CHAPTER 2

Concepts of Financial Economics and Asset

Price Dynamics

In the last chapter, we observe how the application of the no arbitrage argu-
ment enforces the forward price of a forward contract. The forward price is
not given by the expectation of the asset price at maturity of the forward,
that is, it is independent on the asset price dynamics over the life of the
contract. The no arbitrage argument turns out to be the basis of the pricing
models for various types of derivatives considered in this text. We also ob-
serve that a call can be replicated by a put and a forward [see Eq. (1.3.2)].
Indeed, it will be shown in Sec. 3.1 that an option can be replicated dynam-
ically by a portfolio containing the underlying asset and the riskless bond.
Assuming frictionless market and no premature termination of the option
contract, suppose the option’s payoff matches with that of the replicating
portfolio at maturity, one can show by no arbitrage argument that the value
of the option is equal to the value of the replicating portfolio at all times
throughout the life of the option. If every derivative can be replicated by a
portfolio of the fundamental assets in the market, then the market is said
to be complete. In other words, we price a derivative based on the prices of
other marketed assets that replicate the derivative.
From the theory of financial economics, we show that the condition of no
arbitrage is equivalent to the existence of an equivalent martingale measure.
Under the equivalent martingale measure, all discounted price processes of
the risky assets are martingales. Further, if the market is complete (all con-
tingent claims can be replicated), then the equivalent martingale measure
is unique. The above statements are the essence of the Fundamental The-
orem of Asset Pricing. It can be shown that the replication based price of
any contingent claim can be obtained by calculating the discounted expected
value of its terminal payoff under the equivalent martingale probability mea-
sure (Harrison and Kreps, 1979). This approach has come to be known as
the risk neutral pricing. The term risk neutrality is used since all assets in
the market offer the same return as the riskfree security under this proba-
bility, so an investor who is neutral to risk and faces with this probability
would be indifferent among various assets. The concepts of replicable contin-
gent claims, absence of arbitrage and risk neutrality form the cornerstones of
modern option pricing theory.
38 2 Concepts of Financial Economics and Asset Price Dynamics

In the first two sections, we limit our discussion of securities model to the
discrete framework. We start with the single period securities models in Sec.
2.1. The notions of the law of one price, non-dominant trading strategy, linear
pricing measure and absence of arbitrage are discussed. Every attempt has
been made to have the financial economics concepts self contained. The use
of the Separating Hyperplane Theorem leads to the identification of the risk
neutral measure for the valuation of contingent claims under the assumption
of no arbitrage. In Sec. 2.2, the discussion is extended to multi-period secu-
rities models. The readers will be shown how to construct the information
structures of securities models. Various definitions in probability theory will
be presented, like filtrations, measurable random variables, conditional ex-
pectations and martingales. In multi-period situation, the risk neutral prob-
ability measure is defined in terms of martingales. The highlight of the first
two sections is the derivation of the Fundamental Theorem of Asset Pricing.
More detailed exposition on the related concepts of financial economics can
be found in the books by Pliska (1997) and LeRoy and Werner (2001).
In general, the price of a derivative depends primarily on the stochastic
process of the price of the underlying asset. Most asset price processes are
modeled by the Ito processes. For equity prices, they are fairly described by
the Geometric Brownian processes, a popular class of Ito processes. In Sec.
2.3, we provide a brief exposition on the Brownian process. We start with the
discrete random walk model and treat the Brownian process as the continu-
ous limit of the random walk process. The forward Fokker-Planck equation
that governs the transition density function for Brownian processes is also
developed. In the last section, we introduce some basic tools in stochastic
calculus, in particular, the notion of stochastic integrals and stochastic dif-
ferentials. We explain the non-differentiability of Brownian paths. We provide
an intuitive proof of the Ito lemma, which is an essential tool in performing
calculus operations on functions of stochastic state variables. We also discuss
the Feynman-Kac representation, Radon-Nikodym derivative and the Gir-
sanov Theorem. The Girsanov Theorem provides an effective tool to trans-
form Ito processes with general drifts into martingales. All these preliminaries
in stochastic calculus are essential to develop the option pricing theory and
to derive option price formulas in later chapters.

2.1 Single period securities models

The no arbitrage approach is one of the cornerstones in the development of

pricing theory of financial derivatives. In simple language, arbitrage refers
to the possibility of making an investment gain with no chance of loss (the
rigorous definition of arbitrage will be given later). In the theoretical develop-
ment of pricing models, it is commonly assumed that there are no arbitrage
opportunities in well functioning and competitive financial markets.
 2.1 Single period securities models 39

In this section, we discuss the concepts of no arbitrage principle via single

period securities models, where investment decisions on a finite set of M
securities are made at initial time t = 0 and the payoff is attained at terminal
time t = 1. Though single period models appear to be not quite realistic
representation of the complex world of investment activities, however, a lot
of fundamental concepts in financial economics can be revealed from the
analysis of single period securities models. Also, single period investment
models approximate quite well the buy-and-hold investment strategies.

2.1.1 Law of one price, dominant trading strategies and linear

pricing measures

In single period securities models, the initial prices of M risky securities,

denoted by S1 (0), · · · , SM (0), are positive scalars that are known at t = 0.
However, their values at t = 1 are random variables. These random variables
are defined with respect to a sample space Ω = {ω1, ω2, · · · , ωK } of K possible
outcomes (or states of world). At t = 0, the investors know the list of all
possible outcomes, but which outcome does occur is revealed only at the end
of the investment period t = 1. Further, a probability measure P satisfying
P (ω) > 0, for all ω ∈ Ω, is defined on Ω.
We use S to denote the price process {S(t) : t = 0, 1}, where S(t) is the
row vector S(t) = (S1 (t) S2 (t) · · · SM (t)). The possible values of the asset
price process at t = 1 are listed in the following K × M matrix
 
S1 (1; ω1) S2 (1; ω1) · · · SM (1; ω1)
 S (1; ω2) S2 (1; ω2) · · · SM (1; ω2) 
S(1; Ω) =  1 . (2.1.1)
··· ··· ··· ···
S1 (1; ωK ) S2 (1; ωK ) · · · SM (1; ωK )
Since the assets are limited liability securities, the entries in S(1; Ω) are non-
negative scalars.
We also assume the existence of a strictly positive riskless security or bank
account, whose value is denoted by S0 . Without loss of generality, we take
S0 (0) = 1 and the value at time 1 to be S0 (1) = 1 + r, where r ≥ 0 is the
deterministic interest rate over one period. The reciprocal of S0 (1) is called
the discount factor over the period. We define the discounted price process
by
S ∗ (t) = S(t)/S0 (t), t = 0, 1, (2.1.2a)
that is, we use the riskless security as the numeraire or accounting unit.
Accordingly, the payoff matrix of the discounted price processes of the M
risky assets and the riskless security can be expressed in the form
 
1 S1∗ (1; ω1) · · · SM ∗
(1; ω1)
∗ ∗
 1 S1 (1; ω2) · · · SM (1; ω2) 
Sb∗ (1; Ω) =  . (2.1.2b)
··· ··· ··· ···
1 S1∗ (1; ωK ) · · · SM
∗
(1; ωK )
40 2 Concepts of Financial Economics and Asset Price Dynamics

The first column in Sb∗ (1; Ω) (all entries are equal to one) represents the
discounted payoff of the riskless secuirty at all states of world. Also, we define
the vector of discounted price processes associated with the riskless security
and the M risky securities by
Sb∗ (t) = (1 S1∗ (t) · · · SM
∗
(t)), t = 0, 1. (2.1.2c)
An investor adopts a trading strategy by selecting a portfolio of the assets
at time 0. The number of units of asset m held in the portfolio from t = 0
to t = 1 is denoted by hm , m = 0, 1, · · ·, M . The scalars hm can be positive
(long holding), negative (short selling) or zero (no holding).
Let V = {Vt : t = 0, 1} denote the value process that represents the total
value of the portfolio over time. It is seen that
M
X
Vt = h0S0 (t) + hm Sm (t), t = 0, 1. (2.1.3)
m=1

The gain due to the investment on the mth risky security is given by
hm [Sm (1) − Sm (0)] = hm ∆Sm , m = 1, · · ·, M . Let G be the random vari-
able that denotes the total gain generated by investing in the portfolio. We
then have
XM
G = h0 r + hm ∆Sm . (2.1.4)
m=1
If there is no withdrawal or addition of funds within the investment horizon,
then
V1 = V0 + G. (2.1.5)
Suppose we use the bank account as the numeraire, and define the discounted
value process by Vt∗ = Vt /S0 (t) and discounted gain by G∗ = V1∗ − V0∗ , we
then have
M
X
Vt∗ = h0 + ∗
hm S m (t), t = 0, 1; (2.1.6a)
m=1
M
X
G∗ = V1∗ − V0∗ = ∗
hm ∆Sm . (2.1.6b)
m=1

Dominant trading strategies

Let H denote the trading strategy that involves the choice of the number of
units of assets held in the portfolio. A trading strategy is said to be dominant
if there exists another trading strategy H b such that

V0 = Vb0 and V1 (ω) > Vb1 (ω) for all ω ∈ Ω. (2.1.7)

Here, Vb0 and Vb1 denote the portfolio value of H

b at t = 0 and t = 1, respec-
tively. Financially speaking, both strategies H and H b start with the same
 2.1 Single period securities models 41

initial investment amount but the dominant strategy H leads to a higher

gain with certainty.
Suppose H dominates H, b we define a new trading strategy H e = H − H. b
e e e
Let V0 and V1 denote the portfolio value of H at t = 0 and t = 1, respectively.
From Eq. (2.1.7), we then have Ve0 = 0 and Ve1 (ω) > 0 for all ω ∈ Ω. This
trading strategy is dominant since it dominates the strategy which starts
with zero value and does no investment at all. A securities model that allows
the existence of a dominant trading strategy is not realistic since an investor
starting with no money should not be guaranteed of ending up with positive
returns by adopting a particular trading strategy. Equivalently, one can show
that a dominant trading strategy is one that can transform a strictly negative
wealth at t = 0 into a non-negative wealth at t = 1 (see Problem 2.1). Later,
we show how the non-existence of dominant strategies is closely related to
the existence of a linear pricing measure.
The two important concepts in the analysis of securities pricing are lin-
earity and positivity. In simple words, linearity means if two portfolios A and
B have payoff vectors as represented by pA and pB , and their portfolio values
at the current time are VA and VB , respectively, then the current value of the
portfolio with payoff vector αpA + βpB will be given by αVA + βVB , where
α and β are scalar constants. Linearity of pricing is related to the law of one
price [see Eq. (2.1.8)]. Positivity of pricing refers to the positivity of state
prices, and this relates to the concepts of linearly pricing measure [see Eq.
(2.1.9)]. In most of the subsequent expositions, we use the riskless security as
the numeraire and consider discounted value processes of the risky securities.
With this choice of numeraire, the linear pricing measure can be interpreted
as a probability measure [see Eqs. (2.1.10a,b)].
Asset span, law of one price and state prices
Consider the following numerical example, where the number of possible
states is taken to be 3. First, we consider
  two risky securities
  whose dis-
1 3
∗   ∗  
counted payoff vectors are S1 (1) =  2  and S2(1) =  1 . The payoff
3 2
 
1 3
 
vectors are used to form the payoff matrix S ∗ (1) =  2 1 . Let the cur-
3 2
rent discounted prices be represented by the row vector S∗ (0) = (1 2). We
write h as the column vector whose entries are the weights of the securities in
the portfolio. The current portfolio value and the discounted portfolio pay-
off are given by S∗(0)h and S ∗ (1)h, respectively. As S0∗ (0) = 1, the current
portfolio value and discounted portfolio value are the same.
The set of all portfolio payoffs via different holding of securities is called
the asset span S. The asset span is seen to be the column space of the payoff
42 2 Concepts of Financial Economics and Asset Price Dynamics

matrix S∗ (1).

 In this example,
 the asset span consists of all vectors of the
1 3
   
form h1  2  + h2  1 , where h1 and h2 are scalars.
3 2
To these two securities in the portfolio, we may add a third security
or even more securities. The newly added securities may or may not fall
within the asset span. If the added security lies inside S, then its payoff can
be expressed as a linear combination of S∗1 (1) and S∗2 (1). In this case, it is
said to be a redundant security. Since there are only 3 possible states, the
dimension of the asset span cannot be more than 3, that is, the maximal
number of non-redundant securitiesis 3. Suppose we add the third security
1
 
whose discounted payoff is S∗3(1) =  3 , it can be easily checked that it is
4
a non-redundant security. The new asset span [the subspace in R3 spanned
by S∗1(1), S∗2(1) and S∗3 (1)] will be the whole R3 . Any further security added
must be redundant since its discounted payoff vector must lie inside the new
asset span. A securities model is said to be complete if every payoff vector
lies inside the asset span. This occurs if and only if the dimension of the asset
span equals the number of possible states.
The law of one price states that all portfolios with the same payoff have
the same price. Consider two portfolios with different portfolio weights h and
h0. Suppose these two portfolios have the same discounted payoff, that is,
S ∗ (1)h = S ∗ (1)h0, then the law of one price infers that S∗(0)h = S∗ (0)h0. It
is quite straightforward to show that a necessary and sufficient condition for
the law of one price to hold is that a portfolio with zero payoff must have zero
price. Also, if the law of one price fails, then it is possible to have two trading
strategies h and h0 such that S ∗ (1)h = S ∗ (1)h0 but S∗ (0)h > S∗ (0)h0. Let
G∗ (ω) and G∗0 (ω) denote the respective discounted gain corresponding to
the trading strategies h and h0. We then have G∗0(ω) > G∗ (ω) for all ω ∈
Ω, so there exists a dominant trading strategy. Hence, the non-existence of
dominant trading strategy implies the law of one price. However, the converse
statement does not hold (see Problem 2.4).
Given a discounted portfolio payoff x that lies inside the asset span, the
payoff can be generated by some linear combination of the securities in the
securities model. We have x = S ∗ (1)h for some h ∈ RM . The current value
of the portfolio is S∗(0)h, where S∗ (0) is the price vector. We may consider
S∗ (0)h as a pricing functional F (x) on the payoff x. If the law of one price
holds, then the pricing functional is single-valued. Furthermore, it is a linear
functional, that is,

F (α1x1 + α2x2 ) = α1F (x1 ) + α2F (x2) (2.1.8)

for any scalars α1 and α2 and payoffs x1 and x2 (see Problem 2.5).
 2.1 Single period securities models 43

Let ek denote the kth coordinate vector in the vector space RK , where ek
assumes the value 1 in the kth entry and zero in all other entries. The vector
ek can be considered as the discounted payoff vector of a security, and it is
called the Arrow security of state k. Suppose the securities model is complete
and the law of one price holds, then the pricing functional F assigns unique
value to each Arrow security. We write sk = F (ek ), which is called the state
price of state k (see Problem 2.6).
Linear pricing measure
We consider securities models with the inclusion of the riskfree security. A
non-negative row vector q = (q(ω1) · · · q(ωK )) is said to be a linear pricing
measure if for every trading strategy we have
K
X
V0∗ = q(ωk )V1∗ (ωk ). (2.1.9)
k=1

The linear pricing measure exhibits the following properties. First, suppose
we take the holding amount of each risky security to be zero, thereby h1 =
h2 = · · · = hM = 0, then

X
K
V0∗ = h0 = q(ωk )h0 (2.1.10a)
k=1

so that
X
K
q(ωk ) = 1. (2.1.10b)
k=1

Next, by taking the portfolio weights to be zero except for the mth security,
we have
K
X
∗ ∗
Sm (0) = q(ωk )Sm (1; ωk), m = 1, · · · , M. (2.1.11)
k=1

Since we have taken q(ωk ) ≥ 0, k = 1, · · ·, K, and their sum is one, we

may interpret q(ωk ) as a probability measure on the sample space Ω. Note
that q(ωk ) is not related to the actual probability of occurrence of the state
k, though the current discounted security price is given by the expectation
of the discounted security payoff one period later under the linear pricing
measure [see Eq. (2.1.11)]. In matrix, form, Eq. (2.1.11) can be expressed as
b ∗ (0) = qSb∗ (1; Ω),
S q ≥ 0. (2.1.12)

As a numerical example, we consider a securities model with 2 risky se-

curities and the riskfree security, and there are 3 possible states. The current
discounted price vector S b ∗ (0) is (1 4 2) and the discounted payoff matrix
44 2 Concepts of Financial Economics and Asset Price Dynamics
 
