Example: European call options

European call option gives the Buyer the right to buy one share of an equity
at a pre-specified time t = 1 (the expiration date) for an amount K (called the
strike price) in cash.
The strike price K is written into the contract at t = 0. Buyer is not obliged to
exercise at expiration.
The call has a premium V0 at t = 0 that Buyer pays the Seller.
Example: European call options

European call option gives the Buyer the right to buy one share of an equity
at a pre-specified time t = 1 (the expiration date) for an amount K (called the
strike price) in cash.
The strike price K is written into the contract at t = 0. Buyer is not obliged to
exercise at expiration.
The call has a premium V0 at t = 0 that Buyer pays the Seller.
It follows that the value of the call option at expiration is
max{S1 − K , 0} := (S1 − K )0 .
Example: European call options

European call option gives the Buyer the right to buy one share of an equity
at a pre-specified time t = 1 (the expiration date) for an amount K (called the
strike price) in cash.
The strike price K is written into the contract at t = 0. Buyer is not obliged to
exercise at expiration.
The call has a premium V0 at t = 0 that Buyer pays the Seller.
It follows that the value of the call option at expiration is
max{S1 − K , 0} := (S1 − K )0 .
The market value of the call, V0 , is uncertain as it depends on the Equity.
Example: European call options–two-scenario market
Example: European call options–two-scenario market

Assume there is a riskless bond with rate of return r .

Example: European call options–two-scenario market

Assume there is a riskless bond with rate of return r .

Suppose there are only two possible market scenarios: ω1 , ω2 , that is, S1 (ωi ) = di
(i = 1, 2).
Example: European call options–two-scenario market

Assume there is a riskless bond with rate of return r .

Suppose there are only two possible market scenarios: ω1 , ω2 , that is, S1 (ωi ) = di
(i = 1, 2).
By EMH: d1 ⩽ K , S0 e r ⩽ d2 ; otherwise arbitrage appears.
Example: European call options–two-scenario market

Assume there is a riskless bond with rate of return r .

Suppose there are only two possible market scenarios: ω1 , ω2 , that is, S1 (ωi ) = di
(i = 1, 2).
By EMH: d1 ⩽ K , S0 e r ⩽ d2 ; otherwise arbitrage appears.
It follows that the value of the call option at expiration is
max{S1 − K , 0} := (S1 − K )0 .
Example: European call options–two-scenario market

Assume there is a riskless bond with rate of return r .

Suppose there are only two possible market scenarios: ω1 , ω2 , that is, S1 (ωi ) = di
(i = 1, 2).
By EMH: d1 ⩽ K , S0 e r ⩽ d2 ; otherwise arbitrage appears.
It follows that the value of the call option at expiration is
max{S1 − K , 0} := (S1 − K )0 .
The market value of the call, V0 , is uncertain as it depends on the Equity.
Example: European call options–two-scenario market

Assume there is a riskless bond with rate of return r .

Suppose there are only two possible market scenarios: ω1 , ω2 , that is, S1 (ωi ) = di
(i = 1, 2).
By EMH: d1 ⩽ K , S0 e r ⩽ d2 ; otherwise arbitrage appears.
It follows that the value of the call option at expiration is
max{S1 − K , 0} := (S1 − K )0 .
The market value of the call, V0 , is uncertain as it depends on the Equity.

Mini theorem 2 In an arbitrage-free market, the fair premium is

S0 e r − d1 −r
V0 = (d2 − K ) ·e .
d2 − d1
Financed Call Option strategy
r
S e −d
Fair value: V0 = (d2 − K ) d0 −d 1 · e −r .
2 1

Consider the case V0 < v .

At t = 0 sell a = (d2 − K )/(d2 − d1 ) shares. Use V0 to buy 1 call option and

invest aS0 − V0 (the remainder) in the Bond.
At t = 1, you must return a shares and you will exercise the option should ω2 take
place but not ω1 .
ω1 You owe ad1 to repay the a shares, but your cash on hand is
(aS0 − V0 )e r = ad1 + (aS0 − ad1 e −r )e r − V0 e r
= ad1 + (v − V0 )e r > ad1 .
ω2 You owe ad2 to repay the a shares but your cash on hand (from selling the option
(= d2 − K ) and from selling the bond) is
d2 − K + (aS0 − V0 )e r > d2 − K + (aS0 − v )e r
= d2 − K + (aS0 − a(S0 − d1 )e −r ))e r
= d2 − K + ad1 = ad2 .

