BKF4310 Financial

Modelling II
Brownian Motion
Dr Paul Magro

Introduction
BKF4310 is not a mathematics course, so we will not concern
ourselves with the calculus underpinning Brownian Motion.
Our focus is on understanding Brownian Motion, being able to
apply its properties in simple calculations by hand, and being
able to implement it in practice.
The knowledge of Brownian Motion gained here will be needed
later in the semester, most notably when covering Monte Carlo
Analysis and the Black-Scholes Option Pricing Model.
BKF4310 - FINANCIAL MODELLING
II

BROWNIAN MOTION

Introduction
Brownian Motion is the physical phenomenon named after the
English botanist who discovered it in 1827.
Brownian motion is the zig-zagging motion exhibited by a small
particle, such as a grain of pollen, immersed in a liquid or a
gas. Albert Einstein gave the first explanation of this
phenomenon in 1905.
It has widely become one of the most famous and fundamental
of Stochastic Processes. Since then the abstracted process has
been used for modelling the stock market and in quantum
mechanics.
BKF4310 - FINANCIAL MODELLING II
BROWNIAN MOTION
3

Brownian Motion
Interestingly, Brownian motion was independently introduced
in 1900 by the French mathematician Louis Bachelier, who
used it in his doctoral dissertation to model the price
movements of stocks and commodities.
Anticipating by 70 years developments inoptions pricing
theory, Bachelier mathematically defined Brownian motion and
proposed it as a model for asset price movements.
However, Brownian motion appears to have 2 major flaws
when used to model stock of commodity prices.
The price of a stock is a normal random variable, it can
theoretically become negative.
The assumption that a price difference over an interval of
BKF4310fixed
- FINANCIAL
MODELLING has
II
BROWNIAN MOTION
length
the same
normal distribution no matter4

Brownian Motion
Brownian Motion is a process that:
Involves chance;
Knowledge of todays stock price doesnt tell us what the
price will be tomorrow in a stochastic model, but it does tell
us something about the probabilities of various prices
occurring;
The opposite is a deterministic process (knowledge of the
todays price determines all future prices);

BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

Random Walk
A path of prices formed by taking random incremental steps at
each point in time.
In a simple random walk, we might allow the price to go up by
1 or down by 1 each day with equal probability, for example.
Increments of +/-1
Note: Graph taken from the spreadsheet which accompanies this
lecture. Press F9 in the spreadsheet to recalculate and
generate new paths [Excel Skills: Rand() and If() functions].
BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

Random Walk

BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

Standard Brownian Motion B(t)

Also referred to as a Wiener Process W(t).
The continuous-time version of a Random Walk.

drawn from the

same (timeinvariant)
distribution

A random walk with stationary independent increments.

Instead of the increments being fixed, we now draw them from
future
a normal distribution with mean 0 and standard deviation
1.
increments
this is a Standard Normal Distribution
dont depend on
Normal (Gaussian) Distribution:
characterised by mean and standard deviation;
symmetrical about the mean;
BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

past increments
(the probability
of a 10% return
today is the
same whether
yesterdays
return was 10%
or 0%)

Wiener Process
Norbert Wiener showed in the early 1920s, it can be described
directly in terms of a probability measure over a space of
continuous paths.
As Wiener showed, it is legitimate to talk about a random realvalued continuous function W on [0, ) such that:
Every increment W(t)W(s) over an interval of length ts is
normally distributed with mean 0 and variance t s, that is:
W(t) W(s) N(0, t s). for each t > 0, W(t) is Gaussian
with mean zero and variance t; and
if the intervals [t1, t2] and [u1, u2] do not overlap, then the
random variables W(t2) W(t1) and W(u2) W(u1) are
BKF4310 - FINANCIAL MODELLING II
9
BROWNIAN MOTION
independent.

Normal Distribution
Probability Density Function:

BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

X ~ N(, 2) means X is
drawn from a Normal
Distribution with mean
and variance 2
for example X ~ N(0.1, 0.3)

10

Standard Brownian Motion

Independent normally-distributed increments mean that:
Over 1 time step, the change is simply distributed as N(0,1):
Over 2 time steps the change is the sum of two independent
changes distributed as N(0,1):
To add two independent normally distributed variables, just
add the means and add the variances;
So the change over 2 time steps is distributed as N(0,2), the
change over 3 time steps N(0,3), and so on;
BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

11

Standard Brownian Motion B(t)

With increments N(0,1):

BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

12

Standard Brownian Motion B(t)

Calculating the probability of a particular value of B(t).
We can use the Excel NORMDIST function to calculate the
probability that B(t) x for particular values x and t).
IMPORTANT t is the variance of B(t). The NORMDIST function
requires us to input the standard deviation of the distribution,
which is the square root of the variance.
Prob [B(t) x ] = NORMDIST (x, 0, sqrt(t),TRUE)
Prob [B(t) > x] = 1 - Prob [B(t) x ]
BKF4310 - FINANCIAL
MODELLING II

BROWNIAN MOTION

13

Standard Brownian Motion B(t)

Example 1
What is Probability [B(5) 0.2] ?
NORMDIST (0.2, 0, sqrt(5), TRUE) = 0.5356
so the probability is 53.56%.
Intuitive check:
for x > 0 the probability will be > 50%;
for x < 0 the probability will be < 50%;
BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

14

Standard Brownian Motion B(t)

Example 2 scaling Standard Brownian Motion
We might want to consider B(t) instead of just B(t)
the advantage of this is that the variance is now
proportional to t rather than equal to t - useful if we want
to consider more/less volatile assets.

