MAST 397A

Introduction to Quantitative Finance

Slides - Chapter 2

Frédéric Godin

Concordia University

Fall 2020

Frédéric Godin (2020) MAST 397A Fall 2020 1 / 87

 Option pricing in discrete time

Introduction

Chapter objectives: present discrete-time option pricing models based on

discrete-time trees.

Good first step to build intuition for more complicated models seen later.

Frédéric Godin (2020) MAST 397A Fall 2020 2 / 87

 Option pricing in discrete time One period binomial model

Model assumptions

Two time steps: t ∈ {0, 1}.

The sample space has two states:

Ω = {ω1 , ω2 }.

The probability measure P, called the physical measure, represents the

probability of sampling any these states:

P({ω1 }) = p, P({ω2 }) = 1 − p.

where 1 > 0p > 0.

Frédéric Godin (2020) MAST 397A Fall 2020 3 / 87

 Option pricing in discrete time One period binomial model

Assets in the model

Two assets are traded in the market.

The first one is a risk-free bank account type of asset. Its price grows
determinstically at the risk-free rate r > 0. Its time-t price, denoted Bt is

B0 = 1, B1 = e r

Frédéric Godin (2020) MAST 397A Fall 2020 4 / 87

 Option pricing in discrete time One period binomial model

Assets in the model

The second asset is risky. Its initial price is: S0 > 0. Its final price is
random:
S1 (ω1 ) = S0 u, S1 (ω2 ) = S0 d
where u > d are two positive constants.

The risky asset price dynamics can be represented by the following tree:

S0 u (with probability p)
S0 %
&
S0 d (with probability 1 − p)

Frédéric Godin (2020) MAST 397A Fall 2020 5 / 87

 Option pricing in discrete time One period binomial model

Trading strategy

Definition
A trading strategy is a couple θ = (θ(S) , θ(B) ).
θ(S) is the number of risky asset shares in the portfolio,
θ(B) is the number of risk-free asset shares in the portfolio.

θ(S) < 0: short sale of the risky asset.

θ(B) < 0: borrowing money.

Definition
The time-t value of the portfolio associated with θ is

Vtθ = θ(S) St + θ(B) Bt .

Frédéric Godin (2020) MAST 397A Fall 2020 6 / 87

 Option pricing in discrete time One period binomial model

Conditions for the existence of arbitrage

We are interested in relations between values of r , u, d and the existence

of arbitrage opportunities.

Proposition
There is no arbitrage opportunity if and only if u > e r > d.

Frédéric Godin (2020) MAST 397A Fall 2020 7 / 87

 Option pricing in discrete time One period binomial model

Proof: Conditions for the existence of arbitrage

Case 1: Suppose that e r ≥ u > d. Then the following portfolio is an
arbitrage opportunity:
θ(S) = −1, θ(B) = S0 .

Indeed, for this portfolio, we have V0θ = θ(S) S0 + θ(B) B0 = 0. Moreover,

(
S0 (e r − u) if ω1 ,
V1θ = θ(S) S1 + θ(B) B1 = −S1 + S0 e r =
S0 (e r − d) if ω2 .
Then, this strategy is an arbitrage opportunity since S0 (e r − u) ≥ 0 and
S0 (e r − d) > 0. Thus,
P[V1θ ≥ 0] = P[{ω1 , ω2 }] = 1
P[V1θ > 0] ≥ P[{ω2 }] > 0.

Frédéric Godin (2020) MAST 397A Fall 2020 8 / 87

 Option pricing in discrete time One period binomial model

Proof: Conditions for the existence of arbitrage

Case 2: Suppose that u > d ≥ e r . Then the following portfolio is an
arbitrage opportunity:
θ(S) = 1, θ(B) = −S0 .

Indeed, for this portfolio, we have V0θ = θ(S) S0 + θ(B) B0 = 0. Moreover,

(
S0 (u − e r ) if ω1 ,
V1θ = θ(S) S1 + θ(B) B1 = S1 − S0 e r =
S0 (d − e r ) if ω2 .
Then, this strategy is an arbitrage opportunity since S0 (u − e r ) > 0 and
S0 (d − e r ) ≥ 0. Thus,
P[V1θ ≥ 0] = P[{ω1 , ω2 }] = 1
P[V1θ > 0] ≥ P[{ω1 }] > 0.

Frédéric Godin (2020) MAST 397A Fall 2020 9 / 87

 Option pricing in discrete time One period binomial model

Proof: Conditions for the existence of arbitrage

Case 3: Suppose that u > e r > d.

Consider a portfolio with a null initial value. Then, θ(B) = −θ(S) S0 . The
terminal portfolio value is therefore
(
θ (S) (B) (S) r θ(S) S0 (u − e r ) if ω1 ,
V1 = θ S1 + θ B1 = θ (S1 − S0 e ) =
θ(S) S0 (d − e r ) if ω2 .

Therefore, the portfolio is not an arbitrage strategy:

if θ(S) = 0, then P[V1θ = 0] = 1,
if θ(S) > 0, then P[V1θ < 0] = P[{ω2 }] > 0,
if θ(S) < 0, then P[V1θ < 0] = P[{ω1 }] > 0.
Thus, all portfolios such that V0θ = 0 are not arbitrage opportunities.


