Documente Academic
Documente Profesional
Documente Cultură
Jonathan Ziveyi1
1 UNSW
Australia
Risk and Actuarial Studies, UNSW Business School
j.ziveyi@unsw.edu.au
1/77 Version 2017. Copyright UNSW School of Risk and Actuarial Studies
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
2/77 Hedging
Continuous Time Derivative Valuation
Learning Outcome
Apply contingent-claim pricing techniques to value and manage
the risks of embedded options and guarantees.
3/77
What to Expect in this Module
4/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
5/77 Hedging
Change of Measure - Binomial model
We firstly stay in a (two step) binomial tree setting, and
explore the relationship between P and Q in a bit more
detail.
With each possible path we can attach a P probability :
6/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
7/77 Hedging
Equivalent Probability Measures
7/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
8/77 Hedging
dQ
Expectations and dP
So,
X
EQ [X ] = i xi
i
X
i
= i xi
i
i
dQ
= EP X
dP
8/77
Disussion Question
h i
dQ
What is EP dP =?
9/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
10/77 Hedging
Radon-Nikodym as a Process
10/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
11/77 Hedging
Radon-Nikodym Summary
We also have
for 0 t T ,
EQ [X |Ft ] = 1 (t) EP [ (T ) X |Ft ]
where
dQ
(t) = EP |Ft .
dP
11/77
Foundations for Continuous Time Models - Brownian
Motion
12/77
Brownian Motion cont...
Let Xi be a series of IID random variables that takes value
1
We have, for i = 1, 2, .., nt:
i i 1 X
Wn = Wn + i
n n n
and in particular
Xnt
X
Wn (t) = 0 + i
i=1
n
nt
!
X X
= t i
i=1
nt
Hence
Wn (t) N (0, t)
n
14/77
Brownian Motion cont...
Definition
Definition: A stochastic process {W (t) , t 0} is said to be a
P-Brownian motion process if and only if:
1. W (t) is continuous and W (0) = 0;
2. The value of W (t) is distributed, under P, as N (0, t)
3. The increment W (s + t) W (s) is distributed, under P, as
N (0, t), and is independent of Fs
15/77
Brownian Motion cont...
16/77
Application for Stock Prices
Brownian motion on its own is not a good choice for modelling stock
prices.
It is missing (amongst others) two important properties that stock
prices tend to have:
Stock Prices tend to increase in the real world (P)
Stock Prices do not become negative
S (t) = eW (t)+t
17/77
Methods of Specifying a Stochastic Process
18/77
Stochastic Processes
Definition:
In the Brownian motion framework a stochastic process
X (t) is a continuous process (on t) such that
Z t Z t
X (t) = X (0) + (s) dW (s) + (s) ds
0 0
2
19/77 e.g., are known at time t, and are bounded
Stochastic Processes cont...
20/77
Discussion
Let X (0) = x
1. What is the solution to
dX (t) = dW (t) + dt
21/77
Stochastic Calculus
dH (x) = H (x) dx
F (W (t)) = eW (t)
G (W (t) , t) = eW (t)+t ?
22/77
Key observation
Denote Zi as IID N(0, 1) random variables. Observe that
Z t X n 2
2 ti t(i 1)
(dW (u)) W W
0 n n
i=1
n r !2
X t
2
= (Zi )
n
i=1
n
!
X Z2
i
= t
n
i=1
t
as n .
Hence we have
Z t Z t
(dW (u))2 t = du
0 0
or in other words, a key result we often use in stochastic calculus
is
(d(W (t)))2 dt
23/77
Discussion: Itos lemma - Differentiation Rule with
Stochastic Processes
Given
(dW (t))2 = dt,
discuss why
1
dF (W (t)) = F (W (t)) dW (t) + F (W (t)) dt
2
(Hint: Use Taylor series expansion)
This is known as Itos lemma.
24/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
25/77 Hedging
Itos Lemma with more General Processes
25/77
Itos Lemma cont...
