Section 7.

2 Itô Process

2400

2200

2000

Past (real data)

1800

(generated data)
1600

1400

Future
1200
-1 -0.8 -0.6 -0.4 -0.2 0 0.2

Figure 7.2: One year daily closing price of Infosys Technology (labeled ‘Past’) and the
stock price process (labeled ‘Future’) simulated using the Monte Carlo method (7.10)
with parameters given in Example 7.1.8.

Project 7.1.
Write a python code to read a CSV file consisting of the past 200 days (at least)
daily NIFTY50 data (can be downloaded by clicking here), and do the following:
1. Extract the column consisting of ’CLOSE’ points of NIFTY50;
2. Find the 200 days simple moving average and take it as µ;
3. Find the standard deviation using 200 days simple moving average formula
and take it as ω.
4. Predict the next day’s NIFTY50 close price as per the Monte Carlo method.

7.2 Itô Process

In the previous section, we established a stochastic model for a stock price process, sim-
plifying it into a generalized Wiener process with constant drift and volatility. However,
real-world scenarios, such as stock prices and option prices, demand a more detailed
approach, wherein these parameters can fluctuate with time and the underlying process
itself. Thus, a more realistic stochastic model takes the form

dX = µ(X, t)dt + ω(X, t)dW , (7.11)

S. Baskar 177 Spring 2025

 Section 7.2 Itô Process

where W represents a Wiener process, and both the drift and volatility are functions
of X and t. Such a stochastic process, denoted by X = {Xt } governed by the above
stochastic differential equation, is termed an Itô process. Unlike the straightforward
solution found in the case of a generalized Wiener process, solving this equation is
more involved.
Remark 7.2.1.
While stochastic differential equations offer the advantage of familiarity with basic
calculus notation, they present challenges in interpretation due to the involvement
of non-differentiable Brownian paths. Consequently, these equations are typically
handled in integral form
!t !t
X(t) = X0 + µ(X(s), s)ds + ω(X(s), s)dW ,
0 0
where X0 is a given constant. This form of equation is known as a stochastic
integral equation. The integral in the third term on the right-hand side requires
interpretation in the Itô sense, known as the Itô integral. To ensure meaningful
interpretation of the second and third terms, we assume that µ ∈ L1loc and ω ∈ L2loc .
We omit the details of the Itô integral and will not go further into the theory of
stochastic integral equations.

7.2.1 Itô Lemma

The price of a derivative depends on both the underlying asset and time. Even in the
stock price process, as we have seen, it is common to work with ln S rather than S.
Thus, dealing with functions of an Itô process, such as f (S, t), is inevitable. Caution is
required in such situations as the differential df (S, t) is not exactly analogous to what
we encounter in ordinary calculus. In fact, df (S, t) involves an additional term, and the
precise formula is called the Itô formula. Here, we state the formula and provide only
a brief outline of its proof.
Lemma 7.2.2 [Itô Lemma].
Let X be an Itô process in (7.11) and let Y = F (X, t), where F ∈ C 2,1 (R2 ). Then,
Y is an Itô process satisfying the equation

dY = µF dt + ωF dW ,

where the drift and the volatility processes are given by

∂F ∂F 1 ∂2F
µF = µ + + ω2 ,
∂x ∂t 2 ∂x2
∂F
ωF = ω .
∂x

S. Baskar 178 Spring 2025

 Section 7.2 Itô Process

Note 2
In the ordinary calculus, the expression for µF does not include the term ω 2 12 ωωxF2 .

Proof.
We omit the rigorous proof of the lemma and present only a formal derivation of the
formula. Let us take F = F (X, t), for (X, t) ∈ R2 and assume that F is smooth.
Using Taylor expansion, we can write

∂F ∂F 1 ∂2F 2 ∂2F 1 ∂2F 2

∆F = ∆X + ∆t + ∆X + ∆X∆t + ∆t + · · · (7.12)
∂X ∂t 2 ∂X 2 ∂X∂t 2 ∂t2
Let us consider discrete form of the Itô process

∆X = µ(X, t)∆t + ω(X, t)∆W .

