1.2
1.1.1
1.1.2
Boundary Condition . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.1
1.2.2
Boundary Condition . . . . . . . . . . . . . . . . . . . . . . . . .
2 Ju Model (1998)
2.1
2.2
10
2.2.1
Value-Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
2.2.2
High-Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
11
Chapter 1
Carr-Jarrow-Myneni Model (1992)
Carr, P., R. Jarrow, and R. Myneni, 1992, Alternative Characterizations of American
Puts, Mathematical Finance, Vol. 2, 87-106.
1.1
1.1.1
Consider an American call option on the stock with strike price K and maturity date T .
Let C A (t, St , K, T ) denote the value of the American call option at time t [0, T ]. For
each time t, there exists a critical stock price, Bt , above which the American call option
should be exercised early, i.e.,
C A (t, St , K, T ) = H (t, St , K, T ) 1{St <Bt } + (St K) 1{St Bt } ,
(1.1)
where H (t, St , K, T ) is the value of option with continuing to hold. This boundary is
independent of the current stock price St and is a smooth, non-increasing function of
time t terminating in the strike price, i.e. BT = K.
Given the filtration F0 , the American call value, C A (0, S0 , K, T ), can be decomposed
into the corresponding European call price, C E (0, S0 , K, T ), and the early exercise premium, (0, S0 , K, T ):
C A (0, S0 , K, T ) = C E (0, S0 , K, T ) + C (0, S0 , K, T ) ,
where
C
(0, S0 , K, T ) =
qS0 eqt N (d1,t,S0 ) rKert N (d2,t,S0 ) dt,
(1.2)
d1,u,S0
S0
1 2
ln
+ rq+ u
Bu
2
,
=
u
d2,u,S0 = d1,u,S0 u.
S
dt
+
e
St dWtQ .
t
2
t
St
2 St
St
And, each derivative w.r.t t and St is
CtA
Ht
=
1{St <Bt } ,
t
t
CtA
Ht
=
1{St <Bt } + 1{St Bt } ,
St
St
2 CtA
2 Ht
=
1{St <Bt } .
St2
St2
rK
E
1
dt
{St Bt }
{St Bt }
0
0
0
Z T
E
=C0 +
qS0 eqt N (d1,t,S0 ) rKert N (d2,t,S0 ) dt.
0
1.1.2
Boundary Condition
(1.3)
Once Bt , is obtained, the price of the American call can be calculated easily.
1.2
1.2.1
Consider an American put option on the stock with strike price K and maturity date T .
Let P A (t, St , K, T ) denote the value of the American put option at time t [0, T ]. For
each time t, there exists a critical stock price, Bt , above which the American put option
should be exercised early, i.e.,
P A (t, St , K, T ) = H (t, St , K, T ) 1{St >Bt } + (St K) 1{St Bt } ,
(1.4)
where H (t, St , K, T ) is the value of option with continuing to hold. This boundary is
independent of the current stock price St and is a smooth, non-decreasing function of
time t terminating in the strike price, i.e. BT = K.
Given the filtration F0 , the American put value, P A (0, S0 , K, T ), can be decomposed
into the corresponding European put price, P E (0, S0 , K, T ), and the early exercise premium, (0, S0 , K, T ):
P A (0, S0 , K, T ) = P E (0, S0 , K, T ) + P (0, S0 , K, T ) ,
where
Z
(0, S0 , K, T ) =
rKert N (d2,t,S0 ) qS0 eqt N (d1,t,S0 ) dt
=K 1 erT S0 1 eqT + C (0, S0 , K, T ) ,
d1,u,S0
S0
1 2
+ rq+ u
ln
Bu
2
,
=
u
d2,u,S0 = d1,u,S0 u.
(1.5)
S
t
t
St
2 Pt2
St
And, each derivative w.r.t t and St is
PtA
Ht
=
1{St >Bt } ,
t
t
PtA
Ht
=
1{St >Bt } 1{St Bt } ,
St
St
2 PtA
2 Ht
=
1{St >Bt } .
St2
St2
ert
PtA
St dWtQ .