1 4 3
 
at t = 1 is Sb∗ (1) =  1 3 2 . Here, the law of one price holds since the
1 2 4
only solution to S (1)h = 0 is h = 0. This is because the columns of Sb∗ (1)
b∗

are independent so that the dimension of the nullspace of Sb∗ (1) is zero. We
would like to see whether linear pricing measure exists for the given securities
model. By virtue of Eqs. (2.1.10b) and (2.1.11), the linear pricing probabili-
ties q(ω1 ), q(ω2) and q(ω3), if exist, should satisfy the following equations:
1 = q(ω1 ) + q(ω2) + q(ω3 )
4 = 4q(ω1 ) + 3q(ω2 ) + 2q(ω3)
2 = 3q(ω1 ) + 2q(ω2 ) + 4q(ω3). (2.1.13a)
Solving the above equations, we obtain q(ω1 ) = q(ω2 ) = 2/3 and q(ω3 ) =
−1/3. Since not all the pricing probabilities are non-negative, the linear pric-
ing measure does not exist for this securities model.
Do dominant trading strategies exist for the above securities model? That
is, can we find trading strategy (h1 h2 ) such that V0∗ = 4h1 + 2h2 = 0 but
V1∗ (ωk ) > 0, k = 1, 2, 3? This is equivalent to ask whether there exist h1 and
h2 such that 4h1 + 2h2 = 0 and
4h1 + 3h2 > 0
3h1 + 2h2 > 0
2h1 + 4h2 > 0. (2.1.13b)
In Fig. 2.1, we show the region containing the set of points in the (h1 , h2)-
plane that satisfy inequalities (2.1.13b). The region is found to be lying on
the top right sides above the two bold lines: (i) 3h1 + 2h2 = 0, h1 < 0 and
(ii) 2h1 + 4h2 = 0, h1 > 0. It is seen that all the points on the dotted half
line: 4h1 + 2h2 = 0, h1 < 0 represent dominant trading strategies that start
with zero wealth but end with positive wealth with certainty.
Suppose the initial discounted price vector is changed from (4 2) to
(3 3), the new set of linear pricing probabilities will be determined by
1 = q(ω1 ) + q(ω2) + q(ω3 )
3 = 4q(ω1 ) + 3q(ω2 ) + 2q(ω3)
3 = 3q(ω1 ) + 2q(ω2 ) + 4q(ω3), (2.1.14)
which is seen to have the solution: q(ω1 ) = q(ω2) = q(ω3) = 1/3. Now,
all the pricing probabilities have non-negative values, the row vector q =
(1/3 1/3 1/3) represents a linear pricing measure. Referring to Fig. 2.1,
we observe that the line 3h1 + 3h2 = 0 always lies outside the region above the
two bold lines. Hence, with respect to this new securities model, we cannot
find (h1 h2) such that 3h1 + 3h2 = 0 together with h1 and h2 satisfying
inequalities (2.1.13b). Since a linear pricing measure exists, by virtue of Eq.
 2.1 Single period securities models 45

(2.1.12), we expect that the initial price vector (3 3) can be expressed as

some linear combination of the 3 vectors: (4 3), (3 2) and (2 4) with non-
1 1 1
negative weights. Actually, we have (3 3) = (4 3) + (3 2) + (2 4),
3 3 3
where the weights are the linear pricing probabilities.

Fig. 2.1 The region above the two bold lines represents
trading strategies that satisfy inequalities (2.1.13b). The
trading strategies that lie on the dotted line: 4h1 + 2h2 =
0, h1 < 0 are dominant trading strategies.

Apparently, one may conjecture that the existence of linear pricing mea-
sure is related to the non-existence of dominant trading strategies. Indeed,
we have the following theorem.
Theorem 2.1
There exists a linear pricing measure if and only if there are no dominant
trading strategies.
The above linear pricing measure theorem can be seen to be a direct
consequence of the Farkas Lemma.
Farkas Lemma
There does not exist h ∈ RM such that

Sb∗ (1; Ω)h > 0 and S

b ∗ (0)h = 0
46 2 Concepts of Financial Economics and Asset Price Dynamics

if and only if there exists q ∈ RK such that

b ∗ (0) = qSb∗ (1; Ω) and
S q ≥ 0.

2.1.2 Arbitrage opportunities and risk neutral probability

measures

Suppose S∗ (0) in the above securities model is modified to (3 2) and con-

sider the trading strategy: h1 = −2 and h2 = 3. We observe that V0∗ = 0 and
the possible discounted payoffs at t = 1 are: V1∗ (ω1) = 1, V1∗ (ω2 ) = 0 and
V1∗ (ω3) = 8. This represents a trading strategy that starts with zero wealth,
guarantees no loss, and ends up with a strictly positive wealth in some states
(not necessarily in all states). The occurrence of such investment opportunity
is called an arbitrage opportunity. Formally, we define an arbitrage opportu-
nity to be some trading strategy that has the following properties: (i) V0∗ = 0,
(ii) V1∗ (ω) ≥ 0 and EV1∗ (ω) > 0, where E is the expectation under the ac-
tual probability measure P . Readers should be watchful for the difference
between dominant strategy and arbitrage opportunity, where the existence
of a dominant strategy requires a portfolio with initial zero wealth to end up
with a strictly positive wealth in all states. Therefore, the existence of a dom-
inant trading strategy implies the existence of an arbitrage opportunity, but
the converse is not necessarily true. In other words, the absence of arbitrage
implies the non-existence of dominant trading strategy and in turn implying
that the law of one price holds.
Existence of arbitrage opportunities is unreasonable from the economic
standpoint. The natural question: What would be the necessary and sufficient
condition for the non-existence of arbitrage opportunities? The answer is re-
lated to the existence of a pricing measure, called the risk neutral probability
measure. In financial markets with no arbitrage opportunities, we will show
that every investor should use such risk neutral probability measure (though
not necessarily unique) to find the fair value of a portfolio, irrespective to the
risk preference of the investor.
Risk neutral probability measure
The example just mentioned above represents the presence of an arbitrage
opportunity but non-existence of dominant trading strategy [since V1∗ (ω) = 0
for some ω]. The linear pricing measure vector is found to be (0 1 0),
where some of the linear pricing probabilities are zero. In order to exclude
arbitrage opportunities, we need a bit stronger condition on the linear pricing
probabilities, namely, the probabilities must be strictly positive.
A probability measure Q on Ω is a risk neutral probability measure if it
∗
satisfies (i) Q(ω) > 0 for all ω ∈ Ω, and (ii) EQ [∆Sm ] = 0, m = 1, · · ·, M ,
∗
where EQ denotes the expectation under Q. Note that EQ [∆Sm ] = 0 is equiv-
X
K
∗ ∗
alent to Sm (0) = Q(ωk )Sm (1; ωk), which takes similar form as Eq.(2.1.11).
k=1
 2.1 Single period securities models 47

Indeed, a linear pricing measure becomes a risk neutral probability measure

if the probability masses are all positive.
The existence of a risk neutral measure is directly related to the exclusion
of arbitrage opportunities, the details of which are stated in the following
theorem.
Theorem 2.2
No arbitrage opportunities exist if and only if there exists a risk neutral
probability measure Q.
The proof of Theorem 2.2 requires the Separating Hyperplane Theorem.
A brief intuition of the theorem is given here. First, we present the defi-
nitions of hyperplane and convex sets in a vector space. Let f be a vector
in Rn . The hyperplane H = [f , α] in Rn is defined to be the collection of
those vectors x in Rn whose projection
  onto f has magnitude α. For ex-
x1
 
ample, the collection of vectors  x2  satisfying x1 + 2x2 + 3x3 = 2 is
 x3
1
 
a hyperplane in R3 , where f =  2  and α = 2. A set C in Rn is said
3
to be convex if for any pair of vectors x and y in C, all convex combi-
nations of x and y represented by the  form
 λx  + (1 − λ)y, 0 ≤ λ ≤  1,
 x1 
also lie in C. For example, the set C =  x2  : x1 ≥ 0, x2 ≥ 0, x3 ≥ 0
 
x3
is a convex set in R3. The hyperplane [f , α] separates the sets A and B
in Rn if there exists α such that f · x ≥ α for all
 x ∈ A and f · y < α
1
  
for all y ∈ B. For example, the hyperplane  1  , 0 separates the
  1 
 x 1 
two disjoint convex sets A =  x2  : x1 ≥ 0, x2 ≥ 0, x3 ≥ 0 and B =
 
   x3
 x1 
 x2  : x1 < 0, x2 < 0, x3 < 0 in R3.
 
x3
The Separating Hyperplane Theorem states that if A and B are two non-
empty disjoint convex sets in a vector space V , then they can be separated
by a hyperplane. A pictorial interpretation of the Separating Hyperplane
Theorem for the vector space R2 is shown in Fig. 2.2.
48 2 Concepts of Financial Economics and Asset Price Dynamics

Fig. 2.2 The hyperplane (represented by a line in R2 )

separates the two convex sets A and B in R2 .

Proof of Theorem 2.2

“⇐ part”. Assume a risk neutral probability measure Q exists, that is,
b ∗ (0) = π Sb∗ (1; Ω), where π = (Q(ω1) · · · Q(ωK )). Consider a trading strat-
S
egy h = (h1 · · · hM )T ∈ RM such that S ∗ (1; Ω)h ≥ 0 in all ω ∈ Ω and
with strict inequality in some states. Now consider S b ∗ (0)h = π Sb∗ (1; Ω)h,
b ∗
it is seen that S (0)h > 0 since all entries in π are strictly positive and
entries in Sb∗ (1; Ω)h are either zero or strictly positive. Hence, no arbitrage
opportunities exist.
“⇒ part”. First, we define the subset U in RK+1 which consists of vectors
 b∗ 
− S (0)h
 b∗ 
 S (1; ω1)h 
  b ∗ (1; ωk) is the kth row in Sb∗ (1; Ω) and
of the form  .. , where S
 
 . 
b ∗
S (1; ωK )h
h ∈ RM represents a trading strategy. This subset is seen to be a convex
subspace. Consider another subset RK+1
+ defined by

RK+1
+ = {x = (x0 x1 · · · xK )T ∈ RK+1 : xi ≥ 0 for all 0 ≤ i ≤ K},

which is a convex set in RK+1 . We claim that the non-existence of arbitrage

opportunities implies that U and RK+1+ can only have the zero vector in
common.
Assume the contrary, suppose there exists a non-zero vector x ∈ U ∩RK+1
+ .
Since there is a trading strategy vector h assoicated with every vector in
U , it suffices to show that the trading strategy h associated with x al-
ways represents an arbitrage opportunity. We consider the following two
 2.1 Single period securities models 49

cases: −Sb ∗ (0)h = 0 or −Sb ∗ (0)h > 0. When Sb ∗(0)h = 0, since x 6= 0 and
K+1
x ∈ R+ , then the entries S(1; b ωk )h, k = 1, 2, · · ·K, must be all greater
than or equal to zero, with at least one strict inequality. In this case, h is
seen to represent an arbitrage opportunity. When S b ∗ (0)h < 0, all the entries
b ωk)h, k = 1, 2, · · ·, K must be all non-negative. Correspondingly, h rep-
S(1;
resents a dominant trading strategy (see Problem 2.1) and in turns h is an
arbitrage opportunity.
Since U ∩ RK+1+ = {0}, by the Separating Hyperplane Theorem, there
exists a hyperplane that separates RK+1+ \{0} and U . Let f ∈ RK+1 be the
normal to this hyperplane, then we have f · x > f · y, where x ∈ RK+1 + \{0}
and y ∈ U . [Remark: We may have f · x < f · y, depending on the orientation
of the normal. However, the final conclusion remains unchanged.] Since U is
a linear subspace so that a negative multiple of y ∈ U also belongs to U , the
condition f · x > f · y holds only if f · y = 0 for all y ∈ U . We then have
f · x > 0 for all x in RK+1
+ \{0}. This requires all entries in f to be strictly
positive. Also, from f · y = 0, we have
K
X
b ∗(0)h +
− f0 S b ∗(1; ωk )h = 0
fk S (2.1.15a)
k=1

for all h ∈ RM , where fj , j = 0, 1, · · ·, K are the entries of f . We then deduce

that
K
X
b ∗ (0) =
S b ∗ (1; ωk), where Q(ωk ) = fk /f0 .
Q(ωk )S (2.1.15b)
k=1

Lastly, we consider the first component in the vectors on both sides of the
above equation. They both correspond to the current price and discounted
payoff of the riskless security, and all are equal to one. We then obtain

X
K
1= Q(ωk ). (2.1.15c)
k=1

Here, we obtain the risk neutral probabilities Q(ωk ), k = 1, · · ·, K, whose sum

is equal to one and they are all strictly positive since fj > 0, j = 0, 1, · · ·, K.
Remark
Corresponding to each risky asset, Eq. (2.1.15b) dictates that

X
K
∗ ∗
Sm (0) = Q(ωk )Sm (1; ωk), m = 1, 2, · · ·, M. (2.1.16)
k=1

Hence, the current price of any risky security is given by the expectation of
the discounted payoff under the risk neutral measure Q.
50 2 Concepts of Financial Economics and Asset Price Dynamics

Calculation of risk neutral measures

Consider the earlier securities model
 with the riskfree security and only one
1 4
b Ω) = 
risky security, where S(1;
 b
 1 3  and S(0) = (1 3). The risk neu-
1 2
tral probability measure π = (Q(ω1 ) Q(ω2 ) Q(ω3 )), if exists, will be de-
termined by the following system of equations
 
1 4
(Q(ω1) Q(ω2 ) Q(ω3 ))  1 3  = (1 3). (2.1.17)
1 2
Since there are more unknowns than the number of equations, the solution
is not unique. The solution is found to be π = (λ 1 − 2λ λ), where λ is
a free parameter. In order that all risk neutral probabilities are all strictly
positive, we must have 0 < λ < 1/2. We would expect that uniqueness
of the risk neutral measure is directly related to the completeness of the
securities model.
 Suppose
 we add the second risky security with discounted
3
payoff S∗2(1) =  2  and current discounted value S2∗ (0) = 3. With this new
4
addition the securities model becomes complete (the asset span of the two
risky securities and the riskfree security is the whole R3 space). With the
new equation 3Q(ω1) + 2Q(ω2) + 4Q(ω3 ) = 3 added to the system (2.1.17),
this new securities model is seen to have the unique risk neutral measure
(1/3 1/3 1/3).
Let W be a subspace in RK which consists of discounted gains correspond-
ing to some trading strategy h. In the above securities
  model the discounted
 
4 3 1
gains of the first and second risky securities are  3  −  3  =  0 
      2 3 −1
3 3 0
and  2  −  3  =  −1 , respectively. Therefore, the corresponding dis-
4 3 1
counted gain subspace is given by
     
 1 0 
W = h1  0  + h2  −1  , where h1 and h2 are scalars . (2.1.18)
 
−1 1
For any risk neutral probability measure Q, we have
" M #
XK X
EQ G∗ = Q(ωk ) hm ∆Sm∗
(ωk )
k=1 m=1
M
X
∗
= hm EQ [∆Sm ] = 0, (2.1.19)
m=1
 2.1 Single period securities models 51

∗
where ∆Sm (ωk ) is the discounted gain on the mth risky security when the
state ωk occurs. Therefore, the risk neutral probability vector π must lie in
the orthogonal complement W ⊥ . Since the sum of risk neutral probabilities
must be one and all probability values must be positive, the risk neutral
probability vector π must lie in the following subset

P + = {y ∈ RK : y1 + y2 + · · · + yK = 1 and yk > 0, k = 1, · · · K}.(2.1.20)

Let R denote the set of all risk neutral measures. From the above arguments,
we see that R = P + ∩ W ⊥ .
In the above numerical example, W ⊥ is the line through the origin in R3
which is perpendicular to (1 0 − 1) and (0 − 1 1). The line should
assume the form λ(1 1 1) for some scalar λ. Together with the constraints
that sum of components equals one and each component is positive, we ob-
tain the risk neutral probability vector π = (1/3 1/3 1/3). The risk neu-
tral measure of this securities model is unique since the securities model is
complete.

