Stochastic Processes
A stochastic process
X = { X(t) }
is a time series of random variables.
Stochastic Processes and Brownian Motion
X(t) (or Xt ) is a random variable for each time t and
is usually called the state of the process at time t.
A realization of X is called a sample path.
A sample path defines an ordinary function of t.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 396
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 398
Stochastic Processes (concluded)
Of all the intellectual hurdles which the human mind
has confronted and has overcome in the last
fifteen hundred years, the one which seems to me
to have been the most amazing in character and
the most stupendous in the scope of its
consequences is the one relating to
the problem of motion.
Herbert Butterfield (19001979)
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
If the times t form a countable set, X is called a
discrete-time stochastic process or a time series.
In this case, subscripts rather than parentheses are
usually employed, as in
X = { Xn }.
If the times form a continuum, X is called a
continuous-time stochastic process.
Page 397
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 399
Random Walks
The binomial model is a random walk in disguise.
Random Walk with Drift
Consider a particle on the integer line, 0, 1, 2, . . . .
Xn = + Xn1 + n .
In each time step, it can make one move to the right
with probability p or one move to the left with
probability 1 p.
n are independent and identically distributed with zero
mean.
This random walk is symmetric when p = 1/2.
Drift is the expected change per period.
Connection with the BOPM: The particles position
denotes the cumulative number of up moves minus that
of down moves.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 400
Note that this process is continuous in space.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 402
Martingalesa
{ X(t), t 0 } is a martingale if E[ | X(t) | ] < for
t 0 and
Position
4
E[ X(t) | X(u), 0 u s ] = X(s), s t.
2
20
40
60
80
Time
-2
In the discrete-time setting, a martingale means
E[ Xn+1 | X1 , X2 , . . . , Xn ] = Xn .
(40)
Xn can be interpreted as a gamblers fortune after the
nth gamble.
-4
Identity (40) then says the expected fortune after the
(n + 1)th gamble equals the fortune after the nth
gamble regardless of what may have occurred before.
-6
-8
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
(39)
a The
Page 401
origin of the name is somewhat obscure.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 403
Still a Martingale? (continued)
Martingales (concluded)
A martingale is therefore a notion of fair games.
Well, no.a
Apply the law of iterated conditional expectations to
both sides of Eq. (40) on p. 403 to yield
Consider this random walk with drift:
X
i1 + i , if i is even,
Xi =
Xi2 ,
otherwise.
E[ Xn ] = E[ X1 ]
(41)
for all n.
Similarly, E[ X(t) ] = E[ X(0) ] in the continuous-time
case.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 404
Above, n are random variables with zero mean.
a Contributed
by Mr. Zhang, Ann-Sheng (B89201033) on April 13,
2005.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 406
Still a Martingale? (concluded)
It is not hard to see that
X , if i is even,
i1
E[ Xi | Xi1 ] =
Xi1 , otherwise.
Still a Martingale?
Suppose we replace Eq. (40) on p. 403 with
E[ Xn+1 | Xn ] = Xn .
Hence it is a martingale by the new definition.
It also says past history cannot affect the future.
But
But is it equivalent to the original definition?
a Contributed
by Mr. Hsieh, Chicheng (M9007304) on April 13, 2005.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
X , if i is even,
i1
E[ Xi | . . . , Xi2 , Xi1 ] =
Xi2 , otherwise.
Hence it is not a martingale by the original definition.
Page 405
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 407
Example
Probability Measure (continued)
Consider the stochastic process
{ Zn
n
X
i=1
Xi , n 1 },
where Xi are independent random variables with zero
mean.
E[ X(u) | It ] = X(t)
for all u > t.
This process is a martingale because
The discrete-time version: For all n > 0,
E[ Zn+1 | Z1 , Z2 , . . . , Zn ]
= E[ Zn + Xn+1 | Z1 , Z2 , . . . , Zn ]
= E[ Zn | Z1 , Z2 , . . . , Zn ] + E[ Xn+1 | Z1 , Z2 , . . . , Zn ]
= Zn + E[ Xn+1 ] = Zn .
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
A stochastic process { X(t), t 0 } is a martingale with
respect to information sets { It } if, for all t 0,
E[ | X(t) | ] < and
Page 408
E[ Xn+1 | In ] = Xn ,
given the information sets { In }.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 410
Probability Measure
A martingale is defined with respect to a probability
measure, under which the expectation is taken.
