Sei sulla pagina 1di 17

The American Knock-Out Put Option

Luluwah Al-Fagih

First version: 21 March 2014

Research Report No. 3, 2014, Probability and Statistics Group


School of Mathematics, The University of Manchester
Research Report No. 3, 2014, Probab. Statist. Group Manchester (16 pp)

The American Knock-Out Put Option


Luluwah Al-Fagih

We show that the optimal stopping boundary for the American knock-out
put option with finite horizon can be characterized as the unique solution of a
nonlinear integral equation arising from the early exercise premium representa-
tion. The proof relies on Peskir’s local time-space formula [13] and the approach
is parallel to the one used by Peskir in [14]. Using these results we perform a
financial analysis of the American knock-out put option.

1. Introduction
Barrier options on stocks have been traded in OTC markets for over five decades. Although
barrier options are rather well developed in comparison to other exotic options in terms of
pricing and other features, they are not easily available to the general public. The relatively
inexpensive price of barrier options has led to their extensive use by investors in managing
risks related to commodities, FX and interest rate exposures. Despite this seemingly attractive
feature, an investor buying a barrier option is in fact taking a risk that may be much greater
than that of a vanilla option since the option can be terminated at anytime prior to maturity
or, conversely, inactive for the whole lifetime of the option.
There are two types of barrier options: knock-in and knock-out; both are exactly like vanilla
options except they are activated or terminated (respectively) once the price of the underlying
asset hits a certain level. Within these two categories, we can have: up barriers, where the
barrier level is above the spot (initial) price of the asset, and down barriers, where the barrier
level is below the spot price. If an investor expects the price of the underlying asset to fluctuate
strongly, knock-in barriers can lead to a higher profit as a result of the lower cost in buying
the option. Similarly, investors can reduce costs with a knock-out barrier if they expect the
price of the underlying asset to remain within a stable price range. Barrier options are path-
dependent and can take either American or European form. Barrier options sometimes come
with a rebate, which is cash paid out to the option holder if the option expires worthless as a
result of the option being knocked-out or failing to knock-in. Up barriers can be both above
or below the strike price of the option, although the latter is better known as a reverse barrier
option.
There are a number of established methods for pricing European barrier options and these
have been discussed extensively in the literature (see for example [3], [4], [6], [11], [18] and [19]).
Carr [1] presents a comprehensive and historic overview of the literature surrounding barrier
Mathematics Subject Classification 2010. Primary 91G80, 60G40. Secondary 35R35, 45G10.
Key words and phrases: Barrier put option, American put option, arbitrage-free price, optimal exercise
boundary, geometric Brownian motion, optimal stopping, free- boundary problem, nonlinear integral equation,
local time-space calculus.

1
options and treats some specific variations of European barriers. Mijatović [12] obtains a semi-
analytic solution for a system of Volterra integral equations in the case of constant double
barrier option (a European option with two barriers) and discusses a numerical algorithm for
time-dependent double barriers. Karatzas and Wang [8] obtain a closed-form expression for
the price and optimal boundary of the perpetual American put option in the presence of an
up-and-out barrier using the method of variational inequalities to solve the problem explicitly.
Chang, Kang, Kim and Kim [2] provide a numerical approximation scheme for a finite horizon
American up-and-out barrier option. Of particular interest to us is the work of Gao, Huang
and Subrahmanyam [5] who propose an early exercise premium representation for the American
knock-out call and put options in terms of the optimal exercise boundary. In this paper, we
provide a rigorous derivation of this representation that does not require the assumption that
the value function V is regular on the optimal exercise boundary b . The proof uses a change-
of-variable formula for local time on curves derived by Peskir in [13]. We focus on the knock-out
put option of American type (without rebates) and show that by use of the same change-of-
variable formula the optimal stopping boundary can be characterized as the unique solution of
a nonlinear integral equation arising from the early exercise premium representation. We also
look briefly at a reverse up and out put option (without rebates) and prove that it is always
optimal for the investor to exercise immediately.
The paper is organised as follows. In Section 2, we formulate the American knock-out put
option problem and show the effect of varying the barrier level above or below the strike on
the optimal stopping problem. In Section 3, we present the main result and proof. The results
are in line with Peskir’s results on the American put option in [14]. Finally, using these results
in Section 4 we conduct a financial analysis of the American knock-out put option making
comparisons with the European knock-out and American put options. This analysis provides
further insight into the American knock-out put option and highlights the issue of market
liquidity, thereby helping to lay the path for future work on a new type of barrier option which
may address this problem.

