Continuous Time Finance: Tomas BJ Ork Stockholm School of Economics

Continuous Time Finance
Lisbon 2013
Tomas Bj
ork
Stockholm School of Economics
Tomas Bjork, 2013

Contents
Stochastic Calculus (Ch 4-5).

Black-Scholes (Ch 6-7.
Completeness and hedging (Ch 8-9.
The martingale approach (Ch 10-12).
Incomplete markets (Ch 15).
Dividends (Ch 16).
Currency derivatives (Ch 17).
Stochastic Control Theory (Ch 19)
Martingale Methods for Optimal Investment (Ch 20)
Textbook:
Bj
ork, T: Arbitrage Theory in Continuous Time
Oxford University Press, 2009. (3:rd ed.)
Tomas Bj
ork, 2013 1
Stochastic Integrals
(Ch 4-5)
Tomas Bj
ork
Tomas Bj
ork, 2013 2
Typical Setup
Take as given the market price process, S, of some

underlying asset.
St = price, at t, per unit of underlying asset
Consider a fixed financial derivative, e.g. a European

call option.
Main Problem: Find the arbitrage free price of the

derivative.
Tomas Bj
ork, 2013 3
We Need:
1. Mathematical model for the underlying price

process. (The Black-Scholes model)
2. Mathematical techniques to handle the price

dynamics. (The It
o calculus.)
Tomas Bj
ork, 2013 4
Stochastic Processes
We model the stock price St as a stochastic

process, i.e. it evolves randomly over time.
We model S as a Markov process, i.e. in order

to predict the future only the present value is of
interest. All past information is already incorporated
into todays stock prices. (Market efficiency).
Stochastic variable
Choosing a number at random
Stochastic process
choosing a curve (trajectory) at random.
Tomas Bj
ork, 2013 5
Notation
Xt = any random process,

dt = small time step,
dXt = Xt+dt Xt
We often write X(t) instead of Xt .
dXt is called the increment of X over the interval

[t, t + dt].
For any fixed interval [t, t + dt], the increment dXt

is a stochastic variable.
If the increments dXs and dXt, over the disjoint

intervals [s, s + ds] and [t, t + dt] are independent,
then we say that X has independent increments.
If every increment has a normal distribution we say

that X is a normal, or Gaussian process.
Tomas Bj
ork, 2013 6
The Wiener Process
A stochastic process W is called a Wiener process if

it has the following properties
The increments are normally distributed: For s < t:
Wt Ws N [0, t s]
E[Wt Ws] = 0, V ar[Wt Ws] = t s
W has independent increments.
W0 = 0
W has continuous trajectories.
Continuous random walk
Note: In Hull, a Wiener process is typically denoted

by Z instead of W .
Tomas Bj
ork, 2013 7
A Wiener Trajectory
0.8
0.6
0.4
0.2
0.2
0.4 t
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Tomas Bj
ork, 2013 8
Important Fact
Theorem:
A Wiener trajectory is, with probability one, a
continuous curve which is nowhere differentiable.
Proof. Hard.
Tomas Bj
ork, 2013 9
Wiener Process with Drift
A stochastic process X is called a Wiener process

with drift and diffusion coefficient if it has the
following dynamics
dXt = dt + dWt,
where and are constants.
Summing all increments over the interval [0, t] gives us
Xt X0 = t + (Wt W0 ),
Xt = X0 + t + Wt
Thus
Xt N [X0 + t, 2 t]
Tomas Bj
ork, 2013 10
It
o processes
We say, losely speaking, that the process X is an It

o
process if it has dynamics of the form
dXt = tdt + t dWt,
where t and t are random processes.
Informally you can think of dWt as a random variable

of the form
dWt N [0, dt]
To handle expressions like the one above, we need

some mathematical theory.
First, however, we present an important example,

which we will discuss informally.
Tomas Bj
ork, 2013 11
Example: The Black-Scholes model
Price dynamics: (Geometrical Brownian Motion)
dSt = Stdt + StdWt,
Simple analysis:
Assume that = 0. Then
dSt = Stdt
Divide by dt!
dSt
= St
dt
This is a simple ordinary differential equation with
solution
St = s0 et
Conjecture: The solution of the SDE above is a

randomly disturbed exponential function.
Tomas Bj
ork, 2013 12
Intuitive Economic Interpretation
dSt
= dt + dWt
St
Over a small time interval [t, t + dt] this means:
Return = (mean return)

+ (Gaussian random disturbance)
The asset return is a random walk (with drift).
= mean rate of return per unit time
= volatility
Large = large random fluctuations
Small = small random fluctuations
The returns are normal.
The stock price is lognormal.
Tomas Bj
ork, 2013 13
A GBM Trajectory
10
0 t
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Tomas Bj
ork, 2013 14
Stochastic Differentials and Integrals
Consider an expression of the form
dXt = tdt + t dWt,

X0 = x0
Question: What exactly do we mean by this?
Answer: Write the equation on integrated form as

Z t Z t
Xt = x0 + sds + sdWs
0 0
How is this interpreted?
Tomas Bj
ork, 2013 15
Recall:
Z t Z t
Xt = x0 + sds + sdWs
0 0
Two terms:
Z t
sds
0
This is a standard Riemann integral for each -
trajectory.
Z t
sdWs
0
Stochastic integral. This can not be interpreted

as a Stieljes integral for each trajectory. We need a
new theory for this It o integral.
Tomas Bj
ork, 2013 16
Information
Consider a Wiener process W .

Def:
FtW = The information generated by W
over the interval [0, t]
Def: Let Z be a stochastic variable. If the value of Z

is completely determined by FtW , we write
Z FtW
Ex:
For the stochastic variable Z, defined by
Z 5
Z= Wsds,
0
we have Z F5W .
We do not have Z F4W .
Tomas Bj
ork, 2013 17
Adapted Processes
Let W be a Wiener process.
Definition:
A process X is
adapted to the filtration
FtW : t 0 if
Xt FtW , t 0
An adapted process does not look

into the future
Adapted processes are nice integrands for stochastic

integrals.
Tomas Bj
ork, 2013 18
The process Z t
Xt = Wsds,
0
is adapted.
The process
Xt = sup Ws
st
is adapted.
The process
Xt = sup Ws
st+1
is not adapted.
Tomas Bj
ork, 2013 19
The It
o Integral
We will define the It

o integral
Z b
gsdWs
a
for processes g satisfying
The process g is adapted.
The process g satisfies

Z b 2
E gs ds <
a
This will be done in two steps.
Tomas Bj
ork, 2013 20
Simple Integrands
Definition:
The process g is simple, if
g is adapted.
There exists deterministic points t0 . . . , tn with

a = t0 < t1 < . . . < tn = b such that g is piecewise
constant, i.e.
g(s) = g(tk ), s [tk , tk+1)
For simple g we define
Z b n1
X
gsdWs = g(tk ) [W (tk+1 ) W (tk )]
a k=0
FORWARD INCREMENTS!
Tomas Bj
ork, 2013 21
Properties of the Integral
Theorem: For simple g the following relations hold
The expected value is given by

"Z #
b
E gsdWs = 0
a
The second moment is given by

!
Z b 2 Z b
2
E gsdWs = E gs ds
a a
We have Z b
gsdWs FbW
a
Tomas Bj
ork, 2013 22
General Case
For a general g we do as follows.
1. Approximate g with a sequence of simple gn such

that Z b h i
2
E {gn(s) g(s)} ds 0.
a
2. For each n the integral

Z b
gn(s)dW (s)
a
is a well defined stochastic variable Zn.
3. One can show that the Zn sequence converges to a

limiting stochastic variable.
Rb
4. We define a gdW by
Z b Z b
g(s)dW (s) = lim gn(s)dW (s).
a n a
Tomas Bj
ork, 2013 23
Properties of the Integral
Theorem: For general g following relations hold
The expected value is given by

"Z #
b
E gsdWs = 0
a
We do in fact have
"Z #
b

E gsdWs Fa = 0
a
The second moment is given by

!
Z b 2 Z b
2
E gsdWs = E gs ds
a a
Tomas Bj
ork, 2013 24
We have Z b
gsdWs FbW
a
Tomas Bj
ork, 2013 25
Martingales
Definition: An adapted process is a martingale if
E [Xt| Fs] = Xs, s t
A martingale is a process without drift
Proposition: For any g (sufficiently integrable) he

process Z t
Xt = gsdWs
0
is a martingale.
Proposition: If X has dynamics
dXt = tdt + t dWt
then X is a martingale iff = 0.
Tomas Bj
ork, 2013 26
Stochastic Calculus
(Ch 4-5)
Tomas Bj
ork
Tomas Bj
ork, 2013 27
Stochastic Calculus
General Model:
dXt = tdt + t dWt
Let the function f (t, x) be given, and define the

stochastic process Zt by
Zt = f (t, Xt)
Problem: What does df (t, Xt ) look like?
The answer is given by the It

o formula.
We provide an intuitive argument. The formal proof is

very hard.
Tomas Bj
ork, 2013 28
A close up of the Wiener process
Consider an infinitesimal Wiener increment
dWt = Wt+dt Wt
We know:
dWt N [0, dt]
E[dWt] = 0, V ar[dWt] = dt
From this one can show
E[(dWt)2 ] = dt, V ar[(dWt )2] = 2(dt)2
Tomas Bj
ork, 2013 29
Recall
E[(dWt)2 ] = dt, V ar[(dWt )2] = 2(dt)2
Important observation:
1. Both E[(dWt)2 ] and V ar[(dWt)2 ] are very small

when dt is small .
2. V ar[(dWt)2 ] is negligable compared to E[(dWt)2 ].
3. Thus (dWt )2 is deterministic.
We thus conclude, at least intuitively, that
(dWt )2 = dt
This was only an intuitive argument, but it can be

proved rigorously.
Tomas Bj
ork, 2013 30
Multiplication table.
Theorem: We have the following multiplication table
(dt)2 = 0
dWt dt = 0
2
(dWt) = dt
Tomas Bj
ork, 2013 31
Deriving the It
o formula
dXt = tdt + t dWt
Zt = f (t, Xt)
We want to compute df (t, Xt )
Make a Taylor expansion of f (t, Xt ) including second

order terms:
f f 1 2f 2
df = dt + dXt + (dt)
t x 2 t2
1 2f 2 2f
+ 2
(dXt ) + dt dXt
2 x tx
Plug in the expression for dX, expand, and use the

multiplication table!
Tomas Bj
ork, 2013 32
f f 1 2f 2
df = dt + [dt + dW ] + (dt)
t x 2 t2
1 2f 2 2f
+ 2
[dt + dW ] + dt [dt + dW ]
2 x tx
f f f 1 2f 2
= dt + dt + dW + (dt)
t x x 2 t2
1 2f 2 2 2 2
+ 2
[ (dt) + (dW ) + 2dt dW ]
2 x
2f 2 2f
+ (dt) + dt dW
tx tx
Using the multiplikation table this reduces to:

2

f f 1 2 f
df = + + 2
dt
t x 2 x
f
+ dW
x
Tomas Bj
ork, 2013 33
The It
o Formula
Theorem: With X dynamics given by
dXt = tdt + t dWt
we have

f f 1 2 2f
df (t, Xt) = + + dt
t x 2 x2
f
+ dWt
x
Alternatively
f f 1 2f 2
df (t, Xt) = dt + dXt + 2
(dXt ) ,
t x 2 x
where we use the multiplication table.
Tomas Bj
ork, 2013 34
Example: GBM
dSt = Stdt + St dWt
We smell something exponential!

Natural Ansatz:
St = eZt ,
Zt = ln St
It
o on f (t, s) = ln(s) gives us
f 1 f 2f 1
= , = 0, = 2
s s t s2 s
1 1 1 2
dZt = dSt (dS t )
St 2 St2

1 2
= dt + dWt
2
Tomas Bj
ork, 2013 35
Recall
1
dZt = 2 dt + dWt
2
Integrate!
Z t Z t
1
Zt Z0 = 2 ds + dWs
0 2 0

1
= 2 t + Wt
2
Using St = eZt gives us
( 12 2)t+Wt
St = S0 e
Since Wt is N [0, t], we see that St has a lognormal

distribution.
Tomas Bj
ork, 2013 36
A Useful Trick
Problem: Compute E [Z(T )].
Use It
o to get
dZ(t) = Z (t)dt + Z (t)dWt
Integrate.
Z T Z T
Z(T ) = z0 + Z (t)dt + Z (t)dWt
0 0
Take expectations.
Z T
E [Z(T )] = z0 + E [Z (t)] dt + 0
0
The problem has been reduced to that of computing

E [Z (t)] .
Tomas Bj
ork, 2013 37
The Connection SDE PDE
Given: (t, x), (t, x), (x), T
Problem: Find a function F solving the Partial

Differential Equation (PDE)
F
(t, x) + AF (t, x) = 0,
t
F (T, x) = (x).
where A is defined by
F 1 2 2F
AF (t, x) = (t, x) + (t, x) 2 (t, x)
x 2 x
Tomas Bj
ork, 2013 38
Assume that F solves the PDE.
Fix the point (t, x).
Define the process X by
dXs = (s, Xs)dt + (s, Xs)dWs,

Xt = x,
Apply Ito to the process F (t, Xt)!
F (T, XT ) = F (t, Xt )
Z T
F
+ (s, Xs) + AF (s, Xs) ds
t t
Z T
F
+ (s, Xs) (s, Xs)dWs.
t x
F
By assumption t + AF = 0, and F (T, x) = (x)
Tomas Bj
ork, 2013 39
Thus:
(XT ) = F (t, x)
Z T
F
+ (s, Xs) (s, Xs)dWs.
t x
Take expectations.
F (t, x) = Et,x [ (XT )] ,
Tomas Bj
ork, 2013 40
Feynman-Kac
The solution F (t, x) to the PDE
F F 1 2 2F
+ (t, x) + (t, x) 2 rF = 0,
t x 2 x
F (T, x) = (x).
is given by
F (t, x) = er(T t)Et,x [ (XT )] ,
where X satisfies the SDE
dXs = (s, Xs)dt + (s, Xs)dWs,

Xt = x.
Tomas Bj
ork, 2013 41
Black-Scholes
(Ch 6-7)
Tomas Bj
ork
Tomas Bj
ork, 2013 42
Contents
1. Introduction.
2. Portfolio theory.
3. Deriving the Black-Scholes PDE
4. Risk neutral valuation
5. Appendices.
Tomas Bj
ork, 2013 43
1.
Introduction
Tomas Bj
ork, 2013 44
European Call Option
The holder of this paper has the right
to buy
1 ACME INC
on the date
June 30, 2014
at the price
$100
Tomas Bj
ork, 2013 45
Financial Derivative
A financial asset which is defined in terms of some

underlying asset.
Future stochastic claim.
Tomas Bj
ork, 2013 46
Examples
European calls and puts
American options
Forward rate agreements
Convertibles
Futures
Bond options
Caps & Floors
Interest rate swaps
CDO:s
CDS:s
Tomas Bj
ork, 2013 47
Main problems
What is a reasonable price for a derivative?
How do you hedge yourself against a derivative.
Tomas Bj
ork, 2013 48
Natural Answers
Consider a random cash payment Z at time T .
What is a reasonable price 0 [Z] at time 0?
Natural answers:
1. Price = Discounted present value of future payouts.
0 [Z] = erT E [Z]
2. The question is meaningless.
Tomas Bj
ork, 2013 49
Both answers are incorrect!
Given some assumptions we can really talk about

the correct price of an option.
The correct pricing formula is not the one on the

previous slide.
Tomas Bj
ork, 2013 50
Philosophy
The derivative is defined in terms of underlying.
The derivative can be priced in terms of underlying

price.
Consistent pricing.
Relative pricing.
Before we can go on further we need some simple

portfolio theory
Tomas Bj
ork, 2013 51
2.
Portfolio Theory
Tomas Bj
ork, 2013 52
Portfolios
We consider a market with N assets.
Sti = price at t, of asset No i.
A portfolio strategy is an adapted vector process
ht = (h1t , , hN
t )
where
hit = number of units of asset i,

Vt = market value of the portfolio
N
X
Vt = hitSti
i=1
The portfolio is typically of the form
ht = h(t, St)
i.e. todays portfolio is based on todays prices.
Tomas Bj
ork, 2013 53
Self financing portfolios
We want to study self financing portfolio strategies,

i.e. portfolios where purchase of a new asset must
be financed through sale of an old asset.
How is this formalized?
Definition:
The strategy h is self financing if
N
X
dVt = hitdSti
i=1
Interpret!
See Appendix B for details.
Tomas Bj
ork, 2013 54
Relative weights
Definition:
ti = relative portfolio weight on asset No i.
We have
hitSti
ti =
Vt
Insert this into the self financing condition

N
X
dVt = hitdSti
i=1
We obtain
Portfolio dynamics:
N
X i
i dSt
dVt = Vt t i
i=1
St
Interpret!
Tomas Bj
ork, 2013 55
3.
Deriving the Black-Scholes PDE
Tomas Bj
ork, 2013 56
Back to Financial Derivatives
Consider the Black-Scholes model
dSt = Stdt + StdWt,

dBt = rBtdt.
We want to price a European call with strike price X

and exercise time T . This is a stochastic claim on
the future. The future pay-out (at T ) is a stochastic
variable, Z, given by
Z = max[ST X, 0]
More general:
Z = (ST )
for some contract function .
Main problem: What is a reasonable price, t [Z],

for Z at t?
Tomas Bj
ork, 2013 57
Main Idea
We demand consistent pricing between derivative

and underlying.
No mispricing between derivative and underlying.
No arbitrage possibilities on the market (B, S, )
Tomas Bj
ork, 2013 58
Arbitrage
The portfolio is an arbitrage portfolio if
The portfolio strategy is self financing.
V0 = 0.
VT > 0 with probability one.
Moral:
Arbitrage = Free Lunch
No arbitrage possibilities in an efficient market.
Tomas Bj
ork, 2013 59
Arbitrage test
Suppose that a portfolio is self financing whith

dynamics
dVt = kVtdt
No driving Wiener process
Risk free rate of return.
Synthetic bank with rate of return k.
If the market is free of arbitrage we must have:
k=r
Tomas Bj
ork, 2013 60
Main Idea of Black-Scholes
Since the derivative is defined in terms of the

underlying, the derivative price should be highly
correlated with the underlying price.
We should be able to balance dervative against

underlying in our portfolio, so as to cancel the
randomness.
Thus we will obtain a riskless rate of return k on

our portfolio.
Absence of arbitrage must imply
k=r
Tomas Bj
ork, 2013 61
Two Approaches
The program above can be formally carried out in two

slightly different ways:
The way Black-Scholes did it in the original paper.

