Professional Documents
Culture Documents
Exoticoptions
Exoticoptions
Exotic Options
where (Bt )t∈R+ is a standard Brownian motion under the risk-neutral prob-
ability measure P∗ . In particular the value Vt of a self-financing portfolio
satisfies wT
VT e−rT = V0 + σ ξt St e−rt dBt , t ∈ [0, T ].
0
8.1 Generalities
An exotic option is an option whose payoff may depend on the whole path
{St : t ∈ [0, T ]} of the price process via a “complex” operation such as
averaging or computing a maximum. They are opposed to vanilla options
whose payoff
C = φ(ST ),
where φ is called a payoff function, depends only on the terminal value ST of
the price process.
where fST (y) is the (one parameter) probability density function of ST , which
satisfies wy
P(ST ≤ y) = fST (v)dv, y ∈ R.
−∞
"
N. Privault
Exotic Options
Exotic options, also called path-dependent options, are options whose payoff
C may depend on the whole path
{St : 0 ≤ t ≤ T }
Options on Extrema
We take
C := φ(MT ),
where
MT = max St
t∈[0,T ]
Figure 8.1 represents the running maximum process (Mt )t∈R+ of Brownian
motion (Bt )t∈R+ .
152 "
3
Xt
Bt
2.5
1.5
Bt , X t
0.5
-0.5
-1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 2.0
t
Barrier Options
(K − ST )+ if max St < B,
0≤t≤T
+
C = (K − ST ) 1( ) =
max St < B 0
if max St ≥ B.
0≤t≤T 0≤t≤T
This option is also called a Callable Bear Contract with no residual value, in
which the call price B usually satisfies B ≤ K.
This option is also called a Callable Bull Contract with no residual value, in
which B denotes the call price B ≥ K. It is also called a turbo warrant with
no rebate.
" 153
Lookback Options
The payoff of a floating strike lookback call option on the underlying asset
St with exercise date T is
C = ST − min St .
0≤t≤T
The payoff of a floating strike lookback put option on the underlying asset
St with exercise date T is
C = max St − ST .
0≤t≤T
Options on Average
where
1 wT
St dt
T 0
represents the average of (St )t∈R+ over the time interval [0, T ] and φ : R −→ R
is a payoff function.
3
Xt
Bt
2.5
1.5
Bt , X t
0.5
-0.5
-1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 2.0
t
154 "
Figure 8.2 shows a graph of Brownian motion and its moving average process
Xt .
Asian Options
Asian options are particular cases of options on average, and they were first
traded in Tokyo in 1987. The payoff of the Asian call option on the underlying
asset St with exercise date T and strike K is given by
1 wT
+
C= St dt − K .
T 0
Similarly, the payoff of the Asian put option on the underlying asset St with
exercise date T and strike K is
1 wT
+
C= K− St dt .
T 0
Due to the fact that their dependence on averaged asset prices, Asian op-
tions are less volatile than plain vanilla options whose payoffs depend only
on the terminal value of the underlying asset. Asian options have become
particularly popular in commodities trading.
In order to price barrier options we will have to derive the probability density
of the maximum
MT = max St
t∈[0,T ]
of geometric Brownian motion (St )t∈R+ over a given time interval [0, T ].
where f(MT ,ST ) is the joint probability density function of (MT , ST ), which
satisfies
wx wy
P(MT ≤ x, ST ≤ y) = f(MT ,ST ) (u, v)dudv, x, y ∈ R.
−∞ −∞
" 155
XT = max Bt
t∈[0,T ]
might seem a difficult problem. However this is not the case, due to the
reflection principle. Note that since B0 = 0 we have
XT ≥ 0,
almost surely.
τa = inf{t ∈ R+ : Bt = a}
denote the first time (Bt )t∈R+ hits the level a > 0.
2P(BT > a & τa < T ) = P(τa < T ) = 2P(BT < a & τa < T ),
and to
156 "
Figure 8.3 shows a graph of Brownian motion and its reflected path.
2.5
1.5
1
Bt
0.5
-0.5
-1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 2.0
t
P(XT ≤ a) = P(|BT | ≤ a)
1 w a −x2 /(2T )
= √ e dx
2πT −a
2 w a 2
= √ e−x /(2T ) dx, a ∈ R+ ,
2πT 0
and probability density
r
dP(XT ≤ a) 2 −a2 /(2T )
fXT (a) = = e 1[0,∞) (a), a ∈ R. (8.2)
da πT
" 157
1
Density function
0.8
0.6
density
0.4
0.2
0
-4 -3 -2 -1 0 1 2 3 4
x
Fig. 8.4: Probability density of the maximum of Brownian motion over [0, 1].
2 w∞
r
2
= e−rT φ(x)e−|x| /(2T ) dx.
πT 0
Next we consider
MT = max St
t∈[0,T ]
= S0 max eσBt
t∈[0,T ]
= S0 eσ maxt∈[0,T ] Bt
= S0 eσXT ,
C = φ(MT ),
where
ST = S0 eσBT ,
we have
C = φ(MT ) = φ(S0 eσXT ),
hence
158 "
2 −rT w ∞
r
2
= e φ(S0 eσx )e−x /(2T ) dx.
πT 0
This however is not sufficient since this imposes the condition r = σ 2 /2. In
order to do away with this condition we need to consider the maximum of
drifted Brownian motion, and for this we have to compute the joint density
of XT and BT .
Joint Density
The reflection principle also allows us to compute the joint density of Brow-
nian motion BT and its maximum XT . Indeed, for b ∈ [0, a] we also have
i.e.
P(BT > 2a − b & τa < T ) = P(BT < b & τa < T ),
or, by (8.1),
hence, since 2a − b ≥ a,
" 159
r
2 (2a − b) −(2a−b)2 /(2T )
fXT ,BT (a, b) = e 1{a≥b∨0} (8.4)
πT T
r
2 (2a − b) −(2a−b)2 /(2T )
a > b ∨ 0,
e ,
= πT T
0, a < b ∨ 0.
Density function
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0-1
-0.5
0
0.5 -0.5 -1
1 0
b 1.5 0.5
2 1.5 1
2.5 2 a
3 3 2.5
Fig. 8.5: Joint probability density of B1 and its maximum over [0,1].
On the other hand, B̃t is a standard Brownian motion under the proba-
bility measure P̃ defined from
160 "
dP̃ 2
= e−µBT −µ T /2 , (8.5)
dP
2 wawb
r
2 (2x − y) −(2x−y)2 /(2T )
= 1(−∞,x] (y)eµy−µ T /2 e dxdy,
πT 0 −∞ T
0 ≤ b ≤ a, which yields the joint probability density
r
1 2 2 2
fX̃T ,B̃T (a, b) = 1{a≥b∨0} (2a − b)eµb−(2a−b) /(2T )−µ T /2 (8.6)
T πT
r
1 2 2 2
(2a − b)e−µ T /2+µb−(2a−b) /(2T ) , a > b ∨ 0,
= T πT
0, a < b ∨ 0.
We also find
2 waw∞
r
2 (2x − y) −(2x−y)2 /(2T )
P(X̃T ≤ a) = 1(−∞,x] (y)eµy−µ T /2 e dydx
πT 0 −∞ T
r
2 −µ2 T /2 w a w a (2x − y) −(2x−y)2 /(2T )
= e eµy e dxdy
πT −∞ y∨0 T
1 −µ2 T /2 w a
r
2 2
= e eµy−(2(y∨0)−y) /(2T ) − eµy−(2a−y) /(2T ) dy
2πT −∞
1 w a −(y−µT )2 /(2T )
r
2
= e − e−(y−(µT +2a)) /(2T )+2aµ dy
2πT −∞
1 w a −(y−µT )2 /(2T ) 1 w a −(y−(µT +2a))2 /(2T )
r r
= e dy − e2aµ e dy
2πT −∞ 2πT −∞
1 w a−µT −y2 /(2T ) 1 w −a−µT −y2 /(2T )
r r
= e dy − e2aµ e dy
2πT −∞ 2πT −∞
a − µT −a − µT
=Φ √ − e2µa Φ √ , (8.7)
T T
cf. Corollary 7.2.2 and pages 297-299 of [71] for another derivation. This
yields the density
r
dP(X̃T ≤ a) −a − µT
2 −(a−µT )2 /(2T )
= e − 2µe2µa Φ √ ,
da πT T
of the supremum of drifted Brownian motion, and recovers (8.2) for µ = 0.
µ=0
1.4 µ=-0.5
µ=0.5
1.2
1
density
0.8
0.6
0.4
0.2
0
-1 0 1 2 3 4
x
Note from Figure 8.2 that small values of the maximum are more likely to
occur when µ takes large negative values.
