76 2 Basic Options

V (S, t
k
)

V (S, t
k+1
) for k = K 2, K 3, , 0
successively. This means

V (S, t
k
)

V (S, t
m
) for k m K.
Letting K and noticing that

V (S, t) generates V (S, t) as K
, we arrive at the conclusion
V (S, t

) V (S, t

) if t

,
which can be rewritten as
V (S, t

) Ze
r(Tt

)
V (S, t

) Ze
r(Tt

)
if t

.
b) The answer is No. For k = K 1, K 2, , 0, we can still dene
V (S, t
k
) = max

e
rt

0
V (S

, t
k+1
)G(S

, t
k+1
; S, t
k
)dS

, nS

and show that if V (S, t

k
) V (S, t
k+1
), then
V (S, t
k
) = max

e
rt

0
V (S

, t
k+1
)G(S

, t
k+1
; S, t
k
)dS

, nS

max

e
rt

0
V (S

, t
k
)G(S

, t
k
; S, t
k1
)dS

, nS

= V (S, t
k1
).
However V (S, t
K1
) V (S, t
K
) is not true, so we cannot prove
V (S, t

) V (S, t

) for t

. In fact, V (0, t) = Ze
r(Tt)
is
an increasing function in t.
31. The American call option is the solution of the following linear comple-
mentarity problem on a nite domain:

(V (, ) max(2 1, 0)) = 0,
V

V 0,
V (, ) max(2 1, 0) 0,
V (, 0) = max(2 1, 0),
where
0 1, 0
and
2 Basic Options 77
L

=
1
2

2
()
2
(1 )
2

2

2
+ (r D
0
)(1 )

[r(1 ) +D
0
].
Reformulate this problem as a free-boundary problem if D
0
> 0.
Solution:
Let G(, ) denote max(2 1, 0). In order to reformulate this problem
as a free-boundary problem, we need to determine when the PDE can be
used and when it cannot be used to get a solution satisfying the constraint
V (, ) G(, ). Suppose that is a suciently small positive number
and that V (, ) is known. We want to determine V (, + ). If in
a neighbourhood of a point (, ), V (, ) > G(, ), then at the point,
the value V (, + ) determined by the PDE must be greater than
G(, + ) when is small enough. Thus that the PDE cannot be
used could occur only when V (, ) = G(, ). In this problem, an initial
condition is given. Thus if
G(, )

G(, ) 0,
then the solution determined by
V (, )

V (, ) =
V (, )

G(, ) = 0
satises the constraint. Consequently, that the PDE cannot be used will
occur only when
G(, )

G(, ) > 0.
Because
G(, ) = max(2 1, 0) =

0, if 0 <
1
2
,
2 1, if
1
2
1,
if 0 < <
1
2
, then G(, ) = 0 and
G(, )

G(, ) = 0.
Consider

1
2
, 1

. In this case G(, ) = 2 1 and

G(, )

G(, ) = 2(r D
0
)(1 ) + (r(1 ) +D
0
)(2 1)
= r(1 ) +D
0

= r + (r +D
0
).
78 2 Basic Options
If
1
2
<
r
r +D
0
, then in the interval

1
2
,
r
r +D
0

, we have
G(, )

G(, ) < 0
and in the interval

r
r +D
0
, 1

,
G(, )

G(, ) > 0.
If
1
2

r
r +D
0
, then in the entire domain

1
2
, 1

,
G(, )

G(, ) > 0.
Let

= max

1
2
,
r
r +D
0

.
Therefore if (0,

), then the solution can be determined by the PDE;

and if (

, 1), then it is possible that the PDE cannot be used to

determine the solution. More precisely, when V (, ) = G(, ) in a neigh-
bourhood of a point (, ) with (

, 1), the solution at that point can-

not be determined by the PDE. Therefore a free boundary a boundary
between the two dierent types of regions can appear only in the case
(

, 1). At = 0, in (0,

), the PDE can be used to determine the

solution and in (

, 1), the PDE cannot be used. Consequently, at = 0

a free boundary will appear at =

and it will move in (

, 1). Let
the location of this free boundary be
f
(). Because in the entire interval
(
f
(), 1)
G(, )

