X(s)ds
_
= E
t,x
[(T)] +
_
T
t
E
t,x
[k(s, X(s)] ds
Exercise 5.11 Using the representation formula from Exercise 5.10 we get
F(t, x) = E
t,x
_
2 ln[X
2
(T)]
_
+
_
T
t
E
t,x
[X(s)] ds,
Given
dX(s) = X(s)dW(s).
The rst term is easily computed as in the exercise 5.9 above. Furthermore
it is easily seen directly from the SDE (how?)that E
t,x
[X(s)] = x. Thus we
have the result
F(t, x) = 2 ln(x) (T t) +x(T t)
= ln(x
2
) + (x 1)(T t)
7 Arbitrage Pricing
Exercise 7.1
9
(a) From standard theory we have
(t) = F(t, S(t)), where F solves the Black-Scholes equation.
Using Ito we obtain
d(t) =
_
F
t
+rS(t)
F
s
+
1
2

2
S
2
(t)

2
F
s
2
_
dt +S(t)
F
s
dW(t).
Using the fact that F satises the Black-Scholes equation, and that
F(t, S(t)) = (t) we obtain
d(t) = r(t) dt +S(t)
F
s
dW(t)
and so g(t) = S(t)
F
s
.
(b) Apply Itos formula to the process Z(t) =
(t)
B(t)
and use the result in
(a).
dZ(t) =
1
B(t)
(d(t))
(t)
B
2
(t)
(d(B(t))
=
g(t)
B(t)
dW(t)
= Z(t)
S(t)
(t)
F
s
dW(t)
Z is a martingale since E
t
[Z(T)] = Z(t) for all t < T and its diusion
coecient is given by
Z
(t) =
S(t)
(t)
F
s
.
Exercise 7.4 We have as usual
(t) = e
r(Tt)
E
Q
t,s
_
S

(T)
_
.
We know from earlier exercises (check exercises 3.4 and 4.5) that Y (t) =
S

(t) satises the SDE under Q

dY (t) =
_
r +
1
2
( 1)
2
_
Y (t)dt +Y (t)dW(t).
10
Using the standard technique, we can integrate, take expectations, dieren-
tiate with respect to time and solve by ODE techniques, to obtain
E
Q
t,s
_
S

(T)
_
= s

e
[r+
1
2
(1)
2
](Tt)
,
So,
(t) = s

e
[r(1)+
1
2
(1)
2
](Tt)
.
Exercise 7.5 In the case of the binary option the payo (s) can be
described as
(s) =
_
K, if s [, ]
0, otherwise.
We know that the arbitrage free price (t) is given by
(t) = e
r(Tt)
E
Q
t,s
[(S(T))] .
Without any loss of generality we can normalize K and set K = 1. Given
the specially form of the payo (s) we have
E
Q
t,s
[(T)] = Q[S(T) [, ]]
= Q[ S(T) ] .
We also know, from the properties of a GBM, that
E
Q
t
[ln(S(T))] = ln(S(t)) +r
1
2

2
,
V [S(T)] =
2
(T t) and that lnS(T) is normally distributed. So we have
lnS(T) N
_
ln(S(t)) +r
1
2

2
,

T t
_
.
Thus
E
Q
t,s
[(T)] = Q[ln ln(S(T)) ln]
= Q
_

_
ln lnS(t) (r

2
2
)(T t)

. .
A
Z
ln lnS(t) (r

2
2
)(T t)

. .
B
_

_
11
where Z N [0, 1] . So,we have
(t) = e
r(Tt)
{N [A] N [B]}
where A =
ln ln S(t)(r

2
2
)(Tt)

and B =
ln lnS(t)(r

2
2
)(Tt)

. For K =
1 we just have (t) = Ke
r(Tt)
{N [A] N [B]}.
Exercise 7.6 We want to nd the arbitrage free price process for the claim
for the claim X where X is given by
X =
S(T
1
)
S(T
0
)
(1)
where the times T
0
and T
1
are given and claim is paid out at T
1
.
Under the martingale measure Q
dS = rSds +SdW
s
S(t) = s
Since the stock dynamics is a GBM we can solve for S(T
1
) and S(T
0
) ex-
plicitely.
S(T
1
) = s exp(r
1
2

2
)(T
1
t) +(W(T
1
) W(t)) (2)
S(T
0
) = s exp(r
1
2

2
)(T
0
t) +(W(T
0
) W(t)) (3)
Thus the ratio is
S(T
1
)
S(T
0
)
= exp(r
1
2

2
)(T
1
T
0
) +(W(T
1
) W(T
0
)) (4)
Using the characteristic functions for normal distribution and noticing that
(W(T
1
) W(T
0
)) with zero mean and variance
2
(T
1
T
0
)
E
Q
_
S(T
1
)
S(T
0
)
_
= e
(r
1
2

2
)(T
1
T
0
)
E
Q
_
e
(W(T
1
)W(T
0
))
_
= e
(r
1
2

2
)(T
1
T
0
)
e
1
2
(T
1
T
0
)
= e
r(T
1
T
0
)
The arbitrage free price process is equal then
(t, s) = e
r(T
1
t)
e
r(T
1
T
0
)
= e
r(T
0
t)
(5)
12
Exercise 7.7 The price in SEK of the ACME INC., Z, is dened as Z(t) =
S(t)Y (t) and by Ito has the following dynamics under Q
dZ(t) = rZ(t)dt +Z(t)dW
1
(t) +Z(t)dW
2
(t)
We also have, by using Ito once more, that the dynamics of lnZ
2
are
d lnZ
2
(t) =
_
2r
2

2
_
dt + 2dW
1
(t) + 2dW
2
(t)
which integrating and taking conditioned expectations give us
E
Q
t,z
_
ln[Z
2
(T)]
_
= lnz
2
+
_
2r
2

2
_
(T t)
Since we know that
(t) = F(t, s) = e
r(Tt)
E
Q
t,z
_
ln[Z
2
(T)]
_
,
the arbitrage free pricing function is
(t) = e
r(Tt)
_
lnz
2
+
_
2r
2

