Mathematical Finance

Lecture Note2
2. Risky Assets

Goal: modelling the price development of risky

assets over several periods.
2.1 Dynamics of stock prices

Notations:

• S(t) = price of the stock at time t

 S(0) :the current stock price, a positive
constant
 S(t) for t > 0: future stock price, a pos-
itive random variable
 the family of random variables {S(t)}t>0
is a stochastic process.
• t = time, in this chapter, t = 1, 2, ..., n

• Example:

in probabilistic terms: 4 scenarios ω1, ω2, ω3, ω4

with two random variables S(1) and S(2):
 At t = 1, only scenarios {ω1, ω2} will
occur in which case
S(1)(ω1) = S(1)(ω2) = 110
or only scenarios {ω3, ω4} will occur in
which case
S(1)(ω3) = S(1)(ω4) = 95
  At t = 2, only one of the scenarios ω1, ω2, ω3,
ω4 will occur. So,
S(2)(ω1) = 120, S(2)(ω2) = 105,
S(2)(ω3) = 100, S(2)(ω4) = 85

Formulae for the returns:

• Return over the time interval [n, m]:

S(m) − S(n)
K(n, m) :=
S(n)

• Return over a single time step [n − 1, n]:

S(n) − S(n − 1)
K(n) := K(n − 1, n) =
S(n − 1)

• The relation between K(n, m) and single

time step returns:
Note that 1 + K(n) = S(n−1)
S(n)
which implies
     
1 + K(n + 1) 1 + K(n + 2) ... 1 + K(m)

= S(n+1) S(n+2)
... S(m)
S(n) S(n+1) S(m−1)

= S(m)
S(n)
= 1 + K(n, m)
So,
   
1 + K(n, m) = 1 + K(n + 1) ... 1 + K(m)

Formulae for the logarithmic returns:

• logarithmic return over the time interval

[n, m]:
 S(m) 
k(n, m) := ln
S(n)

• one-step logarithmic return:

S(n) 
k(n) := k(n − 1, n) = ln


S(n − 1)
 • relation between k(n, m) and single time
steps:
 Note
 that
   
k(n + 1) + k(n + 2) + ... + k(m)

= ln S(n+1)
ln
S(m)
   
S(n)
+ ... + S(m−1)

= ln S(m)
 
S(n)
= k(n, m)
So,
     
k(n, m) = k(n) + k(n + 1) + ... + k(m)

Example 1) Consider the price of a risky asset

follows the following path.

Calculate the returns at t = 1 and t = 2 in

each scenario.
Solution:

Scenarios K(1) K(2) K(0, 2)

ω1 10% 9.09% 20%
ω2 10% -4.55% 5%
ω3 -5% 5.26% 0%
ω4 -5% -10.53% -15%
Note that K(0, 2) 6= K(1) + K(2),
but K(0, 2) = (1 + K(1))(1 + K(2)) − 1
Example 2) Calculate the logarithmic returns
in the preceding example.
Solution:
 Scenarios k(1) k(2) k(0, 2)
ω1 9.53% 8.7% 18.23%
ω2 9.53% -4.65% 4.88%
ω3 -5.13% 5.13% 0%
ω4 -5.13% -11.12% -16.25%
Note that k(0, 2) = k(1) + k(2)
Example 3) Consider a two-period model with
three scenarios such that K(2) is either 10%
or 0%, and K(0, 2) takes the values 21%, 10%
and 0%. Find a possible distribution of K(1)
that takes only two dierent values.
Solution: The random variables K(1), K(2)
and K(0, 2) are related by
1 + K(0, 2) = (1 + K(1))(1 + K(2)). The pos-
sible congurations are
Scenarios K(1) K(2) K(0, 2)
ω1 21% 0% 21%
ω2 0% 10% 10%
ω3 0% 0% 0%
 Scenarios K(1) K(2) K(0, 2)
ω1 10% 10% 21%
ω2 0% 10% 10%
ω3 0% 0% 0%
Scenarios K(1) K(2) K(0, 2)
ω1 10% 10% 21%
ω2 10% 0% 10%
ω3 0% 0% 0%
Scenarios K(1) K(2) K(0, 2)
ω1 10% 10% 21%
ω2 10% 0% 10%
ω3 -9.09% 10% 0%
Formula for the expected return:

• The expected return is E(K(n, m)).