2. Problem formulation of the American knock-out put option


1. Let W = (Wt )0≤t≤T be a standard Brownian motion defined on a probability space
(Ω, F, P) , where T > 0 is the expiration date or maturity. For any interest rate r > 0 ,
volatility σ > 0 , strike price K and barrier level � , the arbitrage-free price of the American
knock-out put option is given by
� �
(2.1) V = sup E e−rτ (K − Xτ )+ I (Mτ < �)
0≤τ ≤T

where τ is a stopping time of the geometric Brownian motion X = (Xt )0≤t≤T solving

(2.2) dXt = rXt dt + σXt dWt (X0 = x)

with x < � and Mt = max0≤s≤t Xs denotes the maximum of the stock price process X.
Recall that the unique strong solution to (2.2) is given by
� �
(2.3) Xt = x exp σWt + (r − σ 2 /2) t

2
Figure 1: An illustration of how a stopped process can be used to
model the behaviour of an up-and-out put option. The option knocking
out is financially the same as the price process being ‘stopped’ as soon
as it hits the barrier.

under P for t ∈ [0, T ] where P is the (unique) equivalent martingale measure. The process
X is strong Markov with infinitesimal generator given by

∂ σ2 ∂ 2
(2.4) LX = rx + x2 2 .
∂x 2 ∂x
The option knocking out is (financially) equivalent to the price process being ‘stopped’ as soon
as it hits the barrier. We make use of the Markov structure of X and introduce a new process
X � = (Xt� )t≥0 which represents the process X stopped once it hits the barrier level � . We
define this process by Xt� = (Xt∧τ� )t≥0 where τ� is the first hitting time of the level � , given
by

(2.5) τ� = inf {t ≥ 0 : Xt ≥ �} .

This is particularly helpful since it means that we do not need to monitor the maximum process
M . Moreover, the process X � behaves exactly like the process X for all times t before τ�
which means most of the properties of X follow naturally for X � .
2. Summarising the preceding facts, we see that the American knock-out put option problem

3
(2.1) with finite horizon, where � > K , reduces to solve the following optimal stopping problem

(2.6) V (t, x) = sup E t,x [e−rτ G(Xt+τ



)]
0≤τ ≤T −t

where the supremum is taken over all stopping times τ of X � with values in [0, T − t] and
the expectation Et,x is taken with respect to the (unique) equivalent martingale measure Pt,x
under which Xt = x < � . The gain (payoff) function G is given by

G(x) = (K − x)+

Using the same arguments as in [17, Section 25.2] applied to the process X � we may infer from
general theory of optimal stopping for Markov processes that the continuation set equals

(2.7) C = {(t, x) ∈ [0, T ) × (0, ∞) | V (t, x) > G(x)} ,

and the stopping set equals

(2.8) D = {(t, x) ∈ [0, T ] × (0, ∞) | V (t, x) = G(x)} .

From [17, Corollary 2.9] we conclude that the optimal stopping time in (2.6) given by
� �

(2.9) τD = inf s ∈ [0, T − t] | (t + s, Xt+s )∈D

is optimal in our problem. It follows that there exists a function b : [0, T ] → R (see discussion
in [17, p. 380]) such that the continuation set equals

(2.10) C = {(t, x) ∈ [0, T ) × (0, ∞) | x > b(t)}

and the stopping set D̄ is the closure of the set

(2.11) D = {(t, x) ∈ [0, T ] × (0, ∞) | x < b(t)} ,

so that the stopping time


� �

(2.12) τb = inf s ∈ [0, T − t] : Xt+s ≤ b(t + s)

is optimal in (2.6). Since the gain function G(x) = (K − x)+ is a function of space only (i.e.
does not depend on time), it follows (see [14] and [17]) that the map t �→ V (t, x) is decreasing
on [0, T ] for each x ∈ (0, ∞) and that the boundary t �→ b(t) is increasing on [0, T ] .
Recall from [14] that x �→ V (t, x) is convex on (0, ∞) and (t, x) �→ V (t, x) is continuous
on [0, T ] × (0, ∞) and the boundary b satisfies the properties that b : [0, T ] → (0, K] is
continuous and increasing with b(T ) = K .
3. Standard arguments based on the strong Markov property lead to the following free-
boundary problem for the unknown value function V and the unknown boundary b :

(2.13) Vt + LX V = rV in C

(2.14) V (t, x) = (K − x)+ for x = b(t)

4
Figure 2: The value/payoff functions of the barrier/American put options plotted
against the stock price. We can see that the value function of the American put lies
closely above that of the American knock-out put in the continuation set, i.e. the
American knock-out is cheaper for x > bA (t) .

(2.15) Vx (t, x) = −1 for x = b(t) (smooth fit)


(2.16) V (t, x) > (K − x)+ in C
(2.17) V (t, x) = (K − x)+ in D
(2.18) V (t, x) = 0 for x ∈ [�, ∞).
The relations (2.13)-(2.17) are independently verified in [17] in the case of the American put
option; the results can be translated fully into the present setting. A proof of the smooth fit
condition (2.15) will be shown in the next section. Similarly, the following properties of V and
b also hold:

(2.19) V is continuous on [0, T ] × R+

(2.20) V is C 1,2 in C
(2.21) x �→ V (t, x) is decreasing and convex, with Vx (t, x) ∈ [−1, 0]
(2.22) t �→ V (t, x) is decreasing with V (T, x) = (K − x)+

5
(2.23) t �→ b(t) is increasing and continuous, with 0 < b(0+) < K and b(T −) = K.

