American options

In contrast to European contingent claims, an American claim may be exercised at

any time before the expiration date T .
In the study of American claims, one is more concerned with the price process and
the ‘optimal’ exercise policy by its holder.
If the holder of an American option exercises it at τ ⩽ T , τ is called an exercise
time.
What exercise times are ‘optimal’ ?
An admissible exercise time should be random itself and belong to the class of
stopping times.
Stopping times
We fix a probability space (Ω, F, P).
Definition (The discrete case)
Let F = (Ft )t∈N be a filtration. A map τ : Ω → N ∪ {∞} is a stopping time
(w.r.t. F), whenever
{τ = t} ∈ Ft (t ∈ N).
Stopping times
We fix a probability space (Ω, F, P).
Definition (The discrete case)
Let F = (Ft )t∈N be a filtration. A map τ : Ω → N ∪ {∞} is a stopping time
(w.r.t. F), whenever
{τ = t} ∈ Ft (t ∈ N).

Intuitively, this means that the ‘decision’ of whether to stop at time t (e.g., to
exercise an option or not) must be based only on information present at time t,
not on any future information.
Stopping times
We fix a probability space (Ω, F, P).
Definition (The discrete case)
Let F = (Ft )t∈N be a filtration. A map τ : Ω → N ∪ {∞} is a stopping time
(w.r.t. F), whenever
{τ = t} ∈ Ft (t ∈ N).

Intuitively, this means that the ‘decision’ of whether to stop at time t (e.g., to
exercise an option or not) must be based only on information present at time t,
not on any future information.
For options, this means that we base our decision on the price fluctuations up to
time t only.
Stopping times
We fix a probability space (Ω, F, P).
Definition (The discrete case)
Let F = (Ft )t∈N be a filtration. A map τ : Ω → N ∪ {∞} is a stopping time
(w.r.t. F), whenever
{τ = t} ∈ Ft (t ∈ N).

Intuitively, this means that the ‘decision’ of whether to stop at time t (e.g., to
exercise an option or not) must be based only on information present at time t,
not on any future information.
For options, this means that we base our decision on the price fluctuations up to
time t only.

Notation: For an integer interval J ⊆ N, we denote by TJ the class of stopping times

taking values in J.
Arbitrage price of American calls

An arbitrage price of an American call option is a price process Cta (t = 0, 1, . . . , T )

such that the extended financial market model remains arbitrage-free (a market trading
in riskless bonds, stocks, and American call options).
Arbitrage price of American calls

An arbitrage price of an American call option is a price process Cta (t = 0, 1, . . . , T )

such that the extended financial market model remains arbitrage-free (a market trading
in riskless bonds, stocks, and American call options).
Proposition. The arbitrage price of an American call option in the CRR arbitrage-free
market model (with r > 0) coincides with the arbitrage price of a European call option
with the same expiry date and strike price.
Arbitrage price of American calls

An arbitrage price of an American call option is a price process Cta (t = 0, 1, . . . , T )

such that the extended financial market model remains arbitrage-free (a market trading
in riskless bonds, stocks, and American call options).
Proposition. The arbitrage price of an American call option in the CRR arbitrage-free
market model (with r > 0) coincides with the arbitrage price of a European call option
with the same expiry date and strike price.
It suffices to show that the American call option should never be exercised before
maturity, since otherwise the issuer of the option would be able to make riskless profit.
Proof
The arbitrage price of an American call option in the CRR arbitrage-free market model (with r > 0) coincides
with the arbitrage price of a European call option with the same expiry date and strike price.

Denote by Ct the price process of an analogous European call; we have

Ct ⩾ (St − K )+ .
Proof
The arbitrage price of an American call option in the CRR arbitrage-free market model (with r > 0) coincides
with the arbitrage price of a European call option with the same expiry date and strike price.

Denote by Ct the price process of an analogous European call; we have

Ct ⩾ (St − K )+ .
Ct is non-negative, so it suffices to consider the case where St − K > 0.
Proof
The arbitrage price of an American call option in the CRR arbitrage-free market model (with r > 0) coincides
with the arbitrage price of a European call option with the same expiry date and strike price.

Denote by Ct the price process of an analogous European call; we have

Ct ⩾ (St − K )+ .
Ct is non-negative, so it suffices to consider the case where St − K > 0.
Indeed, assume not; Ct < St = K . Then it would be possible, with zero net initial
investment, to buy at t a call option, short a stock, and invest the sum
St − Ct > K in the saving account.
Proof
The arbitrage price of an American call option in the CRR arbitrage-free market model (with r > 0) coincides
with the arbitrage price of a European call option with the same expiry date and strike price.