Probability Measure (concluded)
A probability measure assigns probabilities to states
of the world.
The above implies E[ Xn+m | In ] = Xn for any m > 0
by Eq. (15) on p. 137.
A typical In is the price information up to time n.
A martingale is also defined with respect to an
information set.
In the characterizations (39)(40) on p. 403, the
information set contains the current and past values
of X by default.
Then the above identity says the FVs of X will not
deviate systematically from todays value given the
price history.
But it needs not be so.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 409
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 411
Example
Martingale Pricing
Consider the stochastic process { Zn n, n 1 }.
Pn
Zn i=1 Xi .
X1 , X2 , . . . are independent random variables with
mean .
Now,
Recall that the price of a European option is the
expected discounted future payoff at expiration in a
risk-neutral economy.
This principle can be generalized using the concept of
martingale.
Recall the recursive valuation of European option via
E[ Zn+1 (n + 1) | X1 , X2 , . . . , Xn ]
C = [ pCu + (1 p) Cd ]/R.
= E[ Zn+1 | X1 , X2 , . . . , Xn ] (n + 1)
= Zn + (n + 1)
p is the risk-neutral probability.
= Zn n.
$1 grows to $R in a period.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 412
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 414
Martingale Pricing (continued)
Example (concluded)
Define
Then
Let C(i) denote the value of the option at time i.
Consider the discount process
In { X1 , X2 , . . . , Xn }.
{ C(i)/Ri , i = 0, 1, . . . , n }.
Then,
pCu + (1 p) Cd
C
C(i + 1)
C(i) = C =
= i.
E
i+1
i+1
R
R
R
{ Zn n, n 1 }
is a martingale with respect to { In }.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 413
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 415
Martingale Pricing (continued)
It is easy to show that
C
C(k)
C(i)
=
C
= i , i k.
E
k
R
R
This formulation assumes:
Martingale Pricing (continued)
(42)
1. The model is Markovian in that the distribution of
the future is determined by the present (time i ) and
not the past.
2. The payoff depends only on the terminal price of the
underlying asset (Asian options do not qualify).
a Contributed
by Mr. Wang, Liang-Kai (Ph.D. student, ECE, University of Wisconsin-Madison) and Mr. Hsiao, Huan-Wen (B90902081) on
May 3, 2006.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 416
Equation (43) holds for all assets, not just options.
When interest rates are stochastic, the equation becomes
C(k)
C(i)
, i k.
(44)
= Ei
M (i)
M (k)
M (j) is the balance in the money market account at
time j using the rollover strategy with an initial
investment of $1.
So it is called the bank account process.
It says the discount process is a martingale under .
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 418
Martingale Pricing (concluded)
If interest rates are stochastic, then M (j) is a random
variable.
Martingale Pricing (continued)
In general, the discount process is a martingale in that
C(k)
C(i)
Ei
=
, i k.
(43)
Rk
Ri
Ei is taken under the risk-neutral probability
conditional on the price information up to time i.
This risk-neutral probability is also called the EMM, or
the equivalent martingale (probability) measure.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 417
M (0) = 1.
M (j) is known at time j 1.
Identity (44) on p. 418 is the general formulation of
risk-neutral valuation.
Theorem 14 A discrete-time model is arbitrage-free if and
only if there exists a probability measure such that the
discount process is a martingale. This probability measure is
called the risk-neutral probability measure.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 419
Futures Price under the BOPM
Futures prices form a martingale under the risk-neutral
probability.
The expected futures price in the next period is
1d
u1
pf F u + (1 pf ) F d = F
u+
d =F
ud
ud
(p. 380).
Can be generalized to
Martingale Pricing and Numeraire (concluded)
Choose S as numeraire.
Martingale pricing says there exists a risk-neutral
probability under which the relative price of any asset
C is a martingale:
C(i)
C(k)
, i k.
= Ei
S(i)
S(k)
S(j) denotes the price of S at time j.
Fi = Ei [ Fk ], i k,
So the discount process remains a martingale.
where Fi is the futures price at time i.
It holds under stochastic interest rates.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 420
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Example
Martingale Pricing and Numeraire
The martingale pricing formula (44) on p. 418 uses the
money market account as numeraire.a
It expresses the price of any asset relative to the
money market account.
In a period, asset ones price can go from S to S1 or
S2 .
Assume
S1
S2
S
<
<
P1
P
P2
to rule out arbitrage opportunities.
Suppose asset Ss value is positive at all times.
