Hedging Exercises
Exercise 13.1. The price evolution of two assets are modelled with Geometrical Brownian motions :
dS
1(t) =
1S
1(t) dt +
1S
1(t) dW
1(t) ;
dS
2(t) =
2S
2(t) dt +
2S
2(t) dW
1(t) +
2p
1
2S
2(t) dW
2(t)
where W
1and W
2are two independent ( ; F ; fF
t: t 0 g ; P ) Brownian motions. We as- sume that the risk free rate r is constant. fF
t: t 0 g is the …ltration generated by the two Brownian motion (with the usual regularity conditions.
The goal is to hedge the contingent claim C:
a) What is the risk neutral market model.
b) Explain why C is attainable.
c) Assume that the time t value of the contingent claim C is a function of the under- lying asset prices and time, that is, f (t; S
1(t) ; S
2(t)) : Express the discounted contingent claim value as a di¤usion process. More precisely,we search for the coe¢ cient a; b and c such that
t
f (t; S
1(t) ; S
2(t)) =
0f (0; S
1(0) ; S
2(0)) +
Z
t 0a (s) ds + Z
t0
b (s) df W
1(s) + Z
t0
c (s) df W
2(s)
where W f
1, f W
2are two independent ( ; F ; fF
t: t 0 g ; Q ) Brownian motions. If you have to add some hypothesis, mention it.
d) Use the Martingale Representation Theorem to justify the existence of predictable processes and such that
t
f (t; S
1(t) ; S
2(t)) = f (0; S
1(0) ; S
2(0)) +
Z
t 0u
d
uS
1(u) + Z
t0
u
d
uS
2(u) : Why are you allowed to use this theorem?
e) Use the results obtained in c) and d) to get
tand
tin therms of partial derivatives of f and the market model components. More precisely, show that
t
= @f
@s
1(t; S
1(t) ; S
2(t)) and
t= @f
@s
2(t; S
1(t) ; S
2(t))
f) Construct the hedging strategy for C, that is, the self-…nancing investment strategy with a time T value of C:
g) If the contingent claim C = S
1(T ) S
2(T ) ; then what is the hedging strategy?
Exercise 13.2. Let
X
t= price of the underlying asset at time t and B
t= bank account value at time t;
B
0= 1:
The model is :
dX
t= ( X
t) dt + dW
tdB
t= r
tB
tdt
where the instantaneous risk free rate f r
t: t 0 g is a time dependent deterministic variable.
a) What is the risk-neutral market model? Justify each step.
b) Find the replicating strategy with a time t value of f (X
t; B
t; t).
Exercise 13.3. The risky asset price dynamics satisfy
dS (t) = r (t) S (t) dt + (t) S (t) dW (t)
where f W (t) : t 0 g is a ( ; F ; fF
t: t 0 g ; Q ) Brownian motion, Q is a martingale mea- sure and f r (t) : t 0 g and f (t) : t 0 g are fF
t: t 0 g predictable processes (careful ! S is not a Geometrical Brownian motion). The time t value B (t) of the bank account satisfy
B (t) = exp Z
t0
r (s) ds ; B (0) = 1:
a) Find an expression for the time t value P (t; T ) of a zero-coupon bond paying one dollar at time T in terms of the risk free rate f r (t) : t 0 g : (Easy question that do not require computation).
Forward contract
A forward contract set at time t and with a maturity of T determines the price at which
the underlying contract will be traded at maturity. The convention says that the delivery
price is the one that make nil the initial value of the forward contract. The delivery price
or T f orward price, determine at time t, is denoted F (t) : It is a F
tmeasurable random
b) Determine F (t).
c) Determine the value V (t) at time t of the forward contract established at time 0:
d) Assuming that the interest rate r is a deterministic function of time, determine the hedging strategy of the forward contract set at time 0:
Change of numeraire
Any asset with positive price may be used as numeraire. The bank account can serve as numeraire. In this case, the risky asset worth
B(t)S(t)shares of the bank account.
The probability measure Q
Nis said to be risk neutral for the numeraire N if the price of any asset, divided by the numeraire value, is a Q
Nmartingale. The risk neutral measure Q is a risk neutral measure for the bank account B.