Frédéric Godin (2020) MAST 397A Fall 2020 10 / 87

 Option pricing in discrete time One period binomial model

Adding a derivative to the market

We add a derivative with payoff C1 among market assets.

What would be possible prices C0 for this derivative so that there is no

arbitrage opportunity?

Idea: find a portfolio whose terminal value is equal to the option payoff
almost surely (a.s.) i.e. with probability 1.
By the law of one price, the initial value of this portfolio must be the
same than the option price (otherwise there would exist an arbitrage
opportunity).

Frédéric Godin (2020) MAST 397A Fall 2020 11 / 87

 Option pricing in discrete time One period binomial model

Replicating portfolio

In other words, if we find a portfolio such that

Payoff Portfolio = Payoff Derivative

Then
Initial price Portfolio = Initial price Derivative

Frédéric Godin (2020) MAST 397A Fall 2020 12 / 87

 Option pricing in discrete time One period binomial model

Replicating portfolio

In what follows, we will use the following notation:

(
C1 (ω1 ) = C (u) ,
C1 (ω2 ) = C (d) .

Theorem
In the binomial model, the following portfolio, called replicating
portfolio, gives P[C1 = V1θ ] = 1:

C (u) − C (d)
θ(S) = ,
S0 (u − d)
uC (d) − dC (u)
θ(B) = .
B1 (u − d)

Frédéric Godin (2020) MAST 397A Fall 2020 13 / 87

 Option pricing in discrete time One period binomial model

Proof: replicating portfolio

Proof:
We seek a replicating portfolio such that V1θ (ω1 ) = C (u) and
V1θ (ω2 ) = C (d) .

This leads to the following system of linear equations:

   (B)   (u) 
B1 S1 (ω1 ) θ C
(S) =
B1 S1 (ω2 ) θ C (d)

The solution to this system of linear equations is

C (u) − C (d)
θ(S) = ,
S0 (u − d)
uC (d) − dC (u)
θ(B) = . 
B1 (u − d)

Frédéric Godin (2020) MAST 397A Fall 2020 14 / 87

 Option pricing in discrete time One period binomial model

Measure Q

Theorem
The initial value V0θ of the replicating portfolio is

V0θ = e −r EQ [C1 ] .

where the probability measure Q is defined by

(e r − d) (u − e r )
Q(ω1 ) = , Q(ω2 ) = .
u−d u−d

Remark
Q is a fictitious probability measure. It has no physical interpretation.

Frédéric Godin (2020) MAST 397A Fall 2020 15 / 87

 Option pricing in discrete time One period binomial model

Proof of the previous theorem

Proof:
The initial value of the replicating portfolio is

C (u) − C (d) uC (d) − dC (u)

V0θ = θ(S) S0 + θ(B) B0 = +
(u − d) e r (u − d)
C (u) (e r − d) C (d) (u − e r )
= + r .
er u − d e u−d
We notice that we can define the following probability measure Q:

(e r − d) (u − e r )
Q({ω1 }) = , Q({ω2 }) = ,
u−d u−d
and thus

V0θ = e −r EQ [C1 ] . 

Frédéric Godin (2020) MAST 397A Fall 2020 16 / 87

 Option pricing in discrete time One period binomial model

Pricing through the measure Q

Corollary
In the one-period binomial model, if there exists no arbitrage opportunities,
then the time-0 price C0 of a derivative of payoff C1 is given be the relation

C0 = e −r EQ [C1 ] .

Proof: By the law of one price, C0 = V0θ . Otherwise, there exists

arbitrage. 

Frédéric Godin (2020) MAST 397A Fall 2020 17 / 87

 Option pricing in discrete time One period binomial model

Risk-neutral measure

Remark
The measure Q is called a risk-neutral measure.
Under Q, we can identify derivatives price by calculating the expected
payoff discounted at the risk-free rate.

Frédéric Godin (2020) MAST 397A Fall 2020 18 / 87

 Option pricing in discrete time Multi-period binomial model

Multi-period binomial model

Consider the multi-period case:

There exists many time steps: t = 0, 1, . . . , T .

Two assets are traded in the market.

The first is risk-free and its deterministic price is

Bt = e rt

where r > 0 is the continuously compounded risk-free rate for one period.

Frédéric Godin (2020) MAST 397A Fall 2020 19 / 87

 Option pricing in discrete time Multi-period binomial model

Risky asset
The risky asset has the following dynamics:
Assumption
Let mj , j = 1, . . . , T be i.i.d. Bernouilli(p) random variables.
Let u > d.
We assume
(
mt+1 1−mt+1 St u if mt+1 = 1,
St+1 = St u d =
St d if mt+1 = 0.

Recall: A random variable X has a Bernouilli(p) distribution if

(
1 with probability p,
X =
0 otherwise.

Frédéric Godin (2020) MAST 397A Fall 2020 20 / 87

 Option pricing in discrete time Multi-period binomial model

Risky asset

We can therefore represent the model with the following tree:

S0 u 3
S0 u 2 %
& ...
S0 u %
& S0 u2d
S0 %
& S0 ud %
& ...
S0 d %
& S0 ud 2
S0 d 2 %
& ...
S0 d3

The tree is recombining: a upward followed by a downward movement

produces a price identical to a downward movement followed by an upward
movement.