2.1.3 Valuation of contingent claims

A contingent claim can be considered as a random variable Y that represents

a terminal payoff whose value depends on the occurrence of a particular state
ωk , where ωk ∈ Ω. Suppose the holder of the contingent claim is promised
to receive the preset payoff, how much should the writer of such contingent
claim charge at t = 0 so that the price is fair to both parties.
Consider the securities model with the riskfree security whose values at
t = 0 and t = 1 are S0 (0) = 1 and  S0 (1)= 1.1, respectively, and a risky se-
4.4
curity with S1 (0) = 3 and S1(1) =  3.3 . The set of t = 1 payoffs that can
2.2    
1.1 4.4
be generated by certain trading strategy is given by h0 1.1  + h1 3.3 
1.1   2.2
5.5
for some scalars h0 and h1 . For example, the contingent claim  4.4  can be
3.3
generated bythe  trading strategy: h0 = 1 and h1 = 1, while the other contin-
5.5
gent claim  4.0  cannot be generated by any trading strategy associated
3.3
with the given securities model. A contingent claim Y is said to be attainable
if there exists some trading strategy h, called the replicating portfolio, such
that V1 = Y for all possible states occurring at t = 1.
52 2 Concepts of Financial Economics and Asset Price Dynamics

 What  should be the price at t = 0 of the attainable contingent claim

5.5
 4.4 ? One may propose that the price at t = 0 of the replicating portfolio
3.3
is given by V0 = h0S0 (0)+h1S1 (0) = 1×1+1×3 = 4. As discussed in previous
sub-section, suppose there are no arbitrage opportunities (equivalent to the
existence of a risk neutral probability measure), then the law of one price
holds and so V0 is unique. The price at t = 0 of the contingent claim Y
is simply V0 , the price that is implied by the arbitrage pricing theory. If
otherwise, suppose the price p of the contingent claim at t = 0 is greater
than V0 , an arbitrageur can lock in a riskfree profit of amount p − V0 by
shorting selling the contingent claim and buying the replicating portfolio.
The arbitrage strategy is reversed if p < V0. In this securities model, we
have shown earlier that risk neutral probability measures do exist (though
not
 unique).
 By the above argument, the initial price of the contingent claim
5.5
 4.4  is unique and it is found to be V0 = 4.
3.3
Consider a given attainable contingent claim Y which is generated by
certain trading strategy. The associated discounted gain G∗ of the trading
XM
∗
strategy is given by G∗ = hm ∆Sm . Now, suppose a risk neutral proba-
m=1
bility measure Q associated with the securities model exists, we have

V0 = EQ V0∗ = EQ [V1∗ − G∗ ]. (2.1.21a)

Since EQ [G∗ ] = 0 and V1∗ = Y /S0(1), we obtain

V0 = EQ [Y /S0(1)]. (2.1.21b)

Recall that the existence of the risk neutral probability measure implies the
law of one price. Does EQ [Y /S0(1)] assume the same value for every risk
neutral probability measure Q? This must be true by virtue of the law of
one price since we cannot have two different values for V0 corresponding to
the same contingent claim Y . This gives the risk neutral valuation principle:
The price at t = 0 of an attainable claim Y is given by the expectation under
any risk neutral measure Q of the discounted value of the contingent claim.
Actually, one can show that a rather strong result: If EQ [Y /S0(1)] takes the
same value for every Q, then the contingent claim Y is attainable [for proof,
see Pliska’s text (1997)].
Readers are reminded that if the law of one price does not hold for a given
securities model, we cannot define a unique price for an attainable contingent
claim (see Problem 2.12).
 2.1 Single period securities models 53

State prices
Suppose we take Y to be the following contingent claim: Y ∗ = Y /S0(1) equals
one if ω = ωk for some ωk ∈ Ω and zero otherwise. This is just the Arrow
security ek corresponding to the state ωk . We then have
EQ [Y /S0(1)] = π ek = Q(ωk ). (2.1.22)
The price of the Arrow security with discounted payoff ek is called the state
price for state ωk ∈ Ω. The above result shows that the state price for ωk is
equal to the risk neutral probability for the same state.
Any contingent claim Y can be written as a linear combination of these
XK
basic Arrow securities. Suppose Y ∗ = Y /S0 = αk ek , then the price at
k=1
K
X
t = 0 of the contingent claim is equal to αk Q(ωk ). For example, suppose
k=1
   
5 1 4
Y∗ = 4 and Sb∗ (1; Ω) =  1 3  , (2.1.23a)
3 1 2
we have seen that the risk neutral probability is given by
π = (λ 1 − 2λ λ), where 0 < λ < 1/2. (2.1.23b)
The price at t = 0 of the contingent claim is given by
V0 = 5λ + 4(1 − 2λ) + 3λ = 4, (2.1.23c)
which is independent of λ. This verifies the earlier claim that EQ [Y /S0(1)]
assumes the same value for any risk neutral measure Q.
Complete markets
Recall that a securities model is complete if every contingent claim Y lies in
the asset span, that is, Y can be generated by some trading strategy. Consider
the augmented terminal payoff matrix
 
S0 (1; ω1) S1 (1; ω1) · · · SM (1; ω1)
b Ω) = 
S(1; 
.. .. .. 
 , (2.1.24)
. . .
S0 (1; ωK ) S1 (1; ωK ) · · · SM (1; ωK )
we deduce from linear algebra theory that Y always lies in the asset span if
b Ω) is equal to RK . Since the dimension
and only if the column space of S(1;
b Ω) cannot be greater than M + 1, therefore
of the column space of S(1;
a necessary condition for market completeness is that M + 1 ≥ K. Under
market completeness, if the set of risk neutral probability measures is non-
empty, then it must be a singleton (see Problem 2.11). Furthermore, when
b Ω) has independent columns and the asset span is the whole RK , then
S(1;
54 2 Concepts of Financial Economics and Asset Price Dynamics

M + 1 = K. In this case, the trading strategy that generates Y must be

unique since there are no redundant securities. On the other hand, when the
asset span is the whole RK but some securities are redundant, the trading
strategy that generates Y would not be unique. However, the price at t = 0
of the contingent claim is unique under arbitrage pricing, independent of the
chosen trading strategy. This is a consequence of the law of one price, which
holds since risk neutral measure exists.
When the dimension of the column space S(1; b Ω) is less than K, then
not all contingent claims can be attainable. In this case, a non-attainable
contingent claim cannot be priced using arbitrage pricing theory. However,
we may specify an interval (V− (Y ), V+ (Y )) where a reasonable price at t = 0
of the contingent claim should lie. The lower and upper bounds are given by

V+ (Y ) = inf{EQ [Ye /S0(1)] : Ye ≥ Y and Ye is attainable} (2.1.24a)

V− (Y ) = sup{EQ[Ye /S0 (1)] : Ye ≤ Y and Ye is attainable}. (2.1.24b)

Here, V+ (Y ) is the minimum value among all prices of attainable contin-

gent claims that dominate the non-attainable claim Y , while V− (Y ) is the
maximum value among all prices of attainable contingent claims that are
dominated by Y . Suppose V (Y ) > V+ (Y ), then an arbitrageur can lock in
riskless profit by selling the contingent claim to receive V (Y ) and use V+ (Y )
to construct the replicating portfolio that generates the attainable Ye as de-
fined in Eq. (2.1.24a). The upfront positive gain is V (Y ) − V+ (Y ). At t = 1,
the payoff from the replicating portfolio always dominates that of Y so that
no loss at expiry is also ensured.

2.1.4 Principles of binomial option pricing model

We would like to illustrate the risk neutral valuation of contingent claims

using the renowned binomial option pricing model. In the binomial model,
the asset price movement is simulated by a discrete binomial random walk
model (see Sec. 2.3.1 for a more detailed discussion on random walk models).
Here, we limit our discussion to the one-period binomial model, and defer
the analysis of the multi-period binomial model later (see Sec. 2.2.4). We will
show that the option price obtained from the binomial model depends only
on the riskless interest rate but independent on the actual expected rate of
return of the asset price.
Formulation of the replicating portfolio
We follow the derivation of the discrete binomial model presented by Cox,
Ross and Rubinstein (1979). They showed that by buying the asset and bor-
rowing cash (in the form of riskless investment) in appropriate proportions,
one can replicate the position of a call. Let S denote the current asset price.
Under the binomial random walk model, the asset price after one period ∆t
will be either uS or dS with probability q and 1 − q, respectively (see Fig.
 2.1 Single period securities models 55

2.3). We assume u > 1 > d so that uS and dS represent the up-move and
down-move of the asset price, respectively. The jump parameters u and d
will be related to the asset price dynamics, the detailed discussion of which
will be relegated to Sec. 7.1.1. Let R denote the growth factor of riskless
investment over one period so that $1 invested in a riskless money market
account will grow to $R after one period. In order to avoid riskless arbitrage
opportunities, we must have u > R > d (see Problem 2.14).
Suppose we form a portfolio which consists of α units of asset and cash
amount B in the form of riskless investment (money market account). After
one period of time 4t, the value of the portfolio becomes (see Fig. 2.3)

αuS + RM with probability q
αdS + RM with probability 1 − q.
The portfolio is used to replicate the long position of a call option on
a non-dividend paying asset. As there are two possible states of the world:
asset price goes up or down, the call is thus a contingent claim. Suppose
the current time is only one period 4t prior to expiration. Let c denote the
current call price, and cu and cd denote the call price after one period (which
is the expiration time in the present context) corresponding to the up-move
and down-move of the asset price, respectively. Let X denote the strike price
of the call. The payoff of the call at expiry is given by

cu = max(uS − X, 0) with probability q
cd = max(dS − X, 0) with probability 1 − q.

Fig. 2.3 Evolution of the asset price S and money mar-

ket account M after one time period under the binomial
model. The risky asset value may either go up to uS or
go down to dS, while the riskless investment amount M
grows to RM .

The above portfolio containing the risky asset and money market account
is said to replicate the long position of the call if and only if the values of the
portfolio and the call option match for each possible outcome, that is,
αuS + RM = cu and αdS + RM = cd . (2.1.25)
The unknowns are α and M in the above linear system of equations. It occurs
that the number of unknowns (related to the number of units of asset and
56 2 Concepts of Financial Economics and Asset Price Dynamics

cash amount) and the number of equations (two possible states of the world
under the binomial model) are equal. Solving the equations, we obtain
cu − cd ucd − dcu
α= ≥ 0, M = ≤ 0. (2.1.26)
(u − d)S (u − d)R
Since M is always non-positive, the replicating portfolio involves buying the
asset and borrowing cash in the proportions given by Eq. (2.1.26). The num-
ber of units of asset held is seen to be the ratio of the difference of call values
cu − cd to the difference of asset values uS − dS.
Under the present one-period binomial model of asset price dynamics, we
observe that the call option can be replicated by a portfolio of basic securities:
risky asset and riskfree money market account.
Binomial option pricing formula
By the principle of no arbitrage, the current value of the call must be the same
as that of the replicating portfolio. What happens if it were not? Suppose the
current value of the call is less than the portfolio value, then we could make
a riskless profit by buying the cheaper call and selling the more expensive
portfolio. The net gain from the above two transactions is secured since the
portfolio value and call value will cancel each other off one period later.
The argument can be reversed if the call is worth more than the portfolio.
Therefore, the current value of the call is given by the current value of the
portfolio, that is,
R−d u−R
c
u−d u
+ c
u−d d
c = αS + M =
R (2.1.27)
pcu + (1 − p)cd R−d
= where p= .
R u−d
Note that the probability q, which is the subjective probability about upward
or downward movement of the asset price, does not appear in the call value
formula (2.1.27). The parameter p can be shown to be 0 < p < 1 since
u > R > d and so p can be interpreted as a probability. Further, from the
relation
R−d u−R
puS + (1 − p)dS = uS + dS = RS, (2.1.28)
u−d u−d
one can interpret the result as follows: the expected rate of returns on the
asset with p as the probability of upside move is just equal to the riskless
interest rate. Let S ∆t be the random variable that denotes the asset price
one period later. We may express Eq. (2.1.28) as
1 ∗ ∆t
S= E (S |S), (2.1.29)
R
where E ∗ is expectation under this probability measure. According to the
definition given in Sec. 2.1.2, we may view p as the risk neutral probability.
 2.2 Filtrations, martingales and multi-period models 57

Similarly, the call price formula (2.1.27) can be interpreted as the expectation
of the payoff of the call option at expiry under the risk neutral probability
measure discounted at the riskless interest rate [see Eq. (2.1.21b) for com-
parison]. The binomial call value formula (2.1.27) can be expressed as
1 ∗ ∆t

c= E c |S , (2.1.30)
R
where c denotes the call value at the current time, and c∆t denotes the random
variable representing the call value one period later.
Besides applying the principle of replication of claims, the binomial option
pricing formula can also be derived using the riskless hedging principle or via
the concept of state prices (see Problems 2.15 and 2.16).

2.2 Filtrations, martingales and multi-period models

In this section, we extend our discussion of securities models to multi-

period, where there are T + 1 trading dates: t = 0, 1, · · ·, T, T > 1. Sim-
ilar to an one-period model, we have a finite sample space Ω of K ele-
ments, Ω = {ω1, ω2, · · · , ωK }, which represents the possible states of the
world. There is a probability measure P defined on the sample space with
P (ω) > 0 for all ω ∈ Ω. The securities model consists of M risky securities
whose price processes are non-negative stochastic processes, as denoted by
Sm = {Sm (t); t = 0, 1, · · ·, T }, m = 1, · · · , M . In addition, there is a riskfree
security whose price process S0 (t) is deterministic, with S0 (t) strictly posi-
tive and possibly non-decreasing. We may consider S0 (t) as a money market
S0 (t) − S0 (t − 1)
account, and the quantity rt = , t = 1, · · · , T , is visualized
S0 (t − 1)
as the interest rate over the time interval (t − 1, t).
In this section, we would like to show that the concepts of arbitrage op-
portunity and risk neutral valuation can be carried over from single-period
models to multi-period models. However, we need to specify how the in-
vestors learn about the true state of the world on intermediate trading dates
in a multi-period model. Accordingly, we have to construct some information
structure that models how information is revealed to investors in terms of
the subsets of the sample space Ω. We show how information structure can
be described by a filtration and understand how security price processes can
be adapted to a given filtration. After then, we introduce martingales, which
are defined with reference to conditional expectations. In the multi-period
setting, the risk neutral probability measures are defined in terms of mar-
tingales. The highlight of this section is the derivation of the Fundamental
Theorem of Asset Pricing. The last part of this section will be devoted to the
multi-period binomial models.
58 2 Concepts of Financial Economics and Asset Price Dynamics

2.2.1 Information structures and filtrations

Consider the sample space Ω = {ω1, ω2, · · · , ω10} with 10 elements. We can
construct various partitions of the set Ω. A partition of Ω is a collection
P = {B1 , B2 , · · ·Bn } such that Bj , j = 1, · · · , n, are subsets of Ω and Bi ∩
[ n
Bj = φ, i 6= j, and Bj = Ω. Each of the sets B1 , · · ·, Bn is called an atom
j=1
of the partition. For example, we may form the partitions as

P0 = {Ω}
P1 = {{ω1, ω2, ω3, ω4}, {ω5, ω6, ω7, ω8, ω9, ω10}}
P2 = {{ω1, ω2}, {ω3, ω4}, {ω5, ω6}, {ω7, ω8, ω9}, {ω10}}
P3 = {{ω1}, {ω2}, {ω3}, {ω4}, {ω5}, {ω6}, {ω7}, {ω8}, {ω9}, {ω10}}.

In the above, we have defined a finite sequence of partitions of Ω, which have

the property that they are nested with successive refinements of one another.
Each set belonging to Pk splits into smaller sets which are elements of Pk+1.