4. To gain a better understanding of our problem, we apply Itô’s formula to e−rs V (t +



s, Xt+s ) and take the Pt,x expectation. The martingale term then vanishes by the optional
sampling theorem so that by (2.13), (2.14) and (2.17), and setting s = T − t , we get:
� �
(2.24) V (t, x) = e−r(T −t) E t,x G(XT� )
� T −t
� � � ��
− e−ru E t,x H(t + u, Xt+u

) I Xt+u ≤ b(t + u) du
0

for all (t, x) ∈ [0, T ] × R+ where we set G(x) = (K − x)+ and H = Gt + LX G − rG so that
H = −rK for x < b(t) . We obtain the following early exercise premium representation of the
value function:
� T −t
−r(T −t)
� � � � �
(2.25) V (t, x) = e �
E t,x G(XT ) + rK e−ru Pt,x Xt+u ≤ b(t + u) du.
0

Inserting x = b(t) , we obtain the following free-boundary equation


� T −t
−r(T −t)
� � �
(2.26) K − b(t) = e E t,b(t) (K − XT� )+ + rK e−ru Pt,b(t) Xt+u ≤ b(t + u) du
0

for all t ∈ [0, T ] . The first term on the RHS is the well known arbitrage-free price of the
European knock-out put option VE at the point (t, b(t)) , and can be written explicitly (see
e.g. [7]) as
� � � � ��
−r(T −t) 1 K σ2
(2.27) VE (t, x) = Ke Φ √ log − (r − )(T − t)
σ T −t x 2
� � � � 2
��
1 K σ
−xΦ √ log − (r + )(T − t)
σ T −t x 2
� �( 2r2 −1) � � � � ��
−r(T −t) � σ 1 xK σ2
− Ke Φ √ log − (r − )(T − t)
x σ T −t �2 2
� �( 2r2 +1) � � � � ��
� σ 1 xK σ2
+x Φ √ log − (r + )(T − t)
x σ T −t �2 2

for t ∈ [0, T ) and x ∈ (0, �) . Writing


� �
� �

(2.28) P Xt+u ≤ b(t + u) = P Xt+u ≤ b(t + u), max Xt+s < �
0≤s≤u

and using (2.3) more explicitly, while recalling (see e.g. [9]) that the joint density function of a
drifted Brownian motion Wuλ and its maximum Suλ under P is given by

2 (2s − w) − (2s−w)2 +λ(w− λu )
(2.29) f (u, w, s) = e 2u 2
π u3/2

6
for all u > 0 , s ≥ 0 and w ≤ s , we obtain the expression
(2.30) K − b(t) = e−r(T −t) E t,b(t) (K − XT� )+
� T � � � � ��
−r(v−t) 1 b(v) σ2
+ rK e Φ √ log − r− (v − t) dv
t σ v−t b(t) 2
� T � � 2r � � � � ��
−r(v−t) � σ2 −1 1 b(v)b(t) σ2
− rK e Φ √ log − r − (v − t) dv
t b(t) σ v−t �2 2
where
� x
1 z2
(2.31) Φ(x) = √ e− 2 dz
2π −∞

for x ∈ R denotes the standard normal distribution function. This is a nonlinear Volterra
integral equation of the second kind.
5. We turn to the American up-and-out put with a reverse barrier, i.e. the case where
� < K . The payoff here is discontinuous since the option is knocked out while it is in the
money. The arbitrage-free price is thus given by

(2.32) V (t, x) = sup E t,x [e−rτ (K − Xt+τ



)+ I (Xt+τ

< �)]
0≤τ ≤T −t

This is the same as writing

(2.33) V (t, x) = E t,x [e−rτ∗ (K − Xt+τ




)+ I (Xt+τ


< �) [I (τ∗ < τ� ) + I (τ∗ ≥ τ� )]]

where τ∗ is an optimal stopping time. If τ∗ < τ� , then we have

V (t, x) = E t,x [e−rτ∗ (K − Xt+τ∗ )+ I (Xt+τ




< �)]
σ2
= E t,x [(e−rτ∗ K − xeσWτ∗ − τ
2 ∗

) I (Xt+τ ∗
< �)]
= (KEe−rτ∗ − x) I (Xt+τ


< �)

≤ (K − x) I (Xt+τ ∗
< �) ≤ K − x
where the third equality follows from the optional sampling theorem and using the fact that
2
x exp(σWs − σ2 s) is a martingale under the measure Pt,x . On the other hand, if τ∗ ≥ τ� ,
then the option would have already been knocked-out and V (t, x) = 0 . We conclude from this
brief analysis that one can make profit by exercising at any time and thus it is always optimal
to stop immediately in the case of a reverse up-and-out put option with no rebates.