Frédéric Godin (2020) MAST 397A Fall 2020 21 / 87

 Option pricing in discrete time Multi-period binomial model

Risky asset

Remark
We can recursively obtain

St = S0 u Mt d t−Mt
Xt
Mt = mj
j=1

Remark
Mt follows a Binomial(t, p) distribution.

Frédéric Godin (2020) MAST 397A Fall 2020 22 / 87

 Option pricing in discrete time Multi-period binomial model

Stochastic process

Definition
A stochastic process is a sequence of random variables indexed by time:
X0 , . . . , XT . Such variables can be vectorial (i.e. multivariate).

Frédéric Godin (2020) MAST 397A Fall 2020 23 / 87

 Option pricing in discrete time Multi-period binomial model

Trading strategy

Definition
A trading strategy is a stochastic process
(S) (B)
θ = {θt = (θt , θt )|t = 1, . . . , T }

which is predictable i.e. where ∀t, θt exclusively function of variables

S0 , . . . , St−1 which represent the information available at time t − 1.

By convention we set θ0 := θ1 .

(S) (B)
θt and θt represent respectively the number of shares of the risky and
risk-free assets in a portfolio between times t − 1 and t.1

1 (S) (B)
θt and θt are therefore known at time t − 1.
Frédéric Godin (2020) MAST 397A Fall 2020 24 / 87
 Option pricing in discrete time Multi-period binomial model

Self-financing portfolios

Definition
The portfolio value corresponding to the trading strategy θ is
(S) (B)
Vtθ = θt St + θt Bt

Definition
The trading strategy θ is self-financing if
(S) (B) (S) (B)
∀t, θ St + θ Bt = θt St + θt Bt
|t+1 {z t+1 } | {z }
portfolio value after rebalancing at t portfolio value before rebalancing at t

Frédéric Godin (2020) MAST 397A Fall 2020 25 / 87

 Option pricing in discrete time Multi-period binomial model

Absence of arbitrage under the multi-period binomial

model

Theorem
Under the multi-period binomial model where only assets B and S are
traded, there doesn’t exist any arbitrage opportunity if and only if
u > e r > d.
Proof: Omitted. See the MACF 401 class.

Assumption
We suppose that u > e r > d.

Frédéric Godin (2020) MAST 397A Fall 2020 26 / 87

 Option pricing in discrete time Multi-period binomial model

Notation for portfolios in the multi-period binomial model

Definition
Consider the notation
Vtθ (k): time-t portfolio value for strategy θ,
Ct (k): time-t derivative price,
(S)
θt+1 (k): number of risky asset shares in the portfolio between time t
and t + 1,
(B)
θt+1 (k): number of risk-free asset shares in the portfolio between
time t and t + 1,
conditionally on St = S0 u k d t−k . a

Moreover, define St (k) = S0 u k d t−k .

a
i.e. when k upward movements of the risky asset occured during the t first
time periods.

Frédéric Godin (2020) MAST 397A Fall 2020 27 / 87

 Option pricing in discrete time Multi-period binomial model

Example of option pricing

We shall illustrate how to price an option in an example before deriving

the general theory.
This example is taken from the exercises list.

Example
We want to find the price for a European call option on the risky asset
with strike K = 100 and maturity T = 2.
u = 1.12, d = 0.93,
r = 0.02,
S0 = 100,
p = 0.6,
No dividends.

Frédéric Godin (2020) MAST 397A Fall 2020 28 / 87

 Option pricing in discrete time Multi-period binomial model

Strategy for pricing

We will construct a self-financing replicating strategy θ which gives

P[VTθ = CT ] = 1.

By the law of one price,

Option price = Replicating portfolio value

⇒ to price the option, we seek the value of the replicating portfolio.

Frédéric Godin (2020) MAST 397A Fall 2020 29 / 87

 Option pricing in discrete time Multi-period binomial model

Underlying asset price at each node

First, we can compute the underlying asset price at each node of the tree:

S2 (2) = S0 u 2 = 125.44
S1 (1) = S0 u = 112 %
&
S0 (0) = S0 = 100 %
& S2 (1) = S0 ud = 104.16
S1 (0) = S0 d = 93 %
&
S2 (0) = S0 d 2 = 86.49

Frédéric Godin (2020) MAST 397A Fall 2020 30 / 87

 Option pricing in discrete time Multi-period binomial model

Pricing problem

We seek the replicating portfolio value and the option price at each
node:
v2 (2) =?
v1 (1) =? %
&
v0 (0) =? %
& v2 (1) =?
%
v1 (0) =? &
v2 (0) =?

C2 (2) =?
C1 (1) =? %
&
C0 (0) =? %
& C2 (1) =?
C1 (0) =? %
&
C2 (0) =?

