Lecture 11

Models of asset dynamics

Reading: Luenberger Chapter 11
True multiperiod investments fluctuate in value, pay out random dividends, exist
in an environment of variable interest rates, and are subject to a continuing
variety of other uncertainties.
No investment principles in this chapter. Introduce mathematical models that
form the foundation for analyses developed later. The goal is to develop the
mathematical framework by means of which we can generate realistic stock
price movement.
Primary model types used to represent asset dynamics:
Binomial lattices are widely applicable. Many real investment
problems can be formulated and solved using the binomial lattice
framework. Textbook: 80% of the materials in later chapters are
presented in terms of binomial models.
Ito processes are more realistic than binomial lattice models in the
sense that they have a continuum of possible stock prices at each
period.

11.1 Binomial Lattice Model
t Period length
S Price at beginning of the period
Two possibilities of change
u multiple up, with u > 1
d multiple down, (usually) d < 1
Probabilities of change
p from S upwards to uS
1p from S downwards to dS
Model continues for several periods
Complete specification of the model
requires u, d and p.
Must capture the true stochastic nature
of the stock as faithfully as possible.
Will do so in terms of expected value
and of variance.
Multiplicative model, with u > 0 and d > 0 implies that the price will never be
negative. We can therefore use the logarithm of the price as the fundamental
variable.

Define as the expected yearly growth rate
0
ln
T
S
v E
S

=



For a deterministic process
0
ln
T
S
v E
S

=



0
vT
T
S S e =
Define as the yearly standard deviation
2
0
var ln
T
S
S

=



With t << 1, the parameters of the binomial lattice can be selected as
1 1
, ,
2 2
t t
v
p t u e d e

= + = =



With this choice, the binomial model will closely match the values of and .
The expected growth rate of ln S in the binomial model will be nearly , and
the variance of that rate will be nearly
2
. The closeness of the match improves
as t is made smaller, and is exact in the limit.
Example 11.1 (A volatile stock)
Consider a stock with the parameters = 15% and = 30%.
We wish to make a binomial model based on weekly periods.

0.30/ 52
1/ 52
1.04248
1/ 0.95925
1 0.15 1
1
2 0.30 52
t
u e
d u
p
=
= =
= =

= + = 0.534669



11.2 The Additive Model
Now consider models for which price can range over a continuum.
First discrete-time models, later continuous-time models defined by Ito
processes.
N + 1 time points indexed by k, k = 0, 1, 2, . . . , N.
Price of asset at k denoted by S(k)
Additive model is
2
(1) (0) (0)
(2) (1) (1) (0) (0) (1)
( ) (0) Sum of random variables
k
S aS u
S aS u a S au u
S k a S
= +
= + = + +

= +
M

It is sometimes assumed that u(k), k = 0, 1, 2, . . . , N-1 are independent normal
random variables with common variance
2
.
Then S(k) is a normal random variable.
If the expected values of all the u(k)s are zero then
( )
( ) (0)
k
E S k a S =


a > 1 The expected values grows geometrically
The additive model is structurally simple and easy to work with. The expected
values of the price grows geometrically and all the prices are normal random
variables.

Model is seriously flawed because it lacks realism:
Normal random variables can be negative

prices might be negative

Standard deviation should increase as the price increases
Now consider the multiplicative model, where we will use our understanding of
the additive model.
11.3 The Multiplicative Model
( 1) ( ) ( ), 0, , 1 S k u k S k k N + = = L
with u(k), k = 0, 1, 2, . . . , N-1 mutually independent random variables.
The relative change ( ) ( 1) / ( ) u k S k S k = + is independent of the units and of the
magnitude of S(k).
When we take the natural logarithm
ln ( 1) ln ( ) ln ( ) S k S k u k + = +
This is an additive model in terms of the logarithm of the price.
Specify random disturbances directly in terms of the ln ( ) u k s. Let
( ) ln ( ), 0, , 1 w k u k k N = = L
be mutually independent normal random variables each with expected value
( ) w k v = and variance
2
.
The original multiplicative disturbances
( )
( ) , 0, , 1
w k
u k e k N = = L
are lognormal random variables, since their logarithms are normal random
variables.
Negative values are no longer a problem: The w(k)s may be negative, but the
corresponding u(k)s are not. Prices remain positive.
Lognormal prices
1 1
0 0
( ) ( 1) ( 2) (0) (0)
ln ( ) ln (0) ln ( ) ln (0) ( )
k k
i i
S k u k u k u S
S k S u i S w i