On substitution, we have
26/77
Example: Finding the SDE Satisfied by a Process
(with closed form solution)
Suppose, for S(0) = 1, we have
S (t) = eW (t)+t
We can use Ito to help us find dS(t). Let
X (t) = W (t) + t
then we have
S (t) = eX (t) = f (X (t)) .
Hence
f (X (t)) = f (X (t))
= f (x (t))
= eX (t)
= S (t)
27/77
Example cont...
Simplifying yields
f 1 2f
dS (t) = dX (t) + (dX (t))2
x 2 x 2
1 2
= S(t) dW (t) + + dt
2
28/77
Finding a Closed Form Solution of a Process from its
SDE
29/77
Example: Geometric Brownian Motion (GBM)
with S(0) = s?
30/77
Summary of Pricing Steps - Binomial Tree Model
S(t)
1. Identify a measure Q such that Z (), with Z (t) = B(t) , is a
Q martingale.
2. Observe that both Z () and Y (), with
X
Y (t) = EQ |F(t)
B(T )
are Q martingales.
By the martingale representation theorem, we know that
there is a previsible process () such that
t
X
Y (t) = Y (0) + (k ) (Z (k ) Z (k 1))
k =1
31/77
Summary of Pricing Steps cont...
32/77
Reflection
33/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
34/77 Hedging
Martingales in continuous time
3
34/77 subject to technical conditions e.g. the expectations are well defined
Examples of Martingales
35/77
Examples cont...
7 The process M() with
Z t
M(t) = (s)dW (s)
0
36/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
37/77 Hedging
Change of Measure - the Continuous Radon-Nikodym
Derivative
dQ
dP
This can be interpreted as the ratio of the probabilities
between Q and P on each possible path of outcomes.
For any event A, we have the following relationship
dQ
Q(A) = EP 1A
dP
37/77
Change of Measure - cont...
In continuous time, heuristically we can also consider the
(continuous) Radon-Nikodym derivative as the ratios of the
probabilities.
Heuristically, for a path , consider discrete time points
{t1 , t2 , .., tn 1, tn }, with tn = T , and a P Brownian motion
with corresponding path {x1 , .., xn }, the density (on these
time points) are represented by
We can consider
dQ f n (x1 , .., xn )
() = limn Qn
dP fP (x1 , .., xn )
where is a constant.
We can define an arbitrary measure Q (note: this is a
generic result at this stage, and not necessary the Q we
want for financial modelling) by
Q [A] = EP [Y (T ) 1A ]
39/77
Discussion - cont...
Recall that the moment generating function of a normally
distributed (under say measure P) random variable X with
mean and variance 2 can be represented by
h i 1 2 2
EP eX = e+ 2
40/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
41/77 Hedging
Change of Measure: Cameron-Martin-Girsanov
Theorem
is a Q Brownian motion.
4
subject to technical conditions - e.g. if () was bounded, or even
41/77 constant
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
42/77 Hedging
Discussion Example
X (t) = W (t) + t
dQ Rt Rt 2
= e 0
dW (s) 21 0 ( ) ds
?
dP
42/77
Discussion Example
and
dX (t) = X (t) (dWQ (t) + dt) .
43/77
Fundamental Theorem of Asset Pricing
44/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
45/77 Hedging
Martingale Representation Theorem: Brownian Motion
Based Model
5
45/77 technically - pre-visible, square integrable, (essentially) unique
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
46/77 Hedging
Replicating Portfolios - Trading Strategies (1)
.
What properties do we want them to have?
46/77
Replicating Portfolios - cont...
(1) Previsible
( () , ()) should be previsible - to remove insider
trading.
Note that previsibility in continuous time means that (t) is
known at time t - this is different in notation to the discrete
time case.
47/77
Replicating Portfolios - cont...
48/77
Replicating Portfolios - cont...
6
49/77 subject to technical conditions
Example
50/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
51/77 Hedging
Pricing Steps: Continuous Time Model
S(t)
1. Identify a measure Q such that Z (), with Z (t) = B(t) , is a
Q martingale.
2. Observe that both Z () and Y (), with
h i
Y (t) = EQ B(T )1 X |Ft
are Q martingales.