This leads to some logical problems.
A more conceptually satisfying way, first presented

by Merton.
Here we use the Merton method. You will find the

original BS method in Appendix C at the end of this
lecture.
Tomas Bj
ork, 2013 62
Formalized program a la Merton
Assume that the derivative price is of the form
t [Z] = f (t, St).
Form a portfolio based on the underlying S and the

derivative f , with portfolio dynamics

dSt df
dVt = Vt tS + tf
St f
Choose S and f such that the dW -term is wiped

out. This gives us
dVt = Vt kdt
Absence of arbitrage implies
k=r
This relation will say something about f .
Tomas Bj
ork, 2013 63
Back to Black-Scholes
dSt = Stdt + StdWt,

t [Z] = f (t, St)
It
os formula gives us the f dynamics as
2

f f 1 2 2 f
df = + S + S 2
dt
t s 2 s
f
+ S dW
s
Write this as
df = f f dt + f f dW
where
2
f
t + S f
s + 1 2 2 f
2 S s2
f =
f
S f
s
f =
f
Tomas Bj
ork, 2013 64
df = f f dt + f f dW

dS df
dV = V S + f
S f
S f

= V (dt + dW ) + (f dt + f dW )
S f
S f

dV = V + f dt + V + f dW
Now we kill the dW -term!

Choose ( S , f ) such that
S + f f = 0
S + f = 1
Linear system with solution

f
S = , f =
f f
Plug into dV !
Tomas Bj
ork, 2013 65
We obtain
S f

dV = V + f dt
This is a risk free synthetic bank with short rate

S f

+ F
S f

+ f = r
Plug in the expressions for S , f , f and simplify.

This will give us the following result.
Tomas Bj
ork, 2013 66
Black-Scholes PDE
The price is given by

t [Z] = f (t, St)
where the pricing function f satisfies the PDE (partial differential equation)

f f 1 2 2 2f
(t, s) + rs (t, s) + s (t, s) rf (t, s) = 0
t s 2 s2

f (T, s) = (s)
There is a unique solution to the PDE so there is a unique arbitrage free

price process for the contract.
Tomas Bj
ork, 2013 67
Black-Scholes PDE ctd

f f 1 2 2 2f
+ rs + s rf = 0
t s 2 s2

f (T, s) = (s)
The price of all derivative contracts have to satisfy

the same PDE
f f 1 2 2 2f
+ rs + s rf = 0
t s 2 s2
otherwise there will be an arbitrage opportunity.
The only difference between different contracts is in

the boundary value condition
f (T, s) = (s)
Tomas Bj
ork, 2013 68
Data needed
The contract function .
Todays date t.
Todays stock price S.
Short rate r.
Volatility .
Note: The pricing formula does not involve the mean

rate of return !
??
Tomas Bj
ork, 2013 69
Black-Scholes Basic Assumptions
Assumptions:
The stock price is Geometric Brownian Motion
Continuous trading.
Frictionless efficient market.
Short positions are allowed.
Constant volatility .
Constant short rate r.
Flat yield curve.
Tomas Bj
ork, 2013 70
Black-Scholes Formula
European Call
T =date of expiration,
t=todays date,
X=strike price,
r=short rate,
s=todays stock price,
=volatility.
f (t, s) = sN [d1] er(T t)XN [d2 ] .
N []=cdf for N (0, 1)-distribution.
s
1 1 2
d1 = ln + r + (T t) ,
T t X 2

d2 = d1 T t.
Tomas Bj
ork, 2013 71
Black-Scholes
European Call,
X = 100, = 20%, r = 7%, T t = 1/4
25
20
15
10
0 s
80 85 90 95 100 105 110 115 120
Tomas Bj
ork, 2013 72
Dependence on Time to Maturity
25
13 weeks
7 weeks
1 week
maturity
20
15
C
10
0
80 85 90 95 100 105 110 115 120
S
Tomas Bj
ork, 2013 73
Dependence on Volatility
25
sigma=0.2
sigma=0.4
sigma=0.6
maturity
20
15
C
10
0
80 85 90 95 100 105 110 115 120
S
Tomas Bj
ork, 2013 74
4.
Risk Neutral Valuation
Tomas Bj
ork, 2013 75
Risk neutral valuation
Appplying Feynman-Kac to the Black-Scholes PDE we

obtain
Q
[t; X] = er(T t)Et,s [X]
Q-dynamics:

dSt = rStdt + StdWtQ ,

dBt = rBt dt.
Price = Expected discounted value of future

payments.
The expectation shall not be taken under the

objective probability measure P , but under the
risk adjusted measure (martingale measure) Q.
Note: P Q
Tomas Bj
ork, 2013 76
Concrete formulas
Z
[0; ] = erT (sez )f (z)dz

( 2 )
1 2
1 z (r 2 )T
f (z) = exp
2T 2 2T
Tomas Bj
ork, 2013 77
Interpretation of the risk adjusted
measure
Assume a risk neutral world.
Then the following must hold
s = S0 = ertE [St]
In our model this means that
dSt = rStdt + StdWtQ
The risk adjusted probabilities can be intrepreted as

probabilities in a fictuous risk neutral economy.
Tomas Bj
ork, 2013 78
Moral
When we compute prices, we can compute as if we

live in a risk neutral world.
This does not mean that we live (or think that we

live) in a risk neutral world.
The formulas above hold regardless of the investors

attitude to risk, as long as he/she prefers more to
less.
The valuation formulas are therefore called

preference free valuation formulas.
Tomas Bj
ork, 2013 79
Properties of Q
P Q
For the price pricess of any traded asset, derivative

or underlying, the process
t
Zt =
Bt
is a Q-martingale.
Under Q, the price pricess of any traded asset,

derivative or underlying, has r as its local rate of
return:
dt = rtdt + tdWtQ
The volatility of is the same under Q as under P .
Tomas Bj
ork, 2013 80
A Preview of Martingale Measures
Consider a market, under an objective probability

measure P , with underlying assets
B, S 1, . . . , S N
Definition: A probability measure Q is called a

martingale measure if
P Q
For every i, the process
Sti
Zti =
Bt
is a Q-martingale.
Theorem: The market is arbitrage free iff there exists

a martingale measure.
Tomas Bj
ork, 2013 81
5.
Appendices
Tomas Bj
ork, 2013 82
Appendix A: Black-Scholes vs Binomial
Consider a binomial model for an option with a fixed

time to maturity T and a fixed strike price X.
Build a binomial model with n periods for each

n = 1, 2, ....
Use the standard formulas for scaling the jumps:

u = e t
d=e t
t = T /n
For a large n, the stock price at time T will then

be a product of a large number of i.i.d. random
variables.
More precisely
ST = S0 Z1 Z2 Zn ,
where n is the number of periods in the binomial

model and Zi = u, d
Tomas Bj
ork, 2013 83
Recall
ST = S0 Z1 Z2 Zn ,
The stock price at time T will be a product of a

large number of i.i.d. random variables.
The return will be a large sum of i.i.d. variables.
The Central Limit Theorem will kick in.
In the limit, returns will be normally distributed.
Stock prices will be lognormally distributed.
We are in the Black-Scholes model.
The binomial price will converge to the Black-

Scholes price.
Tomas Bj
ork, 2013 84
Binomial convergence to Black-Scholes
11
BS
Bin
10.5
10
9.5
Price
8.5
7.5
0 5 10 15 20 25 30 35 40 45 50
N
Tomas Bj
ork, 2013 85
Binomial Black-Scholes
The intuition from the Binomial model carries over to

Black-Scholes.
The B-S model is just a binomial model where

we rebalance the portfolio infinitely often.
The B-S model is thus complete.
Completeness explains the unique prices for options

in the B-S model.
The B-S price for a derivative is the limit of the

binomial price when the number of periods is very
large.
Tomas Bj
ork, 2013 86
Appendix B: Portfolio theory
We consider a market with N assets.
A portfolio strategy is an adapted vector process
ht = (h1t , , hN
t )
where
hit = number of units of asset i,

Vt = market value of the portfolio
N
X
Vt = hitSti
i=1
The portfolio is typically of the form
ht = h(t, St)
i.e. todays portfolio is based on todays prices.
Tomas Bj
ork, 2013 87
We want to study self financing portfolio strategies,

i.e. portfolios where
There is now external infusion and/or withdrawal of

money to/from the portfolio.
Purchase of a new asset must be financed through

sale of an old asset.
How is this formalized?
Problem: Derive an expression for dVt for a self

financing portfolio.
We analyze in discrete time, and then go to the

continuous time limit.
Tomas Bj
ork, 2013 88
Discrete time portfolios
We trade at discrete points in time t = 0, 1, 2, . . ..
Price vector process:
Sn = (Sn1 , , SnN ), n = 0, 1, 2, . . .
Portfolio process:
hn = (h1n, , hN
n ), n = 0, 1, 2, . . .
Interpretation: At time n we buy the portfolio hn at

the price Sn, and keep it until time n + 1.
Value process:
N
X
Vn = hinSni = hnSn
i=1
Tomas Bj
ork, 2013 89
The self financing condition
At time n 1 we buy the portfolio hn1 at the price

Sn1.
At time n this portfolio is worth hn1Sn.
At time n we buy the new portfolio hn at the price

Sn .
The cost of this new portfolio is hnSn.
The self financing condition is the budget

constraint
hn1Sn = hnSn
Tomas Bj
ork, 2013 90
The self financing condition
Recall:
Vn = hnSn
Definition: For any sequence x1, x2, . . . we define the

sequence xn by
xn = xn xn1
Problem: Derive an expression for Vn for a self

financing portfolio.
Lemma: For any pair of sequences x1, x2 , . . . and

y1 , y2, . . . we have the relation
(xy)n = xn1yn + ynxn
Proof: Do it yourself.
Tomas Bj
ork, 2013 91
Recall
Vn = hnSn
From the Lemma we have
Vn = (hS)n = hn1Sn + Snhn
Recall the self financing condition
hn1Sn = hnSn
which we can write as
Snhn = 0
Inserting this into the expression for Vn gives us.
Proposition: The dynamics of a self financing portfolio

are given by
Vn = hn1Sn
Note the forward increments!
Tomas Bj
ork, 2013 92
Portfolios in continuous time
Price process:
Portfolio:
ht = (h1t , , hN
t )
Value process
N
X
Vt = hitSti
i=1
From the self financing condition in discrete time
Vn = hn1Sn
we are led to the following definition.

Definition: The portfolio h is self financing if and only
if
XN
dVt = hitdSti
i=1
Tomas Bj
ork, 2013 93
Relative weights
Definition:
ti = relative portfolio weight on asset No i.
We have
hitSti
ti =
Vt
Insert this into the self financing condition

N
X
dVt = hitdSti
i=1
We obtain
Portfolio dynamics:
N
X i
i dSt
dVt = Vt t i
i=1
St
Interpret!
Tomas Bj
ork, 2013 94
Appendix C:
The original Black-Scholes PDE
argument
Consider the following portfolio.
Short one unit of the derivative, with pricing

function f (t, s).
Hold x units of the underlying S.
The portfolio value is given by
V = f (t, ST ) + xSt
The object is to choose x such that the portfolio is

risk free for an infinitesimal interval of length dt.
We have dV = df + xdS and from It
o we obtain
2

f f 1 2 2 f
dV = + S + S 2
dt
t s 2 s
f
S dW + xSdt + xSdW
s
Tomas Bj
ork, 2013 95

f f 1 2 2 2f
dV = xS S S dt
t s 2 s2

f
+ S x dW
s
We obtain a risk free portfolio if we choose x as
f
x=
s
and then we have, after simplification,

2

f 1 2 2 f
dV = S 2
dt
t 2 s
Using V = f + xS and x as above, the return dV /V

is thus given by
2
dV f
t
1 2 2 f
2 S s2
= f
dt
V f + S s
Tomas Bj
ork, 2013 96
We had
1 2 2 2f
dV f
t 2 S s2
= f
dt
V f + S s
This portfolio is risk free, so absence of arbitrage

implies that
1 2 2 2f
f
t 2 S s2
=r
f + S f
s
Simplifying this expression gives us the Black-Scholes

PDE.
f f 1 2 2 2f
+ rs + s rf = 0,
t s 2 s2
f (T, s) = (s).
Tomas Bj
ork, 2013 97
Completeness and Hedging
(Ch 8-9)
Tomas Bj
ork
Tomas Bj
ork, 2013 98
Problems around Standard Black-Scholes
We assumed that the derivative was traded. How

do we price OTC products?
Why is the option price independent of the expected

rate of return of the underlying stock?
Suppose that we have sold a call option. Then we

face financial risk, so how do we hedge against that
risk?
All this has to do with completeness.
Tomas Bj
ork, 2013 99
Definition:
We say that a T -claim X can be replicated,
alternatively that it is reachable or hedgeable, if
there exists a self financing portfolio h such that
VTh = X, P a.s.
In this case we say that h is a hedge against X.

Alternatively, h is called a replicating or hedging
portfolio. If every contingent claim is reachable we say
that the market is complete
Basic Idea: If X can be replicated by a portfolio h

then the arbitrage free price for X is given by
t [X] = Vth .
Tomas Bj
ork, 2013 100
Trading Strategy
Consider a replicable claim X which we want to sell at

t = 0..
Compute the price 0 [X] and sell X at a slightly

(well) higher price.
Buy the hedging portfolio and invest the surplus in

the bank.
Wait until expiration date T .
The liabilities stemming from X is exactly matched

by VTh, and we have our surplus in the bank.
Tomas Bj
ork, 2013 101
Completeness of Black-Scholes
Theorem: The Black-Scholes model is complete.
Proof. Fix a claim X = (ST ). We want to find

processes V , uB and uS such that

dBt S dSt
dVt = Vt uB
t + ut
Bt St
VT = (ST ).
i.e.
B
dVt = Vt ut r + ut dt + VtuSt dWt,
S
VT = (ST ).
Tomas Bj
ork, 2013 102
Heuristics:
Let us assume that X is replicated by h = (uB , uS )
with value process V .
Ansatz:
Vt = F (t, St)
Ito gives us

1 2 2
dV = Ft + SFs + S Fss dt + SFsdW,
2
Write this as
( )
1 2 2
Ft + SFs + 2 S Fss SFs
dV = V dt + V dW.
V V
Compare with
B
dV = V u r + u dt + V uS dW
S
Tomas Bj
ork, 2013 103
Define uS by
StFs(t, St)
uSt = ,
F (t, St)
This gives us the eqn
( )
1 2 2
Ft + 2 S Fss
dV = V r + uS dt + V uS dW.
rF
Compare with
B
dV = V u r + u dt + V uS dW
S
Natural choice for uB is given by

1 2 2
B Ft + 2 S Fss
u = ,
rF
Tomas Bj
ork, 2013 104
The relation uB + uS = 1 gives us the Black-Scholes
PDE
1 2 2
Ft + rSFs + S Fss rF = 0.
2
The condition
VT = (ST )
gives us the boundary condition
F (T, s) = (s)
Moral: The model is complete and we have explicit

formulas for the replicating portfolio.
Tomas Bj
ork, 2013 105
Main Result
Theorem: Define F as the solution to the boundary
value problem

F + rsF + 1 2 s2F rF = 0,
t s ss
2
F (T, s) = (s).
Then X can be replicated by the relative portfolio
F (t, St) StFs(t, St)

uB
t = ,
F (t, St)
StFs(t, St)
uSt = .
F (t, St)
The corresponding absolute portfolio is given by
F (t, St) StFs(t, St)

hB
t = ,
Bt
hSt = Fs(t, St),
and the value process V h is given by
Vth = F (t, St).
Tomas Bj
ork, 2013 106
Notes
Completeness explains unique price - the claim is

superfluous!
Replicating the claim P a.s. Replicating the

claim Q a.s. for any Q P . Thus the price only
depends on the support of P .
Thus (Girsanov) it will not depend on the drift of

the state equation.
The completeness theorem is a nice theoretical

result, but the replicating portfolio is continuously
rebalanced. Thus we are facing very high
transaction costs.
Tomas Bj
ork, 2013 107
Completeness vs No Arbitrage
Question:
When is a model arbitrage free and/or complete?
Answer:
Count the number of risky assets, and the number of
random sources.
R = number of random sources

N = number of risky assets
Intuition:
If N is large, compared to R, you have lots of
possibilities of forming clever portfolios. Thus lots
of chances of making arbitrage profits. Also many
chances of replicating a given claim.
Tomas Bj
ork, 2013 108
Meta-Theorem
Generically, the following hold.
The market is arbitrage free if and only if
N R
The market is complete if and only if
N R
Example:
The Black-Scholes model. R=N=1. Arbitrage free
and complete.
Tomas Bj
ork, 2013 109
Parity Relations
Let and be contract functions for the T -claims

Z = (ST ) and Y = (ST ). Then for any real
numbers and we have the following price relation.
t [ + ] = t [] + t [] .
Proof. Linearity of mathematical expectation.