St
At maturity, the payoffs of American and European are the same, i.e.
PTA = PTE = (K St ) 1{ST <K} .
Thus, the value of American put option at time 0 is
Z T
Z T
A
Q
Q
A
rT E
rt
rt Pt
St dWt
P0 = E0 e PT +
e (rK qSt ) 1{St Bt } dt +
e
St
0
0
Z T
Q
E
rt
=P0 + E0
e (rK qSt ) 1{St Bt } dt
0
Z Tn
1 2
o
E
qt Q
2 t+ tt
=P0 +
rK EQ
1
qS
e
E
e
1
dt
0
{St Bt }
{St Bt }
0
0
0
Z T
E
=P0 +
rKert N (d2,t,S0 ) qS0 eqt N (d1,t,S0 ) dt.
0
1.2.2
Boundary Condition
=P E (t, Bt , K, T ) + K 1 erT S0 1 eqT
Z T
+
qBt equ N (d1,u,Bt ) rKeru N (d2,u,Bt ) du.
t
Once Bt , is obtained, the price of the American put can be calculated easily.
(1.6)
Chapter 2
Ju Model (1998)
Ju, N., 1998, Pricing an American Option by Approximating Its Early Exercise Boundary as a Multi-Piece Exponential Function, Review of Financial Studies, Vol. 11, 627646.
2.1
Assume Bt to be an exponential function a exp (bt) for the interval [t1 , t2 ]. Given the
filtration F0 , t1 0, consider an integral
Z t2
I1 (t1 , t2 , a, b) =
reru N (d2,u,S0 ) du
(2.1)
t1
x2
x2
rt1
rt2
e
=e
N x1 t2 +
N x1 t1 +
t1
t2
x2 (x3 x1 )
e
x1
x2
x2
+
+ 1 N x3 t2 +
N x3 t 1 +
2
x3
t2
t1
1 N x3 t 2
N x3 t 1
,
2
x3
t2
t1
where
1
r q b 2
2 ,
x1 =
S0
a ,
ln
x2 =
q
x3 = x21 + 2r.
1 rt1
x2
x2
rt2
N x1 t1 +
N x1 t2 +
= e
e
r
t1
t2
x2 (x3 x1 ) Z t2
1
1
x1 1 x 2 3
e
n x3 u 2 + x2 u 2
u 2 u 2 du.
+
r
2
2
t1
Recall the property of normal distribution,
x
1
1
1
1
x2 3
3 1
d N x3 u 2 + x2 u 2 = n x3 u 2 + x2 u 2
u 2 u 2 du,
2
2
x
1
1
1
1
x2 3
3 1
d N x3 u 2 x2 u 2 = n x3 u 2 x2 u 2
u 2 + u 2 du
2
2
1
1
1
x2 3
x
3
=e2x2 x3 n x3 u 2 + x2 u 2
u 2 + u 2 du.
2
2
Thus, we can easily get
Z t2
x
1
1 1
21
2
u 2 du
n x3 u + x2 u
2
t1
Z
x
1
1
x1 t2
3 1
u 2 du
=
n x3 u 2 + x2 u 2
x3 t1
2
Z t2
x e2x2 x3 Z t2
1
1
x1
1
12
12
2
2
=
d N x3 u + x2 u
+
d N x3 u x2 u
2x3 t1
2x3
t1
x1
x2
x1
x2
=
N x3 t2 +
N x3 t1 +
2x3
2x3
t2
t1
2x2 x3
2x
x
2
3
x1 e
x2
x2
x1 e
+
N x3 t2
N x3 t1
,
2x3
2x3
t2
t1
and
Z
t2
x
1
1
1 1
n x3 u 2 + x2 u 2
u 2 du
2
t1
Z
1 Z t2
1
1
1
1
e2x2 x3 t2
=
d N x3 u 2 x2 u 2
d N x3 u 2 + x2 u 2
2
2 t1
t1
e2x2 x3
e2x2 x3
x2
x2
=
N x3 t2
N x3 t1
2
2
t2
t
1
1
x2
1
x2
N x3 t2 +
+ N x3 t1 +
.