Since W satisfies the Wiener process, we have

√
∆W = # ∆t.
Therefore, we have (dropping the arguments for the sake of notational convenience)
√
∆X = µ∆t + ω# ∆t,

Squaring both sides, we get

∆X 2 = ω 2 #2 ∆t + o(∆t).
Substituting these expressions in (7.12), we get
" #
∂F ∂F 1 ∂2F 2 2 ∂F
∆F = µ + + ω # ∆t + ω ∆W + o(∆t).
∂X ∂t 2 ∂X 2 ∂X

Since # ∼ N (0, 1), we have

Var(#2 ∆t) = ∆t2 E(#4 ) − ∆t2 .
Since # ∼ N (0, 1), we have from the basic probability theory that the fourth moment
of the standard normal distribution is 3. That is, E(#4 ) = 3. This implies that
Var(#2 ∆t) = 2∆t2 .
Recall from the basic probability theory, that if a random variable has Var(X) = 0,
then X = E(X) a.s. Since Var(#2 ∆t) → 0 as ∆t → 0, we see that #2 ∆t tends to dt
a.s. as ∆t → 0.
Thus, neglecting o(∆t) terms and then taking ∆t → 0, we get
" #
∂F ∂F 1 ∂2F ∂F
dF = µ + + ω2 dt + ω dW .
∂X ∂t 2 ∂X 2 ∂X

S. Baskar 179 Spring 2025

 Section 7.3 Option Pricing Model

Example 7.2.3.
Consider the stochastic differential equation (7.9)
dS = µ̃Sdt + ωSdW .
This equation can be viewed as Itô process with µ(S, t) = µ̃S and ω(S, t) = ωS.
Now let us take F (S, t) = ln S. Then, we have

∂F 1 ∂F ∂2F 1
= , = 0, 2
= − 2.
∂S S ∂t ∂X S
Therefore, by Itô lemma, we have
$ $ %%
1 1 1 1
d ln S = µ̃S × + 0 + ω 2 S 2 − 2 dt + ωS × dW .
S 2 S S
This leads to the stochastic differential
$ equation
% equivalent to the above equation as
1 2
d ln S = µ̃ − ω dt + ωdW ,
2
which is the same as (7.8).

7.3 Option Pricing Model

In previous sections, we have derived a continuous time model for the underlying asset.
We now study the continuous time model for option premium.

Recall, in Theorem 6.2.8, we have obtained a formula for option premium using an n-
(n)
step binomial model. Let us assign to each n, the corresponding price as H0 . Then we
(n)
have a sequence of option prices {H0 }, where the nth term of the sequence corresponds
to the price of the option at time t = 0 computed using the partition Tn of [0, T ] with
∆tn = T /n, for n = 1, 2, . . .. With this notation, we pose the following question: ‘does
the price sequence converge as n → ∞?’. This question is equivalent to asking, ‘is the
pricing model stable?’. It can be shown that the binomial model is stable, i.e. the
price sequence converges as n → ∞, and in the limiting case, we obtain the well-known
Black-Scholes model.

For a given number of partitions n = 1, 2, . . ., let us denote the up and down movements
of the stock as un , and dn , respectively. For the sake of simplicity, we consider the
following parameter calibration:
& √ ' ⎫
un = exp µ∆tn + ω ∆tn ⎪
⎪
⎪
& ' ⎬
√
dn = exp µ∆tn − ω ∆tn (7.13)
⎪
⎪
⎪
pn = 12 . ⎭

S. Baskar 180 Spring 2025

 Section 7.3 Option Pricing Model

Assume that the risk-free instrument is a continuous compounding scheme with interest
rate r.

Let us assume that the market is viable and complete. Then, by Theorem 6.3.22, there
exists a unique EMM, P∗n ({un }) = p∗n , which, in the binomial context, is the risk-neutral
probability
er∆tn − dn
p∗n = .
un − dn

First, let us observe that the probability p∗ is independent of µ as n → ∞.

Lemma 7.3.1.
If un and dn are given by (7.13), then
1
lim p∗ = .
n→∞ n 2

Proof.
By definition, we have
√
er∆tn − e(−σ ∆tn +µ∆tn )
p∗n = √ √ .
e(σ ∆tn +µ∆tn ) − e(−σ ∆tn +µ∆tn )
Therefore, we have
√ √
2er∆tn − e(σ ∆tn +µ∆tn ) − e(−σ ∆tn +µ∆tn )
2p∗n − 1 = √ √ .
e(σ ∆tn +µ∆tn ) − e(−σ ∆tn +µ∆tn )
Using Taylor expansion for each exponential, we get
& '
σ2
r− 2
− µ ∆tn + o(∆tn )
2p∗n − 1 = √ , as n → ∞. (7.14)
ω ∆tn (1 + o(1))

We can see that the RHS tends to zero as n → ∞. Hence the result.

7.3.1 Black-Scholes Formula

We consider a European option with strike K and period [0, T ]. From Theorem 6.2.8
we can see that for a given n = 1, 2, . . ., the price of the option is given by
(n) (n)
H0 = E∗n (e−rT HT ),
(n)
where HT is the payoff of the option.

