THE AMERICAN STRADDLE CLOSE TO EXPIRY

GHADA ALOBAIDI AND ROLAND MALLIER

Received 23 August 2005; Revised 26 December 2005; Accepted 22 March 2006
We address the pricing of American straddle options. We use a technique due to Kim
(1990) to derive an expression involving integrals for the price of such an option close
to expiry. We then evaluate this expression on the dual optimal exercise boundaries to
obtain a set of integral equations for the location of these exercise boundaries, and solve
these equations close to expiry.
Copyright 2006 G. Alobaidi and R. Mallier. This is an open access article distributed
under the Creative Commons Attribution License, which permits unrestricted use, dis-
tribution, and reproduction in any medium, provided the original work is properly cited.
1. Introduction
One of the classic problems of mathematical nance is the pricing of American options
and the behavior of the optimal exercise boundary close to expiry. For the uninitiated,
nancial derivatives are securities whose value is based on the value of some other under-
lying security, and options are an example of derivatives, carrying the right but not the
obligation to enter into a specied transaction in the underlying security. A call option
allows the holder to buy the underlying security at a specied strike price E, a put option
allows the holder to sell the underlying at the price E, while a straddle, which we consider
in the current study, allows the holder the choice of either buying or selling (but not both)
the security. If S is the price of the underlying, then the payo at expiry is max(S E, 0)
for a call, max(E S, 0) for a put, and max(S E, E S) for a straddle. From this payo, it
would appear at rst glance that the holder of a straddle is holding a call and a put on the
same underlying with the same strike and the same expiry, but is only allowed to exercise
one. This is true for a European straddle, which can be exercised at expiry, because if the
call is in the money, the put must be out of the money and vice versa, and the holder will
naturally exercise whichever of the call and the put is in the money at expiry, unless he is
unlucky enough that S = E so that they are both exactly at the money, and the payo is
zero.
Hindawi Publishing Corporation
Boundary Value Problems
Volume 2006, Article ID 32835, Pages 114
DOI 10.1155/BVP/2006/32835
2 The American straddle close to expiry
However, an American straddle is not simply the combination of an American call
and put. Unlike Europeans, which can be exercised only at expiry, American options may
be exercised at any time at or before expiry. The payo from immediate exercise is the
same as the payo at expiry, namely, max(S E, 0) for a call, max(E S, 0) for a put, and
max(S E, E S) for a straddle. Naturally, a rational investor will choose to exercise early
if that maximizes his return, and it follows that there will be regions where it is optimal to
hold the option, and others where exercise is optimal, with a free boundary known as the
optimal exercise boundary separating these regions. It is because the free boundaries for
the straddle dier from those for the call and the put so that an American straddle diers
from the combination of an American call and put.
For vanilla Americans, there have been numerous studies of this free boundary, but
a closed form solution for its location remains elusive, as does a closed form expression
for the value of an American option. One popular approach [11, 16, 20] has been used
to reformulate the problem as an integral equation for the location of the free boundary,
which can be solved using either asymptotics or numerics, although for problems with
a single free boundary, such as vanilla American calls and puts, it may be simpler just
to apply asymptotics directly to the Black-Scholes-Merton partial dierential equation
using the methods developed by Tao [23] for Stefan problems, and this has been done for
the American call and put [1, 8, 18]. For the American straddle, which we have previously
studied using partial Laplace transforms [2], there are two free boundaries not one: an
upper one on which an exercise as a call occurs, and a lower one on which an exercise as a
put occurs, and applying Taos method is somewhat harder, making the integral equation
approach more attractive.
We should also mention the work of Kholodnyi [14, 15] on American-style options
with general payos of which the American straddle considered here is one particular
example. In [14], a new formulation in terms of the semilinear evolution equation in the
entire domain of the state variables was introduced, while in [15], the foreign exchange
option symmetry was introduced.
In the present study, we will use an approach originally developed for physical Stefan
problems [17] and later applied to economics [20], and applied to vanilla Americans with
great success by Kim [16] and Jacka [11], who independently derived the same results:
Kim both by using McKeans formula and by taking the continuous limit of the Geske-
Johnson formula [9] which is a discrete approximation for American options, and thereby
demonstrating that those two approaches led to the same result, and Jacka by applying
probability theory to the optimal stopping problem. P. Carr et al. [6] later used these
results to show how to decompose the value of an American option into intrinsic value
and time value. The approach in [11, 16, 20] leads to an integral equation for the location
of the free boundary, which was solved numerically by [10] and by approximating the
free boundary as a multipiece exponential function by [13].
Our contribution is to extend the analysis of [11, 16, 20] to the American straddle and
obtain a set of integral equations for the locations of the free boundaries. These equations
are then solved asymptotically to nd the locations of the free boundaries close to expiry,
and the results are compared to the results for American calls and puts. Before we proceed
with our analysis, we note that [24] proved that the free boundary was regular for vanilla
Americans, and his analysis can be carried over to the straddle.
G. Alobaidi and R. Mallier 3
2. Analysis
In this section, we follow the approach taken in [16]. Our starting point is the Black-
Scholes-Merton partial dierential equation [5, 21] governing the price V of an equity
derivative,
V =
_