3. The result and proof


The approach used in the proof below is analogous to Peskir’s work on the American put
option (see [14] and [17]). The main difference is the fact that this problem is based on the
stopped process X � whereas the American put option looks only at the original process X .
Below, we will make use of the following function:
� � � � ��
1 z σ2
(3.1) J (t, x, v, z) = Φ √ log − r − (v − t)
σ v−t x 2

7
� � 2r2 −1 � � � � ��
� σ 1 z∧x σ2
− Φ √ log 2 − r − (v − t)
x σ v−t � 2

for t ∈ [0, T ), x > 0, z > 0 and v ∈ (t, T ) .


Theorem 1. The optimal stopping boundary can be characterized as the unique solution to
the nonlinear Volterra integral equation (2.30) in the class of continuous increasing functions
b : [0, T ] �→ R+ satisfying 0 < b(t) < K for all 0 < t < T . The arbitrage-free price of an up
and out American put option admits the following early-exercise premium representation:
� T −t
−r(T −t)
(3.2) V (t, x) = e �
E t,x G(XT ) + rK e−r(v−t) J(t, x, v, b(v)) dv
0

for all (t, x) ∈ [0, T ] × (0, �) , where the first term is the arbitrage-free price of the European
up-and-out put and the second term is the early-exercise premium.
Proof. We first show that the local time-space formula is applicable to our problem. We
then show that (3.2) cannot have other (continuous) solutions.
1. Let V : [0, T ] × (0, ∞) → R and b : [0, T ] → R+ denote the (unique) solution to our
free-boundary problem. Set

(3.3) C1 = {(t, x) ∈ [0, T ) × (0, ∞) | x > b(t)}

and

(3.4) C2 = {(t, x) ∈ [0, T ) × (0, ∞) | x < b(t)} .

Let
σ2 2
(3.5) LX V (t, x) = rxVx (t, x) + x Vxx (t, x)
2
for (t, x) ∈ C1 ∪ C2 . Then it can be shown (cf. [14] and [17]) that V and b are continuous
functions satisfying the following conditions: (i) V is C 1,2 on C1 ∪ C2 , (ii) b is of bounded
variation, (iii) P(Xs = b(s)) = 0 for all s ∈ (0, t] , (iv) Vt + LX V − rV is locally bounded on
C1 ∪ C2 , (v) x �→ V (t, x) is convex on (0, ∞) for every t ∈ [0, T ] and (vi) t �→ Vx (t, b(t)±)
is continuous on [0, T ] . From these conditions, we can see that the local time-space formula
in [13] is applicable to (s, y) �→ e−rs V (t + s, xy) with t ∈ [0, t), x > 0 given and fixed. This
yields

(3.6) e−rs V (t + s, xXs� ) = V (t, x) + Msb


� s
+ e−ru (Vt + LX V − rV )(t + u, xXu� ) I (xXu� �= b(t + u)) du
0

1 s −ru
+ e (Vx (t + u, b(t + u)+) − Vx (t + u, b(t + u)−)) dLbu (X � )
2 0
where
� s
(3.7) Msb =σ e−ru xXu� Vx (t + u, xXu� ) I (xXu� �= b(t + u)) dWu
0

8
is a continuous martingale for s ∈ [0, T − t] and

(3.8) Lb (X � ) = Lbu (X � )0≤u≤s

is the local time of X � = (xXu� )0≤u≤s on the curve b for s ∈ [0, T − t] . To show that
M b = (Msb )0≤s≤T −t is a martingale for t ∈ [0, T ) note that X � ≤ � and by (2.21) we get
� s
� b �
(3.9) b
E M ,M s = σ x E 2 2
e−2ru (Xu� )2 Vx2 (t + u, xXu� ) I (xXu� �= b(t + u)) du
�0 �
−1
≤ σ 2 x2 � 2 (e−rs + 1) < ∞
r
from where it follows that M b is a continuous martingale as claimed. Applying the optional
sampling theorem to (3.6), we get
� T −t
−r(T −t)
� � �
(3.10) V (t, x) = e �
E t,x G(XT ) + rK e−ru Pt,x Xt+u ≤ b(t + u) du
0