Based on the relation min B̃t = − max (−B̃t ), the joint density fR̃T ,B̃T
t∈[0,T ] t∈[0,T ]
of the minimum
R̃T = min B̃t = min (Bt + µt)
t∈[0,T ] t∈[0,T ]
of the drifted Brownian motion B̃t := Bt + µt and its value B̃T at time T
can similarly be computed as follows, letting a ∧ b := min(a, b):
162 "
r
1 2 2 2
fR̃T ,B̃T (a, b) = 1{a≤b∧0} (b − 2a)eµb−(2a−b) /(2T )−µ T /2 (8.8)
T πT
r
1 2 2 2
(b − 2a)e−µ T /2+µb−(2a−b) /(2T ) , a < b ∧ 0,
= T πT
0, a > b ∧ 0.
General Case
Using the joint density of B̃T and X̃T we are able to price any exotic option
with payoff φ(B̃T , X̃T ), as
" #
e−r(T −t) IE φ(X̃T , B̃T ) Ft ,
with in particular
h i w∞ w∞
e−rT IE φ(X̃T , B̃T ) = e−rT φ(x, y)dP(X̃T = x, B̃T = y).
−∞ y∨0
C = φ(MT , ST ),
where 2
ST = S0 eσBT −σ T /2+rT
= S0 eσB̃T ,
with µ = −σ/2 + r/σ and B̃T = BT + µT , and
MT = max St
t∈[0,T ]
2
= S0 max eσBt −σ t/2+rt
t∈[0,T ]
= S0 max eσB̃t
t∈[0,T ]
= S0 eσ maxt∈[0,T ] B̃t
= S0 eσX̃T ,
" 163
we have
C = φ(ST , MT )
2
= φ(S0 eσBT −σ T /2+rT
, MT )
= φ(S0 eσB̃T , S0 eσX̃T ),
hence
h i
e−rT IE[C] = e−rT IE φ(S0 eσB̃T , S0 eσX̃T )
w∞ w∞
= e−rT φ(S0 eσy , S0 eσx )dP(X̃T = x, B̃T = y)
−∞ y∨0
2 −rT w ∞ w ∞
r
1 2 2
= e φ(S0 eσy , S0 eσx )(2x − y)e−µ T /2+µy−(2x−y) /(2T ) dxdy
T πT −∞ y∨0
2 w∞w∞
r
1 2 2
= e−rT φ(S0 eσy , S0 eσx )(2x − y)e−µ T /2+µy−(2x−y) /(2T ) dxdy
T πT 0 y
2 w0 w∞
r
1 2 2
+ e−rT φ(S0 eσy , S0 eσx )(2x − y)e−µ T /2+µy−(2x−y) /(2T ) dxdy.
T πT −∞ 0
We can distinguish 8 different versions of barrier options according to the
following table.
164 "
+
down-and-out (ST − K) 1( )
min St > B
0≤t≤T
+
down-and-in (ST − K) 1( )
min St < B
0≤t≤T
+
up-and-in (ST − K) 1( )
max St > B
0≤t≤T
+
down-and-out (K − ST ) 1( )
min St > B
0≤t≤T
+
down-and-in (K − ST ) 1( )
min St < B
0≤t≤T
+
up-and-in (K − ST ) 1( )
max St > B
0≤t≤T
We have the following obvious relations between the prices of barrier and
vanilla call and put options:
Cup−in (t) + Cup−out (t) = C(t) = e−r(T −t) IE∗ [(ST − K)+ ],
Cdown−in (t) + Cdown−out (t) = C(t) = e−r(T −t) IE∗ [(ST − K)+ ],
" 165
where C(t), resp. P (t) denotes the price of a European call, resp. put option
with strike K as obtained from the Black-Scholes formula. Consequently, in
the sequel we will only compute the prices of the up-and-out call and put,
and down-and-out barrier call and put options.
S −K if max St ≤ B,
T
0≤t≤T
+
C = (ST − K) 1( ) =
max St ≤ B 0
if max St > B,
0≤t≤T 0≤t≤T
of the up-and-out call option with maturity T , strike K and barrier B is given
by
where
166 "
τ 1 1
δ± (s) = √ log s + r ± σ 2 τ , s > 0. (8.10)
σ τ 2
Note that taking B = +∞ in the above identity (8.9) recovers the Black-
Scholes formula for the price of a European call option, and that the price of
the up-and-out barrier call option is 0 when B < K.
The following graph represents the up-and-out call option price given the
value St of the underlying and the time t ∈ [0, T ] with T = 220 days.
16
14
12
10
8
6
4
2
0
220
200
180
160 Time in days
140
50 55 60 65 70 120
75 80 100
85 90
underlying
hence
" #
+
e−r(T −t) IE∗ [C | Ft ] = e−r(T −t) IE∗ (ST − K) 1{MT ≤B} Ft
" #
−r(T −t) ∗ +
=e IE (ST − K) 1{MT ≤ B } Ft
+
= e−r(T −t) IE∗ (ST − K) 1{Mt ≤ B } 1( Ft
)
max Sr ≤ B
t≤r≤T
" 167
−r(T −t) ∗ +
=e 1{Mt ≤ B } IE (ST − K) 1( Ft
)
max Sr ≤ B
t≤r≤T
+
−r(T −t) ∗ ST
=e 1{Mt ≤ B } IE x −K 1( )
St
x max Sr /St ≤ B
t≤r≤T
x=St
+
ST −t
= e−r(T −t) 1{Mt ≤ B } IE∗ x −K 1( ) .
S0
x max Sr /S0 ≤ B
0≤r≤T −t
x=St
It suffices to compute
h i
+
e−rτ IE∗ [C] = e−rτ IE∗ (Sτ − K) 1{Mτ ≤B}
+
= e−rτ IE∗ S0 eσB̃τ − K 1{S0 eσX̃τ ≤B}
w∞ w∞
+
= e−rτ (S0 eσy − K) 1{S0 eσx ≤B} dP(X̃τ = x, B̃τ = y)
−∞ y∨0
2 w σ−1 log(B/S0 )
r
1
= e−rτ
τ πτ 0
w∞ 2 2
+
(S0 eσy − K) 1{S0 eσx ≤B} (2x − y)e−µ τ /2+µy−(2x−y) /(2τ ) dxdy
y
2 w0
r
1
+ e−rτ
τ πτ −∞
w∞ 2 2
+
(S0 eσy − K) 1{S0 eσx ≤B} (2x − y)e−µ τ /2+µy−(2x−y) /(2τ ) dxdy
0
2 w σ−1 log(B/S0 )
r
1
= e−rτ
τ πτ 0
w∞ 2 2
+
(S0 eσy − K) 1{x≤σ−1 log(B/S0 )} (2x − y)e−µ τ /2+µy−(2x−y) /(2τ ) dxdy
y
2 w0
r
1
+ e−rτ
τ πτ −∞
w∞ 2 2
+
(S0 eσy − K) 1{x≤σ−1 log(B/S0 )} (2x − y)e−µ τ /2+µy−(2x−y) /(2τ ) dxdy
0
2 w σ−1 log(B/S0 )
r
1
= e−rτ
τ πτ 0
168 "
w σ−1 log(B/S0 ) 2
+ τ /2+µy−(2x−y)2 /(2τ )
(S0 eσy − K) (2x − y)e−µ dxdy
y
2 w0
r
1
+ e−rτ
τ πτ −∞
w σ−1 log(B/S0 ) 2
+ τ /2+µy−(2x−y)2 /(2τ )
(S0 eσy − K) (2x − y)e−µ dxdy
0
2 w σ−1 log(B/S0 )
r
1 −rτ
= e
τ πτ σ−1 log(K/S0 )
w σ−1 log(B/S0 ) 2 2
(S0 eσy − K) (2x − y)e−µ τ /2+µy−(2x−y) /(2τ ) dxdy
y∨0
2 w σ−1 log(B/S0 )
r
1 2 2
= e−rτ −µ τ /2 (S0 eσy − K) eµy−y /(2τ )
τ πτ σ−1 log(K/S0 )
w σ−1 log(B/S0 )
(2x − y)e2x(y−x)/τ dxdy,
y∨0
" 169
2 1 w b µy−y2 /(2τ )
−Ke−τ (r+µ /2)
√e dy
2πτ c
2 2 1 w b 2
+Ke−τ (r+µ /2)−2b /τ √ ey(µ+2b/τ )−y /(2τ ) dy.
