MARTINGALES, WIENER

PROCESSES & ITO’S LEMMA

Sankarshan Basu
Professor of Finance
Indian Institute of Management Bangalore
MARTINGALES
Martingales

• A martingale is a zero-drift stochastic process.

• A variable θ follows a martingale if its process
has the form
dθ = σ dz
where dz is a Wiener process.
• The expected value of a martingale at any
future time is equal to its value today.
E(θT) = θ0
Market Price of Risk

• The market price of risk of a variable θ is given

by
λ = (μ – r) / σ
where
μ = expected rate of return
r = risk-free rate
σ = volatility
Money Market Account

• The money market account is a security that is

worth 1 at time zero and earns the
instantaneous risk-free rate r at any given
time.
• If g is a money market account then
dg = r g dt
Equivalent Martingale Measure Result

• Suppose that f and g are the prices of two

traded securities dependent on a single source
of uncertainty and provide no income during
the time period under consideration.
• Let φ = f/g
• f/g can be considered as a price of f per unit of
g.
• The price of g is referred to as the numeraire.
Equivalent Martingale Measure Result

• The equivalent martingale measure result

shows that, without arbitrage opportunities, if
the market price of risk of f is set equal to the
volatility of g, then the ratio f/g is a martingale
for all security prices of f.
• Then λ of security f = σg
• μ=r+λσ
Equivalent Martingale Measure Result - Proof
• df = μ f dt + σ f dz
• or, df = (r + λ σ) f dt + σ f dz
• Hence, df = (r + σg σf) f dt + σf f dz
and dg = (r + σg2) g dt + σg g dz
• Using ITO’s lemma,
dlnf = (r + σg σf - σf2/2) dt + σf dz
and dlng = (r + σg2/2) dt + σg dz
• Hence, d(ln f/g) = - (σf – σg)2/2 dt +(σf – σg) dz
• With reverse ITO’s lemma then the process for f/g is
d(f/g) = (σf – σg) f/g dz
• So, f/g is a martingale i.e., E(fT/gT) = f0/g0
Wiener Processes and Itô’s Lemma

9
• Y (Price) = aX (time) + b
– a > 0; b > 0 – feasible solution  price is a monotonically increasing function of time
– a > 0; b < 0 – feasible in parts, infeasible in parts
– a < 0; b > 0 – infeasible in parts, feasible in parts
– a < 0; b < 0 – infeasible solution
• Know S0 – want to find ST
• Assume normality;
– Min(St) = -  ; Max(St) = + 
• St = K + at + b Random Term
– Random Term – Brownian Motion / Weiner Process
– It is defined as Z where z = t;   N(0, 1)
• ST = S0 + St = S0 + (at + bz)
• St = a t +  Random Term = at + bz
– Change in St = function (Change in time; change in the Random Term)
– Assume normality on St
– Also assume rate of return to be constant and write the equation as
• St = a St t + b St z  (St / St)= at + bz
• (dSt / St)= adt + bdz; integrate this equation between (0, T) and reorder – you will get a function of ST
• ln ST – ln S0 = aT + bdz  ln ST = (ln S0 + aT) + bdz = Constant + bdz
• (dSt / St)= dt + dz – Black Scholes Price Equation

• ln St  N(, 2)  Min(ln(St)) = -   exp(Min(ln(St))) = exp(- )  Min(exp(ln(St))) = exp(- )

•  Min(St) = 0

10
Stochastic Process
• Any variable whose value changes over time in an
uncertain way is said to follow a stochastic process.
• Continuous-time, continuous variable stochastic
process means that the underlying variable can take
any value within a certain range and that can take
place at any time.
• The continuous time, continuous variable process
proves to be the most useful for the purposes of
valuing derivatives.

11
Markov Processes
• A Markov process is a particular type of stochastic
process where only the present value of a variable
is relevant for predicting the future.
• The past history of the variable and the way in
which the present has emerged from the past are
irrelevant.
• We assume the Markov property for stock price
implying that the probability distribution of the
price at any particular future time depends only
on the current stock price.