1 T −t −ru
− e (Vx (t + u, b(t + u)+) − Vx (t + u, b(t + u)−)) du E t,x (Lbu (X � ))
2 0
We can see that the last term on the right-hand side vanishes if we can prove that x �→ V (t, x)
is C 1 at b(t) , i.e. that the smooth-fit condition (2.15) holds. For this, let us fix a point
(t, x) ∈ (0, T ) × (0, ∞) lying on the boundary b so that x = b(t) . Then x < K and for all
ε > 0 such that x + ε < K we have V (t, x + ε) > G(x + ε) since V (t, x) = G(x) . We make
use of the mean value theorem:
V (t, x + ε) − V (t, x) G(x + �) − G(x)
(3.11) ≥ .
ε ε
Taking the limit as ε ↓ 0 , we get
∂ +V
(3.12) (t, x) ≥ G� (x) = 1,
∂x
where the left-hand derivative exists by convexity of x �→ V (t, x) on (0, ∞) . The converse
inequality can be achieved by considering the stopping time τε = τ∗ (t, x + ε) being optimal for
V (t, x + ε) . Denoting
σ2
(3.13) Yt = eσWt +(r− 2
)t
,
we have
V (t, x + ε) − V (t, x) 1 � � −rτε � � ��
(3.14) ≤ E e (K − (x + ε)Yτε )+ − E e−rτε (K − xYτε )+
ε ε
1 � � −rτε � � ��
≤ E e (K − (x + ε)Yτε )+ − (K − xYτε )+ I ((x + ε)Yτε < K)
ε � �
= −E e−rτε Yτε I ((x + ε) Yτε < K) .

If we can show that τε → 0 as ε ↓ 0 then we will have


� �
(3.15) E e−rτε Yτε I ((x + ε) Yτε < K) → 1 as ε ↓ 0.

9
For this, consider the stopping time

(3.16) ρ = inf {s ∈ [0, T − t] : Xt+s ≥ x}

under the measure Pt,x+ε . The process X started at x + ε at time t will always hit the
boundary b before it hits the level x since s �→ b(s) is increasing on [t, T ] . Hence, 0 ≤ τε ≤ ρ
and thus it is enough to show that ρ → 0 under Pt,x+ε as ε ↓ 0 . Note that

(3.17) ρ ≤ inf {s ∈ [0, T − t] : (x + ε)Xs ≥ x}


� � � �
1 σ2
= inf s ∈ [0, T − t] : Ws ≥ log (x/x + ε) − r − /σ s ,
σ 2
letting ε ↓ 0 and using the fact from Blumenthal’s 0-1 law that
� 2

r − σ2
(3.18) s �→ − s
σ
is a lower function of B at zero, we see that ρ → 0 under Pt,x+ε and hence τε → 0 as
claimed. Taking the limit in (3.14) as ε ↓ 0 , by the dominated convergence theorem we
conclude that x �→ V (t, x) is differentiable at b and Vx (t, b(t)) = −1 . Using this result, we
can see that (3.10) yields the representation given in the theorem and since V (t, b(t)) = G(b(t))
for all t ∈ [0, T ] , we establish that b solves our equation. We now turn to uniqueness.
2. We show that the optimal stopping boundary is the unique solution to (3.2) in the class
of continuous functions 0 ≤ b(t) ≤ K for all t ∈ [0, T ] . Motivated by the representation
(2.24) above, define the function U a : [0, T ] × (0, ∞) → R by setting

(3.19) U a (t, x) = e−r(T −t) E t,x G(XT� )


� T −t
� � � ��
− e−ru E t,x H(t + u, Xt+u

) I Xt+u ≤ a(t + u) du
0

for (t, x) ∈ [0, T ] × (0, ∞) and H = Gt + LX G − rG so that H = −rK for x < a(t) .
Observe that a solving (3.2) means exactly that U a (t, a(t)) = G(a(t)) for all t ∈ [0, T ] .
(i) We first show that U a (t, x) = G(x) for all (t, x) ∈ [0, T ] × (0, ∞) such that x ≤ a(t) .
Take any such (t, x) and observe that the Markov property of X � implies that
� s
−rs a
(3.20) �
e U (t + s, Xt+s ) + rK e−ru I (Xt+u

≤ a(t + u)) du
0

is a continuous martingale under the measure Pt,x . Given that we are only looking at the
case of x ≤ a(t) for this part of the proof, we can work with the process X instead of
X � since a(t) lies below � by definition of the option. Consider the stopping time σa =
inf {s ∈ [0, T − t] : Xt+s ≥ a(t + s)} under the measure Pt,x+ε . Firstly note that the local
time-space formula applied to (s, y) �→ e−rs G(xy) yields:
� s �
−rs −ru K 1 s −ru K
(3.21) e G(Xt+s ) = G(x) − rK e I (Xt+u < K) du + Ms + e dLu (X)
0 2 0

10
�s
upon using that MsK = − 0 e−ru σXt+u I (Xt+u < K) dWu is a martingale and that LX G−rG
equals −rK on (0, K) and 0 on (K, ∞) . Replacing s by σa , taking E on both sides and
applying the optional sampling theorem, we have:
� σa
−rσa
(3.22) Ee G(Xt+σa ) = G(x) − rK Ee−ru I (Xt+u < K) du.
0