2πτ c
Using the relation
1 w b γy−y2 /(2τ ) −c + γτ −b + γτ
2
√ e dy = eγ τ /2 Φ √ −Φ √ ,
2πτ c τ τ
we find
h i
+
e−rτ IE∗ [C] = e−rT IE∗ (ST − K) 1{MT ≤B}
−c + (σ + µ)τ −b + (σ + µ)τ
2 2
= S0 e−τ (r+µ /2)+(σ+µ) τ /2 Φ √ −Φ √
τ τ
2 2 2
−S0 e−τ (r+µ /2)−2b /τ +(σ+µ+2b/τ ) τ /2
−c + (σ + µ + 2b/τ )τ −b + (σ + µ + 2b/τ )τ
× Φ √ −Φ √
τ τ
−c −b
+ µτ + µτ
−Ke−rτ Φ √ −Φ √
τ τ
2 2 2
+Ke−τ (r+µ /2)−2b /τ +(µ+2b/τ ) τ /2
−c + (µ + 2b/τ )τ −b + (µ + 2b/τ )τ
× Φ √ −Φ √
τ τ
τ S0 τ S0
= S0 Φ δ+ − Φ δ+
K B
2
−τ (r+µ2 /2)−2b2 /τ +(σ+µ+2b/τ )2 τ /2 τ B τ B
−S0 e Φ δ+ − Φ δ+
KS0 S0
S0 S0
−Ke−rτ Φ δ− τ
− Φ δ− τ
K B
2
−τ (r+µ2 /2)−2b2 /τ +(µ+2b/τ )2 τ /2 B B
+Ke Φ δ− − Φ δ− ,
KS0 S0
τ
0 ≤ x ≤ B, where δ± (s) is defined in (8.10). Given the relations
and
this yields
h i
+
e−rτ IE∗ [C] = e−rτ IE∗ (Sτ − K) 1{Mτ ≤B} (8.11)
170 "
τ S0 τ S0
= S0 Φ δ+ − Φ δ+
K B
−rτ τ S0 τ S0
−e K Φ δ− − Φ δ−
K B
2r/σ2 2
B τ B τ B
−B Φ δ+ − Φ δ+
S0 KS0 S0
1−2r/σ2 2
S0 B B
+e−rτ K Φ δ− τ
− Φ δ− τ
B KS0 S0
τ S0 τ S0
= S0 Φ δ+ − Φ δ+
K B
1+2r/σ2 2
B τ B τ B
−S0 Φ δ+ − Φ δ+
S0 KS0 S0
S0 S0
−e−rτ K Φ δ− τ
− Φ δ− τ
K B
1−2r/σ2 2
S0 B B
−e−rτ K Φ δ− τ
− Φ δ− τ
,
B KS0 S0
and this yields the result of Proposition 8.1, cf. § 7.3.3 pages 304-307 of [71]
for a different calculation. This concludes the proof of Proposition 8.1.
The price
+
−r(T −t) ∗ ST −t
e 1{Mt ≤ B } IE K − x 1(
)
S0
x max Sr /S0 ≤ B
0≤r≤T −t
x=St
" 171
1−2r/σ2 2 !
St T −t B
− Φ δ− −1
B KSt
1+2r/σ2 2 !
T −t St B T −t B
= St 1{Mt ≤ B } −Φ −δ+ + Φ −δ+
K St KSt
−Ke−r(T −t)
1−2r/σ2 2 !
T −t St St T −t B
×1{Mt ≤ B } −Φ −δ− + Φ −δ− ,
K B KSt
if B > K, and
St
e−r(T −t) IE∗ [P | Ft ] = St 1{Mt ≤ B } Φ δ+ T −t
−1
B
1+2r/σ2 !
B T −t B
− Φ δ+ −1
St St
T −t St
−e−r(T −t) K1{Mt ≤ B } Φ δ− −1
B
1−2r/σ 2 !
St T −t B
− Φ δ− −1
B St
1+2r/σ2 !
T −t St B T −t B
= St 1{Mt ≤ B } −Φ −δ+ + Φ −δ+
B St St
−Ke−r(T −t)
1−2r/σ2 !
T −t St St T −t B
×1{Mt ≤ B } −Φ −δ− + Φ −δ− ,
B B St
(8.12)
if B < K.
172 "
12
10
4 220
200
2 180
160 Time in days
0 50 140
55 60 65 120
70 75 80 85 90 100
-2 underlying
Fig. 8.8: Graph of the up-and-out put option price with B > K.
50
45
40
35
30
25
20
15 220
10 200
180
5 160 Time in days
0 50 140
55 60 120
65 70 75 80
underlying
85 90 100
Fig. 8.9: Graph of the up-and-out put option price with K > B.
S −K if min St > B,
T
0≤t≤T
+
C = (ST − K) 1( ) =
min St > B 0
if min St ≤ B,
0≤t≤T 0≤t≤T
" 173
We have
St > B, 0 ≤ t ≤ T , and
+
IE∗ (ST − K) 1( Ft = 1{mint∈[0,T ] St >B} g(t, St ),
)
min St > B
0≤t≤T
174 "
1.6
1.4
1.2
0.8
0.6
140
0.4 135
130
125
0.2 120 Time in days
115
0 50 110
55 60 105
65 70 75 80 85 90 100
underlying
Fig. 8.10: Graph of the down-and-out call option price with B < K.
60
50
40
30
20
220
200
10 180
160 Time in days
0 50 140
55 60 120
65 70 75 80
underlying
85 90 100
Fig. 8.11: Graph of the down-and-out call option price with K > B.
" 175
14
12
10
6
220
4
200
2 180
160 Time in days
0 50 140
55 60 120
65 70
-2 75 80
underlying
85 90 100
Fig. 8.12: Graph of the down-and-out put option price with K > B.
Note that although Figures 8.8 and 8.10, resp. 8.9 and 8.11, appear to share
some symmetry property, the functions themselves are not exactly symmetric.
Concerning 8.7 and 8.12 the pricing function is actually the same, but the
conditions B < K and B > K play opposite roles.
PDE Method
176 "
= g(t, St , Mt ),
Vt = g(t, St , Mt ), t ∈ R+ .
∂g ∂g 1 ∂2g
rg(t, x, y) = (t, x, y) + ry (t, x, y) + x2 σ 2 2 (t, x, y), (8.18)
∂t ∂x 2 ∂x
t > 0, x > 0, 0 < y < B, and ξt is given by
∂g
ξt = (t, St , Mt ), t ∈ [0, T ], (8.19)
∂x
provided Mt < B.
Proof. By (8.17) the price at time t of the down-and-out call barrier option
discounted to time 0 is given by
+
e−rt g(t, St , Mt ) = 1{Mt ≤B} e−rT IE∗ (ST − K) 1( St = x
)
max Sr ≤ B
t≤r≤T
−rT ∗ +
=e IE (ST − K) 1{Mt ≤B} 1( St = x
)
max Sr ≤ B
t≤r≤T
" 177
−rT ∗ +
=e IE (ST − K) 1( St = x ,
)
max Sr ≤ B
0≤r≤T
S −K if max St ≤ B,
T
0≤t≤T
+
C = (ST − K) 1( ) =
max St ≤ B 0
if max St > B,
0≤t≤T 0≤t≤T
with B > K.
Proposition 8.3. The price of any self-financing portfolio of the form Vt =
g(t, St ) hedging the up-and-out barrier call option satisfies the Black-Scholes
PDE
∂g ∂g 1 ∂2g
rg(t, x) =
(t, x) + rx (t, x) + x2 σ 2 2 (t, x),
∂t ∂x 2 ∂x
g(t, x) = 0, x ≥ B, t ∈ [0, T ],
g(T, x) = (x − K)+ 1{x<B} ,
g(t, B) = 0. (8.20)
Condition (8.20) holds since the price of the claim at time t is 0 whenever
St = B, cf. e.g. [23].
178 "
x x
T −t T −t
g(t, x) = x Φ δ+ − Φ δ+ (8.21)
K B
x 1−2r/σ2 2
B
B
T −t T −t
−x Φ δ+ − Φ δ+
B Kx x
x x
T −t T −t
−Ke−r(T −t) Φ δ− − Φ δ−
K B
x 1−2r/σ2 2
B
B
T −t T −t
+Ke−r(T −t) Φ δ− − Φ δ− ,
B Kx x
0 < x ≤ B, 0 ≤ t ≤ T .
We note that the expression (8.21) can be rewritten using the standard
Black-Scholes formula
S S
τ
BSc (S, K, r, σ, τ ) = SΦ δ+ − Ke−rτ Φ δ− τ
K K
0 < x ≤ B, 0 ≤ t ≤ T .
Figure 8.13 represents the value of Delta obtained from (8.19) for the
up-and-out call option, cf. Exercise 8.3-(1).
" 179
1
0.8
0.6 100
0.4 120
0.2
140
0
160
Time in days
180
90 200
85 80 75 70 220
65 60 55 50
underlying
g(t, x) = 0, x > B.
180 "
+∞ if s > 1,
0
δ± (s) = −∞ × 1{s<1} + ∞ × 1{s>1} = 0 if s = 1,
−∞ if s < 1,
Similarly the price g(t, St ) at time t of the down-and-out barrier call option
satisfies the Black-Scholes PDE
" 181
∂g ∂g 1 ∂2g
rg(t, x) = (t, x) + rx (t, x) + x2 σ 2 2 (t, x),
∂t ∂x 2 ∂x
g(t, B) = 0, t ∈ [0, T ],
g(T, x) = (x − K)+ 1{x≥B} ,
Let
mts = inf Su
u∈[s,t]
and
Mst = sup Su ,
u∈[s,t]
0 ≤ s ≤ t ≤ T , and let Mts be either mts or Mst . In the lookback option case
the payoff φ(ST , MT0 ) depends not only on the price of the underlying asset
at maturity but it also depends on all price values of the underlying asset
over the period which starts from the initial time and ends at maturity.