G(, ) > 0,
it is impossible to appear a new free boundary in (
f
(), 1). For an Amer-
ican option, V (, + ) V (, ) is always true. Thus in (0,
f
()) it
is impossible to appear a new free boundary and
f
( + ) >
f
().
Consequently, the free boundary starting at the point (

, 0) is the only
free boundary and

f
()

= max

1
2
,
r
r +D
0

for any . Consequently, the corresponding free-boundary problem is

2 Basic Options 79

V = 0, 0
f
(), 0 ,
V (, 0) = max(2 1, 0), 0
f
(0),
V (
f
(), ) = 2
f
() 1, 0 ,
V (
f
(), )

= 2, 0 ,

f
(0) = max

1
2
,
r
r +D
0

.
32. The American put option is the solution of the following linear comple-
mentarity problem:

2
u
x
2

[u(x, ) g
p
(x, )] = 0,
u

2
u
x
2
0,
u(x, ) g
p
(x, ) 0,
u(x, 0) = g
p
(x, 0),
where
< x < , 0
and
g
p
(x, ) = max(e
2r /
2
e
x+(2D0/
2
+1)
, 0).
Find the domain where a free boundary may appear and the domain
where it is impossible for a free boundary to appear, show that there is
only one free boundary at = 0 and give the starting location of this free
boundary.
Solution:
In this problem an initial value is given, so in order to nd where a free
boundary may appear and where it is impossible for a free boundary to
appear, we need to determine where
g
p
(x, )

2
g
p
(x, )
x
2
> 0,
and where
g
p
(x, )

2
g
p
(x, )
x
2
0.
If e
2r /
2
< e
x+(2D0/
2
+1)
, i.e.,
x > [2(r D
0
)/
2
1] ,
80 2 Basic Options
then
g
p
(x, ) = 0
and
g
p
(x, )

2
g
p
(x, )
x
2
= 0.
If
x < [2(r D
0
)/
2
1] ,
then
g
p
(x, ) = e
2r /
2
e
x+(2D0/
2
+1)
and
g
p
(x, )

2
g
p
(x, )
x
2
= 2r/
2
e
2r /
2
(2D
0
/
2
+ 1)e
x+(2D0/
2
+1)
+e
x+(2D0/
2
+1)
= 2r/
2
e
2r /
2
2D
0
/
2
e
x+(2D0/
2
+1)
= 2D
0
/
2
e
(2D0/
2
+1)

r/D
0
e
[2(rD0)/
2
1]
e
x

.
Therefore if r < D
0
, then when
[2(r D
0
)/
2
1] + ln(r/D
0
) < x < [2(r D
0
)/
2
1] ,
we have
g
p
(x, )

2
g
p
(x, )
x
2
< 0
and when
x < [2(r D
0
)/
2
1] + ln(r/D
0
),
we have
g
p
(x, )

2
g
p
(x, )
x
2
> 0.
If r D
0
, then
g
p
(x, )

2
g
p
(x, )
x
2
> 0
when
x < [2(r D
0
)/
2
1] .
Put them together, we see that if
x < [2(r D
0
)/
2
1] + min

0, ln
r
D
0

,
then the following holds:
g
p
(x, )

2
g
p
(x, )
x
2
> 0;
2 Basic Options 81
and if
x > [2(r D
0
)/
2
1] + min

0, ln
r
D
0

,
then we have
g
p
(x, )

2
g
p
(x, )
x
2
0.
In a region where
g
p
(x, )

2
g
p
(x, )
x
2
> 0,
if on one port of this region u(x, ) > g
p
(x, ) and on another port
u(x, ) = g
p
(x, ), then the boundary between the two ports is a free
boundary because on one port the PDE can be used and on the another
port the PDE cannot be used. Therefore it is possible for a free boundary
to appear in this region. In a region where
g
p
(x, )

2
g
p
(x, )
x
2
0,
we can use the PDE to have solutions satisfying the constraint and no
free boundary can appear in this case.
At = 0, u(x, 0) = g
p
(x, 0). Thus g
p
(x, ) itself completely determines
where the PDE can or cannot be used. Consequently at = 0, when
x < min