2
S
2
2

2
S
2
(r D)S

S
+r
_
V =0, (2.1)
where S is the price of the underlying stock and = T t is the time remaining until
expiry. In our analysis, the volatility , risk-free interest rate r, and dividend yield D are
assumed constant. For European options, the value of the option can be written as
V
E
(S, ) =
_

0
V
E
(Z, 0)G(S, Z, )dZ, (2.2)
where V
E
(S, 0) is the payo at expiry, and we have introduced Greens function,
G(S, Z, ) =
e
r
Z

2
exp
_

_
ln(S/Z) +r
()

_
2
2
2

_
, (2.3)
with r
()
= r D
2
/2 and r
(+)
= r D +
2
/2. Using (2.2), the price of a European
straddle with strike E and payo at expiry V
E
(S, 0) =max(S E, E S) is
V
E
(S, ) =Se
D
erf
_
ln(S/E) +r
(+)

2
_
Ee
r
erf
_
ln(S/E) +r
()

2
_
. (2.4)
The value of a European straddle is simply the sum of the values of a European call and
put. In the above, erf is the error function, with erfc the complementary error function.
To derive our analytic expression for an American straddle, two paths may be taken.
We may follow[16] and approximate an American option by a Bermudan option with ex-
ercise opportunities at m for 0 mn, and then take the limit 0, so that we can
neglect certain terms, with n to recover the value of the American option. In this
procedure, we write the value of the Bermudan option at time m in terms of its value at
(m1), where the option will be exercised if its value falls below that from immediate
exercise. For the Bermudan option, the upper and lower optimal exercise boundaries at
time m, S =S
u
(m) and S =S
l
(m) will be the solutions of V
B
(S, m) =S E and
V
B
(S, m) =E S, respectively, and we will hold the option if S
l
(m) < S < S
u
(m)
but exercise as a call if S > S
u
(m) and as a put if S < S
l
(m). To leading order in ,
we nd
V
B
(S, n) =V
E
(S, n) +
n1
_
m=1

S
f
(m)
(DZ rE)G
_
S, Z, (nm)
_
dZ
+
n1
_
m=1

_
S
l
(m)
0
(rE DZ)G
_
S, Z, (nm)
_
dZ.
(2.5)
4 The American straddle close to expiry
To go to the continuous exercise American case, we follow [16] and take the limit 0
with n , using S = S
u
() and S
l
() to denote upper and lower optimal exercise
boundaries, respectively,
V
A
(S, ) =V
E
(S, ) +
_

0
_

S
u
()
(DZ rE)G(S, Z, )dZd
+
_

0
_
S
l
()
0
(rE DZ)G(S, Z, )dZd.
(2.6)
An alternative approach is to apply a more general formula. For American-style options
with early exercise features, it follows from the work of [4, 11, 12, 16, 17, 20] that if such
an option obeys (2.1), where it is optimal to hold the option and the payo at expiry is
V(S, 0) while that from immediate exercise is P(S, ), then we can write the value of the
option as the sum of the value of the corresponding European option V
(e)
(S, ) together
with another term representing both the premium from an early exercise, a technique
introduced for the American call and put by [4, 12],
V(S, ) =V
(e)
(S, ) +
_

0
_

0
(Z, )G(S, Z, )dZd, (2.7)
with (S, ) 0 where it is optimal to hold the option while where exercise is optimal
(S, ) is the result of substituting the early exercise payo P(S, ) into (2.1), (S, ) =
P. For the straddle, =DS rE when we exercise as a call, and =rE DS when we
exercise as a put.
Using either of these approaches, we nd that
V
A
(S, ) =Se
D
erf
_
ln(S/E) +r
(+)

2
_
Ee
r
erf
_
ln(S/E) +r
()

