The Infinite Actuary’s

Chin Textbook Problems for the

QFI Quant Exam

Zak Fischer, FSA CERA

Ken Qian, FSA

2023
QFI Quant Online Seminar – Chin Practice Questions 2023

Please read the below information as background before starting the problem set!!!
Reading these two pages will only take a few minutes and will set you up for success
to make sure you are approaching this problem set in the best way possible.

First, I wanted to add a little bit of background on why this problem set exists to give some
context. Obviously, problem sets in general are great ways to solidify your knowledge of quantitative
material, but there’s additional reasoning for this problem set.

The Chin textbook is quite dense with many technical quantitative problems. The SOA specifically
lists out the Chin textbook questions they expect you to do on the QFI QF syllabus. Although I
believe the Chin textbook is written much better than MFD, the Chin textbook still has many typos.
If you spend time working through the Chin textbook, you will have to parse through these many
mistakes. This can take up costly study time and cause confusion. Therefore, we have made our TIA
problem set. We have solved the Chin problems, but added additional clarifications/corrections as
applicable to streamline your study process. This also makes things easier as you can simply solve
through this PDF instead of jumping around to the different page ranges of the Chin textbook.
Please note that this problem set works through exactly the same problems as those flagged on the
SOA syllabus.

Note that, of course, we at TIA are not perfect and given the length and complexity of this problem
set - it’s almost certain there are some typos in this PDF too. But the good news is we can refine
and improve this file each sitting, compared pointing people to a “static” resource we can’t update.

At the beginning of each problem, we denote the problem by its corresponding position in the
textbook. For example, 1.2.2.7 is Question 7 in Section 1.2.2. of the Chin textbook. If you post
any questions on the discuss forum, please title them in the format of the question prefix at the
beginning of each question (e.g. “1.2.2.7”) to keep things organized.

One other recommendation – if you are bogged down by the number of practice problems, feel free
to only work through a portion of the questions (e.g. odds only) on your first pass. This is an
imperfect approach, since you’ll notice that some questions reference prior questions, but feel free
to work through a portion of the problem set on your first pass if you are short on time. You can
always work on the remaining questions closer to the exam date if you have time.

Lastly, some additional information:

• This problem set is difficult! I highly recommend going through the following
BEFORE attempting this problem set to make sure you have the fundamentals
down first:
◦ Finish the MFD problem set and drill problems in the practice tab of the TIA seminar
◦ Watch all videos and read the detailed study manual for Section 1 and 1b
◦ Watch all review videos related to Section 1 topics
◦ Skim through the supplemental FAQ DSG and read through any content related to
Section 1

© 2023 The Infinite Actuary, LLC Page 1

QFI Quant Online Seminar – Chin Practice Questions 2023

• There are many times where the Chin textbook is correct but they do not provide detailed
step-by-step solutions, and one of the purposes of this problem set is to give more detailed
step-by-step solutions so you can follow them more easily

• The Chin textbook also sometimes references results proved in prior questions to solve certain
questions, and there are many times the Chin textbook will use a prior off syllabus question
to solve one of the problems you are responsible for. This can be frustrating and confusing, so
we have streamlined things for you and offered alternative solutions to those questions that
use the tools you are already familiar with. We think you will appreciate this and it makes
things much easier to follow!

• Note that for certain questions, hints are given to you (e.g. you may be given a relevant
equation to point you in the right direction). Please note that all hints are proven at the end
of this problem set for your reference

• The relevant chapter/question are noted by each question, but please note that these are not
straight word-for-word copies from the text. I have often made adjustments and clarifications.
In many cases, I broke down problems into sub-parts (e.g. part a, b, c) to break down the
problem into manageable parts. So keep in mind these questions are definitely motivated
from the text, but there are some intentional differences!

• The convention this author uses (as well as many other books on the QFI track) is that
when they write “log”, they are referring to natural logarithm “ln”, unless specifically told
otherwise

• Some, but not all, questions have a title. In the cases the Chin textbook gives a title to a
question, it’s included in italics at the beginning of each question

• A lot of the questions in this problem set are algebraically lengthy. While some of these
questions can be tough, keep in mind it’s essentially applying the same set of a few tools over
and over again to different examples. It can be easy to feel overwhelmed with the notation and
number of computations in some of these questions. When this happens, I highly recommend
taking a step back and summarizing in words what you are trying to do (e.g. for this problem
I need to prove a martingale so I need to prove these three properties, for another problem
I’m applying Ito’s Lemma so I’ll need to calculate the following partial derivatives). Breaking
down large problems into steps and summarizing the key takeaways can really help you focus
on the big picture to make sure you are absorbing the key themes

Please note that the below problem in the Chin textbook has particularly substantial errors, and so
we have significantly re-worded both the question and solutions for to correct the derivations. See
below for more details.

Page 134-135 (Problem 3.2.2.11)

• This question has a few mistakes in the Chin textbook

• The textbook incorrectly formulates Zt as eκYt instead of eκt · Yt . Additionally, the Chin
textbook has a mistake for V(XT |Xt = x)

• See 3.2.2.11(g)(ii) has the corrected variance formula

© 2023 The Infinite Actuary, LLC Page 2

QFI Quant Online Seminar – Chin Practice Questions 2023

Acknowledgements

• Thank you to Eric Chin, Dian Nel and Sverrir Olafsson for writing the Problems and Solutions
in Mathematical Finance textbook which includes these practice problems. This is a great
source of stochastic calculus practice for QFI QF!

• Also thank you to all students who beta tested this problem set and gave feedback

Errata List

• All errata will be posted here

• Please let me know if you believe you have found any errors (even if they are small typos)!

• You can email me at [email protected]

• Revision History
◦ V1 was posted on 3/10/2023

Good luck and enjoy! ,

© 2023 The Infinite Actuary, LLC Page 3

QFI Quant Online Seminar – Chin Practice Questions Chapter 1 - Practice Problems

CHIN CHAPTER 1 - PRACTICE PROBLEMS

© 2023 The Infinite Actuary, LLC Page 4

QFI Quant Online Seminar – Chin Practice Questions Chapter 1 - Practice Problems

1.2.2.7. (Normal Distribution Property). Given X ∼ N(µ, σ 2 ), show that for δ ∈ {−1, 1}:

δ(µ + σ 2 − logK)
   
X µ+ 21 σ 2 δ(µ − logK)
E[max(δ(e − K), 0)] = δe Φ − δKΦ
σ σ

Where K > 0 and Φ(·) denotes the cumulative standard normal distribution function.

© 2023 The Infinite Actuary, LLC Page 5

QFI Quant Online Seminar – Chin Practice Questions Chapter 1 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 1, Pages 18-19, Question 7

Note that the first problem here is fairly computationally lengthy. There’s no difficult stochastic
calculus involved, but it does involve remembering some basic calculus skills that are assumed as
background knowledge on the QFI track.

Throughout this question, it will be helpful to recall the following property for a standard normal
random variable:
P r(X ≥ a) = 1 − P r(X ≤ a) = 1 − Φ(a) = Φ(−a)
We start with the case where δ = 1. Notice that we can re-write the expectation as an integral, and
note that we only need to worry about integrating over non-zero values, which is when eX − K >
0 ⇒ X ≥ log K.

Thus,
R∞
E[max(eX − K, 0)] = logK (e
x − K)fX (x)dx

Next, we can plug in the standard normal PDF:

Z ∞ Z ∞
R∞ − 21 ( x−µ )
2 1 − 12 ( x−µ )
2 1 1 x−µ 2
x √1
= logK (e − K) σ 2π e σ dx = √ e σ
+x
dx −K √ e− 2 ( σ ) dx
logK σ 2π logK σ 2π
| {z } | {z }
I II

Looking at the above equation, we now have two integrals to evaluate. We will tackle them one-
by-one below. To evaluate the first integral, set w = x−µ 1
σ change the variable we are integrating :

2
− 12 ( x−µ 1 2
R∞ R∞
σ ) √1 e− 2 w +σw+µ dw
√1 +x
I: logK σ 2π e dx = logK−µ
2π
σ

1 2 R∞ 1 2 1 2 1 2 R∞ 1 2
= eµ+ 2 σ logK−µ √1 e− 2 w +σw− 2 σ dw = eµ+ 2 σ logK−µ √1 e− 2 (w−σ) dw
σ 2π σ 2π

Now use the property mentioned at the beginning of the solution to get:
h  i  
1 2
µ+ 21 σ 2 µ+σ 2 −logK
= eµ+ 2 σ Φ − logK−µ
σ − σ = e Φ σ

And for the second integral, again use the property mentioned at the start of this solution:
x−µ 2
   
− 21 (
R∞ ) dx = Φ − logK−µ = Φ µ−logK
II : √1
logK σ 2π e
σ
σ σ

Putting this all together gives the desired result:

µ + σ 2 − logK
   
X µ+ 21 σ 2 µ − logK
E[max(e − K, 0)] = e Φ − KΦ
σ σ

Therefore, we have proved the identity for δ = 1.

1 x−µ
The idea here is to standardize to a N(0,1). Recall from calculus that using the substitution w = σ
gives
dw = dx
σ
and x = wσ + µ

© 2023 The Infinite Actuary, LLC Page 6

QFI Quant Online Seminar – Chin Practice Questions Chapter 1 - Practice Problems

Next, we will go through a similar process for δ = −1.