Frédéric Godin (2020) MAST 397A Fall 2020 31 / 87

 Option pricing in discrete time Multi-period binomial model

Law of one price: consequence

Since the replicating portfolio is such that P[VTθ = CT ] = 1, the law of

one price stipulates that
Ct (k) = vt (k).
Otherwise, an arbitrage opportunity would exist.

v2 (2) = C2 (2)
v1 (1) = C1 (1) %
&
v0 (0) = C0 (0) %
& v2 (1) = C2 (1)
v1 (0) = C1 (0) %
&
v2 (0) = C2 (0)

Frédéric Godin (2020) MAST 397A Fall 2020 32 / 87

 Option pricing in discrete time Multi-period binomial model

Terminal option value

The terminal option value is its payoff: C2 = max(0, S2 − K ).

C2 (2) = max(0, S0 u 2 − K ) = 25.44

C1 (1) =? %
&
C0 (0) =? %
& C2 (1) = max(0, S0 ud − K ) = 4.16
C1 (0) =? %
&
C2 (0) = max(0, S0 d 2 − K ) = 0

Frédéric Godin (2020) MAST 397A Fall 2020 33 / 87

 Option pricing in discrete time Multi-period binomial model

Solution of the tree

By the law of one price: v2 (k) = C2 (k).

v2 (2) = C2 (2) = 25.44

v1 (1) =? %
&
v0 (0) =? %
& v2 (1) = C2 (1) = 4.16
v1 (0) =? %
&
v2 (0) = C2 (0) = 0

Frédéric Godin (2020) MAST 397A Fall 2020 34 / 87

 Option pricing in discrete time Multi-period binomial model

Solution of the tree

We now want to calculate v1 (1). We focus on the following part of the

tree.

Calculations are analogous to a binomial tree for a single period.

Frédéric Godin (2020) MAST 397A Fall 2020 35 / 87

 Option pricing in discrete time Multi-period binomial model

Solution of the tree

(S) (B)
We now want to calculate θ2 (1) and θ2 (1) i.e. the composition of the
portfolio in the red box.

We want
V2θ = v2 (2) if the underlying goes up,
V2θ = v2 (1) if the underlying goes down.

(S) (B)
Since V2θ = θ2 S2 + θ2 B2 , we have
(S) (B)
v2 (2) = θ2 (1)S0 u 2 + θ2 (1)B2 ,
(S) (B)
v2 (1) = θ2 (1)S0 ud + θ2 (1)B2 .

Frédéric Godin (2020) MAST 397A Fall 2020 36 / 87

 Option pricing in discrete time Multi-period binomial model

Solution of the tree

The solution to the system of equations is

(S) v2 (2) − v2 (1)

θ2 (1) = = 1,
S0 u(u − d)
(B) uv2 (1) − dv2 (2)
θ2 (1) = = −96.08.
B2 (u − d)

Frédéric Godin (2020) MAST 397A Fall 2020 37 / 87

 Option pricing in discrete time Multi-period binomial model

Solution of the tree

We want the option price C1 (1) = v1 (1).

(S) (B)
We have V1θ = θ2 S1 + θ2 B1 .

(S) (B)
C1 (1) = v1 (1) = θ2 (1)S0 u + θ2 (1)B1 ,
= 1 × 112 − 96.08 × 1.0202 = 13.98.

Frédéric Godin (2020) MAST 397A Fall 2020 38 / 87

 Option pricing in discrete time Multi-period binomial model

Solution of the tree

This gives
C2 (2) = 25.44
C1 (1) = 13.98 %
&
C0 (0) =? %
& C2 (1) = 4.16
%
C1 (0) =? &
C2 (0) = 0

Frédéric Godin (2020) MAST 397A Fall 2020 39 / 87

 Option pricing in discrete time Multi-period binomial model

Solution of the tree

Through analogous calculations we obtain C1 (0) = 1.936.

C2 (2) = v2 (2) = 25.44

C1 (1) = v1 (1) = 13.98 %
&
C0 (0) = v0 (0) =? %
& C2 (1) = v2 (1) = 4.16
C1 (0) = v1 (0) = 1.936 %
&
C2 (0) = v2 (0) = 0

Frédéric Godin (2020) MAST 397A Fall 2020 40 / 87

 Option pricing in discrete time Multi-period binomial model

Solution of the tree

We want the option price C0 (0) = v0 (0).

(S) (B)
We have V1θ = θ2 S1 + θ2 B1 .

(S) (B)
C1 (1) = v1 (1) = θ2 (1)S0 u + θ2 (1)B1 ,
= 1 × 112 − 96.08 × 1.0202 = 13.98.

Frédéric Godin (2020) MAST 397A Fall 2020 41 / 87

 Option pricing in discrete time Multi-period binomial model

Solution of the tree

We now want to calculate v0 (0). We focus on this part of the tree.

Calculations analogous to the one-period binomial tree on this section of

the multi-period tree.

Frédéric Godin (2020) MAST 397A Fall 2020 42 / 87

 Option pricing in discrete time Multi-period binomial model

Solution of the tree

(S) (B)
We now want to calculate θ1 (0) and θ1 (0) i.e. composition of the
replicating portfolio in the red box.

We want
V1θ = v1 (1) if the underlying goes up,
V1θ = v1 (0) if the underlying goes down.

(S) (B)
Since V1θ = θ1 S1 + θ1 B1 , we have
(S) (B)
v1 (1) = θ1 (0)S0 u + θ1 (0)B1 ,
(S) (B)
v1 (0) = θ1 (0)S0 d + θ1 (0)B1 .