• If the one-step return K(n + 1),..., K(m)

are independent, then
 1 + E(K(n, m))

= [1 + E(K(n + 1))] × (3.1)

[1 + E(K(n + 2))]...[1 + E(K(m))]

Example 1) Assuming that all four scenarios

have equal probability in the preceding Exam-
ple 1 (i.e.P (ω1) = P (ω2) = P (ω3) = P (ω4) =
0.25), calculate the expected returns.

Solution: Recall that

So, E(K(1)) = 12 10% + 12 (−5%) = 2.5%

E(K(2)) ≈ 1
4 9.09% + 1 (−4.55%) + 1 5.26%+
4 4

1 (−10.53%) ≈ −0.18%
+4

E(K(0, 2)) = 1
4 20% + 1 5% + 1 0% + 1 (−15%)
4 4 4

= 2.5%.

Note that [1+E(K(1))][1+E(K(2))] ≈ 1.0231

which is not the same as 1 + E(K(0, 2)), be-
cause K(1) and K(2) are not independent.
Example 2) Assume that K(1), K(2), K(3), K(4)
denote quarterly returns and are independent
and identically distributed. What is the ex-
pected full year return if the expected return
for the rst half of the year is 5%?
Solution: Since the quarterly returns are iden-
tically distributed, we have
E(K(1)) = E(K(2)) = E(K(3)) = E(K(4)).
By independence, we thus obtain
1 + E(K(0, 2)) = (1 + E(K(1)))2,

1 + E(K(0, 4)) = (1 + E(K(1)))4.

Setting E(K(0, 2)) = 5% yields
√
E(K(1)) = 1.05 − 1 ≈ 2.469%.
Hence, we conclude
E(K(0, 4)) = [1 + E(K(1))]4 − 1 = 10.25%.
2.2 Binomial tree model

Model Assumptions:

• All the one-step returns K(1), ..., K(n) have

the same distribution
u with probability p


K(n) = (3.2)
d with probability 1 − p
where −1 < d < u and p ∈ (0, 1).

• The one-step returns K(n), n = 1, 2, ... are

independent.

• The risk-free asset has a constant one-step

return r, which satises d < r < u.

Note:
• (3.2) implies S(n) goes up or down by a
factor of either 1 + u or 1 + d at each step.

• −1 < d < u implies that starting with a

positive S(0), stock prices remain positive.

• Independent returns implies previous returns

do not aect the current return.

• We say that the market model is arbitrage

free if there do not exist any arbitrage op-
portunities.

• d < r < u implies no-arbitrage condition;

we will see this in Chapter 3, but already
note this is equivalent to Proposition 1.1.
black: prices S(n) with S(0) = 1; for the general case multiply by

S(0)

red: probabilities for the up/down movement of S(t) from t = n−1

to t = n, n = 1, 2, ...

green: probabilities that S(n) takes the particular values from t=0

to t=n

Note:

• In general for j = 0, 1, ..., n,

S(n) = S(0)(1 + u)j (1 + d)n−j
 with probability n j
.
 
p (1 − p) n−j
j

• The number U of upward movements has

a binomial distribution:
n
P (U = j) = pj (1 − p)n−j
j

Example 1) Assume a binomial tree model

such that the stock prices are S(1) = 110 and
S(2) = 121 in the best case scenario. Find
S(0).

Solution: The values of S(0) and u are the

solution of
S(0)(1 + u) = 110,

S(0)(1 + u)2 = 121

This yields 1+u = 121
110 = 1.1 and hence S(0) =
110 = 100.
1.1
Example 2) Assume that S(2) takes the val-
ues 121, 99 and 81. Find S(0), u and d?

Solution: The values of S(0), u and d are the

solution to
S(0)(1 + u)2 = 121,

S(0)(1 + u)(1 + d) = 99,

S(0)(1 + d)2 = 81.

Because the middle equation is redundant (square
root of the product of the rst and last equa-
tions), the system is equivalent to
S(0)(1 + u)2 = 121,

S(0)(1 + d)2 = 81,

which are only two equations for three un-
knowns so that S(0), u and d are not uniquely
determined. They are related by
11 9
u=q − 1, d=q − 1.
S(0) S(0)
Formula for the expected price:

Proposition 2.1 The expected price E(S(n))

is given by
E(S(n)) = S(0)[1 + E(K(1))]n.