Fig. 2.4 Information tree of a three-period securities

model with 10 possible states. The partitions form a se-
quence of successively finer partitions.
Consider a three-period securities model that consists of the above se-
quence of successively finer partitions: {Pk : k = 0, 1, 2, 3}. The pair (Ω, Pk )
is called a filtered space, which consists of a sample space Ω and a sequence of
partitions of Ω. The filtered space is used to model the unfolding of informa-
tion through time. At time t = 0, the investors know only the set of all possible
 2.2 Filtrations, martingales and multi-period models 59

outcomes, so P0 = {Ω}. At time t = 1, the investors get a bit more informa-

tion: the actual state ω is in either {ω1, ω2, ω3, ω4} or {ω5 , ω5, ω7, ω8, ω9, ω10}.
In the next trading date t = 2, more information is revealed, say, ω is in the
set {ω7, ω8, ω9}. On the last date t = 3, we have P3 = {{ωi }, i = 1, · · · , 10}.
Each set of P3 consists of a single element of Ω, so the investors have full in-
formation of which particular state has occurred. The information submodel
of this three-period securities model can be represented by the information
tree shown in Fig. 2.4.
Algebra
Let Ω be a finite set and F be a collection of subsets of Ω. The collection F
is an algebra on Ω if
(i) Ω ∈ F
(ii) B ∈ F ⇒ B c ∈ F
(iii)B1 and B2 ∈ F ⇒ B1 ∪ B2 ∈ F.
Given an algebra F on Ω, one can always find a unique collection of
disjoint subsets Bn such that each Bn ∈ F and the union of the subsets
equals Ω. The algebra F generated by a partition P = {B1 , · · · , Bn} is a
set of subsets of Ω. Actually, when Ω is a finite sample space, there is a
one-to-one correspondence between partitions of Ω and algebras on Ω. The
information structure defined by a sequence of partitions can be visualized
as a sequence of algebras. We define a filtration F = {Fk ; k = 0, 1, · · ·, T } to
be a nested sequence of algebras satisfying Fk ⊆ Fk+1.
As an example, given the algebra F = {φ, {ω1}, {ω2, ω3}, {ω4}, {ω1, ω2,
ω3}, {ω2, ω3, ω4}, {ω1, ω4}, {ω1, ω2, ω3, ω4}}, the corresponding partition P is
found to be {{ω1}, {ω2, ω3}, {ω4}}. The atoms of P are B1 = {ω1 }, B2 =
{ω2, ω3} and B3 = {ω4}. A non-empty event whose occurrence to be revealed
through revelation of P would be an union of atoms in P. Take the event A =
{ω1, ω2, ω3}, which is the union of B1 and B2 . Given that B2 = {ω2, ω3} of P
has occurred, we can decide whether A or its complement Ac has occurred.
However, for another event A e = {ω1, ω2}, even though we know that B2 has
occurred, we cannot determine whether A e or Aec has occurred.
Next, we define a probability measure P defined on an algebra F. The
probability measure P is a function
P : F → [0, 1]
such that
1. P (Ω) = 1.
2. If B1 , B2, · · · are pairwise disjoint sets belonging to F, then
P (B1 ∪ B2 ∪ · · ·) = P (B1) + P (B2 ) + · · · .
Equipped with a probability measure, the elements of F are called measurable
events. Given the sample space Ω, an algebra F and a probability measure
P defined on Ω, the triplet (Ω, F, P ) together with the filtration F is called
a filtered probability space.
60 2 Concepts of Financial Economics and Asset Price Dynamics

Equivalent measures
Given two probability measures P and P 0 defined on the same measurable
space (Ω, F), suppose that

P (ω) > 0 ⇐⇒ P 0(ω) > 0, for all ω ∈ Ω,

then P and P 0 are said to be equivalent measures. In other words, though

the two equivalent measures may not agree on the assignment of probability
values to individual events, but they always agree as to which events are
possible or impossible.
Measurability of random variables
Consider an algebra F generated by a partition P = {B1 , · · ·, Bn }, a random
variable X is said to be measurable with respect to F (denoted by X ∈ F) if
X(ω) is constant for all ω ∈ Bi , Bi is any element in P. For example, consider
the algebra F1 generated by P1 = {{ω1, ω2, ω3, ω4}, {ω5, ω6, ω7, ω8, ω9, ω10}}.
If X(ω1 ) = 3 and X(ω4 ) = 5, then X is not measurable with respect to F1.
Consider an example where P = {{ω1, ω2}, {ω3, ω4}, {ω5}} and X is mea-
surable with respect to the algebra F generated by P. Let X(ω1 ) = X(ω2 ) =
3, X(ω3) = X(ω4 ) = 5 and X(ω5 ) = 7. Suppose the random experiment as-
sociated with the random variable X is performed, giving X = 5. This tells
the information that the event {ω3, ω4} has occurred. One can argue that the
information of outcome from the random experiment is revealed through the
random variable X. We may say that F is being generated by X.
A stochastic process Sm = {Sm (t); t = 0, 1, · · ·, T } is said to be adapted
to the filtration F = {Ft ; t = 0, 1, · · ·, T } if the random variables Sm (t) is
Ft-measurable for each t = 0, 1, · · ·, T . For the bank account process S0 (t),
the interest rate is normally known at the beginning of the period so that
S0 (t) is Ft−1-measurable, t = 1, · · ·, T . In this case, we say that the process
S0 (t) is predictable.

2.2.2 Conditional expectations and martingales

Consider the filtered probability space defined by the triplet (Ω, F, P ) to-
gether with the filtration F. Recall that a random variable is a mapping
ω → X(ω) that assigns a real number X(ω) to each ω ∈ Ω. A random
variable is said to be simple if X can be decomposed into the form
n
X
X(ω) = aj IBj (ω) (2.2.1)
j=1

where {B1 , · · ·Bn } is a finite partition of Ω with each Bj ∈ F and the

indicator of Bj is defined by
n
1 if ω ∈ Bj
IBj (ω) = . (2.2.2)
0 if otherwise
 2.2 Filtrations, martingales and multi-period models 61

The expectation of X with respect to the probability measure P is defined

as
X n n
X
E[X] = aj E[IBj (ω)] = aj P [Bj ], (2.2.3)
j=1 j=1

where P (Bj ) is the probability that a state ω contained in Bj occurs. The

conditional expectation of X given that event B has occurred is defined to
be
X
E[X|B] = xP [X = x|B]
x
X
= xP [X = x, B]/P [B]
x
1 X
= X(ω)P [ω]. (2.2.4)
P [B]
ω∈B

Fig. 2.5 The asset price process of a two-period securities

model.
As a numerical example, consider the sample space Ω = {ω1, ω2, ω3, ω4}
and the algebra is generated by the partition P = {{ω1, ω2}, {ω3, ω4}}. The
probabilities of occurrence of the states are given by P [ω1] = 0.2, P [ω2] =
0.3, P [ω3] = 0.35 and P [ω4] = 0.15. Consider the two-period price process S
whose values are given by
S(1; ω1 ) = 3, S(1; ω2 ) = 3, S(1; ω3 ) = 5, S(1; ω4 ) = 5,
S(2; ω1 ) = 4, S(2; ω2 ) = 2, S(2; ω3 ) = 4, S(2; ω4 ) = 6,
where the tree representation is shown in Fig. 2.5. The conditional expecta-
tions E[S(2)| S(1) = 3] and E[S(2)|S(1) = 5] are computed using Eq. (2.2.4)
as follows:
S(2; ω1)P [ω1] + S(2; ω2 )P [ω2]
E[S(2)|S(1) = 3] =
P [ω1] + P [ω2]
= (4 × 0.2 + 2 × 0.3)/0.5 = 2.8; (2.2.5a)
S(2; ω3)P [ω3] + S(2; ω4 )P [ω4]
E[S(2)|S(1) = 5] =
P [ω3] + P [ω4]
= (4 × 0.35 + 6 × 0.15)/0.5 = 4.6. (2.2.5b)
62 2 Concepts of Financial Economics and Asset Price Dynamics

Interpretation of E[X|F]
It is quite often that we would like to consider all conditional expectations of
the form E[X|B] where the event B runs through the algebra F. We define
the quantity E[X|F] by

E[X|F]IB = E[X|B] for all B ∈ F. (2.2.6)

We see that E[X|F] is actually a random variable that is measurable with

respect to the algebra F. In the above numerical example, suppose we write
F1 = {{ω1, ω2}, {ω3, ω4}}, then

2.8 if ω1 or ω2 occurs
E[S(2)|F1] = . (2.2.7)
4.6 if ω3 or ω4 occurs

Since E[X|F] is a random variable, we may compute its expectation. We find

that
X X X
E[E[X|F]] = E[X|B]P [B] = X(ω)(P [ω]/P [B])P [B]
B∈F B∈F ω∈B
X X
= X(ω)P [ω] = E[X]. (2.2.8a)
B∈F ω∈B

The above result can be generalized as follows. If F1 ⊂ F2, then

E[E[X|F2]|F1] = E[X|F1]. (2.2.8b)

If we condition first on the information up to F2 and later on the information

F1 at an earlier time, then it is the same as conditioning originally on F1.
This is called the tower property of conditional expectations.
Suppose that the random variable X is F-measurable, we would like to
show E[XY |F] = XE[Y X |F] for any random variable Y . Using Eq. (2.2.1),
we may write X = aj IBj , where P is the partition corresponding to the
Bj ∈P
algebra F. Further, by Eq. (2.2.6), we obtain
X X
E[XY |F] = E[XY |Bj ]IBj = E[aj Y |Bj ]IBj
Bj ∈P Bj ∈P
X
= aj E[Y |Bj ]IBj = XE[Y |F]. (2.2.9)
Bj ∈P

When we take the conditional expectation with respect to the filtration F,

we can treat X as constant if X is known with regard to the information
provided by F. The proofs of other properties on conditional expectations
are relegated as exercises (see Problem 2.18).
 2.2 Filtrations, martingales and multi-period models 63

Martingales
The term martingale has its origin in gambling. It refers to the gambling tac-
tics of doubling the stake when losing in order to recoup oneself. In the stud-
ies of stochastic processes, martingales are defined in relation to an adapted
stochastic process.
Consider a filtered probability space with filtration F = {Ft; t = 0, 1, · · ·,
T }. An adapted stochastic process S = {S(t); t = 0, 1 · · ·, T } is said to be
martingale if it observes

E[S(t + s)|Ft] = S(t) for all t ≥ 0 and s ≥ 0. (2.2.10)

We define an adapted stochastic process S to be a supermartingale if

E[S(t + s)|Ft] ≤ S(t) for all t ≥ 0 and s ≥ 0; (2.2.11a)

and a submartingale if

E[S(t + s)|Ft] ≥ S(t) for all t ≥ 0 and s ≥ 0. (2.2.11b)

It is straightforward to deduce the following properties:

1. All martingales are supermartingales, but not vice versa. The same ob-
servation is applied to submartingales.
2. An adapted stochastic process S is a submartingale if and only if −S is a
supermartingale; S is a martingale if and only if it is both a supermartin-
gale and a submartingale.
Martingales are related to models of fair gambling. For example, let Xn
represent the amount of money a player possesses at stage n of the game.
The martingale property means that the expected amount of the player would
have at stage n + 1 given that Xn = αn, is equal to αn, regardless of his past
history of fortune.
It must be emphasized that a martingale is defined with respect to a
filtration (information set) and with respect to some probability measure.
In later sections we will show that martingales are found to be very useful
tools in option pricing theory. Indeed, the risk neutral valuation approach in
option pricing theory is closely related to the theory of martingales. In Sec.
2.2.3, we show that the necessary and sufficient condition for the exclusion of
arbitrage opportunities in a securities model is the existence of a risk neutral
pricing measure constructed from the martingale property of the asset price
processes.
Martingale transforms
Suppose S is a martingale and H is a predictable process with respect to the
filtration F = {Ft; t = 0, 1, · · ·, T }, we define the process
t
X
Gt = Hu∆Su , (2.2.12)
u=1
64 2 Concepts of Financial Economics and Asset Price Dynamics

where ∆Su = Su − Su−1. One then deduces that ∆Gu = Gu − Gu−1 =

Hu∆Su . If S and H represent the asset price process and trading strategy,
respectively, then G can be visualized as the gain process. Note that trad-
ing strategy is a predictable process, that is Ht is Ft−1-measurable. This
is because the number of units held for each security is determined at the
beginning of the trading period by taking into account all the information
available up to that time.
We call G to be the martingale transform of S by H, as G itself is also a
martingale. To show the claim, it suffices to show that E[Gt+s|Ft] = Gt , t ≥
0, s ≥ 0. We consider

E[Gt+s|Ft] = E[Gt+s − Gt + Gt |Ft]

= E[Ht+1∆St+1 + · · · + Ht+s∆St+s |Ft] + E[Gt|Ft]
= E[Ht+1∆St+1 |Ft] + · · · + E[Ht+s∆St+s |Ft] + Gt . (2.2.13)

Consider the typical term E[Ht+u∆St+u |Ft], we can express it as E[E[Ht+u

∆St+u |Ft+u−1]|Ft], by the result in Eq. (2.2.8b). Further, since Ht+u is
Ft+u−1-measurable and S is a martingale, by virtue of Eqs. (2.2.9,10), we
have

E[Ht+u∆St+u|Ft+u−1] = Ht+uE[∆St+u|Ft+u−1] = 0. (2.2.14)

Collecting all the calculations, we obtain the desired result.

2.2.3 Multi-period securities models

Once we are equipped with the knowledge of filtrations, adapted stochastic

processes and martingales, we now move to the discussion of the fundamentals
of financial economics of multi-period securities models. In particular, we
will consider the relation between exclusion of arbitrage opportunities and
existence of the martingale measure (risk neutral probability measure).
We start with the prescription of a discrete n-period securities model with
M risky securities. Like the discrete single-period model, there is a sample
space Ω = {ω1, ω2, · · · , ωK } of K possible states of the world. Let S denote
the asset price process {S(t); t = 0, 1, · · ·, n}, where S(t) is the row vector
S(t) = (S1 (t) S2 (t) · · · SM (t)) and whose components are security prices.
Also, there is a bank account process S0 (t), whose value is given by

S0 (t) = (1 + r0)(1 + r1 ) · · · (1 + rt−1), (2.2.15)

where ru is the interest rate applied over one time period starting at time
u, u = 0, 1, · · ·, t − 1. A trading strategy is the rule taken by an investor
that specifies the investor’s position in each security at each time and in
each state of the world based on the available information as prescribed by a
filtration. Hence, one can visualize a trading strategy as an adapted stochas-
tic process. We prescribe a trading strategy by a vector stochastic process
 2.2 Filtrations, martingales and multi-period models 65

H(t) = (h1(t) h2 (t) · · · hM (t))T , t = 1, 2, · · ·, n (represented as a column vec-

tor), where hm (t) is the number of units held in the portfolio for the mth
security from time t − 1 to time t. Also, the amount of bank account held at
time t−1 is given by h0(t)S0 (t). Note that hm (t) should be Ft−1-measurable,
m = 0, 1, · · ·, M .
The value of the portfolio is a stochastic process given by
M
X
V (t) = h0(t)S0 (t) + hm (t)Sm (t), t = 1, 2, · · ·, n, (2.2.16)
m=1

which gives the portfolio value at the moment right after the asset prices are
observed but before changes in portfolio weights are made.
We write ∆Sm (t) = Sm (t) − Sm (t − 1) as the change in value of one unit
of the mth security between times t−1 and t. The cumulative gain associated
with investing in the mth security from time zero to time t is given by
t
X
hm (u)∆Sm (u).
u=1

We define the gain process G(t) to be the total cumulative gain in holding
the portfolio consisting of the M risky securities and the bank account up to
time t. The value of G(t) is found to be

X
t X
M X
t
G(t) = h0 (u)∆S0(u) + hm (u)∆Sm (u), t = 1, 2, · · ·, n. (2.2.17)
u=1 m=1 u=1

∗
If we define the discounted price process Sm (t) by
∗
Sm (t) = Sm (t)/S0 (t), t = 0, 1, · · ·, n, m = 1, 2, · · ·, M, (2.2.18a)
∗ ∗ ∗
and write ∆Sm (t) = Sm (t) − Sm (t − 1), then the discounted value process
V ∗ (t) and discounted gain process G∗(t) are given by
M
X
V ∗ (t) = h0 (t) + ∗
hm (t)Sm (t), t = 1, 2, · · ·n, (2.2.18b)
m=1
X
M X
t
G∗ (t) = ∗
hm (u)∆Sm (u), t = 1, 2, · · ·, n. (2.2.18c)
m=1 u=1

Once the asset prices, Sm (t), m = 1, 2, · · ·, M , are revealed to the investor,

he changes the trading strategy from H(t) to H(t+1) as a response to the ar-
rival of the new information. In general, there may be addition or withdrawal
of fund from the portfolio. However, if there were no such addition and with-
drawal, then the portfolio value remains the same; and correspondingly, we
observe
66 2 Concepts of Financial Economics and Asset Price Dynamics