Let
Q b (A) = 1
N (0) E
QN (T )
B (T ) I
Afor each event A 2 F
Twhere I
Ais the indicator function that worth 1 if ! 2 A and zero otherwise.
e) Show that Q b is a probability measure.
f) Knowing that for all F
Tmeasurable random variable X , E
Qb[ X j F
t] = N (0) B (t)
N (t) E
QX N (T )
N (0) B (T ) F
t; show that Q b is risk neutral for the numeraire N .
g) Consider the case where the numeraire is the zero-coupon bond of maturity T .
Determine the value of the risky asset S in terms of this numeraire and justify intuitively why
the risk neutral measure Q
P( ;T)for the numeraire P ( ; T ) is called the measure T forward.
Solutions
1 Exercise 13.1
a) Under the risk neutral measure
dS
1(t) = rS
1(t) dt +
1S
1(t) df W
1(t) ;
dS
2(t) = rS
2(t) dt +
2S
2(t) df W
1(t) +
2p
1
2S
2(t) df W
2(t) :
b) The martingale measure is unique, son the market model is complete. All contin-
gent claims are attainable.
c) If the function f is twicely continuously di¤erentiable, then df (t; S
1(t) ; S
2(t))
= @f
@t (t; S
1(t) ; S
2(t)) dt + @f
@s
1(t; S
1(t) ; S
2(t)) dS
1(t) + @f
@s
2(t; S
1(t) ; S
2(t)) dS
2(t) + 1
2
@
2f
@s
21(t; S
1(t) ; S
2(t)) d h S
1i (t) + 1 2
@
2f
@s
22(t; S
1(t) ; S
2(t)) d h S
2i (t) + @
2f
@s
1@s
2(t; S
1(t) ; S
2(t)) d h S
1; S
2i (t)
= @f
@t (t; S
1(t) ; S
2(t)) dt + @f
@s
1(t; S
1(t) ; S
2(t)) rS
1(t) dt + @f
@s
1(t; S
1(t) ; S
2(t))
1S
1(t) df W
1(t) + @f
@s
2(t; S
1(t) ; S
2(t)) rS
2(t) dt + @f
@s
2(t; S
1(t) ; S
2(t))
2S
2(t) df W
1(t) + @f
@s
2(t; S
1(t) ; S
2(t))
2p
1
2S
2(t) df W
2(t) + 1
2
@
2f
@s
21(t; S
1(t) ; S
2(t))
21S
12(t) dt + 1 2
@
2f
@s
22(t; S
1(t) ; S
2(t))
22S
22(t) dt + @
2f
@s
1@s
2(t; S
1(t) ; S
2(t))
1 2S
1(t) S
2(t) dt
= 2 6 6 6 4
@f
@t
(t; S
1(t) ; S
2(t)) +
@s@f1
(t; S
1(t) ; S
2(t)) rS
1(t) +
@s@f2
(t; S
1(t) ; S
2(t)) rS
2(t) +
12@@s2f21
(t; S
1(t) ; S
2(t))
21S
12(t) +
12@@s2f22
(t; S
1(t) ; S
2(t))
22S
22(t) +
@s@2f1@s2
(t; S
1(t) ; S
2(t))
1 2S
1(t) S
2(t)
3 7 7 7 5 dt
+ @f
@s
1(t; S
1(t) ; S
2(t))
1S
1(t) + @f
@s
2(t; S
1(t) ; S
2(t))
2S
2(t) df W
1(t) + @f
@s
2(t; S
1(t) ; S
2(t))
2p
1
2S
2(t) df W
2(t) : The multiplication rule leads to
d
tf (t; S
1(t) ; S
2(t)) =
tdf (t; S
1(t) ; S
2(t)) + f (t; S
1(t) ; S
2(t)) d
t=
tdf (t; S
1(t) ; S
2(t)) rf (t; S
1(t) ; S
2(t))
tdt
a (t) =
t2 6 6 6 6 6 4
@f
@t
(t; S
1(t) ; S
2(t)) +
@s@f1
(t; S
1(t) ; S
2(t)) rS
1(t) +
@s@f2
(t; S
1(t) ; S
