You are on page 1of 4

# Stochastic Calculus for Finance I

-
some Solutions to Chapter V
Matthias Thul

September 4, 2010
Exercise 5.5
(i) We know that the symmetric random walk M
n
is a martingale. Assume that we ﬁrst
reach the level m at time τ
m
≤ n. Due to the martingale property and conditional
on reaching level m at τ
m
, all paths leading to levels M
n
= m − α have the same
probability as the paths leading to M
n
= m + α with alpha ∈ {0, 2, . . . , n − τ
m
}.
Obviously, all paths leading to M
n
+ α also have τ
m
≤ n. Thus, the two sets
{W
m
≥ m∧ W
n
= m−α} and {τ
m
≤ n ∧ W
n
= m−α} are equivalent and have the
same probability as {W
n
= m + α}. Now let b = m−α, then α = m−b and we get
P{τ
m
≤ n ∧ W
n
= b} = P{W
n
= 2m−b}
The values of M
n
are binomially distributed with
P{M
n
= k} =

n
k

1
2

n
where k ∈ {0, 1, . . . , n}. At step n, M
n
= n is thus represented by either k
M
n
=n
= 0
or k
M
n
=n
= n. Similarly, M
n
= −n refers to the index k
M
n
=−n
= n−k
M
n
=n
. Without
loss of generality, we choose k
M
n
=n
= n. I.e. k represents the number of up-steps.
We then get for k and n −k

The author can be contacted via <<firstname>>.<<lastname>>@googlemail.com and
http://www.matthiasthul.com.
1
k =
n + 2m−b)
2
=
n −b
2
+ m
n −k =
n + b
2
−m
Plugging back into the binomial density and using all previous results yields
P{W
m
≥ m∧ W
n
= b} =

n!

n−b
2
+ m

!

n+b
2
−m

!

1
2

n
q.e.d.
(ii) If the random walk is asymmetric, then it is no longer a martingale. Conditional on
τ
m
≤ n, the paths leading to W
n
= m−α have probability
P{τ
m
≤ n ∧ W
n
= m−α} = P{W
τ
m
= m}

ˆ n
ˆ
k

p
ˆ
k
q
(ˆ n−
ˆ
k)
Here, ˆ n = n − τ
n
is the number of steps between the ﬁrst hitting time τ
m
of level
m and n. Again, we have used the fact that α ∈ {0, 2, . . . , n − τ
m
}. α = 0 is thus
equivalent to an equal amount of up- and down-steps between τ
m
and n and thus
translates into
ˆ
k =
n−τ
m
2
. Similarly, α = n − τ
m
translates into
ˆ
k = 0. In general,
we thus get for the number of up-steps
ˆ
k required from M
τ
m
to m−α:
ˆ
k =
n−τ
m
−α
2
.
We now analyse the probability of the reﬂected path. Conditional on τ
m
≤ n, the
n
= m + α have probability
P{W
n
= m + α} = P{τ
m
≤ n ∧ W
n
= m + α} = P{W
τ
m
= m}

˜ n
˜
k

p
ˆ
k
q
(ˆ n−
ˆ
k)
Here, ˜ n = ˆ n and by similar arguments as used to construct
ˆ
k, we ﬁnd that
˜
k =
n−τ
m

2
. We then see that

ˆ n
ˆ
k

=
ˆ n!
ˆ
k!(ˆ n −
ˆ
k)!
=
(n −τ)!

n−τ
m
−α
2

!