The payoff of such of an option is of the form φ(ST , MT0 ) with φ(x, y) =
x − y in the case of lookback call options, and φ(x, y) = y − x in the case of
lookback put options. We let
The standard lookback put option gives its holder the right to sell the un-
derlying asset at its historically highest price. In this case the strike is M0T
and the payoff is
C = M0T − ST .
Our goal is to prove the following pricing formula for lookback put options.
Proposition 8.4. The price at time t ∈ [0, T ] of the lookback put option with
payoff M0T − ST is given by
182 "
Figure 8.14 represents the lookback put price as a function of St and M0t , for
different values of the time to maturity T − t.
80
60
40
20
0
80
60
Mt 40
20
0 0 40 60 80
20
St
and
Next, we have
" 183
!
M0t MT
P(M0t > MtT | Ft ) = P > t Ft
St St
!
MT
=P x> t Ft
St
x=M0t /St
!
M0T −t
=P <x .
S0
x=M0t /St
Hence
!
M0T −t
P(M0t > MtT ) =P <x
S0
x=M0t /St
−1+2r/σ2
M0t
t
T St T M0
= Φ −δ− − Φ −δ− .
M0t St St
Next, we have
" #
MtT
IE∗ [MtT 1{MtT >M0t } | Ft ] = St IE∗ 1 T t Ft
St {Mt /St >M0 /St }
" #
Sr
= St IE∗ max 1{maxr∈[t,T ] Sr /St >x} Ft
r∈[t,T ] St
x=M0t /St
Sr
= St IE∗ max 1{maxr∈[0,T −t] Sr /S0 >x} ,
r∈[0,T −t] S0 x=M0t /St
and
Sr
IE∗ max 1{maxr∈[0,τ ] Sr /S0 >x}
r∈[0,τ ] S0
= IE∗ max eσB̃r 1{maxr∈[0,τ ] eσB̃r >x}
r∈[0,τ ]
h i
∗
= IE eσ maxr∈[0,τ ] B̃r 1{maxr∈[0,τ ] B̃r >σ−1 log x}
184 "
h i
= IE∗ eσX̃τ 1{X̃τ >σ−1 log x}
w∞
= −1 eσx fX̃τ (z)dz
σ log x
w∞ r
2 −(z−µτ )2 /(2τ )
−z − µτ
!
= −1 eσz e − 2µe2µz Φ √ dz
σ log x πτ τ
2 w∞ w∞
r
−z − µτ
2
= eσz−(z−µτ ) /(2τ ) dz − 2µ −1 ez(σ+2µ) Φ √ dz.
πτ σ −1 log x σ log x τ
−1
with a = σ log x. We let
−z − µτ
u(z) = Φ √ and v 0 (z) = ez(σ+2µ)
τ
which satisfy
1 2 1
u0 (z) = − √ e−(z+µτ ) /(2τ ) and v(z) = ez(σ+2µ) ,
2πτ σ + 2µ
and
w∞
−z − µτ
w∞
ez(σ+2µ) Φ √ dz = v 0 (z)u(z)dz
a τ a
" 185
w∞
= u(+∞)v(+∞) − u(a)v(a) − v(z)u0 (z)dz
a
−a − µτ
1
=− ea(σ+2µ) Φ √
σ + 2µ τ
1 w ∞
z(σ+2µ) −(z+µτ )2 /(2τ )
+ √ e e dz
(σ + 2µ) 2πτ a
−a − µτ
1
=− ea(σ+2µ) Φ √
σ + 2µ τ
1 w∞
(τ (σ+µ)2 −µ2 τ )/2 2
+ √ e e−(z−τ (σ+µ)) /(2τ ) dz
(σ + µ) 2πτ a
−a − µτ
1
=− ea(σ+2µ) Φ √
σ + 2µ τ
1 w∞
(τ (σ+µ)2 −µ2 τ )/2 −z 2 /2
+ √ e √ e dz
(σ + 2µ) 2π (a−τ (σ+µ))/ τ
−a − µτ
1
=− ea(σ+2µ) Φ √
σ + 2µ τ
−a + τ (σ + µ)
1 2 2
+ e(τ (σ+µ) −µ τ )/2 Φ √
σ + 2µ τ
−(r/σ − σ/2)τ − σ −1 log x
2r 2r/σ 2
= − (x) Φ √
σ τ
+τ (r/σ + σ/2) − σ −1 log x
2r στ (σ+2µ)/2
+ e Φ √
σ τ
σ rτ τ 1 σ 2r/σ2 τ
= e Φ δ+ − x Φ −δ− (x) ,
2r x 2r
cf. pages 317-319 of [71] for a different derivation using double integrals.
Hence we have
" #
∗ Sr
IE MtT 1{MtT >M0t } Ft = St IE∗ max 1{maxr∈[0,T −t] Sr /S0 >x}
r∈[0,T −t] S0 x=M0t /St
T −t St µσ T −t S t
= 2St er(T −t) Φ δ+ − St er(T −t) Φ δ+
M0t r M0t
2r/σ2
µσ M0t
t
T −t M0
+St Φ −δ− ,
r St St
186 "
t 2r/σ2 t
T −t St M0 T −t M0
= M0t Φ −δ− t − St Φ −δ−
M0 St St
r(T −t) T −t St
+2St e Φ δ+
M0t
2
σ T −t St
−St 1 − er(T −t) Φ δ+
2r M0t
2
t 2r/σ2 t
σ M0 T −t M0
+St 1 − Φ −δ−
2r St St
σ2
t T −t St r(T −t) T −t St
= M0 Φ −δ− + St e 1 + Φ δ +
M0t 2r M0t
2
t 2r/σ2 t
σ M0 T −t M0
−St Φ −δ− ,
2r St St
hence
e−r(T −t) IE∗ [M0T − ST | Ft ] = e−r(T −t) IE∗ [M0T | Ft ] − e−r(T −t) IE∗ [ST | Ft ]
= e−r(T −t) IE∗ [M0T | M0t ] − St
T −t St T −t St
= M0t e−r(T −t) Φ −δ− t − St Φ −δ+ t
M0 M0
2
2
t 2r/σ2 t
σ T −t St σ −r(T −t) M0 T −t M0
+St Φ δ+ − S t e Φ −δ− .
2r M0t 2r St St
PDE Method
" 187
Vt = f (t, St , M0t ), t ∈ R+ ,
2 2
for some f ∈ C ((0, ∞) × (0, ∞) ).
Then the function f (t, x, y) satisfies the Black-Scholes PDE
∂f ∂f 1 ∂2f
rf (t, x, y) = (t, x, y) + rx (t, x, y) + x2 σ 2 2 (t, x, y), t, x, y > 0,
∂t ∂x 2 ∂x
(8.22)
under the boundary conditions
−r(T −t)
f (t, 0, y) = e
y, 0 ≤ t ≤ T, y ∈ R+ , (8.23a)
∂f
(t, x, y)x=y = 0, 0 ≤ t ≤ T, y > 0, (8.23b)
∂y
f (T, x, y) = y − x, 0 ≤ x ≤ y. (8.23c)
from the time homogeneity of the asset price process (St )t∈R+ . Applying the
change of variable formula to the discounted portfolio value
188 "
∂f ∂f
= −re−rt f (t, St , M0t )dt + e−rt (t, St , M0t )dt + re−rt St (t, St , M0t )dt
∂t ∂x
1 ∂2f
+ e−rt σ 2 St2 2 (t, St , M0t )dt
2 ∂x
∂f ∂f
+e−rt (t, St , M0t )dM0t + e−rt σSt (t, St , M0t )dBt .
∂y ∂x
Since IE∗ [φ(ST , M0T ) | Ft ] t∈[0,T ] is a P-martingale and (M0t )t∈[0,T ] has finite
∂f
df (t, St , M0t ) = σSt (t, St , M0t )dBt , t ∈ [0, T ], (8.26)
∂x
and the function f (t, x, y) satisfies the equation
∂f ∂f
(t, St , M0t )dt + rSt f (t, St , M0t )dt
∂t ∂x
1 2 2 ∂2f ∂f
+ σ St 2 (t, St , M0t )dt + (t, St , M0t )dM0t = rf (t, St , M0t )dt,
2 ∂x ∂y
which implies
∂f ∂f 1 ∂2f
(t, St , M0t ) + rSt (t, St , M0t ) + σ 2 St2 2 (t, St , M0t ) = rf (t, St , M0t ),
∂t ∂x 2 ∂x
which is (8.22), and
∂f
(t, St , M0t )dM0t = 0,
∂y
because M0t increases only on a set of zero measure (which has no isolated
points). This implies
∂f
(t, St , St ) = 0,
∂y
which shows the boundary condition (8.23b), since M0t hits St when M0t
increases. On the other hand, (8.26) shows that
wT ∂f
φ(ST , M0T ) = IE∗ [φ(ST , M0T )] + σ St (t, x, M0t )|x=St dBt ,
0 ∂x
0 ≤ t ≤ T , which implies (8.24) as in the proof of Proposition 5.2.