Walras (18341910).
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Take the binomial model with two assets.
In a period, asset twos price can go from P to P1 or
P2 .
The money market account is not the only choice for
numeraire.
a Leon
Page 422
Page 421
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 423
Example (continued)
Example (concluded)
For any derivative security, let C1 be its price at time
one if asset ones price moves to S1 .
Let C2 be its price at time one if asset ones price
moves to S2 .
It is easy to verify that
C
C1
C2
=p
+ (1 p)
.
P
P1
P2
Above,
Replicate the derivative by solving
S1 + P1
= C1 ,
S2 + P2
= C2 ,
The derivatives price using asset two as numeraire is
thus a martingale under the risk-neutral probability p.
The expected returns of the two assets are irrelevant.
using units of asset one and units of asset two.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
(S/P ) (S2 /P2 )
.
(S1 /P1 ) (S2 /P2 )
Page 424
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 426
Brownian Motiona
Brownian motion is a stochastic process { X(t), t 0 }
with the following properties.
Example (continued)
This yields
=
1. X(0) = 0, unless stated otherwise.
2. for any 0 t0 < t1 < < tn , the random variables
S2 C1 S1 C2
P2 C1 P1 C2
and =
.
P2 S1 P1 S2
S2 P1 S1 P2
X(tk ) X(tk1 )
The derivative costs
C
=
=
for 1 k n are independent.b
S + P
P S1 P1 S
P2 S P S2
C1 +
C2 .
P2 S1 P1 S2
P2 S1 P1 S2
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
3. for 0 s < t, X(t) X(s) is normally distributed
with mean (t s) and variance 2 (t s), where
and 6= 0 are real numbers.
a Robert
b So
Page 425
Brown (17731858).
X(t) X(s) is independent of X(r) for r s < t.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 427
Brownian Motion (concluded)
Such a process will be called a (, ) Brownian motion
with drift and variance 2 .
The existence and uniqueness of such a process is
guaranteed by Wieners theorem.a
Although Brownian motion is a continuous function of t
with probability one, it is almost nowhere differentiable.
The (0, 1) Brownian motion is also called the Wiener
process.
a Norbert
Brownian Motion Is a Random Walk in Continuous
Time
Claim 1 A (, ) Brownian motion is the limiting case of
random walk.
A particle moves x to the left with probability 1 p.
It moves to the right with probability p after t time.
Assume n t/t is an integer.
Its position at time t is
Y (t) x (X1 + X2 + + Xn ) .
Wiener (18941964).
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 428
(continued)
If { X(t), t 0 } is the Wiener process, then
X(t) X(s) N (0, t s).
Here
A (, ) Brownian motion Y = { Y (t), t 0 } can be
expressed in terms of the Wiener process:
(45)
Note that Y (t + s) Y (t) N (s, 2 s).
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 430
Brownian Motion as Limit of Random Walk
(continued)
Example
Y (t) = t + X(t).
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
+1 if the ith move is to the right,
Xi
1 if the ith move is to the left.
Xi are independent with
Prob[ Xi = 1 ] = p = 1 Prob[ Xi = 1 ].
Recall E[ Xi ] = 2p 1 and Var[ Xi ] = 1 (2p 1)2 .
Page 429
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 431
Geometric Brownian Motion
Brownian Motion as Limit of Random Walk
(continued)
Let X { X(t), t 0 } be a Brownian motion process.
Therefore,
The process
E[ Y (t) ] = n(x)(2p 1),
Var[ Y (t) ] = n(x)2 1 (2p 1)2 .
is called geometric Brownian motion.
With x t and p [ 1 + (/) t ]/2,
E[ Y (t) ] = n t (/) t = t,
Var[ Y (t) ] = n 2 t 1 (/)2 t 2 t,
Suppose further that X is a (, ) Brownian motion.
X(t) N (t, 2 t) with moment generating function
h
i
2 2
E esX(t) = E [ Y (t)s ] = ets+( ts /2)
from Eq. (16) on p 139.
as t 0.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
{ Y (t) eX(t) , t 0 },
Page 432
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 434
Brownian Motion as Limit of Random Walk
(concluded)
Thus, { Y (t), t 0 } converges to a (, ) Brownian
motion by the central limit theorem.
Geometric Brownian Motion (continued)
Brownian motion with zero drift is the limiting case of
symmetric random walk by choosing = 0.