Notice how we will integrate over non-zero values when K ≥ eX , which is when X ≤ logK.
1 x−µ 2
(K − ex ) σ√12π e− 2 ( ) dx
R logK
E[max(K − eX , 0)] = −∞
σ

1 x−µ 2 1 x−µ 2
√1 e− 2 ( σ ) √1 e− 2 ( σ ) +x dx
R logK R logK
=K −∞ σ 2π
dx − −∞ σ 2π

Now, use the results from the prior page to simplify and get:
  logK−µ
logK−µ 1 2 1 2
− eµ+ 2 σ √1 e− 2 (w−σ) dw
R σ
= KΦ σ −∞ 2π
  1 2
 
logK−µ logK−µ
= KΦ σ − eµ+ 2 σ Φ σ −σ

Putting this all together proves the identity for δ = −1:

logK − (µ + σ 2 )
   
logK − µ 1 2
E[max(K − eX , 0)] = KΦ − eµ+ 2 σ Φ
σ σ

© 2023 The Infinite Actuary, LLC Page 7

QFI Quant Online Seminar – Chin Practice Questions Chapter 1 - Practice Problems

1.2.3.4. (Change of Measure). Let Ω be a probability space and let P and Q be two probability
measures on Ω. Let Z(ω) denote the Radon-Nikodym derivative:

Q(ω)
Z(ω) = P(Z > 0) = 1
P(ω)

Show that the following relationships hold:

EQ (X) = EP (XZ)
 
X
EP (X) = EQ
Z

Where EQ (X) and EP (X) are the expectations under measures Q and P, respectively.

© 2023 The Infinite Actuary, LLC Page 8

QFI Quant Online Seminar – Chin Practice Questions Chapter 1 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 1, Pages 43-44, Question 4

This one is rather simple. We simply employ the definition of Z(ω) we are given and substitute
into the first principles definition of the expectation term.

Note: The logic used for the step of the blue equalities is using the first equation in the question
stem

X X
EQ (X) = X(ω)Q(ω) = X(ω)Z(ω)P (ω) = EP (XZ)
ω∈Ω ω∈Ω

Similarly:

X X(ω)  
Q X
X
P
E (X) = X(ω)P (ω) = Q(ω) = E
Z(ω) Z
ω∈Ω ω∈Ω

© 2023 The Infinite Actuary, LLC Page 9

QFI Quant Online Seminar – Chin Practice Questions Chapter 1 - Practice Problems

1.2.3.5. (Conditional Probability). Let (Ω, F , P) be a probability space and let G be a sub-σ-
algebra of F (sets in G are also in F ).

If IA is an indicator for a random event A, defined as:

(
1 If ω ∈ A
IA =
0 Otherwise

Show that:

E(IA |G ) = P(A|G )

© 2023 The Infinite Actuary, LLC Page 10

QFI Quant Online Seminar – Chin Practice Questions Chapter 1 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 1, Pages 43-44, Question 5

Since E(IA |G ) is G measurable, we are being asked to prove the following:

Z Z
E(IA |G )dP = P(A|G )dP B∈G
B B

Focusing on the left hand side of the equation, we can see that from first principles, the expectation
term can be reduced as follows:

Z Z
E(IA |G )dP = IA dP
B B

Now we focus on the right-hand side of the equation and note that we can break it down as follows:

Z Z Z
P(A|G )dP = P(A ∩ B) = IA∩B dP = IA dP
B Ω B

This identity holds since I is an indicator variable so all instances of overlap between A and B are
captured by integrating IA over B.

Since we have shown that both the left hand side and the right hand side of the first equation in
this solution reduce to the same term, we can say:

Z
E(IA |G ) = IA dP = P(A|G )
B

Note that this proof should also make intuitive sense since IA is simply a Bernoulli variable. The
expected value of a Bernoulli variable is simply the probability of its occurrence.

© 2023 The Infinite Actuary, LLC Page 11

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

CHIN CHAPTER 2 - PRACTICE PROBLEMS

© 2023 The Infinite Actuary, LLC Page 12

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.1.1. Let (Ω, F , P) be a probability space and consider a symmetric random walk with jth
step Zj defined as:

(
1 with probability 0.5
Zj =
−1 with probability 0.5

Zi is independent of Zj for all i 6= j

If we define Mk as follows with 0 = k0 ≤ k1 ≤ · · · ≤ kt :

ki
X
Mki = Zj i = 1, 2, ..., t M0 = 0
j=1

(a) Show that the symmetric random walk has independent increments (each term below is
independent of the rest)

Mk1 − Mk0 , Mk2 − Mk1 , ..., Mki − Mki−1

(b) Prove the following:

E(Mki+1 − Mki ) = 0
V(Mki+1 − Mki ) = ki+1 − ki

© 2023 The Infinite Actuary, LLC Page 13

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 55-68, Question 1

(a) Let’s start by defining each increment2 of M using the notation Si :

ki ki−1 ki
X X X
Si = Mki − Mki−1 = Zj − Zj = Zj
j=1 j=1 j=ki−1 +1

We need to show that the Si are independent of each other. This should be fairly obvious,
because we are summing up Z’s associated with non-overlapping increments.
To formally prove independence, we need to show that the conditional probability of attaining
a value of Si is identical to the unconditional probability. Consider two values of S given by
Sn and Sm where we have that mP< n for values of m and n in 1, 2, . . . , t. Now let A
kn
denote the event Mkn − Mkn−1 = j=kn−1 +1 Zj = Sn . Similarly, let B denote the event
Pkm
Mkm − Mkm−1 = j=km−1 +1 Zj = Sm . Then,

P(A ∩ B) P(A) · P(B)

P(A|B) = = = P (A)
P(B) P(B)

Where in the middle step we know that P(A ∩ B) = P(A) · P(B) since m < n and each
Zi is independent, there are no overlapping events between the intervals [kn−1 + 1, kn ] and
[km−1 + 1, km ]. Therefore, the probability of the intersection in the numerator is simply a
product of these two independent events. Summarizing, we showed the conditional probability
equals the unconditional probability:

P(A|B) = P (A) ⇒ P(Mkn − Mkn−1 = Sn |Mkm − Mkm−1 = Sm ) = P(Mkn − Mkn−1 = Sn )

Thus, we can deduce that Mk1 − Mk0 , Mk2 − Mk1 , ..., Mki − Mki−1 are independent

(b) First, we can calculate the expected value and variance of Zj as follows:
E(Zj ) = (1 · 0.5) + (−1 · 0.5) = 0
E(Zj2 ) = 1 since Zj2 always equals 1
V(Zj ) = E(Zj2 ) − E(Zj )2 = 1 − 02 = 1
Second, since successive Zj terms are independent, we can deduce:

ki
X
E(Mki − Mki−1 ) = E(Zj ) = 0
j=ki−1 +1

ki
X ki
X
V(Mki − Mki−1 ) = V(Zj ) = 1 = ki − ki−1
j=ki−1 +1 j=ki−1 +1

2
Note that this problem is all about understanding the intuition of increments. If you haven’t already, I highly
recommend watching the TIA review videos on increments.

© 2023 The Infinite Actuary, LLC Page 14

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.1.2. Let (Ω, F , P) be a probability space. For a symmetric random walk:

k
X
Mk = Zi M0 = 0
i=1
(
1 with probability 0.5
Zj =
−1 with probability 0.5

Show that Mk is a martingale

© 2023 The Infinite Actuary, LLC Page 15

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 55-68, Question 2

We need to prove the following three properties:

(i) E(Xt |Fs ) = Xs for all 0 ≤ s ≤ t

(ii) E(|Xt |) < ∞ for finite t

(iii) Xt is F -adapted

To prove the first property, we utilize the fact that the expected value of future changes in the
variable Mk is 0:

E(Mk |Fj ) = E(Mj |Fj ) + E(Mk − Mj |Fj )

E(Mk |Fj ) = Mj + 0

Where we know the latter expectation equals 0 because each Zj has an expected value of 0.

For the second property, we can see that3 :

k
X k
X k
X
|Mk | = Zi ≤ |Zi | = 1=k<∞
i=1 i=1 i=1

Thus,

E(|Mk |) < ∞
for finite k.

Finally, Mk is clearly Fk adapted since all values are conditional on Fk . That is, the information
at time k contains Z1 , . . . , Zk and thus implies the value of Mk .

Therefore, all 3 of our conditions are satisfied and we can state that Mk is a martingale

3
The inequality can be derived using triangle inequality.

© 2023 The Infinite Actuary, LLC Page 16

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.1.3. (Donsker Theorem). Let (Ω, F , P) be a probability space. For a symmetric random walk:

k
X 1
Mk = Zi M0 = 0 P(Zi = 1) = P(Zi = −1) =
2
i=1

Also let:

bntc
(n) 1 1 X
Wt = √ Mbntc = √ Zi
n n
i=1

For a fixed time t, show that:

bntc
(n) 1 X
lim Wt = lim √ Zi → N(0, t)
n→∞ n→∞ n
i=1

© 2023 The Infinite Actuary, LLC Page 17

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 55-68, Question 3

This problem asks us to show the limit converges to a N (0, t) random variable. First, we will show
that the expected value is indeed 0 and the variance is indeed t. Additionally, since we are summing
up many independently and identically distributed random variables, we can use the central limit
theorem to deduce that the limit converges a normal distribution.

We know that for all i = 1, 2, ...,

E(Zi ) = 0 V(Zi ) = 1

and that Zi is independent of Zj for all i 6= j. Therefore, we can conclude that:

bntc
(n) 1 X
E(Wt ) = √ E(Zi ) = 0
n
i=1
bntc bntc
(n) 1X 1X bntc
V(Wt ) = V(Zi ) = 1=
n n n
i=1 i=1

Therefore, taking the limit of both, we can conclude that:

(n)
lim E(Wt ) = 0
n→∞

(n) bntc
lim V(Wt ) = lim =t
n→∞ n→∞ n

Therefore, from the central limit theorem, we can see that:

(n)
lim Wt → N(0, t)
n→∞

To summarize, we have shown that Wt ∼ N (0, t). This is a very important result that pops up on
every QFI QF exam. Additionally, we showed we can theoretically construct a Wiener process as
the scaled sum infinitely many random innovations. As we crank up n, we get a finer partition and
finally converge to a normal distribution in the limit as n → ∞.