Denote by Ct the price process of an analogous European call; we have

Ct ⩾ (St − K )+ .
Ct is non-negative, so it suffices to consider the case where St − K > 0.
Indeed, assume not; Ct < St = K . Then it would be possible, with zero net initial
investment, to buy at t a call option, short a stock, and invest the sum
St − Ct > K in the saving account.
By holding this portfolio unchanged up to maturity, we’d make a riskless profit.
Indeed, CT − ST + (1 + r )T −t (St − Ct ) > (ST − K )+ − ST + (1 + r )T −t K ⩾ 0.
Proof
The arbitrage price of an American call option in the CRR arbitrage-free market model (with r > 0) coincides
with the arbitrage price of a European call option with the same expiry date and strike price.

Denote by Ct the price process of an analogous European call; we have

Ct ⩾ (St − K )+ .
Ct is non-negative, so it suffices to consider the case where St − K > 0.
Indeed, assume not; Ct < St = K . Then it would be possible, with zero net initial
investment, to buy at t a call option, short a stock, and invest the sum
St − Ct > K in the saving account.
By holding this portfolio unchanged up to maturity, we’d make a riskless profit.
Indeed, CT − ST + (1 + r )T −t (St − Ct ) > (ST − K )+ − ST + (1 + r )T −t K ⩾ 0.
As we are in the arbitrage-free setting, Ct ⩾ St − K .
Proof, ctd

Using Ct ⩾ (St − K )+ , we will employ no-arbitrage assumption to get the

conclusion.
Proof, ctd

Using Ct ⩾ (St − K )+ , we will employ no-arbitrage assumption to get the

conclusion.
Assume that the issuer of an American call is able to sell the option at time 0 at
the price C0a > C0 . In order to profit from this transaction, the option writer
establishes a dynamic portfolio ψ that replicates the value process of the European
call and invests the remaining funds in the riskless asset.
Proof, ctd

Using Ct ⩾ (St − K )+ , we will employ no-arbitrage assumption to get the

conclusion.
Assume that the issuer of an American call is able to sell the option at time 0 at
the price C0a > C0 . In order to profit from this transaction, the option writer
establishes a dynamic portfolio ψ that replicates the value process of the European
call and invests the remaining funds in the riskless asset.
Suppose that the holder of the option decides to exercise it at instant t < T
(before the expiry date).
Proof, ctd

Using Ct ⩾ (St − K )+ , we will employ no-arbitrage assumption to get the

conclusion.
Assume that the issuer of an American call is able to sell the option at time 0 at
the price C0a > C0 . In order to profit from this transaction, the option writer
establishes a dynamic portfolio ψ that replicates the value process of the European
call and invests the remaining funds in the riskless asset.
Suppose that the holder of the option decides to exercise it at instant t < T
(before the expiry date).
Proof, ctd

Then the issuer of the option locks in a riskless profit, since the value of his
portfolio at time t satisfies

CT − (St − K )+ + (1 + r )t (C0a − C0 ) > 0.

Proof, ctd

Then the issuer of the option locks in a riskless profit, since the value of his
portfolio at time t satisfies

CT − (St − K )+ + (1 + r )t (C0a − C0 ) > 0.

Thus, the European and American call options are equivalent from the point of view of
arbitrage pricing theory. Both options have the same price and an American call should
never be exercised by its holder before expiry.
Proof, ctd

Then the issuer of the option locks in a riskless profit, since the value of his
portfolio at time t satisfies

CT − (St − K )+ + (1 + r )t (C0a − C0 ) > 0.

Thus, the European and American call options are equivalent from the point of view of
arbitrage pricing theory. Both options have the same price and an American call should
never be exercised by its holder before expiry.
The assumption r ⩾ 0 was necessary.
American options valuation
Consider the American put option on a stock with strike K and expiry T .
American options valuation
Consider the American put option on a stock with strike K and expiry T .

Theorem
The risk-neutral price process (Pta )t∈[0,T ] is given by

Pta = max EQ [(1 + r )−(τ −t) (K − Sτ )+ |Ft ] (t ⩽ T ).

τ ∈T[t,T ]

Moreover, for t ⩽ T , the stopping time τt∗ realising the maximum is given by

τt∗ = min{u ⩾ t : Pua = (K − Su )+ }.

American options valuation
Consider the American put option on a stock with strike K and expiry T .