2
_
+
1
2
_

0
_
SDe
D()
_
erf
_
ln
_
S/S
u
()
_
+r
(+)
( )

_
2( )
_
+erf
_
ln
_
S/S
l
()
_
+r
(+)
( )

_
2( )
__
rEe
r()
_
erf
_
ln
_
S/S
u
()
_
+r
()
( )

_
2( )
_
+erf
_
ln
_
S/S
l
()
_
+r
()
( )

_
2( )
___
d,
(2.8)
which is an expression for the value of the American straddle. If we compare our results
to the expressions for the American call and put in [11, 16], the expressions involving
S
u
() appear in the expression for the call while those with S
l
() appear in that for the
put, and at rst glance, it looks as though the value of an American straddle is the sum of
the values of an American call and an American put, as was the case with the Europeans,
although of course this is not really the case as the free boundaries for the straddle will
dier from those for the call and the put.
G. Alobaidi and R. Mallier 5
3. Integral equations
The integral equations for the location of the upper and lower free boundaries S =S
u
()
and S = S
l
() are derived by substituting the expression for the American straddle (2.8)
into the conditions at the free boundaries, and requiring that the value must be contin-
uous across the boundaries, so that V
A
=S E at S =S
u
() and V
A
=E S at S =S
l
(),
together with the high contact or smooth pasting condition [22] that (V
A
/S) = 1 at
S
u
() and (V
A
/S) =1 at S
l
(). These four conditions will give us four integral equa-
tions, which as discussed in [16, 17] are Volterra equations of the second kind.
For the upper boundary, the condition that V
A
=S E at S =S
u
() yields
S
u
()
_
1 e
D
erf
_
ln
_
S
u
()/E
_
+r
(+)

2
__
E
_
1 e
r
erf
_
ln
_
S
u
()/E
_
+r
()

2
__
=
_

0
_
S
u
()De
D()
2
_
erf
_
ln
_
S
u
()/S
u
()
_
+r
(+)
( )

_
2( )
_
+erf
_
ln
_
S
u
()/S
l
()
_
+r
(+)
( )

_
2( )
__

rEe
r()
2
_
erf
_
ln
_
S
u
()/S
u
()
_
+r
()
( )

_
2( )
_
+erf
_
ln
_
S
u
()/S
l
()
_
+r
()
( )

_
2( )
___
d,
(3.1)
while for the lower boundary, the condition that V
A
=E S at S =S
l
() yields
S
l
()
_
1 +e
D
erf
_
ln
_
S
l
()/E
_
+r
(+)

2
__
+E
_
1 +e
r
erf
_
ln
_
S
l
()/E
_
+r
()

2
__
=
_

0
_
S
l
()De
D()
2
_
erf
_
ln
_
S
l
()/S
u
()
_
+r
(+)
( )

_
2( )
_
+erf
_
ln
_
S
l
()/S
l
()
_
+r
(+)
( )

_
2( )
__

rEe
r()
2
_
erf
_
ln
_
S
l
()/S
u
()
_
+r
()
( )

_
2( )
_
+erf
_
ln
_
S
l
()/S
l
()
_
+r
()
( )

_
2( )
___
d,
(3.2)
6 The American straddle close to expiry
while the condition (V
A
/S) =1 at S =S
u
() gives
1 e
D
_
erf
_
ln
_
S
u
()/E
_
+r
(+)

2
_

exp
_

_
ln
_
S
u
()/E
_
+r
(+)

_
2
2
2

__
+
E

2e
r
S
u
()

exp
_

__
S
u
()/E
_
+r
()

_
2
2
2

_
=
_

0
_
De
D()
2
_
erf
_
ln
_
S
u
()/S
u
()
_
+r
(+)
( )

_
2( )
_
+erf
_
ln
_
S
u
()/S
l
()
_
+r
(+)
( )

_
2( )
__
+
De
D()

_
2( )

_
exp
_

_
ln
_
S
u
()/S
u
()
_
+r
(+)
( )
_
2
2
2
( )
_
+exp
_

_
ln
_
S
u
()/S
l
()
_
+r
(+)
( )
_
2
2
2
( )
__

rEe
r()
S
u
()
_
2( )

_
exp
_

_
ln
_
S
u
()/S
u
()
_
+r
()
( )
_
2
2
2
( )
_
+exp
_

_
ln
_
S
u
()/S
l
()
_
+r
()
( )
_
2
2
2
( )
___
d,
(3.3)
and the condition (V
A
/S) =1 at S
l
() gives
1 e
D
_
erf
_
ln
_
S
l
()/E
_
+r
(+)

