Compiled Notes

Module 5
MScFE 620
Discrete-time Stochastic Processes

Revised: 07/21/2020
MScFE 620 Discrete-time Stochastic Processes − Notes Module 5

Module 5: The Binomial Model

Contents
Unit 1: Modeling on a Tree 3

Unit 2: Pricing on a Tree 5

Unit 3: Hedging 7

Unit 4: Exotic Derivatives 9

Problem Set 11

c 2019 - WorldQuant University − All rights reserved. 1
MScFE 620 Discrete-time Stochastic Processes − Summary Module 5

Summary
This module discusses a popular discrete-time model for asset prices – the binomial model or binomial
tree – which is used to represent asset dynamics for discrete processes as the asset values change
randomly. The module begins by defining the binomial model in detail and then proceeds by pricing
and hedging different kinds of derivatives in this model such as vanilla European derivatives and
Exotic European options.

c 2019 - WorldQuant University − All rights reserved. 2
MScFE 620 Discrete-time Stochastic Processes – Notes (1) Module 5: Unit 1

Unit 1: Modeling on a Tree

We describe a special market ((Ω, F, F, P), X) with trading occurring at times t = 0, 1, . . . , T , where
T is a positive integer. First, we will assume that the market has only one risky asset X (i.e, d = 1).
Thus, the primary tradeable assets can be represented as a vector (1, X) where 1 is the risk-free asset.

The binomial model or binomial tree assumes a very simple form: at each time t, the price of X at
the next time step (t + 1) can take only two values, both of which are multiples of Xt . To be precise,
let u and d be positive real numbers with d < u. We will assume that for every t ∈ {1, . . . T },

Xt = Xt−1 uZt d(1−Zt ) ,

where the Zt ’s are i.i.d Bernoulli random variables. Thus,

(
uXt−1 if Zt = 1
Xt =
dXt−1 if Zt = 0.

We can interpret this as follows: If we know the value of Xt−1 , the value of Xt can either jump ‘up’
to uXt−1 or jump ‘down’ to dXt−1 . We have not yet assigned probabilities to these jumps.
If we proceed inductively, we can write (for each t ≥ 1)
t
Y  u Pti=1 Zi  u  Yt
Zi (1−Zi )
Xt = X 0 u d = X0 dt = X0 dt ,
i=1
d d
Pt
where Yt := i=1 Zi has a binomial distribution.

With these heuristics, we can now formally define all the components of the market ((Ω, F, F, P), X).
First, let

Ω := {ω = (ω1 , . . . , ωT ) : ωi ∈ {0, 1}} .

Define F := 2Ω and assume that P ({ω}) > 0 for each ω ∈ Ω. Define the sequence {Zt : t = 1, . . . , T }
as
Zt (ω) := ωt , ω ∈ Ω, 1 ≤ t ≤ T.
Define the filtration F = (Ft )t≥0 by

F0 := {∅, Ω} , and Ft := σ ({Z1 , . . . , Zt }) t ≥ 1.

We can then define the price process X = {Xt : t = 0, 1, . . . , T } as

t
Y  u Pti=1 Zi
Zi (1−Zi )
X0 = constant, and Xt = X0 u d = X0 dt , t ≥ 1,
i=1
d

where 0 < d < u are fixed constants.

We now consider martingale measures on this market. If P∗ is a martingale measure, then for every
t ≥ 1,
Xt−1 = E∗ (Xt |Ft−1 ) = Xt−1 E∗ uZt d(1−Zt ) |Ft−1 ,

c 2019 - WorldQuant University – All rights reserved. 3
MScFE 620 Discrete-time Stochastic Processes – Notes (1) Module 5: Unit 1

which implies that

E∗ uZt d(1−Zt ) |Ft−1 = 1.

If we let p∗ (Z1 , . . . , Zt−1 ) := P (Zt = 1|Ft−1 ), then

up∗ (Z1 , . . . , Zt−1 ) + (1 − p∗ (Z1 , . . . , Zt−1 )) d = 1,

which implies that

1−d
p∗ (Z1 , . . . , Zt−1 ) =
=: p∗ .
u−d
It is left as an exercise to show that this implies that the random variables Zt are i.i.d Bernoulli
random variables with parameter p∗ , provided d < 1 < u. Thus, the market has a unique EMM if
and only if d < 1 < u. We shall henceforth make this assumption.

c 2019 - WorldQuant University – All rights reserved. 4
MScFE 620 Discrete-time Stochastic Processes – Notes (2) Module 5: Unit 2

Unit 2: Pricing on a Tree

We now want to price contingent claims in the market constructed in the previous section.