Theorem
The risk-neutral price process (Pta )t∈[0,T ] is given by

Pta = max EQ [(1 + r )−(τ −t) (K − Sτ )+ |Ft ] (t ⩽ T ).

τ ∈T[t,T ]

Moreover, for t ⩽ T , the stopping time τt∗ realising the maximum is given by

τt∗ = min{u ⩾ t : Pua = (K − Su )+ }.

τt∗ is called the rational exercise time of an American put option assuming, the
option is alive at time t.
American options valuation
Consider the American put option on a stock with strike K and expiry T .

Theorem
The risk-neutral price process (Pta )t∈[0,T ] is given by

Pta = max EQ [(1 + r )−(τ −t) (K − Sτ )+ |Ft ] (t ⩽ T ).

τ ∈T[t,T ]

Moreover, for t ⩽ T , the stopping time τt∗ realising the maximum is given by

τt∗ = min{u ⩾ t : Pua = (K − Su )+ }.

τt∗ is called the rational exercise time of an American put option assuming, the
option is alive at time t.
Bellman’s principle (adapted to our setting) can reduce the optimal stopping
problem to an explicit recursive procedure for the value process.
American put/calls pricing
Theorem (Corollary (Bellman’s principle in our setting))
Let us define the process (Ut )t=0,...,T recursively backwards by UT = (K − ST )+ and

Ut = max{(K − St )+ , (1 + r )−1 EQ [Ut+1 |Ft ]} (t < T ).

The the risk-neutral price Pta of the American put option at t equals Ut and the
rational exercise time after t is given by

τt∗ = min{u ⩾ t : Uu − (K − Su )+ }.

Therefore, τT∗ = T and

τt∗ = t1{Ut =(K −St )+ } + τt+1

∗
1{Ut >(K −St )+ } (t < T ).
American put/calls pricing
Theorem (Corollary (Bellman’s principle in our setting))
Let us define the process (Ut )t=0,...,T recursively backwards by UT = (K − ST )+ and

Ut = max{(K − St )+ , (1 + r )−1 EQ [Ut+1 |Ft ]} (t < T ).

The the risk-neutral price Pta of the American put option at t equals Ut and the
rational exercise time after t is given by

τt∗ = min{u ⩾ t : Uu − (K − Su )+ }.

Therefore, τT∗ = T and

τt∗ = t1{Ut =(K −St )+ } + τt+1

∗
1{Ut >(K −St )+ } (t < T ).

This is an extension of the backward recursive approach to the valuation of European

contingent claims.
American put/calls valuation

One can show that

Pta = max{(K − St )+ , (1 + r )−1 EQ [Pt+1

a
|Ft ]} (t < T )

subject to PTa = (K − ST )+ .
American put/calls valuation

One can show that

Pta = max{(K − St )+ , (1 + r )−1 EQ [Pt+1

a
|Ft ]} (t < T )

subject to PTa = (K − ST )+ .
In the case of CRR, this formula reduces the valuation problem to the simple
single-period case.
In CRR, the risk-neutral price of the American put reduces to the following
recursive scheme:

Pta = max{(K − St )+ , (1 + r )−1 (qPt+1

au ad
+ (1 − q)Pt+1 )} (t < T )

subject to PTa = (K − ST )+ .
au , P ad represent the values of the American put in the next step corresponding
Pt+1 t+1
to the upward/downward movements of the stock given from a given note of the
binomial tree.
American put/calls pricing: Example

We consider here the CRR binomial model with the horizon date T = 2 and the
risk-free rate r = 0.2.
The stock price S for t = 0 and t = 1 equals

S0 = 10, Su = 13.2, Sd = 10.8.

Let X a be the American call option with maturity date T = 2 and the following
payoff process
g (St , t) = (St − Kt )+ .

The strike Kt is variable and satisfies

K0 = 9, K1 = 9.9, K2 = 12.
American put/calls pricing: example, ctd.