2
_

exp
_

_
ln
_
S
l
()/E
_
+r
(+)

_
2
2
2

__
+
E

2e
r
S
l
()

exp
_

_
ln
_
S
l
()/E
_
+r
()

_
2
2
2

_
=
_

0
_
De
D()
2
_
erf
_
ln
_
S
l
()/S
u
()
_
+r
(+)
( )

_
2( )
_
+erf
_
ln
_
S
l
()/S
l
()
_
+r
(+)
( )

_
2( )
__
+
De
D()

_
2( )
G. Alobaidi and R. Mallier 7

_
exp
_

_
ln
_
S
l
()/S
u
()
_
+r
(+)
( )
_
2
2
2
( )
_
+exp
_

_
ln
_
S
l
()/S
l
()
_
+r
(+)
( )
_
2
2
2
( )
__

rEe
r()
S
l
()
_
2( )

_
exp
_

_
ln
_
S
l
()/S
u
()
_
+r
()
( )
_
2
2
2
( )
_
+exp
_

_
ln
_
S
l
()/S
l
()
_
+r
()
( )
_
2
2
2
( )
__
d.
(3.4)
The expressions (3.1)(3.4), valid for 0, are the integral equations for the location of
the free boundaries for the straddle, and each involves the locations of both free bound-
aries, so that the boundaries are coupled together. If we compare our results to the cor-
responding expressions for the American call and put, we see that (3.1) appears at rst
glance to be the sum of the corresponding equations for the call and the put evaluated at
the upper boundary while (3.2) appears to be the same expression evaluated at the lower
boundary, with a similar relation between (3.3) and (3.4) and the equations coming from
the deltas of the call and the put. Once again, we would stress that since the free bound-
aries for the straddle dier from those for the call and the put, the relation between these
equations is not quite so straightforward.
4. Solution of the integral equations close to expiry
We will solve the above integral equations (3.1)(3.4) close to expiry to nd expressions
for the location of the free boundaries in the limit 0, writing S
u
() = S
u0
e
x
u
()
and
S
l
() = S
l0
e
x
l
()
, where S
u0
and S
l0
are the locations of the upper and lower free bound-
aries at expiry, which can be deduced by considering the behavior of (V
A
/) at expiry.
Initially, we will try a solution of the form
x
u
()

_
n=1
x
un

n/2
,
x
l
()

_
n=1
x
ln

n/2
,
(4.1)
which is motivated both by earlier work on American options and by the work of [23]
on Stefan problems in general. Since there will be several terms of the form ln(S
u0
/E) and
ln(S
l0
/E), we would expect the behavior when S
u0
= E to dier from that when S
u0
= E,
and similarly with S
l0
, and this suggests that we consider several cases separately.
Case 4.1 (D < r). The free boundary starts at S
l0
=E and S
u0
=rE/D. If we take the limit
0, we can drop certain terms and (3.1)(3.4) decouple, so that we have two pairs of
equations: one involving only x
u
but not x
l
, and a second pair involving only x
l
but not
x
u
. Because close to expiry, the value of the put-like element of the straddle is not felt
at S
u0
= rE/D, the pair of equations involving only x
u
is identical to the pair of integral
8 The American straddle close to expiry
equations for an American call with D < r. Recalling that we can decompose an American
option into European component and an early exercise component, the pair of equations
involving only x
l
is similar to but not quite identical to the pair of integral equations for
an American put with D < r in the limit 0. Because close to expiry, the value of the
early exercise component of the call-like element of the straddle (2.8) is not felt at S
l0
=E
but the European component most denitely is.
This decoupling only happens when we take the limit 0 and for larger times, the
two boundaries remain coupled together. The eect of this decoupling, however, is that
for the case D < r, in the limit 0, the upper free boundary x
u
, which starts above the
strike price E at rE/D, behaves exactly like the free boundary for the call while the lower
free boundary x
l
, which starts at the strike price E, behaves at leading order like the free
boundary for the put. We would stress that this only holds true in the 0 when the
boundaries are decoupled. This also holds true in the case D > r, with the roles of put
and call and upper and lower boundaries reversed, but not for the case D =r when both
boundaries start from E. Because of this, for D = r, as 0 the value of an American
straddle at leading order is the sum of the values of an American call and an American
put. We would stress that for larger times, the two boundaries are coupled together, and
the value of an American straddle will be less than the sum of the values of an American
call and an American put.
Proceeding with the analysis, we now substitute the series (4.1) into (3.1)(3.4) and
expand and collect powers of . To evaluate the integrals on the right-hand sides of (3.1)
(3.4), we make the change of variable =, which enables us to pull the dependence
outside of the integrals when we expand. Considering rst the upper boundary, from
(3.1), (3.3), at leading order we nd
_
1
0
_