Proof: 1) Rewrite E(S(n)) in terms of E(K(0, n))

S(n) − S(0)
K(0, n) =
S(0)
which implies
S(n) = S(0)(1 + K(0, n))
so that
E[S(n)] = S(0)(1 + E[K(0, n)])

2) By independence and equation 3.1,

1+E(K(0, n)) = [1+E(K(1))]...[1+E(K(n))].
3) Because K(j), j = 1, 2, ..., n are identically
distributed, we have E(K(j)) = E(K(1)) for
all j .
Combining 1) − 3) yields the result.
Stock investment is risky and a typical investor
is risk averse =⇒ usually E(K(1)) > r.
Risk-neutral probability:

• Denition: A probability with strictly

P∗
positive weights in all scenarios is called
risk neutral if E∗[K(n)] = r, where E∗[K(n)]
is the expected value of K(n) under the
probability P∗.

• Properties:
 In general, P∗ 6= market probability P
  P∗ is an abstract mathematical object
 P∗ is of great importance in pricing op-
tions

• Calculation: Denote by p∗the probability

under P∗ for an upward movement:
p∗ = P∗(K(n) = u)

E∗[K(n)] = uP∗(K(n) = u)+dP∗(K(n) = d)

= up∗ + d(1 − p∗) = d + p∗(u − d)

Solving E∗[K(n)] = r yields
r−d
p∗ = Works only for Binomial tree model
u−d
Note that the assumption d < r < u implies
0 < p∗ < 1.
• Interpretation: the risk-neutral probabil-
ity balances the upward and downward move-
ments so that their weighted average equals
r.

• In more detail:
 consider points d < r < u as coordinates
on the real axis,
 the risk-neutral probability puts masses
p∗ and 1 − p∗ on u and d, respectively,
such that the centre of mass is at r.
Conditional expectation:

• Consider a two-period model with

S(0) = 100, u = 0.2, d = 0 and r = 0.05,
r−d = 1
p∗ = u−d 4

Note that E∗[S(n)] = (1 + r)nS(0) by the

denition of risk-neutral probability.

• Consider the situation at time 1.

1) Assume that S(1) = 120.
The conditional expectation of S(2) given
S(1) = 120 is

E∗[S(2)|S(1)

= 120] 
= 144 · P∗ S(2) = 144 S(1) = 120 +
 
+ 120 · P∗ S(2) = 120 S(1) = 120
1 3
= 144 · + 120 · = 126
4 4

which equals S(1)(1 + r) with S(1) = 120.

2) Assume that S(1) = 100.
The conditional expectation of S(2) given
S(1) = 100 is
E∗[S(2)|S(1)

= 100] 
= 120 · P∗ S(2) = 120 S(1) = 100 +
 
+ 100 · P∗ S(2) = 100 S(1) = 100
1 3
= 120 · + 100 · = 105
4 4

which equals S(1)(1 + r) with S(1) = 100.

 =⇒ We have
E∗[S(2) S(1) = 120] = S(1)(1 + r)
and
E∗[S(2) S(1) = 100] = S(1)(1 + r), which
can be written as
E∗[S(2) S(1)] = S(1)(1 + r)

• In general, let P be a probability, and X, Y

be discrete random variables taking values
x1, ..., xn and y1, ..., ym, respectively. The
conditional expectation of Y given the event
X = xj equals
m
X
E[Y X = xj ] = yiP (Y = yi X = xj )
i=1
where the right-hand side can be seen as a
function of xj :
m
X
f (xj ) := yiP (Y = yi X = xj )
i=1
 Then the conditional expectation of Y given
X is
E(Y |X) = f (X),
which means E(Y |X = xj ) = f (xj ) for all
xj . =⇒ This is prediction of Y assuming
X is known.