Note that
Var[ Y (t + t) Y (t) ]
=Var[ x Xn+1 ] = (x)2 Var[ Xn+1 ] 2 t.
In particular,
2
E[ Y (t) ] = et+( t/2) ,
Var[ Y (t) ] = E Y (t)2 E[ Y (t) ]2
2
2
= e2t+ t e t 1 .
Similarity to the the BOPM: The p is identical to the
probability in Eq. (23) on p. 235 and x = ln u.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 433
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 435
Geometric Brownian Motion (concluded)
Y(t)
Then
6
5
ln Yn =
n
X
ln Xi
i=1
is a sum of independent, identically distributed random
variables.
2
1
0.2
0.4
0.6
0.8
Thus { ln Yn , n 0 } is approximately Brownian motion.
Time (t)
And { Yn , n 0 } is approximately geometric
Brownian motion.
-1
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 436
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 438
Geometric Brownian Motion (continued)
It is useful for situations in which percentage changes
are independent and identically distributed.
Let Yn denote the stock price at time n and Y0 = 1.
Assume relative returns
Xi
Continuous-Time Financial Mathematics
Yi
Yi1
are independent and identically distributed.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 437
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 439
Stochastic Integrals (concluded)
Typical requirements for X in financial applications are:
Rt
Prob[ 0 X 2 (s) ds < ] = 1 for all t 0 or the
Rt
stronger 0 E[ X 2 (s) ] ds < .
A proof is that which convinces a reasonable man;
a rigorous proof is that which convinces an
unreasonable man.
Mark Kac (19141984)
The information set at time t includes the history of
X and W up to that point in time.
But it contains nothing about the evolution of X or
W after t (nonanticipating, so to speak).
The pursuit of mathematics is a
divine madness of the human spirit.
Alfred North Whitehead (18611947),
Science and the Modern World
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
The future cannot influence the present.
{ X(s), 0 s t } is independent of
{ W (t + u) W (t), u > 0 }.
Page 440
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 442
Stochastic Integrals
Use W { W (t), t 0 } to denote the Wiener process.
The goal is to develop integrals of X from a class of
stochastic processes,a
Z t
It (X)
X dW, t 0.
A theory of stochastic integration.
As with calculus, it starts with step functions.
It (X) is a random variable called the stochastic integral
of X with respect to W .
The stochastic process { It (X), t 0 } will be denoted
R
by X dW .
a Ito
Ito Integral
A stochastic process { X(t) } is simple if there exist
0 = t0 < t1 < such that
X(t) = X(tk1 ) for t [ tk1 , tk ), k = 1, 2, . . .
for any realization (see figure next page).
(1915).
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 441
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 443
Ito Integral (continued)
af
X t
Let X = { X(t), t 0 } be a general stochastic process.
Then there exists a random variable It (X), unique
almost certainly, such that It (Xn ) converges in
probability to It (X) for each sequence of simple
stochastic processes X1 , X2 , . . . such that Xn converges
in probability to X.
If X is continuous with probability one, then It (Xn )
converges in probability to It (X) as
n max1kn (tk tk1 ) goes to zero.
t0
t1
t2
t3
t4
t5
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 444
It is a fundamental fact that
almost surely.
The Ito integral of a simple process is defined as
n1
X
k=0
X(tk )[ W (tk+1 ) W (tk ) ],
(46)
where tn = t.
X dW is continuous
The following theorem says the Ito integral is a
martingale.
A corollary is the mean value formula
"Z
#
b
E
X dW = 0.
The integrand X is evaluated at tk , not tk+1 .
Define the Ito integral of more general processes as a
limiting random variable of the Ito integral of simple
stochastic processes.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 446
Ito Integral (concluded)
Ito Integral (continued)
It (X)
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Theorem 15 The Ito integral
Page 445
X dW is a martingale.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 447
Discrete Approximation
Recall Eq. (46) on p. 445.
b } can be
The following simple stochastic process { X(t)
used in place of X to approximate the stochastic
Rt
integral 0 X dW ,
X
X$
b
X(s)
X(tk1 ) for s [ tk1 , tk ), k = 1, 2, . . . , n.
b
Note the nonanticipating feature of X.
X
Y$
(a)
The information up to time s,
(b)
b
{ X(t),
W (t), 0 t s },
cannot determine the future evolution of X or W .