1
2
exp
_

x
2
u1
(1

)
2
2
(1 +

)
_

x
u1
2
erfc
_
x
u1

1 +

__
d =0,
_
1
0
_
x
u1

2
(1 )
exp
_

x
2
u1
(1

)
2
2
(1 +

)
_
erfc
_
x
u1

1 +

__
d =0.
(4.2)
These two equations (4.2) have a numerical root x
u1
= 0.6388 which agrees with the
value in [1] for the call and also with the one in [8], where this coecient was rst re-
ported. Continuing with our expansion, at the next order, we nd a numerical value of
x
u2
= 0.2898 (r D), again in very good agreement with the value reported in [1] for
the call. Since the decoupled equations for the upper free boundary are identical to those
for the call, it follows that all of the coecients in the series for x
u
will be identical to
those for the call.
Turning now to the lower boundary, we will attempt to use the series (4.1) for x
l
and
again make the change of variable = in (3.2), (3.4). At leading order, we nd

2
exp
_

x
2
l1
2
2
_
+
x
l1
2
erfc
_

x
l1

2
_
=0,
erfc
_

x
l1

2
_
=0.
(4.3)
G. Alobaidi and R. Mallier 9
We note that each of these equations occurs a power of earlier than that for the upper
boundary, which occurs because we cannot replace the error functions on the left-hand
sides of (3.2), (3.4) as we did with (3.1), (3.3). For (4.3) to have a solution requires that
x
l1
= , which suggests that the series (4.1) is inappropriate for x
l
(), and we will in-
stead suppose that
x
l
()

_
n=0
x
ln
()
n/2
, (4.4)
so that the coecients in the series are functions of , and in turn expand the x
ln
()
themselves as series in an unknown function f (), which is assumed to be small,
x
l1
()
_
f ()

_
m=0
x
(m)
l1
_
f ()
_
m
. (4.5)
We need to solve for f () as part of the solution process. Using this newseries (4.4), (4.5),
we need to balance the leading-order terms from (3.2), which tells us that

1/2
(Dr)
2
_
1
0
erfc
_

x
l1
()

x
l1
()

_
2(1 )
_
d
=

2
_
x
l1
()

2
erfc
_

x
l1
()

2
_
+exp
_

x
2
l1
()
2
2
__
.
(4.6)
Since we expect x
l1
() < 0, as 0, the right-hand side of (4.6) tends to

2x
(0)2
l1
f ()
exp

x
(0)2
l1
f ()
2
2

x
(0)
l1
x
(1)
l1

. (4.7)
To evaluate the left-hand side of (4.6), we make the change of variable =1 / f () to
enable us to strip the dependence out of the integral, and we note that
_
1
0
d becomes
_
1/ f ()
0
d/ f ()
_

0
d/ f (). In the limit, the left-hand side of (4.6) becomes

1/2
(Dr)
2 f ()
_

0
erfc

x
(0)
l1
_

d
2
1/2

2
(Dr)
x
(0)2
l1
f ()
, (4.8)
and from (4.7), (4.8), our leading-order equation is therefore
exp

x
(0)2
l1
f ()
2
2

2(r D)

exp

x
(0)
l1
x
(1)
l1

. (4.9)
This has a solution x
(0)
l1
=, f () =ln, and x
(1)
l1
= ln
_
2

2(r D)/
_
. This log-
arithmic behavior is exactly what we would expect, since it is well known [3, 18] that this
is the behavior for the put with D < r, and indeed both x
(0)
l1
and x
(1)
l1
are the same as the
10 The American straddle close to expiry
values for a put, so at leading order, the lower boundary behaves like that of the put. In
a sense, this is a little surprising: if we decompose the European straddle component of
(2.8) into a European put and call, for the case D < r, we would expect the put to be too far
out of the money to contribute on the upper boundary close to expiry, which is why the
upper boundary is identical to that of a call for r < D, but since the lower boundary starts
at the strike price E, there should be a contribution from the call on the lower bound-
ary, but our analysis indicates that such a contribution is not at leading order; however,
we believe it will enter at a later power of since the decoupled equations for the lower
boundary dier slightly from those for the put.
Similarly, balancing the leading-order terms from (3.4) yields the same equation (4.9)
as above. Hence for D < r, close to expiry, the free boundary is of the form
x
u
() x
u1

