You are on page 1of 26

STOCK PRICE

MOVEMENT
The Basics

STOCK PRICE PROCESS

The most prevalent model

Geometric Brownian motion model:

S (t )

1
( 2 ) t ( t ) t
2
S ( 0) e

where
: price of stock at time t.
S (t:)price of stock at time 0.
S (0:)N(0,1)
(t )mean differential return per unit time.
std. dev. of differential return per unit time.

Geometric Brownian
Model

How? And Why? (Open the black box)


Background:

Discrete time process multiplicative


model, random walks.
Continuous time process Wieners
process, generalized Wieners process,
Ito process.

Binomial Lattice Model.

DISCRETE TIME
MODELS

Multiplicative Model:
S(k+1) = u(k)*S(k),

where k = 0, 1,,N-1 are discrete points in


time.
u(k) are mutually independent random
variables, for k = 0, 1,,N-1.

Observe the memory-less property of


stock price movements.

Multiplicative Model

S(k+1) = u(k)*S(k).
Ln [S(k+1)] = Ln [u(k)] + Ln [S(k)].

Let, w(k) = Ln [u(k)], and


w(k) ~ N(
u(k) = ew(k), is lognormal.

Lognormal Prices

Since,
S(k) = u(k-1)*S(k-1),
S(k) = u(k-1)*u(k-2)u(0)*S(0).
Ln [S(k)] = Ln [u(k-1)] + + Ln [u(0)]
+ Ln [S(0)].
= w(k-1) ++ w(0) + Ln [S(0)].

Lognormal Prices

Ln [S(k)] = w(k-1) ++ w(0) + Ln [S(0)].

Since, each w(k) is independent, Ln [S(k)]


is normally distributed for all k.
Hence, stock prices remain lognormal at
all future points in time.
Nonnegativity of stock prices are
ensured.

Lognormal Prices

If E [w(k)] = and Var [w(k)] = for all


k, then:
E [Ln S(k)] = Ln S(0) + k
Var [Ln S(k)] = k
Mean and Variance change linearly
with time.
Empirically, stock prices seem to
subscribe to such a movement.

Lognormal Random
Variables
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0

1.924999952

3.849999905

5.774999857

7.699999809

Lognormal Random
Variables
Suppose,
u - Lognormal random variable.
w

ue ,
2

where w ~ N (v , ).

Is E[u] e ?
1 2
v
E [ u] e 2 .

DISCRETE TIME
MODELS

Stock prices display a memory-less


property.
Stock prices are random, and they
follow a log-normal distribution at
all times.
Log of stock prices follow a normal
distribution whose mean and
variance change linearly with time.

Stock Price Process

Recall that in the multiplicative


model:
Ln [S(k+1)] - Ln [S(k)] = Ln [u(k)]
= w(k).
In the continuous-time version:
d Ln [S(t)] = dt + dz.

Generalized Wiener
Process

It is a stochastic process described as:


dx(t) = a dt + b dz,
where a and b are constants, and z
defines a Wiener process.
The generalized Wiener process has
the following analytical solution:
x(t) = x(0) + a t + b z(t).

Stock Price Process

Below is a Generalized Wiener Process


d Ln [S(t)] = dt + dz.
The analytical solution to the above is:
Ln [S(t)] = Ln [S(0)] + t + z(t).
Hence we get,

t ( t ) t

S (t ) S (0)e

- A Geometric Brownian motion.

Geometric Brownian
Motion
CHARACTERISTICS:

i ) E[ S (t )]

1 2
(v ) t
S ( 0) e 2

1 2
Let, .
2

Then,

E[ S (t )] S (0)e

Geometric Brownian
Motion
CHARACTERISTICS:

ii ) St .Dev[ S (t )] S (0)e
1 2
.
2

( )t

(e

2t

1/ 2

1)

Ito Process

The Ito Process is:


dx(t) = a(x,t) dt + b(x,t) dz. (1)
Clearly, the stock price process below,
d S(t) = S(t)dt + S(t)dz.
(2)
is an Ito process.
Using Itos Lemma, the geometric
Brownian motion is derived from (2).

Binomial Lattice Model

A Discrete Approximation Model.

t=0 t
2 t
3t
T
Suppose at t=0 stock price is known
to be S.
In the next time period, t, only
two possible prices: uS or dS, with
u>1, 0<d<1.

Binomial Lattice Model

A Discrete Approximation Model.

t=0 t
2 t
3t
T
Since u>1, it represents upward
movement of stock price, while d<1
represents downward movement.
Prob.(u) = p, Prob.(d) = 1-p.This
process holds true for all time
periods.

Binomial Lattice Model


u3S
u2S
uS
S

u2dS
udS

ud2S

u4S
u3dS
u2d2S

dS
ud3S

dS
2

d3S

d4S

Binomial Lattice
Approximation

Approximation by matching the


means and variances of the
binomial lattice and the
multiplicative models,
respectively.
Matching is done with respect to
the natural log (Ln) of the price
change.

Binomial Lattice
Approximation
Suppose that at time t=0, S(0) = 1.
At time t = t (call it t = 1):
Binomial Process:
E[ Ln S(1)] = p Ln u + (1-p) Ln d.
Var[Ln S(1)] = p(1-p)(Ln u Ln d)2.

Binomial Lattice
Approximation
Recall that for
Multiplicative Model:
E[ Ln S(1)] = t.
Var[Ln S(1)] = 2 t.

Binomial Lattice
Approximation

Let U = Ln u, and D = Ln d.
Matching the mean and variance,
we get:
p U + (1-p) D = t.
p(1-p)(U D)2 = 2 t.
Unknown parameters U, D,and p.
Hence, set U = - D.

Binomial Lattice
Approximation

Solving for the two equations we


1
get:
i)

ii )

1
p
2

2 /( 2 t ) 1

Ln u t (t )
2

iii ) Ln d t (t )

Binomial Lattice
Approximation

When t becomes very small then:

i)

ii )

1 1
p ( ) t
2 2

ue

iii ) d e

You might also like