Consider the following basic contract functions.
S (x) = x,
B (x) 1,
C,K (x) = max [x K, 0] .
Prices:
t [S ] = St,
t [B ] = er(T t),
t [C,K ] = c(t, St; K, T ).
Tomas Bj
ork, 2013 110
If we have
n
X
= S + B + iC,Ki ,
i=1
then
n
X
t [] = t [S ] + t [B ] + it [C,Ki ]
i=1
We may replicate the claim using a portfolio

consisting of basic contracts that is constant over
time, i.e. a buy-and hold portfolio:
shares of the underlying stock,
zero coupon T -bonds with face value $1,
i European call options with strike price Ki, all

maturing at T .
Tomas Bj
ork, 2013 111
Put-Call Parity
Consider a European put contract
P,K (s) = max [K s, 0]
It is easy to see (draw a figure) that
P,K (x) = C,K (x) s + K

= P,K (x) S (x) + B (x)
We immediately get
Put-call parity:
p(t, s; K) = c(t, s; K) s + Ker(T t)
Thus you can construct a synthetic put option, using

a buy-and-hold portfolio.
Tomas Bj
ork, 2013 112
Delta Hedging
Consider a fixed claim
X = (ST )
with pricing function
F (t, s).
Setup:
We are at time t, and have a short (interpret!) position
in the contract.
Goal:
Offset the risk in the derivative by buying (or selling)
the (highly correlated) underlying.
Definition:
A position in the underlying is a delta hedge against
the derivative if the portfolio (underlying + derivative)
is immune against small changes in the underlying
price.
Tomas Bj
ork, 2013 113
Formal Analysis
1 = number of units of the derivative product

x = number of units of the underlying
s = todays stock price
t = todays date
Value of the portfolio:
V = 1 F (t, s) + x s
A delta hedge is characterized by the property that
V
= 0.
s
We obtain
F
+x =0
s
Solve for x!
Tomas Bj
ork, 2013 114
Result:
We should have
F
x=
s
shares of the underlying in the delta hedged portfolio.
Definition:
For any contract, its delta is defined by
F
= .
s
Result:
We should have
x
=
shares of the underlying in the delta hedged portfolio.
Warning:
The delta hedge must be rebalanced over time. (why?)
Tomas Bj
ork, 2013 115
Black Scholes
For a European Call in the Black-Scholes model we

have
= N [d1 ]
NB This is not a trivial result!
From put call parity it follows (how?) that for a

European Put is given by
= N [d1 ] 1
Check signs and interpret!
Tomas Bj
ork, 2013 116
Rebalanced Delta Hedge
Sell one call option a time t = 0 at the B-S price F .
Compute and by shares. (Use the income

from the sale of the option, and borrow money if
necessary.)
Wait one day (week, minute, second..). The stock

price has now changed.
Compute the new value of , and borrow money in

order to adjust your stock holdings.
Repeat this procedure until t = T . Then the value

of your portfolio (B+S) will match the value of the
option almost exactly.
Tomas Bj
ork, 2013 117
Lack of perfection comes from discrete, instead of
continuous, trading.
You have created a synthetic option.

(Replicating portfolio).
Formal result:
The relative weights in the replicating portfolio are
S
uS = ,
F
F S
uB =
F
Tomas Bj
ork, 2013 118
Portfolio Delta
Assume that you have a portfolio consisting of

derivatives
i (STi ), i = 1, , n
all written on the same underlying stock S.
Fi (t, s) = pricing function for i:th derivative

Fi
i =
s
hi = units of i:th derivative
Portfolio value: n
X
= hiFi
i=1
Portfolio delta:
n
X
= h i i
i=1
Tomas Bj
ork, 2013 119
Gamma
A problem with discrete delta-hedging is.
As time goes by S will change.
F
This will cause = s to change.
Thus you are sitting with the wrong value of delta.
Moral:
If delta is sensitive to changes in S, then you have

to rebalance often.
If delta is insensitive to changes in S you do not

need to rebalance so often.
Tomas Bj
ork, 2013 120
Definition:
Let be the value of a derivative (or portfolio).
Gamma () is defined as

=
s
i.e.
2
=
s2
Gamma is a measure of the sensitivity of to changes
in S.
Result: For a European Call in a Black-Scholes model,
can be calculated as
N 0[d1]
=
S T t
Important fact:
For a position in the underlying stock itself we have
=0
Tomas Bj
ork, 2013 121
Gamma Neutrality
A portfolio is said to be gamma neutral if its

gamma equals zero, i.e.
= 0
Since = 0 for a stock you can not gamma-hedge

using only stocks. item Typically you use some
derivative to obtain gamma neutrality.
Tomas Bj
ork, 2013 122
General procedure
Given a portfolio with underlying S. Consider two

derivatives with pricing functions F and G.
xF = number of units of F
xG = number of units of G
Problem:
Choose xF and xG such that the entire portfolio is
delta- and gamma-neutral.
Value of hedged portfolio:
V = + xF F + xG G
Tomas Bj
ork, 2013 123
Value of hedged portfolio:
V = + xF F + xG G
We get the equations
V
= 0,
s
2V
= 0.
s2
i.e.
+ xF F + xGG = 0,
+ xF F + xGG = 0
Solve for xF and xG!
Tomas Bj
ork, 2013 124
Particular Case
In many cases the original portfolio is already

delta neutral.
Then it is natural to use a derivative to obtain

gamma-neutrality.
This will destroy the delta-neutrality.
Therefore we use the underlying stock (with zero

gamma!) to delta hedge in the end.
Tomas Bj
ork, 2013 125
Formally:
V = + xF F + xS S
+ xF F + xS S = 0,
+ xF F + xS S = 0
We have
= 0,
S = 1
S = 0.
i.e.
+ xF F + xS = 0,
+ xF F = 0

xF =
F
F
xS =
F
Tomas Bj
ork, 2013 126
Further Greeks

= ,
t

V = ,

=
r
V is pronounced Vega.
NB!
A delta hedge is a hedge against the movements in
the underlying stock, given a fixed model.
A Vega-hedge is not a hedge against movements of

the underlying asset. It is a hedge against a change
of the model itself.
Tomas Bj
ork, 2013 127
The Martingale Approach
I: Mathematics
(Ch 10-12)
Tomas Bj
ork
Tomas Bj
ork, 2013 128
Introduction
In order to understand and to apply the martingale

approach to derivative pricing and hedging we will
need to some basic concepts and results from measure
theory. These will be introduced below in an informal
manner - for full details see the textbook.
Many propositions below will be proved but we will

also present a couple of central results without proofs,
and these must then be considered as dogmatic truths.
You are of course not expected to know the proofs of
such results (this is outside the scope of this course)
but you are supposed to be able to use the results in
an operational manner.
Tomas Bj
ork, 2013 129
Contents
1. Events and sigma-algebras
2. Conditional expectations
3. Changing measures
4. The Martingale Representation Theorem
5. The Girsanov Theorem
Tomas Bj
ork, 2013 130
1.
Events and sigma-algebras
Tomas Bj
ork, 2013 131
Events and sigma-algebras
Consider a probability measure P on a sample space

. An event is simply a subset A and P (A) is
the probability that the event A occurs.
For technical reasons, a probability measure can only

be defined for a certain nice class F of events, so for
A F we are allowed to write P (A) as the probability
for the event A.
For technical reasons the class F must be a sigma-

algebra, which means that F is closed under the usual
set theoretic operations like complements, countable
intersections and countable unions.
Interpretation: We can view a -algebra F as

formalizing the idea of information. More precisely: A
-algebra F is a collection of events, and if we assume
that we have access to the information contained in F,
this means that for every A F we know exactly if A
has occured or not.
Tomas Bj
ork, 2013 132
Borel sets
Definition: The Borel algebra B is the smallest

sigma-algebra on R which contains all intervals. A set
B in B is called a Borel set.
Remark: There is no constructive definition of B, but

almost all subsets of R that you will ever see will in
fact be Borel sets, so the reader can without danger
think about a Borel set as an arbitrary subset of R.
Tomas Bj
ork, 2013 133
Random variables
An F-measurable random variable X is a a mapping
X:R
such that {X B} = { : X() B} belongs

to F for all Borel sets B. This guarantees that we are
allowed to write P (X B). Instad of writing X is
F-measurable we will often write X F.
This means that if X F then the value of X is

completely determined by the information contained in
F.
If we have another -algebra G with G F then we

interpret this as G contains less information than F .
Tomas Bj
ork, 2013 134
2.
Conditional Expectation
Tomas Bj
ork, 2013 135
Conditional Expectation
If X F and if G F then we write E [X| G] for

the conditional expectation of X given the information
contained in G. Sometimes we use the notation EG [X].
The following proposition contains everything that we

will need to know about conditional expectations within
this course.
Tomas Bj
ork, 2013 136
Main Results
Proposition 1: Assume that X F, and that G F .

Then the following hold.
The random variable E [X| G] is completely determined by

the information in G so we have
E [X| G] G
If we have Y G then Y is completely determined by G so

we have
E [XY | G] = Y E [X| G]
In particular we have
E [Y | G] = Y
If H G then we have the law of iterated expectations
E [E [X| G]| H] = E [X| H]
In particular we have
E [X] = E [E [X| G]]
Tomas Bj
ork, 2013 137
3.
Changing Measures
Tomas Bj
ork, 2013 138
Changing Measures
Consider a probability measure P on (, F ), and

assume that L F is a random variable with the
properties that
L0
and
E P [L] = 1.
For every event A F we now define the real number

Q(A) by the prescription
Q(A) = E P [L IA]
where the random variable IA is the indicator for A,

i.e. (
1 if A occurs
IA =
0 if Ac occurs
Tomas Bj
ork, 2013 139
Recall that
Q(A) = E P [L IA]
We now see that Q(A) 0 for all A, and that
Q() = E P [L I] = E P [L 1] = 1
We also see that if A B = then
Q(A B) = E P [L IAB ] = E P [L (IA + IB )]

= E P [L IA] + E P [L IB ]
= Q(A) + Q(B)
Furthermore we see that
P (A) = 0 Q(A) = 0
We have thus more or less proved the following
Tomas Bj
ork, 2013 140
Proposition 2: If L F is a nonnegative random
variable with E P [L] = 1 and Q is defined by
Q(A) = E P [L IA]
then Q will be a probability measure on F with the

property that
P (A) = 0 Q(A) = 0.
I turns out that the property above is a very important

one, so we give it a name.
Tomas Bj
ork, 2013 141
Absolute Continuity
Definition: Given two probability measures P and Q

on F we say that Q is absolutely continuous w.r.t.
P on F if, for all A F , we have
P (A) = 0 Q(A) = 0
We write this as
Q << P.
If Q << P and P << Q then we say that P and Q

are equivalent and write
QP
Tomas Bj
ork, 2013 142
Equivalent measures
It is easy to see that P and Q are equivalent if and

only if
P (A) = 0 Q(A) = 0
or, equivalently,
P (A) = 1 Q(A) = 1
Two equivalent measures thus agree on all certain

events and on all impossible events, but can disagree
on all other events.
Simple examples:
All non degenerate Gaussian distributions on R are
equivalent.
If P is Gaussian on R and Q is exponential then

Q << P but not the other way around.
Tomas Bj
ork, 2013 143
Absolute Continuity ctd
We have seen that if we are given P and define Q by
Q(A) = E P [L IA]
for L 0 with E P [L] = 1, then Q is a probability

measure and Q << P . .
A natural question is now if all measures Q << P

are obtained in this way. The answer is yes, and the
precise (quite deep) result is as follows. The proof is
difficult and therefore omitted.
Tomas Bj
ork, 2013 144
The Radon Nikodym Theorem
Consider two probability measures P and Q on (, F ),

and assume that Q << P on F . Then there exists a
unique random variable L with the following properties
1. Q(A) = E P [L IA] , A F
2. L 0, P a.s.
3. E P [L] = 1,
4. LF
The random variable L is denoted as
dQ
L= , on F
dP
and it is called the Radon-Nikodym derivative of Q

w.r.t. P on F , or the likelihood ratio between Q and
P on F .
Tomas Bj
ork, 2013 145
A simple example
The Radon-Nikodym derivative L is intuitively the local

scale factor between P and Q. If the sample space
is finite so = {1, . . . , n} then P is determined by
the probabilities p1, . . . , pn where
pi = P (i ) i = 1, . . . , n
Now consider a measure Q with probabilities
qi = Q(i ) i = 1, . . . , n
If Q << P this simply says that
pi = 0 qi = 0
and it is easy to see that the Radon-Nikodym derivative

L = dQ/dP is given by
qi
L(i) = i = 1, . . . , n
pi
Tomas Bj
ork, 2013 146
If pi = 0 then we also have qi = 0 and we can define
the ratio qi /pi arbitrarily.
If p1 , . . . , pn as well as q1, . . . , qn are all positive, then

we see that Q P and in fact
1
dP 1 dQ
= =
dQ L dP
as could be expected.
Tomas Bj
ork, 2013 147
Computing expected values
A main use of Radon-Nikodym derivatives is for the

computation of expected values.
Suppose therefore that Q << P on F and that X is

a random variable with X F . With L = dQ/dP on
F then have the following result.
Proposition 3: With notation as above we have
E Q [X] = E P [L X]
Proof: We only give a proof for the simple example

above where = {1, . . . , n}. We then have
n
X n
X qi
E Q [X] = X(i )qi = X(i) pi
i=1 i=1
pi
Xn
= X(i )L(i )pi = E P [X L]
i=1
Tomas Bj
ork, 2013 148
The Abstract Bayes Formula
We can also use Radon-Nikodym derivatives in order to

compute conditional expectations. The result, known
as the abstract Bayes Formula, is as follows.
Theorem 4: Consider two measures P and Q with

Q << P on F and with
dQ
LF = on F
dP
Assume that G F and let X be a random variable

with X F . Then the following holds

Q E L X G
P F
E [X| G] =
E P [LF | G]
Tomas Bj
ork, 2013 149
Dependence of the -algebra
Suppose that we have Q << P on F with
dQ
F
L = on F
dP
Now consider smaller -algebra G F . Our problem
is to find the R-N derivative
dQ
LG = on G
dP
We recall that LG is characterized by the following

properties
P
G

1. Q(A) = E L IA A G
2. LG 0
P
G

3. E L =1
4. LG G
Tomas Bj
ork, 2013 150
A natural guess would perhaps be that LG = LF , so
let us check if LF satisfies points 1-4 above.
By assumption we have
P
F

Q(A) = E L IA A F
Since G F we then have
P
F

Q(A) = E L IA A G
so point 1 above is certainly satisfied by LF . It is

also clear that LF satisfies points 2 and 3. It thus
seems that LF is also a natural candidate for the R-N
derivative LG , but the problem is that we do not in
general have LF G.
This problem can, however, be fixed. By iterated

expectations we have, for all A G,
P
F
P
P
F

E L IA = E E L IA G
Tomas Bj
ork, 2013 151
Since A G we have
P
F
P
F
E L IA G = E L G IA

Let us now define LG by
G P

F
L =E L G
We then obviously have LG G and
P
G

Q(A) = E L IA A G
It is easy to see that also points 2-3 are satisfied so we

have proved the following result.
Tomas Bj
ork, 2013 152
A formula for LG
Proposition 5: If Q << P on F and G F then,

with notation as above, we have
G P

F
L =E L G
Tomas Bj
ork, 2013 153
The likelihood process on a filtered space
We now consider the case when we have a probability
measure P on some space and that instead of just
one -algebra F we have a filtration, i.e. an increasing
family of -algebras {Ft}t0 .
The interpretation is as usual that Ft is the information
available to us at time t, and that we have Fs Ft
for s t.
Now assume that we also have another measure Q,
and that for some fixed T , we have Q << P on FT .
We define the random variable LT by
dQ
LT = on FT
dP
Since Q << P on FT we also have Q << P on Ft
for all t T and we define
dQ
Lt = on Ft 0tT
dP
For every t we have Lt Ft, so L is an adapted
process, known as the likelihood process.
Tomas Bj
ork, 2013 154
The L process is a P martingale
We recall that
dQ
Lt = on Ft 0tT
dP
Since Fs Ft for s t we can use Proposition 5 and

deduce that
Ls = E P [Lt| Fs] stT
and we have thus proved the following result.
Proposition: Given the assumptions above, the

likelihood process L is a P -martingale.
Tomas Bj
ork, 2013 155
Where are we heading?
We are now going to perform measure transformations

on Wiener spaces, where P will correspond to the
objective measure and Q will be the risk neutral
measure.
For this we need define the proper likelihood process L

and, since L is a P -martingale, we have the following
natural questions.
What does a martingale look like in a Wiener driven

framework?
Suppose that we have a P -Wiener process W and

then change measure from P to Q. What are the
properties of W under the new measure Q?
These questions are handled by the Martingale

Representation Theorem, and the Girsanov Theorem
respectively.
Tomas Bj
ork, 2013 156
4.
The Martingale Representation Theorem
Tomas Bj
ork, 2013 157
Intuition
Suppose that we have a Wiener process W under

the measure P . We recall that if h is adapted (and
integrable enough) and if the process X is defined by
Z t
Xt = x0 + hsdWs
0
then X is a a martingale. We now have the following

natural question:
Question: Assume that X is an arbitrary martingale.