1/2
+x
u2
+ ,
x
l
()
1/2
_
ln
_
x
(0)
l1
+x
(1)
l1
(ln)
1
+
_
+ .
(4.10)
As might be expected, close to expiry, the upper boundary behaves exactly like the call
boundary, while the lower boundary behaves at leading order like the put boundary.
Case 4.2 (D > r). The free boundary starts at S
l0
= rE/D and S
u0
= E, which is the op-
posite of the case D < r. As with that case, if we take the limit 0, the four equations
(3.1)(3.4) decouple into a pair of equations involving only x
u
but not x
l
, and a pair of
equations involving only x
l
but not x
u
. From our analysis for the case D < r and from
symmetry, in the limit 0 we expect the lower boundary to behave exactly like the
boundary for the put, and the upper boundary to behave at leading order like the bound-
ary for the call, and this is exactly what happens. The actual analysis for this case is es-
sentially the same as for D < r, but with the roles of put and call and upper and lower
boundaries reversed, and so the details are omitted.
For D < r, close to expiry, the free boundary is of the form
x
l
() x
l1

1/2
+x
l2
+ ,
x
u
()
1/2
_
ln
_
x
(0)
u1
+x
(1)
u1
(ln)
1
+
_
+ .
(4.11)
For the lower boundary, the numerical root x
l1
=0.6388 for D > r is minus the value of
x
u1
found for D < r, which agrees with the value for the put with D > r reported in the
literature, while the numerical root of x
l2
=0.2898(r D) is the same as the value of x
u2
found earlier for D < r, although of course r D will be negative when D > r.
For the upper boundary, x
(0)
u1
=, f () =ln, and x
(1)
u1
= ln
_
2

2(Dr)/
_
, so
that x
(0)
u1
is minus the value found for x
(0)
l1
when D < r, and r and Dare interchanged in x
(1)
u1
compared to x
(1)
l1
. This matches the behavior reported for the call with D > r and indicates
that the contribution from the put on the upper boundary is not at leading order. As for
D < r, the upper and lower boundaries behave like those for the call and put.
Case 4.3 (D =r). The free boundary starts at S
l0
=S
u0
=E, and (3.1)(3.4) no longer de-
couple in the limit 0. We recall that for D < r, we found x
(1)
l1
= ln
_
2

2(r D)/
_
,
while for D > r, we had x
(1)
u1
= ln
_
2

2(D r)/
_
; both of which are problematic
G. Alobaidi and R. Mallier 11
when D =r. For this case, intuitively we would expect both the put to inuence the up-
per boundary and the call to inuence the lower boundary.
We will attempt to proceed with the analysis as before, and substitute the series (4.1)
into (3.1)(3.4). At leading order, we once again arrive at (4.3) and their counterparts on
the upper boundary, and these equations have a solution x
l1
= and x
u1
= . Once
again, this is not possible and again suggests that we need to use the series (4.4), (4.5) for
x
l
and a similar series for x
u
.
At the upper boundary, we need to balance the leading-order terms from (3.1), which
tells us that
r
_
1
0

x
u1
()

erfc

x
u1
()

x
u1
()

_
2(1 )

+erfc

x
u1
()

x
l1
()

_
2(1 )

+

_
1

exp

_
x
u1
()

x
u1
()
_
2
2
2
(1 )

+ exp

_
x
u1
()

x
l1
()
_
2
2
2
(1 )

d
=

_
x
u1
()

2
erfc
_
x
u1
()

2
_
exp
_

x
2
u1
()
2
2
__
.
(4.12)
On the lower boundary, balancing the leading-order terms from (3.2) gives (4.12) but
with x
u1
replaced by x
l1
and vice versa, which again tells us that x
l1
= x
u1
, as might
be expected from put-call symmetry for American options [7, 15, 19]. Using this, (4.12)
becomes
r
_
1
0

x
u1
()
2

erfc

x
u1
()

x
u1
()

_
2(1 )

+erfc

x
u1
() +

x
u1
()

_
2(1 )

+

_
1

exp

_
x
u1
()

x
u1
()
_
2
2
2
(1 )

+ exp

_
x
u1
() +

x
u1
()
_
2
2
2
(1 )

d
=

_
x
u1
()

