Lecture 6

Arbitrage in multiperiod models

In multiperiod models arbitrage is dened similarly.
Denition
1) An arbitrage opportunity is some trading strategy H such that:
(a) V
0
= 0
(b) V
T
0
(c) EV
T
> 0
(d) H is self-nancing
2) A self nancing strategy H is an arbitrage opportunity i:
(a) V

0
= 0 or (a) G

T
0
(b) V

T
0 or (b) EG

T
> 0
(c) EV

T
> 0 or (c) V

0
= 0
But the risk neutral measure the denition is a bit dierent:
Denition
A risk neutral measure (martingale measure) is a probability measure Q such that:
1) Q() > 0 for all
2) S

n
is a martingale under Q for every n = 1, 2, ..., N
E
Q
[S

n
(t +s)|F
t
] = S

n
(t), t, s 0
or
E
Q
[S

n
(t +s) S

n
(t)|F
t
] = 0
or
E
Q
[S
n
(t +s)/B(t +s)|F
t
] = S
n
(t)/B(t)
=E
Q
[B
t
S
n
(t +s)/B(t +s)|F
t
].
First fundamental theorem of nance:
There are no arbitrage opportunities if and only if there exists a martingale measure Q.
Proposition
1
If the multiperiod model does not have any arbitrage opportunitity, then none of the underlying
single period models has any arbitrage opportunities in the single period sense.
Example
Consider a 2-period problem with = {
1
,
2
, ...,
5
} and one risky security:
S
0
() S
1
() S
2
()

1
6 5 3

2
6 5 4

3
6 5 8

4
6 7 6

5
6 2 8
Find: V
t
, V

t
, G

t
, G
t
for all strategies (H
0
, H
1
) and check for the existance of martingale measures
or linear pricing measure. Find all of them. Keep in mind B
t
= (
10
9
)
t
.
Is the strategy: H
0
(1)() = 2 for all

s a self nancing startegy?

H
1
(1)() = 3 for all

s
H
0
(2)() =

1, for =
1
,
2
,
3
2, for =
4
,
5
H
1
(2)() =

29/9, for =
1
,
2
,
3
4, for =
4
,
5
The binomial model
Is a partocular case of the multiperiod model. Each period there are 2 possibilities: the security
price either goes up by the factor u (u > 1) orgoes down by a factor d (0 < d < 1). The probability
of an up move is equal to p, and the moves over time are independent of each other. Hence
the binomial model is related to the Binomial process (Bernoulli process) from probability in the
following fashion.
S
n+1
= S
n
Z
n
where {Z
n
}
nN
are iid with the distribution:
z =

u, with probability p
d, with probability 1 p
This is a multiplicative model.The reason we prefer multiplicative models to additive ones is because
stock prices have an exponential behavious that could be explained by the multiplicative models.
Also, additive models may result in negative prices.
2
Let k = # of up moves. in general, if

# up steps = n
total # steps = t t n
then the # of path to reach state m =

t
m

P(S
t
= Su
n
d
tn
) =

t
m

p
n
(1 p)
tn
n = 0, 1, ..., t
The self nancing equation now is
H
0
(t)(1 +R) +H
1
(t)S
t
= H
0
(t + 1) +H
1
(t + 1)S
t
What equation does the risk neutral probability verify here?
S
0
=
1
1 +r
E
Q
[S
1
|F
0
], in general S
t
=
1
1 +r
E
Q
[S
t+1
|F
t
]
q[
u 1 r
1 +r
] + (1 q)[
d 1 r
1 +r
] = 0 q =
1 +r d
u d
In general q is the conditional probability the next move is an up move given the information F
t
at time t. So q =
1+rd
ud
for all F
t
and t.
Also remark that in order for Q to exist we need 0 < q < 1 d < 1 +r < u
The martingale measure will be given by
Q() = q
n
(1 q)
Tn
where is any state corresponding to n ups and T n downs.
Hence the probability distribution of S
t
under the risk neutral probability measure is given for all
t by
Q(S
t
= S
0
u
n
d
tn
) =

t
m

q
n
(1 q)
tn
, n = 0, 1, ..., t.
A contingent claim is a r.v. X that represents the time T pay o from a seller to a buyer. The
contingent claim is simple if X = (S(T)).
Example
European call option X = max{S(T) K, 0}. If

k
t = 0 t = 1 t = 2

1
S
0
= 5 S
1
= 8 S
2
= 9

2
S
0
= 5 S
1
= 8 S
2
= 6
3

3
S
0
= 5 S
1
= 4 S
2
= 6

4
S
0
= 5 S
1
= 4 S
2
= 3
then X = max{S
2
K, 0}, for K = 5
X() =

4, =
1
1, =
2
,
3
0, =
4
Example of a contingent claim (derivative security) that is not simple:
X = max{
S
0
+S
1
+S
2
3 5
, 0}
This is called an Asian or averaging option.
X() =