2
_
(T t)
_
= e
r(Tt)
_
2 ln(sy) +
_
2r
2

2
_
(T t)
_
,
where, as usual, z = Z(t), s = S(t) and y = Y (t).
Exercise 7.9 The forward price, i.e. the amount of money to be payed out
at time T, but decided at the time t is
F(t, T) = E
Q
t
[X] .
Note that the forward price is not the price of the forward contract on the
T-claim X which is what we are looking for.
Take for instance the long position: at time T, the buyer of a forward
contract receives X and pays F(t, T). Hence, the price at time t of that
position is
(t; X F(t, T)) = E
Q
t
_

_
e
r(Tt)
_
_
_
_
X F(t, T)
. .
E
Q
t
[X]
_
_
_
_
_

_
= 0.
13
At time s > t, however, the underlying asset may have changed in value, in
a way dierent from expectations, so then the price of a forward contract
can be dened as
(s; X F(t, T)) = E
Q
s
_
e
r(Ts)
(X F(t, T))
_
= e
r(Ts)
_

_
E
Q
s
[X]
F(t,T)
..
E
Q
t
[X]
_

_
.
Remark: For the special case where the contract is on one share S we get:
(s) = e
r(Ts)
_

_
E
Q
s
[S(T)] S(t)e
r(Tt)
. .
E
Q
t
[S(T)]
_

_
.
We can also easily calculate E
Q
s
[S(T)] since
E
Q
s
[S(T)] = S(t) +r
_
s
t
S(u)du
. .
S(s)
+r
_
T
s
E
Q
s
[S(u)] du
so,
E
Q
s
[S(T)] = S(s)e
r(Ts)
and, therefore, the free arbitrage pricing function at time s > t is
(s) = S(s) S(t)e
r(st)
.
14
8 Completeness and Hedging
Exercise 8.2 We have F(t, s, z) be dened by
F
t
+r s F
s
+
1
2

2
s
2
F
ss
+gF
z
= rF
F(T, s, z) = (s, z)
and the dynamics under Q for S and Z
dS(u) = rS(u)du +S(u)dW(u)
dZ(u) = g(u, S(u))du
We want to show that F(t, S(t), Z(t)) = e
r(Tt)
E
Q
t,s,z
[(S(T), Z(T))].
For that we nd , by Ito, the dynamics of (t) = F(t, S(t), Z(t)), the arbi-
trage free pricing process
d(t) = F
t
dt +F
s
[(rS(t)dt +S(t)dW(t)] +F
z
g(t, S(t))dt +
1
2
F
ss

2
S
2
(t)dt
=
_
F
t
+r S(t) F
s
+
1
2

2
S
2
(t)F
ss
+g(t, S(t))F
z
_
. .
r(t)
+S(t)F
s
dW(t)
Integrating we have
(T) = (t) +r
_
T
t
(u)du +
_
T
t
S(u)F
s
dW(u)
Hence
E
Q
t,z,s
[(T)] = (t) +r
_
T
t
E
Q
t,z,s
[(u)] du
So, using the usual trick of setting m(u) = E
Q
t,z,s
[(u)] and using tech-
niques of ODE we nally get
(t) = F(t, S(t), Z(t)) = e
r(Tt)
E
Q
t,s,z
[(S(T), Z(T))] .
(Remember that (T) = F(T, S(T), Z(T)) = (S(T), Z(T)).)
15
Exercise 8.3 The price arbitrage free price is given by (note that this time
our claim is not simple, i.e. it is not of the form X= (S(T))).
(t) = e
r(T
2
t)
E
Q
t
[X]
= e
r(T
2
t)
1
T
2
T
1
_
T
2
T
1
E
Q
t
[S(u)] du
We know that under Q
dS(u) = rS(u)du +S(u)dW(u)
S(t) = s
So,
E
Q
t
[S(u)] = se
r(ut)
1
T
2
T
1
_
T
2
T
1
se
r(ut)
du =
1
T
2
T
1
s
r
_
e
r(T
2
t)
e
r(T
1
t)
_
The price to the mean contract is thus
(t) =
s
r(T
2
T
1
)
_
1 e
r(T
2
T
1
)
_
.
16
9 Parity Relations and Delta Hedging
Exercise 9.1 The T-claim X given by:
X =
_

_
K, if S(T) A
K +A S(T), if A < S(T) < K +A
0, otherwise.
,
has then following contract function (recall that X = S(T))
(x) =
_

_
K, if x A
K +Ax, if A < x < K +A
0, otherwise.
,
which can be decomposed into the following basic contract functions writ-
ten
(x) = K 1
..

B
(x)
max [0, x A]
. .

c,A
(x)
+max [0, x A K]
. .

c,A+K
(x)
.
Having this T-claimX is then equivalent to having the following (replicating)
portfolio at time T:
* K in monetary units
* short (position in) a call with strike A
* long (position in) a call with strike A+K
Given the decomposition of the contract function into basic contract func-
tions, we immediately have that the arbitrage free pricing process is
(t) = K
B(t)
..
e
r(Tt)
c(s, A, T) +c(s, A +K, T)
where c(s, A, T) and c(s, A + K, T) stand for the prices of European call
options on S and maturity T with strike prices A and A+ K, respectively.
The Black-Scholes formula give us both c(s, A, T) and c(s, A +K, T) .
17
The hedge portfolio thus consists of a reverse position in the above compo-
nents, i.e., borrow e
r(Tt)
K, buy a call with strike K and sell a call with
strike A+K.
Exercise 9.4 We apply, once again, the exact same technique. The T-claim
X given by:
X =
_

_
0, if S(T) < A
S(T) A, if A S(T) B
C S(T), if B < S(T) C
0, if S(T) > C.
where B =
A+C
2
, has a contract function that can be written as
(x) = max [0, x A]
. .

c,A
(x)
+max [0, x C]
. .

c,C
(x)
2 max [0, x B]
. .

c,B
(x)
Having this buttery is then equivalent to having the following constant(replicating)
portfolio at time T:
* long (position in) a call option with strike A
* long (position in) a call option with strike C
* short (position in) a call option with strike B
The arbitrage free pricing process follows immediately from the decom-
position of the contract function and is given by
(t) = c(s, A, T) +c(s, C, T) 2c(s, B, T)
where c(s, A, T), c(s, B, T) and c(s, C, T) stand for the prices of European
call options on S, with maturity T and strike prices A, B and C, respectively,
and can be computed using the Black-Scholes formula.
The hedge portfolio consists of a reverse position in the above components,
i.e., sell two call options one with strike A and other with strike B and buy
other two both with strike B.
18
Exercise 9.5 We have a portfolio P and two derivatives F and G. In order
to delta-hedge our portfolio we need to combine the two derivatives in a way
such that
u
F
.
F
+u
G
.
G
=
P
,
since, in addition we what to gamma-hedge we also need
u
F
.
F
+u
G
.
G
=
F
Where u
F
and u
G
are the quantities of the derivatives F and G, respectively,
that should be bought (or sold if negative in value).
So we just need to solve the system
_
_
_
u
F
+ 5u
G
= 2
2u
F
2u
G
= 3
_