In other words, the price of the lookback put option takes the form
" #
f (t, St , Mt ) = e−r(T −t) IE∗ M0T − ST Ft ,
" 189
σ2
T −t T −t
f (t, x, y) = ye−r(T −t) Φ −δ−
(x/y) + x 1 + Φ δ+ (x/y)
2r
σ2 2r/σ 2 T −t
−x e−r(T −t) (y/x)
Φ −δ− (y/x) − x.
2r
τ > 0, hence
∂f
(t, x, x) = C(T − t), t ∈ [0, T ],
∂x
while we also have
∂f
(t, x, y)y=x = 0, 0 ≤ x ≤ y.
∂y
190 "
" #
f (t, x, y) = e−r(T −t) IE∗ M0T − ST St = x, Mt = y
" #
−r(T −t) ∗
=e IE M0t ∨ MtT − ST St = x, Mt = y
" #
M0t MtT
= e−r(T −t) x IE∗ ∨ −1 St = x, Mt = y
St St
" #
y MtT
= e−r(T −t) x IE∗ ∨ −1 St = x, Mt = y
x St
" #
y
= e−r(T −t) x IE∗ M0t ∨ MtT − 1 St = 1, Mt =
x
= xf (t, 1, y/x)
= xg(T − t, x/y),
where we let
σ2
1 −rτ τ
τ
g(τ, z) := e Φ −δ− (z) + 1 + Φ δ+ (z)
z 2r
2r/σ2
σ2 1
− e−rτ τ
Φ(−δ− (1/z)) − 1,
2r z
∂g
(τ, 1) = 0, τ > 0, (8.27a)
∂z
g(0, z) = 1 − 1,
z ∈ (0, 1]. (8.27b)
z
The next Figure 8.15 shows a graph of the function g(τ, z).
" 191
1.2
1
0.8
0.6
0.4
0.2
0 1
0.6 0.8
0.7 0.6
z 0.8 0.4 τ
0.9 0.2
1 0
Letting
S S
BSp (S, K, r, σ, τ ) = Ke−rτ Φ −δ−
τ τ
− SΦ −δ+
K K
denote the standard Black-Scholes formula for the price of a European put
option, we observe that the lookback put option price satisfies
i.e.
" #
St
e−r(T −t) IE∗ M0T − ST Ft = BSp (St , M0t , r, σ, T − t) + St hp T − t, t
M0
192 "
for the standard Black-Scholes put formula, we observe that f (t, x, y) satisfies
i.e.
f (t, x, y) = xg(T − t, x/y),
with
g(τ, z) = BSp (1, 1/z, r, σ, τ ) + hp (τ, z), (8.29)
where hp (τ, z) is the function given by (8.28), and (x, y) 7→ xhp (T − t, x/y)
also satisfies the Black-Scholes PDE (8.22), i.e. (τ, z) 7→ BSp (1, 1/z, r, σ, τ )
and hp (τ, z) both satisfy the PDE
∂hp ∂hp 1 ∂ 2 hp
(τ, z) = z r + σ 2 (τ, z) + σ 2 z 2 (τ, z), (8.30)
∂τ ∂z 2 ∂z 2
τ ∈ R+ , z ∈ [0, 1], under the boundary condition
hp (0, z) = 0, 0 ≤ z ≤ 1.
The next Figures 8.16 and 8.17 show the decompositions (8.29) of the normal-
ized lookback put option price g(τ, z) in Figure 8.15 into the Black-Scholes
put function BSp (1, 1/z, r, σ, τ ) and hp (τ, z).
1.2
1
0.8
0.6
0.4
0.2
0 1
0.6 0.8
0.7 0.6
z 0.8 0.4 τ
0.9 0.2
1 0
" 193
h(τ,x)
1.2
1
0.8
0.6
0.4
0.2
0 1
0.6 0.8
0.7 0.6
z 0.8 0.4 τ
0.9 0.2
1 0
Note that in Figures 8.16-8.17 the condition hp (0, z) = 0 is not fully respected
as z → 1 due to numerical error in the approximation of the function Φ.
The standard Lookback call option gives the right to buy the underlying asset
at its historically lowest price. In this case the strike is mT0 and the payoff is
C = ST − mT0 .
The following result gives the price of the lookback call option, cf. e.g. Propo-
sition 9.5.1, page 270 of [15].
Proposition 8.6. The price at time t ∈ [0, T ] of the lookback call option
with payoff ST − mT0 is given by
Figure 8.18 represents the price of the lookback call option as a function of
mt0 and St for different values of the time to maturity T − t.
194 "
90
80
70
60
50
40
30
20 80
10 60
0
80 40 St
60
40 20
mt 20
0 0
e−r(T −t) IE∗ [ST − mT0 | Ft ] = e−r(T −t) IE∗ [ST | Ft ] − e−r(T −t) IE∗ [mT0 | Ft ],
and
and
" 195
Sr
IE∗ [mTt 1{mt0 >mTt } | Ft ] = St IE∗ min 1{mt0 /St >mTt /St }
r∈[t,T ] St x=mt0 ,y=St
Sr
= St IE∗ min 1{minr∈[t,T ] Sr /St <x}
r∈[t,T ] St x=mt0 /St
Sr
= St IE∗ min 1{minr∈[0,T −t] Sr /S0 <x}
r∈[0,T −t] S0 x=mt0 /St
T −t St µσ T −t St
= 2St er(T −t) Φ −δ+ t − St er(T −t) Φ −δ+ t
m0 r m0
2r/σ2
µσ mt0
t
T −t m 0
+St Φ δ− .
r St St
Letting
196 "
S S
τ
BSc (S, K, r, σ, τ ) = SΦ δ+ − Ke−rτ Φ δ−
τ
K K
denote the standard Black-Scholes formula for the price of a European call
option, we observe that the lookback call option price satisfies
i.e.
St
e−r(T −t) IE∗ [ST − mT0 | Ft ] = BSc (St , mt0 , r, σ, T − t) + St hc T − t, t
m0
σ2 2
hc (τ, z) = hp (τ, z) − 1 − e−rτ z −2r/σ , τ ∈ R+ , z ∈ R+ ,
2r
2
where (z, τ ) 7→ e−rτ z −2r/σ also solves the PDE (8.30).
∂f ∂f 1 ∂2f
rf (t, x, y) = (t, x, y) + rx (t, x, y) + x2 σ 2 2 (t, x, y), t, x > 0,
∂t ∂x 2 ∂x
under the boundary conditions
lim f (t, x, y) = x, 0 ≤ t ≤ T, x > 0, (8.32a)
y&0
∂f
(t, x, y)x=y = 0, 0 ≤ t ≤ T, y > 0, (8.32b)
∂y
f (T, x, y) = x − y, 0 ≤ y ≤ x, (8.32c)
" 197
In other words, the price of the lookback call option takes the form
T −t x T −t x
f (t, x, y) = xΦ δ+ − e−r(T −t) yΦ δ− (8.34)
y y
2
σ 2 y 2r/σ T −t y
T −t x
+e−r(T −t) x Φ δ− − er(T −t) Φ −δ+
2r x x y
2
T −t x σ T −t x
= x − ye−r(T −t) Φ δ− −x 1+ Φ −δ+
y 2r y
2 2r/σ 2
−r(T −t) σ y T −t y
+xe Φ δ− .
2r x x
We have
f (t, x, x) = xC(T − t),
with
198 "
σ2 σ2
C(τ ) = 1 − e−rτ Φ δ−
τ τ
(1) + e−rτ Φ δ−
τ
(1) − 1 + Φ −δ+ (1) ,
2r 2r
τ > 0, hence
∂f
(t, x, x) = C(T − t), t ∈ [0, T ],
∂x
while we also have
∂f
(t, x, y)y=x = 0, 0 ≤ x ≤ y.
∂y
= xf (t, 1, y/x),
hence letting
σ2
1
g(τ, z) = 1 − e−rτ Φ δ−τ τ
(z) − 1 + Φ −δ+ (z)
z 2r
σ2 2
+ e−rτ z −2r/σ Φ(δ− τ
(1/z)),
2r
we have g(τ, 1) = C(T − t), and
" 199
∂g
(τ, 1) = 0, τ > 0, (8.35a)
∂z
g(0, z) = 1 − 1 ,
z ≥ 1. (8.35b)
z
The next Figure 8.19 shows a graph of the function g(τ, z).
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1 0
50
03 100
2.5 150 t
2
z 1.5 200
1
The next Figure 8.20 represents the path of the underlying asset price used
in Figure 8.19.
200 "
50
40
30
20
10
0
0 50 100 150 200
t
Fig. 8.21: Graph of the lookback call option price.
The next Figure 8.22 represents the corresponding underlying asset price and
its running minimum.