The following properties are important:

N (0, t ) t N (0,1)

B(t) ~ N (0,t)
B(t) ~ N(0, t)

N (0, 2t ) t N (0,1)

BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

15

Standard Brownian Motion B(t)

Example 2 scaling Standard Brownian Motion
What is Prob [3.5 B(15) > 1]?
Prob [3.5 B(15) > 1] = Prob [B(15) > 1/3.5]
= 1 Prob [B(15] 1/3.5]
NORMDIST (1/3.5, 0, sqrt(15), TRUE) = 0.5294
so the probability is 1 52.94% = 47.06%
Alternatively, we could have said that 3.5 B(15) is distributed
as N(0,3.52 15).
NORMDIST (1, 0, sqrt(3.52 15), TRUE) = 0.5294 so the
probability is 47.06% as before.
BKF4310 - FINANCIAL MODELLING
II

BROWNIAN MOTION

16

Using a Standard Normal Table

Standard Normal Tables give the probabilities for the standard
Normal distribution N(0,1).
Given B(t) ~ N(0,t), B(t) / ( sqrt(t)) ~ N(0,1).
Using this, we can look up the probabilities the old-fashioned
way using a standard normal distribution table if we dont have
Excel to hand (e.g. in an exam.. ).
Example 1:
we want to look up Z= 0.2 / sqrt(5) = 0.08944
nearest z value in the table is 0.09 => 0.46414
note this gives the area ABOVE, so we want 1-0.46414 =
0.53586.
BKF4310 - FINANCIAL
MODELLING II

BROWNIAN MOTION

17

Using a Standard Normal Table

B(t) / ( sqrt(t)) ~ N(0,1)
Example 2:
we want to look up z = 1 / (3.5 sqrt(15)) = 0.0738.
nearest z value in the table is 0.07 => 0.4721.
so the probability is 47.21%.
Note that we do get some rounding error when using the table
rather than Excel.
BKF4310 - FINANCIAL MODELLING
II

BROWNIAN MOTION

18

Arithmetic Brownian Motion

Brownian Motion with Drift
in modelling financial assets we often need to include a drift
term as well as a random component.
A(t) = A(0) + t + B(t)
A(0) is our starting value
is the drift term
scales our Standard Brownian Motion (allowing us, for
example, to make some assets returns more volatile than
others we have already seen examples of scaled Standard
Brownian Motion)
BKF4310 - FINANCIAL
MODELLING II

BROWNIAN MOTION

19

Arithmetic Brownian Motion A(t)

Example generated with A(0) = 10, = 0.05, = 0.3

BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

20

Arithmetic Brownian Motion A(t)

Calculating the probability of a particular value of A(t)
essentially we just work back to the required value of B(t)
and calculate the probability as before:
A(t) = A(0) + t + B(t) so A(t) A(0) = t + B(t)
note that A(t) A(0) ~ N (t, 2t)
so A(t) A(0) - t ~ N (0, 2t)
and (A(t) A(0) t) / ( sqrt(t)) ~ N (0, 1)
Lets say were interested in the distribution of A(25) where
A(0) = 10, = 0.05 and = 0.3, for example
A(10) A(0) ~ N (t, 2 t)
= N(0.05 25, 0.32 25)
BKF4310 - FINANCIAL MODELLING
= N (1.25, 2.25)
BROWNIAN MOTION
21
II

Arithmetic Brownian Motion A(t)

Example 1
What is Prob [A(25) 9]?
A(25) 9 implies A(25) A(0) 9 10 = -1.
A(25) A(0) is normally distributed with mean 1.25 and
variance 2.25.
NORMDIST (-1, 1.25, sqrt(2.25),TRUE) = 0.0668
so the probability is 6.68%.
BKF4310 - FINANCIAL MODELLING
II

BROWNIAN MOTION

22

Arithmetic Brownian Motion A(t)

Example 1:
Alternatively, (A(t) A(0) t) / ( sqrt(t)) is standard normal if we
calculate the required z we can look up the answer in the standard
normal table.
required z = (9 10 0.05 25) / (0.3 sqrt(25)) = -1.50.
Note: that the value for z = -1.5 is the same as 1 - the value for 1.50.
the standard normal distribution is symmetrical about 0.
think in terms of the area under the curve.
for z = 1.5 the table gives a probability of 0.06681.
for z = 1.5 the probability is 1 0.06681 but remember the table
gives values for > whereas we want so the probability is 6.681%.
BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