Example 1) Roll a pair of fair dice. What is

the prediction for the sum of dice if we know
the number of the rst die?
Solution: Let X = number of the rst die and
Y = sum of both dice. We are interested in
E(Y |X). We derive

E(Y |X = x) = 3.5 + x
for x = 1, 2, ..., 6 so that E(Y |X) = 3.5 + X .
Example 2) What is the predicted number of
ones if we roll a die until we get a 6?
Solution: Let X =number of rolls until we get
a 6,
Y = number of ones until we get a 6.
We are interested in E(Y |X). We rst derive
E(Y |X = 1) = 0,
1 · 1 = 1,
E(Y |X = 2) = 5 5
E(Y |X = 3) = 1
5 · 2 = 2
5
. . .,
5 (x − 1) for x = 1, 2, ...
E(Y |X = x) = 1
so that E(Y |X) = 15 (X − 1).
Example 3) Consider a two-period binomial
tree model with S(0) = 100, u = 0.2, d = −0.1
and r = 0. Let C(2) be the payo of a (Euro-
pean) call option with strike price 100. What
are the values of E∗[C(2)] and E∗[C(2)|S(1)]?
Solution: First p∗ = u−d
r−d = 1/3.
Thus, E∗[C(2)|S(1) = 120] = 13 44 + 23 8 = 20
18 + 20 = 8
E∗[C(2)|S(1) = 90] = 3 3 3

E∗[C(2)] = 1
9 44 + 2 · 2 · 8 + 4 0 = 76
9 9 9
Or alternatively using ,
 
E[Y ] = E E[Y |X]
 
E∗[C(2)] = E∗ E∗[C(2)|S(1)]
= 13 E∗ [C(2)|S(1) = 120] + 2 E [C(2)|S(1) =
3 ∗
90]
=13 20 + 2 2 = 76
3 9
Martingale property:

• The discounted price process S̃(n) is de-

ned by
S̃(n) = S(n)(1 + r)−n

Theorem 2.2: E∗[S̃(n + 1)|S̃(n)] = S̃(n)

i.e. the discounted prices S̃(n) are a mar-
tingale under the risk-neutral probability.
Proof: S̃(n + 1) = S(n + 1)(1 + r)−n−1

(1 + u)S(n)(1 + r)−n−1 = 1+u S̃(n)
= 1+r
(1 + d)S(n)(1 + r)−n−1 = 1+d S̃(n)
1+r
Thus, if S̃(n) = x for any x, then
with P∗ -prob.

 1+u x p∗
1+r
S̃(n + 1) = 1+d

1+r x with P∗ -prob. 1 − p∗
Thus, E∗[S̃(n + 1)|S̃(n)]
= p∗ 1+u
1+r x + (1 − p ∗ ) 1+d
1+r x
x (1 + d + (u − d)p )
= 1+r ∗
x (1 + d + (u − d) r−d ) = x = S̃(n)
= 1+r u−d

Hence, E∗[S̃(n + 1)|S̃(n)] = S̃(n)

• Martingale property:
 For every n, the conditional expectation
of the process at time n+1 equals its
value at time n.
 Mathematical formulation of a fair game:
on average, one does not gain or lose
anything.
2.3 Trinomial tree model
Model assumptions:

• The one-step returns K(n) have all the

same distribution
with prob. p

u

with prob. q

K(n) = s
with prob. 1 − p − q



d
where −1 < d < s < u and 0 < p, q, p+q < 1.

• The one-step returns K(n), n = 1, 2, ... are

independent.

• The risk-free asset has a constant one-step

return r, which satises d < r < u.

• The dierence to the binomial model is

that at each time step, stock prices can
 move up or down in three (rather than two)
ways: by a factor
1 + u, 1 + s, or 1 + d.

• Often, s = 0, meaning a stable price in this

scenario.
Risk-neutral probabilities:

• Denition: Risk-neutral probabilities are

two numbers p∗, q∗ such that 0 < p∗, q∗, p∗ +
q∗ < 1 and E∗[K(n)] = r which is equivalent
to
(u − r)p∗ + (s − r)q∗ + (d − r)(1 − p∗ − q∗) = 0.
As opposed to the binomial tree model,
there is no unique risk-neutral probability
since we have two parameters p∗, q∗ and
only one equation.

• Interpretation:
 consider points d < r, s < u as coordi-
nates on the real axis,
 a risk-neutral probability puts masses p∗, q∗
and 1−p∗ −p∗ on u, s and d, respectively,
such that the centre of mass is at r.
 • To check the positivity condition of the
risk-neutral probabilities, it is enough to
check if p∗, q∗ > 0 and 1 − p∗ − q∗ > 0.

Example 1) What are the risk-neutral proba-

bilities if we have u = 0.2, s = 0.1, d = 0 and
r = 0.05?