Doing the same to (3.20) and using the above result, we find that
� σa
� −rσa a �
a
U (t, x) = E e U (t + σa , Xt+σa ) + rKEe−ru I (Xt+u ≤ a(t + u)) du
0
� σa
� −rσa �
=E e G(Xt+σa ) + rKEe−ru I (Xt+u ≤ a(t + u)) du
0
� σa
= G(x) − rKEe−ru I (Xt+u < K) du
� 0σa
+ rKEe−ru I (Xt+u ≤ a(t + u)) du
0
= G(x),

since Xt+u < K and Xt+u ≤ a(t + u) is true for all u ∈ [0, σa ) .
(ii) We show that U a (t, x) ≤ V (t, x) for all (t, x) ∈ [0, T ] × (0, ∞) . For this, take any
such (t, x) and consider the stopping time τa = inf {s ∈ [0, T − t] : Xt+s ≤ a(t + s)} . One
can verify that U a satisfies the conditions (i)-(vi) from 1. above, we can therefore write
� s
(3.23) −rs a � a
e U (t + s, Xt+s ) = U (t, x) − rKe−ru I (Xt+u

≤ a(t + u))du + M �sa
0

where M�a is a martingale. Replacing s by τa , taking E on both sides and applying the
s
optional sampling theorem, we get
� �
(3.24) U a (t, x) = E e−rτa U a (t + τa , Xt+τ�
a
)
� �
= E e−rτa G(Xt+τ �
a
)
� �
= E e−rτa (K − Xt+τ �
a
)+
≤ V (t, x)

for all (t, x) ∈ [0, T ) × (0, ∞) .


(iii) Let us now show that a(t) ≥ b(t) on [0, T ] . For this, recall that by the same arguments
as for U a above, we also have
� s
(3.25) −rs �
e V (t + s, Xt+s ) = V (t, x) − rKe−ru I (Xt+u

≤ b(t + u)) du + M �b ,
s
0

where (M �b )0≤s≤T −t is a martingale under Pt,x . Fix (t, x) ∈ (0, T ) × (0, ∞) such that x <
s
a(t) ∧ b(t) and consider the stopping time
� �

(3.26) σb = inf s ∈ [0, T − t] : Xt+s ≥ b(t + s) .

11
Figure 3: A computer drawing showing how the optimal stopping boundary
changes as we vary the barrier level. We see that the introduction of a barrier
raises the optimal boundary.

Inserting σb in place of s in (3.23) and (3.25) and taking Pt,x -expectation we get
� σb
� −rσ a �
(3.27) E e b �
U (t + σb , Xt+σb ) = G(x) − rKEe−ru I (Xt+u

≤ a(t + u)) du
0
� σb
� −rσ �
(3.28) E e b �
V (t + σb , Xt+σb ) = G(x) − rKEe−ru I (Xt+u

≤ b(t + u)) du
0

Hence, by part (ii) we see that


� σb � σb
−ru
(3.29) �
Ee I (Xt+u ≤ a(t + u)) du ≥ Ee−ru I (Xt+u

≤ b(t + u)) du
0 0

from where it follows from the continuity of a and b that a(t) ≥ b(t) for all t ∈ [0, T ] .
(iv) Finally, let us show that a must be equal to b . For this, assume that there is t ∈ (0, T )
such that a(t) > b(t) , and pick x ∈ (b(t), a(t)) . Inserting the stopping time τb in (2.12) in
place of s in (3.23) and (3.25) and taking Pt,x -expectation we get:

� τb
� −rτb

(3.30) E e �
G(Xt+τ b
) a
= U (t, x) − E rKe−ru I (Xt+u

≤ a(t + u)) du
0

12
Time (months) 0 2 4 6 8 10 12
Exercise at 8 (Knock-out � = 11) 343% 343% 343% 343% 343% 343% 343%
Exercise at 8 (Knock-out � = 13) 189% 189% 189% 189% 189% 189% 189%
Exercise at 8 (Knock-out � = 15) 171% 171% 171% 171% 171% 171% 171%
Exercise at 8 (American put) 167% 167% 167% 167% 167% 167% 167%
Exercise at b11 (Knock-out � = 11) 453% 452% 448% 440% 420% 366% 0%
Exercise at b13 (Knock-out � = 13) 309% 301% 290% 273% 248% 205% 0%
Exercise at b15 (Knock-out � = 15) 286% 276% 264% 247% 224% 186% 0%
Exercise at bA (American put) 281% 271% 258% 242% 219% 182% 0%
Exercise at 6 (Knock-out � = 11) 687% 687% 687% 687% 687% 687% 687%
Exercise at 6 (Knock-out � = 13) 378% 378% 378% 378% 378% 378% 378%
Exercise at 6 (Knock-out � = 15) 342% 342% 342% 342% 342% 342% 342%
Exercise at 6 (American put) 335% 335% 335% 335% 335% 335% 335%
Exercise at 4 (Knock-out � = 11) 1030% 1030% 1030% 1030% 1030% 1030% 1030%
Exercise at 4 (Knock-out � = 13) 568% 568% 568% 568% 568% 568% 568%
Exercise at 4 (Knock-out � = 15) 513% 513% 513% 513% 513% 513% 513%
Exercise at 4 (American put) 502% 502% 502% 502% 502% 502% 502%
Exercise at 2 (Knock-out � = 11) 1373% 1373% 1373% 1373% 1373% 1373% 1373%
Exercise at 2 (Knock-out � = 13) 757% 757% 757% 757% 757% 757% 757%
Exercise at 2 (Knock-out � = 15) 683% 683% 683% 683% 683% 683% 683%
Exercise at 2 (American put) 669% 669% 669% 669% 669% 669% 669%