Does it then follow that X has the form
Z t
Xt = x0 + hsdWs
0
for some adapted process h?
In other words: Are all martingales stochastic integrals

w.r.t. W ?
Tomas Bj
ork, 2013 158
Answer
It is immediately clear that all martingales can not be

written as stochastic integrals w.r.t. W . Consider for
example the process X defined by
(
0 for 0 t < 1
Xt =
Z for t 1
where Z is an random variable, independent of W ,

with E [Z] = 0.
X is then a martingale (why?) but it is clear (how?)

that it cannot be written as
Z t
Xt = x0 + hsdWs
0
for any process h.
Tomas Bj
ork, 2013 159
Intuition
The intuitive reason why we cannot write

Z t
Xt = x0 + hsdWs
0
in the example above is of course that the random

variable Z has nothing to do with the Wiener process
W . In order to exclude examples like this, we thus need
an assumption which guarantees that our probability
space only contains the Wiener process W and nothing
else.
This idea is formalized by assuming that the filtration

{Ft}t0 is the one generated by the Wiener
process W .
Tomas Bj
ork, 2013 160
The Martingale Representation Theorem
Theorem. Let W be a P -Wiener process and assume

that the filtation is the internal one i.e.
Ft = FtW = {Ws; 0 s t}
Then, for every (P, Ft)-martingale X, there exists a

real number x and an adapted process h such that
Z t
Xt = x + hsdWs,
0
i.e.
dXt = htdWt.
Proof: Hard. This is very deep result.
Tomas Bj
ork, 2013 161
Note
For a given martingale X, the Representation Theorem

above guarantees the existence of a process h such that
Z t
Xt = x + hsdWs,
0
The Theorem does not, however, tell us how to find

or construct the process h.
Tomas Bj
ork, 2013 162
5.
The Girsanov Theorem
Tomas Bj
ork, 2013 163
Setup
Let W be a P -Wiener process and fix a time horizon

T . Suppose that we want to change measure from P
to Q on FT . For this we need a P -martingale L with
L0 = 1 to use as a likelihood process, and a natural
way of constructing this is to choose a process g and
then define L by
(
dLt = gtdWt
L0 = 1
This definition does not guarantee that L 0, so we

make a small adjustment. We choose a process and
define L by
(
dLt = Ltt dWt
L0 = 1
The process L will again be a martingale and we easily

obtain Rt
dW 1 R t 2ds
s s
Lt = e 0 2 0 s
Tomas Bj
ork, 2013 164
Thus we are guaranteed that L 0. We now change
measure form P to Q by setting
dQ = LtdP, on Ft, 0 t T
The main problem is to find out what the properties

of W are, under the new measure Q. This problem is
resolved by the Girsanov Theorem.
Tomas Bj
ork, 2013 165
The Girsanov Theorem
Let W be a P -Wiener process. Fix a time horizon T .
Theorem: Choose an adapted process , and define
the process L by
(
dLt = Ltt dWt
L0 = 1
Assume that E P [LT ] = 1, and define a new mesure Q

on FT by
dQ = LtdP, on Ft, 0 t T
Then Q << P and the process W Q , defined by

Z t
WtQ = Wt sds
0
is Q-Wiener. We can also write this as
dWt = tdt + dWtQ
Tomas Bj
ork, 2013 166
Changing the drift in an SDE
The single most common use of the Girsanov Theorem
is as follows.
Suppose that we have a process X with P dynamics
dXt = tdt + t dWt
where and are adapted and W is P -Wiener.

We now do a Girsanov Transformation as above, and
the question is what the Q-dynamics look like.
From the Girsanov Theorem we have
dWt = tdt + dWtQ
and substituting this into the P -dynamics we obtain

the Q dynamics as
dXt = {t + tt} dt + tdWtQ
Moral: The drift changes but the diffusion is

unaffected.
Tomas Bj
ork, 2013 167
The Converse Girsanov Theorem
Let W be a P -Wiener process. Fix a time horizon T .
Theorem. Assume that:
Q << P on FT , with likelihood process
dQ
Lt = , on Ft 0, t T
dP
The filtation is the internal one .i.e.
Ft = {Ws; 0 s t}
Then there exists a process such that

(
dLt = Ltt dWt
L0 = 1
Tomas Bj
ork, 2013 168
The Martingale Approach
II: Pricing and Hedging

(Ch 10-12)
Tomas Bj
ork
Tomas Bj
ork, 2013 169
Financial Markets
Price Processes:

St = St0, ..., StN
Example: (Black-Scholes, S 0 := B, S 1 := S)
dSt = Stdt + St dWt,

dBt = rBt dt.
Portfolio:
ht = h0t , ..., hN
t
hit = number of units of asset i at time t.
Value Process:
N
X
Vth = hitSti = htSt
i=0
Tomas Bj
ork, 2013 170
Self Financing Portfolios
Definition: (intuitive)
A portfolio is self-financing if there is no exogenous
infusion or withdrawal of money. The purchase of a
new asset must be financed by the sale of an old one.
Definition: (mathematical)
A portfolio is self-financing if the value process
satisfies
XN
dVt = hitdSti
i=0
Major insight:
If the price process S is a martingale, and if h is
self-financing, then V is a martingale.
NB! This simple observation is in fact the basis of the

following theory.
Tomas Bj
ork, 2013 171
Arbitrage
The portfolio u is an arbitrage portfolio if
The portfolio strategy is self financing.
V0 = 0.
VT 0, P a.s.
P (VT > 0) > 0
Main Question: When is the market free of arbitrage?
Tomas Bj
ork, 2013 172
First Attempt
Proposition: If St0, , StN are P -martingales, then

the market is free of arbitrage.
Proof:
Assume that V is an arbitrage strategy. Since
N
X
dVt = hitdSti,
i=0
V is a P -martingale, so
V0 = E P [VT ] > 0.
This contradicts V0 = 0.
True, but useless.
Tomas Bj
ork, 2013 173
Example: (Black-Scholes)

dBt = rBt dt.
(We would have to assume that = r = 0)
We now try to improve on this result.
Tomas Bj
ork, 2013 174
Choose S0 as numeraire
Definition:
The normalized price vector Z is given by
St 1 N

Zt = 0 = 1, Zt , ..., Zt
St
The normalized value process V Z is given by
N
X
VtZ = hitZti .
0
Idea:
The arbitrage and self financing concepts should be
independent of the accounting unit.
Tomas Bj
ork, 2013 175
Invariance of numeraire
Proposition: One can show (see the book) that
S-arbitrage Z-arbitrage.
S-self-financing Z-self-financing.
Insight:
If h self-financing then
N
X
dVtZ = hitdZti
1
Thus, if the normalized price process Z is a P -

martingale, then V Z is a martingale.
Tomas Bj
ork, 2013 176
Second Attempt
Proposition: If Zt0, , ZtN are P -martingales, then

the market is free of arbitrage.
True, but still fairly useless.
Example: (Black-Scholes)

dBt = rBt dt.
dZt1 = ( r)Zt1 dt + Zt1dWt ,

dZt0 = 0dt.
We would have to assume risk-neutrality, i.e. that

= r.
Tomas Bj
ork, 2013 177
Arbitrage
Recall that h is an arbitrage if
h is self financing
V0 = 0.
VT 0, P a.s.
P (VT > 0) > 0
Major insight
This concept is invariant under an equivalent change
of measure!
Tomas Bj
ork, 2013 178
Martingale Measures
Definition: A probability measure Q is called an

equivalent martingale measure (EMM) if and only
if it has the following properties.
Q and P are equivalent, i.e.
QP
The normalized price processes
Sti
Zti = 0, i = 0, . . . , N
St
are Q-martingales.
Wan now state the main result of arbitrage theory.
Tomas Bj
ork, 2013 179
First Fundamental Theorem
Theorem: The market is arbitrage free
iff
there exists an equivalent martingale measure.
Tomas Bj
ork, 2013 180
Comments
It is very easy to prove that existence of EMM

imples no arbitrage (see below).
The other imnplication is technically very hard.
For discrete time and finite sample space the hard

part follows easily from the separation theorem for
convex sets.
For discrete time and more general sample space we

need the Hahn-Banach Theorem.
For continuous time the proof becomes technically

very hard, mainly due to topological problems. See
the textbook.
Tomas Bj
ork, 2013 181
Proof that EMM implies no arbitrage
Assume that there exists an EMM denoted by Q.
Assume that P (VT 0) = 1 and P (VT > 0) > 0.
Then, since P Q we also have Q(VT 0) = 1 and
Q(VT > 0) > 0.
Recall:
N
X
dVtZ = hitdZti
1
Q is a martingale measure
V Z is a Q-martingale

V0 = V0Z =E Q
VTZ >0
No arbitrage
Tomas Bj
ork, 2013 182
Choice of Numeraire
The numeraire price St0 can be chosen arbitrarily. The

most common choice is however that we choose S 0 as
the bank account, i.e.
St0 = Bt
where
dBt = rtBt dt
Here r is the (possibly stochastic) short rate and we

have
Rt
Bt = e 0 rs ds
Tomas Bj
ork, 2013 183
Example: The Black-Scholes Model

dBt = rBt dt.
Look for martingale measure. We set Z = S/B.
dZt = Zt( r)dt + ZtdWt ,
Girsanov transformation on [0, T ]:

(
dLt = LttdWt,
L0 = 1.
dQ = LT dP, on FT
Girsanov:
dWt = tdt + dWtQ,
where W Q is a Q-Wiener process.
Tomas Bj
ork, 2013 184
The Q-dynamics for Z are given by
dZt = Zt [ r + t ] dt + ZtdWtQ .
Unique martingale measure Q, with Girsanov kernel

given by
r
t = .

Q-dynamics of S:
dSt = rStdt + StdWtQ .
Conclusion: The Black-Scholes model is free of

arbitrage.
Tomas Bj
ork, 2013 185
Pricing
We consider a market Bt, St1, . . . , StN .
Definition:
A contingent claim with delivery time T , is a random
variable
X FT .
At t = T the amount X is paid to the holder of the
claim.
Example: (European Call Option)
X = max [ST K, 0]
Let X be a contingent T -claim.
Problem: How do we find an arbitrage free price

process t [X] for X?
Tomas Bj
ork, 2013 186
Solution
The extended market
Bt, St1, . . . , StN , t [X]
must be arbitrage free, so there must exist a martingale

measure Q for (St, t [X]). In particular
t [X]
Bt
must be a Q-martingale, i.e.

t [X] T [X]
= EQ Ft
Bt BT
Since we obviously (why?) have
T [X] = X
we have proved the main pricing formula.
Tomas Bj
ork, 2013 187
Theorem: For a T -claim X, the arbitrage free price is

given by the formula
h i
RT

t [X] = E Q e t rs ds
X Ft
Tomas Bj
ork, 2013 188
Example: The Black-Scholes Model
Q-dynamics:
dSt = rStdt + StdWtQ .
Simple claim:
X = (ST ),
t [X] = er(T t)E Q [(ST )| Ft]
Kolmogorov
t [X] = F (t, St)
where F (t, s) solves the Black-Scholes equation:

F 1 2 2 2F
t + rs F
s + 2 s s2 rF = 0,

F (T, s) = (s).
Tomas Bj
ork, 2013 189
Problem
Recall the valuation formula
h i
RT

X Ft
What if there are several different martingale measures

Q?
This is connected with the completeness of the

market.
Tomas Bj
ork, 2013 190
Hedging
Def: A portfolio is a hedge against X (replicates

X) if
h is self financing
VT = X, P a.s.
Def: The market is complete if every X can be

hedged.
Pricing Formula:
If h replicates X, then a natural way of pricing X is
t [X] = Vth
When can we hedge?
Tomas Bj
ork, 2013 191
Existence of hedge
Existence of stochastic integral

representation
Tomas Bj
ork, 2013 192
Fix T -claim X.
If h is a hedge for X then
X
VTZ = BT
h is self financing, i.e.
K
X
dVtZ = hitdZti
1
Thus V Z is a Q-martingale.

Z Q X
Vt = E Ft
BT
Tomas Bj
ork, 2013 193
Lemma:
Fix T -claim X. Define martingale M by

X
Mt = E Q Ft
BT
Suppose that there exist predictable processes

h1, , hN such that
N Z
X t
Mt = x + hisdZsi ,
i=1 0
Then X can be replicated.
Tomas Bj
ork, 2013 194
Proof
We guess that
N
X
Mt = VtZ = hB
t 1+ hitZti
i=1
Define: hB by
N
X
hB
t = Mt hitZti.
i=1
We have Mt = VtZ , and we get

N
X
dVtZ = dMt = hitdZti,
i=1
so the portfolio is self financing. Furthermore:

X X
VTZ = MT = E Q FT = .
BT BT
Tomas Bj
ork, 2013 195
Second Fundamental Theorem
The second most important result in arbitrage theory

is the following.
Theorem:
The market is complete
iff
the martingale measure Q is unique.
Proof: It is obvious (why?) that if the market

is complete, then Q must be unique. The other
implication is very hard to prove. It basically relies on
duality arguments from functional analysis.
Tomas Bj
ork, 2013 196
Black-Scholes Model
Q-dynamics
dSt = rSt dt + StdWtQ,

dZt = ZtdWtQ
Q
rT

Mt = E e X Ft ,
Representation theorem for Wiener processes

there exists g such that
Z t
Mt = M (0) + gsdWsQ.
0
Thus Z t
Mt = M0 + h1s dZs,
0
gt
with h1t = Zt .
Tomas Bj
ork, 2013 197
Result:
X can be replicated using the portfolio defined by
h1t = gt/Zt,
hB
t = Mt h1t Zt.
Moral: The Black Scholes model is complete.
Tomas Bj
ork, 2013 198
Special Case: Simple Claims
Assume X is of the form X = (ST )
Q
rT
Mt = E e (ST ) Ft ,
Kolmogorov backward equation Mt = f (t, St)

( 2
f f 1 2 2 f
t + rs s + 2 s s2 = 0,
f (T, s) = erT (s).
o
It
f
dMt = St dWtQ,
s
so
f
gt = St ,
s
Replicating portfolio h:
f
hB
t = f St ,
s
f
h1t = Bt .
s
Interpretation: f (t, St) = VtZ .
Tomas Bj
ork, 2013 199
Define F (t, s) by
F (t, s) = ertf (t, s)
so F (t, St) = Vt. Then

F (t,St )St F
s (t,St )
hB
t = Bt ,
h1t = F
s (t, St )
where F solves the Black-Scholes equation
(
F 1 2 2 2F
t + rs F
s + 2 s s2 rF = 0,
F (T, s) = (s).
Tomas Bj
ork, 2013 200
Main Results
The market is arbitrage free There exists a

martingale measure Q
The market is complete Q is unique.
Every X must be priced by the formula

h RT i

t [X] = E Q e t rsds X Ft
for some choice of Q.
In a non-complete market, different choices of Q

will produce different prices for X.
For a hedgeable claim X, all choices of Q will

produce the same price for X:
h RT i

t [X] = Vt = E Q e t rsds X Ft
Tomas Bj
ork, 2013 201
Completeness vs No Arbitrage
Rule of Thumb
Question:
When is a model arbitrage free and/or complete?
Answer:
Count the number of risky assets, and the number of
random sources.
R = number of random sources

N = number of risky assets
Intuition:
If N is large, compared to R, you have lots of
possibilities of forming clever portfolios. Thus lots
of chances of making arbitrage profits. Also many
chances of replicating a given claim.
Tomas Bj
ork, 2013 202
Rule of thumb
Generically, the following hold.
The market is arbitrage free if and only if
N R
The market is complete if and only if
N R
Example:
The Black-Scholes model.

dBt = rBt dt.
For B-S we have N = R = 1. Thus the Black-Scholes

model is arbitrage free and complete.
Tomas Bj
ork, 2013 203
Stochastic Discount Factors
Given a model under P . For every EMM Q we define

the corresponding Stochastic Discount Factor, or
SDF, by Rt
0 rs ds
Dt = e Lt ,
where
dQ
Lt = , on Ft
dP
There is thus a one-to-one correspondence between
EMMs and SDFs.
The risk neutral valuation formula for a T -claim X can

now be expressed under P instead of under Q.
Proposition: With notation as above we have
1 P
t [X] = E [DT X| Ft]
Dt
Proof: Bayes formula.
Tomas Bj
ork, 2013 204
Martingale Property of S D
Proposition: If S is an arbitrary price process, then

the process
StDt
is a P -martingale.
Proof: Bayes formula.
Tomas Bj
ork, 2013 205
Dividends,
Forwards, Futures, and Futures Options