2(t)) rS
2(t) +
12@@s2f21
(t; S
1(t) ; S
2(t))
21S
12(t) +
12@@s2f2 2(t; S
1(t) ; S
2(t))
22S
22(t) +
@s@2f1@s2
(t; S
1(t) ; S
2(t))
1 2S
1(t) S
2(t) rf (t; S
1(t) ; S
2(t))
3 7 7 7 7 7 5
b (t) =
t@f
@s
1(t; S
1(t) ; S
2(t))
1S
1(t) + @f
@s
2(t; S
1(t) ; S
2(t))
2S
2(t) c (t) =
t@f
@s
2(t; S
1(t) ; S
2(t))
2p
1
2S
2(t) :
d) The discounted value of the contingent claim is a Q martingale. Moreover, the discounted value of any primary asset are also Q martingales with positive di¤usion coe¢ cients, that is,
t
S
1(t) = S
1(0) + Z
t0
1 u
S
1(u)
| {z }
>0
df W
1(u)
t
S
2(t) = S
2(0) + Z
t0
2 u
S
2(u)
| {z }
>0
df W
1(u) + Z
t0 2
p 1
2 uS
2(u)
| {z }
>0
df W
2(u) :
Moreover,
E
QZ
T0 2 1
2
u
S
12(u) du
=
21Z
T0
E
Q 2uS
12(u) du
=
21Z
T0
S
12(0) E
Q"
exp ( rT ) exp r
2 1
2 u +
1W
u2
# du
=
21Z
T0
S
12(0) E
Qexp
21u + 2
1W
udu
=
21S
12(0) Z
T0
exp
21u + 2
21u du
=
21S
12(0) Z
T0
exp
21u du = S
12(0) e
21T1 < 1 :
Similarly,
E
QZ
T0 2 2 2 2
u
S
22(u) du < 1 and E
QZ
T 02
2
1
2 2uS
22(u) du < 1 :
Consequently, the Martingale Representation Theorem implies the existence of predictable processes and such that
t
f (t; S
1(t) ; S
2(t))
=
0f (0; S
1(0) ; S
2(0)) +
Z
t 0u
d
uS
1(u) + Z
t0
u
d
uS
2(u)
= f (0; S
1(0) ; S
2(0)) +
Z
t 0u 1 u
S
1(u) df W
1(u) +
Z
t 0u 2 u
S
2(u) df W
1(u) + Z
t0 u 2
p 1
2 uS
2(u) df W
2(u)
= f (0; S
1(0) ; S
2(0)) +
Z
t 0[
u 1 uS
1(u) +
u 2 uS
2(u)] df W
1(u) +
Z
t 0u 2
p 1
2 uS
2(u) df W
2(u) :
e) Since the equations in c) and in d) are both di¤usion processes for the discounted contingent claim value,
tf (t; S
1(t) ; S
2(t)), then their di¤usion coe¢ cients must be equal.
Therefore, the coe¢ cients of W f
2have to be equal :
t
@f
@s
2(t; S
1(t) ; S
2(t))
2p
1
2S
2(t) =
t 2p
1
2 tS
2(t) which is veri…ed if and only if
t
= @f
@s
2(t; S
1(t) ; S
2(t)) : The coe¢ cients of f W
1have to be equal:
t
@f
@s
1(t; S
1(t) ; S
2(t))
1S
1(t) + @f
@s
2(t; S
1(t) ; S
2(t))
2S
2(t)
= [
t 1 tS
1(t) +
t 2 tS
2(t)] :
This equality is satis…ed if and only if
t
=
@f
@s1
(t; S
1(t) ; S
2(t))
1S
1(t) +
@s@f2
(t; S
1(t) ; S
2(t))
2S
2(t)
t 2S
2(t)
1
S
1(t)
=
@f
@s1
(t; S
1(t) ; S
2(t))
1S
1(t) +
@s@f2
(t; S
1(t) ; S
2(t))
2S
2(t)
@s@f2
(t; S
1(t) ; S
2(t))
2S
2(t)
1
S
1(t)
= @f
@s
1(t; S
1(t) ; S
2(t))
f) Let
t
= the number of share of asset 1
t
= the number of share of asset 2
and
t= the number of share of the riskless asset.