X
M
V (t) = h0 (t + 1)S0 (t) + hm (t + 1)Sm (t). (2.2.19)
m=1

The purchase of additional units of one particular security is financed by

the sales of other securities. In this case, the trading strategy is said to be
self-financing.
If there were no addition or withdrawal of funds at all trading times, then
the cumulative change of portfolio value V (t) − V (0) should be equal to the
gain G(t) associated with price changes of the securities on all trading dates.
Hence, we expect that a trading strategy H is self-financing if and only if

V (t) = V (0) + G(t). (2.2.20)

To show the claim, we rewrite Eqs. (2.2.16a) and (2.2.19) as

M
X
V (u) = h0(u)S0 (u) + hm (u)Sm (u) (2.2.21a)
m=1
M
X
V (u − 1) = h0(u)S0 (u − 1) + hm (u)Sm (u − 1), (2.2.21b)
m=1

respectively, then subtract the two equations to obtain

X
M
V (u) − V (u − 1) = h0(u)∆S0 (u) + hm (u)∆Sm (u). (2.2.21c)
m=1

Summing the above equation from u = 1 to u = t and applying the result

in Eq. (2.2.17), we obtain the result in Eq. (2.2.20). In a similar manner, we
can use Eqs. (2.2.18b,c) to show that H is self-financing if and only if

V ∗ (t) = V ∗ (0) + G∗ (t). (2.2.22)

No arbitrage principle and martingale measure

The definition of arbitrage opportunity for single period securities model
(see Sec. 2.1.2) is extended to multi-period models. A trading strategy H
represents an arbitrage opportunity if and only if the value process V (t) and
H satisfy the following properties:
(i) V (0) = 0,
(ii) V (T ) ≥ 0 and EV (T ) > 0, and
(iii) H is self-financing.
Equivalently, we may state that the self-financing trading strategy H is an
arbitrage opportunity if and only if (i) G∗ (T ) ≥ 0 and (ii) EG∗(T ) > 0. Like
that in single period models, we expect that arbitrage opportunity does not
exist if and only if there exists a risk neutral probability measure. In multi-
period models, risk neutral probabilities are defined in terms of martingales.
 2.2 Filtrations, martingales and multi-period models 67

Martingale measure
The measure Q is called a martingale measure (or called risk neutral proba-
bility measure) if it has the following properties:
1. Q(ω) > 0 for all ω ∈ Ω.
∗
2. Every discounted price process Sm in the securities model is a martingale
under Q, m = 1, 2, · · ·, M , that is,
∗ ∗
EQ [Sm (t + s)|Ft] = Sm (t) for all t ≥ 0 and s ≥ 0. (2.2.23)
∗
We call the discounted price process Sm (t) to be a Q-martingale.
As a numerical example, we determine the martingale measure Q asso-
ciated with the two-period securities model shown in Fig. 2.5. Let r ≥ 0 be
the constant riskless interest rate over one period, and write Q(ωj ) as the
martingale measure associated with the state ωj , j = 1, 2, 3, 4. By invoking
Eq. (2.2.23), we obtain the following equations for Q(ω1), · · · , Q(ω4):
(i) t = 0 and s = 1
3 5
4= [Q(ω1) + Q(ω2)] + [Q(ω3) + Q(ω4)] (2.2.24a)
1+r 1+r
(ii) t = 0 and s = 2
4 2
4= 2
Q(ω1) + Q(ω2 )
(1 + r) (1 + r)2
4 6
+ Q(ω3) + Q(ω4 ) (2.2.24b)
(1 + r)2 (1 + r)2
(iii) t = 1 and s = 1
4 Q(ω1 ) 2 Q(ω2 )
3= + (2.2.24c)
1 + r Q(ω1) + Q(ω2 ) 1 + r Q(ω1) + Q(ω2 )
4 Q(ω3 ) 6 Q(ω4 )
5= + . (2.2.24d)
1 + r Q(ω3) + Q(ω4 ) 1 + r Q(ω3) + Q(ω4 )
It may be quite tedious to solve the above equations simultaneously. The
calculation procedure can be simplified by observing that Q(ωj ) is given by
the product of the conditional probabilities along the path from the node at
t = 0 to the node ωj at t = 2. First, we start with the conditional probability
p associated with the upper branch {ω1, ω2}. The corresponding conditional
probability p is given by
3 5
4= p+ (1 − p) (2.2.25a)
1+r 1+r
1 − 4r
so that p = . Similarly, the conditional probability p0 associated with
2
the branch {ω1} from the node {ω1, ω2} is given by
68 2 Concepts of Financial Economics and Asset Price Dynamics

4 0 2
3= p + (1 − p0 ) (2.2.25b)
1+r 1+r
1 − 3r
giving p0 = . In a similar manner, the conditional probability p00 as-
2
1 − 5r
sociated with {ω3} from {ω3, ω4} is found to be . The martingale
2
probabilities are then found to be
1 − 4r 1 − 3r
Q(ω1) = pp0 = ,
2 2
1 − 4r 1 + 3r
Q(ω2) = p(1 − p0 ) = ,
2 2
1 + 4r 1 − 5r
Q(ω3) = (1 − p)p00 =
2 2
00 1 + 4r 1 + 5r
Q(ω4) = (1 − p)(1 − p ) = . (2.2.26)
2 2
These martingale probabilities can be shown to satisfy Eqs. (2.2.24a-d). In
order that the martingale probabilities remain positive, we have to impose
the restriction: r < 0.2.
Martingale property of value processes
Suppose H is a self-financing trading strategy and Q is a martingale measure
with respect to a filtration F, then the value process V (t) is a Q-martingale.
To show the claim, we use the relation

V ∗ (t) = V ∗ (0) + G∗ (t) (2.2.27a)

since H is self-financing [see Eq. (2.2.22)]; and deduce that

V ∗ (t + 1) − V ∗ (t) = G∗(t + 1) − G∗(t)

= [S ∗ (t + 1) − S ∗ (t)]H(t + 1). (2.2.27b)

As H is a predictable process, V ∗ (t) is the martingale transform of the Q-

martingale S ∗ (t). Hence, V ∗ (t) itself is also a Q-martingale.
The above result can be applied to show that the existence of martingale
measure Q implies the non-existence of arbitrage opportunities. To prove the
claim, suppose H is any self-financing trading strategy with V ∗ (T ) ≥ 0 and
EV ∗ (T ) > 0. As Q(ω) > 0, we then have EQ V ∗ (T ) > 0. Furthermore, since
V ∗ (T ) is a Q-martingale, we have V ∗ (0) = EQ V ∗ (T ) > 0. Therefore, H
cannot be an arbitrage opportunity.
The converse of the above claim remains to be valid, that is, the non-
existence of arbitrage opportunities implies the existence of a martingale
measure. The intuition behind the proof can be outlined as follows. If there
are no arbitrage opportunities in the multi-period model, then there will
be no arbitrage opportunities in any underlying single period. Since each
single period does not admit arbitrage opportunities, one can construct the
 2.2 Filtrations, martingales and multi-period models 69

one-period risk neutral conditional probabilities. The martingale probability

measure Q(ω) is then obtained by multiplying all the risk neutral conditional
probabilities along the path from the node at t = 0 to the terminal node
(T, ω). The rigorous proofs of these arguments are quite technical, the details
of which can be found in the text by Bingham and Kiesel (1998).
We summarize the result into the following theorem.
Theorem 2.3
A multi-period securities model is arbitrage free if and only if there exists
a probability measure Q such that the discounted asset price processes are
Q-martingales.
Most of the results on valuation of contingent claims in the single period
models can be extended to the multi-period models. First, the martingale
measure is unique if and only if the multi-period securities model is com-
plete. Here, completeness implies that all contingent claims (FT -measurable
random variables) can be replicated by a self-financing trading strategy. In
an arbitrage free complete market, the arbitrage price of a contingent claim
is then given by the discounted expected value under the martingale measure
of the unique portfolio that replicates the claim. Let Y denote a contingent
claim at maturity T and V (t) denote the arbitrage price of the contingent
claim at time t, t < T . We then have
S0 (t)
V (t) = EQ [Y |Ft], (2.2.28)
S0 (T )
where S0 (t) is the riskless asset and the ratio S0 (t)/S0 (T ) is the discount
factor over the period from t to T . Rigorous justification to these results can
be found in Bingham and Kiesel’s text (1998).

2.2.4 Multi-period binomial models

We extend the one-period binomial model to its multi-period version. We

start with the two-period binomial model. The corresponding dynamics of
the binomial process for the asset price and the call price are shown in Fig.
2.6. It is assumed that the jump ratios of the asset price, u and d, stay the
same value for all binomial steps.
Let cuu denote the call value at two periods beyond the current time
with two consecutive upward moves of the asset price and similar notational
interpretation for cud and cdd . Based on a similar argument as depicted in Eq.
(2.1.27), the call values cu and cd are related to cuu, cud and cdd as follows:
pcuu + (1 − p)cud pcud + (1 − p)cdd
cu = and cd = . (2.2.29a)
R R
Subsequently, by substituting the above results into Eq. (2.1.27), the call
value at the current time which is two periods from expiry is found to be
70 2 Concepts of Financial Economics and Asset Price Dynamics

p2 cuu + 2p(1 − p)cud + (1 − p)2 cdd

c= , (2.2.29b)
R2
where the corresponding terminal payoff values are given by

cuu = max(u2 S − X, 0), cud = max(udS − X, 0), cdd = max(d2 S − X, 0).

(2.2.30)
Note that the coefficients p2 , 2p(1 − p) and (1 − p)2 represent the respective
risk neutral probability of having two up jumps, one up jump and one down
jump, and two down jumps in two moves of the binomial process.

Fig. 2.6 Dynamics of the asset price and call price in a

two-period binomial model.
The extension of the binomial model to the n-period case should be
quite straightforward. With n binomial steps, the risk neutral probability
of having j up jumps and n − j down jumps is given by Cjnpj (1 − p)n−j ,
n!
where Cjn = is the binomial coefficient. The corresponding ter-
j!(n − j)!
minal payoff when j up jumps and n − j down jumps occur is seen to be
max(uj dn−j S − X, 0). The call value obtained from the n-period binomial
model is given by
X
n
Cjn pj (1 − p)n−j max(uj dn−j S − X, 0)
j=0
c= . (2.2.31)
Rn
We define k to be the smallest non-negative integer such that uk dn−kS ≥
X
ln Sd n
X, that is, k ≥ . It is seen that
ln ud

j n−j 0 when j < k
max(u d S − X, 0) = (2.2.32)
uj dn−j S − X when j ≥ k.
 2.2 Filtrations, martingales and multi-period models 71

The integer k gives the minimum number of upward moves required for the
asset price in the multiplicative binomial process in order that the call expires
in-the-money. The call formula can then be simplified as
X
n
uj dn−j X n
c=S Cjnpj (1 − p)n−j n
− XR−n Cjnpj (1 − p)n−j . (2.2.33)
R
j=k j=k

The last term in above equation can be interpreted as the expectation value
of the payment made by the holder at expiration discounted by the factor
Xn
R−n, and Cjn pj (1 − p)n−j is seen to be the probability (under the risk
j=k
neutral measure) that the call will expire in-the-money. The above probability
is related to the complementary binomial distribution function defined by
n
X
Φ(n, k, p) = Cjnpj (1 − p)n−j . (2.2.34)
j=k

Note that Φ(n, k, p) gives the probability for at least k successes in n trials
of a binomial experiment, where p is the probability of success in each trial.
up d(1 − p)
Further, if we write p0 = so that 1 − p0 = , then the call price
R R
formula for the n-period binomial model can be expressed as

c = SΦ(n, k, p0) − XR−n Φ(n, k, p). (2.2.35)

The first term gives the discounted expectation of the asset price at expiration
given that the call expires in-the-money and the second term gives the present
value of the expected cost incurred by exercising the call.
Using the argument of discounted expectation of the payoff of a contingent
claim under the risk neutral measure, the call price for the n-period binomial
model can be expressed in the following canonical form
1 ∗ 1
c= n
E (cT ) = n E ∗ [max(ST − X, 0)] , T = t + n∆t, (2.2.36)
R R
where cT is the payoff, max(ST − X, 0), of the call at expiration time T and
1
is the discount factor over n periods. Here, the expectation operator
R∗n
E is taken under the risk neutral measure rather than the true probability
measure associated with the actual (physical) asset price process.

Numerical implementation
The n-period binomial model can be represented schematically by a n-step
tree structure (see Fig. 2.7 for a 3-step tree). The binomial tree will be sym-
metrical about S if ud = 1, skewed upward if ud > 1 and skewed downward
if ud < 1. At the time level that is m time steps marching forward from
the current time in the binomial tree, there are m + 1 nodes. The asset
72 2 Concepts of Financial Economics and Asset Price Dynamics

price at the node obtained by j upward moves and m − j downward moves

equals Suj dm−j , j = 0, 1, · · ·m. The possible option values at expiration are
known since the payoff function at expiry is given in the option contract.
Rather than using the multiplicative binomial formula (2.2.31), the following
stepwise backward induction procedure is more effective in numerical imple-
mentation. First, we compute option values at the nodes that are one time
step from expiration using the binomial formula (2.1.27). Once option values
at one time step from expiration are known, we proceed two time steps from
expiration and repeat the same numerical procedure. This backward induc-
tion procedure is similar in spirit as the procedure used in the derivation of
Eqs. (2.2.28-29). After performing n backward steps in the tree, we come to
the starting node (tip of the tree) at which the option value is desired.

Fig. 2.7 Illustration of the binomial calculations with 3

time steps for a European call with strike price X = 70.
The top figures are asset prices and the bottom figures
are option values.

As a numerical example, suppose we have chosen the following values for

the binomial parameters: u = 1.25, d = 0.8, and the discount factor for one
period = 1/R = 0.95. According to Eq. (2.1.27), we have
 
R−d 1
p= = − 0.8 (1.25 − 0.8) = 0.5614. (2.2.37)
u−d 0.95
The strike price of the call is taken to be 70 and the asset price S at the
current time is 120. The binomial tree with three time steps is illustrated in
 2.3 Asset price dynamics and stochastic processes 73

Fig. 2.7. The upper and lower figures at the nodes denote the asset prices
and option values, respectively. For example, the option values at nodes P
and Q are, respectively, max(150 − 70, 0) = 80 and max(96 − 70, 0) = 26. The
option value at node Y is computed by
1
cY = [pcP + (1 − p)cQ ]
R
= 0.95(0.5614 × 80 + 0.4386 × 26)
= 53.50 (2 decimal places). (2.2.38)

Working backward from the expiration time to the current time, the current
option value at S = 120 is found to be 60.61 (see Fig. 2.7).

2.3 Asset price dynamics and stochastic processes

In this section, we discuss the stochastic models for the simulation of the
asset price movement. The asset price movement is said to follow a stochastic
process if its value changes over time in an uncertain manner. The study of
stochastic processes is concerned with the investigation of the structure of
families of random variables Xt , where t is a parameter (t is usually inter-
preted as the time parameter) running over some index set T . If the index
set T is discrete, then the stochastic process {Xt , t ∈ T } is called a discrete
stochastic process, and if the index set T is continuous, then {Xt , t ∈ T } is
called a continuous stochastic process. In other words, a discrete-time stochas-
tic process for the asset price is one where the asset price can change at some
discrete fixed times while the asset price which follows a continuous-time
stochastic process can change its value at any time. Further, the value taken
by the random variable Xt can be either discrete or continuous, and the cor-
responding stochastic process is called discrete-valued or continuous-valued,
respectively. In reality, stock prices can change only at discrete values and
during periods when the stock exchange is open. However, for simplicity,
we assume continuous-valued, continuous-time stochastic processes for asset
price movement models for most of the later exposition so that analytic tools
in stochastic calculus can be employed.
A Markovian process is a stochastic process that, given the value of Xs ,
the value of Xt , t > s, depends only on Xs but not on the values taken by
Xu , u < s. If the asset prices follow a Markovian process, then only the present
asset prices are relevant for predicting their future values. This Markovian
property of asset prices is consistent with the weak form of market efficiency,
which assumes that the present value of an asset price already impounds all
information in past prices and the particular path taken by the asset price to
reach the present value is irrelevant. If the past history was indeed relevant,
that is, a particular pattern might have a higher chance of price increases,
74 2 Concepts of Financial Economics and Asset Price Dynamics

then investors would bid up the asset price once such a pattern occurs and
the profitable advantage would be eliminated.
We start with the discussion of the discrete random walk model and de-
duce its continuum limit. We obtain the Fokker-Planck equation that governs
the probability density function of the continuous random walk motion. We
then present the formal definition of a Brownian motion and discuss some of
the properties of Brownian processes.