Ch 16 & 26
Tomas Bj
ork
Tomas Bj
ork, 2013 206
Contents
1. Dividends
2. Forward and futures contracts
3. Futures options
Tomas Bj
ork, 2013 207
1. Dividends
Tomas Bj
ork, 2013 208
Dividends
Black-Scholes model:

dBt = rBt dt.
New feature:
The underlying stock pays dividends.
Dt = The cumulative dividends over

the interval [0, t]
Interpretation:
Over the interval [t, t + dt] you obtain the amount dDt
Two cases
Discrete dividends (realistic but messy).
Continuous dividends (unrealistic but easy to

handle).
Tomas Bj
ork, 2013 209
Portfolios and Dividends
Consider a market with N assets.
Sti = price at t, of asset No i

Dti = cumulative dividends for S i over
the interval [0, t]
hit = number of units of asset i
Vt = market value of the portfolio h at t
Assumption: We assume that D has continuous

trajctories.
Definition: The value process V is defined by
N
X
Vt = hitSti
i=1
Tomas Bj
ork, 2013 210
Recall:
N
X
Vt = hitSti
i=1
Definition: The strategy h is self financing if
N
X
dVt = hitdGit
i=1
where the gain process Gi is defined by
dGit = dSti + dDti
Interpret!
Note: The definitions above rely on the assumption

that D is continuous. In the case of a discontinuous
D, the definitions are more complicated.
Tomas Bj
ork, 2013 211
Relative weights
uit = the relative share of the portfolio value, which is

invested in asset No i.
hitSti
uit =
Vt
N
X
dVt = hitdGit
i=1
Substitute!
N
X i
i dGt
dVt = Vt ut i
i=1
St
Tomas Bj
ork, 2013 212
Continuous Dividend Yield
Definition: The stock S pays a continuous dividend

yield of q, if D has the form
dDt = qSt dt
Problem:
How does the dividend affect the price of a European
Call? (compared to a non-dividend stock).
Answer:
The price is lower. (why?)
Tomas Bj
ork, 2013 213
Black-Scholes with Cont. Dividend Yield

dDt = qStdt
Gain process:
dGt = ( + q)Stdt + StdWt
Consider a fixed claim
X = (ST )
and assume that
t [X] = F (t, St)
Tomas Bj
ork, 2013 214
Standard Procedure
Assume that the derivative price is of the form
t [X] = F (t, St).
Form a portfolio based on underlying S and

derivative F , with portfolio dynamics

dGt dF
dVt = Vt uSt + uF
t
St F
Choose uS and uF such that the dW -term is wiped

out. This gives us
dVt = Vt ktdt
kt = r
This relation will say something about F .
Tomas Bj
ork, 2013 215
Value dynamics:

S dG F dF
dV = V u +u ,
S F
dG = S( + q)dt + SdW.
From It
o we obtain
dF = F F dt + F F dW,
where
2

1 F F 1 2 2 F
F = + S + S ,
F t s 2 s2
1 F
F = S .
F s
Collecting terms gives us

S F

dV = V u ( + q) + u F dt
S F

+ V u + u F dW,
Tomas Bj
ork, 2013 216
Define uS and uF by the system
uS + uF F = 0,
uS + uF = 1.
Tomas Bj
ork, 2013 217
Solution
F
uS = ,
F

uF = ,
F
Value dynamics
S F

dV = V u ( + q) + u F dt.
uS ( + q) + uF F = r,
We get
F F 1 2 2 2F
+ (r q)S + S 2
rF = 0.
t s 2 s
Tomas Bj
ork, 2013 218
Pricing PDE
Proposition: The pricing function F is given as the

solution to the PDE
F F 1 2 2 2F
+ (r q)s + s rF = 0,
t s 2 s2
F (T, s) = (s).
We can now apply Feynman-Kac to the PDE in order

to obtain a risk neutral valuation formula.
Tomas Bj
ork, 2013 219
The pricing function has the representation
Q
F (t, s) = er(T t)Et,s [(ST )] ,
where the Q-dynamics of S are given by
dSt = (r q)St dt + StdWtQ.
Question: Which object is a martingale under the

meausre Q?
Tomas Bj
ork, 2013 220
Martingale Property
Proposition: Under the martingale measure Q the

normalized gain process
Z t
GZ
t =e
rt
St + erudDu
0
is a Q-martingale.
Proof: Exercise.
Note: The result above holds in great generality.
Interpretation:
In a risk neutral world, todays stock price should be
the expected value of all future discounted earnings
which arise from holding the stock.
Z t
S0 = E Q erudDu + ertSt ,
0
Tomas Bj
ork, 2013 221
Pricing formula
Pricing formula for claims of the type
Z = (ST )
We are standing at time t, with dividend yield q.

Todays stock price is s.
Suppose that you have the pricing function
F 0(t, s)
for a non dividend stock.
Denote the pricing function for the dividend paying

stock by
F q (t, s)

F q (t, s) = F 0 t, seq(T t)
Tomas Bj
ork, 2013 222
Moral
Use your old formulas, but replace todays stock price

s with seq(T t).
Tomas Bj
ork, 2013 223
European Call on Dividend-Paying-Stock
F q (t, s) = seq(T t)N [d1] er(T t)XN [d2] .
s
1 1
d1 = ln + r q + 2 (T t)
T t X 2

d2 = d1 T t.
Tomas Bj
ork, 2013 224
Martingale Analysis
Basic task: We have a general model for stock price S

and cumulative dividends D, under P . How do we find
a martingale measure Q, and exactly which objects will
be martingales under Q?
Main Idea: We attack this situation by reducing it

to the well known case of a market without dividends.
Then we apply standard techniques.
Tomas Bj
ork, 2013 225
The Reduction Technique
Consider the self financing portfolio where you keep

1 unit of the stock and invest all dividends in the
bank. Denote the portfolio value by V .
This portfolio can be viewed as a traded asset

without dividends.
Now apply the First Fundamental Theorem to the

market (B, V ) instead of the original market (B, S).
Thus there exists a martingale measure Q such that

t
Bt is a Q martingale for all traded assets (underlying
and derivatives) without dividends.
In particular the process
Vt
Bt
is a Q martingale.
Tomas Bj
ork, 2013 226
The V Process
Let ht denote the number of units in the bank account,

where h0 = 0. V is then characterized by
Vt = 1 St + htBt (1)
dVt = dSt + dDt + htdBt (2)
From (1) we obtain
dVt = dSt + htdBt + Btdht
Comparing this with (2) gives us
Btdht = dDt
Integrating this gives us

Z t
1
ht = dDs
0 Bs
Tomas Bj
ork, 2013 227
We thus have
Z t
1
Vt = St + Bt dDs (3)
0 Bs
and the first fundamental theorem gives us the

following result.
Proposition: For a market with dividends, the

martingale measure Q is characterized by the fact
that the normalized gain process
Z t
Z St 1
Gt = + dDs
Bt 0 Bs
is a Q martingale.
Quiz: Could you have guessed the formula (3) for V ?
Tomas Bj
ork, 2013 228
Continuous Dividend Yield
Model under P

dDt = qStdt
We recall Z t
St 1
GZ
t = + dDs
Bt 0 Bs
Easy calculation gives us
dGZ
t = Zt ( r + q) dt + Zt dWt
where Z = S/B.
Girsanov transformation dQ = LdP , where
dLt = Ltt dWt
We have
dWt = tdt + dWtQ
Insert this into dGZ
Tomas Bj
ork, 2013 229
The Q dynamics for GZ are
Q
dGZ
t = Zt ( r + q + t ) dt + ZtdWt
Martingale condition
r + q + t = 0
Q-dynamics of S
dSt = St ( + ) dt + StdWtQ
Using the martingale condition this gives us the Q-

dynamics of S as
dSt = St (r q) dt + StdWtQ
Tomas Bj
ork, 2013 230
Theorem: For a T -claim X, the price process t [X]

is given by
t [X] = er(T t)E Q [X| Ft] ,
where the Q-dynamics of S are given by
dSt = (r q)St dt + StdWtQ.
Tomas Bj
ork, 2013 231
2. Forward and Futures Contracts
Tomas Bj
ork, 2013 232
Forward Contracts
A forward contract on the T -claim X, contracted

at t, is defined by the following payment scheme.
The holder of the forward contract receives, at time

T , the stochastic amount X from the underwriter.
The holder of the contract pays, at time T , the

forward price f (t; T, X) to the underwriter.
The forward price f (t; T, X) is determined at time

t.
The forward price f (t; T, X) is determined in such

a way that the price of the forward contract equals
zero, at the time t when the contract is made.
Tomas Bj
ork, 2013 233
General Risk Neutral Formula
Suppose we have a bank account B with dynamics
dBt = rtBt dt, B0 = 1
with a (possibly stochastic) short rate rt. Then

Rt
Bt = e 0 rs ds
and we have the following risk neutral valuation for a

T -claim X
h RT i

t [X] = E Q e t rsds X Ft
Setting X = 1 we have the price, at time t, of a zero

coupon bond maturing at T as
h i
rs ds
RT
p(t, T ) = E Q e t Ft
Tomas Bj
ork, 2013 234
Forward Price Formula
Theorem: The forward price of the claim X is given

by
1 h RT i

f (t, T ) = E Q e t rsds X Ft
p(t, T )
where p(t, T ) denotes the price at time t of a zero

coupon bond maturing at time T .
In particular, if the short rate r is deterministic we have
f (t, T ) = E Q [X| Ft]
Tomas Bj
ork, 2013 235
Proof
The net cash flow at maturity is X f (t, T ). If the

value of this at time t equals zero we obtain
t [X] = t [f (t, T )]
We have
h i
RT

X Ft
and, since f (t, T ) is known at t, we obviously (why?)

have
t [f (t, T )] = p(t, T )f (t, T ).
This proves the main result. If r is deterministic then

p(t, T ) = er(T t) which gives us the second formula.
Tomas Bj
ork, 2013 236
Futures Contracts
A futures contract on the T -claim X, is a financial

asset with the following properties.
(i) At every point of time t with 0 t T , there exists

in the market a quoted object F (t; T, X), known as
the futures price for X at t, for delivery at T .
(ii) At the time T of delivery, the holder of the contract

pays F (T ; T, X) and receives the claim X.
(iii) During an arbitrary time interval (s, t] the holder

of the contract receives the amount F (t; T, X)
F (s; T, X).
(iv) The spot price, at any time t prior to delivery, for

buying or selling the futures contract, is by definition
equal to zero.
Tomas Bj
ork, 2013 237
Futures Price Formula
From the definition it is clear that a futures contract
is a price-dividend pair (S, D) with
S 0, dDt = dF (t, T )
From general theory, the normalized gains process

Z t
St 1
GZ
t = + dDs
Bt 0 Bs
is a Q-martingale.
Since S 0 and dDt = dF (t, T ) this implies that
1
dF (t, T )
Bt
is a martingale increment, which implies (why?) that

dF (t, T ) is a martingale increment. Thus F is a
Q-martingale and we have
F (t, T ) = E Q [F (T, T )| Ft] = E Q [X| Ft]
Tomas Bj
ork, 2013 238
Theorem: The futures price process is given by
F (t, T ) = E Q [X| Ft] .
Corollary. If the short rate is deterministic, then the

futures and forward prices coincide.
Tomas Bj
ork, 2013 239
3. Futures Options
Tomas Bj
ork, 2013 240
Futures Options
We denote the futures price process, at time t with

delivery time at T by
F (t, T ).
When T is fixed we sometimes suppress it and write

Ft, i.e. Ft = F (t, T )
Definition:
A European futures call option, with strike price X and
exercise date T , on a futures contract with delivery date
T1 will, if exercised at T , pay to the holder:
The amount F (T, T1 ) X in cash.
A long postition in the underlying futures contract.
NB! The long position above can immediately be

closed at no cost.
Tomas Bj
ork, 2013 241
Institutional fact:
The exercise date T of the futures option is typcally
very close to the date of delivery of the underlying T1
futures contract.
Tomas Bj
ork, 2013 242
Why do Futures Options exist?
On many markets (such as commodity markets)

the futures market is much more liquid than the
underlying market.
Futures options are typically settled in cash. This

relieves you from handling the underlying (tons of
copper, hundreds of pigs, etc.).
The market place for futures and futures options is

often the same. This facilitates hedging etc.
Tomas Bj
ork, 2013 243
Pricing Futures Options Black-76
We consider a futures contract with delivery date T1

and use the notation Ft = F (t, T1 ). We assume the
following dynamics for F .
dFt = Ft dt + FtdWt
Now suppose we want to price a derivative with exercise

date T with the T1 -futures price F as underlying, i.e.
a claim of the form
(FT )
This turns out to be quite easy.
Tomas Bj
ork, 2013 244
From risk neutral valuation we know that the price
process t [] is of the form
t [] = f (t, Ft)
where f is given by
Q
f (t, F ) = er(T t)Et,F [(FT )]
so it only remains to find the Q-dynamics for F .
We now recall
Proposition: The futures price process Ft is a Q-

martingale.
Thus the Q-dynamics of F are given by
dFt = FtdWtQ
Tomas Bj
ork, 2013 245
We thus have
Q
f (t, F ) = er(T t)Et,F [(FT )]
with Q-dynamics
dFt = FtdWtQ
Now recall the formula for a stock with continuous

dividend yield q.
Q
f (t, s) = er(T t)Et,s [(ST )]
with Q-dynamics
dSt = (r q)St + StdWtQ
Note: If we set q = r the formulas are identical!
Tomas Bj
ork, 2013 246
Pricing Formulas
Let f 0(t, s) be the pricing function for the contract

(ST ) for the case when S is a stock without dividends.
Let f (t, F ) be the pricing formula for the claim (FT ).
f (t, F ) = f 0(t, F er(T t))
Moral: Reset todays futures price F to F er(T t)

and use your formulas for stock options.
Tomas Bj
ork, 2013 247
Black-76 Formula
The price of a futures option with exercise date T and

exercise price X is given by
c = er(T t) {F N [d1] XN [d2]} .

1 F 1 2
d1 = ln + (T t) ,
T t X 2

d2 = d1 T t.
Tomas Bj
ork, 2013 248
Currency Derivatives
Ch 17
Tomas Bj
ork
Tomas Bj
ork, 2013 249
Pure Currency Contracts
Consider two markets, domestic (England) and foreign

(USA).
rd = domestic short rate

rf = foreign short rate
X = exchange rate
NB! The exchange rate X is quoted as
units of the domestic currency

unit of the foreign currency
Tomas Bj
ork, 2013 250
Simple Model (Garman-Kohlhagen)
The P -dynamics are given as:
dXt = Xt X dt + Xt X dWt,
dBtd = rd Btddt,
dBtf = rf Btf dt,
Main Problem:
Find arbitrage free price for currency derivative, Z, of
the form
Z = (XT )
Typical example: European Call on X.
Z = max [XT K, 0]
Tomas Bj
ork, 2013 251
Naive idea
For the European Call, use the standard Black-Scholes

formula, with S replaced by X and r replaced by rd.
Is this OK?
Tomas Bj
ork, 2013 252
NO!
WHY?
Tomas Bj
ork, 2013 253
Main Idea
When you buy stock you just keep the asset until
you sell it.
When you buy dollars, these are put into a bank

account, giving the interest rf .
Moral:
Buying a currency is like buying a dividend-paying
stock with dividend yield q = rf .
Tomas Bj
ork, 2013 254
Technique
Transform all objects into domestically traded

asset prices.
Use standard techniques on the transformed model.
Tomas Bj
ork, 2013 255
Transformed Market
1. Investing foreign currency in the foreign bank gives

value dynamics in foreign currency according to
dBtf = rf Btf dt.
2. Bf units of the foreign currency is worth X Bf in

the domestic currency.
3. Trading in the foreign currency is equivalent to

trading in a domestic market with the domestic
price process
Btf = Btf Xt
4. Study the domestic market consisting of
f ,
B Bd
Tomas Bj
ork, 2013 256
Market dynamics
dXt = Xt X dt + XtX dW
tf = Btf Xt
B
Using It
o we have domestic market dynamics

tf
dB f f tf X dWt
= Bt X + r dt + B
dBtd = rdBtddt
Standard results gives us Q-dynamics for domestically

traded asset prices:
tf
dB tf rd dt + B
= B tf X dWtQ
dBtd = rd Btddt
It tf /Btf :
o gives us Q-dynamics for Xt = B
dXt = Xt(rd rf )dt + Xt X dWtQ
Tomas Bj
ork, 2013 257
Risk neutral Valuation
Theorem: The arbitrage free price t [] is given by

t [] = F (t, Xt) where
d Q
F (t, x) = er (T t)
Et,x [(XT )]
The Q-dynamics of X are given by
dXt = Xt(rd rf )dt + Xt X dWtQ
Tomas Bj
ork, 2013 258
Pricing PDE
Theorem:The pricing function F solves the boundary

value problem
F d f F 1 2 2 2F
+ x(r r ) + x X 2 rdF = 0,
t x 2 x
F (T, x) = (x)
Tomas Bj
ork, 2013 259
Currency vs Equity Derivatives
Proposition: Introduce the notation:
F 0(t, x) = the pricing function for the claim Z =