Table 1: Returns observed upon exercising the American knock-out put option below the
strike price K with barrier levels � = 11, 13, 15 . The returns are expressed as a percentage of
the original option price paid by the buyer (rounded to the nearest integer), i.e. R(t, x)/100 =
(K − x)+ /V (0, K) . The parameter set is K = 10 , T = 1 , r = 0.1 , σ = 0.4 and the
initial stock price equals K .

� �
(3.31) E e−rτb G(Xt+τ

b
) = V (t, x).

Hence, by part (ii) we see that


� τb
(3.32) E rKe−ru I (Xt+u

≤ a(t + u)) du ≤ 0,
0

from which it follows from the continuity of a and b that such a point x cannot exist. Thus
a(t) = b(t) for all t ∈ [0, T ] and the proof is complete. �

4. Financial analysis of the American knock-out put option


In this section, we present a numerical example to draw a comparison between the American
knock-out put option and its European counterpart with the American put option. Using an
approach similar to the one used in [15], we mainly address the question as to what the return
would be if the underlying process enters the given region at a given time. Such an analysis
remains skeletal since we do not discuss the probability of the latter event nor do we account
for any risk associated with its occurrence. However, we are able to highlight the features and
drawbacks of the option irrespective of whether the stock price model is correct or not.

13
Time (months) 0 2 4 6 8 10 12
Sell at 12 (American put) 53% 46% 39% 30% 20% 8% 0%
Sell at 12 (American knock-out) 29% 27% 25% 22% 17% 8% 0%
Sell at 12 (European knock-out) 29% 28% 26% 23% 18% 9% 0%
Sell at 11 (American put) 73% 66% 58% 49% 37% 21% 0%
Sell at 11 (American knock-out) 61% 59% 55% 49% 40% 24% 0%
Sell at 11 (European knock-out) 62% 60% 57% 52% 43% 26% 0%
Sell at K (American put) 100% 94% 86% 77% 65% 49% 0%
Sell at K (American knock-out) 100% 97% 92% 84% 73% 55% 0%
Sell at K (European knockout) 100% 98% 95% 89% 78% 60% 0%
Sell at 9 (American put) 137% 132% 125% 118% 109% 97% 84%
Sell at 9 (American knock-out) 148% 144% 139% 133% 123% 110% 95%
Sell at 9 (European knock-out) 146% 145% 142% 138% 130% 118% 105%

Table 2: Returns observed upon selling the American knock-out, European knock-out
and American put option at/above the strike price K . The returns are expressed as
a percentage of the original option price paid by the buyer (rounded to the nearest
integer), i.e. Rs (t, x)/100 = V (t, x)/V (0, K) , RE (t, x)/100 = VE (t, x)/VE (0, K) and
RA (t, x)/100 = (K − x)+ /VA (0, K) respectively. The parameter set is the same as in
Table 1 above ( K = 10 , T = 1 , r = 0.1 , σ = 0.4 , � = 13 ) and the initial stock price
equals K .

1. Figure 3 illustrates how the optimal boundary of the American knock-out put option
changes as one varies the barrier level � . We can see that the introduction of a barrier raises
the optimal boundary, this is made more apparent as the barrier is lowered further towards the
strike. Closer to the end of the lifetime of the option, we note that the optimal boundary for
all barrier levels is almost identical; the reason behind this could be the decreasing likeliness
of the stock price hitting the barrier before maturity, given that it has not hit it well into
into the contract. Hence, one can say that the optimal boundary is not much affected by
the presence of a barrier as the time draws nearer to maturity. In addition, when the initial
point is fixed at K = 10 , we see that if we choose a barrier level of 15 , while keeping the
remaining parameters fixed, the American knock-out put option behaves comparably to the
American put option. This can also been seen in Table 1 where the returns of the American
put in comparison with the American knock-out show a difference of a maximum of 11% .
This is logical since it becomes highly unlikely for the stock price to hit the barrier. In fact, as
the barrier rises to infinity, the American knock-out put option converges to the American put
option. In Table 1, we can see that the returns observed on exercising below the strike for a
lower barrier are more than double those of a higher barrier. This is obtained, however, at a
greater risk since the option is more likely to be knocked-out. We note in addition that setting
V = V�,K and G = G�,K to indicate dependence on the strike price K and barrier level � ,
we have V�,K (t, x) = KV � ,1 (t, x/K) and G�,K (t, x) = KG � ,1 (t, x/K) , so that the return does
K K
not depend on the size of the strike price (when properly scaled). The same fact is valid for the
returns of the European knock-out and the American put options considered in this section.
2. Table 2 shows the returns an investor can extract upon selling the American and European
knock-out options and the American put options (at the arbitrage-free price) at and above the