Let H be a contingent claim. Since the market is complete, the unique no-arbitrage price of H is
given by π(H) = E∗ (H). From the previous section, we can write P∗ as
T
Y
P∗ ({ω}) = p∗ ωt (1 − p∗ )1−ωt , ω = (ω1 , . . . , ωT ) ∈ Ω.
t=1

Hence,
X T
Y
π(H) = E (H) = ∗
H(ω) p∗ ωt (1 − p∗ )1−ωt .
ω∈Ω t=1

If H is a derivative that depends only on the terminal value of X, H = h(XT ), then we can write
the price as
T  


X
T ∗y
∗ ∗
π(H) = E (H) = E h X0 u d YT T −YT
= y T −y
h X0 u d p (1 − p∗ )T −y ,
y=0
y
PT
where YT := i=1 Zi has a binomial distribution with parameters n = T and p = p∗ .

Let us look at a concrete example of a call option. A call option is a derivative H with payoff function
H = h(XT ) = (XT − K)+ , where K is the strike price. The price is therefore given by
T  
X
+ T ∗ y
πC := π(H) = E (H) = ∗
X0 u d y T −y
−K p (1 − p∗ )T −y
y=0
y

With the following values X0 = 100, K = 110, u = 1/d = 1.2, T = 2, we get

1−d 45 5
p∗ = = =
u−d 99 11
and the price is
 2     2
−2

+ 6 + 5 6 2

+ 5
πC = 100 × 1.2 − 110 + 2 (100 − 110) + 100 × 1.2 − 110
11 11 11 11
 2
5 850
= 34 = .
11 121
Another popular example is the put option with a payoff function of H = (K − XT )+ , where K is
again called the strike price. The price is given by
T  
y T −y + T
X
∗
p∗ y (1 − p∗ )T −y


πP := π(H) = E (H) = K − X0 u d
y=0
y

Using the same parameters as above (X0 = 100, K = 110, u = 1/d = 1.2, T = 2), we get
 2     2
−2 +

6 + 5 6 2 +

5
πP = 110 − 100 × 1.2 + 2 (110 − 100) + 110 − 100 × 1.2
11 11 11 11

c 2019 - WorldQuant University – All rights reserved. 5
MScFE 620 Discrete-time Stochastic Processes – Notes (2) Module 5: Unit 2

850
. = 10 +
121
Now consider the derivative H which is the difference between the call option and the put option.
Then the payoff of H is
(
+ + XT − K if XT > K
H = (XT − K) − (K − XT ) = = XT − K.
XT − K if XT ≤ K

Hence
E∗ (XT − K)+ − (K − XT )+ = E∗ (XT − K) ,

which implies that

πC − πP = X0 − K.
This relationship between the call price πC and the put price πP is called the put-call parity. We can
also verify for the prices calculated above as
 
850 850
− 10 + = −10 = 100 − 110.
121 121

c 2019 - WorldQuant University – All rights reserved. 6
MScFE 620 Discrete-time Stochastic Processes – Notes (3) Module 5: Unit 3

Unit 3: Hedging
We now look at the problem of replication in a binomial tree.

Since the market is complete, we know that the unique EMM P∗ has PRP, in the sense that for every
(F, P∗ )-martingale M , there exists a predictable process ϕM such that for every 1 ≤ t ≤ T ,
t
X
Mt − M0 = ϕM
k (Xk − Xk−1 ) .
k=1

The process ϕM advertised above can be found as follows. First note that for each t ≥ 1,

Mt = Mt−1 + ϕM
t (Xt − Xt−1 ) ,

which implies that

Mt − Mt−1
ϕM
t =
Xt − Xt−1
since Xt 6= Xt−1 .

Now let H be a contingent claim. Define the martingale V H = VtH : 0 ≤ t ≤ T by

VtH := E∗ (H|Ft ), 0 ≤ t ≤ T.

Note that V0H = E∗ (H) = π(H).

Now since V H is an (F, P∗ )-martingale, it follows that we can find a predictable process ϕH such that
t
X
VtH − V0H = ϕH
k (Xk − Xk−1 ) ,
k=1

which implies that

t
X
VtH = π(H) + ϕH
k (Xk − Xk−1 ) .
k=1

Furthermore,
VtH − Vt−1
H
ϕH
t = .
Xt − Xt−1
Notice that since
t
X
VtH = π(H) + ϕH
k (Xk − Xk−1 ) ,
k=1

it follows that V is the value of the replicating strategy ϕH , and VTH = E∗ (H|FT ) = H.
H

Let us look at an example. Consider the example discussed in the previous section with the following
values: X0 = 100, u = 1/d = 1.2, T = 2. Consider a call option with strike price K = 110. We want
to calculate ϕH , where H = (XT − 110)+ .