In our case, the riskless asset is B (t) = exp (rt) : We have shown that
t
= @f
@s
1(t; S
1(t) ; S
2(t)) and
t= @f
@s
2(t; S
1(t) ; S
2(t)) : It su¢ ces to set
t
=
tf (t; S
1(t) ; S
2(t))
t tS
1(t)
t tS
2(t)
=
tf (t; S
1(t) ; S
2(t)) S
1(t) @f
@s
1(t; S
1(t) ; S
2(t)) S
2(t) @f
@s
2(t; S
1(t) ; S
2(t)) :
g) Since C = S
1(T ) S
2(T ) ; then its time t value is f (t; S
1(t) ; S
2(t))
= 1
t
E
Q[
TS
1(T ) S
2(T ) jF
t]
= e
rtE
Q2 6 6 4
e
rTS
1(t) exp h
r
221(T t) +
1(W
1(T ) W
1(t)) i S
2(t) exp
"
r
222(T t) +
2(W
1(T ) W
1(t)) +
2p
1
2(W
2(T ) W
2(t))
# jF
t3 7 7 5
= exp [ r (T t)] S
1(t) S
2(t) E
Q2 6 6 4
exp h
2r
221 222(T t) i exp [(
1+
2) (W
1(T ) W
1(t))]
exp h
2
p 1
2(W
2(T ) W
2(t)) i jF
t3 7 7 5
= exp r
2 1
2
2 2
2 (T t) S
1(t) S
2(t) E
Q[exp [(
1+
2) (W
1(T ) W
1(t))]] E
Qh
exp h
2
p 1
2(W
2(T ) W
2(t)) ii
= exp r
2 1
2
2 2
2 (T t) S
1(t) S
2(t) exp
"
(
1+
2)
2(T t) 2
# exp
2
2
(1
2)
2 (T t)
= S
1(t) S
2(t) exp [(r +
1 2) (T t)] : Consequently,
t
= @f
@s
1(t; S
1(t) ; S
2(t)) = S
2(t) exp [(r +
1 2) (T t)]
t
= @f
@s
2(t; S
1(t) ; S
2(t)) = S
1(t) exp [(r +
1 2) (T t)]
and
t
=
tf (t; S
1(t) ; S
2(t))
t tS
1(t)
t tS
2(t)
= exp ( rt) S
1(t) S
2(t) exp [(r +
1 2) (T t)]
S
2(t) exp [(r +
1 2) (T t)] exp ( rt) S
1(t) S
1(t) exp [(r +
1 2) (T t)] exp ( rt) S
2(t)
= exp ( rt) S
1(t) S
2(t) exp [(r +
1 2) (T t)] :
Veri…cation. At time T , we hold S
2(T ) shares of asset 1, S
1(T ) shares of asset 2 and we are short of exp ( rt) S
1(T ) S
2(T ) shares of the bank account. The time t value of this investment strategy is
V
t=
tS
1(t) +
tS
2(t) +
tB (t)
= S
2(t) exp [(r +
1 2) (T t)] S
1(t) + S
1(t) exp [(r +
1 2) (T t)] S
2(t) exp ( rt) S
1(t) S
2(t) exp [(r +
1 2) (T t)] exp ( rt)
= S
1(t) S
2(t) exp [(r +
1 2) (T t)] : In particular,
V
T= S
1(T ) S
2(T ) :
Other veri…cation. a (t) has to be equal to 0. But 2
6 6 6 6 6 4
@f
@t
(t; S
1(t) ; S
2(t)) +
@s@f1
(t; S
1(t) ; S
2(t)) rS
1(t) +
@s@f2
(t; S
1(t) ; S
2(t)) rS
2(t) +
12@@s2f21
(t; S
1(t) ; S
2(t))
21S
12(t) +
12@@s2f22
(t; S
1(t) ; S
2(t))
22S
22(t) +
@s@2f1@s2
(t; S
1(t) ; S
2(t))
1 2S
1(t) S
2(t) rf (t; S
1(t) ; S
2(t))
3 7 7 7 7 7 5
= 2 6 6 6 6 4
(r +
1 2) s
1s
2exp [(r +
1 2) (T t)]
+s
2exp [(r +
1 2) (T t)] rs
1+ s
1exp [(r +
1 2) (T t)] rs
20 + 0
+ exp [(r +
1 2) (T t)]
1 2s
1s
2rs
1s
2exp [(r +