Solution: The risk-neutral probabilities satisfy

0.15p∗ + 0.05q∗ − 0.05(1 − p∗ − q∗) = 0
⇐⇒ 0.1q∗ + 0.2p∗ − 0.05 = 0
⇐⇒ q∗ = 0.5 − 2p∗. Moreover, p∗, q∗ > 0 and
1 − p∗ − p∗ > 0
which is equivalent to 0 < p∗ < 0.25.
Hence, all risk-neutral probabilities are given by
P∗(p∗, 0.5 − 2p∗, p∗ + 0.5) for 0 < p∗ < 0.25.

Example 2) Assume that there are three se-

curities:
1) bank account with interest rate r = 0,
2) stock 1 with independent returns K1(n),
3) stock 2 with independent returns K2(n).
At each step, there are three scenarios with
equal probability 1/3. K1(n) takes the values
1/4, 0 and -1/4.
a) Is there a risk-neutral probability P∗ if K2(n)
takes the values 1, -1/4 and -1/2 (with the
same order of scenarios as for K1(n)?
Solution: Let E∗[K1(n)] = r = 0 and
E∗[K2(n)] = r = 0.

 1 p + 0 · q − 1 (1 − p − q ) = 0
∗ ∗ ∗ ∗
⇐⇒ 4 4
1 · p∗ − 1 q∗ − 1 (1 − p∗ − q∗) = 0
4 2
 
− 1 + 1 p + 1 q = 0
⇐⇒ 4 2 ∗ 4 ∗
− 1 + 3 p∗ + 1 q∗ = 0
2 2 4

1 , q = 2 − 6p = 1 .
⇐⇒ p∗ = 4 ∗ ∗ 2

As p∗ > 0, q∗ > 0 and p∗ + q∗ = 34 < 1, we have

a unique risk neutral probability P∗ = ( 14 , 12 , 14 ).
b) Is there a risk-neutral probability P∗ if K2(n)
takes the values 1, 1/2 and -1/2?
Solution: Let E∗[K1(n)] = r = 0 and
E∗[K2(n)] = r = 0.

 1 p + 0 · q − 1 (1 − p − q ) = 0
∗ ∗ ∗ ∗
⇐⇒ 4 4
1 · p∗ + 1 q∗ − 1 (1 − p∗ − q∗) = 0
2 2

− 1 + 1 p + 1 q = 0
⇐⇒ 4 2 ∗ 4 ∗
− 1 + 3 p∗ + 1q∗ = 0
2 2
⇐⇒ p∗ = 1, q∗ = 1
2 − 3 p = −1.
2 ∗ Since q∗ < 0,
there does not exist a risk-neutral probability.
c) Is there a risk-neutral probability P∗ if K2(n)
takes the values 1, 1/4 and -1/2?
Solution: Let E∗[K1(n)] = r = 0 and
E∗[K2(n)] = r = 0.

 1 p + 0 · q − 1 (1 − p − q ) = 0
∗ ∗ ∗ ∗
⇐⇒ 4 4
1 · p∗ + 1 q∗ − 1 (1 − p∗ − q∗) = 0
4 2

− 1 + 1 p + 1 q = 0
⇐⇒ 4 2 ∗ 4 ∗
− 1 + 3 p∗ + 3 q∗ = 0
2 2 4

1
⇐⇒ 0 = 4 which is a contradiction. Hence,
there does not exist a risk-neutral probability.
d) Is there a risk-neutral probability P∗ if K2(n)
takes the values 1/2, 0 and -1/2?
Solution: Let E∗[K1(n)] = r = 0 and
E∗[K2(n)] = r = 0.

 1 p + 0 · q − 1 (1 − p − q ) = 0
⇐⇒ 14 ∗ ∗ 4 ∗ ∗
 · p∗ + 0 · q∗ − 1 (1 − p∗ − q∗) = 0
2 2
These are identical equations. So,
−14 p ∗ + 1p + 1q = 0
2 ∗ 4 ∗
⇐⇒ q∗ = 1 − 2p∗
As p∗ > 0, q∗ > 0 and p∗ + q∗ < 1, we have
p∗ > 0, 1 − 2p∗ > 0 which implies 0 < p∗ < 1/2.