100
St
mt
90
80
70
60
50
40
30
20
0 50 100 150 200
t
Fig. 8.22: Running minimum of the underlying asset price.
for the standard Black-Scholes call formula, we observe that f (t, x, y) satisfies
" 201
i.e.
f (t, x, y) = xg(T − t, x/y),
with
g(τ, z) = BSc (1, 1/z, r, σ, τ ) + hc (τ, z), (8.36)
where hc (τ, z) is the function given by (8.31), and (x, y) 7→ xhc (T − t, x/y)
also satisfies the Black-Scholes PDE (8.22), i.e. (τ, z) 7→ BSc (1, 1/z, r, σ, τ )
and hc (τ, z) both satisfy the PDE (8.30) under the boundary condition
hc (0, z) = 0, z ≥ 1.
The next Figures 8.23 and 8.24 show the decomposition of g(t, z) in (8.36) and
Figures 8.19-8.20 into the sum of the Black-Scholes call function BSc (1, 1/z, r, σ, τ )
and h(t, z).
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1 0
50
03 100
2.5 150 t
2
z 1.5 200
1
Fig. 8.23: Black-Scholes call price in the decomposition (8.36) of the normalized
lookback call option price g(τ, z).
202 "
h(T-t,x)
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1 0
50
03 100
2.5 150 t
2
z 1.5 200
1
Fig. 8.24: Function hc (τ, z) in the decomposition (8.36) of the normalized lookback
call option price g(τ, z).
σ2
T −t St T −t St
ξt = Φ δ + − Φ −δ + (8.37)
mt0 2r mt0
" 203
2r/σ2
mt0 σ2
t
T −t m0
+e−r(T −t) − 1 Φ δ− , t ∈ [0, T ].
St 2r St
Proof. We need to differentiate
x
f (t, x, y) = BSc (x, y, r, σ, T − t) + xhc T − t,
y
σ2 2
τ
(z) − e−rτ z −2r/σ Φ δ−
τ
hc (τ, z) = − Φ −δ+ (1/z)
2r
is given by (8.31) First we note that the relation
∂ τ x
BSc (x, y, r, σ, τ ) = Φ δ+
∂x y
and
σ2
∂hc ∂ 2 ∂
τ
(z) − e−rτ z −2r/σ τ
(τ, z) = − Φ −δ+ Φ(δ− (1/z))
∂z 2r ∂x ∂z
σ 2 2r −rτ −1−2r/σ2
τ
− e z Φ(δ − (1/z))
2r σ 2
σ 1 τ 2
= √ exp − δ+ (z)
2rz 2πτ 2
−rτ −2r/σ 2 σ 1 τ 2
−e z √ exp − (δ− (1/z))
2rz 2πτ 2
2r 2
+ 2 e−rτ z −1−2r/σ Φ(δ− τ
(1/z)) .
σ
204 "
and
∂
x
x
y 2r/σ2 y
xhc τ, = hc τ, − e−rτ τ
Φ δ− ,
∂x y y x x
and the quantity of the riskless asset ert in the portfolio is given by
1−2r/σ2 t !
T −t St St T −t m0
ηt = −mt0 e−rT Φ δ− + Φ δ −
mt0 mt0 St
≤ 0,
Vt = ξt St + ηt ert , t ∈ R+ ,
and one has to constantly borrow from the riskless account in order to hedge
the lookback option.
As we will see below there exists no easily tractable closed form solution for
the price of an arithmetically averaged Asian option.
" 205
General Results
C = φ(YT , ST ),
where wT wT
2
YT = S0 eσBu +ru−σ u/2
du = Su du, T ∈ R+ .
0 0
+
• For example when φ(y, x) = (y/T − K) this yields the Asian call option
with payoff
w + +
1 T YT
Su du − K = −K , (8.39)
T 0 T
which is a path-dependent option whose price at time t ∈ [0, T ] is given
by
1 wT
" + #
−r(T −t) ∗
e IE Su du − K Ft . (8.40)
T 0
wT
" #
e−r(T −t) IE∗ [φ(YT , ST ) | Ft ] = e−r(T −t) IE∗ φ Yt + Su du, ST Ft
t
wT S
" #
−r(T −t) ∗ u ST
=e IE φ y + x du, x Ft
t St St
y=Yt ,x=St
w T −t S ST −t
−r(T −t) ∗ u
=e IE φ y + x du, x . (8.41)
0 S0 S0 y=Yt ,x=St
First we note that the numerical computation of Asian option prices can
be done using the joint probability density ψYT −t ,BT −t of (YT −t , BT −t ), as
206 "
follows:
f (t, x, y) =
w∞w∞ 2
e−r(T −t) φ y + xz, xeσu+r(T −t)−σ (T −t)/2 ψYT −t ,BT −t (z, u)dzdu.
0 −∞
has been computed in the case σ = 2, cf. also [52]. In the next proposition
we restate this result for an arbitrary variance parameter σ after rescaling.
Let θ(v, τ ) denote the function defined as
1 + eσu
σu/2 2
σ −pσu/2−p2 σ2 t/8 4e σ t dz
= e exp −2 2
θ , du,
2 σ z σ2 z 4 z
u ∈ R, z > 0.
The expression of this probability density can then been used for the pricing
of options on average such as (8.41), as
w T −t S ST −t
v
f (t, x, y) = e−r(T −t) IE∗ φ y + x dv, x
0 S0 S0
= e−r(T −t)
w∞ 2
w T −t S
v
× φ y + xz, xeσu+r(T −t)−σ (T −t)/2 P dv ∈ dz, BT −t ∈ du
0 0 S0
σ 2 2
w∞w∞ 2
= e−r(T −t)+p σ (T −t)/8 φ y + xz, xeσu+r(T −t)−σ (1+p)(T −t)/2
2 0 −∞
! !
2 2
1 + eσu−pσ (T −t)/2 p 4eσu/2−pσ (T −t)/4 σ 2 (T − t) dz
× exp −2 − σu θ , du
σ2 z 2 σ2 z 4 z
w∞w∞
−r(T −t)−p2 σ 2 (T −t)/8 2 r(T −t)−σ 2 (T −t)/2
=e φ y + x/z, xv e
0
2
0
4vz σ 2 (T − t)
1+v dz
×v −1−p exp −2z θ , dv ,
σ2 σ2 4 z
" 207
which actually stands as a triple integral due to the definition (8.42) of θ(v, τ ).
Note that here the order of integration between du and dz cannot be ex-
changed without particular precautions, at the risk of wrong computations.
We have
1 wT
" + #
e−r(T −t) IE∗ Su du − K Ft
T 0
wT
" + #
1
= e−r(T −t) IE∗ Yt + Su du − K Ft
T t
wT S
" + #
−r(T −t) ∗ 1 u
=e IE y+x du − K Ft
T t St
x=St , y=Yt
w T −t S
" + #
1 u
= e−r(T −t) IE∗ y+x du − K .
T 0 S0
x=St , y=Yt
1 wT
" + #
f (t, St , Yt ) = e−r(T −t) IE∗ Su du − K Ft ,
T 0
w T −t S
" + #
1 u
f (t, x, y) = e−r(T −t) IE∗ y+x du − K
T 0 S0
" + #
1 x
= e−r(T −t) IE∗ y+ YT −t − K .
T S0
Probabilistic Approach
First we note that the numerical computation of Asian option prices can be
done using the probability density of
wT
YT = St dt.
0
208 "
wt 2
Yt = eσBs −pσ s/2
ds,
0
as follows:
w
t 2
P eσBs −pσ s/2 ds ∈ du
0
σ p2 σ2 t/8 w ∞
2
! 2
!
1 + eσv−pσ t/2 p 4eσv/2−pσ t/4 σ 2 t
= e exp −2 − σv θ , dvdu
2u −∞ σ2 u 2 σ2 u 4
2 2
w∞
1 + v2
4v σ 2 t
du
= e−p σ t/8 v −1−p exp −2 2 θ , dv ,
0 σ u σ2 u 4 u
2 2
w ∞
1+v 2
2
4v σ t
du
= e−p σ t/8 v −1−p exp −2 2 θ , dv ,
0 σ u σ2 u 4 u
2
where St = S0 eσBt −pσ t/2 and p = 1 − 2r/σ 2 . This probability density can
then be used for the pricing of Asian options, as
" + #
1 x
f (t, x, y) = e−r(T −t) IE∗ y+ YT −t − K (8.44)
T S0
w ∞ y + xz +
= e−r(T −t) −K P(YT −t /S0 ∈ dz)
0 T
σ 2 2
w ∞ ∞w
y + xz
+
= e−r(T −t) e−p σ (T −t)/8 −K
2 0 0 T
1 + v2 4v σ 2 (T − t)
−1−p dz
×v exp −2 2 θ , dv
σ z σ2 z 4 z
1 −r(T −t)−p2 σ2 (T −t)/8 w ∞ w∞
= e (xz + y − KT )
T 0∨(KT −y)/x 0
2 2
4v σ (T − t)
1+v dz
× exp −2 2 θ , dv
σ z σ2 z 4 z
4x −r(T −t)−p2 σ2 (T −t)/8 w ∞ w ∞ 1 σ 2 (KT − y)
+
= 2 e −
σ T 0 0 z 4x
1 + v2 σ 2 (T − t)
−1−p dz
×v exp −z θ vz, dv ,
2 4 z
cf. the Theorem in § 5 of [9], which is actually a triple integral due to the
definition (8.42) of θ(v, t). Note that since the integrals are not absolutely
convergent, here the order of integration between dv and dz cannot be ex-
" 209
The time Laplace transform of Asian option prices has been computed in
[29], and this expression can be used for pricing by numerical inversion of the
Laplace transform. The following Figure 8.25 represents Asian option prices
computed by the Geman-Yor [29] method.