n−τ
m

2

!
=
˜ n
(˜ n −
˜
k)!
˜
k!
=

˜ n
˜
k

and
p
˜
k
q
(˜ n−
˜
k)
= p
(
˜ n+α
2
)
q
(
˜ n−α
2
)
= p
(
˜ n−α
2
)
p
α
q
(
˜ n+α
2
)
q
−α
=

p
α
q
−α

p
ˆ
k
q
ˆ n−
ˆ
k

2
Thus
P{W
n
= m + α} = P{τ
m
≤ n ∧ W
n
= m + α}
= P{W
τ
m
= m}

˜ n
˜
k

p
ˆ
k
q
(ˆ n−
ˆ
k)
= P{W
τ
m
= m}

ˆ n
ˆ
k

p
α
q
−α

p
ˆ
k
q
ˆ n−
ˆ
k

=

p
α
q
−α

P{τ
m
≤ n ∧ W
n
= m−α}
Substituting b = m− α yields the “modiﬁed” reﬂection principle for a asymmetric
random walk
P{τ
m
≤ n ∧ W
n
= b} =

p
m−b
q
b−m

P{W
n
= 2m−b}
=

p
m−b
q
b−m

n
n−b
2
+ m

p
(
n−b
2
+m
)
q
(
n+b
2
−m
)
=

n!

n−b
2
+ m

!

n+b
2
−m

!

p
(
n+b
2
)
q
(
n−b
2
)
As expected, for p = q =
1
2
the above equation simpliﬁes to the result from (i).
Exercise 5.9
(i) We ﬁrst substitute 2
p
for v(s) to get
s
p
=
2
5
2
p
s
p
+
2
5

1
2

p
s
p
⇔ s
p

1 −
2
5
2
p

2
5

1
2

p

= 0
We thus need to solve for
1 −
2
5
2
p

2
5

1
2

p
= 0
We make a substitution and set x = 2
p
and multiply by −
5
2
x to get
x
2

5
2
x + 1 = 0 ⇒ x ∈

2,
1
2

⇒ p ∈ {1, −1} q.e.d.
where we have solved the quadratic equation by standard methods.
3
(ii)
lim
s→∞
v(s) = lim
s→∞
¸
As +
B
s

=

0 A = 0
sgn(A) · ∞ otherwise
q.e.d.
Here, we have used that lim
s→∞
B
s
= 0.
(iii) After some simple transformations, we get
B
s
−(4 −s) = 0 ⇔ (s −2)
2
= 4 −B
This equation obviously only has real solutions if 4 −B ≥ 0 or B ≤ 4 such that the
square-root is deﬁned. For B ≤ 4, the solutions are s = 2 ±

4 −B and especially
s = 2 if B = 4.
(iv) While the lower part of the piecewise deﬁned function f
l
(s) = 4 −s is independent
of B, the upper part f
u
(s) =
B
s
increases with B. We thus choose B = 4, i.e.
set it to the max. value it can attain. We furthermore note, that the function
f
u
(s) =
4
s
≥ f
l
(s) = 4 − s for all s > 0 and thus especially for all attainable s.
A ﬁrst idea would thus be to max. the option value by choosing s
B
as small as
possible. But this does not guarantee that f
u
(s
B
) = f
l
(s
B
). We thus choose s
B
in
accordance with (iii) such that
4 −s
B
=
4
s
B
⇔ s
B
= 2
(v) Give B = 4, we equate derivatives and solve for s
B
to obtain
−1 = −
4
s
2
B
⇔ s
B
= ±2
This shows that for s
B
= 2, both the function values and the derivatives coincide
(smooth pasting condition).
4