2.3.1 Random walk models

We describe the unrestricted, one-dimensional discrete random walk and con-

sider the continuum limit of the discrete random walk problem to yield the
continuous random walk model. Suppose a particle starts at the origin of the
x-axis and it jumps either to the left or the right of the same length δ. We
define xi to be a random variable which takes the value δ or −δ when the
particle at the ith step moves to the right or the left, respectively. Assume
that the jump probabilities are stationary, that is, these probabilities are the
same at all times. We then write the probabilities as

P [xi = δ] = p, P [xi = −δ] = q, (2.3.1)

where p + q = 1, p and q are independent of i. The individual jumps are

assumed to be independent of each other so that the random variables xi are
independent. This discrete random walk problem is then a discrete Markovian
process (see Fig. 2.8).
Define the discrete sum process

Xn = x 1 + x 2 + · · · + x n , (2.3.2)

which gives the position of the particle at the nth step. Since the expected
value of xi is

E[xi] = δp − δq = (p − q)δ, i = 1, 2, · · ·, n, (2.3.3a)

therefore " #
X
n X
n
E[Xn] = E xi = E[xi] = (p − q)δn. (2.3.3b)
i=1 i=1

As xi ’s are independent, we have

var(Xn ) = n var(xi). (2.3.3c)

The variance of xi is

var(xi ) = [δ 2 p + (−δ)2 q] − (E[xi])2 = δ 2 − (p − q)2δ 2 = 4pqδ 2 (2.3.4a)

so that
var(Xn ) = 4pqδ 2n. (2.3.4b)
 2.3 Asset price dynamics and stochastic processes 75

We call Xn+k −Xn an increment of the discrete random walk model. Since Xn
is a sum process of independent and identically distributed (iid) random vari-
ables, it observes the properties of stationary and independent increments.

q p

0 ( k − 1)δ kδ ( k + 1)δ

Fig. 2.8 A pictorial representation of the discrete random

walk model. Suppose the particle is at position x = kδ
after i − 1 steps, |k| ≤ i − 1. In the ith step, it moves to
the right with probability p or the left with probability q.

Continuum limit
Next, we take the continuum limit of infinitesimally small step size of the
above discrete model to yield the continuous random walk model. Suppose
there are r steps per unit time, then according to Eqs. (2.3.3b, 2.3.4b), the
mean displacement of the particle per unit time µ is (p−q)δr and the variance
of the observed displacement around the mean position per unit time σ2 is
4pqδ 2r. Let λ = 1/r, which is the time interval between two successive steps,
and let u(x, t) denote the probability that the particle takes the position x
at time t. Now, we write Xn = x and nλ = t so that

u(x, t) = P [Xn = x] at t = nλ. (2.3.5)

The probability function u(x, t) satisfies the recurrence relation

u(x, t + λ) = pu(x − δ, t) + qu(x + δ, t). (2.3.6)

In the continuum limit, we take δ → 0 and r → ∞ so that λ → 0. Now,

consider the Taylor expansion of relation (2.3.6):
 
∂u 2 ∂u δ2 ∂2u 3
u(x, t)+λ (x, t) + O(λ ) = p u(x, t) − δ (x, t) + (x, t) + O(δ )
∂t ∂x 2 ∂x2
 
∂u δ2 ∂2u 3
+ q u(x, t) + δ (x, t) + (x, t) + O(δ ) , (2.3.7a)
∂x 2 ∂x2
and upon simplification, we obtain
     
∂u δ ∂u 1 δ 2 ∂ 2 u δ3
= (q − p) + + O(λ) + O (q − p) . (2.3.7b)
∂t λ ∂x 2 λ ∂x2 λ
We take the limits δ, λ → 0 such that the mean displacement and variance
per unit time are given by
76 2 Concepts of Financial Economics and Asset Price Dynamics

δ δ2
(p − q) = µ and 4pq = σ2, (2.3.8)
λ λ
where µ and σ2 are finite quantities. One can deduce from detailed asymp-
totic analysis that the probability values p and q should not be infinitesimal
quantities, that is, p = O(1), q = O(1) and p + q = 1. Consequently,
  we can
δ2 δ 2
δ 1
deduce from 4pq = σ2 that = O(1) and = O . Further, from
λ λ λ δ
δ
(p − q) = µ and p + q = 1, the asymptotic expansion of p and q up to O(δ)
λ
must take the following forms:
1 1
p≈ (1 + kδ) and q ≈ (1 − kδ) (2.3.9)
2 2
for some k to be determined. We then have 4pq ≈ 1 and so
δ2
lim = σ2 . (2.3.10)
δ,λ→0 λ

δ µ
Lastly, from (p−q) = µ and condition (2.3.10), one deduces that p−q ≈ 2 δ
λ σ
µ
and so k = 2 . The asymptotic expansion of p and q are then found to be
σ
1 µ  1 µ 
p≈ 1 + 2 δ and q ≈ 1− 2δ . (2.3.11)
2 σ 2 σ
1 1
Note that p → and q → when taking the asymptotic limit δ → 0;
2 2
δ2
otherwise, the drift rate would become infinite. Since = O(1), the last term
  λ
3
δ
in Eq. (2.3.7b) becomes O (q − p) = O(λ). Consequently, by taking the
λ
limits δ, λ → 0 in Eq. (2.3.7b), we obtain the following partial differential
equation
∂u ∂u σ2 ∂ 2 u
= −µ + (2.3.12)
∂t ∂x 2 ∂x2
for the probability density function u(x, t) of the continuous random walk
motion with drift.
The above differential equation is called the forward Fokker-Planck equa-
tion. The drift rate is µ and the diffusion rate is σ2 . In time t, the mean
displacement of the particle is µt and the variance of the observed displace-
ment around the mean position is σ2 t.
From the Central Limit Theorem in probability theory, one can show
that the continnum limit of the probability density of the discrete random
variable Xn defined in Eq. (2.3.2) tends to that of a normal random variable
with the same mean and variance. The probability density function of the
normal random variable X with mean µt and variance σ2 t is given by
 2.3 Asset price dynamics and stochastic processes 77
   
1 (x − µt)2 x − µt
fX (x, t) = √ exp − = n √ , (2.3.13)
2πσ2 t 2σ2 t σ t
where the function
1 2
n(x) = √ e−x /2 (2.3.14a)
2π
is called the standard normal density function. The cumulative normal dis-
tribution function N (x) is defined to be
Z x
1 2
N (x) = √ e−t /2 dt. (2.3.14b)
−∞ 2π
From partial differential equation theory, fX (x, t) satisfies the following initial
value problem
∂u ∂u σ2 ∂ 2 u
+µ − = 0, −∞ < x < ∞, t > 0, (2.3.15)
∂t ∂x 2 ∂x2
with initial condition: u(x, 0+ ) = δ(x), where u(x, 0+ ) signifies lim u(x, t).
t→0+
Here, δ(x) represents the Dirac function.
The above result has the following probabilistic interpretation. Condi-
tional on the event that the particle starts at the position x = 0 ini-
tially, fX (x, t)∆x gives the probability that the particle would stay within
[x, x + ∆x] at some future time t. This is why fX (x, t) is usually called the
transition density function. The initial condition u(x, 0+) = δ(x) indicates
that the particle stays at x = 0 almost surely. Also, the continuous random
walk model inherits the properties of stationary and independent increments
from the discrete random walk model.

2.3.2 Brownian motions

The Brownian motion refers to the ceaseless, irregular random motion of

small particles immersed in a liquid or gas, as observed by R. Brown in
1827. The phenomena can be explained by the perpetual collisions of the
particles with the molecules of the surrounding medium. Brownian motion is
sometimes known as the Wiener stochastic process.
The formal definition of a Brownian motion with drift is presented below.

Definition
The Brownian motion with drift is a stochastic process {X(t); t ≥ 0} with
the following properties:
(i) Every increment X(t + s) − X(s) is normally distributed with mean µt
and variance σ2t; µ and σ are fixed parameters.
(ii) For every t1 < t2 < · · · < tn , the increments X(t2 ) − X(t1 ), · · · , X(tn ) −
X(tn−1 ) are independent random variables with distributions given in
(i).
78 2 Concepts of Financial Economics and Asset Price Dynamics

(iii) X(0) = 0 and the sample paths of X(t) are continuous.

Note that X(t+s)−X(s) is independent of the past history of the random

path, that is, the knowledge of X(τ ) for τ < s has no effect on the probability
distribution for X(t + s) − X(s). This is precisely the Markovian character
of the Brownian motion.

Standard Brownian motion

For the particular case µ = 0 and σ2 = 1, the Brownian motion is called the
standard Brownian motion (or standard Wiener process). The corresponding
probability distribution for the standard Wiener process {Z(t); t ≥ 0} is given
by [see Eq. (2.3.13)]

P [Z(t) ≤ z|Z(t0 ) = z0 ] = P [Z(t) − Z(t0 ) ≤ z − z0 ]

Z z−z0  
1 x2
= p exp − dx
2π(t − t0 ) −∞ 2(t − t0 )
 
z − z0
=N √ .
t − t0
(2.3.16)

Some other properties

(a) E[Z(t)2 ] = var(Z(t)) + E[Z(t)]2 = t. (2.3.17a)
(b) E[Z(t)Z(s)] = min(t, s). (2.3.17b)
To show the result in (b), we assume t > s (without loss of generality)
and consider

E[Z(t)Z(s)] = E[{Z(t) − Z(s)}Z(s) + Z(s)2 ]

= E[{Z(t) − Z(s)}Z(s)] + E[Z(s)2 ]. (2.3.18a)

Since Z(t) − Z(s) and Z(s) are independent and both Z(t) − Z(s) and Z(s)
have zero mean, so

E[Z(t)Z(s)] = E[Z(s)2 ] = s = min(t, s). (2.3.18b)

Overlapping Brownian increments

When t > s, the correlation coefficient ρ between the two overlapping Brow-
nian increments Z(t) and z(s) is given by
r
E[Z(t)Z(s)] s s
ρ= p p = √ = . (2.3.19)
var(Z(t)) var(Z(s)) st t

Since both Z(t) and Z(s) are normally distributed with zero mean and vari-
ance t and s, respectively, the probability distribution of the overlapping
Brownian increments is given by the bivariate normal distribution function.
 2.3 Asset price dynamics and stochastic processes 79
√ √
If we define X1 = Z(t)/ t and X2 = Z(s)/ s, then X1 and X2 become
standard normal random variables. We then have
√ √
P [Z(t) ≤ zt , Z(s) ≤ zs ] = P [X1 ≤ zt / t, X2 ≤ zs / s]
√ √ p
= N2 (zt / t, zs / s; s/t) (2.3.20a)

where the bivariate normal distribution function is given by

Z x2 Z x1
1
N2 (x1, x2; ρ) = p
2
−∞ −∞ 2π 1 − ρ
 2 
ξ − 2ρξ1 ξ2 + ξ22
exp − 1 dξ1dξ2 . (2.3.20b)
2(1 − ρ2 )

Geometric Brownian motion

Let X(t) denote the Brownian motion with drift parameter µ ≥ 0 and vari-
ance parameter σ2 . The stochastic process defined by

Y (t) = eX(t) , t ≥ 0, (2.3.21)

is called the Geometric Brownian motion. Obviously, the value taken by Y (t)
is non-negative. Since X(t) = ln Y (t) is a Brownian motion, by properties (i)
and (ii), we deduce that ln Y (t) − ln Y (0) is normally distributed with mean
Y (t)
µt and variance σ2t. For common usage, is said to be lognormally
Y (0)
distributed. From the density function of X(t) given in Eq. (2.3.13), the
Y (t)
density function of is deduced to be
Y (0)
 
1 (ln y − µt)2
fY (y, t) = √ exp − . (2.3.22)
y 2πσ2t 2σ2 t

The mean of Y (t) conditional on Y (0) = y0 is found to be

E[Y (t)|Y (0) = y0 ]

Z ∞
= y0 yfY (y, t) dy
0
Z ∞  
ex (x − µt)2
= y0 √ exp − dx, x = ln y,
−∞ 2πσ2t 2σ2 t
Z ∞  
1 [x − (µt + σ2 t)]2 − 2µtσ2 t − σ4t2
= y0 √ exp − dx
−∞ 2πσ2t 2σ2 t
 
σ2 t
= y0 exp µt + . (2.3.23a)
2
Similarly, the variance of Y (t) conditional on Y (0) = y0 is found to be
80 2 Concepts of Financial Economics and Asset Price Dynamics

var(Y (t)|Y (0) = y0 )

Z ∞   2
2 2 σ2 t
= y0 y fY (y, t) dy − y0 exp µt +
0 2
Z ∞  2

2 1 [x − (µt + 2σ t)]2 − 4µtσ2 t − 4σ4 t2
= y0 √ exp − dx
−∞ 2πσ2 t 2σ2t
  2 )
σ2 t
− exp µt +
2
= y02 exp(2µt + σ2 t)[exp(σ2 t) − 1]. (2.3.23b)

Further, for every t1 < t2 < · · · < tn, the successive ratios Y (t2)/Y (t1 ), · · ·,
Y (tn )/Y (tn−1) are independent random variables, that is, the percentage
changes over non-overlapping time intervals are independent.

2.4 Stochastic calculus: Ito’s Lemma and Girsanov’s

Theorem

The price of a derivative is a function of the underlying asset price, and the
asset price process is modeled by a stochastic state variable. In order to con-
struct pricing models for derivatives, it is necessary to develop calculus tools
that allow us to perform mathematical operations, like composition, differen-
tiation, integration, etc. on functions of stochastic random variables. In this
section, we define stochastic integrals and stochastic differentials of functions
that involve the Brownian random variables. In particular, we develop the
Ito differentiation rule that computes the differentials of functions of random
variables. The Feynman-Kac representation formula is derived, which gives
a stochastic representation to the solution of a parabolic partial differential
equation. We also discuss the notion of Radon-Nikodym derivatives and the
Girsanov Theorem that effect the change of equivalent probability measures.