_
u
F
=
19
8
u
G
=
7
8
the hedging strategy is then to short
19
8
of derivative F and
7
8
of derivative
G.
Exercise 9.10 From the put-call parity we have that
p(t, s) = Ke
r(Tt)
+c(t, s) S
where p(t, s) and c(t, s) stand for the price of a put and a call option on S
with maturity T and strike price K.
The delta measures the variation in the price of a derivative with respect to
changes in the value of the underlying. Dierentiating the put-call parity
w.r.t. S we have
p(t, s)
S
. .
put
=

S
_
Ke
r(Tt)
_
+
c(t, s)
S
. .

call

S
S

put
=
call
1
19
Since,
call
= N [d1]
put
= N [d1] 1.
To nd the result on the gamma we dierentiate one more time (so two
times) w.r.t. S the put-call parity and get

2
p(t, s)
S
2
. .

call
=

2
c(t, s)
S
2
. .
put
From the fact that
call
=
(d1)
s

Tt
it follows that
put
=
(d1)
s

Tt
.
20
13 Several Underlying Assets
Exercise 13.3
Consider the T-claim X = max [S
1
(T) S
2
(T); 0] where S
1
and S
2
are de-
ned by
dS
1
(t) =
1
S
1
(t)dt +
1
S
1
(t)d

W
1
dS
2
(t) =
2
S
2
(t)dt +
2
S
2
(t)d

W
2
and
d

W
1
.d

W
2
= dt.
The contract function associated with it, is homogeneous of degree 1, and
thus, it can be rewritten as a contract function on the normalized variable
z =
s
1
s
2
.
(s
1
, s
2
) = max [s
1
s
2
; 0] = s
2
max
_

_
s
1
s
2
..
z
1; 0
_

_
= s
2
max [z 1; 0]
. .
(z)
.
So, the pricing function F(t, s
1
, s
2
) = s
2
G(t, z) and from
F
t
= s
2
G
t
F
1
= s
2
G
z
z
s
1
= s
2
G
z
1
s
2
= G
z
F
2
= G+s
2
G
z
z
s
2
= G +s
2
G
z
_

s
1
s
2
2
_
= GzG
z
F
11
= G
zz
z
s
1
=
1
s
2
G
zz
F
22
= G
z
z
s
2

z
s
2
G
z
+zG
zz
z
s
2
_
=
1
s
2
z
2
G
zz
F
12
= G
zz
z
s
2
=
1
s
2
zG
zz
21
we realize that G satises the (much easier) PDE,
G
t
+
1
2
z
2
G
zz
(
2
1
+
2
2
2
1

2
) = 0
G(T, z) = max [z 1, 0] ,
that we recognize as the Black-Scholes equation and, therefore we have that
G is the price of a call with underlying Z, exercise price K = 1 and with,
r = 0, and =
_

2
1
+
2
2
2
1

2
.
So, we can use Black-Scholes formula to concretize the pricing function
(t) = F(t, S
1
(t), S
2
(t)):
(t) = S
2
(t)G(t, Z(t))
= S
2
(t)
_

_
Z(t)
..
S
1
(t)
S
2
(t)
N
_
d

N
_
d

_
= S
1
(t)N
_
d

S
2
(t)N
_
d

where
d

1
=
1
_

2
1
+
2
2
2
1

T t
_
ln
_
S
1
(t)
S
2
(t)
_
+
1
2
_

2
1
+
2
2
2
1

2
_
(T t)
_
and d

2
= d

2
1
+
2
2
2
1

T t.
Exercise 13.4 For this special maximum option we have X = max [aS
1
(T); bS
2
(T)]
where S
1
and S
2
are dened by
dS
1
(t) =
1
S
1
(t)dt +
1
S
1
(t)d

W
1
dS
2
(t) =
2
S
2
(t)dt +
2
S
2
(t)d

W
2
and

W
1
is independent from

W
2
.
22
The contract function is homogeneous of degree 1 and can be rewritten
in terms of the normalized variable z =
s
1
s
2
:
(s
1
, s
2
) = max [as
1
; bs
2
] = bs
2
max
_
as
1
bs
2
; 1
_
= = bs
2
_
max
_
as
1
bs
2
1; 0
_
+ 1
_
= bs
2
_

_
a
b
max
_

_
z
..
s
1
s
2

b
a
; 0
_

_ + 1
_

_
= bs
2
+as
2
max
_
z
b
a
; 0
_
. .
(z)
Thus, the pricing function F is given by F(t, s
1
, s
2
) = bs
2
+as
2
G(t, z) where
G solves the boundary problem So we have
G
t
+
1
2
z
2
G
zz
(
2
1
+
2
2
) = 0
G(T, z) = max
_
z
b
a
, 0
_
.
Once again, the above expression is the Black-Scholes equation, G is the
price of a call option on Z with strike price K =
b
a
, r = 0, and =
_

2
1
+
2
2
We can use Black-Scholes formula to concretize the pricing function (t) =
F(t, S
1
(t), S
2
(t):
(t) = bS
2
(t) +aS
2
(t)G(t, Z(t)) = bS
2
(t) +aS
2
(t)
_
S
1
(t)
S
2
(t)
N
_
d

b
a
N
_
d

_
= bS
2
(t) +aS
1
(t)N
_
d

bS
2
(t)N
_
d

= (1 N
_
d

)bS
2
(t) +aN
_
d

S
1
(t)
where
d

1
=
1
_

2
1
+
2
2

T t
_
ln
_
aS
1
(t)
bS
2
(t)
_
+
1
2
_

2
1
+
2
2
_
(T t)
_
and d

2
= d

2
1
+
2
2

T t.
23
16 Incomplete Markets
Exercise 16.1 Given a claim X = (X(T))) and since the dynamics of X
under the Q-measure are
dX(t) = [(t, X(t)) (t, X(t))(t, X(t))] dt +(t, X(t))dW
Q
(t),
we can nd the Q-dynamics of the pricing function F(t, X(t)) using the Ito
formula
dF(t, X(t)) =
F
t
dt +
F
x
dX(t) +
1
2