© 2023 The Infinite Actuary, LLC Page 18

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.1.4. (Covariance of Two Standard Wiener Processes). Let (Ω, F , P) be a probability space
and {Wt : t ≥ 0 } be a standard Wiener process. Prove the following two identities:

(a) Cov(Ws , Wt ) = min{s, t}

q
min{s,t}
(b) ρ = max{s,t} , where ρ is the correlation between Ws and Wt

Tip: We will prove how to derive these properties in the solutions, but also make a mental note of
these equations. These do pop up on QFI QF exams, so I think they are worth being familiar with
– especially part (a). In fact, we will use (a) to create the matrix for the next problem 2.2.1.5

© 2023 The Infinite Actuary, LLC Page 19

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 55-68, Question 4

(a) Note: There are multiple ways to approach this problem and two potential solutions you could
use are shown below! Try reading through both and seeing if one or the other is more intuitive
to you.
Solution #1:
Using the fact that Wt ∼ N(0, t) and Ws ∼ N(0, s) we can try to come up with an expression
for the covariance.

Cov(Ws , Wt ) = E(Ws Wt ) − E(Wt )E(Ws ) = E(Ws Wt )

To break down this expression, we can use the following expansion4 :

Ws · Wt = Ws · Wt − Ws2 + Ws2 = Ws (Wt − Ws ) + Ws2

Now, let us temporarily assume s ≤ t and take expectations of both sides5 of the equation
above:

E(Ws Wt ) = E(Ws (Wt − Ws ) + Ws2 ) = E(Ws )E(Wt − Ws ) + E(Ws2 ) = 0 + s = s

This gives us that

Cov(Ws , Wt ) = s

Similar steps show that, if s > t, we get Cov(Ws , Wt ) = t. Therefore, we have shown the
desired result that:
Cov(Ws , Wt ) = min{s, t}

Solution #2:
There is a faster way to derive this by leveraging the covariance property:

Cov(X, Y + Z) = Cov(X, Y ) + Cov(X, Z)

Without loss of generality, assume s ≤ t such that:

Cov(Ws , Wt ) = Cov(Ws , Ws + (Wt − Ws )) = Cov(Ws , Ws ) + Cov(Ws , Wt − Ws ) = s

| {z } | {z }
Var(Ws )=s 0

So we have that for s ≤ t, Cov(Ws , Wt ) = s. Similar steps show that, if s > t, we get
Cov(Ws , Wt ) = t. Therefore, we have shown the desired result that:

Cov(Ws , Wt ) = min{s, t}
4
If you are having any trouble seeing the intuition involved in this step, please review the TIA review video on
increments!
5
Here we used the fact that Ws and Wt − Ws are independent increments.

© 2023 The Infinite Actuary, LLC Page 20

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

(b) By definition:

Cov(Ws , Wt ) min{s, t}
ρ= p = √
V(Ws )V(Wt ) st

When s ≤ t:
r
s s
ρ= √ =
st t

On the flip side, when s > t:

r
t t
ρ= √ =
st s

Thus, we know that:

s
min{s, t}
ρ=
max{s, t}

© 2023 The Infinite Actuary, LLC Page 21

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.1.5. (Joint Distribution of Standard Wiener Processes). Let (Ω, F , P) be a probability space
and {Wt : t ≥ 0 } be a standard Wiener process.

(a) Find the moment generating function of the joint distribution (Wt1 , Wt2 , . . . Wtn ) and its
corresponding probability density function given t1 < t2 < · · · < tn

(b) Show that:  

T x2 −2txy+ty 2
1 − 12 t(T −t)
fWt ,WT (x, y) = p e
2π t(T − t)

(c) Prove the following two identities for the following conditional distributions given t < T :
 
t(T −t)
(i) Wt |WT = y ∼ N yt T , T

(ii) WT |Wt = x ∼ N(x, T − t)

===============

Note: This is a tough question. Chin does not provide hints, but I think for many people it may
be helpful to get the following hints:

T x2 −2txy+ty 2 y2 (x− yt )2
• Hint 1: t(T −t) − T = T
t(T −t)
T

T x2 −2txy+ty 2 x2 (y−x)2
• Hint 2: t(T −t) − t = (T −t)

Additionally, it may help to recall the basics of 2x2 matrix algebra stated below:

Given matrix
 
a b
Z=
c d

The determinant of Z is given by |Z| = ad − bc

And, assuming |Z| =

6 0, the inverse is given by:
 
−1 1 d −b
Z =
ad − bc −c a

© 2023 The Infinite Actuary, LLC Page 22

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 55-68, Question 5

(a) By definition, the moment generating function of (Wt1 , Wt2 ...Wtn ) is given as:

MWt1 ,Wt2 ,...Wtn (θ1 , θ2 , ...θn ) = E(eθ1 Wt1 +θ2 Wt2 +...+θn Wtn )

Given Wt1 , Wt2 −Wt1 , ..., Wtn −Wtn−1 are independent and normally distributed, we can write
the right-hand side in terms of independent increments:

θ1 Wt1 +θ2 Wt2 +...+θn Wtn = (θ1 +θ2 +...+θn )Wt1 +(θ2 +...+θn )(Wt2 −Wt1 )+...+θn (Wtn −Wtn−1 )

Thus,

E(eθ1 Wt1 +θ2 Wt2 +...+θn Wtn ) = E[e(θ1 +θ2 +...+θn )Wt1 +(θ2 +...+θn )(Wt2 −Wt1 )+...+θn (Wtn −Wtn−1 ) ]

= E[e(θ1 +θ2 +...+θn )Wt1 ]E[e(θ2 +...+θn )(Wt2 −Wt1 ) ] × ... × E[eθn (Wtn −Wtn−1 ) ]
And then using the properties shown in the moment generating function review video, we can
evaluate each expectation to simplify the above expression to:
1 2 t + 1 (θ +···+θ )2 (t −t )+...+ 1 θ 2 (t −t
= e 2 (θ1 +θ2 +...+θn ) 1 2 2 n 2 1 2 n n n−1 )

Or, equivalently, we can write this more compactly in matrix form:

1 T
= e2θ Σθ

where θT = (θ1 , θ2 , ..., θn ) and Σ is the covariance matrix for the Wiener process (where
the off-diagonal entries are computed using 2.2.1.4 where Cov(Wt1 Wt2 ) = E(Wt1 Wt2 ) =
min{t1 , t2 } = t1 ):

E(Wt21 )
   
E(Wt1 Wt2 )
... E(Wt1 Wtn ) t1 t1 ... t1
 E(Wt Wt )
2 1 E(Wt22 )
... E(Wt2 Wtn ) t1 t2 ... t2 
Σ=  =  .. ..
   
.. .. .. .. .. .. 
 . . . .  . . . .
E(Wtn Wt1 ) E(Wtn Wt2 ) ... 2
E(Wt2 ) t1 t2 ... tn

Thus, the moment generating function is:

1 T
MWt1 ,Wt2 ,...Wtn (θ1 , θ2 , ...θn ) = e 2 θ Σθ

We can then write the probability density function for the joint distribution (Wt1 , Wt2 , ...Wtn )
as:

1 1 T −1
fWt1 ,Wt2 ,...Wtn (x) = n 1 e− 2 xΣ x
(2π) |Σ|
2 2

where xT = (x1 , x2 , ..., xn )

© 2023 The Infinite Actuary, LLC Page 23

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

(b) Below I show two ways you could derive this equation.
Approach #1
Start by using the results from part (a):
1 1 T −1
fWt ,WT (x, y) = 2 1 e− 2 xΣ x
(2π) |Σ|
2 2

Where we have that:

 
t t
Σ=
t T

Next will be a couple steps using matrix algebra (assumed background knowledge, see hints
for more info). First, we will compute the determinant:

|Σ| = T t − t2 = t(T − t)

Next, invert the matrix:

 
−1 1 T −t
Σ =
t(T − t) −t t

Thus:
      
 −1 x  T −t x  T x − ty T x2 −2txy+ty 2
xT Σ−1 x 1 1
  
= x y Σ = t(T −t) x y = t(T −t) x y = t(T −t)
y −t t y −tx + ty

Putting this all together gives that:

 
T x2 −2txy+ty 2
1 − 12 t(T −t)
fWt ,WT (x, y) = p e
2π t(T − t)

Approach #2
Another way to derive this without needing matrices is to use the bivariate normal PDF.
  2     2 
1√ 1 x−µX x−µX y−µY y−µY
f (x, y) = 2
exp − 2(1−ρ2 ) σX − 2ρ σX σY + σY
2πσX σY 1−ρ
q
t
Note that Wt ∼ N(0, t) and WT ∼ N(0, T ). From the prior problem, ρ = T. Thus,
   q      2 
2
1
= √ √ t exp − 2(1− t ) 1 x
√ − 2 Tt √xt √y + √yT
2π tT 1− T T t T
 h 2 i
y2
= √1 exp − 2(TT−t) xt − 2 xy T + T
2π t(T −t)

Thus, simplifying, we get:

 
T x2 −2txy+ty 2
1 − 12 t(T −t)
fWt ,WT (x, y) = p e
2π t(T − t)

© 2023 The Infinite Actuary, LLC Page 24

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

(c) (i) The main equation we are going to use is

fWt ,WT (x, y)
fWt |WT (x|y) =
fWT (y)
.
To start off with, we will have to get formulas for the numerator and denominator of the
equation on the right hand side. Luckily we just computed the numerator in part (b), so that
part is done. The denominator is straightforward to determine:

1 y2
fWT (y) = √ e− 2T
2πT
Now we have calculated the numerator from (b) and the denominator above, so we can plug
them in and simplify using Hint 1:
" #
yt 2

T x2 −2txy+ty 2
 (x− )
√ −1 − 12 T
t(T −t)
fWt ,WT (x,y) 2 t(T −t)
fWt |WT (x|y) = fWT (y) = √2πT × e
y2
= r 1 e T
(2π) t(T −t) − t(T −t)
e 2T 2π T

Inspecting the right hand side, we can see the equation is written in the form of the PDF of
a normal distribution, which gives us the desired conclusion:

 
yt t(T − t)
Wt |WT = y ∼ N ,
T T

(ii) The result for this part should be intuitive using your understanding of increments. We
know that WT |Wt = x for t < T is equal to x plus the increment WT − Wt ∼ N(0, T − t).
Thus, WT |Wt = x ∼ N(x, T − t).
We can prove this formally using a similar approach to c(i). Here, we need to compute:
fWt ,WT (x, y)
fWT |Wt (y|x) =
fWt (x)
We can leverage the numerator already calculated in part (b):
 