1 2) (T t)]
3 7 7 7 7 5
= 2 6 6 4
(r +
1 2) s
1s
2exp [(r +
1 2) (T t)]
+rs
1s
2exp [(r +
1 2) (T t)] + rs
1s
2exp [(r +
1 2) (T t)]
+
1 2s
1s
2exp [(r +
1 2) (T t)]
rs
1s
2exp [(r +
1 2) (T t)]
3 7 7 5
= 0:
2 Exercise 13.2
a) The multiplication rule implies that
dB
t 1X
t= B
t 1dX
t+ X
tdB
t 1+ d X; B
1 t= B
t 1( ( X
t) dt + dW
t) + X
tr
tB
t1dt
= B
t 1(r
t+ ) B
t 1X
tdt + dW
tLet us introduce the process : W
t= W
t+ R
t0 s
ds and
dB
t 1X
t= (
t) B
t 1(r
t+ ) B
t 1X
tdt + B
t 1d W
t+ Z
t0 s
ds
= (
t) B
t 1(r
t+ ) B
t 1X
tdt + dW
t: To cancel the drift term,
(
t) B
t 1(r
t+ ) B
t1X
t= 0 ,
t= (r
t+ ) X
t+ :
Assuming that the Novikov condition
1E
Ph
exp
12R
T 02 t
dt i
is satis…ed, we can apply Girsanov theorem and state that there exists a measure Q under which W is a Brownian
1
EP
"
exp 1 2
Z T 0
2 tdt
!#
= EP
"
exp 1 2
Z T 0
(rt+ )Xt
+
2
dt
!#
= EP
"
exp 1 2
Z T 0
(rt+ )Xt 2
2(rt+ )Xt
+
2! dt
!#
= exp 1 2
2
T
! EP
"
exp 1 2
Z T 0
(rt+ )Xt 2
2(rt+ )Xt ! dt
!#
exp 1 2
2
T
! vuutEP
"
exp Z T
0
(rt+ )Xt 2
dt
!#vuutEP
"
exp 2 Z T
0
(rt+ )Xt
dt
!#
(Cauchy-Schwartz inequality)
= exp 1 2
2
T
! vuutEP
"
exp Z T
0
(rt+ )Xt 2
dt
!#vuutEP
"
exp 2 Z T
0
(rt+ )Xt
dt
!#
TO BE COMPLETED
motion. Let …nd the SDE of X in function of the Q Brownian motion:
dX
t= ( X
t) dt + dW
t= ( X
t) dt + d W
tZ
t0 s
ds
= (
tX
t) dt + dW
t= (r
t+ ) X
t+ X
tdt + dW
t= r
tX
tdt + dW
t: b) According to Itô’s lemma
df (X
t; B
t; t)
= @f
@x (X
t; B
t; t) dX
t+ @f
@b (X
t; B
t; t) dB
t+ @f
@t (X
t; B
t; t) dt + 1
2
@
2f
@x
2(X
t; B
t; t) d h X i
t+ 1 2
@
2f
@b
2(X
t; B
t; t) d h B i
t+ @
2f
@x@b (X
t; B
t; t) d h X; B i
t= @f
@x (X
t; B
t; t) (r
tX
tdt + dW
t) + @f
@b (X
t; B
t; t) r
tB
tdt + @f
@t (X
t; B
t; t) dt + 1
2
@
2f
@x
2(X
t; B
t; t)
2dt
= r
tX
t@f
@x (X
t; B
t; t) + r
tB
t@f
@b (X
t; B
t; t) + @f
@t (X
t; B
t; t) +
2
2
@
2f
@x
2(X
t; B
t; t) dt + @f
@x (X
t; B
t; t) dW
tdB
t 1f (X
t; B
t; t)
= r
tB
t1X
t@f@x(X
t; B
t; t) + r
t@f@b(X
t; B
t; t) + B
t1@f@t(X
t; B
t; t)
+
22B
t 1@@x2f2(X
t; B
t; t) r
tB
t1f (X
t; B
t; t) dt + B
t 1@f
@x (X
t; B
t; t) dW
tAccording the Martingale Representation Theorem,there is a predictable process such that
dB
t 1f (X
t; B
t; t) =
tdB
t1X
t=
tB
t 1dW
t1
Therefore
0 = r
tB
t1X
t@f
@x (X
t; B
t; t) + r
t@f
@b (X
t; B
t; t) + B
t1@f
@t (X
t; B
t; t) +
2