Since T = 2, the sample space is

Ω = {(1, 1) , (1, 0) , (0, 1) , (0, 0)} .

c 2019 - WorldQuant University – All rights reserved. 7
MScFE 620 Discrete-time Stochastic Processes – Notes (3) Module 5: Unit 3

For convenience, we will replace 1 with u and 0 with d and write

Ω = {(u, u) , (u, d) , (d, u) , (d, d)} .

The values of V can be summarised in the following table:

ω V0H (ω) V1H (ω) V2H (ω) XT (ω) H(ω) = (XT (ω) − 110)+
850 170
(u, u) 121 11
34 100u2 34
850 170
(u, d) 121 11
0 100 0
850
(d, u) 121
0 0 100 0
850
(d, d) 121
0 0 100d2 0

The hedging strategy ϕH is given by

H
VtH − Vt−1
ϕH
t = .
Xt − Xt−1
The values are summarized below:
ω ϕH
1 (ω) ϕH
2 (ω)
51 17
(u, u) 121 22
51 17
(u, d) 121 22
51
(d, u) 121
0
51
(d, d) 121
0
850
The interpretation of the strategy is as follows. At time 0, starting with an initial capital of V0H = 121 ,
51 −4250
buy ϕH1 = 121
units of X and invest the remainder η H
1 = V 0
H
− ϕ H
1 X 0 = 121
. So, the transaction
involves borrowing 4250
121
from the bank. Hold these quantities until time 1.

At time 1, if the share goes up from 100 to 120, then the value of the portfolio will be V1H ((u, u)) =
V1H ((u, d)) = −4250
121
51
+ 121 × 120 = 170
11
, and in that case, we will change our holding in X to be
ϕH
2 ((u, u)) = ϕ H
2 ((u, d)) = 17
22
and the bank account investment to η2H = 170
11
− 17
22
× 120 = −850
11
.

Likewise, if the stock falls to 1.2−1 × 100, the value of the portfolio will be V1H ((d, u)) = V1H ((d, d)) =
−4250
121
51
+ 121 ×(1.2−1 × 100) = 0, and there will be no investment in the stock (ϕH H
2 ((d, u)) = ϕ2 ((d, d)) =
0), and no investment in the bank account η2 = V2 − ϕH 2 X2 = 0.

At time 2, if the share goes up again (to 144), the value of the portfolio will be V2 ((u, u)) = −850
11
+
17 H
22
× 144 = 34. For all other cases, V is equal to 0, which corresponds to the value of the derivative.

c 2019 - WorldQuant University – All rights reserved. 8
MScFE 620 Discrete-time Stochastic Processes – Notes(4) Module 5: Unit 4

Unit 4: Exotic Derivatives

In this section we will give examples of some common exotic derivatives and show how to price them.
We will continue with the market used in the previous section. Recall that this market has the
following parameters:

T = 2, X0 = 100, u = 1/d = 1.2 and p∗ = 5/11.

The first example we consider is an Asian option. The payoff of an arithmetic average Asian call
option with strike price K is
T
! T
!+
1 X 1 X
H := max Xt − K, 0 = Xt − K .
T + 1 t=0 T + 1 t=0

Its payoff is similar to that of a vanilla (ordinary) call option, but instead of using the terminal value
of the stock price XT , one uses the average of X over the lifetime of the option. The are variations
on the time points used to calculate the average.
In our example, the payoff looks like this:
1
P2
+
ω X0 (ω) X1 (ω) X2 (ω) H(ω) = 3 t=0Xt (ω) − 110
(u, u) 100 120 144 34/3
(u, d) 100 120 100 0
(d, u) 100 100 × 1.2−1 100 0
(d, d) 100 100 × 1.2−1 100 × 1.2−2 0

The price can then be calculated as

34 850
π(H) = × p∗ 2 = .
3 363
The price of an Asian put option can be calculated in a similar way. One can also replace the
arithmetic average with a geometric average, and the corresponding option is called a geometric
average Asian option.
Next, we loot at a lookback option. The payout of a lookback put option is

H := max Xt − XT .
0≤t≤T

For our example, the payoff values are tabulated below

ω X0 (ω) X1 (ω) X2 (ω) H(ω) = max0≤t≤2 Xt (ω) − X2 (ω)

(u, u) 100 120 144 0
(u, d) 100 120 100 20
(d, u) 100 100 × 1.2−1 100 0
(d, d) 100 100 × 1.2−1 100 × 1.2−2 275
9

The price of H is calculated as follows:

 2
5 6 275 6
π(H) = 20 × × + × ≈ 14.0496
11 11 9 11

c 2019 - WorldQuant University – All rights reserved. 9
MScFE 620 Discrete-time Stochastic Processes – Notes(4) Module 5: Unit 4

The last example we look at is a barrier option. A barrier option makes a payment if the stock prices
hits or misses a barrier. As an example, we consider an up-and-out call option whose payoff is
(
+ 0 if max0≤t≤T Xt ≥ B
H := (XT − K) I{max0≤t≤T Xt ≥B } = +
(XT − K) otherwise.