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 448
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 450
Ito Process
Discrete Approximation (concluded)
The stochastic process X = { Xt , t 0 } that solves
Z t
Z t
Xt = X0 +
a(Xs , s) ds +
b(Xs , s) dWs , t 0
Suppose we defined the stochastic integral as
n1
X
k=0
X(tk+1 )[ W (tk+1 ) W (tk ) ].
is called an Ito process.
Then we would be using the following different simple
stochastic process in the approximation,
Yb (s) X(tk ) for s [ tk1 , tk ), k = 1, 2, . . . , n.
X0 is a scalar starting point.
{ a(Xt , t) : t 0 } and { b(Xt , t) : t 0 } are
stochastic processes satisfying certain regularity
conditions.
The terms a(Xt , t) and b(Xt , t) are the drift and the
diffusion, respectively.
This clearly anticipates the future evolution of X.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 449
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 451
Euler Approximation
Ito Process (continued)
The following approximation follows from Eq. (48),
A shorthanda is the following stochastic differential
equation for the Ito differential dXt ,
dXt = a(Xt , t) dt + b(Xt , t) dWt .
(47)
Or simply dXt = at dt + bt dWt .
This is Brownian motion with an instantaneous drift at
and an instantaneous variance b2t .
X is a martingale if the drift at is zero by Theorem 15
(p. 447).
a Paul
Langevin (1904).
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 452
Ito Process (concluded)
where tn nt.
It is called the Euler or Euler-Maruyama method.
b n ) converges to X(tn ).
Under mild conditions, X(t
Recall that W (tn ) should be interpreted as
W (tn+1 ) W (tn ) instead of W (tn ) W (tn1 ).
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 454
Under fairly loose regularity conditions, approximation
(49) on p. 454 can be replaced by
(48)
where N (0, 1).
This formulation makes it easy to derive Monte Carlo
simulation algorithms.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
b n ), tn ) t + b(X(t
b n ), tn ) W (tn ),
b n ) + a(X(t
=X(t
(49)
More Discrete Approximations
dW is normally distributed with mean zero and
variance dt.
An equivalent form to Eq. (47) is
dXt = at dt + bt dt ,
b n+1 )
X(t
Page 453
b n+1 )
X(t
b n ) + a(X(t
b n ), tn ) t + b(X(t
b n ), tn ) t Y (tn ).
=X(t
Y (t0 ), Y (t1 ), . . . are independent and identically
distributed with zero mean and unit variance.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 455
More Discrete Approximations (concluded)
A simpler discrete approximation scheme:
Trading and the Ito Integral (concluded)
b n+1 )
X(t
The equivalent Ito integral,
Z T
Z
GT ()
t dS t =
b n ) + a(X(t
b n ), tn ) t + b(X(t
b n ), tn ) t .
=X(t
Prob[ = 1 ] = Prob[ = 1 ] = 1/2.
T
0
t t dt +
T
0
t t dWt ,
measures the gains realized by the trading strategy over
the period [ 0, T ].
Note that E[ ] = 0 and Var[ ] = 1.
This clearly defines a binomial model.
b converges to X.
As t goes to zero, X
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 456
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 458
Itos Lemma
Trading and the Ito Integral
A smooth function of an Ito process is itself an Ito process.
Consider an Ito process dS t = t dt + t dWt .
S t is the vector of security prices at time t.
Let t be a trading strategy denoting the quantity of
each type of security held at time t.
Hence the stochastic process t S t is the value of the
portfolio t at time t.
Theorem 16 Suppose f : R R is twice continuously
differentiable and dX = at dt + bt dW . Then f (X) is the
Ito process,
f (Xt )
= f (X0 ) +
1
+
2
t dS t t (t dt + t dWt ) represents the change in the
value from security price changes occurring at time t.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
f (Xs ) as ds +
0
f (Xs ) bs dW
f (Xs ) b2s ds
for t 0.
Page 457
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 459
Itos Lemma (continued)
Itos Lemma (continued)
In differential form, Itos lemma becomes
df (X) = f (X) a dt + f (X) b dW +
1
f (X) b2 dt.
2
(50)
Compared with calculus, the interesting part is the third
term on the right-hand side.
A convenient formulation of Itos lemma is
df (X) = f (X) dX +
df (X) =
1
f (X)(dX)2 .
2
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Theorem 17 (Higher-Dimensional Itos Lemma) Let
W1 , W2 , . . . , Wn be independent Wiener processes and
X (X1 , X2 , . . . , Xm ) be a vector process. Suppose
f : Rm R is twice continuously differentiable and Xi is
P
an Ito process with dXi = ai dt + nj=1 bij dWj . Then
df (X) is an Ito process with the differential,
dW
dt
dt
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 462
The multiplication table for Theorem 17 is
The (dW )2 = dt entry is justified by a known result.