The case V0 < v is to some extent the converse. At t = 0, sell 1 call option, borrow aS0 − V0
and use the sum to buy a shares (if V0 > aS0 , no need to borrow anything.)
FToAP in action
r
S e −d
V0 = (d2 − K ) d0 −d 1 · e −r .
2 1
FToAP in action
r
S e −d
V0 = (d2 − K ) d0 −d 1 · e −r .
2 1

There is a probability measure π on the two scenarios that determines prices by

discounted expectation
FToAP in action
r
S e −d
V0 = (d2 − K ) d0 −d 1 · e −r .
2 1

There is a probability measure π on the two scenarios that determines prices by

discounted expectation
S0 = π(ω1 )e −r d1 + π(ω2 )e −r d2 .
FToAP in action
r
S e −d
V0 = (d2 − K ) d0 −d 1 · e −r .
2 1

There is a probability measure π on the two scenarios that determines prices by

discounted expectation
S0 = π(ω1 )e −r d1 + π(ω2 )e −r d2 .
Recall that we must have d1 ⩽ S0 e r ⩽ d2 .
FToAP in action
r
S e −d
V0 = (d2 − K ) d0 −d 1 · e −r .
2 1

There is a probability measure π on the two scenarios that determines prices by

discounted expectation
S0 = π(ω1 )e −r d1 + π(ω2 )e −r d2 .
Recall that we must have d1 ⩽ S0 e r ⩽ d2 .

As we have only two states:

π(ω1 ) = (d2 − S0 e r )/(d2 − d1 )

π(ω2 ) = (S0 e r − d1 )/(d2 − d1 ).

Risk-neutral measures: the set-up.
Risk-neutral measures: the set-up.
Let us consider a market M in which there are freely traded assets:

risk-less instrument B,
risky instruments A1 , . . . , AK ,

whose share prices in market scenario ωi are StB (ωi ) and St1 (ωi ), . . . , StK (ωi ) at time t,
respectively, where Ω = {ω1 , . . . , ωk } is the sample space of market scenarios.
Risk-neutral measures: the set-up.
Let us consider a market M in which there are freely traded assets:

risk-less instrument B,
risky instruments A1 , . . . , AK ,

whose share prices in market scenario ωi are StB (ωi ) and St1 (ωi ), . . . , StK (ωi ) at time t,
respectively, where Ω = {ω1 , . . . , ωk } is the sample space of market scenarios. The
riskless asset B’s price equals B0 = 1 and B1 = 1 + r for some constant r > −1.
Risk-neutral measures: the set-up.
Let us consider a market M in which there are freely traded assets:

risk-less instrument B,
risky instruments A1 , . . . , AK ,

whose share prices in market scenario ωi are StB (ωi ) and St1 (ωi ), . . . , StK (ωi ) at time t,
respectively, where Ω = {ω1 , . . . , ωk } is the sample space of market scenarios. The
riskless asset B’s price equals B0 = 1 and B1 = 1 + r for some constant r > −1.
Investor’s personal beliefs about the future behaviour of the stock prices are represented
by a probability measure P on Ω with P({ωi }) > 0.
Risk-neutral measures: the set-up.
Let us consider a market M in which there are freely traded assets:

risk-less instrument B,
risky instruments A1 , . . . , AK ,

whose share prices in market scenario ωi are StB (ωi ) and St1 (ωi ), . . . , StK (ωi ) at time t,
respectively, where Ω = {ω1 , . . . , ωk } is the sample space of market scenarios. The
riskless asset B’s price equals B0 = 1 and B1 = 1 + r for some constant r > −1.
Investor’s personal beliefs about the future behaviour of the stock prices are represented
by a probability measure P on Ω with P({ωi }) > 0.
A derivative security/contingent claim is a tradeable asset whose value V1 at t = 1 is a
function of the market scenarios ωi (put simply, a random variable)
In which circumstances a general single-period market model M = (B, S1 , . . . , SK ) is
arbitrage-free?
Risk-neutral measures
Definition
A measure π on Ω is risk-neutral, whenever
π({ωi }) > 0 for all i (that is, π is equivalent to P)
Eπ (S1j ) = (1 + s)S0j (= e −r S0j ).

A trading strategy in M is defined as (x, ψ 1 , . . . , ψ K ) ∈ RK +1 , where x is the initial

wealth to be invested in the portfolio (ψ 1 , . . . , ψ K ) at t = 0. If adopted, the cash value
of the portfolio at t = 1 is
K K
ψ j S0j )(1 + s) + ψ j S1j
X X
V1 (x, ψ1 , . . . , ψK ) = (x −
j=1 j=1

Gain process: G1 = V1 − V0 (a random process, can be negative!)