(XT ), where we interpret X as the price of an
ordinary stock without dividends.
F (t, x) = the pricing function of the same claim

when X is interpreted as an exchange rate.
Then the following holds

f

F (t, x) = F0 t, xer (T t) .
Tomas Bj
ork, 2013 260
Currency Option Formula
The price of a European currency call is given by
f d
F (t, x) = xer (T t)
N [d1] er (T t)
KN [d2] ,
where

1 x 1
d1 = ln + rd rf + X
2
(T t)
X T t K 2

d2 = d1(t, x) X T t
Tomas Bj
ork, 2013 261
Martingale Analysis
Qd = domestic martingale measure

Qf = foreign martingale measure
dQf dQd dQf

Lt = , Ldt = , Lft =
dQd dP dP
P -dynamics of X
dXt = Xt tdt + Xtt dWt
where and are arbitrary adapted processes and W

is P -Wiener.
Problem: How are Qd and Qf related?
Tomas Bj
ork, 2013 262
Main Idea
Fix an arbitrary foreign T -claim Z.
Compute foreign price and change to domestic

currency. The price at t = 0 will be
h RT f i
Qf 0 rs ds
0 [Z] = X0 E e Z
This can be written as

d
h RT f i
0 [Z] = X0 E Q LT e 0 rs dsZ
Change into domestic currency at T and then

compute arbitrage free price. This gives us
d
h RT d i
0 [Z] = E Q e 0 rs dsXT Z
These expressions must be equal for all choices of

Z FT .
Tomas Bj
ork, 2013 263
We thus obtain
d
h RT d i d
h RT f i
E Q e 0 rs dsXT Z = X0 E Q LT e 0 rs dsZ
for all T -claims Z. This implies the following result.
Theorem: The exchange rate X is given by

Rt d f
Xt = X0 e 0 (rs rs )ds Lt
alternatively by
Dtf
Xt = X0 d
Dt
where Dtd is the domestic stochastic discount factor
etc.
Proof: The last part follows from
f f

dQ dQ dQd
L= d
=
dQ dP dP
Tomas Bj
ork, 2013 264
Qd-Dynamics of X
In particular, since L is a Qd-martingale the Qd

dynamics of L are of the form
dLt = Ltt dWtd
where W d is Qd-Wiener. From

Rt d f
Xt = X0 e 0 (rs rs )ds Lt
the Qd-dynamics of X follows as
dXt = (rtd rtf )Xt dt + Xt tdWtd
so the Girsanov kernel equals the exchange rate

volatility and we have the general Qd dynamics.
Theorem: The Qd dynamics of X are of the form
dXt = (rtd rtf )Xt dt + Xtt dWtd
Tomas Bj
ork, 2013 265
Market Prices of Risk
Recall
0t rsdds d
R
Dtd = e Lt
We also have
dLdt = LdtdtdWt
where dt = d is the domestic market price of risk

and similar for f etc. From
Dtf
Xt = X0 d
Dt
we now easily obtain

d f
dXt = Xt tdt + Xt t t dWt,
where we do not care about the exact shape of . We

thus have
Theorem: The exchange rate volatility is given by
t = dt ft
Tomas Bj
ork, 2013 266
Siegels Paradox
Assume that the domestic and the foreign markets are
risk neutral and assume constant short rates. We now
have the following surprising (?) argument.
A: Let us consider a T claim of 1 dollar. The arbitrage
free dollar value at t = 0 is of course
r f T
e
so the Euro value at at t = 0 is given by
r f T
X0e .
The 1-dollar claim is, however, identical to a T -claim

of XT euros. Given domestic risk neutrality, the Euro
value at t = 0 is then
r dT
e E P [XT ] .
We thus have
f d
X0 er T
= er T E P [XT ]
Tomas Bj
ork, 2013 267
Siegels Paradox ctd
B: We now consider a T -claim of one Euro and

compute the dollar value of this claim. The Euro
value at t = 0 is of course
d
er T
so the dollar value is

1 rdT
e .
X0
The 1-Euro claim is identical to a T -claim of XT1

Euros so, by foreign risk neutrality, we obtain the
dollar price as
r f T P 1
e E
XT
which gives us

1 r d
T f 1
e = er T
EP
X0 XT
Tomas Bj
ork, 2013 268
Siegels Paradox ctd
Recall our earlier results
r f T r d T
X0 e = e E P [XT ]

1 rdT f 1
e = er T E P
X0 XT
Combining these gives us

P 1 1
E = P
XT E [XT ]
which, by Jensens inequality, is impossible unless XT

is deterministic. This is sometimes referred to as (one
formulation of) Siegels paradox.
It thus seems that Americans cannot be risk neutral at

the same time as Europeans.
What is going on?
Tomas Bj
ork, 2013 269
Formal analysis of Siegels Paradox
Question: Can we assume that both the domestic and

the foreign markets are risk neutral?
Answer: Generally no.
Proof: The assumption would be equivalent to

assuming the P = Qd = Qf i.e.
dt = ft = 0
However, we know that
t = dt ft
so we would need to have t = 0 i.e. a non-stochastic

exchange rate.
Tomas Bj
ork, 2013 270
Moral
The previous slide gave us the mathematical result, but

the intuitive question remains why Americans cannot
be risk neutral at the same time as Europeans.
The solution is roughly as follows.
Risk neutrality (or risk aversion) is always defined

in terms of a given numeraire.
It is not an attitude towards risk as such.
You can therefore not be risk neutral w.r.t two

different numeraires at the same time unless the
ratio between them is deterministic.
In particular we cannot have risk neutrality w.r.t.

Dollars and Euros at the same time.
Tomas Bj
ork, 2013 271
Change of Numeraire
Ch 26
Tomas Bj
ork
Tomas Bj
ork, 2013 272
Recap of General Theory
Consider a market with asset prices
St0, St1, . . . , StN
Theorem: The market is arbitrage free

iff
there exists an EMM, i.e. a measure Q such that
Q and P are equivalent, i.e.
QP
The normalized price processes
St0 St1 StN

, ,..., 0
St0 St0 St
are Q-martingales.
Tomas Bj
ork, 2013 273
Recap continued
Recall the normalized market

St0 St1 StN
Zt0, Zt1, . . . Zt
N
= , ...,
St0 St0 St0
We obviously have
Zt0 1
Thus Z 0 is a risk free asset in the normalized

economy.
Z 0 is a bank account in the normalized economy.
In the normalized economy the short rate is zero.
Tomas Bj
ork, 2013 274
Dependence on numeraire
The EMM Q will obviously depend on the choice

of numeraire, so we should really write Q0 to
emphasize that we are using S 0 as numeraire.
So far we have only considered the case when the

numeraire asset is the bank account, i.e. when
St0 = Bt. In this case, the martingale measure
QB is referred to as the risk neutral martingale
measure.
Henceforth the notation Q (without upper case

index) will only be used for the risk neutral
martingale measure, i.e. Q = QB .
We will now consider the case of a general

numeraire.
Tomas Bj
ork, 2013 275
General change of numeraire.
Consider a financial market, including a bank

account B.
Assume that the market is using a fixed risk neutral

measure Q as pricing measure.
Choose a fixed asset S as numeraire, and denote

the corresponding martingale measure by QS .
Problems:
Determine QS , i.e. determine
dQS
Lt = , on Ft
dQ
Develop pricing formulas for contingent claims using

QS instead of Q.
Tomas Bj
ork, 2013 276
Constructing QS
Fix a T -claim X. From general theory we know that

Q X
0 [X] = E
BT
Since QS is a martingale measure for the numeraire S,

the normalized process
t [X]
St
is a QS -martingale. We thus have

0 [X] S T [X] S X Q X
=E =E = E LT
S0 ST ST ST
From this we obtain

Q X S0
0 [X] = E LT ,
ST
Tomas Bj
ork, 2013 277
For all X FT we thus have

Q X Q X S0
E = E LT
BT ST
Recall the following basic result from probability theory.
Proposition: Consider a probability space (, F, P )

and assume that
E [Y X] = E [Z X] , for all Z F.
Then we have
Y = Z, P a.s.
From this result we conclude that

1 S0
= LT
BT ST
Tomas Bj
ork, 2013 278
Main result
Proposition: The likelihood process
dQS
Lt = , on Ft
dQ
is given by
St 1
Lt =
Bt S0
Tomas Bj
ork, 2013 279
Easy exercises
1. Convince yourself that L is a Q-martingale.
2. Assume that a process At has the property that

At/Bt is a Q martingale. Show that this implies
that At/St is a QS -martingale. Interpret the result.
Tomas Bj
ork, 2013 280
Pricing
Theorem: For every T -claim X we have the pricing

formula
S X

t [X] = StE Ft
St
Proof: Follows directly from the QS -martingale

property of t [X] /St.
Note 1: We observe St directly on the market.
Note 2: The pricing formula above is particularly

useful when X is of the form
X = ST Y
In this case we obtain
t [X] = StE S [Y | Ft]
Tomas Bj
ork, 2013 281
Important example
Consider a claim of the form

X = ST0 , ST1
We assume that is linearly homogeneous, i.e.
(x, y) = (x, y), for all > 0
Using Q0 we obtain
" #
ST0 , ST1
t [X] = St0E 0 Ft
ST0

1
0 0 ST
t [X] = t [X] = St E 1, 0 Ft
ST
Tomas Bj
ork, 2013 282
Important example cntd
Proposition: For a claim of the form

X= ST0 , ST1 ,
where is homogeneous, we have
t [X] = St0 E 0 [ (ZT )| Ft]
where
St1
(z) = [1, z] , Zt = 0
St
Tomas Bj
ork, 2013 283
Exchange option
Consider an exchange option, i.e. a claim X given by

X = max ST1 ST0 , 0
Since (x, y) = max [x y, 0] is homogeneous we

obtain
t [X] = St0E 0 [max [ZT 1, 0]| Ft]
This is a European Call on Z with strike price K.
Zero interest rate.
Piece of cake!
If S 0 and S 1 are both GBM, then so is Z, and the

price will be given by the Black-Scholes formula.
Tomas Bj
ork, 2013 284
Identifying the Girsanov Transformation
Assume the Q-dynamics of S are known as
dSt = rtStdt + StvtdWtQ
St
Lt =
S0 Bt
From this we immediately have
dLt = LtvtdWtQ .
and we can summarize.
Theorem: The Girsanov kernel is given by the

numeraire volatility vt, i.e.
dLt = LtvtdWtQ .
Tomas Bj
ork, 2013 285
Recap on zero coupon bonds
Recall: A zero coupon T -bond is a contract which
gives you the claim
X1
at time T .
The price process t [1] is denoted by p(t, T ).
Allowing a stochastic short rate rt we have
dBt = rtBt dt.
This gives us Rt
Bt = e 0 rs ds ,
and using standard risk neutral valuation we have

h RT i

p(t, T ) = E Q e t rsds Ft
Note:
p(T, T ) = 1
Tomas Bj
ork, 2013 286
The forward measure QT
Consider a fixed T .
Choose the bond price process p(t, T ) as numeraire.
The corresponding martingale measure is denoted

by QT and referred to as the T -forward measure.
For any T claim X we obtain

QT T [X]
t [X] = p(t, T )E Ft
p(T, T )
We have
T [X] = X, p(T, T ) = 1
Theorem: For any T -claim X we have

T
t [X] = p(t, T )E Q [X| Ft]
Tomas Bj
ork, 2013 287
A general option pricing formula
European call on asset S with strike price K and maturity T .
X = max [ST K, 0]
Write X as
X = (ST K) I {ST K} = ST I {ST K} KI {ST K}
Use QS on the first term and QT on the second.
0 [X] = S0 QS [ST K] K p(0, T ) QT [ST K]
Tomas Bj
ork, 2013 288
Incomplete Markets
Ch 15
Tomas Bj
ork
Tomas Bj
ork, 2013 289
Derivatives on Non Financial Underlying
Recall: The Black-Scholes theory assumes that the

market for the underlying asset has (among other
things) the following properties.
The underlying is a liquidly traded asset.
Shortselling allowed.
Portfolios can be carried forward in time.
There exists a large market for derivatives, where the

underlying does not satisfy these assumptions.
Examples:
Weather derivatives.
Derivatives on electric energy.
CAT-bonds.
Tomas Bj
ork, 2013 290
Typical Contracts
Weather derivatives:
Heating degree days. Payoff at maturity T is
given by
Z = max {XT 30, 0}
where XT is the (mean) temperature at some place.
Electricity option:
The right (but not the obligation) to buy, at time
T , at a predetermined price K, a constant flow of
energy over a predetermined time interval.
CAT bond:
A bond for which the payment of coupons and
nominal value is contingent on some (well specified)
natural disaster to take place.
Tomas Bj
ork, 2013 291
Problems
Weather derivatives:
The temperature is not the price of a traded asset.
Electricity derivatives:
Electric energy cannot easily be stored.
CAT-bonds:
Natural disasters are not traded assets.
We will treat all these problems within a factor model.
Tomas Bj
ork, 2013 292
Typical Factor Model Setup
Given:
An underlying factor process X, which is not the

price process of a traded asset, with dynamics under
the objective probability measure P as
dXt = (t, Xt ) dt + (t, Xt ) dWt .
A risk free asset with dynamics
dBt = rBt dt,
Problem:
Find arbitrage free price t [Z] of a derivative of the
form
Z = (XT )
Tomas Bj
ork, 2013 293
Concrete Examples
Assume that Xt is the temperature at time t at the

village of Peniche (Portugal).
Heating degree days:
(XT ) = 100 max {XT 30, 0}
Holiday Insurance:

1000, if XT < 20
(XT ) =

0, if XT 20
Tomas Bj
ork, 2013 294
Question
Is the price t [] uniquely determined by the P -

dynamics of X, and the requirement of an arbitrage
free derivatives market?
Tomas Bj
ork, 2013 295
NO!!
WHY?
Tomas Bj
ork, 2013 296
Stock Price Model Factor Model
Black-Scholes:
dSt = Stdt + StdWt,

dBt = rBtdt.
Factor Model:
dXt = (t, Xt)dt + (t, Xt)dWt ,

dBt = rBt dt.
What is the difference?
Tomas Bj
ork, 2013 297
Answer
X is not the price of a traded asset!
We can not form a portfolio based on X.
Tomas Bj
ork, 2013 298
1. Rule of thumb:
N = 0, (no risky asset)

R = 1, (one source of randomness, W )
We have N < R. The exogenously given market,

consisting only of B, is incomplete.
2. Replicating portfolios:
We can only invest money in the bank, and then sit
down passively and wait.
We do not have enough underlying assets in order

to price X-derivatives.
Tomas Bj
ork, 2013 299
There is not a unique price for a particular
derivative.
In order to avoid arbitrage, different derivatives

have to satisfy internal consistency relations.
If we take one benchmark derivative as given,

then all other derivatives can be priced in terms of
the market price of the benchmark.
We consider two given claims (XT ) and (XT ). We

assume they are traded with prices
t [] = f (t, Xt)
t [] = g(t, Xt )
Tomas Bj
ork, 2013 300
Program:
Form portfolio based on and . Use It

o on f and
g to get portfolio dynamics.

df dg
dV = V uf + ug
f g
Choose portfolio weights such that the dW term

vanishes. Then we have
dV = V kdt,
(synthetic bank with k as the short rate)
k=r
Read off the relation k = r!
Tomas Bj
ork, 2013 301
From It
o:
df = f f dt + f f dW,
where (
ft +fx+ 12 2 fxx
f = f ,
fx
f = f .
Portfolio dynamics

df dg
dV = V uf + ug .
f g
Reshuffling terms gives us

f g
f g

dV = V u f + u g dt + V u f + u g dW.
Let the portfolio weights solve the system

uf + ug = 1,
uf f + ug g = 0.
Tomas Bj
ork, 2013 302
f g
u = ,
f g
f
ug = ,
f g
Portfolio dynamics
f g

dV = V u f + u g dt.
i.e.
g f f g
dV = V dt.
f g
Absence of arbitrage requires
g f f g
=r
f g
which can be written as

g r f r
= .
g f
Tomas Bj
ork, 2013 303
g r f r
= .
g f
Note!
The quotient does not depend upon the particular
choice of contract.
Tomas Bj
ork, 2013 304
Result
Assume that the market for X-derivatives is free of

arbitrage. Then there exists a universal process ,
such that
f (t) r
= (t, Xt),
f (t)
holds for all t and for every choice of contract f .
NB: The same for all choices of f .