Frédéric Godin (2020) MAST 397A Fall 2020 43 / 87

 Option pricing in discrete time Multi-period binomial model

Solution of the tree

Solution to the system of equations:

(S) v1 (1) − v1 (0)

θ1 (0) = = 0.6339,
S0 (u − d)
(B) uv1 (0) − dv1 (1)
θ1 (0) = = −55.89.
B1 (u − d)

Frédéric Godin (2020) MAST 397A Fall 2020 44 / 87

 Option pricing in discrete time Multi-period binomial model

Solution of the tree

We want the option price C0 (0) = v0 (0).

(S) (B)
We have V0θ = θ1 S0 + θ1 B0 .

(S) (B)
C0 (0) = v0 (0) = θ1 (0)S0 + θ1 (0)B0 ,
= 0.6339 × 100 − 55.89 × 1 = 7.502.

Frédéric Godin (2020) MAST 397A Fall 2020 45 / 87

 Option pricing in discrete time Multi-period binomial model

Multi-period binomial model replication portfolio

Theorem
Consider a European-type derivative whose time-T payoff is CT = f (ST )
for some function f .

There exists a self-financing trading strategy θ such that P[VTθ = CT ] = 1.

Frédéric Godin (2020) MAST 397A Fall 2020 46 / 87

 Option pricing in discrete time Multi-period binomial model

Multi-period binomial model replication portfolio

Theorem (Continued)
The portfolio value associated with this strategy θ can be calculated
recursively for all t by
(
CT = f (ST (k)) if t = T ,
Vtθ (k) = (1)
e −r qu Vt+1
θ (k + 1) + q V θ (k) if t < T ,
 
d t+1

where
er − d
qu = and qd = 1 − qu .
u−d

Frédéric Godin (2020) MAST 397A Fall 2020 47 / 87

 Option pricing in discrete time Multi-period binomial model

Multi-period binomial model replication portfolio

Theorem (Continued)
Furthermore, the composition of the replicating portfolio during time
period [t, t + 1) given k previous upward movements have occurred is
given by
θ (k + 1) − V θ (k)
Vt+1
(S) t+1
θt+1 (k) = ,
St (k)(u − d)
θ (k) − dV θ (k + 1)
uVt+1
(B) t+1
θt+1 (k) = .
Bt+1 (u − d)

Frédéric Godin (2020) MAST 397A Fall 2020 48 / 87

 Option pricing in discrete time Multi-period binomial model

Replicating portfolio

Proof (by induction):

Initial step: Trivially, a portfolio whose final value is VTθ = CT a.s.2

allows replicating the derivative payoff.

Thus, conditionally on the event {MT = k}, a portfolio which has the
property

VTθ = VTθ (k) = CT (k) = f (ST (k))

allows replicating the derivative payoff.

2
a.s. mean with probability one.
Frédéric Godin (2020) MAST 397A Fall 2020 49 / 87
 Option pricing in discrete time Multi-period binomial model

Proof (continued)

Induction step:

Induction hypothesis: Suppose that at time t + 1, for all i, holding a

θ (i) = v
portfolio whose value Vt+1 t+1 (i) when the underlying asset is worth
St+1 (i) allows constructing a self-financing portfolio such that
P[VTθ = CT ] = 1.

Frédéric Godin (2020) MAST 397A Fall 2020 50 / 87

 Option pricing in discrete time Multi-period binomial model

Proof (continued)

Suppose that at time t the underlying asset had k upward movements

previously i.e. St = St (k).

We would like to identify the portfolio composition θt+1 which would give
vt+1 (k + 1) if the underlying asset goes up and vt+1 (k) if the underlying
asset goes down.

Such a portfolio would allow obtaining P[VTθ = CT ] = 1 (as per induction

hypothesis).

Frédéric Godin (2020) MAST 397A Fall 2020 51 / 87

 Option pricing in discrete time Multi-period binomial model

Proof (continued)
This situation is analogous to the one-period binomial case. We seek a

θ
portfolio such that Vt+1 = vt+1 (k + 1) if the underlying asset goes up and
θ
Vt+1 = vt+1 (k) if it goes down.

This leads to the following system of linear equations:

  " (B) #  
Bt+1 St (k)u θt+1 vt+1 (k + 1)
(S) =
Bt+1 St (k)d θ vt+1 (k)
t+1

The solution to this system of equations is given by

(S) vt+1 (k + 1) − vt+1 (k)
θt+1 = ,
St (k)(u − d)
(B) uvt+1 (k) − dvt+1 (k + 1)
θt+1 = .
Bt+1 (u − d)
Frédéric Godin (2020) MAST 397A Fall 2020 52 / 87
 Option pricing in discrete time Multi-period binomial model

Proof (continued)

The portfolio we sought3 therefore exists. Its value is given by

(S) (B)
Vtθ (k) = θt+1 (k)St (k) + θt+1 (k)Bt
= ...
= e −r [qu vt+1 (k + 1) + qd vt+1 (k)] .