T x2 −2txy+ty 2
1 − 12 t(T −t)
fWt ,WT (x, y) = p e
2π t(T − t)
The denominator is straightforward to determine:
1 − x2
fWt (x) = √ e 2t
2πt
Taking the ratio of these quantities and applying Hint 2 gives that:
 
(y−x)2
fWt ,WT (x, y) 1 − 12 (T −t)
fWT |Wt (y|x) = =p e
fWt (x) 2π(T − t)
Notice how the right hand side is written in terms of the PDF of a normal distribution. Thus,
we can see:
WT |Wt = x ∼ N(x, T − t)

© 2023 The Infinite Actuary, LLC Page 25

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.1.6. (Reflection). Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener
process. Show that under reflection, Bt = −Wt is also a standard Wiener process

© 2023 The Infinite Actuary, LLC Page 26

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 55-68, Question 6

Reminder on Wiener Processes

Remember that in order to prove that a function, Xt is a standard Wiener process, we have
to prove the following:

X0 = 0
Xt+s − Xt ∼ N(0, s)
Xt+s − Xt is independent of Xt
Note that the last condition can be proved by showing that E[(Xt+s − Xt )Xt ] = 0 since this
implies no covariance between the incremental movement, Xt+s and Xt . In this case, since
we are working with joint normal distributions, showing that there is no covariance
between increments is sufficient to prove independence.
These conditions will come up several times over the next few questions so use this as a
reference going forward.

Let’s start by listing some of the facts we know about Bt :

B0 = −W0 = 0

Bt+s − Bt = −(Wt+s − Wt ) ∼ N(0, s)

E[(Bt+s − Bt )Bt ] = Cov(−Wt+s + Wt , −Wt ) = Cov(Wt+s − Wt , Wt ) = 0

Therefore, we have shown that the starting value of Bt is 0 and that increments of Bt are indepen-
dent and identically distributed. We also showed that increments of the process follow a normal
distribution with an expected value of 0 and a variance equal to the length of the interval. These are
all of the necessary characteristics of a Wiener process. Therefore, we have proved that Bt = −Wt
is a standard Wiener process6

6
Note how the result of this question connects closely to the concepts discussed in the Innovation Term Sign review
video.

© 2023 The Infinite Actuary, LLC Page 27

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.1.7. (Time Shifting). Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard
Wiener process. Show that under time shifting, Bt = Wt+u − Wu is also a standard Wiener process.

© 2023 The Infinite Actuary, LLC Page 28

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 55-68, Question 7

To show we have a standard Wiener process7 , we will see if we can satisfy the same 3 conditions
as in Question 6.

B0 = Wu − Wu = 0

For the second property, note that:

• Bt+s = Wt+s+u − Wu

• Bt = Wt+u − Wu

Thus, we get:

Bt+s − Bt = Wt+s+u − Wt+u ∼ N (0, s)

Finally, for the third property:

E[(Bt+s − Bt ) · Bt ] = E [(Wt+s+u − Wt+u ) · (Wt+u − Wu )] = E(Wt+s+u − Wt+u ) · E(Wt+u − Wu ) = 0 · 0 = 0

Where the last step follows because we have written in terms of independent increments, each with
expectation zero.

Therefore, our usual conditions have all been met and Bt = Wt+u − Wu is also a standard Wiener
process

7
I did want to note that the Chin textbook solution for this question has mistakes, but we have corrected them
in our solution

© 2023 The Infinite Actuary, LLC Page 29

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.1.8. (Normal Scaling). Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard
Wiener process. Show that under normal scaling, Bt = cW t is also a standard Wiener process.
c2

© 2023 The Infinite Actuary, LLC Page 30

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 55-68, Question 8

At this point, hopefully you know the drill. Let’s run through each of the three properties we need
to show one-by-one in order to prove we have a standard Wiener process.

The first property is straightforward to show:

B0 = cW0 = 0

Next, for the second property, I’ll show two ways you could prove this. For both approaches, you’ll
want to note that:
Bt+s − Bt = c(W t+s − W t )
c2 c2

Approach #1

The mean is given by:

E(Bt+s − Bt ) = cE(W t+s ) − cE(W t ) = 0
c2 c2

The variance can be computed as:

V(Bt+s − Bt ) = V[c(W t+s − W t )]

c2 c2

= V(cW t+s ) + V(cW t ) − 2Cov(cW t+s , cW t )

c2 c2 c2 c2

= c2 V(W t+s ) + c2 V(W t ) − 2c2 Cov(W t+s , W t )

c2 c2 c2 c2

t+s t
= c2 + c2 − 2c2 × min{ t+s , t)



c2 c2 c2 c2

= t + s + t − 2t

=s

Note that in order to compute the covariance above, we used 2.2.1.4. Thus, we have shown the
second property:
Bt+s − Bt = c(W t+s − W t ) ∼ N(0, s)
c2 c2

Approach #2

Note that an alternative, quicker way to prove the second property is to denote A = W t+s − W t .
c2 c2
It is clear that A is an increment of length cs2 . Since we have written in terms of an increment
of W , we know we have a normal distribution. The expected value will be zero and the variance
equals the length of the increment. Thus, A ∼ N(0, cs2 ) and using V(cA) = c2 V(A) = s we have the
desired conclusion that Bt+s − Bt = cA ∼ N(0, s).

© 2023 The Infinite Actuary, LLC Page 31

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

Finally, we want to show that Bt+s − Bt is independent of Bt :

E[(Bt+s − Bt )Bt ] = E[(cW t+s − cW t )cW t ] = E[cW t+s − cW t ] · E[cW t ] = 0 · 0 = 0

c2 c2 c2 c2 c2 c2

Where the last step follows because we have written in terms of non-overlapping increments, which
are independent with expected value of zero

Therefore, our usual conditions have all been met and Bt = cW t is also a standard Wiener process
c2

© 2023 The Infinite Actuary, LLC Page 32

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.1.9. (Time Inversion). Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard
Wiener process. Show that under time inversion, the following function is also a standard Wiener
process.

(
0 if t = 0
Bt =
tW 1 if t 6= 0
t

© 2023 The Infinite Actuary, LLC Page 33

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 55-68, Question 9

Once again, we attempt to prove the usual 3 conditions:

B0 = 0 by definition
Bt+s − Bt = (t + s)W 1 − tW 1
t+s t

The mean can be computed as:

E(Bt+s − Bt ) = E[(t + s)W 1 ] − E[tW 1 ] = 0
t+s t

And the variance equals:

V(Bt+s − Bt ) = V[(t + s)W 1 ] + V[tW 1 ] − 2Cov[(t + s)W 1 , tW 1 ]

t+s t t+s t

= (t + s)2 V(W 1 ) + t2 V(W 1 ) − 2t(t + s) × Cov[W 1 , W1 ]

t+s t t+s t

We can use 2.2.1.4 to compute the covariance to get:

1
= (t + s)2 ( t+s ) + t2 ( 1t ) − 2t(t + s) × min{ t+s
1 1
, t } = (t + s) + t − 2t(t + s) × 1
t+s =s

Thus, we have shown that the expected value is zero and the variance is s. To additionally prove
normality, note that:
Bt+s − Bt = (t + s)W 1 − tW 1 = sW 1 − t(W 1 − W 1 )
t+s t t+s t t+s

Or, in other words, Bt+s −Bt is normally distributed because we re-wrote it in terms of independent
increments; summing up independent normals gives a normal distribution.

Therefore, we have proved the second property that:

Bt+s − Bt ∼ N(0, s)

Finally, we must show the independence8 between Bt and Bt+s − Bt :

E[(Bt+s − Bt )Bt ] = E(Bt+s Bt ) − E(Bt2 ) = Cov(Bt+s , Bt ) − E(Bt2 )
Now let us compute each of the pieces on the right hand side where, as usual, we use 2.2.1.4 to
compute the covariance:
1 1 1
Cov(Bt+s , Bt ) = Cov((t+s)W 1 , tW 1 ) = t(t+s)×Cov(W 1 , W 1 ) = t(t+s)×min{ , } = t(t+s)× =t
t+s t t+s t t+s t t+s
1
E(Bt2 ) = E(t2 W 12 ) = t2 V(W 1 ) = t2 × = t
t t t
Plugging in these values to the right hand side, we have that:
E[(Bt+s − Bt )Bt ] = t − t = 0
Therefore, the necessary conditions have been met and Bt = tW 1 , t 6= 0 is also a standard Wiener
t
process
8
Here, we used the fact that Cov(X, Y ) = E(XY ) − E(X)E(Y ). And when E(X) = E(Y ) = 0, this simplifies to
Cov(X, Y ) = E(XY )

© 2023 The Infinite Actuary, LLC Page 34

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.1.10. (Time Reversal). Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard
Wiener process. Show that under time reversal, the function Bt = W1 − W1−t is also a standard
Wiener process.

© 2023 The Infinite Actuary, LLC Page 35

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 55-68, Question 10

You know the drill by now! Let’s get right into it and show each of the three properties. The first
one is easy:
B0 = W1 − W1 = 0
Next, note that:

Bt+s − Bt = (W1 − W1−(t+s) ) − (W1 − W1−t ) = W1−t − W1−(t+s)

The mean and variance of this increment9 can be given as:

E(Bt+s − Bt ) = E(W1−t ) − E(W1−(t+s) ) = 0

V(Bt+s − Bt ) = V(W1−t ) + V(W1−(t+s) ) − 2Cov(W1−t , W1−(t+s) )
V(Bt+s − Bt ) = (1 − t) + (1 − (t + s)) − 2min{1 − t, 1 − (t + s)} = 1 − t + 1 − t − s − 2 × (1 − t − s) = s

Note that Bt = W1 − W1−t and Bt+s − Bt = W1−t − W1−(t+s) are both written as increments so
they are normally distributed. Summarizing, we have the result that:

Bt+s − Bt ∼ N(0, s)

Now we attempt to show independence between Bt+s − Bt and Bt :

E[(Bt+s −Bt )Bt ] = E[(W1−t −W1−(t+s) )(W1 −W1−t )] = E[(W1−t −W1−(t+s) )]E[(W1 −W1−t )] = 02 = 0

Note that above, we we were able to re-write in terms of non-overlapping increments of W . Non-
overlapping increments are independent, and also have expected value of zero.