K K
S0j )s ψ j ∆S1j ,
X X
j
G1 (x, ψ1 , . . . , ψK ) = (x − ψ +
j=1 j=1
Risk-neutral measures
Definition
A measure π on Ω is risk-neutral, whenever
π({ωi }) > 0 for all i (that is, π is equivalent to P)
Eπ (S1j ) = (1 + s)S0j (= e −r S0j ).

A trading strategy in M is defined as (x, ψ 1 , . . . , ψ K ) ∈ RK +1 , where x is the initial

wealth to be invested in the portfolio (ψ 1 , . . . , ψ K ) at t = 0. If adopted, the cash value
of the portfolio at t = 1 is
K K
ψ j S0j )(1 + s) + ψ j S1j
X X
V1 (x, ψ1 , . . . , ψK ) = (x −
j=1 j=1

Gain process: G1 = V1 − V0 (a random process, can be negative!)

K K
S0j )s ψ j ∆S1j ,
X X
j
G1 (x, ψ1 , . . . , ψK ) = (x − ψ +
j=1 j=1
Arbitrages
A trading strategy/portfolio v ∈ RK +1 is an arbitrage, when G1 (v )(ωi ) > 0 for all
i = 1, . . . , K .
V0 (v ) ⩽ 0 ⇒ V1 (v )(ωi ) > 0,
V0 (v ) < 0 ⇒ V1 (ωi ) ⩾ 0
for all i = 1, . . . , K .
Arbitrages
A trading strategy/portfolio v ∈ RK +1 is an arbitrage, when G1 (v )(ωi ) > 0 for all
i = 1, . . . , K .
V0 (v ) ⩽ 0 ⇒ V1 (v )(ωi ) > 0,
V0 (v ) < 0 ⇒ V1 (ωi ) ⩾ 0
for all i = 1, . . . , K .
Fundamental theorem of asset pricing

Theorem (Fundamental theorem of asset pricing (Harrison & Pliska, 1981))

A general single-period model M = (B, S1 , . . . , SK ) is arbitrage-free if and only if there
exists a risk-neutral probability measure for M.

We have to show that if the market does not admit arbitrages, then it has a risk-neutral
measure π:
Fundamental theorem of asset pricing

Theorem (Fundamental theorem of asset pricing (Harrison & Pliska, 1981))

A general single-period model M = (B, S1 , . . . , SK ) is arbitrage-free if and only if there
exists a risk-neutral probability measure for M.

We have to show that if the market does not admit arbitrages, then it has a risk-neutral
measure π: S0j = e −r Ki=1 π(ωi )S1j .
P
Fundamental theorem of asset pricing

Theorem (Fundamental theorem of asset pricing (Harrison & Pliska, 1981))

A general single-period model M = (B, S1 , . . . , SK ) is arbitrage-free if and only if there
exists a risk-neutral probability measure for M.

We have to show that if the market does not admit arbitrages, then it has a risk-neutral
measure π: S0j = e −r Ki=1 π(ωi )S1j .
P

Minkowski’s hyperplane separation theorem. Let A and B be two disjoint

non-empty closed, convex subsets of Rd one of each is compact. Then there exist
a non-zero vector v and c ∈ R such that

⟨x, v ⟩ > c and ⟨y , v ⟩ < c

for all x ∈ A and y ∈ B; i.e., the hyperplane ⟨·, v ⟩ = c (for which v the normal vector)
separates A from B.
Illustration to the hyperplane separation theorem
Proof of the FToAP

Suppose that π is a risk-neutral measure. For any v , we have

K
X
V0 (v ) = ⟨v , S⟩ = e −r V1 (v )(ωi ).
i=1

If V1 (v )(ωi ) > 0 [ < 0] for every i (necessary for arbitrage), then V0 (v ) > 0 [ < 0], so
there is no arbitrage.
Proof of the FToAP
Convention: for a measure µ on Ω and x ∈ Ω we write µ(x) instead of µ({x})

We seek a probability measure π on Ω s.t. S0j = e −r K j

P
i=1 π(ωi )S1 for any j.
(We don’t have to care about B as it’s price is deterministic.)
Proof of the FToAP
Convention: for a measure µ on Ω and x ∈ Ω we write µ(x) instead of µ({x})

We seek a probability measure π on Ω s.t. S0j = e −r K j

P
i=1 π(ωi )S1 for any j.
(We don’t have to care about B as it’s price is deterministic.)
j
E = {y ∈ RK : ∃µ∈Prob Ω ∀j⩽K yj = e −r K
P
i=1 µ(ωi )S1 }
Proof of the FToAP
Convention: for a measure µ on Ω and x ∈ Ω we write µ(x) instead of µ({x})