= =
=
= + = +

L

If each w(i) has expected value ( ) w i v = and variance
2
, and all are mutually
independent, then
2
[ln ( )] ln (0)
var[ln ( )]
E S k S vk
S k k
= +
=

Both expected value and variance increase linearly with k.
Real Stock Distributions See Figure 11.3, p 302
Price distributions of most stocks are close to lognormal. Typically the tails are
fatter, which imply that large price changes tend to occur more frequently than
predicted by a normal distribution with the same variance.
11.4 Typical Parameter Values
Typical values for stock/share prices are
= 12%, = 15% over one year
If the period length is p part of a year then
,
p p
v pv p = =
With N+1 time points of price data, spanning N periods, we have the single
period v estimates
1 1
0 0
2
1
2
0
1 ( 1) 1 1 ( )
ln [ln ( 1) ln ( )] ln
( ) (0)
1 ( 1)
ln
1 ( )
N N
k k
N
k
S k S N
v S k S k
N S k N N S
S k
v
N S k

= =

=
+
= = + =



+
=

The error in these estimates is characterized by their variances
2
4
2
var( )
2
var( )
1
v
N
N

=
=

Suppose =0.12 and =0.15. Estimate with 10 years of data.
0.15
( ) 0.0474
10
v
N

= = =
With 10 years of data the standard deviation of the estimate of is reduced to
0.05, which is still about 40% of the true value.
Estimate
2
with 1 year of weekly data
2 4 2 4 4 2 2
2 2 2 2
2
52 2 52 2 2 2
var( ) 52 var( ) ( )
1 52 51 51 5
51
w
w
N



= = = = =


11.5 Lognormal Random Variables
Consider a few properties of lognormal random variables.
The distribution is defined for positive values of the random variable.
The shape is slightly skewed.
Suppose
2
~ ( , ) w N w . What is the expected value of
w
u e = ?
2
1
2
w
u e
+
=
The extra term
1
2
2
is small for low-volatility stocks.
For stocks of high volatility this correction can be significant.


11.6 Random Walks and Wiener Processes
Discrete Prices and Discrete Time: Binomial Lattice
Continuum of Prices and Discrete Time
Additive model ( 1) ( ) ( ) S k aS k u k + = +
Multiplicative model ( 1) ( ) ( ) S k u k S k + =
0
ln ( 1) ln ( ) ln ( )
( ) ln ( ), ( ) ln ( )
ln (0) ( )
k
i
S k u k S k
w k S k w k u k
S w i
=
+ = +
= + =
= +

2
[ ( )]
var[ ( )]
E w k v
w k
=
=

2
[ln ( )] ln (0)
var[ln ( )]
E S k S vk
S k k
= +
=

[ln ( ) ln ( )] ( ), E S k S j v k j j k = <
Section 11.7: The limit as the period
length goes to zero will give
continuous-time model.
First introduce special random
functions of time:
Random walks and Wiener processes
Random Walk

1
0
1
( ) ( ) ( ) , 0, ,
( ) 0
k k k
k k
z t z t t t k N
z t
t t t

+
+
= + =
=
= +
L

Standardized normal random variable ( ) ~ (0,1)
k
t N
Mutually uncorrelated [ ( ) ( )] 0, for
j k
E t t j k =
Consider the difference random variables ( ) ( ),
k j
z t z t j k <
1
1 1
2 2
1
2
( ) ( ) ( )
[ ( ) ( )] 0
var[ ( ) ( )] [ ( ) ] { [ ( )] }
[ ( )] ( )
k
k j i
i j
k j
k k
k j i i
i j i j
k
i k j
i j
z t z t t t
E z t z t
z t z t E t t E t t
t E t t k j t t