= Risk premium per unit of volatility
= Market Price of Risk (cf. CAPM).
= Sharpe Ratio
Slogan:
On an arbitrage free market all X-derivatives have
the same market price of risk.
The relation
f r
=
f
is actually a PDE!
Tomas Bj
ork, 2013 305
Pricing Equation

f + { } f + 1 2 f rf = 0
t x xx
2
f (T, x) = (x),
P -dynamics:
dX = (t, X)dt + (t, X)dW.
Can we solve the PDE?
Tomas Bj
ork, 2013 306
No!!
Why??
Tomas Bj
ork, 2013 307
Answer
Recall the PDE

f + { } f + 1 2 f rf = 0
t x xx
2
f (T, x) = (x),
In order to solve the PDE we need to know .
is not given exogenously.
is not determined endogenously.
Tomas Bj
ork, 2013 308
Question:
Who determines ?
Tomas Bj
ork, 2013 309
Answer:
THE MARKET!
Tomas Bj
ork, 2013 310
Interpreting
Recall that the f dynamics are
df = f f dt + f f dWt
and is defined as
f (t) r
= (t, Xt),
f (t)
measures the aggregate risk aversion in the

market.
If is big then the market is highly risk averse.
If is zero then the market is risk netural.
If you make an assumption about , then you

implicitly make an assumption about the aggregate
risk aversion of the market.
Tomas Bj
ork, 2013 311
Moral
Since the market is incomplete the requirement of

an arbitrage free market will not lead to unique
prices for X-derivatives.
Prices on derivatives are determined by two main

factors.
1. Partly by the requirement of an arbitrage free
derivative market. All pricing functions satisfies
the same PDE.
2. Partly by supply and demand on the market.
These are in turn determined by attitude towards
risk, liquidity consideration and other factors. All
these are aggregated into the particular used
(implicitly) by the market.
Tomas Bj
ork, 2013 312
We recall the PDE

f + { } f + 1 2 f rf = 0
t x xx
2
f (T, x) = (x),
Using Feynman-Kac we obtain a risk neutral valuation

formula.
Tomas Bj
ork, 2013 313
Q
f (t, x) = er(T t)Et,x [(XT )]
Q-dynamics:
dXt = { } dt + dWtQ
Price = expected value of future payments
The expectation should not be taken under the

objective probabilities P , but under the risk
adjusted probabilities Q.
Tomas Bj
ork, 2013 314
Interpretation of the risk adjusted
probabilities
The risk adjusted probabilities can be interpreted as

probabilities in a (fictuous) risk neutral world.
When we compute prices, we can calculate as if

we live in a risk neutral world.
This does not mean that we live in, or think that

we live in, a risk neutral world.
The formulas above hold regardless of the attitude

towards risk of the investor, as long as he/she prefers
more to less.
Tomas Bj
ork, 2013 315
Diversification argument about
If the risk factor is idiosyncratic and diversifiable,

then one can argue that the factor should not be
priced by the market. Compare with APT.
Mathematically this means that = 0, i.e. P = Q,

i.e. the risk neutral distribution coincides with
the objective distribution.
We thus have the actuarial pricing formula
f (t, x) = er(T t)Et,x

P
[(XT )]
where we use the objective probabiliy measure P .
Tomas Bj
ork, 2013 316
Modeling Issues
Temperature:
A standard model is given by
dXt = {m(t) bXt} dt + dWt,
where m is the mean temperature capturing

seasonal variations. This often works reasonably
well.
Electricity:
A (naive) model for the spot electricity price is
dSt = St {m(t) a ln St} dt + StdWt
This implies lognormal prices (why?). Electricty

prices are however very far from lognormal, because
of spikes in the prices. Complicated.
CAT bonds:
Here we have to use the theory of point processes
and the theory of extremal statistics to model
natural disasters. Complicated.
Tomas Bj
ork, 2013 317
Martingale Analysis
Model: Under P we have
dXt = (t, Xt ) dt + (t, Xt ) dWt ,

dBt = rBt dt,
We look for martingale measures. Since B is the only

traded asset we need to find Q P such that
Bt
=1
Bt
is a Q martingale.
Result: In this model, every Q P is a martingale

measure.
Girsanov
dLt = Ltt dWt
Tomas Bj
ork, 2013 318
P -dynamics
dXt = (t, Xt ) dt + (t, Xt ) dWt ,
dLt = Ltt dWt

dQ = LtdP on Ft
Girsanov:
dWt = tdt + dWtQ
Martingale pricing:
F (t, x) = er(T t)E Q [Z| Ft]
Q-dynamics of X:
dXt = { (t, Xt ) + (t, Xt) t} dt + (t, Xt ) dWtQ,
Result: We have t = t, i.e,. the Girsanov kernel

equals minus the market price of risk.
Tomas Bj
ork, 2013 319
Several Risk Factors
We recall the dynamics of the f -derivative
df = f f dt + f f dWt
and the Market Price of Risk

f r
= , i.e. f r = f .
f
In a multifactor model of the type

n
X
dXt = (t, Xt ) dt + i (t, Xt ) dWti ,
i=1
it follows from Girsanov that for every risk factor W i

there will exist a market price of risk i = i such
that n
X
f r = i i
i=1
Compare with CAPM.
Tomas Bj
ork, 2013 320
Stochastic Control Theory
Ch 19
Tomas Bj
ork
Tomas Bj
ork, 2013 321
Contents
1. Dynamic programming.
2. Investment theory.
Tomas Bj
ork, 2013 322
1. Dynamic Programming
The basic idea.
Deriving the HJB equation.
The verification theorem.
The linear quadratic regulator.
Tomas Bj
ork, 2013 323
Problem Formulation
"Z #
T
max E F (t, Xt, ut)dt + (XT )
u 0
subject to
dXt = (t, Xt, ut) dt + (t, Xt , ut) dWt

X0 = x0,
ut U (t, Xt), t.
We will only consider feedback control laws, i.e.

controls of the form
ut = u(t, Xt)
Terminology:
X = state variable
u = control variable
U = control constraint
Note: No state space constraints.
Tomas Bj
ork, 2013 324
Main idea
Embedd the problem above in a family of problems

indexed by starting point in time and space.
Tie all these problems together by a PDE: the

Hamilton Jacobi Bellman equation.
The control problem is reduced to the problem of

solving the deterministic HJB equation.
Tomas Bj
ork, 2013 325
Some notation
For any fixed vector u Rk , the functions u, u

and C u are defined by
u(t, x) = (t, x, u),

u(t, x) = (t, x, u),
C u(t, x) = (t, x, u)(t, x, u)0 .
For any control law u, the functions u , u, C u(t, x)

and F u(t, x) are defined by
u(t, x) = (t, x, u(t, x)),

u (t, x) = (t, x, u(t, x)),
C u(t, x) = (t, x, u(t, x))(t, x, u(t, x))0 ,
F u (t, x) = F (t, x, u(t, x)).
Tomas Bj
ork, 2013 326
More notation
For any fixed vector u Rk , the partial differential

operator Au is defined by
n
X X n
u u 1 u 2
A = i (t, x) + Cij (t, x) .
i=1
xi 2 i,j=1 xi xj
For any control law u, the partial differential

operator Au is defined by
n
X X n
1 2
A =
u
i (t, x)
u
+ Cij (t, x)
u
.
i=1
x i 2 i,j=1
xi xj
For any control law u, the process X u is the solution

of the SDE
dXtu = (t, Xtu, ut) dt + (t, Xtu, ut) dWt,
where
ut = u(t, Xtu)
Tomas Bj
ork, 2013 327
Embedding the problem
For every fixed (t, x) the control problem Pt,x is defined

as the problem to maximize
"Z #
T
Et,x F (s, Xsu, us)ds + (XTu ) ,
t
given the dynamics
dXsu = (s, Xsu , us) ds + (s, Xsu , us) dWs,

Xt = x,
and the constraints
u(s, y) U, (s, y) [t, T ] Rn.
The original problem was P0,x0 .
Tomas Bj
ork, 2013 328
The optimal value function
The value function
J : R+ Rn U R
is defined by
"Z #
T
J (t, x, u) = E F (s, Xsu, us)ds + (XTu )
t
given the dynamics above.
The optimal value function
V : R+ Rn R
is defined by
V (t, x) = sup J (t, x, u).

uU
We want to derive a PDE for V .
Tomas Bj
ork, 2013 329
Assumptions
We assume:
There exists an optimal control law u

.
The optimal value function V is regular in the sense

that V C 1,2 .
A number of limiting procedures in the following

arguments can be justified.
Tomas Bj
ork, 2013 330
Bellman Optimality Principle
Theorem: If a control law u is optimal for the time

interval [t, T ] then it is also optimal for all smaller
intervals [s, T ] where s t.
Proof: Exercise.
Tomas Bj
ork, 2013 331
Basic strategy
To derive the PDE do as follows:
Fix (t, x) (0, T ) Rn.
Choose a real number h (interpreted as a small

time increment).
Choose an arbitrary control law u on the time inerval

[t, t + h].
Now define the control law u? by

? u(s, y), (s, y) [t, t + h] Rn
u (s, y) =
(s, y),
u (s, y) (t + h, T ] Rn.
In other words, if we use u? then we use the arbitrary

control u during the time interval [t, t + h], and then
we switch to the optimal control law during the rest of
the time period.
Tomas Bj
ork, 2013 332
Basic idea
The whole idea of DynP boils down to the following

procedure.
Given the point (t, x) above, we consider the

following two strategies over the time interval [t, T ]:
.
I: Use the optimal law u
II: Use the control law u? defined above.
Compute the expected utilities obtained by the

respective strategies.
Using the obvious fact that u is least as good

as u?, and letting h tend to zero, we obtain our
fundamental PDE.
Tomas Bj
ork, 2013 333
Strategy values
:
Expected utility for u
J (t, x, u
) = V (t, x)
Expected utility for u?:
The expected utility for [t, t + h) is given by

"Z #
t+h
Et,x F (s, Xsu, us) ds .
t
Conditional expected utility over [t + h, T ], given

(t, x):
Et,x V (t + h, Xt+h) .
u
Total expected utility for Strategy II is

"Z #
t+h
Et,x F (s, Xsu , us) ds + V (t + h, Xt+h
u
) .
t
Tomas Bj
ork, 2013 334
Comparing strategies
We have trivially
"Z #
t+h
V (t, x) Et,x F (s, Xsu, us) ds + V (t + h, Xt+h
u
) .
t
Remark
We have equality above if and only if the control law
.
u is the optimal law u
Now use It
o to obtain
V (t + h, Xt+h
u
) = V (t, x)
Z t+h
V
+ (s, Xsu) + AuV (s, Xsu ) ds
t t
Z t+h
+ xV (s, Xsu) u dWs,
t
and plug into the formula above.
Tomas Bj
ork, 2013 335
We obtain
"Z #
t+h
V
Et,x F (s, Xsu, us) + (s, Xsu ) + AuV (s, Xsu) ds 0.
t t
Going to the limit:

Divide by h, move h within the expectation and let h tend to zero.
We get
V
F (t, x, u) + (t, x) + AuV (t, x) 0,
t
Tomas Bj
ork, 2013 336
Recall
V
F (t, x, u) + (t, x) + AuV (t, x) 0,
t
This holds for all u = u(t, x), with equality if and only
.
if u = u
We thus obtain the HJB equation
V
(t, x) + sup {F (t, x, u) + AuV (t, x)} = 0.
t uU
Tomas Bj
ork, 2013 337
The HJB equation
Theorem:
Under suitable regularity assumptions the follwing hold:
I: V satisfies the HamiltonJacobiBellman equation
V
(t, x) + sup {F (t, x, u) + AuV (t, x)} = 0,
t uU
V (T, x) = (x),
II: For each (t, x) [0, T ] Rn the supremum in the

HJB equation above is attained by u = u (t, x), i.e. by
the optimal control.
Tomas Bj
ork, 2013 338
Logic and problem
Note: We have shown that if V is the optimal value

function, and if V is regular enough, then V satisfies
the HJB equation. The HJB eqn is thus derived
as a necessary condition, and requires strong ad hoc
regularity assumptions, alternatively the use of viscosity
solutions techniques.
Problem: Suppose we have solved the HJB equation.

Have we then found the optimal value function and
the optimal control law? In other words, is HJB a
sufficient condition for optimality.
Answer: Yes! This follows from the Verification

Theorem.
Tomas Bj
ork, 2013 339
The Verification Theorem
Suppose that we have two functions H(t, x) and g(t, x), such
that
H is sufficiently integrable, and solves the HJB equation

8
< H (t, x) + sup {F (t, x, u) + AuH(t, x)} = 0,
>
t uU
>
: H(T , x) = (x),
For each fixed (t, x), the supremum in the expression
sup {F (t, x, u) + AuH(t, x)}

uU
is attained by the choice u = g(t, x).
Then the following hold.
1. The optimal value function V to the control problem is given

by
V (t, x) = H(t, x).
, and in fact
2. There exists an optimal control law u
(t, x) = g(t, x)
u
Tomas Bj
ork, 2013 340
Handling the HJB equation
1. Consider the HJB equation for V .
2. Fix (t, x) [0, T ] Rn and solve, the static optimization
problem
u
max [F (t, x, u) + A V (t, x)] .
uU
Here u is the only variable, whereas t and x are fixed
parameters. The functions F , , and V are considered as
given.
3. The optimal u, will depend on t and x, and on the function
V and its partial derivatives. We thus write u
as
=u
u (t, x; V ) . (4)
4. The function u (t, x; V ) is our candidate for the optimal

control law, but since we do not know V this description is
incomplete. Therefore we substitute the expression for u
into
the PDE , giving us the highly nonlinear (why?) PDE
V
(t, x) + F u (t, x) + Au (t, x) V (t, x) = 0,
t
V (T , x) = (x).
5. Now we solve the PDE above! Then we put the solution V

into expression (4). Using the verification theorem we can
identify V as the optimal value function, and u
as the optimal
control law.
Tomas Bj
ork, 2013 341
Making an Ansatz
The hard work of dynamic programming consists in

solving the highly nonlinear HJB equation
There are no general analytic methods available

for this, so the number of known optimal control
problems with an analytic solution is very small
indeed.
In an actual case one usually tries to guess a

solution, i.e. we typically make a parameterized
Ansatz for V then use the PDE in order to identify
the parameters.
Hint: V often inherits some structural properties

from the boundary function as well as from the
instantaneous utility function F .
Most of the known solved control problems have,

to some extent, been rigged in order to be
analytically solvable.
Tomas Bj
ork, 2013 342
The Linear Quadratic Regulator
"Z #
T
min E {Xt0QXt + u0tRut } dt + XT0 HXT ,
uRk 0
with dynamics
dXt = {AXt + But } dt + CdWt.
We want to control a vehicle in such a way that it stays

close to the origin (the terms x0Qx and x0Hx) while
at the same time keeping the energy u0 Ru small.
Here Xt Rn and ut Rk , and we impose no control

constraints on u.
The matrices Q, R, H, A, B and C are assumed to be

known. We may WLOG assume that Q, R and H are
symmetric, and we assume that R is positive definite
(and thus invertible).
Tomas Bj
ork, 2013 343
Handling the Problem
The HJB equation becomes

V 0 0

(t, x) + inf uR k {x Qx + u Ru + [x V ](t, x) [Ax + Bu]}
t
1P 2V 0
+ 2 i,j x x (t, x) [CC ]i,j = 0,

i j

V (T, x) = x0Hx.
For each fixed choice of (t, x) we now have to solve the static unconstrained
optimization problem to minimize
x0 Qx + u0Ru + [xV ](t, x) [Ax + Bu] .
Tomas Bj
ork, 2013 344
The problem was:
min x0Qx + u0Ru + [xV ](t, x) [Ax + Bu] .

u
Since R > 0 we set the gradient to zero and obtain
2u0R = (xV )B,
which gives us the optimal u as
1 1 0
= R B (x V )0.
u
2
Note: This is our candidate of optimal control law,

but it depends on the unkown function V .
We now make an educated guess about the structure

of V .
Tomas Bj
ork, 2013 345
From the boundary function x0 Hx and the term x0Qx
in the cost function we make the Ansatz
V (t, x) = x0P (t)x + q(t),
where P (t) is a symmetric matrix function, and q(t) is

a scalar function.
With this trial solution we have,
V
(t, x) = x0P x + q,

t
xV (t, x) = 2x0 P,
xxV (t, x) = 2P
= R1B 0 P x.
u
Inserting these expressions into the HJB equation we

get
n o
x0 P + Q P BR1 B 0 P + A0P + P A x
+q + tr[C 0P C] = 0.
Tomas Bj
ork, 2013 346
We thus get the following matrix ODE for P
(
P = P BR1 B 0 P A0P P A Q,
P (T ) = H.
and we can integrate directly for q:

(
q = tr[C 0P C],
q(T ) = 0.
The matrix equation is a Riccati equation. The

equation for q can then be integrated directly.
Final Result for LQ:

Z T
0
V (t, x) = x P (t)x + tr[C 0P (s)C]ds,
t
(t, x) = R1B 0P (t)x.
u
Tomas Bj
ork, 2013 347
2. Investment Theory
Problem formulation.
An extension of HJB.
The simplest consumption-investment problem.
The Merton fund separation results.
Tomas Bj
ork, 2013 348
Recap of Basic Facts
We consider a market with n assets.
Sti = price of asset No i,

hit = units of asset No i in portfolio
wti = portfolio weight on asset No i
Xt = portfolio value
ct = consumption rate
We have the relations

n
X i i n
X
h t St
Xt = hitSti, i
wt = , wti = 1.
i=1
Xt i=1
Basic equation:
Dynamics of self financing portfolio in terms of relative
weights
Xn i
i dSt
dXt = Xt wt i ct dt
i=1
St
Tomas Bj
ork, 2013 349
Simplest model
Assume a scalar risky asset and a constant short rate.
dSt = Stdt + StdWt

dBt = rBt dt
We want to maximize expected utility of consumption

over time
"Z #
T
max E F (t, ct)dt
w 0,w 1,c 0
Dynamics

dXt = Xt wt0r + wt1 dt ctdt + wt1 XtdWt ,
Constraints
ct 0, t 0,
wt0 + wt1 = 1, t 0.
Tomas Bj
ork, 2013 350
Nonsense!
Tomas Bj
ork, 2013 351
What are the problems?
We can obtain unlimited utility by simply consuming

arbitrary large amounts.
The wealth will go negative, but there is nothing in

the problem formulations which prohibits this.
We would like to impose a constratin of type Xt 0

but this is a state constraint and DynP does not
allow this.
Good News:
DynP can be generalized to handle (some) problems
of this kind.
Tomas Bj
ork, 2013 352
Generalized problem
Let D be a nice open subset of [0, T ]Rn and consider

the following problem.
Z
max E F (s, Xsu, us)ds + (, Xu) .
uU 0
Dynamics:
dXt = (t, Xt, ut ) dt + (t, Xt , ut) dWt,

X0 = x0,
The stopping time is defined by
= inf {t 0 |(t, Xt ) D} T.
Tomas Bj
ork, 2013 353
Generalized HJB
Theorem: Given enough regularity the follwing hold.
1. The optimal value function satisfies
V (t, x) + sup {F (t, x, u) + AuV (t, x)} = 0, (t, x) D

t uU
V (t, x) = (t, x), (t, x) D.
2. We have an obvious verification theorem.
Tomas Bj
ork, 2013 354
Reformulated problem
Z
max E F (t, ct)dt + (XT )
c0, wR 0
The ruin time is defined by
= inf {t 0 |Xt = 0} T.
Notation:
w1 = w,
w0 = 1 w
Thus no constraint on w.
Dynamics
dXt = wt [ r] Xtdt + (rXt ct ) dt + wXtdWt ,
Tomas Bj
ork, 2013 355
HJB Equation
( )
2
V V V 1 V
+ sup F (t, c) + wx( r) + (rx c) + x2 w 2 2 = 0,
t c0,wR x x 2 x2
V (T , x) = 0,
V (t, 0) = 0.
We now specialize (why?) to

F (t, c) = et c ,
and for simplicity we assume that
= 0,
so we have to maximize
t V V 1 2 2 2 2V
e c + wx( r) + (rx c) + x w ,
x x 2 x2
Tomas Bj
ork, 2013 356
Analysis of the HJB Equation
In the embedded static problem we maximize, over c

and w,
t 1 2 2 2
e c + wx( r)Vx + (rx c)Vx + x w Vxx,
2
First order conditions:
c1 = et Vx,
Vx r
w = 2
,
x Vxx
Ansatz:
V (t, x) = eth(t)x ,
Because of the boundary conditions, we must demand
that
h(T ) = 0. (5)
Tomas Bj
ork, 2013 357
Given a V of this form we have (using to denote the
time derivative)
et hx ,
Vt = ethx
Vx = ethx1,
Vxx = ( 1)ethx2.
giving us
r
w(t,
b x) = ,
2 (1 )
c(t, x) = xh(t)1/(1) .
b
Plug all this into HJB!
Tomas Bj
ork, 2013 358
After rearrangements we obtain
n o

x h(t) + Ah(t) + Bh(t)/(1) = 0,
where the constants A and B are given by
( r)2 1 ( r)2
A = 2
+ r 2

(1 ) 2 (1 )
B = 1 .
If this equation is to hold for all x and all t, then we

see that h must solve the ODE

h(t) + Ah(t) + Bh(t)/(1) = 0,
h(T ) = 0.
An equation of this kind is known as a Bernoulli

equation, and it can be solved explicitly.
We are done.
Tomas Bj
ork, 2013 359
Mertons Mutal Fund Theorems
1. The case with no risk free asset
We consider n risky assets with dynamics
dSi = Siidt + SiidW, i = 1, . . . , n
where W is Wiener in Rk . On vector form:
dS = D(S)dt + D(S)dW.
where

1 1
= .. = ..
n n
D(S) is the diagonal matrix
D(S) = diag[S1, . . . , Sn].
Tomas Bj
ork, 2013 360
Formal problem
Z
max E F (t, ct)dt
c,w 0
given the dynamics
dX = Xw0dt cdt + Xw0dW.
and constraints
e0w = 1, c 0.
Assumptions:
The vector and the matrix are constant and
deterministic.
The volatility matrix has full rank so 0 is positive

definite and invertible.
Note: S does not turn up in the X-dynamics so V is

of the form
V (t, x, s) = V (t, x)
Tomas Bj
ork, 2013 361
The HJB equation is
c,w

Vt (t, x) + sup {F (t, c) + A V (t, x)} = 0,
0
e w=1, c0
V (T, x) = 0,

V (t, 0) = 0.
where
1
Ac,w V = xw0Vx cVx + x2 w0w Vxx,
2
The matrix is given by
= 0 .
Tomas Bj
ork, 2013 362
The HJB equation is
8
1
ff
0 2 0
V + sup F (t, c) + (xw c)V + x w wVxx = 0,
>
t x
>
>
>
< w 0e=1, c0 2
>
> V (T , x) = 0,
>
>
: V (t, 0) = 0.
where = 0.
If we relax the constraint w0e = 1, the Lagrange function for the static
optimization problem is given by
1
L = F (t, c) + (xw0 c)Vx + x2w0wVxx + (1 w0e) .
2
Tomas Bj
ork, 2013 363
L = F (t, c) + (xw0 c)Vx
1 2 0
+ x w wVxx + (1 w0e) .
2
The first order condition for c is
Fc = Vx.
The first order condition for w is
x0 Vx + x2 Vxxw0 = e0,
so we can solve for w in order to obtain

xVx
= 1 2
w e 2 .
x Vxx x Vxx
Using the relation e0w = 1 this gives as
x2Vxx + xVx e01

= 0 1
,
e e
Tomas Bj
ork, 2013 364
Inserting gives us, after some manipulation,
0 1
1 Vx e
= 0 1 1e +
w 1 0 1 e .
e e xVxx e e
We can write this as
w(t)
= g + Y (t)h,
where the fixed vectors g and h are given by
1 1
g = 0 1
e,
e e
0 1
e
h = 1 0 1 e ,
e e
whereas Y is given by
Vx(t, X(t))
Y (t) = .
X(t)Vxx (t, X(t))
Tomas Bj
ork, 2013 365
We had
w(t)
= g + Y (t)h,
Thus we see that the optimal portfolio is moving
stochastically along the one-dimensional optimal
portfolio line
g + sh,
in the (n 1)-dimensional portfolio hyperplane ,
where
= {w Rn |e0w = 1} .
If we fix two points on the optimal portfolio line, say
wa = g + ah and wb = g + bh, then any point w on
the line can be written as an affine combination of the
basis points wa and wb. An easy calculation shows
that if ws = g + sh then we can write
ws = wa + (1 )wb ,
where
sb
= .
ab
Tomas Bj
ork, 2013 366
Mutual Fund Theorem
There exists a family of mutual funds, given by

ws = g + sh, such that
1. For each fixed s the portfolio ws stays fixed over

time.
2. For fixed a, b with a 6= b the optimal portfolio w(t)

is, obtained by allocating all resources between the
fixed funds wa and wb, i.e.
= a(t)wa + b(t)wb ,
w(t)
Tomas Bj
ork, 2013 367
The case with a risk free asset
Again we consider the standard model
dS = D(S)dt + D(S)dW (t),
We also assume the risk free asset B with dynamics
dB = rBdt.
We denote B = S0 and P consider portfolio weights

n
(w0, w1, . . . , wn)0 where 0 wi = 1. We then
eliminate w0 by the relation
n
X
w0 = 1 wi ,
1
and use the letter w to denote the portfolio weight

vector for the risky assets only. Thus we use the
notation
w = (w1 , . . . , wn)0 ,
Note: w Rn without constraints.
Tomas Bj
ork, 2013 368
HJB
We obtain
dX = X w0( re)dt + (rX c)dt + X w0dW,
where e = (1, 1, . . . , 1)0.
The HJB equation now becomes

c,w
Vt (t, x) + sup {F (t, c) + A V (t, x)} = 0,

c0,wR n
V (T, x) = 0,
V (t, 0) = 0,
where
Ac V = xw0( re)Vx (t, x) + (rx c)Vx (t, x)

1 2 0
+ x w wVxx(t, x).
2
Tomas Bj
ork, 2013 369
First order conditions
We maximize
0 1 2 0
F (t, c) + xw ( re)Vx + (rx c)Vx + x w wVxx
2
with c 0 and w Rn.
The first order conditions are
Fc = Vx,
Vx 1
=
w ( re),
xVxx
with geometrically obvious economic interpretation.
Tomas Bj
ork, 2013 370
Mutual Fund Separation Theorem
1. The optimal portfolio consists of an allocation

between two fixed mutual funds w0 and wf .
2. The fund w0 consists only of the risk free asset.
3. The fund wf consists only of the risky assets, and

is given by
wf = 1( re).
Tomas Bj
ork, 2013 371
The Martingale Approach to Optimal

Investment Theory
Ch 20
Tomas Bj
ork
Tomas Bj
ork, 2013 372
Contents
Decoupling the wealth profile from the portfolio

choice.
Lagrange relaxation.
Solving the general wealth problem.
Example: Log utility.
Example: The numeraire portfolio.
Tomas Bj
ork, 2013 373
Problem Formulation
Standard model with internal filtration
dSt = D(St )t dt + D(St )t dWt,

dBt = rBt dt.
Assumptions:
Drift and diffusion terms are allowed to be arbitrary
adapted processes.
The market is complete.
We have a given initial wealth x0
Problem:
max E P [(XT )]
hH
where
H = {self financing portfolios}
given the initial wealth X0 = x0.
Tomas Bj
ork, 2013 374
Some observations
In a complete market, there is a unique martingale

measure Q.
Every claim Z satisfying the budget constraint
erT E Q [Z] = x0 ,
is attainable by an h H and vice versa.
We can thus write our problem as
max E P [(Z)]
Z
subject to the constraint
erT E Q [Z] = x0 .
We can forget the wealth dynamics!
Tomas Bj
ork, 2013 375
Basic Ideas
Our problem was
max E P [(Z)]
Z
subject to
erT E Q [Z] = x0 .
Idea I:
We can decouple the optimal portfolio problem into:

1. Finding the optimal wealth profile Z.
find the replicating portfolio.

2. Given Z,
Idea II:
Rewrite the constraint under the measure P .
Use Lagrangian techniques to relax the constraint.
Tomas Bj
ork, 2013 376
Lagrange formulation
Problem:
max E P [(Z)]
Z
subject to
erT E P [LT Z] = x0 .
Here L is the likelihood process, i.e.
dQ
Lt = , on Ft, 0tT
dP
The Lagrangian of the problem is
P
rT P

L = E [(Z)] + x0 e E [LT Z]
i.e.
P rT
L=E (Z) e LT Z + x0
Tomas Bj
ork, 2013 377
The optimal wealth profile
Given enough convexity and regularity we now expect,

given the dual variable , to find the optimal Z by
maximizing
P
rT

L=E (Z) e LT Z + x0
over unconstrained Z, i.e. to maximize

Z
rT

(Z()) e LT ()Z() dP ()

This is a trivial problem!
We can simply maximize Z() for each separately.
rT

max (z) e LT z
z
Tomas Bj
ork, 2013 378
The optimal wealth profile
Our problem:
rT

max (z) e LT z
z
First order condition
0(z) = erT LT
The optimal Z is thus given by

Z = G erT LT
where
1
G(y) = [0] (y).
The dual varaiable is determined by the constraint

h i
rT
e E P
LT Z = x0.
Tomas Bj
ork, 2013 379
Example log utility
Assume that
(x) = ln(x)
Then
1
g(y) =
y
Thus
1 rT 1

Z = G erT
LT = e LT

Finally is determined by
h i
erT E P LT Z = x0.
i.e.
1
erT E P LT erT L1
T = x0 .

so = x1
0 and
Z = x0erT L1
T
Tomas Bj
ork, 2013 380
The optimal wealth process
We have computed the optimal terminal wealth

profile
bT = x0erT L1
Zb = X T
bt look like?
What does the optimal wealth process X
We have (why?)
h i
bt = er(T t)E Q
X bT Ft
X
so we obtain
rt Q
1
b
Xt = x0e E LT Ft
But L1 is a Q-martingale (why?) so we obtain
bt = x0 ertL1
X t .
Tomas Bj
ork, 2013 381
The Optimal Portfolio
We have computed the optimal wealth process.
How do we compute the optimal portfolio?
Assume for simplicity that we have a standard Black-

Scholes model
dSt = Stdt + StdWt,

dBt = rBtdt
Recall that
bt = x0 ertL1
X t .
Tomas Bj
ork, 2013 382
Basic Program
bt to compute dX
1. Use Ito and the formula for X bt like
bt = X
dX bt ( )dt + X
bttdWt
where we do not care about ( ).
2. Recall that

bt = X
bt dBt dSt
dX (1 u
t ) +u
t
Bt St
which we write as
bt = X
dX bt { } dt + X
bt u
tdWt
3. We can identify u
as
t
u
t =

Tomas Bj
ork, 2013 383
We recall
bt = x0 ertL1
X t .
We also recall that
dLt = LtdWt ,
where
r
=

From this we have
dL1
t = 2 1
L t dt L 1
t dWt
and we obtain
bt = X
X bt { } dt X
btdWt
Result: The optimal portfolio is given by

r
u
t =
2
Note that u is a myopic portfolio in the sense that

it does not depend on the time horizon T .
Tomas Bj
ork, 2013 384
A Digression: The Numeraire Portfolio
Standard approach:
Choose a fixed numeraire (portfolio) N .
Find the corresponding martingale measure, i.e. find QN s.t.
B S
, and
N N
are QN -martingales.
Alternative approach:
Choose a fixed measure Q P .
Find numeraire N such that Q = QN .
Special case:
Set Q = P
Find numeraire N such that QN = P i.e. such that
B S
, and
N N
are QN -martingales under the objective measure P .

This N is called the numeraire portfolio.
Tomas Bj
ork, 2013 385
Log utility and the numeraire portfolio
Definition:
The growth optimal portfolio (GOP) is the portfolio
which is optimal for log utility (for arbitrary terminal
date T .
Theorem:
Assume that X is GOP. Then X is the numeraire
portfolio.
Proof:
We have to show that the process
St
Yt =
Xt
is a P martingale.
We have
St
= x1
0 e
rt
St L t
Xt
which is a P martingale, since x1

0 e
rt
St is a Q
martingale.
Tomas Bj
ork, 2013 386

Continuous Time Finance: Tomas BJ Ork Stockholm School of Economics

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Continuous Time Finance: Tomas BJ Ork Stockholm School of Economics

Încărcat de

Drepturi de autor:

Formate disponibile

Continuous Time Finance

Tomas Bjork, 2013

Stochastic Calculus (Ch 4-5).

Take as given the market price process, S, of some

St = price, at t, per unit of underlying asset

Consider a fixed financial derivative, e.g. a European

Main Problem: Find the arbitrage free price of the

1. Mathematical model for the underlying price

2. Mathematical techniques to handle the price

We model the stock price St as a stochastic

We model S as a Markov process, i.e. in order

Xt = any random process,

We often write X(t) instead of Xt .

dXt is called the increment of X over the interval

For any fixed interval [t, t + dt], the increment dXt

If the increments dXs and dXt, over the disjoint

If every increment has a normal distribution we say

A stochastic process W is called a Wiener process if

The increments are normally distributed: For s < t:

E[Wt Ws] = 0, V ar[Wt Ws] = t s

W has independent increments.

W has continuous trajectories.

Continuous random walk

Note: In Hull, a Wiener process is typically denoted

A stochastic process X is called a Wiener process

where and are constants.

Summing all increments over the interval [0, t] gives us

We say, losely speaking, that the process X is an It

dXt = tdt + t dWt,

where t and t are random processes.

Informally you can think of dWt as a random variable

dWt N [0, dt]

To handle expressions like the one above, we need

First, however, we present an important example,

Price dynamics: (Geometrical Brownian Motion)

dSt = Stdt + StdWt,

Conjecture: The solution of the SDE above is a

Return = (mean return)

The asset return is a random walk (with drift).

= mean rate of return per unit time

Large = large random fluctuations

Small = small random fluctuations

The returns are normal.

The stock price is lognormal.

Consider an expression of the form

dXt = tdt + t dWt,

Question: What exactly do we mean by this?

Answer: Write the equation on integrated form as

How is this interpreted?

Stochastic integral. This can not be interpreted

Consider a Wiener process W .

Def: Let Z be a stochastic variable. If the value of Z

Let W be a Wiener process.

An adapted process does not look

Adapted processes are nice integrands for stochastic

We will define the It

for processes g satisfying

The process g is adapted.

The process g satisfies

This will be done in two steps.

There exists deterministic points t0 . . . , tn with

g(s) = g(tk ), s [tk , tk+1)

For simple g we define