We seek a probability measure π on Ω s.t. S0j = e −r K j

P
i=1 π(ωi )S1 for any j.
(We don’t have to care about B as it’s price is deterministic.)
j
E = {y ∈ RK : ∃µ∈Prob Ω ∀j⩽K yj = e −r K
P
i=1 µ(ωi )S1 }
E is closed and convex (why?)
Proof of the FToAP
Convention: for a measure µ on Ω and x ∈ Ω we write µ(x) instead of µ({x})

We seek a probability measure π on Ω s.t. S0j = e −r K j

P
i=1 π(ωi )S1 for any j.
(We don’t have to care about B as it’s price is deterministic.)
j
E = {y ∈ RK : ∃µ∈Prob Ω ∀j⩽K yj = e −r K
P
i=1 µ(ωi )S1 }
E is closed and convex (why?)
PK PK
y , w ∈ E, t ∈ [0, 1]: yj = e −r i=1 µ(ωi )S1j , wj = e −r i=1 ν(ωi )S1j so
K
tyj + (1 − t)wj = e −r i=1 [tµ(ωi ) + (1 − t)ν(ωi )]S1j
P
Proof of the FToAP
Convention: for a measure µ on Ω and x ∈ Ω we write µ(x) instead of µ({x})

We seek a probability measure π on Ω s.t. S0j = e −r K j

P
i=1 π(ωi )S1 for any j.
(We don’t have to care about B as it’s price is deterministic.)
j
E = {y ∈ RK : ∃µ∈Prob Ω ∀j⩽K yj = e −r K
P
i=1 µ(ωi )S1 }
E is closed and convex (why?)
PK PK
y , w ∈ E, t ∈ [0, 1]: yj = e −r i=1 µ(ωi )S1j , wj = e −r i=1 ν(ωi )S1j so
K
tyj + (1 − t)wj = e −r i=1 [tµ(ωi ) + (1 − t)ν(ωi )]S1j
P

Claim: S = (S01 , . . . , S0K ) ∈ E. Contrapositively, if S ∈

/ E, then there are arbitrages.
Proof of the FToAP
Convention: for a measure µ on Ω and x ∈ Ω we write µ(x) instead of µ({x})

We seek a probability measure π on Ω s.t. S0j = e −r K j

P
i=1 π(ωi )S1 for any j.
(We don’t have to care about B as it’s price is deterministic.)
j
E = {y ∈ RK : ∃µ∈Prob Ω ∀j⩽K yj = e −r K
P
i=1 µ(ωi )S1 }
E is closed and convex (why?)
PK PK
y , w ∈ E, t ∈ [0, 1]: yj = e −r i=1 µ(ωi )S1j , wj = e −r i=1 ν(ωi )S1j so
K
tyj + (1 − t)wj = e −r i=1 [tµ(ωi ) + (1 − t)ν(ωi )]S1j
P

Claim: S = (S01 , . . . , S0K ) ∈ E. Contrapositively, if S ∈

/ E, then there are arbitrages.
By the hyperplane separation theorem there are v ∈ RK and c ∈ R s.t.
⟨y , v ⟩ > c > ⟨S, v ⟩ for all y ∈ E. This means that
PK j PK j
i=1 vj S1 (ωi ) > c > j=1 vj S0 .
The portfolio (−c, v1 , . . . , vn ) is an arbitrage!
Hedging

A portfolio θ = (θ1 , . . . , θK ) comprising assets A1 , . . . , AK is replicating for B,

whenever for any i we have
K
X
B
S1 (ωi ) = θj S j (ωi ).
j=1
Hedging

A portfolio θ = (θ1 , . . . , θK ) comprising assets A1 , . . . , AK is replicating for B,

whenever for any i we have
K
X
B
S1 (ωi ) = θj S j (ωi ).
j=1

Replicating portfolios enable financial institutions that sell asset B (for example, a call
option) to hedge: for each share of asset B sold, they buy θj shares of Aj and hold
them to t = 1. Then at t = 1, net gain = net loss = 0. The financial institution makes
its money (usually) by charging the buyer a transaction fee/premium at t = 0.
Complete markets

A risk-neutral measure for a market (if it exists) need not be unique. Markets that
admit a unique risk-neutral measure are termed complete.
Theorem (Completeness theorem)
Let us consider a single-period model M = (B, S1 , . . . , SK ). Suppose that M is
arbitrage-free. If for every contingent claim there is a replicating portfolio in
S1 , . . . , SK , then M is complete. Conversely, if M is complete, and if the unique
equilibrium measure π gives positive probability to every market scenario ωi , then for
every contingen claim there is a replicating portfolio in S1 , . . . , SK .