=

= =

=
=
=
= =
= = =

The variance of ( ) ( )
k j
z t z t is exactly equal to the time difference
k j
t t .
This explains the use of t .
Wiener Process
Wiener process obtained from random walk by letting t0.
Symbolically
( ) , ( ) ~ (0,1), [ ( ') ( ")] 0 for ' " dz t dt t N E t t t t = =
This intuitive description not rigorous.
Required properties of Wiener process or Brownian motion
1. ( ) ( ) ~ (0, ), z s z t N s t s t >
2.
2 1
( ) ( ) z t z t and
4 3
( ) ( ) z t z t are
uncorrelated for
1 2 3 4
t t t t < <
3.
0
Prob[ ( ) 0] 1 z t = =
Figure 11.6
A Wiener process is not differentiable, roughly motivated by
2
2
( ) ( ) 1
as
( )
z s z t s t
E s t
s t s t s t


= =




dz
dt
is white noise, Afrikaans: wit ruis


Generalized Wiener Processes and Ito Processes
The Wiener process is the fundamental building block for more general
processes, obtained by inserting white noise in an ordinary differential equation.
Generalized Wiener Process
( ) dx t adt bdz = +
The coefficients a and b are constants.
If we integrate both sides we obtain the explicit solution
( ) (0) ( ) x t x at bz t = + +
For the Ito process
( ) ( , ) ( , ) dx t a x t dt a x t dz = +
the coefficients a(x, t) and b(x, t) may depend on x and t.
A special form of the Ito process is frequently used to describe share price
processes.
11.7 A Stock Price Process
Extend multiplicative model to continuous-time model
Multiplicative model
ln ( 1) ln ( ) ( ) S k S k w k + =
with the ( ) w k s uncorrelated normal random variables and
2
( ) ~ ( , ) w k N v
Continuous-time version
ln ( ) d S t vdt dz = +
and constants, with 0 and z is a standard Wiener process.
The whole right-hand side plays the role of ( ) w k in the discrete-time model.
Right-hand side is a constant plus a normal random variable with zero mean.
Distinction differentials dt and dz, and small changes t and z.
Mean value
Right-hand side vdt , proportional to dt
Discrete-time [ln ( ) ln ( )] ( ), E S k S j v k j j k = <
Standard deviation
Right-hand side times standard deviation of dz,
order of magnitude dt
Discrete-time stdev[ln ( )] S k k =
Explicit solution of continuous-time model (a generalized Wiener process)
ln ( ) ln (0) ( )
[ln ( )] [ln (0)]
S t S vt z t
E S t E S vt
= + +
= +

Since the expected value of the logarithm grows linearly, like a continuously
compounding savings account, this process is called geometric Brownian
motion.

Lognormal Prices
The Geometric Brownian motion process is a lognormal process.
2
ln ( ) ln (0) ( ) ~ (ln (0) , ) S t S vt z t N S vt t = + + +
( ) S t has a lognormal distribution.
We can write
ln ( ) ( )
( ) (0)
S t vt z t
S t e S e
+
= =
but note that
2
1
( )
2
[ ( )] (0)
v t
E S t S e
+
=
Define
1
2
2
v = +
so that
2
1/2
[ ( )] (0) stdev[ ( )] (0) ( 1)
t t t
E S t S e S t S e e

= , =

Standard Ito Form
Random process in terms of ln ( ) S t generalized the multiplicative model.
It would be nice to express the process in terms of ( ) S t itself.
Calculus:
( )
ln ( )
( )
dS t
d S t
S t
=
Transformation of variables in Ito processes requires an extra term.
In the next section we will see that Itos lemma generalizes the chain rule.
The standard Ito form for price dynamics is
2
( ) 1
( )
( ) 2
dS t
v dt dz dt dz
S t
= + + = +
The term / dS S can be interpreted as the instantaneous rate of return.
The standard form is therefore an equation for the instantaneous return.