2
F
x
2
(dX(t))
2
=
F
t
dt +
F
x
[(t, X(t)) (t, X(t))(t, X(t))] dt +(t, X(t))dW
Q
(t)
+
1
2

2
F
x
2

2
(t, X(t))dt
=
_
F
t
+
F
x
((t, X(t)) (t, X(t))(t, X(t))) +
1
2

2
F
x
2

2
(t, X(t))
_
. .
rF(t,X(t))
dt
+(t, X(t))dW
Q
(t)
= rF(t, X(t)) +(t, X(t))dW
Q
(t)
where the last step results from the fact that F has to satisfy the pricing
PDE:
F
t
+
F
x
((t, X(t)) (t, X(t))(t, X(t))) +
1
2

2
F
x
2

2
(t, X(t)) = rF.
24
17 Dividends
Exercise 17.2 We know that when there is a continuous dividend being
paid we have the following dynamics under Q for the asset S and the dividen
structure D
dS(t) = [r (S(t))] S(t)dt + (S(t)) S(t)dW(t)
dD(t) = S(t) (S(t)) dt
Just by rewritting and rearranging terms we have
dS(t) = rS(t)dt (S(t)) S(t)dt
. .
dD(t)
+ (S(t)) S(t)dW(t)
e
rt
dS(t) e
rt
rS(t)dt = e
rt
dD(t) +e
rt
(S(t)) S(t)dW(t)
d
_
e
rt
S(t)
_
= e
rt
dD(t) +e
rt
(S(t)) S(t)dW(t).
Integratting the last expression we get
e
rt
S(t) = e
r0
S(0)
_
t
0
e
ru
dD(u) +
_
t
0
e
ru
(S(u))S(u)dW(u)
S(0) = e
rt
S(t) +
_
t
0
e
ru
dD(u)
_
t
0
e
ru
(S(u))S(u)dW(u),
and taking E
Q
0
[.] expectations we nally get the results
S(0) = E
Q
0
_
e
rt
S(t) +
_
t
0
e
ru
dD(u)
_
.
Exercise 17.6 In the Black-Scholes model with a constant continuous div-
idend yield we have, under the Q-measure we have
dS(t) = (r ) S(t)dt +S(t)dW
Q
(t).
From the standard call-put parity, which must be valid, we have the fol-
lowing relation between call and put options with the same maturity T and
exercise price K:
p(t, x) = c(t, x) (t, S(T)) +e
r(Tt)
K.
25
But we also know that
(t, S(T)) = e
r(Tt)
E
Q
t
[S(T)]
= e
r(Tt)
S(t)e
(r)(Tt)
= S(t)e
(Tt)
.
So,
p(t, x) = c(t, x) S(t)e
(Tt)
+e
r(Tt)
K.
26
18 Currency Derivatives
Exercise 18.1 We have, under the objective probability measure, the fol-
lowing processes for the spot exchange rate X (units of domestic currency d
per foreign currency unit f), and the domestic , B
d
, and foreign, B
f
, riskless
assets:
dX(t) =
x
X(t)dt +
x
X(t)dW(t)
dB
d
(t) = r
d
B(t)
d
dt
dB
f
(t) = r
f
B(t)
d
dt,
where r
d
and r
f
are the domestic and foreign short rates which are assumed
to be deterministic. Hence the Q-dynamics of X are given by
dX(t) = (r
d
r
f
)X(t)dt +
x
X(t)dW(t).
From the standard call-put parity, which must be valid, we have the fol-
lowing relation between call and put options with the same maturity T and
exercise price K:
p(t, x) = c(t, x) (t, X(T)) +e
r
d
(Tt)
K.
But we also know that
(t, X(T)) = e
r
d
(Tt)
E
Q
t
[X(T)]
= e
r
d
(Tt)
X(t)e
(r
d
r
f
)(Tt)
= X(t)e
r
f
(Tt)
.
So,
p(t, x) = c(t, x) X(t)e
r
f
(Tt)
+e
r
d
(Tt)
K.
(Remark: Compare this exercise with exercise 11.6 in the previous sec-
tion, and see that the foreign risk-free rate can be treated as a continuous
dividend-yield on the spot exchange rate.)
27
Exercise 18.2 The binary option on the exchange rate X is a T-claim, Z,
of the form
Z = 1
[a,b]
(X(T)) ,
i.e. if a X(T) b then the holder of this claim will obtain one unit of
domestic currency, otherwise gets nothing. The dynamics under Q of the
exchange rate X are given by
dX(t) = (r
d
r
f
)X(t)dt +
x
X(t)dW(t).
where r
d
and r
f
are the domestic and foreign short rates which are assumed
to be deterministic.
Integrating the above expression we get
X(T) = X(t) +
_
T
t
(r
d
r
f
)X(u)du +
_
t
t

x
X(u)dW(u)
and so we have (by solving the SDE)
X(T) = X(t)e
_
r
d
r
f

2
x
2
_
(Tt)+x(W(T)W(t))
,
and we see from taking the logarithm that
ln(X(T)) N
_
ln(X(t)) +
_
r
d
r
f


2
x
2
_
(T t),
x

T t
_
.
So, the price, (t) of the binary option is given by
(t) = e
r
d
(Tt)
E
Q
t
[Z]
= e
r
d
(Tt)
Q(a X(T) b)
= e
r
d
(Tt)
Q(ln(a) ln(X(T)) ln(b))
= e
r
d
(Tt)
Q(d
a
z d
b
)
where z N [0, 1] and d
a
=
ln
_
a
X(t)
_

_
r
d
r
f

2
x
2
_
(Tt)
x

Tt
and d
b
=
ln
_
b
X(t)
_

_
r
d
r
f

2
x
2
_
(Tt)
x

Tt
and so we have that the price of the binary exchange option Z is:
(t) = e
r
d
(Tt)
[N (d
b
) N (d
a
)] .
28
Exercise 18.3 Under the objective probability measure we have the follow-
ing dynamics for the domestic stock S
d
, the foreign stock S
f
, the exchange
rate X and the domestic B
d
and foreign B
f
riskless assets
dS
d
(t) =
d
S
d
(t)dt +
d
S
d
(t)dW
d
(t),
dS
f
(t) =
f
S
f
(t)dt +
f
S
f
(t)dW
f
(t),
dX(t) =
x
X(t)dt +
x
X(t)dW(t),
dB
d
(t) = r
d
B
d
(t)dt,
dB
f
(t) = r
f
S
d
(t)dt,
where r
d
and r
f
are the domestic and foreign short rates which are assumed
to be deterministic. The domestic stock denominated in terms of the foreign
currency is given by