0, ln
r
D
0

,
we cannot use the PDE to determine the solution and when
x > min

0, ln
r
D
0

,
the PDE can be used to nd the solution. Therefore there is only one free
boundary and the location of the free boundary at time = 0 is
x = min

0, ln
r
D
0

.
33. The price of a one-factor convertible bond is the solution of the linear
complementarity problem

min

V
t
L
S
V, V (S, t) nS

= 0, 0 S, 0 t T,
V (S, T) = max(Z, nS) nS, 0 S,
where
L
S
=
1
2

2
S
2

2
S
2
+ (r D
0
) S

S
r
82 2 Basic Options
and n, Z, , r, and D
0
are constants. Show that if D
0
> 0, then the
solution of a one-factor convertible bond must involve a free boundary
and its location at t = T is S = Z/n. Assuming that for this problem
there is only one free boundary, formulate this problem as a free boundary
problem.
Solution: Let G(S, t) = nS. In this problem a nal condition is given, we
need to determine where
G(S, t)
t
+L
S
G(S, t) 0
and where
G(S, t)
t
+L
S
G(S, t) < 0.
Because G(S, t) = nS, we have
G(S, t)
t
+L
S
G(S, t)
=

t
+
1
2

2
S
2

2
S
2
+ (r D
0
)S

S
r

nS
= (r D
0
)Sn rnS = D
0
Sn < 0 if D
0
> 0.
Thus for any S, it is possible that the PDE cannot be used. The nal
condition is
V (S, T) = max(Z, nS).
In the interval (0, Z/n), V (S, T) = Z > G(S, T) = nS and the par-
tial dierential equation will give a solution satisfying the condition
V (S, T t) > G(S, T t) for a suciently small positive t. In the
interval (Z/n, ), the PDE cannot give a solution satisfying the condition
V (S, T t) > G(S, T t). Therefore there is a free boundary starting
from the point (Z/n, T). Let the location of this free boundary be S
f
(t).
Let us assume that for this problem, this is the only free bounadry. Then
the corresponding free boundary problem is:

V
t
+
1
2

2
S
2

2
V
S
2
+ (r D
0
)S
V
S
rV = 0,
0 S S
f
(t), 0 t T,
V (S, T) = max(Z, nS), 0 S S
f
(T),
V (S
f
(t), t) = nS
f
(t), 0 t T,
V
S
(S
f
(t), t) = n, 0 t T,
S
f
(T) =
Z
n
.
2 Basic Options 83
33

The price of a one-factor convertible bond is the solution of the linear

complementarity problem

min

V
t
L
S
V kZ, V (S, t) nS

= 0, 0 S, 0 t T,
V (S, T) = max(Z, nS) nS, 0 S,
where
L
S
=
1
2

2
S
2

2
S
2
+ (r D
0
)S

S
r,
and k, Z, n, , r and D
0
are constants. Show that if D
0
> 0, then the solu-
tion of a one-factor convertible bond must involve a free boundary and its
location at t = T is S = max

Z
n
,
kZ
D
0
n

. Also, derive the corresponding

free-boundary problem if this problem has only one free boundary.
Solution: At t = T, if S < Z/n, we have V (S, T) = Z > nS = G(S, T)
and the PDE can be used. If S > Z/n, we have V (S, T) = nS = G(S, T)
and in order to determine if the PDE can be used, we need to determine
where
G(S, T)
t
+L
S
G(S, T) +kZ 0
and where
G(S, T)
t
+L
S
G(S, T) +kZ < 0.
In this case we have
G(S, T)
t
+L
S
G(S, T) +kZ
=

t
+
1
2

2
S
2

2
S
2
+ (r D
0
)S

S
r

nS +kZ
= (r D
0
)Sn rnS +kZ = D
0
n

kZ
D
0
n
S

.
Thus if
Z
n
>
kZ
D
0
n
, then
G(S, T)
t
+ L
S
G(S, T) + kZ < 0 holds for
any S > Z/n, which means existing a free boundary and the free
boundary is located at S = Z/n for t = T. If
Z
n
<
kZ
D
0
n
, then
G(S, T)
t
+ L
S
G(S, T) + kZ > 0 holds when S

