Lecture 13 Itô’s Formula

Itô’s formula is at the heart of stochastic calculus. It shows an important distinction from the
Fundamental Theorem of Calculus. We will present the 1D and multidimensional versions of Itô’s
formula and illustrate their applications in a number of examples.

13.1 The 1D case

Definition 13.1 X = {Xt } is called a 1D Itô process if it has an expression

Z t Z t
Xt = X0 + µu du + σu dWu , 0 ≤ t ≤ T, (13.1)
0 0

where the drift process µ = {µt } is adapted to the Brownian filtration, and the diffusion coefficient
σ = {σt } is a H 2 -process (see Lecture 12).

Sometimes an Itô process X is defined as a solution to the stochastic differential equation (SDE)

dXt = µt dt + σt dWt . (13.2)

Note that (13.2) is not well-defined in a strict sense because Brownian motion is non-differentiable.
From now on, whenever we see a differential form (13.2), it should be understood as defined by
(13.1).

Theorem 13.1 Let g(t, x) be continuously differentiable in t and twice continuously differentiable
in x. Define Yt = g(t, Xt ), 0 ≤ t ≤ T . Then Y = {Yt } is an Itô process that satisfies

∂g(t, Xt ) ∂g(t, Xt ) 1 ∂ 2 g(t, Xt )

dYt = dt + dXt + (dXt )2 (13.3)
∂t ∂x" 2 ∂x2 #
∂g(t, Xt ) ∂g(t, Xt ) ∂g(t, Xt ) 1 ∂ 2 g(t, Xt ) 2
= σt dWt + + µt + σt dt.
∂x ∂t ∂x 2 ∂x2

Note:

(a) The following rules are used in the expansion of (dXt )2 = dhX, Xit : (dt)2 = (dt)(dWt ) = 0
(higher-order infinitesimals); (dWt )2 = dhW, W it = dt (quadratic variation of Brownian motion).

(b) The Itô’s formula (13.3) can be expressed as an integral form

Z t " #
∂g(u, Xu ) ∂g(u, Xu ) 1 ∂ 2 g(u, Xu ) 2
Yt = Y0 + + µu + σu du
0 ∂u ∂u 2 ∂x2
Z t
∂g(u, Xu )
+ σu dWu . (13.4)
0 ∂x

1
(c) Theorem 13.1 is proved by Taylor expansion and estimation of the remainder. See the book by
Karatzas and Shreve for details.

13.2 The multidimensional case

For a positive integer m, W = (W (1) , ..., W (m) )0 is called a m-dimensional Brownian motion if
W (1) , ..., W (m) are independent 1D standard Brownian motions. Here A0 denotes the transpose of
matrix A. Let {Ft }t≥0 be the Brownian filtration generated by W .

Definition 13.2 X = (X (1) , ..., X (k) )0 is called a k-dimensional Itô process if it has an expression

dXt = µt dt + σt dWt , 0 ≤ t ≤ T, (13.5)

where W = {Wt } is a m-dimensional Brownian motion, the (vector) drift process µ = {µt } =
(µ(1) , ..., µ(k) )0 is adapted to the Brownian filtration, the diffusion coefficient σ = {σt } is a k × m
matrix
 (11) (1m)

σt · · · σt
 
σt = 
 ··· ,
 (13.6)
(k1) (km)
σt ··· σt

where all entries σ (11) , ..., σ (km) are H 2 -processes.

Theorem 13.2 Let g(t, x), x = (x1 , ..., xk ) be continuously differentiable in t and twice continu-
ously differentiable in x ∈ IRk . Define Yt = g(t, Xt ), 0 ≤ t ≤ T . Then Y = {Yt } is an Itô process
that satisfies

Xk X k
∂g(t, Xt ) ∂g(t, Xt ) (i) 1 ∂ 2 g(t, Xt ) (i) (j)
dYt = dt + dXt + dXt dXt , (13.7)
∂t i=1
∂xi i,j=1
2 ∂xi ∂xj

(i) (j) (i)

where the following rules are adopted in the expansion of dXt dXt : (dt)2 = (dt)(dWt ) = 0
(i)
(higher-order infinitesimals); (dWt )2 = dt (quadratic variation of Brownian motion);
(i) (j)
dWt dWt = 0 for i 6= j (independence).

The following Itô product rule follows from Theorem 13.2.

Corollary 13.1 Let X, Y be two 1D Itô processes. Then

d(Xt Yt ) = Xt dYt + Yt dXt + dXt dYt . (13.8)

2
13.3 Examples
Rt
Example 13.1 Verify 0 Wu dWu = Wt2 /2 − t/2.

Example 13.2 (geometric Brownian motion) Show that St = S0 exp[(µ − σ 2 /2) t + σWt ] satisfies
the SDE dSt /St = µ dt + σ dWt with constants µ ∈ IR and σ > 0.
R t −µ(t−u)
Example 13.3 (Ornstein-Uhlenbeck process) Show that Xt = X0 e−µt +σ 0e dWu satisfies
the SDE dXt = −µXt dt + σ dWt with positive constants µ and σ.

Example 13.4 Consider the SDE

m
X (i) (i) (i)
dXt = (At Xt + µt ) dt + [Bt Xt + σt ] dWt , 0 ≤ t ≤ T, (13.9)
i=1

(i) (i)
where W = (W (1) , ..., W (m) ) is a m-dimensional Brownian motion, and {At }, {µt }, {Bt }, {σt },
i = 1, ..., m are adapted processes. Use Itô’s formula to show the following Xt satisfies (13.9):
" Z t à m
! m Z t
#
X X
Xt = ξt X0 + ξu−1 µu − Bu(i) σu(i) du + ξu−1 σu(i) dWu(i) , (13.10)
0 i=1 i=1 0

where "Z #
m Z t
X m Z t
t 1 X
ξt = exp Au du + Bu(i) dWu(i) − (Bu(i) )2 du
0 i=1 0
2 i=1 0
.
R t −X
Example 13.5 Use Itô’s formula to calculate EYt where Yt = eXt 0e
u dW and X = at + bW
u t t
with constants a ∈ IR and b > 0.