2
2
t2
t1
x2
x2
rt1
rt2
N x1 t1 +
N x1 t2 +
=e
e
t1
t2
ex2 (x3 x1 ) x1
x2
x2
+ 1 N x3 t2 +
N x3 t1 +
+
2
x3
t2
t1
x2 (x3 +x1 )
e
x2
x2
x1
N x3 t1
.
1 N x3 t2
2
x3
t2
t1
y2
y2
qt1
qt2
=e
N y1 t1 +
e
N y1 t2 +
t1
t2
y2 (y3 y1 )
y1
y2
y2
e
+ 1 N y3 t2 +
+
N y3 t1 +
2
y3
t2
t1
y2
y2
ey2 (y3 +y1 ) y1
1 N y3 t2
N y3 t1
,
2
y3
t2
t1
where
1
r q b + 2
2 ,
y1 =
S0
a ,
ln
y2 =
q
y3 = y12 + 2q.
1 qt1
y2
y2
qt2
= e
N y1 t1 +
e
N y1 t2 +
q
t1
t2
y2 (y3 y1 ) Z t2
1
1
e
y1 1 y2 3
+
n y3 u 2 + y2 u 2
u 2 u 2 du.
q
2
2
t1
Recall the propeqty of normal distribution,
1
2
12
d N y3 u + y2 u
1
2
21
= n y3 u + y2 u
y
12
y2 3
u 2 du,
2
y
1
1
1
1
y2 3
3 1
d N y3 u 2 y2 u 2 = n y3 u 2 y2 u 2
u 2 + u 2 du
2
2
1
1
y
y2 3
3 1
=e2y2 y3 n y3 u 2 + y2 u 2
u 2 + u 2 du.
2
2
Thus, we can easily get
Z t2
y
1
1
1 1
n y3 u 2 + y2 u 2
u 2 du
2
t1
Z
y
1
1
y1 t2
3 1
=
n y3 u 2 + y2 u 2
u 2 du
y3 t1
2
Z t2
y e2y2 y3 Z t2
1
1
1
1
y1
1
=
d N y3 u 2 + y2 u 2 +
d N y 3 u 2 y 2 u 2
2y3 t1
2y3
t1
y1
y1
y2
y2
=
N y3 t2 +
N y3 t1 +
2y3
2y3
t2
t1
2y
y
2y2 y3
2
3
y1 e
y1 e
y2
y2
,
+
N y3 t2
N y3 t1
2y3
2y3
t2
t1
and
Z
t2
y
1
1
1 1
n y3 u 2 + y2 u 2
u 2 du
2
t1
Z
1 Z t2
1
1
e2y2 y3 t2
12
21
2
2
=
d N y3 u y2 u
d N y3 u + y2 u
2
2 t1
t1
e2y2 y3
y2
e2y2 y3
y2
N y3 t2
N y3 t1
=
2
2
t2
t
1
1
y2
1
y2
N y3 t2 +
+ N y3 t1 +
.
2
2
t2
t1
Substitute the results into the initial integral, then
Z t2
qequ N (d2,u,S0 ) du
t1
y2
y2
qt2
qt1
e
N y1 t2 +
=e
N y1 t1 +
t1
t2
y2 (y3 y1 )
e
y1
y2
y2
+
+ 1 N y3 t2 +
N y3 t1 +
2
y3
t2
t1
1 N y3 t2
N y3 t1
.
2
y3
t2
t1
n
Consider a n-piece exponential function, aj ebj t j=1 and tj =
jT
n
n
j=0
P
CnA (0, S0 , K, T )
K nj=1 I1 (tj1 , tj , aj , bj ) ,
S0 < a1 ,
S a ,
S 0 a1 .
0
1
9
(2.3)
+S Pn I (t , t , a , b )
0
j j
j=1 2 j1 j
PnA (0, S0 , K, T )
P
n
K j=1 I1 (tj1 , tj , aj , bj )
S > a1 ,
a S ,
S0 a1 .
1
0
2.2
(2.4)
To determine the series of {ai }ni=1 and {bi }ni=1 , Ju (1998) applied the value-match and
the high-contact conditions.