2.4.1 Stochastic integrals

Brownian motions are the continuous limit of discrete random walk mod-
els. Intuitively, one may visualize Brownian paths to be continuous (though
the rigorous mathematical proof of the continuity property is not trivial).
However, Brownian paths are seen to be non-smooth; and in fact, they are
not differentiable. The non-differentiability property can be shown easily by
proving the finiteness of the quadratic variation of a Brownian motion. This
stems from the well known result in elementary calculus that differentiability
implies vanishing of the quadratic variation of the function.
 2.4 Stochastic calculus: Ito’s Lemma and Girsanov’s Theorem 81

Quadratic variation of Brownian motions

Suppose we form a partition π of the time interval [0, T ] by the discrete points

0 = t0 < t1 < · · · < tn = T,

and let δtmax = max(tk − tk−1). We write ∆tk = tk − tk−1, and define the
k
corresponding quadratic variation of the standard Brownian motion Z(t) by
X
n
Qπ = [Z(tk ) − Z(tk−1)]2. (2.4.1)
k=1

Next, we show that the quadratic variation of Z(t) over [0, T ] is given by

Q[0,T ] = lim Qπ = T. (2.4.2)

δtmax →0

To prove the above claim, it suffices to show that

lim E[Qπ ] = T and lim var(Qπ − T ) = 0. (2.4.3)

δtmax →0 δtmax →0

First, we consider

E[Qπ ]
Xn
= E[{Z(tk ) − Z(tk−1 )}2]
k=1
Xn
= var(Z(tk ) − Z(tk−1)) since Z(tk ) − Z(tk−1) has zero mean
k=1
= var(Z(tn ) − Z(t0 )) since Z(tk ) − Z(tk−1 ), k = 1, · · · , n are independent
= tn − t0 = T (2.4.4)

so that the first result in Eq. (2.4.3) is established. Next, we consider

" n n
XX
var(Qπ − T ) = E [Z(tk ) − Z(tk−1 )]2 − ∆tk
k=1 `=1
#
{[Z(t` ) − Z(t`−1 )]2 − ∆t`} . (2.4.5a)

Since the increments [Z(tk ) − Z(tk−1)] − ∆tk , k = 1, · · ·, n are independent,

only those terms corresponding to k = ` in the above series survive, so we
have
82 2 Concepts of Financial Economics and Asset Price Dynamics
" #
X
n
 2
2
var(Qπ − T ) = E [Z(tk ) − Z(tk−1 )] − ∆tk
k=1
n
" #
X
= E {Z(tk ) − Z(tk−1 )}4
k=1
n
X  
− 2∆tk E {Z(tk ) − Z(tk−1)}2 + ∆t2k . (2.4.5b)
k=1

Since Z(tk ) − Z(tk−1) is normally distributed with zero mean and variance
∆tk , its fourth order moment is known to be (see Problem 2.24)

E[{Z(tk ) − Z(tk−1 )}4] = 3∆t2k , (2.4.5c)

so
X
n X
n
var(Qπ − T ) = [3∆t2k − 2∆t2k + ∆t2k ] = 2 ∆t2k . (2.4.5d)
k=1 k=1

In taking the limit δtmax → 0, we observe that var(Qπ − T ) → 0, thus we

obtain the second result in Eq. (2.4.3). By virtue of lim var(Qπ − T ) = 0,
n→∞
we say that T is the mean square limit of Qπ .
Remark
1. In general, the quadratic variation of the Brownian motion with variance
rate σ2 over the time interval [t1, t2 ] is given by

Q[t1 ,t2 ] = σ2 (t2 − t1). (2.4.6)

2. If we write dZ(t) = Z(t) − Z(t − dt), where dt → 0, then we can deduce

from the above calculations that

E[dZ(t)2] = dt and var(dZ(t)2 ) = 2 dt2. (2.4.7)

Since dt2 is a higher order infinitesimally small quantity, we may claim

that the random quantity dZ(t)2 converges in the mean square sense to
the deterministic quantity dt.
Definition of stochastic integration
Let f(t) be an arbitrary function of t and Z(t) be the standard Brownian mo-
Z T
tion. First, we consider the definition of the stochastic integral f(t) dZ(t)
0
as a limit of the following partial sums (defined in the usual Riemann-Stieltjes
sense):
Z T Xn
f(t) dZ(t) = lim f(ξk )[Z(tk ) − Z(tk−1)] (2.4.8)
0 n→∞
k=1
 2.4 Stochastic calculus: Ito’s Lemma and Girsanov’s Theorem 83

where the discrete points 0 < t0 < t1 < · · · < tn = T form a partition of the
interval [0, T ] and ξk is some immediate point between tk−1 and tk . The limit
is taken in the mean square sense. Unfortunately, the limit depends on how
the immediate points are chosen. For example, suppose we take f(t) = Z(t)
and choose ξk = αtk + (1 − α)tk−1, 0 < α < 1, for all k. We consider
" n #
X
E Z(ξk )[Z(tk ) − Z(tk−1)
k=1
X
n
= E [Z(ξk )Z(tk ) − Z(ξk )Z(tk−1)]
k=1
Xn
= [min(ξk , tk ) − min(ξk , tk−1)] [see Eq. (2.3.17)]
k=1
Xn n
X
= (ξk − tk−1) = α (tk − tk−1) = αT, (2.4.9)
k=1 k=1

so that the expected value of the stochastic integral depends on the choice of
immediate points.
A function is said to be non-anticipative with respect to the Brownian
motion Z(t) if the value of the function at time t is determined by the path his-
tory of Z(t) up to time t. In finance, the investor’s action is non-anticipative in
nature since he makes the investment decision before the asset prices move.
It seems natural to define the stochastic integration by taking ξk = tk−1
(left-hand point in each sub-interval) so that integration is taken to be non-
anticipatory. The Ito definition of stochastic integral is given by
Z T X n
f(t) dZ(t) = lim f(tk−1 )[Z(tk ) − Z(tk−1)], (2.4.10)
0 n→∞
k=1

where the limit is taken in the mean square sense and f(t) is non-anticipative
with respect to Z(t).
As an example, consider the evaluation of the Ito stochastic integral
Z T
Z(t) dZ(t). A naive evaluation according to the usual integration rule
0
gives
Z T Z T
1 d Z(T )2 − Z(0)2
Z(t) dZ(t) = [Z(t)]2 dt = , (2.4.11a)
0 2 0 dt 2
which unfortunately gives a wrong result (see explanation below). According
to the definition in Eq. (2.4.10), we have
Z T Xn
Z(t) dZ(t) = lim Z(tk−1)[Z(tk ) − Z(tk−1 )]
0 n→∞
k=1
84 2 Concepts of Financial Economics and Asset Price Dynamics

1X
n
= lim ({Z(tk−1) + [Z(tk ) − Z(tk−1)]}2
n→∞ 2
k=1
− Z(tk−1 )2 − [Z(tk ) − Z(tk−1 )]2)
1
= lim [Z(tn )2 − Z(t0 )2 ]
2 n→∞
Xn
1
− lim [Z(tk ) − Z(tk−1 )]2
2 n→∞
k=1
Z(T )2 − Z(0)2 T
= − [by Eq. (2.4.3)]. (2.4.11b)
2 2
Rearranging the terms, we may rewrite the above result as
Z T Z T Z T
d
2 Z(t) dZ(t) + dt = [Z(t)]2 dt, (2.4.12a)
0 0 0 dt
or in differential form,

2Z(t) dZ(t) + dt = d[Z(t)]2. (2.4.12b)

Unlike the usual differential rule, we have the extra term dt. This comes
from the finiteness of the quadratic variation of the Brownian motion, since
√ Xn
|Z(tk ) − Z(tk−1 )| is of order ∆tk and lim [Z(tk ) − Z(tk−1)]2 remains fi-
n→∞
k=1
nite on taking the limit. Apparently, it is necessary to develop new differential
rules that deal with the computation of differentials of stochastic functions.

2.4.2 Ito’s Lemma and stochastic differentials

Once we have defined stochastic integrals, we can give a formal definition of

a class of continuous stochastic processes, called Ito processes. Let Ft be the
natural filtration generated by the standard Brownian motion Z(t) through
the observation of the trajectory of Z(t). Let µ(t) and σ(t) be adapted to Ft
Z T Z T
with |µ(t)| dt < ∞ and σ2 (t) dt < ∞ (almost surely) for all T , then
0 0
the process X(t) defined by
Z t Z t
X(t) = X(0) + µ(s) ds + σ(s) dZ(s), (2.4.13)
0 0

is called an Ito process. The differential form of the above equation is given
as

dX(t) = µ(t) dt + σ(t) dZ(t). (2.4.14)

 2.4 Stochastic calculus: Ito’s Lemma and Girsanov’s Theorem 85

Ito’s Lemma
Suppose f(x, t) is a deterministic twice continuously differentiable function
and the stochastic process Y is defined by Y = f(X, t), where X(t) is gov-
erned by Eq. (2.4.14). How to compute the differential dY (t)? We have seen
the justification why dZ(t)2 converges in the mean square sense to dt [see Eq.
(2.4.7)]. Hence, the second order term dX 2 also contributes to the differential
dY . The Ito formula of computing the differential of the stochastic function
f(X, t) is given by
 
∂f ∂f σ2 (t) ∂ 2 f
dY = (X, t) + µ(t) (X, t) + dt
∂t ∂x 2 ∂x2
∂f
+ σ(t) (X, t) dZ. (2.4.15)
∂x
The regorous proof of the Ito formula is quite technical, so only a heuristic
proof is provided here. We expand ∆Y by the Taylor series up to the second
order terms as follows:
∂f ∂f
∆Y = ∆t + ∆X
∂t  ∂x 
1 ∂2f 2 ∂2f ∂2f 2
+ ∆t + 2 ∆X∆t + ∆X + O(∆X 3 , ∆t3). (2.4.16a)
2 ∂t2 ∂x∂t ∂x2
In the limit ∆X → 0 and ∆t → 0, we apply the multiplication rules where
dZ 2 = dt, dZdt = 0 and dt2 = 0 so that
∂f ∂f σ2(t) ∂ 2 f
dY = dt + dX + dt. (2.4.16b)
∂t ∂x 2 ∂x2
Writing out in full in terms of dZ and dt, we obtain the Ito formula (2.4.15).
As a simple verification, when we apply the Ito formula to f = Z 2 , we
obtain the result in Eq. (2.4.12b) immediately. As an another example, we
consider the stochastic function
2


r− σ2 t+σZ(t)
S(t) = S0 e . (2.4.17)

Suppose we write
 
σ2
X(t) = r − t + σZ(t) (2.4.18a)
2
so that

S(t) = S0 eX(t) . (2.4.18b)

Now, the respective partial derivatives of S are

∂S ∂S ∂2S
= 0, =S and = S. (2.4.19)
∂t ∂X ∂X 2
86 2 Concepts of Financial Economics and Asset Price Dynamics

By the Ito lemma, we obtain

 
σ2 σ2
dS = r − + S dt + σS dZ (2.4.20a)
2 2
or
dS
= r dt + σ dZ. (2.4.20b)
S
Conversely, we observe that S(t) defined in Eq. (2.4.17) is the  solution
σ2
to the stochastic differential equation (2.4.20b). Since E[X(t)] = r − t
2
S(t)
and var(X(t)) = σ2 t, the mean and variance of ln are found to be
  S0
σ2
r− t and σ2 t, respectively.
2
Multi-dimensional version of Ito’s lemma
Suppose f(x1 , · · · , xn, t) is a multi-dimensional twice continuously differen-
tiable function and the stochastic process Yn is defined by
Yn = f(X1 , · · · , Xn, t), (2.4.21a)
where the process Xj (t) follows the Ito process
dXj (t) = µj (t) dt + σj (t) dZj (t), j = 1, 2, · · ·, n. (2.4.21b)
The Brownian motions Zj (t) and Zk (t) are assumed to be correlated with
correlation coefficient ρjk so that dZj dZk = ρjk dt. In a similar manner, we
expand ∆Yn up to the second order term in ∆Xj :
X ∂f n
∂f
∆Yn = (X1 , · · · , Xn, t) ∆t + (X1 , · · · , Xn, t) ∆Xj
∂t ∂xj
j=1

1 XX
n n 2
∂ f
+ (X1 , · · · , Xn, t) ∆Xj ∆Xk
2 ∂xj ∂xk
j=1 k=1

+ O(∆t∆Xj ) + O(∆t2). (2.4.22)

In the limits ∆Xj → 0, j = 1, 2, · · ·, n, and ∆t → 0, we neglect the higher or-
der terms in O(∆t∆Xj ) and O(∆t2) and observe dXj dXk = σj (t)σk (t)ρjk dt.
We then obtain the following multi-dimensional version of the Ito lemma:

∂f Xn
∂f
dYn =  (X1 , · · · , Xn, t) + µj (t) (X1 , · · ·, Xn t)
∂t ∂xj
j=1

1 XX
n n
∂2f
+ σj (t)σk (t)ρjk (X1 , · · · , Xn, t) dt
2 ∂xj ∂xk
j=1 k=1
n
X ∂f
+ σj (t) (X1 , · · ·, Xn , t) dZj . (2.4.23)
j=1
∂xj
 2.4 Stochastic calculus: Ito’s Lemma and Girsanov’s Theorem 87

Feynman-Kac representation formula

Suppose the Ito process X(t) is governed by the stochastic differential equa-
tion

dX(s) = µ(X(s), s) ds + σ(X(s), s) dZ(s), t ≤ s ≤ T, (2.4.24)

with initial condition: X(t) = x. Consider a smooth function F (X(t), t), by

virtue of the Ito lemma, the differential of which is given by
 
∂F ∂F σ2(X, t) ∂ 2 F ∂F
dF = + µ(X, t) + 2
dt + σ dZ. (2.4.25)
∂t ∂X 2 ∂X ∂X
We define the infinitesimal generator A associated with the Ito process X(t)
by
∂ σ2 (X, t) ∂ 2
A = µ(X, t) + . (2.4.26)
∂X 2 ∂X 2
Suppose F satisfies the parabolic partial differential equation
∂F
+ AF = 0 (2.4.27)
∂t
with terminal condition: F (X(T ), T ) = h(X(T )), then dF becomes
∂F
dF = σ dZ. (2.4.28)
∂X
∂F
Supposing that σ is non-anticipative with the Brownian motion Z(t), we
∂X
can express the above stochastic differential form into the following integral
form
Z s
∂F
F (X(s), s) = F (X(t), t) + σ(X(u), u) (X(u), u) dZ(u). (2.4.29)
t ∂X
The stochastic integral can be viewed as a sum of inhomogeneous consec-
utive Gaussian increments with mean zero, hence it has zero conditional
expectation. By taking the conditional expectation and setting s = T and
F (X(T ), T ) = h(X(T )), we then obtain the following Feynman-Kac repre-
sentation formula

F (x, t) = E x,t[h(X(T ))], t < T, (2.4.30)

where F (x, t) satisfies the partial differential equation (2.4.27). The process
X(t) is initialized at the fixed point x at time t and it follows the Ito process
defined in Eq. (2.4.24).
88 2 Concepts of Financial Economics and Asset Price Dynamics

2.4.3 Change of measure: Girsanov’s Theorem

Consider an Ito process defined either in differential form

dX(t) = µ(t) dt + σ(t) dZ(t) (2.4.31a)

or in integral form
Z t Z t
X(t) = X(0) + µ(s) ds + σ(s) dZ(s) (2.4.31b)
0 0
Z t
with non-zero drift term µ(t). We write M (t) = σ(s) dZ(s) and note that
0
Z T
M (T ) = M (t) + σ(s) dZ(s), T > t. (2.4.32)
t

Suppose we take the conditional expectation of M (T ) given the history up

to the time t (denoted by the operator Et), we obtain

Et[M (T )] = M (t) (2.4.33)

since the stochastic integral in Eq. (2.4.32) has zero conditional expectation.
Hence, M (t) is a martingale. However, X(t) is not a martingale if µ(t) is
non-zero.
In Chap. 3, we will consider the valuation of contingent claims under
the risk neutral measure. With respect to the risk neutral measure, the dis-
counted price of the underlying asset becomes a martingale. The valuation
procedure often requires the transformation of an underlying price process
with drift into a martingale, but under a different measure. Such transforma-
tion can be performed effectively by the use of Girsanov’s Theorem. Before
stating the theorem, we define the Radon-Nikodym derivative which relates
the transformation between two equivalent probability measures.
Radon-Nikodym derivatives
Let Z and Ze denote two Gaussian random variables with unit variance and
respective means 0 and µ. We define dP (ξ) and dPe(ξ) as
 
dξ dξ
dP (ξ) = P ξ − <Z<ξ+ = fZ (ξ) dξ (2.4.34a)
2 2
 
dξ e < ξ + dξ = f (ξ) dξ
dPe(ξ) = P ξ − <Z e
Z
(2.4.34b)
2 2
where
1 2 1 2
fZ (ξ) = √ e−ξ /2 and fZe(ξ) = √ e−(ξ−µ) /2 (2.4.35)
2π 2π
 2.4 Stochastic calculus: Ito’s Lemma and Girsanov’s Theorem 89

e respectively. We set f(ξ) equal the ratio

are the density function of Z and Z,
fZe(ξ)/fZ (ξ) so that
µ2
dPe(ξ) = f(ξ) dP (ξ) = e− 2 +µξ
dP (ξ). (2.4.36)

Note that the multiplication of dP (ξ) by f(ξ) transforms one measure dP

to another measure dPe. The two measures correspond to the two separate
Gaussian distributions with the same variance but different means. The two
measures P and Pe are said to be equivalent since dP (ξ) > 0 if and only if
dPe(ξ)
dPe(ξ) > 0. In probability theory, f(ξ) = is called the Radon-Nikodym
dP (ξ)
derivative of the two equivalent probability measures.
Girsanov Theorem
We state without proof a version of the Girsanov Theorem, which is a useful
tool to effect a change of measure on a stochastic process. The application
of the Girsanov Theorem in identifying an equivalent martingale measure for
pricing contingent claims will be demonstrated in Sec. 3.2.
Theorem 2.4
Consider a stochastic process γ(t) which satisfies the Novikov condition:
Rt 1
γ(s)2 ds
E[e 0 2 ] < ∞, (2.4.37)

and consider the Radon-Nikodym derivative:

dPe
= ρ(t) (2.4.38a)
dP
where
Z t Z t 
1 2
ρ(t) = exp −γ(s) dZ(s) − γ(s) ds . (2.4.38b)
0 2 0

Here, Z(t) is a Brownian motion under the measure P (called P -Brownian

motion). Under the measure Pe, the stochastic process
Z t
e = Z(t) +
Z(t) γ(s) ds (2.4.39)
0

is Pe-Brownian motion.