Thus, we have shown that:

E[(Bt+s − Bt )Bt ] = 0
Therefore, the necessary conditions have been met and Bt = W1 − W1−t is also a standard Wiener
process

9
Note: We showed the textbook approach here, but you could prove the second property quite quickly by noting
that W1−t − W1−(t+s) ∼ N (0, s) because it is written in terms of an increment

© 2023 The Infinite Actuary, LLC Page 36

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.1.12. Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener

√
process.
Rt 3
Show that the pair of random variables (Wt , 0 Ws ds) has a correlation coefficient 2 by proving
they have the following covariance matrix:

1 2
 
t 2t
Σ= 1 2 1 3
2t 3t

© 2023 The Infinite Actuary, LLC Page 37

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 55-68, Question 12
Rt
Remember that by definition, the covariance matrix for the pair (Wt , 0 Ws ds) is:

" Rt #
V(Wt ) Cov(Wt , 0 Ws ds)
Σ= Rt Rt
Cov(Wt , 0 Ws ds) V( 0 Ws ds)

The first term in the matrix is fairly obvious since we know that Wt ∼ N(0, t) so we know that

V(Wt ) = t

Now we try to solve for the next term:

 Z t   Z t  Z t   Z t 
Cov Wt , Ws ds = E Wt Ws ds − E(Wt ) E Ws ds = E Wt Ws ds
0 0 | {z } 0 0
0

Next, note that Wt does not depend on s, so we can bring it inside the integral10 .
Rt
= 0 E[Wt Ws ]ds

And then re-write in terms of increments11 :

Rt
= 0 E[Ws (Wt − Ws ) + Ws2 ]ds
Z t Rt
= E[Ws (Wt − Ws )ds + 0 E[Ws2 ]ds
0
| {z }
0

And we know the first integral evaluates to zero because we have written in terms of independent
increments, thus E[Ws (Wt − Ws )] = E(Ws ) × E(Wt − Ws ) = 02 = 0

t2
Rt Rt
= 0 E[Ws2 ]ds = 0 sds = 2

Thus, we have shown how to derive the two off-diagonal entries of the matrix:
Z t
t2
 
Cov Wt , Ws ds =
0 2

10
Additionally, we can bring the expected value inside the integral. For more information, check out the Fubini’s
Theorem review video
11
If you had any trouble seeing the intuition to re-write it in this form, please check out the increments review
videos in the TIA seminar!

© 2023 The Infinite Actuary, LLC Page 38

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

Rt
Finally, we attempt to solve for the V( 0 Ws ds) term:

"Z 2 # 2
Z t  t  Z t
V Ws ds =E Ws ds − E Ws ds
0 0 0

Again, we can drop the last term since the expected value of Wt is 0.

Z t  Z t  Z t  Z t Z t  Z t Z t 
V Ws ds = E Ws ds Ws ds =E E(Ws Wu )du ds = E min(s, u)du ds
0 0 0 0 0 0 0

Where in the last step, we used the fact from 2.2.1.4 that Cov(Ws , Wu ) = E(Ws Wu ) = min{s, u}
hR R i
12 t t
Now we are left with evaluating the integral E 0 0 min(s, u)du ds . This is a multivariable
calculus problem; to tackle this, we can split into two cases:
(
u, 0 ≤ u ≤ s
min{s, u} =
s, s ≤ u ≤ ∞

In our case, we only need to integrate to t so we will replace our uppermost value with this. If we
consider just the innermost integral, we can simply split into two separate integrals spanning the
ranges defined above:

Z t  Z s=t Z u=s Z u=t 

V Ws ds = u du + s du ds
0 s=0 u=0 u=s

Z t Z t
s2 s2 t2 t3 t3
 
= + s(t − s) ds = st − ds = t − =
0 2 0 2 2 6 3

Therefore, we have proved that:

1 2
 
t 2t
Σ= 1 2 1 3
2t 3t

Rt
Lastly, recall the question also asked us to show that the pair of random variables (Wt , 0 Ws ds)
√
3
has a correlation coefficient 2 .
We can derive this from our correlation matrix:
√
 Rt 
Cov Wt , 0 Ws ds t2
2 3
ρ= q =q =
Rt
V(Wt )V( 0 Ws ds)
3
t · t3 2

12
Note that this part is also solved in Quiz #2 of the double integrals review video, so check that out for more
info! That video shows another way to solve the double integral through a more visual approach.

© 2023 The Infinite Actuary, LLC Page 39

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.1.13. Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener process.
Prove the following two identities:

Z s Z t 
1 1
Cov Wu du, Wv dv = min{s3 , t3 } + |t − s|min{s2 , t2 }
0 0 3 2

s s
min{s3 , t3 } 3 min{s, t}
ρ= + |t − s|
max{s3 , t3 } 2 max{s3 , t3 }

Rs Rt
Note that ρ is the defined as the correlation between the integrals 0 Wu du and 0 Wv dv.

© 2023 The Infinite Actuary, LLC Page 40

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 55-68, Question 13

Let’s start with the covariance:

Z s Z t  Z s Z t  Z s  Z t 
Cov Wu du, Wv dv = E Wu du × Wv dv −E Wu du E Wv dv
0 0 0 0 0 0

Right off the bat we can drop the second term since that involves the expected value of an increment
of Wu and Wv which are both equal to 0. Additionally, we can follow similar steps to the prior
problem to say:
Z s Z t  Z s Z t  Z s Z t Z sZ t
Cov Wu du, Wv dv = E Wu du × Wv dv = E[Wu Wv ] du dv = min{u, v} du dv
0 0 0 0 0 0 0 0

Since we know that E[Wu Wv ] = min{u, v}

Now, if we assume s ≤ t, we can take a similar approach to taking the double integral of a minimum
function that we did previously in 2.2.1.12:
Z sZ t Z s Z s Z t 
= min{u, v} du dv = min{u, v} du + min{u, v} du dv
0 0 0 0 s
Z s Z v Z s  Z sZ t
= u du + v du dv + min{u, v} du dv
0 0 v 0 s

Now for the second term, we can note that the variable, v (outer integral), only goes from 0 to
s while the variable u goes from s to t (inner integral). Therefore, it is clear in this case that
min{u, v} = v:
Z s 2  Z sZ t
v
= + v(s − v) dv + v du dv
0 2 0 s
Z s
s3 s2 s3
= +s − + v(t − s) dv
6 2 3 0

s3 s3 s3 s2
= + − + (t − s)
6 2 3 2
s3 s2 (t − s)
+ =
3 2
Following a similar procedure for the case where s > t, we can show:
Z s Z t
t3 t2 (s − t)

Cov Wu du, Wv dv = +
0 0 3 2

Therefore, we have proved that:

Z s Z t 
1 1
Cov Wu du, Wv dv = min{s3 , t3 } + |t − s|min{s2 , t2 }
0 0 3 2

© 2023 The Infinite Actuary, LLC Page 41

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

Next, we will derive the correlation coefficient:

R 
s Rt
Cov 0 Wu du, 0 Wv dv
ρ= q R
s Rt
V( 0 Wu du)V( 0 Wv dv)

Note that, of course, we just derived the numerator. Next, we focus on the denominator. From
2.2.1.12, we showed that:

t
t3
Z 
V Ws ds =
0 3

Therefore, we can conclude:

s Z
s  Z t  r 3r 3 √
s t s3 t3
V Wu du V Wv dv = =
0 0 3 3 3

Thus,

R 
s Rt
Cov 0 Wu du, 0 Wv dv 1 3 3 1 2 2
3 min{s , t } +√2 |t − s|min{s , t }
ρ= q R =
s Rt s3 t3
V( 0 Wu du)V( 0 Wv dv) 3

We are very close – to complete the proof, we just need to verify the above formula is consistent
with the formula given in the question stem. We can do so by analyzing each of the two cases
below.

When s ≤ t: r
1 3
+ 12 (t − s)s2 s2 + 23 (t − s)s
r
3s s3 3 s
ρ= √ = √ = 3
+ (t − s) 3
s3 t3 st3 t 2 t
3
Similarly, for s > t
r
1 3
+ 12 (s − t)t2 t2 + 23 (s − t)t
r
3t t3 3 t
ρ= √ = √ = 3
+ (s − t)
t3 s3 ts3 s 2 s3
3

Therefore, we have proved that:

s s
min{s3 , t3 } 3 min{s, t}
ρ= + |t − s|
max{s3 , t3 } 2 max{s3 , t3 }

© 2023 The Infinite Actuary, LLC Page 42

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.3.1. Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener process.13
Z ∞ x2
You may use the fact from deterministic calculus that for σ > 0, xe− 2σ2 dx = σ 2 .
0

(a) Calculate E[max(0, Wt )]

(b) Calculate E(|Wt |)

(c) Show that Wt is a martingale. In other words, prove that a standard Wiener process is a
martingale.

13
Note: The Chin textbook only asks you to do part (c), but I think it’s helpful to isolate parts (a) and (b) as
computations here because it will be used for part (c). The Chin textbook actually does not derive the value of
E(|Wt |), but instead provides an upper bound. Unfortunately, the upper bound is derived using Holder’s inequality,
which is defined on an off-syllabus section of the Chin textbook. It’s a bit silly to use an inequality here anyways,
since it’s straightforward to compute the expected value in part (b)!