S. Baskar 181 Spring 2025

 Section 7.3 Option Pricing Model

Let us first consider a European put option, where the payoff is given by
(n) (n)
PT = max(0, K − ST ) (7.15)
(n)
where ST = S0 eXn , with Xn being the random variable on the sample space (Ω, F ∗ )
defined by
n
, n
- (n)
Xn (ω) = log ωk = Yk (ω), (7.16)
k=1 k=1

for each ω = (ω1 , ω2 , . . . , ωn ) ∈ Ω, and

.
(n) un , if ωk = U ,
Yk (ω) = log ω k and ω k = k = 1, 2, . . . , n.
dn , if ωk = D,

With the above notation, the initial option price can be written as
$ & '%
(n)
P0 = E∗n e−rT max 0, K − S0 eXn . (7.17)

We assume that the stock movement during the time period (tk , tk+1 ] is independent of
its movement in (tk−1 , tk ]. With this assumption, we can see that Yk ’s are iid random
variables and we have (for each k = 1, 2, . . . , n)
$ % ⎫
(n) √
P∗n {Yk = ω ∆tn + µ∆tn } = p∗n , ⎪
⎪
⎬
$
√ %
⎪
(7.18)
(n)
P∗n {Yk = −ω ∆tn + µ∆tn } = 1 − p∗n . ⎪
⎭

We now state two important properties of Xn .

Lemma 7.3.2.
" #
We have: ω2
lim E ∗
(X ) = r − T, (7.19)
2
n
n→∞ n

lim Var∗n (Xn ) = ω 2 T . (7.20)

n→∞

Proof.
First let us prove (7.19). From (7.18) we see that, for each k = 1, 2, . . . , n,
$ / % $ / %
(n)
E∗n (Yk ) = p∗n ω ∆tn + µ∆tn + (1 − p∗n ) −ω ∆tn + µ∆tn
/ & '
= (2p∗n − 1)ω ∆tn + ∆tn µp∗n + µ(1 − p∗n )

S. Baskar 182 Spring 2025

 Section 7.3 Option Pricing Model

Using Lemma 7.3.1 and (7.14), we get

& '
σ2
(n)
r− 2
− µ ∆tn + o(∆tn )
E∗n (Yk ) = + µ∆tn , as n → ∞.
(1 + o(1))

This can be written as

" #
(n) ω2
E∗n (Yk ) = r− ∆tn + o(∆tn ), as n → ∞.
2

Observe that the right hand side is independent of k and therefore, we have
." # 0
(n) ω2
E∗n (Xn ) = nE∗n (Y1 ) =n r− ∆tn + o(∆tn ) , as n → ∞.
2

Since ∆tn = T /n, we get

" #
ω2
E∗n (Xn ) = r − T + o(1), as n → ∞,
2

which proves (7.19).

(n)
Let us now prove (7.20). Since Yk ’s are iid, we have
$ %
(n) (n) 2 (n) 2
Var∗n (Xn ) = nVar∗n (Y1 ) =n E∗n (Y1 ) − E∗n (Y1 ) .

But, we have
(n) 2 (n)
E∗n (Y1 ) = (log un + log dn )E∗n (Y1 ) − log un log dn
"" # #
ω2
= (µ + µ)∆tn r− ∆tn + o(∆tn )
2
/ /
−(ω ∆tn + µ∆tn )(−ω ∆tn + µ∆tn )
2
= ω ∆tn + o(∆tn ), as n → ∞.

Substituting this, we get

⎛ "" # #2 ⎞
ω2
Var∗n (Xn ) = n ⎝ω 2 ∆tn + o(∆tn ) − r− ∆tn + o(∆tn ) ⎠
2
= ω 2 T + o(1), as n → ∞.

This proves (7.20).

The next property is a variation of central limit theorem. So, we state the property
and omit the proof.

S. Baskar 183 Spring 2025

 Section 7.3 Option Pricing Model

Lemma 7.3.3.
The sequence of random variables {Xn } defined by (7.16) converges in distribution
to a random variable "" # #
ω2 2
X∼N r− T,ω T .
2

The above two lemmas can be combined to get the following theorem defining the
Black-Scholes price formula for a European put option.

Theorem 7.3.4 [Black-Scholes pricing model for Put].

(n) (n)
Let H0 be the price of a European K-strike put option (denoted by P0 ) with
period [0, T ] partitioned into n sub-periods. Let the parameters un and dn be given
by (7.13). Then the limit
(n)
lim P
n→∞ 0
= P0

exists and we have

P0 = E(e−rT PT ), (7.21)
&& ' '
σ2
where PT = max(0, K − S0 eX ) with X ∼ N r− 2
T , ω2T .