Here, K is the strike price and B > max (K, X0 ) is the barrier. So the option becomes void if X
crosses the barrier B, otherwise it behaves like an ordinary call option, hence the name ‘up and out’.
For parameters B = 115 and K = 95 we get

ω X0 (ω) X1 (ω) X2 (ω) H(ω) = (X2 (ω) − 95)+ I{max0≤t≤2 Xt (ω)≥115}

(u, u) 100 120 144 0
(u, d) 100 120 100 0
(d, u) 100 100 × 1.2−1 100 5
(d, d) 100 100 × 1.2−1 100 × 1.2−2 0

The price of H is calculated as follows:

6 5 150
π(H) = 5 × × = ≈ 1.23967.
11 11 121

c 2019 - WorldQuant University – All rights reserved. 10
MScFE 620 Discrete-time Stochastic Processes – Problem Set Module 5

Problem Set
Problem 1. Consider a binomial tree with the following parameters:

u = 1.4, d = 0.7, T = 2, X0 = 100.

The price of a call option with strike price K = 100 and expiring at time T = 2 is

Solution: First of all, remember the payoff at maturity for a call option,

H := max {XT − K, 0}

The payoff values are tabulated below

ω X0 (ω) X1 (ω) X2 (ω) H(ω) = max {X2 (ω) − K, 0}

(u, u) 100 140 196 96
(u, d) 100 140 98 0
(d, u) 100 70 98 0
(d, d) 100 70 49 0

The second step, is always to compute the P ∗ , from the lecture notes:
1−d 1 − 0.7
p∗ =
= ≈ 0.4286
u−d 1.4 − 0.7
The price of H is calculated as follows (from the table above we know that, H({u, d}) = H({d, u}) =
H({d, d}) = 0) :

π(H) = E∗ (H) = P ∗ ({u, u})H({u, u})+P ∗ ({u, d})H({u, d})+P ∗ ({d, u})H({d, u})+P ∗ ({d, d})H({d, d}) =

= p∗2 × 96 ≈ 17.6327

Problem 2. Consider a binomial tree with the following parameters:

u = 1.25, d = 1/u, T = 3, X0 = 400.

Consider a hedging strategy (η, ϕ) for the derivative H with payoff H = X2 − 319, where ϕ is the
investment in X. Compute the value of ϕ1 .

Solution: The payoff values are tabulated below

ω X0 (ω) X1 (ω) X2 (ω) H(ω) = X2 (ω) − 319

(u, u) 400 500 625 306
(u, d) 400 500 400 81
(d, u) 400 320 400 81
(d, d) 400 320 256 −63
The second step, is always to compute the P ∗ , from the lecture notes:

c 2019 - WorldQuant University – All rights reserved. 11
MScFE 620 Discrete-time Stochastic Processes – Problem Set Module 5

1−d 1 − 4/5 4
p∗ = = = ≈ 0.444
u−d 5/4 − 4/5 9
Notice that since
t
X
VtH = π(H) + ϕH
k (Xk − Xk−1 ) ,
k=1

it follows that V H is the value of the replicating strategy ϕH , and VTH = E∗ (H|FT ) = H.

For instance, we can compute V1H ({u, .}) as follows,

4 5
V1H ({u, .}) = V1H ({u, u}) = V1H ({u, d}) = ∗ 306 + ∗ 81 = 181
9 9
Following the same procedure,
4 5
V1H ({d, .}) = V1H ({d, u}) = V1H ({d, d}) = ∗ 81 − ∗ 63 = 1
9 9
Last step is to compute, V0H ,
4 4 5 5
V0H = ( )2 ∗ 306 + 2 ∗ 81 ∗ ∗ − ( )2 ∗ 63 = 81
9 9 9 9
The values of V can be summarised in the following table:

ω V0H (ω) V1H (ω) V2H (ω) X2 (ω) H(ω) = X2 (ω) − 319
(u, u) 81 181 306 625 306
(u, d) 81 181 81 400 81
(d, u) 81 1 81 400 81
(d, d) 81 1 −63 256 −63
The hedging strategy ϕH is given by

VtH − Vt−1
H
ϕH
t = .
Xt − Xt−1

The problem only asks about ϕH

1 , thus,

V1H − V0H 181 − 81

ϕH
1 = = =1
X1 − X0 500 − 400
The solution is, ϕH
1 = 1.

c 2019 - WorldQuant University – All rights reserved. 12