Académique Documents
Professionnel Documents
Culture Documents
Karol Dziedziul
Gdansk
Both integrals exist: the first in the sense of the Lebesgue integration,
the second in the sense of the Ito integration when the functions a and b
have to fulfill integrability conditions
Z t
(|a(s, Xs )| + |b(s, Xs )|2 )ds, a.s., t ≥ 0.
0
In this way we obtain the piecewise linear sample path of the process. If
sup{|tj − tj−1 | : 0 ≤ j ≤ n} → 0
♣ Use Program 28 to find out that the price of the asstet ST /S0 (the
solution of (2)) in the moment T has lognormal distribution, where S0 is
an initial value of the asset.
Lognormal distribution
m = µ − σ2 , s = σ.
Itô process
More explicitly Z t Z t
Yt = Y0 + as ds + bs dWs .
0 0
and Z T
|bs |2 ds < ∞ a.s.
0
Itô calculus
Theorem (Itô formula)
Let Yti , i = 1, 2 be two Itô processes
Let
F : [0, t] × R 2 → R
be twice differentiable. Then ξt = F (t, Yt1 , Yt2 ) is the Ito process such
that
∂F ∂F ∂F
dξt = (t, Yt1 , Yt2 )dt + (t, Yt1 , Yt2 )dYt1 + (t, Yt1 , Yt2 )dYt2
∂t ∂x1 ∂x2
1 ∂2F 1 2 1 2 1 ∂2F
+ (t, Yt , Yt )(b t ) dt + (t, Yt1 , Yt2 )(bt2 )2 dt
2 ∂x12 2 ∂x22
∂2F
+ (t, Yt1 , Yt2 )bt1 bt2 dt.
∂x1 ∂x2
Karol Dziedziul The stochastic calculus
.
or equivalently
(µ−σ 2 /2)ds+ 0t σdWs
Rt R
St = S0 e 0 .
is the solution of the Geometric Brownian motion, i.e. solution of the
equation
dSt = St µdt + St σdWt . (4)
We define
Z t Z t
2 2
Yt = (µ − σ /2)t + σWt = (µ − σ /2)ds + σdWs .
0 0
St = S0 exp(Yt ).
1
dSt = S0 e Yt dYt + SeYt σ 2 dt
2
1
= S0 e Yt ((µ − σ 2 /2)dt + σdWt ) + S0 e Yt σ 2 dt = St µdt + St σdWt .
2
From the Itô integral theory we get that the stochastic process
Z t
Ut = e θ(s−t) σdWs
0
2
Z t
VarUt = σe −θt e 2θs ds.
0
Hence
e 2θt − 1 1 − e −2θt
VarUt = σ 2 e −2θt = σ2 .
2θ 2θ
Consequetly,
σ2
lim VarXt = . (7)
t→∞ 2θ
♣ Use Program 29 to create a sample path of the Ornstein - Uhlenbeck
process for large n. Note that a trajectories according to (6) return to µ
and their oscilations stabilize according to (7).
Heston model
One of the generalizations of the Samuelson model, is the Heston model.
We assume that the price of an asset is given by
dSt = µSt dt + σt St dZt1 , (8)
where σt is the Ornsteina-Uhlenbeck process
dσt = −βσt dt + δdZt2 . (9)
We assume that the Wiener processes (Zt1 , Zt2 )
are correlated, i.e
Cov (Zt1 , Zt2 ) = ρt.
To obtain the equivalent formula of Heston model let us define
νt = σt2 .
From the Itô formula
1
dνt = 2σt dσt + 2δ 2 dt.
2
√
Taking (9) and puting σt = νt we get
√ √ 1
dνt = 2 νt (−β νt dt + δdZt2 ) + 2δ 2 dt.
2
Karol Dziedziul The stochastic calculus
.
Thus
√ √
dνt = 2δ νt dZt2 + (δ 2 − βνt )dt = 2δ νt dZt2 + β(δ 2 /β − νt )dt
We obtain the system of equations equivalent to (8) i (9),
√
dSt = µSt dt + νt St dZt1
√
dνt = 2β(δ 2 /β − νt )dt + δ νt dZt2 .
Theorem
For any x > 0 the density function of T0 equals
p x
f (t) = A0 (t)φ(−2x/ A(t)) ,
A(t)3/2
where Z t
A(t) = e 2ks ds
0
Proof
Thus Z t
T0OU = inf{t ≥ 0 : x + e ks dWs = 0}.
0
the processes Z t
Mt = f (s)dWs
0
is a martingale.
Mt = W<M>t = WB(t) ,
where Z t
B(t) = f 2 (s)ds.
0
From Dubins Schwarz’s Theorem there is BM W such that
Consequently
where N(t) is a Poisson process with intensity λ and {Yj } i.i.d. and
independent of N(t). Let us consider the stochastic equation with
constants µ and σ
dSt = St− (µdt + σdWt + dXt ),
where St− is left continuous version of St .
Theorem
The process St e −rt , t ≥ 0 is a martingale iff µ + λEY1 = r .