The proof of the above theorem is advanced and we omit it for our course. However,
we make the following remark about the proof of the theorem.
Remark 7.3.5.
Recall that the convergence in distribution of random variables is defined only
in terms of convergence of the respective distributions. Consequently, it is not
necessary that the random variables are defined on the same probability space. If
we denote by (Ωn , Fn , Pn ) the probability space on which Xn is defined and (Ω, F, P)
the probability space on which X is defined, then {Xn } converges in distribution
to X if and only if
lim En (φ(Xn )) = E(φ(X)),
n→∞

for all φ ∈ Cb(Rm )(set of all bounded continuous functions), where En and E are
the expectations taken with respect to the probability Pn and P, respectively.
(n)
We used the above property to conclude the convergence of the sequence {P0 }
(n)
given by (7.17) to P0 . Note that to use the above property, we need PT given by
(7.15) to be a bounded continuous function and this is the reason why we restricted
to put options. We can extend the pricing formula to call options using put-call
parity relation.

S. Baskar 184 Spring 2025

 Section 7.3 Option Pricing Model

Definition 7.3.6 [Black-Scholes Price for Put].

The number P0 defined by (7.21) is called the Black-Scholes price of a European put
option with strike K and maturity T .

Note that (7.21) does not give an explicit formula for the price. However, it is possible
to derive an explicit formula for P0 , which is one of the advantages of the Black-Scholes’
model.
Corollary 7.3.7 [Black-Scholes Formula for European Put].
The Black-Scholes price P0 can be written as

P0 = Ke−rT Φ(−d− ) − S0 Φ(−d+ ),

where Φ is the standard normal distribution function

x
1 ! − s2
Φ(x) = √ e 2 ds, x ∈ R,
2π −∞

and
& ' & '
σ2
log S0
+ r± 2
T
d± = , (7.22)
K
√
ω T
r is the prevailing interest rate.

Proof.
From (7.21), we see that
$ & '%
P0 = e −rT
E max 0, K − S0 e X
.
&& ' '
σ2
Since X ∼ N r− 2
T , ω 2 T , we see that
" #
ω2 √
X = r− T + ω T Z,
2

where Z ∼ N (0, 1) is the standard normal variable. Therefore,

S0 eX < K ⇐⇒ Z < −d− .

S. Baskar 185 Spring 2025

 Section 7.3 Option Pricing Model

We have
$ & '%
E max 0, K − S0 eX = KE(11{ST <K} ) − E(ST 11{ST <K} ).

Let us compute the two quantities on the right hand side.

For the first term, we have
E(11{ST <K} ) = E(11{Z<−d− } )
= Φ(−d− ).

The second term can be simplified to

√
−d−!−σ T
1 s2
E(ST 11{ST <K} ) = erT S0 √ e− 2 ds.
−∞
2π
√
Since d+ = d− + ω T , we obtained the required formula.

We can now obtain the price for a European call option.

Corollary 7.3.8 [Black-Scholes Formula for Call].

The Black-Scholes price C0 for a European call option can be written as

C0 = S0 Φ(d+ ) − Ke−rT Φ(d− ),

where Φ is the standard normal distribution function and d± are given by (7.22).

Proof is left as an exercise.

Problem 7.7.
Use Black-Scholes formula to find the price of the European call option with the
following details:
• 1 contract consists of 100 units;
• 25-strike;
• the underlying is trading at 20 per share;
• 3 months expiration;
• the stock’s volatility is 24%; and
• the risk-less interest rate is 5% continuously compounded.
Answer: 4.24

S. Baskar 186 Spring 2025

 Section 7.3 Option Pricing Model

7.3.2 Black-Scholes Differential Equation

We can derive a parabolic partial differential equation (PDE) as a limiting case of the
binomial model and the PDE is the well known Black-Scholes equation given by

∂H ω 2 S 2 ∂ 2 H ∂H
+ + rS − rH = 0. (7.23)
∂t 2 ∂S 2 ∂S

The starting level of the backward induction at time t = tn (= T ) can be taken as

H(T , S) = HT ,

where HT is the payoff of the European option.

The above two equations form a terminal value problem. Using the change of variable

H(t, S) = u(τ , x),

where τ = T − t and x = log S, we obtain the initial value problem (or the Cauchy
problem)
" #
∂u ω2 ∂u ω 2 ∂ 2 u
− + r− + − ru = 0, (τ , x) ∈ (0, T ] × R,
∂τ 2 ∂x 2 ∂x2
u(0, x) = HT (ex ), x ∈ R.

Deriving the above IVP is a simple calculus exercise.

Problem 7.8.
Let H(t, S) be the European option premium that satisfies the Black-Scholes partial
differential equation for all (t, S) ∈ [0, T ]×(0, ∞). Consider the portfolio Π = (0, s, h),
where s and h are the number of units held in the underlying stock and the option,
respectively. Show that the value of the portfolio V considered as a function of
(t, S) (i.e. V (t, S)) also satisfies the Black-Scholes partial differential equation for all
(t, S) ∈ [0, T ] × (0, ∞).

S. Baskar 187 Spring 2025