30
25
20
15
10
100
95
underlying 90
85 350
250 300
150 200
80 0 50 100
Time in days
Fig. 8.25: Graph of the Asian option price with σ = 0.3, r = 0.1 and K = 90.
We refer to e.g. [2], [9], [20], and references therein for more on Asian op-
tion pricing using the probability density of the averaged geometric Brownian
motion.
Figure 7.1 presents a graph of implied volatility surface for Asian options on
light sweet crude oil futures.
Lognormal approximation
Other numerical approaches to the pricing of Asian options include [49], [73]
which relies on approximations of the average price probability based on
the Lognormal distribution. The lognormal distribution with mean µ and
variance η 2 has the probability density function
1 2 2 dx
g(x) = √ e−(µ−log x) /(2η ) ,
η 2π x
210 "
1 wT
" + #
1 2
e−rT E St dt − K ' eµ̂+σ̂ /2 Φ(d1 ) − KΦ(d2 ), (8.46)
T 0 T
where
√
log(E[ΛT ]/(KT )) T µ̂T + σ̂ 2 T − log(KT )
d1 = √ + σ̂ = √
σ̂ T 2 σ̂ T
and √
√ log(E[ΛT ]/(KT )) T
d2 = d1 − σ̂ T = √ − σ̂ ,
σ̂ T 2
and µ̂, σ̂ are estimated as
E[Λ2T ]
1
σ̂ 2 = log
T (E[ΛT ])2
and
1 1
log E[ΛT ] − σ̂ 2 , µ̂ =
T 2
based on the first two moments of the lognormal distribution, cf. (8.45) below.
The next Figure 8.26 compares the lognormal approximation to a Monte
Carlo estimate of Asian option prices with σ = 0.5, r = 0.05 and K/St = 1.1
0.24
lognormal approximation
Monte Carlo estimate
0.22
0.2
0.18
asian option price
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0 1 2 3 4 5 6 7 8 9 10
time t
" 211
erT − 1
E[ΛT ] = S0 ,
r
and 2
re(2r+σ )T
− (2r + σ 2 )erT + (r + σ 2 )
E[(ΛT )2 ] = 2S02 .
r(r + σ 2 )(2r + σ 2 )
Proof. The computation of the first moment is straightforward. For the
second moment we have, letting p = 1 − 2r/σ 2 ,
wT wT 2 2
E[(ΛT )2 ] = S02 e−pσ a/2−pσ b/2 E[eσBa eσBb ]dbda
0 0
wT wa 2 2 2 2
= 2S02 e−pσ a/2−pσ b/2 eσ (a+b)/2 ebσ dbda
0 0
wT 2
w a 2
= 2S02 e−(p−1)σ a/2 e−(p−3)σ b/2 dbda
0 0
4S02 wT 2 2
= e−(p−1)σ a/2 (1 − e−(p−3)σ a/2 )da
(p − 3)σ 02
4S02 wT 2 4S02 wT 2 2
= e−(p−1)σ a/2 da − e−(p−1)σ a/2 e−(p−3)σ a/2 da
(p − 3)σ 2 0 (p − 3)σ 2 0
8S02 2 4S02 wT 2
= (1 − e−(p−1)σ T /2 ) − e−(2p−4)σ a/2 da
(p − 3)(p − 1)σ 4 (p − 3)σ 0
2
8S02 2 4S02 2
= (1 − e−(p−1)σ T /2 ) − (1 − e−(p−2)σ T )
(p − 3)(p − 1)σ 4 (p − 3)(p − 2)σ 4
2
re(2r+σ )T
− (2r + σ 2 )erT + (r + σ 2 )
= 2S02 ,
r(r + σ 2 )(2r + σ 2 )
The price at time t of the Asian option with payoff (8.39) can be written as
1 wT
" + #
f (t, St , Yt ) = e−r(T −t) IE∗ Su du − K Ft , t ∈ [0, T ].
T 0
(8.47)
Next, we derive the Black-Scholes partial differential equation (PDE) for the
price of a self-financing portfolio. Until the end of this chapter we model the
212 "
where (Wt )t∈R+ is a standard Brownian motion under the historical proba-
bility measure P.
Proposition 8.10. Let (ηt , ξt )t∈R+ be a portfolio strategy such that
(i) (ηt , ξt )t∈R+ is self-financing,
Vt = f (t, Yt , St ), t ∈ R+ ,
∂f ∂f ∂f 1 ∂2f
rf (t, x, y) = (t, x, y) + x (t, x, y) + rx (t, x, y) + x2 σ 2 2 (t, x, y),
∂t ∂y ∂x 2 ∂x
t, x > 0, under the boundary conditions
y +
f (t, 0, y) = e−r(T −t)
−K , 0 ≤ t ≤ T, y ∈ R+ , (8.48a)
T
lim f (t, x, y) = 0, 0 ≤ t ≤ T, x ∈ R+ , (8.48b)
y→−∞
y +
f (T, x, y) = −K , x, y ∈ R+ , (8.48c)
T
and ξt is given by
∂f
ξt =(t, St , Yt ), t ∈ R+ .
∂x
Proof. We note that the self-financing condition implies
∂f ∂f
df (t, St , Yt ) =
(t, St , Yt )dt + St (t, St , Yt )dt
∂t ∂y
∂f 1 2 2 ∂2f ∂f
+µSt (t, St , Yt )dt + St σ (t, St , Yt )dt + σSt (t, St , Yt )dWt .
∂x 2 ∂x2 ∂x
(8.51)
ξt = ∂f (t, St , Yt ),
∂x
i.e.
∂f ∂f ∂f
rf (t, St , Yt ) =
(t, St , Yt ) + St (t, St , Yt ) + rSt (t, St , Yt )
∂t ∂y ∂x
∂2f
1
+ St2 σ 2 2 (t, St , Yt ),
2 ∂x
ξ = ∂f (t, S , Y ).
t t t
∂x
Next we examine two methods which allow one to reduce the Asian option
pricing PDE from two variables to one variable.
1 1 wt
1 Yt
Zt = Su du − K = −K , t ∈ [0, T ].
St T 0 St T
214 "
With this notation, the price of the Asian option at time t becomes
1 wT
" + # " #
−r(T −t) ∗ −r(T −t) ∗
e IE Su du − K Ft = e IE ST (ZT )+ Ft .
T 0
Lemma 8.1. The price (8.40) at time t of the Asian option with payoff
(8.39) can be written as
St g(t, Zt ), t ∈ [0, T ],
where
1 w T −t Su
" + #
g(t, z) = e−r(T −t) IE∗ z+ du (8.52)
T 0 S0
" + #
YT −t
= e−r(T −t) IE∗ z+ .
S0 T
Proof. For 0 ≤ s ≤ t ≤ T , we have
w
1 t St
d (St Zt ) = d Su du − K = dt,
T 0 T
hence
St Zt 1 w t Su
= Zs + du, t ≥ s.
Ss T s Ss
Since for any t ∈ [0, T ], St is positive and Ft -measurable, and Su /St is inde-
pendent of Ft , u ≥ t, we have:
" # " + #
ST
e−r(T −t) IE∗ ST (ZT )+ Ft = e−r(T −t) St IE∗ ZT Ft
St
w
" + #
−r(T −t) ∗ 1 T Su
=e St IE Zt + du Ft
T t St
1 w T Su
" #
+
= e−r(T −t) St IE∗ z+ du Ft
T t St
z=Zt
1 w T −t Su
" + #
−r(T −t) ∗
=e St IE z+ du
T 0 S0
z=Zt
" + #
−r(T −t) ∗ YT −t
=e St IE z+
S0 T
z=Zt
= St g(t, Zt ),
" 215
Note that as in (8.44), g(t, z) can be computed from the density (8.43) of
YT −t , as
" + #
YT −t
g(t, z) = IE∗ z+
S0 T
w∞ u +
= z+ P(Yt /S0 ∈ du)
0 T
−p2 σ 2 t/8
=e
w∞ u + w ∞ −1−p
1 + v2
4v σ 2 (T − t)
du
× z+ v exp −2 2
θ 2
, dv
0 T 0 σ σ u 4 u
2 2
= e−p σ t/8
w∞ u w ∞ −1−p
1 + v2
4v σ 2 (T − t)
du
× z+ v exp −2 θ , dv
(−zT )∨0 T 0 σ2 σ2 u 4 u
2 2
w∞ w∞
1 + v 2
4v σ 2
(T − t)
du
= ze−p σ t/8 v −1−p exp −2 θ , dv
(−zT )∨0 0 σ2 σ2 u 4 u
1 −p2 σ2 t/8 w ∞ w∞ 1 + v2 4v σ 2 (T − t)
−1−p
+ e v exp −2 θ , dvdu.