© 2023 The Infinite Actuary, LLC Page 43

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 71-74, Question 1

(a) Remember the PDF of a normal random variable with µ = 0 is given by:

1 x2
f (x) = √ e− 2σ2
σ 2π
And also: Z
E(X) = xf (x)dx

Let A be the value of a normally distributed random variable with mean 0 and standard deviation
σ, floored at zero. Then: Z ∞
1 x2
E [A] = √ xe− 2σ2 dx
σ 2π 0
And we can simplify this using the hint:
1
= √ · σ2
σ 2π
σ
=√
2π
√
Thus, we have that, E(A) = √σ . Additionally, know V (Wt ) = t, and thus σ = t. Plugging this
2π
in gives us our answer:
r
t
E[max(0, Wt )] =
2π
(b) Note that, by symmetry, we can multiply by 2 to get the answer here. For part (a), we are
only accumulating for positive values of Wt , whereas |Wt | gets equal contributions from both sides
of the tail, so we just multiply our answer to (a) times 2.

r r
t 2t
E(|Wt |) = 2 × =
2π π

Note: The solution to part (c) is shown on the next page.

© 2023 The Infinite Actuary, LLC Page 44

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

(c)

Reminder on Martingale Processes

Remember that in order for a process Xt to be considered a martingale, the following

conditions should be met:
• E(Xt |Fs ) = Xs for all 0 ≤ s ≤ t

• E(|Xt |) < ∞

• Xt is Ft adapted
We will refer to this list for all the questions in this section

Let’s see how the properties above apply in our case. For the first condition:

E(Wt |Fs ) = E(Wt − Ws + Ws |Fs ) = E(Wt − Ws |Fs ) + E(Ws |Fs )

Now we recognize that the first term is equal to 0. We also recognize that since the second term is
a known constant given information set Fs . Therefore, our expected value expression reduces to:

E(Wt |Fs ) = Ws

Which means our first condition is met.

q
2t
For the second condition, use part (b) to see that E(|Wt |) = π < ∞ for finite t

Finally, it is obvious that Wt is Ft adapted . Given the information at time t, we know the value
of Wt . Therefore, our final condition is met and we have shown that Wt is a martingale

© 2023 The Infinite Actuary, LLC Page 45

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.3.2. Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener process.
Show that Xt = Wt2 − t is a martingale

© 2023 The Infinite Actuary, LLC Page 46

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 71-74, Question 2

We will go through testing our usual conditions:

E(Xt |Fs ) = E(Wt2 − t|Fs ) = E[(Wt − Ws + Ws )2 |Fs ] − t

= E[(Wt − Ws )2 |Fs )] + 2E[(Ws (Wt − Ws )|Fs )] + E[Ws2 |Fs ] − t

Now we recognize that the first term is equal to t − s because Wt − Ws ∼ N(0, (t − s)). The second
term reduces to 0 because the expected value of Wt − Ws is 0. The final term is a known constant
because we already have information set Fs . Therefore:

E(Xt |Fs ) = t − s + 0 + Ws2 − t = Ws2 − s = Xs

Which means our first condition is met

For the second condition, start with the triangle inequality14 :

|Wt2 − t| ≤ |Wt2 | + |t| = Wt2 + |t|

Thus,

E[|Xt |] = E[|Wt2 − t|] ≤ E(Wt2 + |t|) = t + |t| < ∞

Therefore, the second condition is met.

Finally, since Xt is a function of Wt , it is clearly Ft - adapted. In other words, given Ft we know

the value of Wt , which gives the value of Xt – so our third condition is met.

Thus, we have shown the conditions to prove that Xt = Wt2 − t is a martingale

14
Check out the review video on the triangle inequality if you haven’t already!

© 2023 The Infinite Actuary, LLC Page 47

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.3.3. Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener process.
1 2t
Show that Xt = eλWt − 2 λ is a martingale

© 2023 The Infinite Actuary, LLC Page 48

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 71-74, Question 3

Given Wt ∼ N(0, t) we can write:

 
1 2 1 2 2
logXt = λWt − λ t ∼ N − λ t, λ t
2 2
 
1 2 2
Xt ∼ log-N − λ t, λ t
2

Now let’s try to prove our usual 3 conditions:

1 2 1 2 1 2
E(Xt |Fs ) = E(eλWt − 2 λ t |Fs ) = e− 2 λ t · E[eλWt |Fs ] = e− 2 λ t · E[eλ(Wt −Ws )+λWs |Fs ]
1 2
= e− 2 λ t · E[eλ(Wt −Ws ) |Fs ] · E[eλWs |Fs ]
1 2 1 2 (t−s)
= e− 2 λ t · e 2 λ · eλWs
1 2s
= eλWs − 2 λ

= Xs

Thus, E(Xt |Fs ) = Xs

Therefore, we have met the first condition

For the second condition, we can start by recognizing that:

1 2 1 2t
|Xt | = |eλWt − 2 λ t | = eλWt − 2 λ

Now we can try to take the expected value of this term:

1 2 1 2
E(eλWt − 2 λ t ) = e− 2 λ t · E(eλWt )
1 2 1 2 1 2
E(|Xt |) = E(eλWt − 2 λ t ) = e− 2 λ t · e 2 λ t = 1 < ∞

Therefore, the second condition is met

Finally, since Xt is a function of Wt , it is clearly Ft - adapted. In other words, given Ft we know

the value of Wt , which gives the value of Xt – so our third condition is met.
1 2t
Thus, we have shown that Xt = eλWt − 2 λ is a martingale

© 2023 The Infinite Actuary, LLC Page 49

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.3.4. Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener process.
Show that Xt = Wt3 − 3tWt is a martingale
Z ∞ x2
You may use the fact from deterministic calculus that for σ > 0, x3 e− 2σ2 dx = 2σ 4 .
0

(a) Calculate E[max(0, Wt3 )]

(b) Calculate E(|Wt3 |)

(c) Show that Xt is a martingale

© 2023 The Infinite Actuary, LLC Page 50

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 71-74, Question 4

(a) Remember the PDF of a normal random variable with µ = 0 is given by:
1 x2
f (x) = √ e− 2σ2
σ 2π
And also: Z
3
E(X ) = x3 f (x)dx

Let A be the value of a normally distributed random variable with mean 0 and standard deviation
σ, floored at zero. Then: Z ∞
1 x2
E [A] = √ x3 e− 2σ2 dx
σ 2π 0
And we can simplify this using the hint:
1
= √ · 2σ 4
σ 2π
r
2
= σ3 ×
π

Finally, we know V (Wt ) = σ 2 = t. Thus,

r
1.5 2
= t ×
π
(b) By symmetry,

r
2
E(|Wt3 |) = 2 × t1.5 ×
π

As a result, we see that for finite t, E(|Wt3 |) < ∞.

(c) Let’s get right into it and start by trying to see if the first condition is met:

E(Xt |Fs )

E(Wt3 − 3tWt |Fs )

= E[Wt (Wt )2 |Fs ] − 3tE[Wt |Fs ]

= E[Wt (Wt − Ws + Ws )2 |Fs ] − 3tE[Wt |Fs ]

= E[Wt (Wt − Ws )2 |Fs ] + 2Ws E[Wt (Wt − Ws )|Fs ] + Ws2 E[Wt |Fs ] − 3tE[Wt |Fs ]

© 2023 The Infinite Actuary, LLC Page 51

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

Now, we basically have four terms that we will need to compute. To make this easier to follow, I
thought color-coding each of the four calculation terms might be helpful so you can see how each
simplifies as we progress:

= E[Wt (Wt − Ws )2 |Fs ] + 2Ws E[Wt (Wt − Ws )|Fs ] + Ws2 E[Wt |Fs ] − 3tE[Wt |Fs ]

= E[(Wt − Ws + Ws )(Wt − Ws )2 |Fs ]+2Ws E[(Wt − Ws + Ws )(Wt − Ws )|Fs ]+Ws2 E[Wt |Fs ]−3tE[Wt |Fs ]

= E[(Wt − Ws )3 |Fs ] + E[(Ws )(Wt − Ws )2 |Fs ]+2Ws E[(Wt − Ws )2 |Fs ] + 2Ws E[Ws (Wt − Ws )|Fs ]+
Ws2 · Ws − 3t · Ws

= 0 + Ws E[(Wt − Ws )2 |Fs ] + 2Ws E[(Wt − Ws )2 |Fs ] + 0 + Ws2 · Ws − 3t · Ws

= Ws (t − s) + 2Ws (t − s) + Ws3 − 3tWs

= Ws3 − 3sWs = Xs

Thus, we have shown that E(Xt |Fs ) = Xs

Note that several of the expected value terms can be broken down by recognizing that Wt − Ws ∼
N(0, t − s) (recall that the expected value of a normal variable to the third power equals 0).

For the second condition, we want to show that:

E|Xt | = E|Wt3 − 3tWt | < ∞

already showed in (b) that for finite t, E(|Wt3 |) < ∞ and also know from 2.2.3.1 that
We have q
E|Wt | = 2t
π . Thus, putting together our results and using triangle inequality gives:

r r
2 2t
E|Xt | = E|Wt3 − 3tWt | ≤ E|Wt3 | + 3|t| · E|Wt | = 2 × t1.5 × + 3|t| <∞
π π

Therefore, we have shown the second condition.

Finally, since Xt is a function of Wt , it is clearly Ft - adapted. In other words, given Ft we know

the value of Wt , which gives the value of Xt – so our third condition is met.

Thus, we have shown that Xt = Wt3 − 3tWt is a martingale

© 2023 The Infinite Actuary, LLC Page 52

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

Alternate Solution to Show Second Property

Note: The TIA solution shows in part (b) how to prove the second property by deriving the expected
value formula. An alternative approach is to not bother calculating the expected value, but instead
find an upper bound for it using a tool called Holder’s inequality, and then showing that upper bound
is finite. Below we show this alternative approach which is how Chin solves this question.