S
d
=
S
d
X
, then by Ito
d

S
d
(t) =
1
X(t)
dS
d
(t)
S
d
(t)
X
2
(t)
dX(t) +
S
d
(t)
X
3
(t)
(dX(t))
2

1
X
2
(t)
dS
d
(t)dX(t)
. .
0 for W
d
and W indep.
=
d

S
d
(t)dt +
d

S
d
(t)dW
d
(t)
x

S
d
(t)dt
x

S
d
(t)dW(t) +
2
x

S
d
(t)dt
=
_

x
+
2
x
_

S
d
(t)dt +
d

S
d
(t)dW
d
(t)
x

S
d
(t)dW(t).
The dynamics above are under the objective probability measure. Under Q
f
the drift term of all assets denominated in the foreign currency must have
r
f
(the risk-free rate on the foreign economy) as the drift. Also, we can use
the properties of the Wiener processes to normalize the diusion part.
It follows that
d

S
d
(t) = r
f

S
d
(t)dt +
_

2
d
+
2
x

S
d
(t)dW
f
(t).
29
20 Bonds and Interest Rates
Exercise 20.1 Forward Rate Agreement
(a) Note that the cash ow to the lenders in a FRA (K at time S and
Ke
R

(TS)
at time T) can be replicated by the following portfolio:
* sell K S-bonds
* buy Ke
R

(TS)
T-bonds
So, at time t < S, the value (t), on the lenders cash ow in a FRA
has to equal to the value of the replicating portfolio and is given buy
(t) = Ke
R

(TS)
p(t, T) Kp(t, S).
(b) If we have that (0) = 0 we must have
Ke
R

(TS)
p(0, T) Kp(0, S) = 0
e
R

(TS)
p(0, T) p(0, S) = 0
R

(T S) = ln
_
p(0, S)
p(0, T)
_
R

=
ln(p(0, S)) ln(p(0, T))
T S
Since by denition the forward rate R(t; S, T) is given by
R(t; S, T) =
ln(p(t, T)) ln(p(t, S))
T S
,
we have that R

= R(0; S, T).
Exercise 20.2 Since f(t, T) =
ln(p(t,T))
T
and p(t, T) satises
dp(t, T) = p(t, T)m(t, T)dt +p(t, T)v(t, T)dW(t),
30
then we have, by Ito, that ln(p(t, T)) has dynamics that are given by
d ln(p(t, T)) = m(t, T)dt +v(t, T)dW(t)
1
2
v(t, T)v(t, T)

dt
=
_
m(t, T)
1
2
v(t, T)v(t, T)

_
dt +v(t, T)dW(t),
and integrating we have
ln(p(t, T)) = ln(p(0, T))+
_
t
0
m(s, T)
1
2
v(s, T)v(s, T)

ds+
_
t
0
v(s, T)dW(s).
Since f(t, T) =
ln(p(t,T))
T
, we also have
f(t, T) =
ln(p(0, T))
T
. .
f(0,T)

_
t
0
_
_
_
_
_
m(s, T)
T
. .
m
T
(s,T)
v(s, T)
v(s, T)
T
. .
v
T
(s,T)

_
_
_
_
_
ds
_
t
0
v(s, T)
T
dW(s)
and from its dierential form we nally get the result
df(t, T) =
_

_
v(t, T)v
T
(t, T)

m
T
(t, T)
. .
(t,T)
_

_
dt v
T
(t, T)
. .
(t,T)
dW(t).
Exercise 20.5 Let {y(0, T; T 0} denote the zero coupon yield curve.
(a) Then we have that
p(0, T) = e
y(0,T)T
.
Using the denition of the instantaneous forward rate f(t, T) =
ln(p(t,T))
T
and the expression above we get the result
f(0, T) =
ln(p(0, T))
T
=
ln
_
e
y(0,T)T
_
T
=
(y(0, T) T)
T
= y(0, T) +T
y(0, T)
T
.
31
(b) If the zero coupon yield curve is an increasing function of T, then we
know that
y(0,T)
T
0, and using the result from (a) we have that
f(0, T) y(0, T) for any T > 0.
It remains to prove that y
M
(0, T) y(0, T). This will follow from the
price of a coupon bond is given by
p
T
(t) = Kp(t, T
n
) +
n

i=1
c
i
p(t, T
i
).
So in particular we have that
p
T
(0) = K e
y(0,Tn)Tn
. .
p(0,Tn)
+
n

i=1
c
i
e
y(0,T
i
)T
i
. .
p(0,T
i
)
.
but it can also we given, since y
M
is the yield to maturity of a coupon
bond, by
p
T
(0) = Ke
y
M
(0,Tn)Tn
+
n

i=1
c
i
e
y
M
(0,Tn)T
i
.
So,
K e
y(0,Tn)Tn
. .
p(0,Tn)
+
n

i=1
c
i
e
y(0,T
i
)T
i
. .
p(0,T
i
)
= Ke
y
M
(0,Tn)Tn
+
n

i=1
c
i
e
y
M
(0,Tn)T
i
By comparing the LHS and the RHS and since c
i
0, and y(0, T
n
)
y(0, T
n1
) y(0, T
1
) (by assumption), we must have that y(0, T
n
)
y
M
(0, T
n
) since it is valid for any T
n
we can write that y(0, T)
y
M
(0, T) for any T.
32
21 Short Rate Models
Exercise 21.2 The object of the exercise is to connect the forward rates to
the risk neutral valuation of bond prices.
(a) Recall from the risk-neutral valuation of bond prices that
p(t, T) = E
Q
t
_
exp
_

_
T
t
r(s)ds
__
and hence, using the denition of a instantaneous forward rate (and
assuming that we can dierentiate under the expectation sign) we get
f(t, T) =
ln(p(t, T))
T
=