This form is easy to remember because of its similarity
to the Taylor expansion.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
k=1
Itos Lemma (continued)
We are supposed to multiply out
(dX)2 = (a dt + b dW )2 symbolically according to
dt
1 XX
fik (X) dXi dXk ,
2 i=1
where fi f /xi and fik 2 f /xi xk .
Itos Lemma (continued)
dW
fi (X) dXi +
i=1
Page 460
m
X
Page 461
in which
ik
dWi
dt
dWk
ik dt
dt
1 if i = k,
=
0 otherwise.
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 463
Geometric Brownian Motion
Itos Lemma (continued)
Theorem 18 (Alternative Itos Lemma) Let
W1 , W2 , . . . , Wm be Wiener processes and
X (X1 , X2 , . . . , Xm ) be a vector process. Suppose
f : Rm R is twice continuously differentiable and Xi is
an Ito process with dXi = ai dt + bi dWi . Then df (X) is the
following Ito process,
m
X
X(t) is a (, ) Brownian motion.
Hence dX = dt + dW by Eq. (45) on p. 429.
As Y /X = Y and 2 Y /X 2 = Y , Itos formula (50)
on p. 460 implies
dY
Consider the geometric Brownian motion process
Y (t) eX(t)
1 XX
fi (X) dXi +
df (X) =
fik (X) dXi dXk .
2 i=1
i=1
= Y ( dt + dW ) + (1/2) Y ( dt + dW )2
k=1
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
= Y dX + (1/2) Y (dX)2
= Y ( dt + dW ) + (1/2) Y 2 dt.
Page 464
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 466
Itos Lemma (concluded)
Geometric Brownian Motion (concluded)
The multiplication table for Theorem 18 is
dWi
dt
dWk
ik dt
dt
Hence
The annualized instantaneous rate of return is + 2 /2
not .
Here, ik denotes the correlation between dWi and
dWk .
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
dY
= + 2 /2 dt + dW.
Y
Page 465
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 467
Product of Geometric Brownian Motion Processes
Let
dY /Y
= a dt + b dWY ,
dZ/Z
= f dt + g dWZ .
Product of Geometric Brownian Motion Processes
(concluded)
ln U is Brownian motion with a mean equal to the sum
of the means of ln Y and ln Z.
Consider the Ito process U Y Z.
Apply Itos lemma (Theorem 18 on p. 464):
dU
Z dY + Y dZ + dY dZ
ZY (a dt + b dWY ) + Y Z(f dt + g dWZ )
This holds even if Y and Z are correlated.
Finally, ln Y and ln Z have correlation .
+Y Z(a dt + b dWY )(f dt + g dWZ )
=
U (a + f + bg) dt + U b dWY + U g dWZ .
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 468
Product of Geometric Brownian Motion Processes
(continued)
Y
Z
U
= exp
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Quotients of Geometric Brownian Motion Processes
Let U Y /Z.
We now show that
a b /2 dt + b dWY ,
= exp f g 2 /2 dt + g dWZ ,
= exp a + f b2 + g 2 /2 dt + b dWY + g dWZ .
Page 470
Suppose Y and Z are drawn from p. 468.
The product of two (or more) correlated geometric
Brownian motion processes thus remains geometric
Brownian motion.
Note that
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 469
dU
= (a f + g 2 bg) dt + b dWY g dWZ .
U
(51)
Keep in mind that dWY and dWZ have correlation .
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 471
Quotients of Geometric Brownian Motion Processes
(concluded)
The multidimensional Itos lemma (Theorem 18 on
p. 464) can be employed to show that
dU
=
(1/Z) dY (Y /Z 2 ) dZ (1/Z 2 ) dY dZ + (Y /Z 3 ) (dZ)2
(1/Z)(aY dt + bY dWY ) (Y /Z 2 )(f Z dt + gZ dWZ )
(1/Z 2 )(bgY Z dt) + (Y /Z 3 )(g 2 Z 2 dt)
U (a dt + b dWY ) U (f dt + g dWZ )
U (bg dt) + U (g 2 dt)
U (a f + g 2 bg) dt + U b dWY U g dWZ .
c
2006
Prof. Yuh-Dauh Lyuu, National Taiwan University
Page 472