As contingent claims form a vector space, the completeness theorem really says that M
is complete if and only if S1 , . . . , SK form a basis of the space of contingent claims.
Conditional Expectation w.r.t. a sub-σ-algebra
Let (Ω, F, P) be a probability space and X : Ω → Rn an integrable random variable.
Consider H ⊆ F, a sub-σ-algebra of F.
Definition of Conditional Expectation
The conditional expectation of X with respect to H, denoted E(X | H), is defined as
any H-measurable function that satisfies the following for all H ∈ H:
Z Z
E(X | H) dP = X dP.
H H

Radon–Nikodym Derivative
The measure µX (F ) = F X dP is absolutely continuous w.r.t. P, leading to:
R

dµX |H
E(X | H) = .
dP|H

Here, µX ◦ h and P ◦ h denote the restrictions of µX and P to H, and the derivatives

Conditional Expectation w.r.t. another random variable

Consider a random variable Y : Ω → U and a measurable space (U, Σ).

Let (Ω, F, P) be a probability space, X = (X1 , . . . , Xn ) : Ω → Rn an integrable random
variable (i.e., E|Xi | < ∞ for each i), and H ⊂ F a sub-σ-algebra.
X need not be H-measurable. A conditional expectation of X w.r.t. H, denoted as
E(X | H), is any H-measurable function Ω → Rn satisfying:
Z Z
E(X | H) dP = X dP (H ∈ H).
H H
Conditional expectations
X and Y are both discrete. Then the conditional expectation of X given the event
Y = y can be considered as function of y for y in the range of Y :
X X P(X = x, Y = y )
E(X | Y = y ) = x P(X = x | Y = y ) = x· .
P(Y = y )
x∈X [Ω] x∈X [Ω]

X is continuous and Y is discrete. Then the conditional expectation is

Z
E(X | Y = y ) = xfX (x | Y = y ) dx,
X [Ω]
f
X ,Y (x,y )
with fX (x | Y = y ) = P(Y =y ) (where fX ,Y is the joint density of X and Y ) being
the conditional density of X given Y = y .
X and Y are continuous. The conditional expectation is
Z
E(X | Y = y ) = x · fX |Y (x | y ) dx,
X [Ω]
fX ,Y (x,y )
where fX |Y (x | y ) = fY (y )
Conditional expectation

Pulling out independent factors:

If X is independent of H, then E(X | H) = EX .
If X is independent of σ(Y , H), then E(X · Y | H) = EEX E(Y | H).
If X , Y are independent, G, H are independent, X is independent of H and Y is
independent of G, then E(E(X · Y | G) | H) = EX · EY = E(E(X · Y | H) | G).
Stability:
If X is H-measurable, then E(X | H) = X .
If Z is a random variable, then E(f (Z ) | Z ) = f (Z ). E.g., E(Z | Z ) = Z .
Pulling out known factors:
If X is H-measurable, then E(X · Y | H) = X · E(Y | H).
If Z is a random variable, then E(f (Z )Y | Z ) = f (Z )E(Y | Z ).
Law of total expectation: E(E(X | H)) = EX .
Tower property:
For sub-σ-algebras H1 ⊂ H2 ⊂ F we have E(E(X | H2 ) | H1 ) = E(X | H1 ).
If Z is a H-measurable, then σ(Z ) ⊂ H and E(E(X | H) | Z ) = E(X | Z ).
Conditional expectation

Doob martingale property: the above with Z = E(X | H) (which is H-measurable), and
using also E(Z | Z ) = Z , gives E(X | E(X | H)) = E(X | H).
For random variables X , Y we have E(E(X | Y ) | f (Y )) = E(X | f (Y )).
For random variables X , Y , Z we have E(E(X | Y , Z ) | Y ) = E(X | Y ).
Linearity: E(X1 + aX2 | H) = E(X1 | H) + aE(X2 | H) for a ∈ R.
Positivity: If X ⩾ 0 then E(X | H) ⩾ 0.
Monotonicity: If X1 ⩽ X2 then E(X1 | H) ⩽ E(X2 | H).
Monotone convergence: If 0 ⩽ Xn ↑ X then E(Xn | H) ↑ E(X | H).
Dominated convergence: If Xn → X and |Xn | ⩽ Y with Y ∈ L1 , then
E(Xn | H) → E(X | H).
Fatou’s lemma: E(lim inf n→∞ Xn | H) ⩽ lim inf n→∞ E(Xn | H) when
E(inf n Xn | H) > −∞
Jensen’s inequality: if f : R → R is a convex function, then f (E(X | H)) ⩽ E(f (X ) | H).