T
ln
_
E
Q
t
_
exp
_

_
T
t
r(s)ds
___
=
E
Q
t
_
r(T) exp
_

_
T
t
r(s)ds
__
E
Q
t
_
exp
_

_
T
t
r(s)ds
__ .
(b) To check that indeed we have f(t, t) = r(t) use the expression above
f(t, T) =
E
Q
t
_
r(T) exp
_

_
T
t
r(s)ds
__
E
Q
t
_
exp
_

_
T
t
r(s)ds
__ ,
and set T = t:
f(t, t) =
E
Q
t
_
r(t) exp
_

_
t
t
r(s)ds
__
E
Q
t
_
exp
_

_
t
t
r(s)ds
__
=
E
Q
t
[r(t) exp{0}]
E
Q
t
[exp{0}]
= r(t).
Exercise 21.3 Recall that in a swap of a xed rate vs. a short rate we
have:
33
A invests K at time 0 and let it grow at a xed rate of interest R over
the time interval [0, T]. A has thus an amount K
A
at time T and at
that time (T) pays the surplus K
A
K to B.
B invests the principal at a stochastic short rate of interest over the
interval [0, T]. B has thus an amount K
B
at time T and at that time
(T) pays the surplus K
B
K to A.
The swap rate for this contract is dened as the value, R, of the xed
rate which gives this contract value zero at t = 0.
At maturity party A has (T) = K
B
Ke
RT
where K
B
= Ke
_
T
0
r(s)ds
.
Thus, the value of this contract at time 0 is given by
(0) = E
Q
0,r
_

_e

_
T
0
r(s)ds
_
_
_Ke
_
T
0
r(s)ds
Ke
RT
. .
(T)
_
_
_
_

_
= E
Q
0,r
_
K Ke
RT
_
T
0
r(s)ds
_
= K E
Q
0,r
_
1 e
RT
e

_
T
0
r(s)ds
_
= K
_
_
_
_
_
_
1 e
RT
E
Q
0,r
_
e

_
T
0
r(s)ds
_
. .
p(0,T)
_
_
_
_
_
_
Since we must have (0) = 0 we have
K
_
1 e
RT
p(t, T)
_
= 0
e
RT
=
1
p(0, T)
R T = ln(p(0, T))
R =
ln(p(0, T))
T
.
34
22 Martingale Models for the Short Rate
Exercise 22.1 In the Vasicek model we have
dr(t) = (b ar(t))dt +dW(t)
with a > 0.
(a) Using the SDE above and multiplying both sides by e
at
we have that
e
at
dr(t) +r(t)ae
at
dt = e
at
bdt +e
at
dW(t)
d(e
at
r(t)) = e
at
bdt +e
at
dW(t)
e
at
r(t) = r(0) +
_
t
0
e
as
bds +
_
t
0
e
as
dW(s)
r(t) = r(0)e
at
+
b
a
_
1 e
at
_
+e
at
_
t
0
e
a
(s)dW(s).
Looking at the solution of the SDE it is immediate that r is Gaussian,
so it is enough to determine the mean and variance.
From the solution of the SDE we see that
E [r(t)] = r(0)e
at
+
b
a
_
1 e
at
_
and
V [r(t)] =
2
e
2at
_
t
0
e
2as
ds
=

2
2a
_
1 e
2at
_
.
(b) As t we have
lim
t
E [r(t)] = r(0) lim
t
e
at
+
b
a
_
1 lim
t
e
at
_
=
b
a
,
and
lim
t
V [r(t)] =

2
2a
_
1 lim
t
e
2at
_
=

2
2a
.
So, N
_
b
a
,

2
2a
_
is the limiting distribution of r.
35
(c) The results in (a) and (b) are based on the assumption that the value
r(0) is known. If, instead we have that r(0) N
_
b
a
,

2
2a
_
, then for any
t we have that
E [r(t)] = E [r(0)] e
at
+
b
a
_
1 e
at
_
=
b
a
,
and
V [r(t)] = e
2at
V [r(0)] +

2
2a
_
1 e
2at
_
=

2
e
2at
2a
+

2
2a
_
1 e
2at
_
=

2
2a
.
Exercise 22.2 From Exercise 17.1 we know that for the Vasicek model, and
t < u we have
r(u) = r(t)e
a(ut)
+
b
a
_
1 e
a(ut)
_
+
_
u
t
e
a(us)
dW(s).
By denition,
p(t, T) = E
Q
t,r
_
e

_
T
t
r(u)du
_
.
Let us dene Z(t, T) =
_
T
t
r(u)du, from exercise 22.1 we know that r is
normally distributed, so Z is also normally distributed since
Z(t, T) = r(t)e
at
_
T
t
e
au
du
b
a
_
1 e
at
_
T
t
e
au
du
_
+
_
T
t
_
u
t
e
a(us)
dW(s)du
=
r(t)
a
_
1 e
a(Tt)
_

b
a
(T t)
b
2
a
_
1 e
a(Tt)
_
+
_
T
t
_
u
t
e
a(us)
dW(s)du
and has
E
Q
t,r
[Z(t, T)] =
_

1
a
_
1 e
a(Tt)
_
_
. .
deterministic function of t and T
r(t)
b
a
(T t)
b
2
a
_
1 e
a(Tt)
_
. .
deterministic function of t and T
and
V
Q
t,r
[Z(t, T)] = V
_

_
u
t
_
T
t
e
a(us)
dudW(s)
_
=
_
u
t
_
_
T
t
e
a(us)
du
_
2
ds
. .
deterministic function of t and T
.
36
Since Z(t, T) is normally distributed
t,T
(from the properties of the normal
distribution), we have that
p(t, T) = E
Q
t,r
_
e
Z(t,T)
_
= e
E
Q
t,r
[Z(t,T)]+
1
2
V
Q
t,r
[Z(t,T)]

ln(p(t, T)) = E
Q
t,r
[Z(t, T)] +
1
2
V
Q
t,r
[Z(t, T)]
=
1
a
_
1 e
a(Tt)
_
. .
B(t,T)
r(t) +
+
_
_
_
_
_
_
_

b
a
(T t)
b
2
a
_
1 e
a(Tt)
_
+
1
2

_
u
t
_
_
T
t
e
a(us)
du
_
2
ds
. .
A(t,T)
_
_
_
_
_
_
_
.
So, the Vasicek model has an ane term structure.
Alternative Solution
From Exercise 22.1 we know that for the Vasicek model, and t < u we have
r(u) = r(t) e
a(ut)
. .
D(u)
+
b
a
_
1 e
a(ut)
_
+
_
u
t
e
a(us)
dW(s)
. .
F(u)
.
where D(u) is deterministic and F(u) is stochastic and both are non depen-
dent on r.
By denition,
p(t, T) = E
Q
t,r
_
e