Ultimately, I think this approach is weaker because why use bounds when you can directly solve
for the expectation. Additionally, it relies on a tool called Holder’s inequality which is derived in
problem 1.2.3.2 and is not one of the problems flagged on the QFI QF syllabus. However, this
approach does still work and is used by Chin so I wanted to show it below for reference

Holder’s inequality, which you can take as a given, is that for a pair of jointly continuous variables
X and Y with constants p, q > 1 such that p1 + 1q = 1, we have that:
1 1
E(|XY |) ≤ (E|X p |) p × (E|Y q |) q

For example, Holder’s inequality using X = Wt2 and Y = Wt and p = q = 2 gives that:
q
E(|Wt | ) ≤ E[(|Wt |2 )2 ] · E(|Wt |2 )
3

Thus, since we know15 that E(Wt4 ) = 3t2 and E(Wt2 ) = t, we have:

q √ √
E(|Wt |3 ) ≤ E(Wt4 ) · E(Wt2 ) = 3t2 · t = 3t3

Now remember we are trying to show that

E|Xt | = E|Wt3 − 3tWt | < ∞

From triangle inequality, we know that:

√
E|Xt | = E|Wt3 − 3tWt | ≤ E|Wt3 | + E|3tWt | ≤ 3t3 + 3|t| · E|Wt |

Now we need to deal withq the E|Wt | term. We have a couple of options. We could use the fact
from 2.2.3.1 that E|Wt | = 2t
π . Or, using Jensen’s inequality
16 seen in Chin Ch 1,

q q √
E|Wt | = E( Wt ) ≤ E(Wt2 ) = t
2

Thus, putting this all together, we finally get that:

√ √
E|Xt | = E(|Wt3 − 3tWt |) ≤ 3t3 + 3|t| t < ∞

Which means our second condition is met!

15
Warning: The Chin textbook incorrectly says in their solution to this problem that E(Wt4 ) = 3t, but keep in
mind E(Wt4 ) = 3t2
16 √
Note that here we are applying Jensen’s inequality to a concave function f (x) = x, so the inequality goes the
opposite direction from Jensen’s inequality for convex functions. We will not go through a formal proof, but the
reasoning is straightforward and a result of the fact that a function g is convex if and only if −g is concave. Put
simply, this negative sign results in the inequality being flipped. That is, if ϕ : R 7→ R is a concave function, then
E[ϕ(X)|G] ≤ ϕ[E(X|G)]

© 2023 The Infinite Actuary, LLC Page 53

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

Summary:

Ok - this has been a really long solution! So feel free to stop if you’d like. But if I still have
your attention, I wanted to pause and reflect on one last part comparing the actual expectations
vs the upper bounds using our inequalities. I thought this would be interesting to highlight. To
summarize, we showed:
q √
E|Wt | = 2t π < t
q
2
√
E|Wt3 | = 2 × t1.5 × π < 3t3

In other words, E|Wt | is approximately .798t.5 , and our inequality gave an upper bound of t.5 .

E|Wt3 | is approximately 1.596t1.5 , and our inequality gave an upper bound of approximately
1.732t1.5

Notice how for both cases, the power of t was the same, but the inequality ends up giving you a
larger coefficient!

In terms of exam day, I think it’s great to have a diverse set of tools in your toolbox. We’ve now
seen how to apply some inequalities (Holder, Jensen, Triangle) to prove finiteness, and we’ve also
seen in some cases you can just go ahead and directly compute expectations.

© 2023 The Infinite Actuary, LLC Page 54

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

2.2.3.5. Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener process.
Show that the following hyperbolic processes are martingales:

1 2
Xt = e− 2 λ t cosh(λWt )
1 2
Yt = e− 2 λ t sinh(λWt )

Note: To solve this problem, it may be helpful to recall the following properties of hyperbolic functions
which you can take as a given:

1  λx 
cosh(λx) = e + e−λx
2
1  λx 
sinh(λx) = e − e−λx
2

© 2023 The Infinite Actuary, LLC Page 55

QFI Quant Online Seminar – Chin Practice Questions Chapter 2 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 2, Pages 71-74, Question 5

This problem may seem intimidating but the key is simply recognizing that:

1  λWt 
cosh(λWt ) = e + e−λWt
2
1  λWt 
sinh(λWt ) = e − e−λWt
2

Therefore, we can rewrite Xt and Yt as:

1  λWt − 1 λ2 t 1 2

Xt = e 2 + e−λWt − 2 λ t
2
1  λWt − 1 λ2 t 1 2

Yt = e 2 − e−λWt − 2 λ t
2

Now, we recognize that each of the components of Xt and Yt are martingales (proved in 2.2.3.3)
and we know that adding or subtracting two martingales results in a martingale so we can conclude
that both Xt and Yt are martingales

© 2023 The Infinite Actuary, LLC Page 56

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

CHIN CHAPTER 3 - PRACTICE PROBLEMS

© 2023 The Infinite Actuary, LLC Page 57

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.1.13. Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener process

(a) Using integration by parts, show that

Z t Z t
Ws ds = (t − s)dWs
0 0
Rt
(b) Let Mt = 0 Ws ds
(i) Compute E(Mt ) and Var(Mt )
(ii) Prove that Mt is normally distributed

(c) Is Bt below a standard Wiener process?

(
0 t=0
Bt = √ Rt
3
t 0 Ws ds t>0

© 2023 The Infinite Actuary, LLC Page 58

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Page 115, Question 13

(a) Recall that with integration by parts,

Z Z
udv = uv − vdu

In our case, u = Ws , v = s. Thus,

Z t t Z t
Ws ds = sWs − sdWs
0 0 0
Z t
= tWt − sdWs
0
Z t Z t
=t dWs − sdWs
0 0
Z t Z t
= tdWs − sdWs
0 0
Z t
= (t − s)dWs
0

Z t Z t
(b) (i) Let Mt = Ws ds = tWt − sdWs
0 0

Since Mt is an Ito Integral, E(Mt ) = 0. Thus Var(Mt ) = E(Mt2 )

Z t 2
Mt2 = (t − s)dWs
0
Z t 2 Z t
By Ito’s Isometry, E(Mt2 ) = E( (t − s)dWs ) = E( (t − s)2 ds)
0 0

t t
t3
Z Z
= (t − s)2 ds = s2 ds =
0 0 3
t3
Thus, E(Mt ) = 0 and Var(Mt ) =
3
(ii) The Chin textbook gives a lengthy proof to this (shown on the next page) that requires
using off-syllabus problems, so I am mixing things up for this part and giving my own solution
which I think is much clearer. First note that:
Z t Z t
Mt = Ws ds = (t − s)dWs
0 0
Rt P
Focusing on 0 (t − s)dWs , we can re-write this as a sum i (t − ui−1 )(Wi − Wi−1 ). We
can see that we are adding up a bunch of constants times independent WienerR increments.
t
Remember, adding up independent normals gives a normal distribution. Thus, 0 (t − s)dWs
is normal, and therefore this completes the proof that Mt is normal!

© 2023 The Infinite Actuary, LLC Page 59

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

Chin solution:
For reference, the Chin textbook solution is below. I recommend not spinning your wheels
or getting bogged down in this solution. It references multiple off-syllabus problems, so based
on the current syllabus I don’t think this is too important. Feel free to just briefly skim or
even skip the part below. The basic premise of their proof is to show that M has the moment
generating function of a normal distribution, and thus must be normal. Also look at Fall 2022
QFI QF 4(a) and notice how they did not require you to go through all of these steps, and
instead a quick proof similar to the one I just presented before was sufficient for full credit.
To start off, we know that:
Z t Z t
Mt = Ws ds = (t − s)dWs
0 0

From 3.2.1.7 and 3.2.1.12 (not on the syllabus), we can deduce that Mt is a martingale and
also that quadratic variance can be computed as:
Z t
t3
hM, M it = (t − s)2 ds =
0 3
Let’s define yet another function, Zt so that:
t3
 
θMt − 12 θ2
Zt = e 3

Now, using a Taylor expansion, we have:

∂Zt ∂Zt 1 ∂ 2 Zt
dZt = dt + dMt + (dMt )2
∂t ∂Mt 2 ∂Mt2
1 1
= − θ2 t2 Zt dt + θZt dMt + θ2 Zt (dMt )2
2 2
1 2 2 1
= − θ t Zt dt + θZt dMt + θ2 Zt t2 dt
2 2
= θZt dMt

Note that this simplification is possible because (dMt )2 = t2 dt

Now if we take the integrals on both sides, we have:
Z t Z t Z t
dZu = θZu dMu ⇒ Zt = Z0 + θ Zu dMu
0 0 0
Taking the expectation of both sides of the equation and noting Z0 = 1, we have:
Z t 
E(Zt ) = 1 + θE Zu dMu = 1
0

Hence,
t3
 
1 2
θMt θ
E[e ]=e 2 3

We recognize this as the moment generating function of a normal distribution with mean zero
3
and variance t3 . Thus:
Z t  3
t
Ws ds ∼ N 0,
0 3

© 2023 The Infinite Actuary, LLC Page 60

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

(c) No, this is not a Wiener process. We will show it is not a Wiener process because we know a
Wiener process must have that V(Bt+u − Bt ) = u. However, upon computing V(Bt+u − Bt )
we will see it is not equal to u, and thus we do not have a Wiener process.
The expected value itself is fine and does equal 0, which you could verify for reference:

E(Bt+u − Bt ) = E(Bt+u ) − E(Bt )

√ Z t+u ! √ Z t !
3 3
=E Ws ds − E Ws ds
t+u 0 t 0
=0

Now let’s try computing the variance. To do so, it will be helpful to recall from problem
2.2.1.13 that:
Z s Z t 
1 1
Cov Wu du, Wv dv = min{s3 , t3 } + |t − s|min{s2 , t2 }
0 0 3 2

So in our case: Z t+u Z t 

1 1
Cov Ws ds, Ws ds = t3 + ut2
0 0 3 2
Now let’s test out the variance of Bt+u −Bt , plugging in the covariance term we pre-computed
above:

V(Bt+u − Bt ) = V(Bt+u ) + V(Bt ) − 2Cov(Bt+u , Bt )

√ Z t+u ! √ Z t ! Z t+u Z t 
3 3
=V Ws ds + V Ws ds − 2Cov Ws ds, Ws ds
t+u 0 t 0 0 0

Now plug in the variance formula we derived in part (b) to get:

(t + u)3 3 t3
 
3 6 1 3 1 2
= + − t + ut
(t + u)2 3 t2 3 t(t + u) 3 2
From here, we simplify:
 
6 1 3 1 2
= (t + u) + (t) − t + ut
t(t + u) 3 2
3
2t + 3ut 2
= 2t + u −
t(t + u)
t + u 2t3 + 3ut2
= (2t + u) · −
t+u t(t + u)
2
2t + 3tu + u 2 2t2 + 3ut
= −
t+u t+u
u2
= 6= u
t+u

Thus, V(Bt+u − Bt ) 6= u and therefore, Bt is not a standard Wiener process.