Example 11.2 (Bond price dynamics)
P(t) is the price of a bond that pays 1 at t = T, assume interest rate constant at r.
Price satisfies
( )
( )
( ) (0) ( )
( )
rt r t T
dP t
rdt P t P e P t e
P t

= = =
Relations for geometric Brownian motion
[ ] { } [ ] { }
[ ] [ ]
2
1/2
ln ( ) / (0) , stdev ln ( ) / (0)
( ) / (0) , stdev ( ) / (0) ( 1)
t t t
E S t S vt S t S t
E S t S e S t S e e

= =
= =

Simulation
Choose a basic period length t, set
0 0
( ) S t S = , a given initial price at
0
t t = .
Eqn (11.18) Standard Ito form
2
1
1
( ) 1
,
( ) 2
( ) ( ) ( ) ( ) ( )
( ) 1 ( ) ( )
k k k k k
k k k
dS t
dt dz v
S t
S t S t S t t S t t t
S t t t t S t



+
+
= + = +
= +

= + +


Eqn (11.15) Log of prices
1
( )
1
ln ( )
ln ( ) ln ( ) ( )
( ) ( )
k
k k k
v t t t
k k
d S t vdt dz
S t S t v t t t
S t e S t

+
+
+
= +
= +
=


11.8 Itos Lemma
There is a systematic way for making transformations of random processes.
Itos Lemma
Suppose that the random process x is defined by the Ito process
( ) ( , ) ( , ) dx t a x t dt b x t dz = +
where z is a standard Wiener process. Suppose also that the process ( ) y t is
defined by ( , ) y F x t = . Then ( ) y t satisfies the Ito equation
2
2
2
1
( )
2
F F F F
dy t a b dt bdz
x t x x

= + + +




where z is the same Wiener process as previously.

Sketch of Proof
Expand y with respect to a change y.
Retain up to first order in t.
Since x is of order t expand up to second order in x.
2
2
2
2
2
2
2
2
2
1
( , ) ( )
2
1
( , ) ( ) ( )
2
1
( , )
2
F F F
y y F x t x t x
x t x
F F F
F x t a t b z t a t b z
x t x
F F F F
F x t a b t b z
x t x x

+ = + + +


= + + + + +


= + + + +




Taking the limit, and using ( , ) y F x t = yields Itos equation.

Example 11.4 (Stock dynamics)
Suppose that S(t) is governed by the geometric Brownian motion
dS Sdt Sdz = +
Use Itos lemma to find the equation governing the process
( ( )) ln ( ) F S t S t =
With
2
2 2
1 1
, , ,
F F
a S b S
S S S S


= = = =

it follows that
2
2
2
1
ln
2
1
2
a b b
d S dt dz
S S S
dt dz

= +



= +




11.9 Binomial Lattice Revisited
The binomial lattice is analogous to the multiplicative model, since at each step
the price is multiplied by a random variable.
This random variable takes only two possible values u and d.
Therefore find values for u, d and p that match the multiplicative model as
closely as possible.
Do this by matching both the expected value of the logarithm of the price
change and the variance of the logarithm of the price change.
Sufficient to ensure that S
1
, the price after the first step, has the correct
properties. Use S
0
= 1, then we require
2
1 1
[ln ] , var[ln ] E S v t S t = =
1
2 2 2
1
2
[ln ] ln (1 )ln
var[ln ] (ln ) (1 )(ln ) [ ln (1 )ln ]
(1 )(ln ln )
E S p u p d
S p u p d p u p d
p p u d
= +
= + +
=

Matching equations become
2 2
(1 )
(1 )( )
pU p D v t
p p U D t
+ =
=

where ln , ln U u D d = = .
These are two equations in three unknowns. One way to use this degree of
freedom is to set D U = , which is equivalent to 1/ d u = .
With this choice the solution is
2 2
2 2
2 2
1 1 1
2 2
/ ( ) 1
ln ( )
ln ( )
p
v t
u t v t
d t v t

= +
+
= +
= +

For small t these equations can be approximated by
1 1
2 2
t
t
v
p t
u e
d e

= +


=
=

which are the values used earlier.