2
erfc
_
x
u1
()

2
_
exp
_

x
2
u1
()
2
2
__
.
(4.13)
This diers somewhat from the equations we found for D < r and D > r because of the
presence of functions involving x
u1
() +

x
u1
() rather than x
u1
()

x
u1
(), so
that these equations contain both x
u1
and x
l1
. However, if we again make the substi-
tution = 1 / f (), then we can show that as 0, the terms in (4.13) involving
x
u1
() +

x
u1
() can be neglected compared to those involving x
u1
()

x
u1
(),
12 The American straddle close to expiry
and therefore the leading-order balance is
exp
_

x
(0)2
u1
f ()
2
2
_

_
rx
(0)
u1

exp
_
x
(0)
u1
x
(1)
u1

2
__

_
f (), (4.14)
which has a solution x
(0)
u1
= , f () = W
L
(
2
), and x
(1)
u1
= ln
_
r

2
_
, where W
L
is
the Lambert W function which obeys W
L
(x)e
W
L
(x)
= x. Since x
l1
= x
u1
, it follows that
x
(0)
l1
= and x
(1)
l1
= ln
_
r

2
_
. If we compare this to the values for a vanilla call or put,
we nd the solution has the same form but that the call and put have an additional factor
of 2 inside the logarithm that is not present in x
(1)
l1
and x
(1)
u1
, and this demonstrates that the
put-like element has inuenced the upper boundary and the call-like element the lower
boundary. Intuitively, one would expect an American straddle to be less likely than an
American call to exercised early as a call, because the payo from continuing to hold is
sweeter for the straddle, and this seems to be true close to expiry for the case D =r.
If we take a similar approach with the integral equations from the delta on the bound-
aries, we arrive at the same balance (4.14) as above. Hence for D =r, close to expiry, the
free boundary is of the form
x
u
() x
l
()
1/2
_
W
L
_

2
__
x
(0)
u1
+x
(1)
u1
_
W
L
(
2
)
_
1
+
_
+ . (4.15)
The upper and lower boundaries behave qualitatively like those for the call and put, re-
spectively, but the coecients dier, indicating that even at leading-order close to expiry,
when r =D, the American straddle is worth less than a put and call combined, which is
what one would expect intuitively since a combination of a put and a call carries the right
to exercise either or both of the options while the straddle carries the right to exercise
only one.
5. Discussion
In the previous sections, we studied the American straddle and were able to use a tech-
nique due to [16] to derive an integral expression for the value of such a straddle similar
to that found in the literature for American calls and puts [6, 11, 16, 20]. An American
straddle is an option which allows the holder to either buy or sell (but not both) the un-
derlying stock at the strike price at or before the expiry of the option, and this right to
early exercise leads to a free boundary problem with not one but two free boundaries, a
lower one S
l
() on which an exercise as a put is optimal and an upper one S
l
() on which
an exercise as a call is optimal. The free boundaries for the straddle dier from those for
the call and the put, so that an American straddle diers from the combination of an
American call and put.
We were also able to derive a set of four coupled (Volterra of the second kind) in-
tegral equations governing the location of the free boundary, and solve these equations
asymptotically to nd expressions for the location of the free boundary. We found that
there were three dierent cases depending on the relative values of the risk-free rate r and
the dividend yield D. Writing S
u
= S
u0
e
x
u
and S
l
= S
l0
e
x
l
, we found that when D < r, x
u
G. Alobaidi and R. Mallier 13
behaved like
1/2
as 0 while x
l
behaved like