3 θ
one giving Vt+1 (k + mt+1 ) at time t + 1.
Frédéric Godin (2020) MAST 397A Fall 2020 53 / 87
 Option pricing in discrete time Multi-period binomial model

Risk-neutral measure

Definition
The risk-neutral measure Q is defined as the probability measure such
that
e r −d
∀t, Q[{mt = 1}] = u−d ,
Q[{mt = 0}] = 1 − Q[{mt = 1}] and
variables mi , i = 1, . . . , T are all independent under Q.

Frédéric Godin (2020) MAST 397A Fall 2020 54 / 87

 Option pricing in discrete time Multi-period binomial model

Uniqueness of option prices in the binomial model

In the previous chapter (without assumptions on the underlying asset): the

absence of arbitrage assumption only led to bounds on option prices.

However, under the additional assumption of the binomial model, the

absence of arbitrage assumption is strong enough to specify the price of
options in a unique manner.
Option price is the replicating portfolio value.

Frédéric Godin (2020) MAST 397A Fall 2020 55 / 87

 Option pricing in discrete time Multi-period binomial model

Law of iterated expectations

Proposition (Law of iterated expectations)

For all random variables X , Y1 , Y2 where X is integrable,ab
E [E[X |Y1 ]] = E[X ]
 
E E[X |Y1 , Y2 ] Y1 = E[X |Y1 ].
a
We say X is integrable if its expectation is finite.
b
E [X |Y ] means the conditional expectation of X with respect to Y .

Frédéric Godin (2020) MAST 397A Fall 2020 56 / 87

 Option pricing in discrete time Multi-period binomial model

Replicating portfolio value

Theorem
For the replicating portfolio θ, we have
   
θ −r Q θ −r (T −t) Q θ
Vt = e E Vt+1 S0 , . . . , St = e E VT S0 , . . . , St .

Frédéric Godin (2020) MAST 397A Fall 2020 57 / 87

 Option pricing in discrete time Multi-period binomial model

Replicating portfolio value

Proof:
The equality  
−r
Vtθ =e E Q θ
Vt+1 S0 , . . . , St

comes from the following equation which was previously derived:

h i
Vtθ (k) = e −r qu Vt+1θ θ
(k + 1) + qd Vt+1 (k) .

Frédéric Godin (2020) MAST 397A Fall 2020 58 / 87

 Option pricing in discrete time Multi-period binomial model

Replicating portfolio value

Proof (continued): Proceeding recursively, we obtain

 
−r
Vtθ = Q θ
e E Vt+1 S0 , . . . , St
   
= e −r EQ e −r EQ Vt+2 θ
S0 , . . . , St+1 S0 , . . . , St
 
−2r Q θ
= e E Vt+2 S0 , . . . , St
 
= . . . = e −r (T −t) EQ VTθ S0 , . . . , St . 

Frédéric Godin (2020) MAST 397A Fall 2020 59 / 87

 Option pricing in discrete time Multi-period binomial model

Price of a derivative

Corollary
In the multi-period binomial model, if there exists no arbitrage
opportunity, the time-t price Ct of a derivative whose payoff is
CT = f (ST ) is given by the relation
   
−r Q −r (T −t) Q
Ct = e E Ct+1 S0 , . . . , St = e E f (ST ) S0 , . . . , St .

Proof: The proof is a consequence of:

the previous theorem, and
the law of one price
I the derivative’s price must be equal to the replicating portfolio value:
Ct = Vtθ a.s.

Frédéric Godin (2020) MAST 397A Fall 2020 60 / 87

 Option pricing in discrete time Multi-period binomial model

Example of option pricing

We revisit the previous example:

we show that using the previous risk-neutral evaluation formula is
much quicker to calculate option prices.

Example
Recall: We want to find the price for a European call option on the risky
asset with strike K = 100 and maturity T = 2.
u = 1.12, d = 0.93,
r = 0.02,
S0 = 100,
p = 0.6,
No dividends.

Frédéric Godin (2020) MAST 397A Fall 2020 61 / 87

 Option pricing in discrete time Multi-period binomial model

Recursive formula for option price

When directly using the risk-neutral valuation formula to calculate option

prices:
not necessary to calculate the replicating portfolio composition.

Risk-neutral valuation formula:

 

Ctθ = e −r EQ Ct+1 S0 , . . . , St  .
| {z }
Info available at t

⇒ Ct (k) = e −r [qu Ct+1 (k + 1) + qd Ct+1 (k)] .

e r −d
where qu = u−d , qd = 1 − qu .

Frédéric Godin (2020) MAST 397A Fall 2020 62 / 87

 Option pricing in discrete time Multi-period binomial model

Back to the example: option price calculation

We want to calculate C1 (1).

Frédéric Godin (2020) MAST 397A Fall 2020 63 / 87

 Option pricing in discrete time Multi-period binomial model

Option price calculation

Calculation of risk-neutral probabilities:

er − d
qu = = 0.4747,
u−d
qd = 1 − qu = 0.5253.