_
T
t
r(u)du
_
= E
Q
t,r
_
e

_
T
t
[r(t)D(u)+F(u)du]
_
=
_

_
e
_

_
T
t
D(u)du
_
r(t)
. .
B
E
Q
t,r
_
e

_
T
t
F(u)du
_
. .
A
_

_
.
In part A, since
_
T
t
F(u)du is normally distributed (note it is a sum of
37
Wiener increments), we know (from the properties of the normal distribu-
tion) that there exist a A(t, T) such that
E
Q
t,r
_
e

_
T
t
F(u)du
_
= e
A(t,T)
.
From part B we immediatelly see that B(t, T) =
_
T
t
D(u)du.
So, the Vasicek model has an ane term structure.
Exercise 22.3 The problem with the Dothans model is that when r follows
a GBM, like
dr(t) = ar(t)dt +r(t)dW(t),
the solution to the SDE is, for any t < u,
r(u) = r(t)e
(a
1
2

2
)(ut)+(W(u)W(t))
,
i.e., in the Dothans model r is lognormally distributed. Since,
p(t, T) = E
Q
t,r
_
e

_
T
t
r(u)du
_
,
to use the same procedure as in exercise 22.2 we would have to compute
the above expected value, i.e., the expected value of an integral (that is a
sum) of lognormally distributed variables, which is a mess!
Exercise 22.8 Take the follwoing CIR model
dY (t) =
_
2aY (t) +
2
_
dt + 2
_
Y (t)dW(t), Y (0) = y
0
,
Then by Ito Z(t) =
_
Y (t) follows
dZ(t) =
1
2
_
Y (t)

1
2
_
dY (t) +
1
2
_

1
4
_
Y (t)

3
2
_
_
(dY (t))
2
=
1
2
_
2aY (t)
1
2
+
2
Y (t)

1
2
_
dt +dW(t)
1
2

2
Y (t)

1
2
= a Y (t)
1
2
. .
Z(t)
dt +dW(t),
which is a linear diusion.
38
23 Forward Rate Models
Exercise 23.1 We know that the Hull-White model
dr(t) = ((t) ar(t))dt +dW(t)
has an ane term structure, i.e., that
p(t, T) = e
A(t,T)B(t,T)r(t)
,
and that in this model we have in particular, B(t, T) =
1
a
_
1 e
a(Tt)
_
.
Furthermore, f(t, T) =
ln p(t,T)
T
=
A(t,T)
T
+
B(t,T)
T
r(t), so in this case
we have
f(t, T) =
A(t, T)
T
+e
a(Tt)
r(t).
By Ito we have that the dynamics under Q of the forward rates is given by
df(t, T) =

t
_
A(t, T)
T
_
dt +ae
a(Tt)
r(t)dt +e
a(Tt)
dr(t)
=

t
_
A(t, T)
T
_
dt +ae
a(Tt)
r(t)dt +e
a(Tt)
[((t) ar(t))dt +dW(t)]
= (t, T)dt +e
a(Tt)
dW(t),
where (t, T) =

t
_
A(t,T)
T
_
+e
a(Tt)
(t).
Exercise 23.2 Take as given an HJM model (under Q) of the form
df(t, T) = (t, T)dt +(t, T)dW(t),
where the volatility (t, T) is a deterministic function of t and T.
(a) By the HJM drift condition, (t, T) = (t, T)
_
T
t

(t, u)du t, T,
which for deterministic (t, T), means that (t, T) is also determinis-
tic.
39
To see the distribution of the forward rates note that
f(t, T) = f(0, T) +
_
t
0
(s, T)ds +
_
t
0
(s, T)dW(s)
= f(0, T) +
_
t
0
(s, T)
_
T
s

(s, u)duds
. .
(t,T)
+
_
t
0
(s, T)dW(s).
Note that f(0, T) is observable and that the double integral is deter-
ministic, so the only stochastic part is
_
t
0
(s, T)dW(s).
Hence f(t, T) N
_
(t, T),
_
t
0

2
(s, T)ds
_
.
Since r(t) = f(t, t) we immediately have that r(t) is also normally
distributed.
(b) Since we have p(t, T) = exp
_

_
T
t
f(t, s)ds
_
, and in (a) we have al-
ready shown that f(t, s) is normally distributed, then p(t, T) is log-
normally distributed.
Exercise 23.3 Recall that in between the vectors of market-price of risks of
the domestic market and the foreign market there is the following relation:

f
(t) =
d
(t)

x
(t).
where,
x
(t) is the vector of volatilities of in the SDE for the exchange-rate
X(denoted in units of domestic currency per unit of foreign currency).
We also know that any process, so in particular forward rates, has under Q
(the domestic martingale measure), the drift term equal to ((t, T) (t, T)
d
(t)),
where (t, T) is the drift term under the objective probability measure and
that the volatility term remains the same. So, in particular, for the foreign
forward rates must have under the domestic martingale measure, Q,
df
f
(t, T) = (
f
(t, T)
f
(t, T)
d
(t)) dt +
f
(t, T)dW(t),
hence,
f
(t, T) =
f
(t, T)
f
(t, T)
d
(t).
40
Likewise, any process, under Q
f
(the foreign martingale measure),has a drift
term equal to ((t, T) (t, T)
f
(t)).
So, in particular, for the foreign forward rates we have under Q
f
df
f
(t, T) = (
f
(t, T)
f
(t, T)
f
(t))
. .

f
(t,T)
dt +
f
(t, T)dW
f
(t).
So, using the relation between
d
and
f
we can also establish a relation
between
f
and
f
,

f
(t, T) =
f
(t, T)
f
(t, T)
f
(t)
=
f
(t, T)
f
(t, T)
_

d
(t)

x
(t)

=
f
(t, T)
f
(t, T)
d
(t)
. .