ln (4.10), with the situation re-
versed when D > r (4.11), and when D =r and both free boundaries start from the strike
price, x
u
and x
l
both behaved like
_
W
L
(
2
) (4.15).
It behooves us to say a few words about this behavior close to expiry. As we mentioned
in the previous section, in this limit, we would expect the behavior of an upper boundary
starting above the strike (the case D < r) to behave like the call as it will not feel the
inuence of the out of the money put close to expiry, and similarly we would expect the
behavior of a lower boundary starting below the strike (the case D > r) to behave like the
put as it will not feel the inuence of the out of the money call close to expiry, and both
of these aspects were observed in the previous section.
However, we would expect the behavior of an upper boundary starting at the strike
(the cases D r) to behave qualitatively but not quantitatively like the call as it should
feel the inuence of the money put close to expiry, with a similar relation between the
behavior of a lower boundary starting at the strike (the cases D r) and the put, and
these eects can indeed be seen in the case D =r where the coecients x
(1)
u1
and x
(1)
l1
dier
from their counterparts for the call and put, although the behavior is qualitatively the
same. However, for the cases D = r, the eect cannot be seen at leading order, which is
slightly surprising but presumably the eect would occur at a subsequent order.
Finally, we would mention that since this problemhas not one but two free boundaries,
our ability to tackle such a problem demonstrates the power of the integral equation
approach.
References
[1] G. Alobaidi and R. Mallier, Asymptotic analysis of American call options, International Journal of
Mathematics and Mathematical Sciences 27 (2001), no. 3, 177188.
[2] , Laplace transforms and the American straddle, Journal of Applied Mathematics 2 (2002),
no. 3, 121129.
[3] G. Barles, J. Burdeau, M. Romano, and N. Samsoen, Critical stock price near expiration, Mathe-
matical Finance 5 (1995), no. 2, 7795.
[4] G. Barone-Adesi and R. J. Elliot, Free boundary problems in the valuation of securities, Working
Paper, University of Alberta, Alberta, 1989.
[5] F. Black and M. Scholes, The pricing of options and corporate liabilities, Journal of Political Econ-
omy 81 (1973), 637659.
[6] P. Carr, R. Jarrow, and R. Myneni, Alternative characterizations of the American put option, Math-
ematical Finance 2 (1992), no. 2, 87106.
[7] M. Chesney and R. Gibson, State space symmetry and two-factor option pricing models, Advances
in Futures and Options Research 8 (1993), 85112.
[8] J. N. Dewynne, S. D. Howison, I. Rupf, and P. Wilmott, Some mathematical results in the pricing
of American options, European Journal of Applied Mathematics 4 (1993), no. 4, 381398.
[9] R. Geske and H. Johnson, The American put option valued analytically, Journal of Finance 39
(1984), 15111524.
[10] J.-Z. Huang, M. G. Subrahamanyan, and G. G. Yu, Pricing and hedging American options: a re-
cursive investigation method, The Review of Financial Studies 9 (1998), no. 1, 277300.
[11] S. D. Jacka, Optimal stopping and the American put, Mathematical Finance 1 (1991), no. 2, 114.
[12] F. Jamshidian, An analysis of American options, Working Paper, Merrill Lynch Capital Markets,
New York, 1989.
14 The American straddle close to expiry
[13] N. Ju, Pricing by American option by approximating its early exercise boundary as a multipiece
exponential function, Review of Financial Studies 11 (1998), no. 3, 627646.
[14] V. A. Kholodnyi, A nonlinear partial dierential equation for American options in the entire do-
main of the state variable, Nonlinear Analysis 30 (1997), no. 8, 50595070.
[15] V. A. Kholodnyi and J. F. Price, Foreign Exchange Option Symmetry, World Scientic, New Jersey,
1998.
[16] I. J. Kim, The analytic valuation of American options, Review of Financial Studies 3 (1990), no. 4,
547572.
[17] I. I. Kolodner, Free boundary problem for the heat equation with applications to problems of change
of phase. I. General method of solution, Communications on Pure and Applied Mathematics 9
(1956), 131.
[18] R. Mallier and G. Alobaidi, The American put option close to expiry, Acta Mathematica Universi-
tatis Comenianae. New Series 73 (2004), no. 2, 161174.
[19] R. McDonald and M Schroder, A parity result for American options, Journal of Computational
Finance 1 (1998), no. 3, 513.
[20] H. P. McKean Jr, Appendix: a free boundary problem for the heat equation arising from a problem
in mathematical economics, Industrial Management Review 6 (1965), 3229.
[21] R. C. Merton, Theory of rational option pricing, Journal of Economic and Management Sciences
4 (1973), 141183.
[22] P. A. Samuelson, Rational theory of warrant pricing, Industrial Management Review 6 (1965),
1331.
[23] L. N. Tao, The Cauchy-Stefan problem, Acta Mechanica 45 (1982), no. 1-2, 4964.
[24] P. van Moerbeke, On optimal stopping and free boundary problems, Archive for Rational Mechan-
ics and Analysis 60 (1975/76), no. 2, 101148.
Ghada Alobaidi: Department of Mathematics and Statistics, College of Arts and Sciences,
American University of Sharjah, P.O. Box 26666, Sharjah, United Arab Emirates
E-mail address: [email protected]
Roland Mallier: Department of Applied Mathematics, University of Western Ontario,
London, Canada, ON N6A 5B7
E-mail address: [email protected]