n − τm }. In general. n = n − τn is the number of steps between the ﬁrst hitting time τm of level ˆ m and n. we ﬁnd that k = ˜ ˆ n−τm +α . we have used the fact that α ∈ {0. Conditional on τm ≤ n. . the paths leading to Wn = m + α have probability n ˆ n ˆ ˜ P {Wn = m + α} = P {τm ≤ n ∧ Wn = m + α} = P {Wτm = m} ˜ pk q (ˆ −k) k ˆ ˜ Here. 2 ˆ Similarly. n−τm −α . α = n − τm translates into k = 0.e. 2 ˆ ˆ we thus get for the number of up-steps k required from Mτm to m − α: k = We now analyse the probability of the reﬂected path.n−b n + 2m − b) = +m 2 2 n+b n−k = −m 2 k = Plugging back into the binomial density and using all previous results yields n! + m ! n+b − m ! 2 1 2 n P {Wm ≥ m ∧ Wn = b} = n−b 2 q. α = 0 is thus equivalent to an equal amount of up. Again. 2 We then see that n ˆ ˆ k n! ˆ = ˆ n − k)! k!(ˆ ˆ (n − τ )! n ˜ = = n−τm +α ! ! (˜ − k)!k! n ˜ ˜ 2 n ˜ ˜ k = n−τm −α 2 and n pk q (˜ −k) = p( ˜ ˜ n+α ˜ 2 ˜ ˜ ˜ ) q ( n−α ) = p( n−α ) pα q ( n+α ) q −α = pα q −α 2 2 2 ˆ pk q n−k ˆ ˆ 2 . then it is no longer a martingale. 2.d. . n = n and by similar arguments as used to construct k. . .and down-steps between τm and n and thus ˆ translates into k = n−τm . the paths leading to Wn = m − α have probability n ˆ n ˆ ˆ P {τm ≤ n ∧ Wn = m − α} = P {Wτm = m} ˆ pk q (ˆ −k) k Here. (ii) If the random walk is asymmetric. Conditional on τm ≤ n.

e. 3 . where we have solved the quadratic equation by standard methods.9 (i) We ﬁrst substitute 2p for v(s) to get 2 2 sp = 2p sp + 5 5 We thus need to solve for 2 2 1 − 2p − 5 5 1 2 p 1 2 p sp ⇔ 2 2 sp 1 − 2p − 5 5 1 2 p =0 =0 We make a substitution and set x = 2p and multiply by − 5 x to get 2 5 x2 − x + 1 = 0 2 1 2 ⇒ x∈ 2.d. ⇒ p ∈ {1.Thus P {Wn = m + α} = P {τm ≤ n ∧ Wn = m + α} n ˆ n ˆ ˜ = P {Wτm = m} ˜ pk q (ˆ −k) k n ˆ ˆ ˆ ˆ = P {Wτm = m} ˆ pα q −α pk q n−k k α −α = p q P {τm ≤ n ∧ Wn = m − α} Substituting b = m − α yields the “modiﬁed” reﬂection principle for a asymmetric random walk P {τm ≤ n ∧ Wn = b} = = = pm−b q b−m P {Wn = 2m − b} n+b n−b n p( 2 +m) q ( 2 −m) pm−b q b−m n−b +m 2 n−b 2 n! + m ! n+b − m ! 2 p( n+b 2 ) q ( n−b ) 2 As expected. Exercise 5. −1} q. for p = q = 1 2 the above equation simpliﬁes to the result from (i).

i. the option value by choosing sB as small as possible. Here. both the function values and the derivatives coincide (smooth pasting condition). the solutions are s = 2 ± 4 − B and especially s = 2 if B = 4. We thus choose sB in accordance with (iii) such that 4 sB 4 − sB = ⇔ sB = 2 (v) Give B = 4. But this does not guarantee that f u (sB ) = f l (sB ). 4 .e. set it to the max. we equate derivatives and solve for sB to obtain 4 s2 B −1 = − ⇔ sB = ±2 This shows that for sB = 2. the upper part f u (s) = B s increases with B. A ﬁrst idea would thus be to max. that the function f u (s) = 4 s ≥ f l (s) = 4 − s for all s > 0 and thus especially for all attainable s. (iv) While the lower part of the piecewise deﬁned function f l (s) = 4 − s is independent of B. value it can attain. we get B − (4 − s) = 0 s ⇔ (s − 2)2 = 4 − B This equation obviously only has real solutions if 4 − B ≥ 0 or B ≤ 4 such that the √ square-root is deﬁned.e.d. We furthermore note. we have used that lims→∞ = 0.(ii) lim v(s) = lim As + s→∞ s→∞ A=0 B =  s sgn(A) · ∞ otherwise B s   0 q. (iii) After some simple transformations. For B ≤ 4. We thus choose B = 4.