T (−zT )∨0 0 σ2 σ2 u 4
The next proposition gives a replicating hedging strategy for Asian options.
Proposition 8.11. Let (ηt , ξt )t∈R+ be a portfolio strategy such that
(i) (ηt , ξt )t∈R+ is self-financing,
Vt = St g(t, Zt ), t ∈ R+ ,
∂2g
∂g 1 ∂g 1
(t, z) + − rz (t, z) + σ 2 z 2 2 (t, z) = 0, (8.53)
∂t T ∂z 2 ∂z
under the terminal condition
g(T, z) = z + ,
216 "
1 1
−µ + σ 2 dt − σdWt ,
d =
St St
hence
1 Yt
dZt = d −K
St T
Yt K
=d −
T St St
1 Yt 1
= d − Kd
T St St
1 dYt Yt 1
= + −K d
T St T St
dt 1
= + St Zt d
T St
dt
+ Zt −µ + σ 2 dt − Zt σdWt .
=
T
By self-financing we have
" 217
ξt = g(t, Zt ) − Zt ∂g (t, Zt ),
∂z
i.e.
∂2g
∂g 1 ∂g 1
(t, z) +
∂t − rz (t, z) + σ 2 z 2 2 (t, z) = 0,
T ∂z 2 ∂z
ξ = g(t, Z ) − Z ∂g (t, Z ),
t t t t
∂z
under the terminal condition
g(T, z) = z + .
We check that
∂f ∂f
ξt = e−r(T −t) σSt f (t, St , Zt ) − σZt f (t, St , Zt )
∂x ∂z
−r(T −t) ∂g
=e −Zt (t, Zt ) + g(t, Zt )
∂z
1 1 wt
!
−r(T −t) ∂g
=e St t, Su du − K + g(t, Zt )
∂x x T 0 |x=St
w
∂ 1 1 t
= xe−r(T −t) g t, Su du − K , t ∈ [0, T ].
∂x x T 0 |x=St
We also find that the amount invested on the riskless asset is given by
∂g
ηt At = Zt St (t, Zt ).
∂z
218 "
Next we note that a PDE with no first order derivative term can be obtained
using time-dependent coefficients.
1 1 wt
1
Ut := (1 − e−r(T −t) ) + e−r(T −t) Su du − K
rT St T 0
1 −r(T −t) −r(T −t)
= (1 − e )+e Zt , t ∈ [0, T ],
rT
i.e.
er(T −t) − 1
Zt = er(T −t) Ut + , t ∈ [0, T ].
rT
We have
1
dUt = − e−r(T −t) dt + re−r(T −t) Zt dt + e−r(T −t) dZt
T
= e−r(T −t) σ 2 Zt dt − e−r(T −t) σZt dWt − (µ − r)e−r(T −t) Zt dt
= −e−r(T −t) σZt dŴt , t ∈ R+ ,
where
µ−r
dŴt = dWt − σdt + dt = dW̃t − σdt
σ
is a standard Brownian motion under
2 ST ∗
dP̂ = eσWT −σ t/2
dP∗ = e−rT dP .
S0
Lemma 8.2. The Asian option price can be written as
1 wT
" + #
St h(t, Ut ) = e−r(T −t) IE∗ Su du − K Ft ,
T 0
Proof. We have
1 wT
1
UT = Su du − K = ZT ,
ST T 0
and
" 219
dP̂|Ft 2 e−rT ST
= eσ(WT −Wt )−σ (T −t)/2 = −rt ,
dP∗|Ft e St
hence the price of the Asian option is
e−r(T −t) IE∗ [ST (ZT )+ | Ft ] = e−r(T −t) IE∗ [ST (UT )+ | Ft ]
" #
e−rT ST
= St IE∗ (U T ) +
F t
e−rt St
" #
∗ dP̂|Ft +
= St IE (UT ) Ft
dP∗|Ft
ˆ
= St IE[(U +
T ) | Ft ].
The next proposition gives a replicating hedging strategy for Asian options.
See § 7.5.3 of [71] and references therein for a different derivation of the
PDE (8.55).
Proposition 8.12. Let (ηt , ξt )t∈R+ be a portfolio strategy such that
(i) (ηt , ξt )t∈R+ is self-financing,
Vt = St h(t, Ut ) t ∈ R+ ,
2 2
for some f ∈ C ((0, ∞) × (0, ∞) ).
Then the function h(t, z) satisfies the PDE
2
1 − e−r(T −t) ∂2h
∂h 1
(t, y) + σ 2 −y (t, y) = 0, (8.55)
∂t 2 rT ∂y 2
under the terminal condition
h(T, z) = z + ,
ξt = h(t, Ut ) − Zt ∂h (t, Ut ),
∂y
and
" 221
2 2
1 2 1 − e−r(T −t)
∂h ∂ h
(t, y) + σ − y (t, y) = 0,
∂t
2 rT ∂y 2
1 − e−r(T −t)
∂h
ξt = h(t, Ut ) +
− Ut (t, Ut ),
rT ∂y
under the terminal condition
h(T, z) = z + .
We also find
1 − e−r(T −t) ∂h
∂h
ηt At = er(T −t) St Ut − (t, Ut ) = St Zt (t, Ut ).
rT ∂y ∂y
Exercises
Exercise 8.2 Recall that the maximum Xt := sups∈[0,t] Bs over [0, t] of stan-
dard Brownian motion (Bs )s∈[0,t] has the probability density
r
2 −x2 /(2t)
ϕXt (x) = e 1[0,∞) (x), x ∈ R.
πt
222 "
dP(YT ≤ a & BT ≤ b)
fYT ,BT (a, b) = , a, b ∈ R,
dadb
of standard Brownian motion BT and its minimum
YT = min Bt .
t∈[0,T ]
S −K if min St > B,
T
0≤t≤T
+
C = (ST − K) 1( ) =
min St > B 0
if min St ≤ B,
0≤t≤T 0≤t≤T
Exercise 8.4 Barrier forward contracts. Compute the price at time t of the
following barrier forward contracts on the underlying asset St with exercise
" 223
S − K if max St > B,
T
0≤t≤T
C = (ST − K) 1( ) =
max St > B 0
if max St ≤ B.
0≤t≤T 0≤t≤T
S − K if max St < B,
T
0≤t≤T
C = (ST − K) 1( ) =
max St < B 0
if max St ≥ B.
0≤t≤T 0≤t≤T
S − K if min St < B,
T
0≤t≤T
C = (ST − K) 1( ) =
min St < B 0
if min St ≥ B.
0≤t≤T 0≤t≤T
S − K if min St > B,
T
0≤t≤T
C = (ST − K) 1( ) =
min St > B 0
if min St ≤ B.
0≤t≤T 0≤t≤T
K − ST if max St > B,
0≤t≤T
C = (K − ST ) 1( ) =
max St > B 0
if max St ≤ B.
0≤t≤T 0≤t≤T
K − ST if max St < B,
0≤t≤T
C = (K − ST ) 1( ) =
max St < B 0
if max St ≥ B.
0≤t≤T 0≤t≤T
224 "
K − ST if min St < B,
0≤t≤T
C = (K − ST ) 1( ) =
min St < B 0
if min St ≥ B.
0≤t≤T 0≤t≤T
K − ST if min St > B,
0≤t≤T
C = (K − ST ) 1( ) =
min St > B 0
if min St ≤ B.
0≤t≤T 0≤t≤T
Exercise 8.6 Lookback options. Compute the hedging strategy of the look-
back put option priced in Proposition 8.4.
Exercise 8.7 Consider the short rate process rt = σBt , where (Bt )t∈R+ is a
standard Brownian motion.
rT
1. Find the probability distribution of the time integral 0 rs ds.
2. Compute the price
wT
" + #
e−rT IE ru du − κ
0
rT
of a caplet on the forward rate 0
rs ds.
Exercise 8.8 Asian call options with negative strike. Consider the asset price
process
" 225
2
St = S0 ert+σBt −σ t/2
, t ∈ R+ ,
where (Bt )t∈R+ is a standard Brownian motion. Assuming that κ ≤ 0, com-
pute the price
1 wT
" + #
e−r(T −t) IE∗ Su du − κ Ft
T 0
Exercise 8.9 Pricing of Asian options by PDEs. Show that the functions
g(t, z) and h(t, y) are linked by the relation
1 − e−r(T −t)
g(t, z) = h t, + e−r(T −t) z , t ∈ [0, T ], z > 0,
rT
and that the PDE (1.35) for h(t, y) can be derived from the PDE (1.33) for
g(t, z) and the above relation.
226 "