There are innitely many risk-neutral probabil-

ities given by P∗(t, 1 − 2t, t) for 0 < t < 1/2.
Example 3) Consider a model with
r = 0, S1(0) = 10, S2(0) = 1, and
if ω1


s
if 0.5 if ω , ω

 
11

ω2 1 2
S1(1) = , S2(1) =

 13 if ω3 1.5 if ω , ω
3 4
12 if ω



4
for some s > 0. Find the risk-neutral probabil-
ities?
Solution: The returns are given by
if ω1

 s−10
 10
if ω2


S1(1) − S1(0) 0.1

K1(n) = =
S1(0) 
0.3 if ω3
if ω4


0.2


if ω1, ω2

S2(1) − S2(0) −0.5
K2(1) = =
S2(0) 0.5 if ω3, ω4
A risk-neutral probability is of the form
P∗(p∗, q∗, r∗, 1 − p∗ − q∗ − r∗) where
P∗(ω1) = p∗, P∗(ω2) = q∗, and P∗(ω3) = r∗.
Let E∗[K1(n)] = r = 0 and E∗[K2(n)] = r = 0.

( s − 1)p + 0.1q + 0.3r


 10 ∗ ∗ ∗

+0.2(1 − p − q − r ) = 0

∗ ∗ ∗
⇐⇒


−0.5p∗ − 0.5q∗ + 0.5r∗+

+0.5(1 − p − q − r ) = 0

∗ ∗ ∗

2 + (s − 12)p − q + r = 0
∗ ∗ ∗
⇐⇒
0.5 − p∗ − q∗ = 0


r = −2 + (12 − s)p + q
∗ ∗ ∗
⇐⇒
q∗ = 0.5 − p∗


r = −1.5 + (11 − s)p
∗ ∗
⇐⇒
q∗ = 0.5 − p∗

p∗ > 0, q∗ > 0 ⇐⇒ 0.5 − p∗ > 0 ⇐⇒ 0.5 > p∗ > 0

r∗ > 0 ⇐⇒ −1.5 + (11 − s)p∗ > 0,
p∗ + q∗ + r∗ < 1 ⇐⇒ 2 + (s − 11)p∗ > 0,

The third condition can hold only if s < 11 so

that the conditions can be written as
1.5 2
0 < p∗ < 0.5, p∗ > , p∗ <
11 − s 11 − s
which implies
1.5 2
< p∗ < min{0.5, }.
11 − s 11 − s

1.5 < 0.5, hence we

This is possible only if 11−s
need 3 < 11 − s and thus s < 8.
In conclusion:

• if s > 8, no risk-neutral probability exists;

• if 0 < s < 8, the risk-neutral probabilities

are given by
 
P∗ = p∗, 0.5−p∗, −1.5+(11−s)p∗, 2+(s−11)p∗

where 1.5 < p <

11−s ∗ min{0.5, 11−s
2 }.
2.4 Continuous-Time Limit

Model assumptions:

• Consider a sequence of binomial tree mod-

els with time step τ = N1 , letting N → ∞
over time interval [0, 1].

• Each upward and downward price move-

ment has 1/2 probability at each step.

• The return is the logarithmic return:

ln(1 + u) , p = 1/2


k(n) = ln 1+K(n) =
 
ln(1 + d) , p = 1/2

• The equivalent continuous compounding risk

free rate r results in a return over a time
step of length τ will be eτ r .
• Let µ and σ be the expectation and stan-
dard deviation of the return on the time
interval [0, 1].

• The logarithmic returns k(1), k(2), . . .,

k(N ) are identically distributed and inde-
pendent so the expectation
µ = E[k(1) + ... + k(N )] = N E[k(n)]
and the variance
σ 2 = Var[k(1) + ... + k(N )] = N Var[k(n)].
These two equations imply that the expec-
tation and standard deviation at each step
are
E[k(n)] = µ/N = µτ
and
√ √
σk(n) = σ/ N = σ τ
We can show that
 √
ln(1 + u) = µτ + σ τ
ln(1 + d) = µτ − σ√τ
Thus
µτ + σ √τ ,

p = 1/2
k(n) = √
µτ − σ τ , p = 1/2

• Dene a sequence of independent random

variables ξ(n) by
 √

ξ(n) =
 τ with p = 1/2
√
− τ with p = 1/2
Thus, we can write
k(n) = µτ + σξ(n).

• Dene w(0) = 0 and

w(n) = ξ(1) + ... + ξ(n)
• Note that ξ(n) = w(n) − w(n − 1).