f
(t,T)
+
f
(t, T)

x
(t).
We also have that, under the foreign martingale measure, Q
f
, the coe-
cients of the foreign martingale measure must satisfy the standard HJM
drift condition so:

f
(t, T) =
f
(t, T)
_
T
t

f
(t, s)ds.
Using teh relation found between
f
and
f
and using the drift condition
above we get

f
(t, T) =
f
(t, T)
_
T
t

f
(t, s)ds

f
(t, T) +
f
(t, T)

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

f
(t, s)ds

f
(t, T) =
f
(t, T)
_
_
T
t

f
(t, s)ds

x
(t)
_
.
41
24 Change of Numeraire
Exercise 24.1 In the Ho-Lee model we have under Q
dr(t) = (t)r(t)dt +dW(t)
And we know that on this model we have an ane term structure for bond-
prices
p(t, T) = e
A(t,T)B(t,T)r(t)
with B(t, T) = T t.
If the Q-dynamics of p(t, T) are given by
dp(t, T) = r(t)p(t, T)dt +v(t, T)p(t, T)dW(t)
then by Ito we have
v(t, T) = B(t, T) = (T t).
The project is to price an European call option with:
date of maturity T
1
strike price K
where the underlying is a zero-coupon bond with date of maturity T
2
and T
1
< T
2
.
Note that our Z-claim is given by
Z = max [p(T
1
, T
2
) K; 0] ,
Hence, we have using a change of measure on the standard arbitrage pricing
formula that
(t; Z) = E
Q
t
_
e

_
T
1
t
r(s)ds
max [p(T
1
, T
2
) K; 0]
_
42
= p(t, T
1
)E
T
1
t
[max [p(T
1
, T
2
) K; 0]]
= p(t, T
1
)E
T
1
t
_

_
max
_

_
p(T
1
, T
2
)
p(T
1
, T
1
)
. .
Z(T
1
)
K; 0
_

_
_

_
.
Note that dividing p(T
1
, T
1
) in the last step of the equation changes noth-
ing(since p(T, T) = 1 for all T) but has the advantage of seeing our claim
as a claim on the process Z, which we know is a martingale under Q
T
1
, and
as long as it has deterministic volatility the Black-Scholes formula can help
us. To check this note that
Z(t) =
p(t, T
2
)
p(t, T
1
)
can also be written as (just using the fact that we have an ATS)
Z(t) = exp{A(t, T
2
) A(t, T
1
) [B(t, T
2
) B(t, T
1
)] r(t)} ,
and, therefore, has the following dynamics under Q(applying Ito formula)
dZ(t) = { } Z(t)dt +Z(t)
z
(t)dW(t)
where
z
(t) = [B(t, T
2
) B(t, T
1
)] = (T
2
T
1
), will be the same as
under Q
T
1
, and is deterministic.
Since under Q
T
1
,
Z(T
1
) = Z(t) (T
2
T
1
)
_
T
1
t
dW(s)
The conditional (on information at time t) distribution of Z is Z(T
1
)
N
_
Z(t),
2
(T
2
T
1
)
2
(T
1
t)
_
.
So, (from the BS formula)
(t) = p(t, T
1
) {Z(t)N [d
1
(t, T
1
)] KN [d
2
(t, T
1
]}
= p(t, T
2
)N [d
1
(t, T
1
)] p(t, T
1
)KN [d
2
(t, T
1
]
43
where
d
1
(t, T
1
) =
ln
Z(t)
K
+
1
2

2
(T
2
T
1
)
2
(T
1
t)
_

2
(T
2
T
1
)
2
(T
1
t)
=
ln
p(t,T
2
)
Kp(t,T
1
)
+
1
2

2
(T
2
T
1
)
2
(T
1
t)
_

2
(T
2
T
1
)
2
(T
1
t)
and d
2
(t, T
1
) = d
1
(t, T
1
)
_

2
(T
2
T
1
)
2
(T
1
t).
Exercise 24.2 Take as given an HJM model of the form
df(t, T) = (t, T) +(t, T)dW(t) +
2
e
a(Tt)
dW
2
(t)
where (t, T) = [
1
(T t)
2
e
a(Tt)
], W(t) =
_
W
1
(t)
W
2
(t)
_
and we have W
1
and W
2
independent Wiener processes and
1
and
2
constants.
(a) From the relation between the dynamics of bond prices and forward
rates we know that
dp(t, T) = r(t)p(t, T)dt +p(t, T)S(t, T)dW(t)
with S(t, T) =
_
T
t
(t, s)ds.
In this case we have then
S(t, T) =
_
T
t
[
1
(s t)
2
e
a(st)
] ds
=
_

1
2
(T t)
2
. .

p
1
(t,T)

2
a
_
1 e
a(Tt)
_
. .

p
2
(t,T)
_
So,
dp(t, T) = r(t)p(t, T)dt +p(t, T)S(t, T)dW(t)
= r(t)p(t, T)dt +p(t, T) [

1
2
(T t)
2

2
a
_
1 e
a(Tt)
_
]
_
dW
1
(t)
dW
2
(t)
_
= r(t)p(t, T)dt +p(t, T) [
p
1
(t, T)
p
2
(t, T) ]
_
dW
1
(t)
dW
2
(t)
_
44
(b) We will use exactly the same technique as in exercise 24.1 to price an
European call option with maturity T
0
on a T
1
-bond.
Note that in this model the Z process Z(t) =
p(t,T
1
)
p(t,T
0
)
is a martingale
under Q
T
0
and its dynamics has deterministic volatility given by

T
1
,T
0
(t) = S(t, T
1
) S(t, T
0
) =
_
T
1
T
0
[
1
(s t)
2
e
a(st)
] ds
Hence (by the same arguments as in exercise 24.1)
Z(T
0
) N
_

_
Z(t),
_
T
0
t
||
T
1
,T
0
(s)||
2
ds
. .

2
T
0
,T
1
_

_
and the pricing formula is given by
(t) = p(t, T
0
) {Z(t)N [d
1
(t, T
0
)] KN [d
2
(t, T
0
]}
= {p(t, T
1
)N [d
1
(t, T
0
)] p(t, T
0
)KN [d
2
(t, T
0
]}
where
d
1
(t, T
0
) =
ln
Z(t)
K
+
1
2

2
T
0
,T
1
_

2
T
0
,T
1
=
ln
p(t,T
2
)
Kp(t,T
1
)
+
1
2

2
T
0
,T
1
_

2
T
0
,T
1
and d
2
(t, T
1
) = d
1
(t, T
1
)
_

2
T
0
,T
1
.
45