2.2.1
Value-Matching
For the American call option, given the filtration F0 , the value-matching condition at
each time ti =
iT
ai+1 ebi+1 n
iT
,
n
i = 0, 1, 2, , n 1, is
(n i) T
E
bi+1 iT
n , K,
K =C
0, ai+1 e
(2.5)
n
ni
ni
X
X
bi+1 iT
n
+ ai+1 e
I2 (tj1 , tj , aj , bj ) K
I1 (tj1 , tj , aj , bj ) .
j=1
j=1
For the American put option, given the filtration F0 , the value-matching condition at
iT
,
n
i = 0, 1, 2, , n 1, is
(ni)T
iT
(n i) T
bi+1 iT
E
b
i+1
n =P
n , K,
K ai+1 e
0, ai+1 e
+ K 1 er n
n
(ni)T
bi+1 iT
q n
n
ai+1 e
1e
each time ti =
+ ai+1 e
bi+1 iT
n
ni
X
I2 (tj1 , tj , aj , bj ) K
j=1
ni
X
j=1
10
(2.6)
I1 (tj1 , tj , aj , bj ) .
2.2.2
High-Contact
I1
e
x2
ert2
x2
1
= n x1 t1 +
(2.7)
n x1 t2 +
S0
S0
t1
t1
t2
t2
ex2 (x3 x1 ) x1
x2
x2
+
+ 1 N x3 t2 +
N x3 t1 +
(x3 x1 )
2S0
x3
t2
t1
ex2 (x3 x1 ) x1
x2
1
x2
1
n x3 t1 +
+
+ 1 n x3 t2 +
2S0
x3
t2
t2
t1
t1
x2 (x3 +x1 )
x2
x2
e
x1
N x3 t1
(x3 + x1 ) ,
+
1 N x3 t2
2S0
x3
t2
t1
+
1 n x3 t2
,
2S0
x3
t2
t2
t1
t1
and the derivative of equation (2.2) is
qt1
I2
e
y2
eqt2
y2
1
= n y1 t1 +
n y1 t2 +
(2.8)
S0
S0
t1
t1
t2
t2
ey2 (y3 y1 ) y1
y2
y2
+ 1 N y3 t2 +
+
N y3 t1 +
(y3 y1 )
2S0
y3
t2
t1
y2
y2
1
1
ey2 (y3 y1 ) y1
n y3 t1 +
+
+ 1 n y3 t2 +
2S0
y3
t2
t2
t1
t1
y2 (y3 +y1 )
y1
y2
y2
e
1 N y3 t2
+
N y3 t1
(y3 + y1 ) ,
2S0
y3
t2
t1
y2
y2
ey2 (y3 +y1 ) y1
1
1
n y3 t1
1 n y3 t2
+
.
2S0
y3
t2
t2
t1
t1
Let the notations I10 =
I1
S0
and I20 =
I2
.
S0
the derivatives of w.r.t. S0 for the LHS and RHS are the same as S0 B0 . For the
American call option, given the filtration F0 , the high-contact condition at each time
ti =
iT
,
n
i = 0, 1, 2, , n 1, is
1 =e
(ni)T
n
N d1, (ni)T ,a
n
+ ai+1 e
bi+1 iT
n
ni
X
iT
b
i+1 e i+1 n
ni
X
I2 (tj1 , tj , aj , bj )
j=1
I20
(tj1 , tj , aj , bj ) K
j=1
ni
X
j=1
11
I10 (tj1 , tj , aj , bj ) .
(2.9)
For the American put option, given the filtration F0 , the high-contact condition at each
time ti =
iT
,
n
1 =e
i = 0, 1, 2, , n 1, is
(ni)T
n
N d1, (ni)T ,a
n
iT
b
i+1 e i+1 n
1e
(ni)T
n
ni
X
I2 (tj1 , tj , aj , bj )
j=1
(2.10)
bi+1 iT
n
+ ai+1 e
ni
X
I20
(tj1 , tj , aj , bj ) K
j=1
ni
X
j=1
12
I10 (tj1 , tj , aj , bj ) .