Z
n
,
kZ
D
0
n

and
G(S, T)
t
+ L
S
G(S, T) + kZ < 0 holds when S

kZ
D
0
n
,

. Thus the
free boundary is located at S =
kZ
D
0
n
. Put the two cases together, we know
84 2 Basic Options
that the free boundary is located at S = max

Z
n
,
kZ
D
0
n

for t = T. How-
ever because max

Z
n
,
kZ
D
0
n

as D
0
0, there is no free boundary
if D
0
= 0. Let the location of this free boundary be S
f
(t) and we further
assume that for this problem this is the only free bounadry. Then the
corresponding free boundary problem is:

V
t
+
1
2

2
S
2

2
V
S
2
+ (r D
0
)S
V
S
rV +kZ = 0,
0 S S
f
(t), 0 t T,
V (S, T) = max(Z, nS), 0 S S
f
(T),
V (S
f
(t), t) = nS
f
(t), 0 t T,
V
S
(S
f
(t), t) = n, 0 t T,
S
f
(T) = max

Z
n
,
kZ
D
0
n

.
33

. Suppose r, D
0
, and are constant.
a) Derive the putcall symmetry relations.
b) Explain the nancial meaning of the symmetry relation.
c) Explain how to use these relations when we write codes if a code for
put options is quite a dierent from a code for call options.
Solution:
a) As we know, the price of an American put option is the solution of
the following LC problem:

min

P
t
L
S
P, P(S, t) max(E S, 0)

= 0, 0 S, t T,
P(S, T) = max(E S, 0), 0 S,
where
L
S
=
1
2

2
S
2

2
S
2
+ (r D
0
)S

S
r.
Let

=
E
2
S
,
C(, t) =
EP(S, t)
S
.
Because
E
S
max(E S, 0) = max( E, 0),
2 Basic Options 85
for C(, t) the payo and constraint are max( E, 0). Noticing
P
t
=
S
E
C
t
,
P
S
=
1
E

C +S
C

E
2
S
2

=
1
E

C
C

2
P
S
2
=

3
E
3

2
C

2
,
we have
P
t
+
1
2

2
S
2

2
P
S
2
+ (r D
0
)S
P
S
rP
=
S
E

C
t
+
1
2

2
C

2
+ (D
0
r)
C

D
0
C

.
Therefore the function C(, t) is the solution of the following American
call option problem:

min

C
t
L

C, C(, t) max( E, 0)

= 0, 0 , t T,
C(, T) = max( E, 0), 0 ,
where
L

=
1
2

2

2

2
+ (D
0
r)

D
0
.
Consequently, an American put problem can be converted into an
American call problem. However in the two probloms, the state vari-
able and the parameters are dierent. From the denitions of L
S
and
L

, we know that the volatilities of the put and call problems are the
same, but the interest rate and the dividend yield of the call problem
are equal to the dividend yield and the interest rate of the put prob-
lem, respectively. In order to explain these facts, we express the depen-
dences of the options on interest rate and dividend yield explicitely.
Let P(S, t; b, a) denote the price of the put option and C(, t; a, b) the
price of the call option, where the rst and second parameters after
the semicolon are the interest rate and the dividend yield, respectively.
From the denition of and C(, t; a, b), we know
P(S, t; b, a) = C (, t; a, b) S/E,
where =
E
2
S
. This can also be rewritten as
P(S, t; b, a) = C

E
2
/S, t; a, b

S/E,
or
86 2 Basic Options
C (, t; a, b) = P

E
2
/, t; b, a

/E,
where we used the relation E/S = /E . In the last relation, we can
use S, instead of , as the state variable. That is, we can write this
relation as
C (S, t; a, b) = P