7/3, =
1
4/3, =
2
0, =
3
,
4
Question:
The question of pricing is similer to the one-period model. Find (t, x) the price of the contingent
claim at time t = 0, 1, ..., T.
Denition
A contingent claim is said to be marketable or attainable if there exists a self-nancing strategy
such that V
T
() = X() for all . The corresponding portfolio or trading startegy H is said
to be replicate or generate X.
Risk neutral valuation principle
The time t value of a marketable contingent claim X is equal to V
t
, the time t of the portfolio
which replicates X. Moreover:
V

t
= V
t
/B
t
= E
Q
[X/B
T
|F
t
], t = 0, 1, ..., T
for all risk neutral probability measures Q.
Binomial model review
Recall that at time T, each possibility for S
T
is parameterized by K = # of ups from t = 0 to T.
So the value of the option at time T, if # ups = k
V
T
(k) = (S
0
u
k
d
Tk
)
We saw that
V
0
=
1
(1 +r)T
E
Q
[X] =
1
(1 +r)T
T

n=0

T
n

q
n
(1 q)
Tn
max{0, S
0
u
n
d
Tnk
}
4
where k is the strike price.
But what is the option price at time 0 t T?
For any t T 1, there are at most t ups.
V
t
(k) =
1
1 +r
E
Q
[V
t+1
|(# ups up to time t) = k]
=
1
1 +r
(q
u
V
t+1
(k + 1) + (1 q
u
)V
t+1
(k))
So, we have proved the recursion relation:
V
T
(k) = (S
0
u
k
d
Tk
), k = 0, 1, ..., T
t T 1, k t:
V
t
(k) =
1
1 +R
(q
u
V
t+1
(k + 1) + (1 q
u
)V
t+1
(k))
The only question we might need to answer is what are the hedging strategies.
For hedging we need to start at t = 0 and work forward by using the 1-period model hedging
strategy to obtain H = (x, y) which hedges the option given by:

V
1
(1), with probability q
u
V
1
(0), with probability q
d
Specically,
x =
1
1 +r
(uV
1
(0) dV
1
(1))
1
u d
y =
1
S
0
(V
1
(1) V
1
(0))
1
u d
This is obtain by solving the system:
(1 +r)x +S
0
uy = V
1
(1)
(1 +r)x +S
0
dy = V
1
(0)
In general: at time t T 1, for a xed k {0, 1, ..., t}
x =
1
1 +r
(uV
t+1
(k) dV
t+1
(k + 1))
1
u d
y =
1
S
0
(V
t+1
(k + 1) V
t+1
(k))
1
u d
Call options on a stock index
5
Options, so far, were dened in term of one underlying security. But in general one can dene
options one two or more underlying securities. For example, given a function g : R
N
R
+
one
can take X to be X = g(S
1
(T), ..., S
N
(T)). Then if you know the joint probability distribution
of the random variables S
1
(T), ..., S
N
(T) under the martingale measure, it is easy to compute the
time 0 value of this contingent claim. In particular, with:
g(S
1
, ..., S
N
) = (a
1
S
1
+... +a
N
S
N
e)
+
for positive scalars a
1
, ..., a
N
, you could have a call option on a stock index.
Or with
g(S
1
, ..., S
N
) = max{S
1
, ..., S
N
, e}
you could have a contingent claim delivering the best of N securities and the cash amount e.
Example
Suppose K = 9, N = 2, T = 2, r = 0 and the price process and information are as bellow. Also this
model (one can show with a few computations) has an unique probability measure Q, computed as
below.
Consider a call option with exercise price 13 on the time T = 2 value of the stock index S
1
(t)+S
2
(t)
What is the time 0 value of such a call option? What is the time 0 value of a contingent claim on
the 2 stocks and 8?
6