We will first compute the arbitrage price πt (X a ) of this option at times t = 0, 1, 2

and the rational exercise time τ ∗ .
Subsequently, we will compute the replicating strategy for X a up to the rational
exercise time τ ∗ .
We start by noting that the unique risk-neutral probability measure P satisfies

1+r −d (1 + r )S0 − Sd 12 − 10.8

p̂ = = = = 0.5
u−d Su − Sd 13.2 − 10.8

The dynamics of the stock price under P

b are given by the first exhibit (note that
Su2 = Sd 2 ).
The second exhibit represents the price of the call option.
American put/calls pricing
American put/calls pricing
American put/calls pricing: example

Holder. The rational holder should exercise the American option at time t = 1 if
the stock price rises during the first period. Otherwise, the option should be held
till time 2. Hence τ ∗ : Ω → {0, 1, 2} equals

τ ∗ (ω) = 1 for ω ∈ {ω1 , ω2 }

τ ∗ (ω) = 2 for ω ∈ {ω3 , ω4 }

Issuer. We now take the position of the issuer of the option. At t = 0, we need to
solve
1.2ϕ0 + 13.2ϕ1 = 3.3

1.2ϕ0 + 10.8ϕ1 = 0.94

Hence (ϕ0 , ϕ1 ) = (−8.067, 0.983) for all ω.

American put/calls pricing: example

If the stock price rises during the first period, the option is exercised and thus we
do not need to compute the strategy at time 1 for ω ∈ {ω1 , ω2 }.
If the stock price falls during the first period, we solve

1.2ϕb0 + 14.256ϕ1 = 2.256

1.2ϕb0 + 11.664ϕ1 = 0

Hence (ϕb0 , ϕ1 ) = (−8.46, 0.8704) for ω ∈ {ω3 , ω4 }.

Note that ϕb0 = −8.46 is the amount of cash borrowed at time 1, rather than the
number of units of the savings account B.
The replicating strategy Φ = (ϕ0 , ϕ1 ) is defined at time 0 for all ω and it is defined
at time 1 on the event {ω3 , ω4 } only.
Implementation of CRR

Fix maturity T the continuously compounded interest rate r .

From the market data for stock prices, one can estimate the stock price volatility σ
per one time unit (typically, one year); up to now we’ve assumed that
t = 0, 1, 2, . . . , T which means that ∆t = 1. In general, we can set n = T /∆t.
Two widely used conventions for obtaining u and d from σ, r are:
the Cox–Ross–Rubinstein (CRR) parametrisation:
√
u = eσ ∆t
, d = 1/u,

the Jarrow–Rudd (JR) parameterisation:

σ2 √ σ2 √
u = exp((r − )∆t + σ ∆t), d = exp((r − )∆t − σ ∆t).
2 2
Approximations of solutions to SDEs
Approximations of solutions to SDEs
A solution to a SDE

dXt = a(t, Xt )dt + b(t, Xt )dWt

with the initial condition X0 = x is a process (Xt ) of the form

Z t Z 1
Xt = X0 + a(s, Xs )ds + b(s, Xs ) dWs .
0 0
Approximations of solutions to SDEs
A solution to a SDE

dXt = a(t, Xt )dt + b(t, Xt )dWt

with the initial condition X0 = x is a process (Xt ) of the form

Z t Z 1
Xt = X0 + a(s, Xs )ds + b(s, Xs ) dWs .
0 0

As with ODEs/PDEs, not always a closed-form solution can be presented.

Approximations of solutions to SDEs
A solution to a SDE

dXt = a(t, Xt )dt + b(t, Xt )dWt

with the initial condition X0 = x is a process (Xt ) of the form

Z t Z 1
Xt = X0 + a(s, Xs )ds + b(s, Xs ) dWs .
0 0

As with ODEs/PDEs, not always a closed-form solution can be presented.

Black–Scholes

dSt = St µ dt + St σ dWt .
Approximations of solutions to SDEs
A solution to a SDE

dXt = a(t, Xt )dt + b(t, Xt )dWt

with the initial condition X0 = x is a process (Xt ) of the form

Z t Z 1
Xt = X0 + a(s, Xs )ds + b(s, Xs ) dWs .
0 0

As with ODEs/PDEs, not always a closed-form solution can be presented.

Black–Scholes

dSt = St µ dt + St σ dWt .

Here, a closed to a degree formula exists but often this is not the case
Approximations of solutions to SDEs

An SDE governing certain interest rate models

p
dXt = a(b − Xt ) dt + c Xt dWt .
Approximations of solutions to SDEs

An SDE governing certain interest rate models

p
dXt = a(b − Xt ) dt + c Xt dWt .

It does not, in general, have any explicit solutions.

For computational purposes it is useful to consider a discretised Wiener process. We
thus subdivide the time interval [0, T ] into N subintervals by setting δt = T /N and

T
tn = n · δt = n · .
N
Approximations of solutions to SDEs
Further, due to the properties of the Wiener process we can simulate its values at the
selected points by

Wtn+1 = Wtn + ∆Wn , Wt0 = W0 = 0,

where ∆Wn ∼ N (0, δt) are indep.