2 B
t 1@
2f
@x
2(X
t; B
t; t) r
tB
t1f (X
t; B
t; t) ; and
t
B
t 1= B
t 1@f
@x (X
t; B
t; t) :
The last equation implies that the number of shares of the risky asset is
t
= @f
@x (X
t; B
t; t) : The number of shares of the riskless asset is,
t
= B
t1f (X
t; B
t; t) @f
@x (X
t; B
t; t) B
t 1X
t:
3 Exercise 13.3
3.1 Solution a)
P (t; T ) = E
QB (t) B (T ) F
t= E
Qexp
Z
T tr (s) ds F
t:
3.2 Solution b)
The delivery price F (t) satis…es E
QB (t)
B (T ) (S (T ) F (t)) F
t= 0:
But,
0 = E
QB (t)
B (T ) (S (T ) F (t)) F
t= B (t) E
QS (T )
B (T ) F
tB (t) F (t) E
Q1 B (T ) F
t= B (t) S (t)
B (t) B (t) F (t) E
Q1
B (T ) F
twhich implies that
F (t) = S (t) E
Qh
B(t)B(T)
F
ti = S (t) P (t; T ) :
3.3 Solution c)
V (t) = B(t)E
QS (T ) F (0) B (T ) F
t= B(t)E
Q"
S (T )
PS(0)(0;T)B (T ) F
t#
= B(t) E
QS (T )
B (T ) F
tS (0)
P (0; T ) E
Q1 B (T ) F
t= B(t) S (t) B (t)
S (0) P (0; T )
P (t; T )
B (t) since S
B is a Q martingale
= S (t) S (0)
P (0; T ) P (t; T ) :
Therefore, V (0) = S (0)
PS(0)(0;T)P (0; T ) = 0 (the initial contract value is nil) and V (T ) = S (T )
PS(0)(0;T)= S (T ) F (0) (At maturity, the contract value corresponds to its cash ‡ow).
3.4 Solution d)
Assuming deterministic interest rate, the time t value of the forward contract is V (t) = S (t) S (0)
P (0; T ) P (t; T )
= S (t) S (0) exp Z
t0
r (s) ds
= S (t) S (0) B (t) :
It is a twicely continuously di¤erentiable function of S and the bank account B. To replicates
this contract, it su¢ ces to hold one share of the risky asset and be short of S (0) shares of
the bank account.
3.5 Solution e)
First, note that Q b (A) 0 since N (T ) > 0, B (T ) > 0, N (0) > 0 and I
A0: Second, Q b ( ) = 1
N (0) E
QN (T )
B (T ) = 1 N (0)
N (0) B (0) = 1
where the second equality is justify by the fact that Q is risk neutral for the bank account, which implies that N=B is a Q martingale. Finally, if A
1; A
2; ::: are disjoint events
Q b [
1 i=1A
i!
= 1
N (0) E
QN (T )
B (T ) I
S1i=1Ai= X
1i=1
1
N (0) E
QN (T )
B (T ) I
Ai= X
1i=1
Q b (A
i) :
3.6 Solution f )
Let f Y (t) : t 0 g be the value of an asset of the model. We have that Y =B is a Q martingale.
We want to show that Y =N is a Q b martingale. First, since Y and N are fF
t: t 0 g adapted, Y =N is. Second, we need to check the integrability condition : E
Qbh
Y(t) N(t)