Then,
qu C2 (2) + qd C2 (1)
C1 (1) =
er
0.4747 × 25.44 + 0.5253 × 4.16
= = 13.98.
e 0.02

Frédéric Godin (2020) MAST 397A Fall 2020 64 / 87

 Option pricing in discrete time Multi-period binomial model

Option price calculation

Analogously, we obtain

qu C2 (1) + qd C2 (0)
C1 (0) =
er
0.4747 × 4.16 + 0.5253 × 0
= = 1.936.
e 0.02

qu C1 (1) + qd C1 (0)
C0 (0) =
er
0.4747 × 13.98 + 0.5253 × 1.936
= = 7.502.
e 0.02

Frédéric Godin (2020) MAST 397A Fall 2020 65 / 87

 Option pricing in discrete time Multi-period binomial model

Absence of impact of the physical measure

The option price Ct depends only on the risk-neutral measure Q.

The physical measure P (i.e. probability p) has no impact on the option

price.

Frédéric Godin (2020) MAST 397A Fall 2020 66 / 87

 Option pricing in discrete time Modifications to the binomial model

Modifications to the binomial model

The binomial model can be slightly modified to incorporate additional

features:
Dividends paid by the underlying asset.
American or exotic options.
Path-dependent options.

Frédéric Godin (2020) MAST 397A Fall 2020 67 / 87

 Option pricing in discrete time Modifications to the binomial model

Presence of dividends: single period

Consider the one-period case where the underlying asset pays dividends.

Suppose that Bt and St follow the binomial model previously described.

Frédéric Godin (2020) MAST 397A Fall 2020 68 / 87

 Option pricing in discrete time Modifications to the binomial model

Presence of dividends: single period

However, the underlying asset pays at time t = 1 a dividend of

D1 := S1 (e δ − 1).

The continuously compounded dividend rate is thus δ.

S1 (e δ − 1) ≈ S1 δ.

The holder of the underlying asset receives at time t = 1 in total

S1 e δ = S1 + S1 (e δ − 1) .
|{z} | {z }
underlying asset dividend

Frédéric Godin (2020) MAST 397A Fall 2020 69 / 87

 Option pricing in discrete time Modifications to the binomial model

Dividend payment

The dividend is paid to the underlying asset holder at an infinitesimal

time before the maturity of the derivative.

The derivative payoff does not include the dividend:

e.g. for a call option C1 = max(0, S1 − K ).

Frédéric Godin (2020) MAST 397A Fall 2020 70 / 87

 Option pricing in discrete time Modifications to the binomial model

Pricing in the presence of a dividend

We want to determine C0 , the derivative’s price,

We seek a portfolio (θ(B) , θ(S) ) such that P[V1θ = C1 ] = 1.

We must therefore solve the following system of equations:

 r
e S0 ue δ
  (B)   (u) 
θ C
r δ (S) = .
e S0 de θ C (d)

The solution to this system of equations is

C (u) − C (d)
θ(S) = ,
S0 e δ (u − d)
uC (d) − dC (u)
θ(B) = .
e r (u − d)

Frédéric Godin (2020) MAST 397A Fall 2020 71 / 87

 Option pricing in discrete time Modifications to the binomial model

Pricing in the presence of a dividend

The initial value of the replicating portfolio must be equal to C0 by the

law of one price. Therefore

C0 = V0θ = θ(S) S0 + θ(B) B0

C (u) (e r −δ − d) C (d) (u − e r −δ )
= + r .
e r | u {z −d } e | u {z −d }
=:qu =:qd

Frédéric Godin (2020) MAST 397A Fall 2020 72 / 87

 Option pricing in discrete time Modifications to the binomial model

Risk-neutral pricing with the dividend

Theorem

C0 = e −r EQ [C1 ] .

where the probability measure Q is defined by

e r −δ − d u − e r −δ
Q(ω1 ) = qu = , Q(ω2 ) = qd = .
u−d u−d

Remark
By setting δ = 0, we recover the pricing formula in absence of dividends.

Frédéric Godin (2020) MAST 397A Fall 2020 73 / 87

 Option pricing in discrete time Modifications to the binomial model

Multi-period binomial model with dividends

We now consider the multi-period binomial with dividends:

for t = 1, . . . , T , the underlying asset pays a dividend
Dt = St (e δt − 1) an infinitesimal time before t.

Theorem
The time-t price Ct of an option is given by
   
−r Q −r (T −t) Q
Ct = e E Ct+1 S0 . . . St = e E CT S0 . . . St .

where, under the probability measure Q, the mt , t = 1, . . . , T are

independent, and

e (r −δt ) − d
Q({mt = 1}) = , Q({mt = 0}) = 1 − Q({mt = 1}).
u−d

Frédéric Godin (2020) MAST 397A Fall 2020 74 / 87

 Option pricing in discrete time Modifications to the binomial model

Multi-period binomial model with dividends

Theorem (continued)
Finally, the replicating portfolio composition is

(S) Ct+1 (k + 1) − Ct+1 (k)

θt+1 (k) = ,
St (k)e δt+1 (u − d)
(B) uCt+1 (k) − d Ct+1 (k + 1)
θt+1 (k) = .
Bt+1 (u − d)

Frédéric Godin (2020) MAST 397A Fall 2020 75 / 87

 Option pricing in discrete time Modifications to the binomial model

Self-financing trading strategy

Remark
In the presence of dividends, the definition of a self-financing trading
(S) (B)
strategy is a process θ = {θt = (θt , θt ), t = 0, . . . , T } such thata
(S) (B) (S) (B)
∀t, θt+1 St + θt+1 Bt = θt (St + Dt ) + θt Bt

where Dt is the dividend paid an infinitesimal amount of time before t.

a
By convention we set θ1 = θ0 and D0 = 0.