E
2
/S, t; b, a

S/E.
Finally, putting them together, we have

C(S, t; a, b) = P

E
2
/S, t; b, a

S/E, or
P(S, t; b, a) = C

E
2
/S, t; a, b

S/E.
Also, the location of free boundary in the latter problem,
cf
(t; a, b),
must be equal to E
2
divided by the location of free boundary of the
former problem, E
2
/S
pf
(t; b, a), because = E
2
/S, i.e.,

cf
(t; a, b) = E
2
/S
pf
(t; b, a)
or
S
cf
(t; a, b) S
pf
(t; b, a) = E
2
,
where in the last relation, instead of
cf
, we use S
cf
as the name of
the function representing the location of the free boundary. From the
derivation above we know that for Euporean options, the following
relations also hold:

c(S, t; a, b) = p

E
2
/S, t; b, a

S/E, or
p(S, t; b, a) = c

E
2
/S, t; a, b

S/E.
b) Suppose that one British pound is worth S U.S. dollars and that E
2
U.S. dollars are worth British pounds. It is clear that = E
2
/S.
Let P be the value of a put option whose holder can always sell one
pound for E dollars if the holder wants. This means that the payo
and constraint of the put option is max(ES, 0) in dollars. Let C be
the value of a call option whose holder can buy E
2
dollars by paying N
pounds if the holder wants. This means that the payo and constraint
of the call option are max(E
2
/S E, 0) = max( E, 0) in pounds.
The holder of the put option has the right to sell one pound for E
U.S. dollars even if S E. The holder of 1/E units of the call option
has the right to buy E dollars by paying one British pound even if
E. The condition S E is equivalent to E
2
/S = E. Thus,
both the holder of one unit of the put option and the holder of 1/E
units of the call options have the right to exchange one pound for E
dollars even if S < E. The two holders have the same rights, so the
value of one unit of the put option and the value of 1/E units of the
call option in U.S. dollars, which is equal to S C/E, should be equal,
i.e.,
2 Basic Options 87
P = S C/E.
Here, we need to notice that P and C have dierent but related volatil-
ities, interest rates, and dividend yields. According to It os lemma, if
dS = Sdt +SdX,
then
d = ( +
2
)dt dX.
Hence, the volatilities of S and = E
2
/S are the same if the volatil-
ities are constants. Suppose that , r, and D
0
are constant and that
the interest rates of the British pound and the U.S. dollar are a and
b, respectively. Then for the call, r = a and D
0
= b and for the put
r = b and D
0
= a, and the volatilities are the same. In this case, the
relation above can be written as
P(S, t; b, a) = C

E
2
/S, t; a, b

S/E.
Let S
pf
(t; b, a) be the position of the free boundary corresponding to
P(S, t; b, a), and let
cf
(t; a, b) be the position of the free boundary
corresponding to C(, t; a, b). Because = E
2
/S,
cf
(t; a, b) should be
equal to E
2
/S
pf
(t; b, a), i.e.,
S
pf
(t; b, a)
cf
(t; a, b) = E
2
.
Moreover, if we use S, instead of , as the state variable for the call
option C, then the relation above can be written as
S
pf
(t; b, a) S
cf
(t; a, b) = E
2
.
c) Suppose that we need codes for pricing American call and put options
and that we already have a code for pricing American call options.
If it is very easy for the code to be modied to a code for pricing
American put options, then we can have another code for put options
by modifying the code we already have. If the code for put options
will be quite a dierent from the code for call options, then we can use
the code for call options to nd C(E
2
/S, t; a, b) rst and then obtain
P(S, t; b, a) by using the relation P(S, t; b, a) = C(E
2
/S, t; a, b) S/E.
34. a) Suppose = (S, t), r = r(t) and D
0
= D
0
(S, t). Show that the
problem of pricing a put option can always be converted into a problem
of pricing a call option.
b) Let the exercise price be E. Suppose that r, D
0
are constants and
= (S). Show
P(S, t; b, a, (S)) = C

E
2
/S, t; a, b, (S)

S/E,
C (S, t; a, b, (S)) = P

E
2
/S, t; b, a, (S)

S/E
88 2 Basic Options
and
S
cf
(t; a, b, (S)) S
pf
(t; b, a, (E
2
/S)) = E
2
.
Here the rst, second and third parameters after the semicolon in
P, C, S
pf
and S
cf
are the interest rate, the dividend yield and the
volatility function, respectively.
Solution:
a) Suppose that the put option is an American option, which is the so-
lution of the following problem:

min

P
t
L
S
P, P(S, t) max(E S, 0)