Example 11.1 (A volatile stock) revisited, using equations (11.25)
Again consider the stock with the parameters = 15% and = 30%. Construct a
binomial lattice using equations (11.25) for quarterly periods, thus with t = 1/4:

2 2
0.154616
0.154616
2 2
ln ( )
0.154616
1.167210
0.866744
1 1 1
2 2
/ ( ) 1
0.621268
u t v t
u e
d e
p
v t

= +
=
= =
= =
= +
+
=


Critical assumptions
Share price described by geometric Brownian motion.
Can constantly adjust hedging ratio with negligible dealing costs.
Share has infinite liquidity, ie we can buy or sell any amount we
want instantly.
The interest rate is known for the duration of the option - reasonable
if over a few months.
The volatility can be estimated accurately.
Come back to these later - meantime assume they are satisfied.
Note that we have not assumed any particular type of financial
derivative, so the equation has very general applicability.

1.
r
2.
3.
4.
5.
6.
7.

BlackScholes
ECG590I Asset Pricing. Lecture 12: Black-Scholes 1
12 Black-Scholes
12.1 The fundamental PDEs
1. Stock process (Geometric Brownian Motion)
dS = S dt + S dz
2. Option process
dF =

S
F
S
+
F
t
+
1
2

2
S
2

2
F
S
2

dt + S
F
S
dz
Denis Pelletier, North Carolina State University Compiled on November 8, 2006 at 09:19
ECG590I Asset Pricing. Lecture 12: Black-Scholes 2
3. Riskless portfolio
The foregoing dF equation is difcult because of the random term.
Convenient to eliminate that term by constructing a riskless portfolio:
Go short one derivative and go long N shares of S
Portfolio value is
V = F + NS
Its change is
dV = dF + NdS
=

S
F
S
+
F
t
+
1
2

2
S
2

2
F
S
2

dt S
F
S
dz
+NS dt + NSdz
To make this portfolio riskless, choose N =
F
S
dV =

F
t
+
1
2

2
S
2

2
F
S
2

dt
Denis Pelletier, North Carolina State University Compiled on November 8, 2006 at 09:19
ECG590I Asset Pricing. Lecture 12: Black-Scholes 3
4. Equality with riskless asset return
The last equation is the return from holding a risk-free portfolio.
Absence of arbitrage guarantee that this return equals the return from
putting the value of the portfolio into the risk-free asset:
dV = rV dt
We therefore must have

F
t
+
1
2

2
S
2

2
F
S
2

dt = r

F +
F
S
S

dt
Which can be written as
1
2

2
S
2

2
F
S
2
+ rS
F
S
rF +
F
t
= 0
This is the fundamental equation for all derivative pricing, known as
the Black-Scholes (-Merton) equation. Solving this gives F.
Denis Pelletier, North Carolina State University Compiled on November 8, 2006 at 09:19
ECG590I Asset Pricing. Lecture 12: Black-Scholes 4
The Black-Scholes equation applies to any derivative asset. What
distinguishes the different assets are the boundary conditions:
At S = 0, we have dS = 0, because dS = S dt + S dz.
The other boundary condition occurs at the expiration (or exercise point
in the case of an American option) and is different for each derivative
asset.
For example, a call options condition is
F(S, T) = max(S
T
K, 0)
Denis Pelletier, North Carolina State University Compiled on November 8, 2006 at 09:19
ECG590I Asset Pricing. Lecture 12: Black-Scholes 5
12.2 Solving the equation
1. Two approaches:
PDE with boundary conditions
Risk-neutral valuation (martingale theory). This approach is more
convenient for this problem.
2. Risk-neutral valuation
(a) Applies here because
There is a nite stopping time (the expiration date)
No element of investor risk preference enters the Black-Scholes
equation.
Note in particular that the expected return on the stock is
absent from the equation. The value of does depend on
risk preferences.
Denis Pelletier, North Carolina State University Compiled on November 8, 2006 at 09:19
ECG590I Asset Pricing. Lecture 12: Black-Scholes 6
(b) Useful implication
If risk preferences do not enter the equation, they do not affect the
solution.
Any set of risk preferences can be used when evaluating F.
We can use risk-neutral preferences
Discounting can be done with the risk-free rate r.
(c) Simple procedure
Assume the expected return from the underlying asset is the risk-free
rate (i.e, replace by r).
Calculate the expected payoff from the option at maturity.
Discount the expected payoff at the risk-free rate.
Denis Pelletier, North Carolina State University Compiled on November 8, 2006 at 09:19
ECG590I Asset Pricing. Lecture 12: Black-Scholes 7
(d) Note:
The assumption of risk-neutral preferences is a convenience. The
resulting solution is valid for any preferences, including those in the
real world.
When we move from risk-neutral to risk-averse preferences, two things
happen:
The expected growth rate of the stock changes
The discount rate changes
It turns out these two effects always exactly offset each other.
Denis Pelletier, North Carolina State University Compiled on November 8, 2006 at 09:19
ECG590I Asset Pricing. Lecture 12: Black-Scholes 8
(e) Simple example of how to use risk-neutral valuation to get a
derivative asset price:
Take a long forward contract with maturity date T and delivery price
K.
Value of the contract at maturity is S
T
K
Therefore, the value at time 0 is
F = e
rT
E