12
 Wiener Process
• A Wiener process is a particular type of Markov
stochastic process with a mean change of Zero and
variance rate of 1 per year.

• A variable z will follow Wiener process if z, the

change in a small interval of time t has the
following properties:

1. z   t where is (0,1)
2. The values of z for any two different (non-
overlapping) short periods of time are
independent.
13
Wiener Process – contd.
• It follows from the first property that δz itself
has a normal distribution with
Mean of δz = 0
Standard deviation of δz = √δt
Variance of δz = δt
• The second property implies that z follows a
Markov process.

14
Wiener Process – contd.
• Consider the increase in the value of z during a relatively long
period of time, T.
• This can be denoted by z(T) - z(0).
• It can be regarded as the sum of the increases in z in N small
time intervals of length δt, where
T
N
t
N
Thus, z(T) - Z(0)    i t
i 1
where, the  i (i  1, 2, ....., N) are random drawing from  (0, 1)

15
Properties of Wiener Process

• From second property of Wiener processes,

the εi’s are independent of each other.
• [z (T ) – z (0)] is normally distributed with
Mean of [z (T ) – z (0)] = 0
Variance of [z (T ) – z (0)] = T
Standard deviation of [z (T ) – z (0)] =√T

16
Generalized Wiener Processes
• The mean change per unit time for a stochastic
process is known as drift rate and the variance per
unit time is variance rate
• A Wiener process has a drift rate of 0 and a variance
rate of 1
• In a GWP the drift rate and the variance rate can be
set equal to any chosen constants.
• The variable x follows a GWP with a drift rate of a
and a variance rate of b2 if
dx=a dt+b dz
where a and b are constants.
17
Generalized Wiener Processes – contd.

• Hence, change in the value of x in any time interval

T is normally distributed with:
Mean change in x = aT
Variance of change in x = b2T
Standard deviation of change in x = b√T

18
 Itô Process
• In an Itô process the drift rate and the
variance rate are functions of time
dx=a(x,t) dt+b(x,t) dz
• The discrete time equivalent is:

x  a( x, t )t  b( x, t ) t
is only true in the limit as t tends to
zero.

19
Itô’s Lemma
• If a variable x follows Itô process, then a
function of the variable say G (x, t ) follows
Itô’s lemma.
• Since a derivative security is a function of the
price of the underlying and time, Itô’s lemma
plays an important part in the analysis of
derivative securities.

20
Itô’s Lemma
If x follows the process dx  a dt  b dz
G G  2G 2  G
Then the process for G is dG   a ½ b dt  b dz
x t x x
2

This is Ito' s Lemma
Thus G also follows Ito process with :
G G  2G 2
Drift rate  a ½ b
x t x 2

2
G  2
Variance rate    b
x 

21
The Process for A Stock Price - The Discrete Time
Equivalent Model

S
 t   t
S
S  St  S t
This equation shows that S/S is normally distributed with
mean t and standard deviation  t
In other words,
S
~  ( t ,  t )
S

22
Application of Ito’s Lemma to a Stock Price Process

The stock price process is

dS  S dt  S dz
For a function G of S and t
 G G  2G 2 2  G
dG   S   ½ 2  S dt  S dz
 S t S  S

23
Application of Ito’s Lemma to a Stock Price Process

We know, G  ln S
Since ,
2
G 1  G 1 G
 ,  , 0
S S S 2
S 2 t
We get,
  2

dG    dt   dz
 2 
 
G follows Generalize d Wiener Process.

24
 Lognormal Property of Stock Price

• It follows from this assumption that

 2  
ln ST  ln S0      T,  T 
 2  
or
  2  
ln ST  ln S0     T,  T 
  2  

• Since the logarithm of ST is normal,

ST is lognormally distributed.
25
Problem No. 1

• Variables X1 and X2 follow Generalized Wiener

processes with drift rates µ1 and µ2 and
variances σ21 and σ22. What process does X1 +
X2 follow if:
a. The changes in X1 and X2 in any short interval of
time are uncorrelated?
b. There is a correlation  between the changes in X1
and X2 in any short time interval?