= 0, 0 S, t T,
P(S, T) = max(E S, 0), 0 S,
where
L
S
=
1
2

2
(S, t)S
2

2
S
2
+ [r(t) D
0
(S, t)]S

S
r(t).
Let =
E
2
S
and C(, t) =
EP(S, t)
S
. Because
C(, T) =
EP(S, T)
S
=
E
S
max(E S, 0) = max( E, 0),
C(, t) has a payo of a call option. Noticing
P
t
=
S
E
C
t
,
P
S
=
1
E

C +S
C

E
2
S
2

=
1
E

C
C

2
P
S
2
=

3
E
3

2
C

2
,
we have
P
t
+
1
2

2
(S, t)S
2

2
P
S
2
+ [r(t) D
0
(S, t)]S
P
S
r(t)P
=
S
E

C
t
+
1
2

E
2

, t

2
C

2
+

D
0

E
2

, t

r(t)

D
0

E
2

, t

.
Therefore the function C(, t) satises

min

C
t
L

C, C(, t) max( E, 0)

= 0, 0 , t T,
C(, t) = max( E, 0), 0 ,
2 Basic Options 89
where
L

=
1
2

E
2

, t

2

2

2
+

D
0

E
2

, t

r(t)

D
0

E
2

, t

.
Therefore any American put problem can be converted into an Amer-
ican call problem. From the derivation, we know that the conclusion
is also correct for European options.
b) From a) we know
P(S, t; b, a, (S)) =
S
E
C

, t; a, b,

E
2
/

where =
E
2
S
. This can also be rewritten as
P(S, t; b, a, (S)) =
S
E
C

E
2
/S, t; a, b, (S)

or
C

, t; a, b, (E
2
/)

= P

E
2
/, t; b, a, (E
2
/)

/E.
In the last relation, we can use S instead of and further use (S)
instead of (E
2
/S). Hence it can be rewritten as
C (S, t; a, b, (S)) = P

E
2
/S, t; b, a, (S)

S/E.
Suppose that for the put option the location of the free boundary is
S
pf
(t; b, a, (S)) and the location of the free boundary for the call
option is
cf

t; a, b, (E
2
/)

. Because the two problems are actually

the same, the following relation holds:

cf

t; a, b, (E
2
/)

=
E
2
S
pf
(t; b, a, (S))
or
S
cf

t; a, b, (E
2
/S)

S
pf
(t; b, a, (S)) = E
2
.
Here S
cf

t; a, b, (E
2
/S)

is the same as
cf

t; a, b, (E
2
/)

, only
the state variable being replaced by S.
35. Suppose that , r, D
0
are constants. In this case we have the following
symmetry relation for European options
p(S, t; b, a) = c

E
2
S
, t; a, b

S/E,
where the rst and second arguments after the semicolon in p and c are
the values of the interest rate and the dividend yield, respectively. For a
European call option, the price is
90 2 Basic Options
c(S, t) = Se
D0(Tt)
N(d
1
) Ee
r(Tt)
N(d
2
),
where
d
1
=
ln
Se
D0(Tt)
Ee
r(Tt)
+
1
2

2
(T t)

T t
,
d
2
=
ln
Se
D0(Tt)
Ee
r(Tt)

1
2

2
(T t)

T t
.
Find the price of a European put option by using the symmetry relation.
Solution:
p(S, t; b, a)
= c(E
2
/S, t; a, b)S/E
=
S
E

E
2
S
e
b(Tt)
N

ln
E
2
/S e
b(Tt)
Ee
a(Tt)
+
1
2

2
(T t)

T t

Ee
a(Tt)
N

ln
E
2
/S e
b(Tt)
Ee
a(Tt)

1
2

2
(T t)

T t

= Ee
b(Tt)
N

ln
Se
a(Tt)
Ee
b(Tt)

1
2

2
(T t)

T t

Se
a(Tt)
N

log
Se
a(Tt)
Ee
b(Tt)
+
1
2

2
(T t)

T t

.
Therefore
p(S, t) = Ee
r(Tt)
N(d
2
) Se
D0(Tt)
N(d
1
).
36. Derive the formulation of the problem for
P
r
and write down the for-
mulation of the problems for
P

and
P
D
0
, where P is the price of an
American put option.
Solution:
Let P
r
denote
P(S, t; r, D
0
, )
r
. Here, in order to see the meaning of
the derivative with respect to r, we explicitly show that the dependence