• Let t = τ n. So we write S(t) and w(t)

instead of S(n) and w(n).
Proposition 2.3
The stock price at time t is given by
S(t) = S(0)eµt+σw(t)

Proof:
S(t) = S(τ n) = S(τ n − τ )ek(n)
= S(τ n − 2τ )ek(n−1)+k(n)
= ...
= S(0)ek(1)+...+k(n−1)+k(n)
= S(0)eµnτ +σ(ξ(1)+...+ξ(n))
= S(0)eµt+σw(t)

• In order to pass to continuous time, 2we use

the approximation of ex ≈ 1 + x + x2 for x
very close to zero.
 Note that
S(nτ + τ )
= ek(n+1)
S(nτ )
1
≈ 1 + k(n + 1) + k(n + 1)2
2
1 2
≈ 1 + µτ + σξ(n + 1) + µτ + σξ(n + 1)
2
1 2
≈ 1 + µτ + σξ(n + 1) + σ τ
2
1 2
= 1 + (µ + σ )τ + σξ(n + 1)
2

So,
1 2
S(nτ + τ ) − S(nτ ) ≈ S(nτ )(µ + σ )τ + S(nτ )σξ(n + 1)
2
Since t = nτ and
ξ(n+1) = w(nτ +τ )−w(nτ ) = w(t+τ )−w(t)
, we have
1 2
S(t + τ ) − S(t) ≈ S(t)(µ + σ )τ + σS(t)(w(t + τ ) − w(t))
2

• Note that the expectation and variance of

w(t) are 0 and t.
• Dene
k(n) − µτ
x(n) := √
σ τ
for n = 1, 2, 3.... Then this is a sequence
of independent and identically distributed
random variables with mean 0 and stan-
dard deviation 1. By the Central Limit
Theorem,
x(1) + x(2) + ... + x(n)
√
n
converges to the standard normal distribu-
tion N (0, 1).
Let t > 0. Because the random walk wN
is only dened at discrete times which are
integer multiples of the step τ = N1 , we
consider wN (tN ), where tN is integer mul-
tiple of τ = N1 nearest to t. Then, clearly,
N tN is an integer for each N and we can
write
p x(1) + x(2) + ... + x(N tN )
wN (tN ) = tN √
N tN
 As N → ∞, we have N tN → ∞ and tN → t
so that
wN (tN ) → W (t)
where W (t) ∼ N (0, t).

• The family {W (t)}t>0 is a Weiner process

or Brownian motion since it is a family
of random variables and it satises
 W (0) = 0.

 W (t) is a continuous path on {t : t > 0}.

 For each t > 0, W (t)
is normally dis-
tributed with mean 0 and variance t.
 For t > s > r > 0, W (t) − W (s) and
W (s) − W (r) are independent.

• As N → ∞, the approximate equation for

t = nτ
1 2
SN (t + τ ) − SN (t) ≈ SN (t)(µ + σ )τ + σSN (t)(wN (t + τ ) − wN (t))
2
 becomes
1 2
dS(t) = S(t)(µ + σ )dt + σS(t)dW (t)
2

and its solution is given by

S(t) = S(0)eµt+σW (t)

• we can make a sample path of the price of

a stock from time t = 0 to time t = 1.

• Indeed, we divide the interval [0, 1] into n

subintervals
0 = t0 < t1 < t2 < ... < tn = 1
of equal length ∆t.

• For each i = 1, ..., n,

S(ti) S(0)eµti+σW (ti)
=
S(ti−1) S(0)eµti−1+σW (ti−1)
 which implies
S(ti) = S(ti−1)eµ(ti−ti−1)+σ[W (ti)−W (ti−1]
∼ S(ti−1)eµ∆t+σN (0,∆t)
√
= S(ti−1 )eµ∆t+σ ∆tN (0,1)

where N (0, 1) is the standard normal dis-

tribution.

• The following is a simulation of stock price

from t = 0 to t = 1 with
(i) initial stock price S0=S(0) = 50
(ii) Expected return µ = 0.1 and risk σ =
0.2

(iii) Deltat=∆t=0.001
Bibliography

Marek Capinski and Tomasz Zastawniak, Math-

ematics for Finance, An Introduction to Finan-
cial Engineering
John C. Hull, Options, Futures, and Other
derivatives, 9th Edition, Pearson