14
strike price K . In practice, the option holder may choose to sell his option at any time during
the term of the contract, and thus one may view his payoff as the price he receives upon selling.
We choose � = 13 since this practically illustrates the influence of the presence of a barrier on
the returns (not too close to K allowing us to analyse at a greater range of prices; and not too
far from K so that we are not left with the case of the American put option). Viewing the
percentage returns as a measure of option performance we observe the following in the region
directly below K (i.e. when option is in the money). The American put naturally produces
lower returns since the investor would have paid a higher premium, while selling the European
knock-out option produces better returns, then followed by the American knock-out option.
On the other hand, in the region above K , selling the American put option produces returns
higher than the American and European knock-out options. This essentially reflects the option
holder’s uneasiness as the stock price approaches the barrier. In reality, the price at which
the option holder is able to sell will depend upon a number of exogenous factors such as his
inability to access the option market due to the friction costs in the form of transaction costs
and/or taxes involved in selling, and the liquidity of the option market itself, and thus it may be
increasingly difficult for the holder to sell the option. We seek to address the issue of liquidity
in the context of barrier options in forthcoming work based on the economic rationale of the
‘British’ option introduced by Peskir and Samee in [15] and [16].

References
[1] Carr, P. (1995). Two extensions to barrier option valuation. Applied Mathematical Fi-
nance 2 (173–209).

[2] Chang, G., Kang, J., Kim, H. S. and Kim, I. J. (2007). An efficient approximation
method for American exotic options. The Journal of Futures Markets 27 (29–59).

[3] Derman, E. and Kani, I. (1996). The ins and outs of barrier options: Part 1. Derivatives
Quarterly Winter (55–67).

[4] Derman, E. and Kani, I. (1997). The ins and outs of barrier options: Part 2. Derivatives
Quarterly Spring (73–80).

[5] Gao, B., Huang, J. and Subrahmanyam, M. (2000). The valuation of American
barrier options using the decomposition technique. Journal of Economic Dynamics and
Control 24 (1783–1827).

[6] Haug, E. G. (1997). The Complete Guide to Option Pricing Formulas. McGraw-Hill,
New York.

[7] Joshi, M. S. (2003). The Concepts and Practice of Mathematical Finance. Cambridge
University Press.

[8] Karatzas, I. and Wang, H. (2000). A barrier option of American type. Applied Math-
ematics and Optimization 42 (259–279).

[9] Karatzas, I. and Shreve, S. E. (1998). Methods of Mathematical Finance. Springer.

15
[10] Lundgren, R. (2007). Structure of optimal stopping domains for American options with
knock out domains. Theory of Stochastic Processes 13 (98–129).

[11] Merton, R. (1973). Theory of rational option pricing. The Bell Journal of Economics
and Management Science 4 (141–183).

[12] Mijatović, A. (2010). Local time and the pricing of time-dependent barrier options.
Finance and Stochastics 14 (13–48).

[13] Peskir, G. (2005). A change-of-variable formula with local time on curves. Journal of
Theoretical Probability 18 (499–535).

[14] Peskir, G. (2005). On the American option problem. Mathematical Finance 15 (169–181).

[15] Peskir, G. and Samee, F. (2011). The British put option. Applied Mathematical Finance
18 (537–563).

[16] Peskir, G. and Samee, F. (2013) The British call option. Quantitative Finance 13
(95–109).

[17] Peskir, G. and Shiryaev, A. N. (2006). Optimal Stopping and Free-Boundary Problems.
Lectures in Mathematics, ETH Zürich. Birkhäuser.

[18] Reiner E. and Rubinstein M. (1991). Breaking down the barriers. Risk 4 (28–35).

[19] Rich, D. R. (1994). The mathematical foundation of barrier option-pricing theory. Ad-
vances in Futures and Options Research 7 (267–312).

Luluwah Al-Fagih
School of Mathematics
Kingston University London
Kingston upon Thames, KT1 2EE
United Kingdom
l.al-fagih@kingston.ac.uk

16

Potrebbero piacerti anche