23

Geometric Brownian Motion

Modellers tend to prefer to assume that:
Stock prices S(t) are lognormal;
so log price ratios ln(S(t)/S(0)) are normal;
rather than price increments S(t)-S(0) being normal as
was the case for Arithmetic Brownian Motion;
(note I am using times t and 0 here rather than t and t-1
simply for continuity with the previous slides);
Geometric Brownian Motion enables us to do this
Well use Geometric Brownian Motion when we do Monte
Carlo Analysis and when we consider the Black Scholes
Option
Pricing
Model.BROWNIAN MOTION
BKF4310 - FINANCIAL
MODELLING
II
24

Geometric Brownian Motion

To make log Price Ratios normal we just set the LHS of our
Brownian Motion accordingly:
ln(Pt / P0)= t + B(t)
We want an expression for S(t).
Take the exponent of both sides of the above equation to get

S (t )

e t B ( t )
S ( 0)
So:

S (t ) S (0)e t B (t )
BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

25

Geometric Brownian Motion

With = 0.50% and = 4.00%:

BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

26

Lognormal Variables
Lognormal distributed variables have:
Expected Value (mean):

E( X ) e
Variance:

Var [ X ] e

1
2
2

1 e

Standard Deviation:

Stdev[ X ] Var [ X ] e
BKF4310 - FINANCIAL
MODELLING II

BROWNIAN MOTION

2 2

1
2
2

1
27

Lognormal Distribution
Probability Density Function:

f ( x)

BKF4310 - FINANCIAL MODELLING II

1
x 2

ln x 2

2 2

BROWNIAN MOTION

,x 0

28

Lognormal Variables
The drift term of our Brownian Motion, , is the continuously
compounded return corresponding to the MEDIAN of S(t);
Remember that S(t) is LOGNORMAL, so the expected stock
value is higher than the median value (mean > median);
The larger is, the longer the tail of the lognormal distribution
will be, and the higher the expected1 value
will be relative to
2

the median value generated by :
2

E( X ) e

BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

29

Lognormal Variables

E( X ) e

1 2

2

We can define =+2/2 as the continuously compounded

expected rate of return on the stock.
Now, in Finance, we often know the expected growth rate of
the stock we are modelling, rather than the drift term in the
Brownian motion we need to model it.
This is why sometimes, we see used as the expected return
2drift
of the stock price rather than the
of the Brownian Motion,

t B ( t )

so we see things like.

2

Pt P0 e

BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

30

Lognormal Variables
For the purposes of BKF4310 we will use to denote the drift of
the Brownian Motion, and to represent the continuously
compounded expected stock return.
= + 2/2
= 2/2
CARE!
If a question gives , this is the drift term to use in the Brownian
Motion.
If a question gives , we need to use = -2/2 as the drift in the
Brownian Motion.
BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

31

Geometric Brownian Motion

We can calculate the probability of S(t) taking certain values in
exactly the same way as we did for Standard Brownian Motion
and Arithmetic Brownian Motion.
Example 1
The price of a stock S(t) is to be modelled as a geometric
Brownian Motion with continuous stock return = 0.1 and
variance 2.
Given S(0) = 100, what is the probability of S(2) 105?

BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

32

Geometric Brownian Motion

Firstly, we calculate = 2/2 = 0.1-1 = -0.9

S (t ) S (0)e
So:

t B ( t )

S (t )
t B ( t )
e
S ( 0)

Taking the natural logarithm of both sides:

S (t )
t B (t )
ln
S ( 0)

BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

33

Geometric Brownian Motion

Rearranging and substituting for B(t):

S (t )
t tB (1)
ln
S ( 0)

So:

S (t )
t
ln
S ( 0)
t

B (1)

The LHS is the z value we need to use in our Normal Distribution Table.
BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

34

Geometric Brownian Motion

Substituting S(t) = 105:

105
ln
2 0.9
100
0.9244
2 2
0.92 corresponds to 0.1787. Remember this is the probability of
being ABOVE the cut-off, so we want 1 0.1787 = 82.13%.
Alternatively, NORMDIST(0.9244,0,1,TRUE) = 82.24% (more
accurate).
BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

35

Summary
This topic is about Geometric Brownian Motion and is commonly
used to model Stock Prices.
The tutorial questions accompanying this lecture provide practice in
simple calculations involving:
Standard Brownian Motion
Arithmetic Brownian Motion
Geometric Brownian Motion
We will use Brownian Motion when we go on to consider Monte
Carlo Methods in Finance.
BKF4310 - FINANCIAL
MODELLING II

BROWNIAN MOTION

36

Reading List
These lecture slides will suffice when revising material.

BKF4310 - FINANCIAL MODELLING II

BROWNIAN MOTION

37