By the martingale representation theorem, we know that
there is a pre-visible process () such that
Z t
Y (t) = Y (0) + (s)dZ (s)
0
51/77
Pricing Steps cont...
52/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
53/77 Hedging
Black-Scholes Model
dB (t) = rB (t) dt
dS (t) = S (t) dt + S (t) dW (t)
53/77
Step 1 - Finding the Q Measure and Dynamics
7
54/77 as usual, subject to technical conditions
Revision Example
55/77
Example cont...
By the CMG theorem, heuristically, we known that under
an equivalent measure Q we have a Q Brownian motion
WQ (), where
hence
r
dZ (t) = Z (t) dWQ (t) + (t) dt .
But to be a Q martingale, we need
57/77
Step 2 - Martingale Representation
Form
1
Y (t) = EQ X |Ft
B (T )
We have 2 Q martingales Z () and Y (), so we know there
exists (t) such that
58/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
59/77 Hedging
Steps 3,4 - Self Financing, Replicating strategies
59/77
Steps 3,4 cont...
60/77
Steps 3,4 cont...
Hence
61/77
Steps 3,4 cont...
On maturity T we have
V (T ) = Y (T ) B (T )
1
= EQ X |FT B (T )
B (T )
= X
62/77
Step 5
V (t) = Y (t)B(t)
1
= EQ X |Ft B (t)
B (T )
h i
= EQ er (T t) X |Ft
63/77
Derivative Valuation in the Black Scholes model
64/77
Distribution of S(T)
65/77
Discussion Example: Black Scholes Formula (1)
Find:
KerT EQ 1S(T )>K =?
66/77
Black-Scholes Formula (2)
To calculate erT EQ S (T ) 1S(T )>K we can complete the
square. Denoting
1 2 2
H = ln (S (T ) /S (0)) N rT T , T
2
we have:
h i Z
EQ S (T ) 1S(T )>K = (S (0) exp {z})
z=ln K
S(0)
2
z rT 21 2 T
1 1
exp dz
2 T 2 2
T
1
Z
rT
= S (0) e
z=ln K 2 T 2
S(0)
2
z rT + 12 2 T
1
exp dz
2 T
ln (S (0) /K ) + r + 12 2 (T )
rT
= S (0) e
T
67/77
Black-Scholes Formula
68/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
69/77 Hedging
Price Dependence
69/77
Discussion: Valuation of other options
There are many other options that
investors/speculators/hedgers (etc) are interested in. Some
examples include
Barrier Options (Knock in)
Barrier Options (Knock out)
Lookback Options
Asian Options
Chooser Options
How would one value the above options under the
Black-Scholes model?
70/77
Plan
Topic: Continuous Time Derivative Valuation
Measure Theory
Radon-Nikodym Derivative
Equivalent Measures
Expectations and the Radon-Nikodym Derivative
Radon-Nikodym as a Process
Radon-Nikodym Summary
Stochastic Processes
Stochastic Calculus
Itos Lemma for processes
Martingales
The Continuous Radon-Nikodym Derivative
Change of Measure Cameron-Martin-Girsanov Theorem
Change of Measure and Processes
Martingale Representation Theorem: Brownian Motion Based Model
Replicating Portfolios - Construction Strategies
Pricing Steps: Continuous Time Model
Black-Scholes Model
Arbitrage Pricing
Price Dependence
71/77 Hedging
Risk Management and Hedging of Options
71/77
Representation of (t)
V (s, t) = er (T t) EQ [f (S (T )) |S (t) = s]
V (s, t)
(t) =
s
ie just the mathematical derivative of the price with respect
to the stock price.
72/77
Extension: Models with Dividends
So far we have assumed that no dividends are payable on
the stocks.
How does dividends affect the option prices?
So need to account for the fact that if you hold 1 unit of the
stock from time 0 to time t, you will also be getting the
dividend that are paid during that time.
S (t) dt
75/77
Numerical Techniques: Simulation
76/77
Numerical Techniques for American Options
Remember that an American option can be thought of as
its European counterpart, but with the right to exercise it
immediately.