[S
T
K]
= e
rT
E

[S
T
] Ke
rT
Under risk-neutrality, the growth rate of the stock equals r, so we can
write
E

[S
T
] = S
0
e
rT
Substitute this result into the preceding expression for F to get
F = S
0
Ke
rT
which agrees with what we had derived earlier in the semester.
Denis Pelletier, North Carolina State University Compiled on November 8, 2006 at 09:19
Call option formula
The formula uses the function ( ) N x , the standard cumulative normal
probability distribution. This is the cumulative distribution of a normal
random variable having mean 0 and variance 1. It can be expressed as
2
2
/ 2
' / 2
1
( )
2
1
( )
2
x
y
x
N x e dy
N x e

=
=

The function ( ) N x is
illustrated in figure 11.1. The
value ( ) N x is the area under
the familiar bell-shaped curve from to x. Particular values are
( ) 0, (0) 0.5, ( ) 1 N N N = = =
Black-Scholes call option formula Consider a European call
option with strike price K and expiration time T. If the underlying stock
pays no dividends during the time [0, ] T and if interest is constant and
continuously compounded at a rate r, the BlackScholes solution is
( , ) ( , ) f S t c S t = , defined by

( )
1 2
( , ) ( ) ( )
r T t
c S t SN d Ke N d

= *
Where
2
1
ln( / ) ( / 2)( ) S K r T t
d
T t

+ +
=

2
2 1
ln( / ) ( / 2)( ) S K r T t
d d T t
T t

+
= =

t=T
1 2
,
,
S K
d d
S K
+ >

= =

<

d ln( / ) S K ( ) 0, ( ) 1 N N = =

,
( , )
0,
S K S K
c S T
S K
>

=

<

T
T = T =
( )
1
, 0
r T t
d e

= =
( , ) c S S =
* ( ) N d

* ( ) N d 1

( ) r T t
S Ke


( ) S PV K
S
( ) PV K
* ( ) N d
( ) N d 0 1 *

( ) N d

ln( / ) S K d
1
d
2

T t

1
( ) N d
2
( ) N d

ECG590I Asset Pricing. Lecture 12: Black-Scholes 16
5. Put option
Similar reasoning established the price for a European put option:
p = Ke
rT
N(d
2
) S
0
N(d
1
)
where d
1
and d
2
are the same as for the call option
An alternative proof is to use the put-call parity:
p + S
0
= c + Ke
rT
p = S
0
N(d
1
) Ke
rT
N(d
2
) + Ke
rT
S
0
= S
0
[1 N(d
1
)] + Ke
rT
[1 N(d
2
)]
= S
0
N(d
1
) + Ke
rT
N(d
2
)
since the normal distribution is symmetric
Denis Pelletier, North Carolina State University Compiled on November 8, 2006 at 09:19