The proof of the Girsanov Theorem can be found in the text by Karatzas
and Shreve (1991). As an illustration, suppose we take γ(t) to be a constant
so that the Radon-Nikodym derivative is
 
dPe γ2T
= exp −γZ(T ) − . (2.4.40)
dP 2
90 2 Concepts of Financial Economics and Asset Price Dynamics

Recall that if X is the normal random variable with mean µ and variance σ2,
then the expectation of exp(αX) is given by [see Problem 2.24]
 
α2 σ 2
E[exp(αX)] = exp αµ + , for any α. (2.4.41)
2
e ) = Z(T ) + γT , we would
Let Z(T ) be P -Brownian motion and consider Z(T
e e
like to show that Z(T ) is P -Brownian motion. By virtue of Eq. (2.4.41),
2

e ))] = exp γ T
it suffices to show that EPe[exp(γ Z(T , thus verifying that
2
e ) has zero mean and variance T under Pe. We consider
Z(T
e ))] = E [exp(γZ(T ) + γ 2 T ]
EPe[exp(γ Z(T e"
P
#
dPe 2
= EP exp(γZ(T ) + γ T )
dP
   
γ2T
= EP exp γZ(T ) − exp(γZ(T ) + γ 2 T )
2
 2 
γ T
= exp , (2.4.42)
2
e
e ) is Pe -Brownian motion. Note that the factor dP is included when
hence Z(T
dP
we effect the change of probability measure from Pe to P in the expectation
calculations.

2.5 Problems

2.1 Show that a dominant trading strategy exists if and only if there exists
a trading strategy satisfying V0 < 0 and V1 (ω) ≥ 0 for all ω ∈ Ω.
Hint: Consider the dominant trading strategy H = (h0 h1 · · · hM )T
satisfying V0 = 0 and V1 (ω) > 0 for all ω ∈ Ω. Take G∗min =
min G∗ (ω) > 0 and define a new trading strategy with b hm =
ω
M
X
hm , m = 1, · · · , M and b
h0 = −G∗min − ∗
hm S m (0).
m=1

2.2 Consider a portfolio with one risky security and the riskfree security.
Suppose the price of the risky asset at time 0 is 4 and the possible
values of the t = 1 price are 1.1, 2.2 and 3.3 (3 possible states of the
world at the end of a single trading period). Let the riskfree interest
rate r be 0.1 and take the price of the riskfree security at t = 0 to be
unity.
 2.5 Problems 91

(a) Show that the trading strategy: h0 = 4 and h1 = −1 is a domi-

nant trading strategy that starts with zero wealth and ends with
positive wealth with certainty.
(b) Find the discounted gain G∗ over the single trading period.
(c) Find a trading strategy that starts with negative wealth and ends
with non-negative wealth with certainty.
2.3 Show that if the law of one price does not hold, then every payoff in
the asset span can be bought at any price.
2.4 Construct a securities model such that it satisfies the law of one price
but admits a dominant trading strategy.
2.5 Define the pricing functional F (x) on the asset span S by F (x) = {y :
y = S∗ (0)h for some h such that x = S ∗ (1)h, where x ∈ S}. Show that
if the law of one price holds, then F is a linear functional.
2.6 Given the discounted terminal payoff matrix
 
1 3 1
S ∗ (1) =  2 1 3 ,
3 2 4

and the current discounted price vector S∗ (0) = (1 2 4), find the
state price of the Arrow security with discounted payoff ek , k = 1, 2, 3.
2.7 Construct a securities model with 2 risky securities and riskfree security
and 3 possible states of world such that the law of one price holds but
there are dominant trading strategies.
2.8 Show that if there exists a dominant trading strategy, then there ex-
ists an arbitrage opportunity. Outline the thought that underlies the
construction of a securities model such that there exists an arbitrary
opportunity but dominant trading strategy does not exist.
2.9 Show that h is an arbitrage if and only if the discounted gain G∗ satisfies
(i) G∗ ≥ 0 and (ii) EG∗ > 0.
2.10 Suppose a betting game has 3 possible outcomes. If a gambler bets on
outcome i, then he receives a net gain of di dollars for one dollar betted,
i = 1, 2, 3. The payoff matrix thus takes the form (discounting is not
required in a betting game)
 
d1 + 1 0 0
S(1; Ω) =  0 d2 + 1 0 .
0 0 d3 + 1
Find the condition on di such that a risk neutral probability measure
exists for the above betting game (visualized as an investment model).
92 2 Concepts of Financial Economics and Asset Price Dynamics

2.11 Suppose the set of risk neutral measures for a given securities model
is non-empty. Based on the property that a contingent claim X is at-
tainable if and only if EQ [X/S0(1)] is invariant under any risk neutral
measure Q, show that a securities model is complete if and only if there
exists a unique risk neutral measure.
2.12 Consider the following securities model with discounted payoffs of the
securities at t = 1 given by the augmented payoff matrix
 
1 2 3 4
Sb∗ (1; Ω) =  1 3 4 5  ,
1 5 6 7
where the first column gives the discounted payoff of the riskfree se-
curity. Let the augmented initial price vector S b ∗ (0) be (1 3 5 9).
Show that the law of one price does not hold for this securities
 model.
6
Show that the contingent claim with discounted payoff  8  is at-
12
tainable and find the set of all possible trading securities that generate
the payoff. Can we find the price at t = 0 of this contingent claim?
2.13 Let P be the true probability measure, where P (ω) denotes the actual
probability that the state ω occurs. Define the state price density by
the random variable L(ω) = Q(ω)/P (ω), where Q is a risk neutral
measure. Use Rm to denote the return of the risky security m, where
Rm = [Sm (1) − Sm (0)]/Sm (0), m = 1, · · ·, M . Show that EQ [Rm ] =
r, m = 1, · · ·, M , where r is the interest over one period. Let EP [Rm]
denote the expectation of Rm under the actual probability measure P ,
show that
EP [Rm ] − r = −cov(Rm , L),
where cov denotes the covariance operator.
2.14 Suppose u > d > R in the discrete binomial model. Show that an
investor can lock in a riskless profit by borrowing cash as much as
possible to purchase the asset, and selling the asset after one period and
returning the loan. When R > u > d, what should be the corresponding
strategy in order to take arbitrage?
2.15 We can also derive the binomial formula using the riskless hedging prin-
ciple (see Sec. 3.1.1). Suppose we have a call which is one period from
expiry and would like to create a perfectly hedged portfolio with a long
position of one unit of the underlying asset and a short position of m
units of call. Let cu and cd denote the payoff of the call at expiry cor-
responding to the upward and downward movement of the asset price,
respectively. Show that the number of calls to be sold short in the
portfolio should be
 2.5 Problems 93

S(u − d)
m=
cu − cd
in order that the portfolio is perfectly hedged. The hedged portfolio
should earn the risk-free interest rate. Let R denote the growth factor
of the value of a perfectly hedged portfolio in one period. Show that
the binomial option pricing formula for the call as deduced from the
riskless hedging principle is given by
pcu + (1 − p)cd R−d
c= where p= .
R u−d

2.16 Let Πu and Πd denote the state prices corresponding to the states of
asset value going up and going down, respectively. The state prices can
also be interpreted as state contingent discount rates. If no arbitrage
opportunities are available, then all securities (including the bond, the
asset and the call option) must have returns with the same state con-
tingent discount rates Πu and Πd . Hence, the respective relations for
the bond price, asset price and call option value with Πu and Πd are
given by
1 = Πu R + Πd R
S = ΠuuS + Πd dS
c = Πucu + Πd cd .
By solving for Πu and Πd from the first two equations and substituting
the solutions into the third equation, show that the binomial call price
formula over one period is given by
pcu + (1 − p)cd R−d
c= where p= .
R u−d

2.17 Consider the sample space Ω = {−3, −2, −1, 1, 2, 3} and the algebra
F = {φ, {−3, −2}, {−1, 1}, {2, 3}, {−3, −2, −1, 1}, {−3, −2, 2, 3}, {−1,
1, 2, 3}, Ω}. For each of the following random variables, determine
whether it is F-measurable:
(i) X(ω) = ω2 , (ii) X(ω) = max(ω, 2).
Find a random variable that is F-measurable.
2.18 Let X be a random variable defined on (Ω, F, P ) and F1 ⊂ F2 are
sub-algebras of F. Prove the following properties on conditional expec-
tations:
(a) E[XIB ] = E[IB E[X|F]] for all B ∈ F,
(b) E[E[X|F1]|F2] = E[X|F1].
2.19 Let X = {Xt; t = 0, 1, · · ·, T } be a stochastic process adapted to the
filtration F = {Ft; t = 0, 1, · · ·, T }. Show that X is a martingale if and
only if Xt = E[XT |Ft] = 0, t = 0, 1, · · ·, T − 1. Does the property:
E[Xt+1 − Xt |Ft] = 0, t = 0, 1, · · ·, T − 1 imply that X is a martingale?
94 2 Concepts of Financial Economics and Asset Price Dynamics

2.20 Consider the binomial experiment with probability of success p, 0 <

p < 1. We let Nk denote the number of successes after k independent
trials. Define the discrete process Yk by Nk − kp, the excess number of
successes above the mean kp. Show that Yk is a martingale.
2.21 Consider the two-period securities model shown in Fig. 2.5. Suppose
the riskless interest rate r violates the restriction r < 0.2, say, r =
0.3. Construct an arbitrage opportunity associated with the securities
model.
2.22 Deduce the price formula for a European put option with terminal
payoff max(X − S, 0) for the n-period binomial model.
Hint: See Eqs. (2.2.33-35).
2.23 Consider the continuous random walk model discussed in Sec. 2.3.1 [see
Eq. (2.3.13)]. Suppose the particle starts initially at x = a0 , find the
probability that the particle stays above x = a at time t. Express your
answer in terms of
Z x
1 2 1 2
n(x) = √ e−x /2 and N (x) = √ e−t /2 dt.
2π 2π −∞

2.24 Let X be a normally distributed random variable with mean µ and vari-
ance σ2 . Show that the higher central moments of the normal random
variable are given by

0, n odd
µn(X; µ) = E[(X − µ)n ] =
(n − 1)(n − 3) · · · 3· 1σn, n even.

For the lognormal random variable Z = eX , show that the correspond-

ing higher moments are
 
n2 σ 2
µn (Z; 0) = E(Z n ) = exp nµ + , n = 1, 2, · · ·.
2

Hint: See Eqs. (2.3.21-22).

2.25 Suppose {X(t), t ≥ 0} is the standard Brownian motion, its correspond-
ing reflected Brownian motion is defined by
Y (t) = |X(t)| , t ≥ 0.
Show that Y (t) is also Markovian and its mean and variance are re-
spectively r
2t
E[Y ] =
π
and  
2
var(Y ) = 1 − t.
π
 2.5 Problems 95

2.26 Suppose Z(t) is a standard Brownian process, show that the following
processes defined by

X1 (t) = kZ(t/k2 ), k > 0



1
X2 (t) = tZ t for t > 0
0 for t = 0
and
X3 (t) = Z(t + h) − Z(h), h>0
are also Brownian processes.
Hint: To show that Xi (t) is a Brownian process, i = 1, 2, 3, it is nec-
essary to show that

Xi (t + s) − Xi (s)

is normally distributed with zero mean, and

E[[Xi (t + s) − Xi (s)]2 ] = t.

Also, the increments over disjoint time intervals are indepen-

dent, and Xi (t) is continuous at t = 0. For X2 (t), we need to
establish   
X(t) 
P lim 
= 0x(0) = 0 = 1.
t→∞ t

2.27 Consider the Brownian motion with drift defined by

X(t) = µt + σZ(t), X(0) = 0, Z(t) is the standard Brownian motion,

find E[X(t)|X(t0 )], var(X(t)|X(t0 )) and cov(X(t1 ), X(t2 )).

2.28 Assume that the price of an asset follows the Geometric Brownian mo-
tion with expected return of 10% per annum and a volatility of 20% per
annum. Suppose the asset price at present is $100, find the expected
value and variance of the asset price half a year from now and its 90%
confidence limits.
2.29 Let Z(t) denote the standard Brownian motion. Show that
(a) dZ(t)2+n = 0, for any positive integer n,
Z t1
1
(b) Z(t)n dZ(t) = [Z(t1 )n+1 − Z(t0 )n+1 ]
t0 n+1
Z
n t1
− Z(t)n−1 dt,
2 t0
for any positive integer n,
(c) E[Z 4 (t)] = 3t2,
2
(d) E[eαZ(t)] = eα t/2.
96 2 Concepts of Financial Economics and Asset Price Dynamics

2.30 Let the stochastic process X(t), t ≥ 0, be defined by

Z t
X(t) = eα(t−u) dZ(u),
0

where Z(t) is the standard Wiener process. Show that

eα(s+t) − eα|s−t|
cov(X(s), X(t)) = , s ≥ 0, t ≥ 0.
2α

2.31 Let Z(t), t ≥ 0, be the standard Wiener process, f(t) and g(t) are
differentiable functions over [a, b]. Show that
"Z Z b #
b
0 0
E f (t)[Z(t) − Z(a)] dt g (t)[Z(t) − Z(a)] dt
a a
Z b
= [f(b) − f(t)][g(b) − g(t)] dt.
a

Hint: Interchange the order of expectation and integration, and observe

E[[Z(t) − Z(a)][Z(s) − Z(a)]] = min(t, s) − a.

2.32 Show that

Z T
σ [Z(u) − Z(t)] du
t

has zero mean and variance σ2 (T − t)3/3.

Hint: Consider
Z T !
var [Z(u) − Z(t)] du
t
"Z Z #
T T
=E [Z(u) − Z(t)][Z(v) − Z(t)] dudv
t t
Z T Z T
= E[{Z(u) − Z(t)}{Z(v) − Z(t)}] dudv
t t
Z T Z T
= [min(u, v) − t] dudv.
t t

2.33 Show that

N2 (a, b; ρ) + N2 (a, −b; −ρ) = N (a),
where N and N2 denote the standard univariate and bivariate normal
distribution functions, respectively. Also, show that
 2.5 Problems 97
Z !
a
b − ρx
N2 (a, b; ρ) = n(x)N p dx,
−∞ 1 − ρ2

where n(x) is the standard normal density function.

2.34 Suppose the stochastic variables S1 and S2 follow the Geometric Brow-
nian processes where
dSi
= µi dt + σi dZi , i = 1, 2.
Si
Let ρ12 denote the correlation coefficient between the Wiener processes
dZ1 and dZ2 . Let f = S1 S2 , show that f also follows the Geometric
Brownian process of the form
df
= µ dt + σ dZ
f
where µ = µ1 + µ2 + ρ12 σ1σ2 and σ2 = σ12 + σ22 + 2ρ12 σ1σ2. Similarly,
S1
let g = , show that
S2
dg
e dt + σ
=µ e dz
g
e = µ1 − µ2 − ρ12 σ1σ2 + σ22 and σ
where µ e2 = σ12 + σ22 − 2ρ12σ1 σ2.
2.35 Suppose the function F (x, t) satisfies

∂F ∂F σ2 (x, t) ∂ 2 F
+ µ(x, t) + − rF = 0
∂t ∂x 2 ∂x2
with terminal condition: F (X(T ), T ) = h(X(T )). Show that

F (x, t) = e−r(T −t) Et[h(X(T ))|X(t) = x], t < T,

where X(t) follows the Ito process

dX(t) = µ(X(t), t) dt + σ(X(t), t) dZ(t).

2.36 Define the discrete random variable X by

(
2 if ω = ω1
X(ω) = 3 if ω = ω2 ,
4 if ω = ω3
where the sample space Ω = {ω1, ω2, ω3}, P [ω1] = P [ω2] = P [ω3] =
1/3. Find a new probability measure Pe such that the mean becomes
EPe[X] = 3.5 while the variance remains unchanged. Is Pe unique?
98 2 Concepts of Financial Economics and Asset Price Dynamics

2.37 Given that S(t) is a Geometric Brownian motion which follows

dS
= µ dt + σ dZ
S
where Z is P -Brownian motion. Find another measure Pe by specifying
dPe
the Radon-Nikodym derivative such that S(t) is governed by
dP
dS
= µ0 dt + σ dZe
S
under the measure Pe , where Ze is Pe-Brownian motion and µ0 is the new
drift rate.