Frédéric Godin (2020) MAST 397A Fall 2020 76 / 87

 Option pricing in discrete time Modifications to the binomial model

American options and the binomial model

The American option pricing algorithm in the binomial model is similar to

the algorithm for a European option.

Difference: for the American option, its value is bounded below by its
immediate exercise value.

Frédéric Godin (2020) MAST 397A Fall 2020 77 / 87

 Option pricing in discrete time Modifications to the binomial model

American option pricing

Consider the multi-period binomial model, and an American option whose

payoff is Ψτ if the option is exercised at τ .

Theorem
Then
 
 
Ct = max  Ψt , e −r EQ Ct+1 S0 , . . . , St  if t < T .
|{z}
Exercice value

Frédéric Godin (2020) MAST 397A Fall 2020 78 / 87

 Option pricing in discrete time Modifications to the binomial model

American option pricing

Heuristic justification:
At time t < T , the option holder has two choices.
Immediately exercise the option. Value =Ψt .
Hold the
 option for at least another period. Value =
e −r EQ Ct+1 S0 , . . . , St ,

The option holder will decide whether or not to exercise according to what
maximizes the value of his option. 

Frédéric Godin (2020) MAST 397A Fall 2020 79 / 87

 Path dependent options

Path dependent options

For some options, the payoff doesn’t depend only on the final underlying
asset price ST , but on its whole path S0 , . . . , ST .
This is the case for instance for lookback, Asian or barrier options.

In this case, the price doesn’t only depend on the number of previous
upward movements k, but rather on the whole sequence of
upward/downward movements.

This generates non-recombining trees.

[Draw picture on the blackboard]

Frédéric Godin (2020) MAST 397A Fall 2020 80 / 87

 Path dependent options

Path dependent options

The theoretical algorithm to compute the price in such framework is

analogous to what was described previously.
The steps in the proof would analogous.

Denote by H a sequence of upward/downward movements.

E.g. H = uud implies two upward movements followed by a
downward one.
Hu or Hd represent the sequence of movements h followed
respectively by an upward or a downward movement.

Frédéric Godin (2020) MAST 397A Fall 2020 81 / 87

 Path dependent options

Price for path dependent options

In this case, the option prices can be expressed analogously as
(
CT (H) if t = T ,
Vtθ (H) =
e −r qu Vt+1
θ (Hu) + q V θ (Hd) if t < T ,


d t+1

where CT (H) is the option payoff when a sequence of movements H of

length T occurs.

Similarly, the replicating portfolio composition during [t, t + 1) is when

sequence of movements H occured by time t is

(S) vt+1 (Hu) − vt+1 (Hd)

θt+1 (H) = ,
St (H)(u − d)
(B) uvt+1 (Hd) − dvt+1 (Hu)
θt+1 (H) = .
Bt+1 (u − d)

Frédéric Godin (2020) MAST 397A Fall 2020 82 / 87

 Diversifiable versus non-diversifiable risks

Pricing of different classes of risk

Derivatives are priced through a risk-neutral expectation (under the

risk-neutral measure Q) of the payoff discounted at the risk-free rate.

However, insurance risks (life insurance, casualty & property insurance)

are priced by taking an expectation under the physical measure P.

Rationale: there is a fundamental difference between both types of risks.

Frédéric Godin (2020) MAST 397A Fall 2020 83 / 87

 Diversifiable versus non-diversifiable risks

Diversifiable insurance risks

Insurance risks are often presumed to be diversifiable due to the limited

dependence between the various risks.
For instance, the death of the various policyholders are often
unrelated events.

Thus an insurance can diversify its risk by holding a large portfolio of

insurance policies.

By the Law of large numbers, the total loss of an insurer should be very
close to its expectation under the physical probability measure P.
justifies the use of the measure P to price insurance risks.

Frédéric Godin (2020) MAST 397A Fall 2020 84 / 87

 Diversifiable versus non-diversifiable risks

Undiversifiable financial market risks

Financial risks related to markets are not completely diversifiable;

presence of a systemic component.

For instance, during a stock market crash, most of the stock prices
simultaneously go down.

The holder of a portfolio of stocks cannot fully diversify its risk by holder
a very large number of stocks.

The rationale based on the Law of large numbers used previously for
insurance risks does not typically apply to financial risks.

Frédéric Godin (2020) MAST 397A Fall 2020 85 / 87

 Diversifiable versus non-diversifiable risks

Absence of arbitrage paradigm

For financial products, it is rather the absence of arbitrage paradigm

which is applied.

This approach ensures that financial products are priced consistently

between themselves so as not to allow for arbitrage opportunities.

The pricing of financial instruments is not absolute, but rather relative

with respect to other financial instruments.

As seen previously, such a method requires the use of a risk-neutral

measure Q.

Frédéric Godin (2020) MAST 397A Fall 2020 86 / 87

 Diversifiable versus non-diversifiable risks

References

Björk, T. (2004). Arbitrage theory in continuous time. Oxford. 2nd

edition.

Frédéric Godin (2020) MAST 397A Fall 2020 87 / 87