26
Problem No. 1 (Ans. – Part a.)

• Suppose that X1 and X2 equal a1 and a2 initially.

• After a time period of length T, X1 has the
probability distribution
Ф(a1 + µ1T, σ1√T)
• and X2 has the probability distribution
Ф(a2 + µ2T, σ2√T)
• From the property of sums of independent normally
distributed variables, X1 + X2 has the probability
distribution
Ф(a1 + µ1T + a2 + µ2T, √[(σ1)2T + (σ2)2T])
or, Ф[a1 + a2 + (µ1 + µ2)T, √{(σ1)2T + (σ2)2T}]

27
Problem No. 1 (Ans. – Part b.)

In this case the change in the value of X 1  X 2 in a

short time interval  t has the probabilit y distributi on :

  (  1   2 )  t, ( 12   22  2  1 2 )  t 


If  1 ,  2 ,  1 ,  2 and  are all constant, arguments show
that the change in a longer period of time T is :

  (  1   2 ) T, ( 12   22  2  1 2 )T 


The variable, X 1  X 2 , therefore follows a generalize d
Wiener process with drift rate  1   2 and variance rate
 12   2
2  2  1 2 .

28
Problem No. 2

• Suppose that a stock price, S, follows

geometric Brownian motion with expected
return μ, and volatility σ:
dS = μSdt + σSdz
what is the process followed by Sn? Show that
Sn also follows geometric Brownian motion.

29
Problem No. 2 (Ans.)

If G(S, t)  Sn , then
G G n 1  2
G n 2
 0,  nS and  n(n  1)S
t S S2
Using Ito' s lemma :
 1 
dG   nG  n(n  1) 2 G  dt   n Gdz
 2 
This shows that G  Sn follows geometric Brownian motion where
1
the expected return is n  n(n  1) 2
2
and the volatility is n .

30
Problem No. 3

• A company's cash position, measured in millions of

dollars, follows a generalized Wiener process with a
drift rate of 0.1 per month and a variance rate of
0.16 per month. The initial cash position is 2.0.
a. What are the probability distributions of the cash position
after one month, six months, and one year?
b. What are the probabilities of a negative cash position at
the end of six months and one year?
c. At what time in the future is the probability of a negative
cash position greatest?

31
Problem No. 3 (Ans.)

Part a :
The probability distributions are :
 (2.0  0.1, 0.16 )   (2.1, 0.4)
 (2.0  0.6, 0.16 * 6 )   (2.6, 0.98)
 (2.0  1.2, 0.16 * 12 )   (3.2, 1.39)

Part b :
The chance of a random sample from  (2.6, 0.98) being negative is
 2.6 
N    N 2.65
 0.98 

32
Problem No. 3 (Ans. – contd.)
Part b :  contd.
N(x)  Cumulative probability that a standardized normal variable i.e.,
a variable with probability distribution  (0, 1) is less than x
Now, N(2.65)  0.0040
Hence, the probability of a negative cash position at the end of six months
 0.04%
 3.2 
Again, N    N(2.30)  0.0107
 1.39 
Hence, the probability of a negative cash position at the end of one year
 1.07%
Part c :
In general the probability distribution of the cash position at the
end of x months is
 (2.0  0.1x, 0.4 x )

33
Problem No. 3 (Ans. – contd.)

Part c :  contd.
The probability of the cash position being negative is maximized when :
2.0  0.1x
is minimized.
0.4 x
1 1
2.0  0.1x 
Define y   5x 2  0.25x 2
0.4 x
3 1
dy  
  2.5x 2  0.125x 2
dx
3

x 2 ( 2.5  0.125x)

2
This is zero when x  20 and it is easy to verify that y
d
2  0 for this value of x.
dx
It therefore gives minimum value of y.
Hence, the probability of a negative cash position is greatest after 20 months.

34
 Thank You!

Indian Institute of Management Bangalore

Bannerghatta Road, Bangalore – 560 076, INDIA

www.iimb.ac.in