© 2023 The Infinite Actuary, LLC Page 61

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.1. Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener process.

(a) Find the SDE for the random process

Xt = Wtn , n ∈ Z+

(b) Show that

Z t  
1
E(Wtn ) = n(n − 1) (n−2)
E Ws ds
2 0

(c) Using the mathematical induction prove that

 n
n!t 2
 n n
n = 2, 4, 6, ...
E(Wtn ) = 22 ( )
2
!
0 n = 1, 3, 5, ...

© 2023 The Infinite Actuary, LLC Page 62

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 1

(a) Start with the given:

Xt = Wtn , n ∈ Z+

Then expand dXt using Baby Ito’s lemma17 :

∂X ∂X 1 ∂2X
dXt = dWt + dt + dt
∂Wt ∂t 2 ∂Wt2

(n−1) 1 (n−2)
dWtn = nWt dWt + n(n − 1)Wt dt
2
∂Xt
(Note that ∂t dt = 0)
Taking the integrals, we have:
Z t Z t Z t
1
dWsn = nWs(n−1) dWs + n(n − 1) Ws(n−2) ds
0 0 2 0
Z t Z t
1
Wtn = nWs(n−1) dWs + n(n − 1) Ws(n−2) ds
0 2 0

(b) Taking the expectation of the solution from part (a) above, we have:

Z t  Z t 
1
E(Wtn ) = nE Ws(n−1) dWs
+ n(n − 1)E Ws(n−2) ds
0 2 0
Z t h
1 i
= n(n − 1) E Ws(n−2) ds
2 0

17
Since Xt is a function of Wt , we know that Baby Ito’s Lemma applies.

© 2023 The Infinite Actuary, LLC Page 63

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

(c) Note that proving this for odd powers of n is trivial because we know by symmetry the
expectation of E(Wtk ) for odd k must equal18 0. Therefore, we will focus on proving the
formula for even powers k using induction below.
For those unfamiliar with induction, I wanted to briefly recap how a proof by induction works.
First, you need to prove the identity holds for a base case (e.g. n = 2). Then you need to
show that if the identity holds for a certain number (e.g. even number n = 2k) then it holds
for the following number (e.g. the next even number n = 2(k + 1)). We will do exactly that
below, which proves the identity by induction.
To prove this identity, first try out the base case for n = 2:

2! · t1
E(Wt2 ) = =t
21 · 1!
We know this to be correct since Wt ∼ N(0, t). Thus, we have proved the base case.
Next, assume that the result is true for any n = 2k:

(2k)!tk
E(Wt2k ) =
2k k!
Then, for n = 2(k + 1), and using our result from part (b) we have:

Z t
2(k+1) 1
E(Wt ) = (2k + 2)(2k + 1) E(Wt2k )ds
2 0
Z t
1 (2k)!sk
= (2k + 2)(2k + 1) ds
2 0 2k k!
(2k + 2)! t k
Z
= s ds
2k+1 k! 0
(2k + 2)!tk+1
=
2k+1 (k + 1)!
2(k+1)
(2(k + 1))!t 2
= 2(k+1)  
2(k+1)
2 2 2 !

Therefore, the identity holds for n = 2(k + 1) if it holds for n = 2k.

Therefore, we have shown the required steps for a proof by induction, and this proves the
formula below:
 n
n!t 2
 n n
n = 2, 4, 6, ...
E(Wtn ) = 22 ( )
2
!
0 n = 1, 3, 5, ...

18
If you desire, you can use similar steps for a proof by induction for odd k.

© 2023 The Infinite Actuary, LLC Page 64

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.2. Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener process.

(a) For constant θ find the SDE for the random process

1 2
Xt = eθWt − 2 θ t

(b) By writing the SDE in integral form, calculate the moment generating function of the standard
Wiener process:
 
E eθWt

© 2023 The Infinite Actuary, LLC Page 65

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 2

(a) Start by using the given:

1 2
Xt = eθWt − 2 θ t

Now, compute dXt with Ito’s Lemma:

∂Xt 1 ∂ 2 Xt
 
∂Xt
dXt = + 2 dt + dWt
∂t 2 ∂Wt ∂Wt
 
1 2 1 2
= − θ Xt + θ Xt dt + θXt dWt
2 2
= θXt dWt

(b) Taking the integrals of part (a) above, we have:

Z t Z t
dXs = θXs dWs
0 0
Z t
Xt − X0 = θXs dWs
0

Now since we know that X0 = 1, we can write:

Z t 
E(Xt ) − 1 = E θXs dWs =0
0
 1 2

E eθWt − 2 θ t = 1

Thus, we have that:

  1 2
E eθWt = e 2 θ t

© 2023 The Infinite Actuary, LLC Page 66

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.3. Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener process.
Consider the process:

Zt = eθWt

where θ is a constant parameter

(a) Use Ito’s formula to find an SDE for Zt

(b) By setting mt = E(eθWt ) show that the integrated SDE can be expressed as

dmt 1 2
− θ mt = 0
dt 2
(c) Given W0 = 0, solve the first order ordinary differential equation to find mt

© 2023 The Infinite Actuary, LLC Page 67

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 3

(a) Start with the given:

Zt = eθWt
Expanding dZt using the Ito’s formula, we have:

1 ∂ 2 Zt
 
∂Zt
dZt = dWt + (dWt2 )
∂Wt 2 ∂Zt2
1
= θeθWt dWt + θ2 eθWt dt
2
1 2
= θZt dWt + θ Zt dt
2

(b) Taking the integrals of part (a) above, we have:

Z t Z t Z t
1 2
dZs = θZs dWs + θ Zs ds
0 0 0 2
Z t Z t
1 2
Zt − Z0 = θZs dWs + θ Zs ds
0 0 2
Now we know that Z0 = 1 and that under the expectation, we can drop the first term on the
right hand side of the equation. Therefore, we can write:
Z t
1 2
E(Zt ) − 1 = θ E(Zs )ds
0 2

Differentiating the integral equation we have:

d t1 2
Z
dE(Zt )
= θ E(Zs )ds
dt dt 0 2

dE(Zt ) 1
= θ2 E(Zt )
dt 2
dmt 1 2
− θ mt = 0
dt 2

© 2023 The Infinite Actuary, LLC Page 68

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

(c) Setting the integrating factor19 as:

1 2 1 2
R
I = e− 2
θ dt
= e− 2 θ t

and multiplying the differential equation with I, we have:

1 2 dmt 1 2 1
e− 2 θ t
− e− 2 θ t θ2 mt = 0
dt 2
Thus,
d  1 2

mt e− 2 θ t = 0
dt
Of course, if the derivative of something is zero, then that something is a constant, so:
1 2
e− 2 θ t E(eθWt ) = C

Where C is a constant.
Now if we use the fact that W0 = 0 then E(eθW0 ) = 1 and C = 1. Our solution to the SDE
then becomes:

1 2
E(eθWt ) = e 2 θ t

19
Please watch the integrating factor review video if you have not done so already!

© 2023 The Infinite Actuary, LLC Page 69

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.4. Consider the following:

Z t
Mt = f (s)dWs
0
1 2
Rt
f (s)2 ds
Xt = eθMt − 2 θ 0

(a) Show that the SDE for Xt is:

dXt = θf (t)Xt dWt

(b) Show that:

 Z t 
2
Mt ∼ N 0, f (s) ds
0

© 2023 The Infinite Actuary, LLC Page 70

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 4

(a) Using the Ito formula for Xt we have:

∂Xt ∂Xt 1 ∂ 2 Xt
dXt = dt + dWt + dWt2
∂t ∂Wt 2 ∂Wt2
   
∂Xt ∂Xt ∂Mt 1 ∂ ∂Xt ∂Mt
= dt + · dWt + · · dt
∂t ∂Mt ∂Wt 2 ∂Wt ∂Mt ∂Wt
    
∂Xt 1 ∂ ∂Xt ∂Mt ∂Xt ∂Mt
= + · · dt + · dWt
∂t 2 ∂Wt ∂Mt ∂Wt ∂Mt ∂Wt

Next, we have a bunch of partial derivatives to compute. To do these, first recall the definitions
θMt − 12 θ2 0t f (s)2 ds
Rt R
we were given in the question stem: Mt = 0 f (s)dWs and Xt = e

Using the definitions above, we can deduce that:

∂Xt 1 2 t
R 2
= θeθMt − 2 θ 0 f (s) ds = θXt
∂Mt
∂Xt 1
= − θ2 f (t)2 Xt
∂t 2
∂Mt
= f (t)
∂Wt
∂Xt ∂Mt
· = θXt f (t)
∂Mt ∂Wt
 
∂ ∂Xt ∂Mt ∂ ∂Xt
· = (θXt f (t)) = θf (t) = θf (t) · [θf (t)Xt ] = θ2 f (t)2 Xt
∂Wt ∂Mt ∂Wt ∂Wt ∂Wt

Therefore, our equation for dXt becomes:

 
1 1
dXt = − θ2 f (t)2 Xt + θ2 f (t)2 Xt dt + θXt · f (t) dWt


2 2
= θf (t)Xt dWt

© 2023 The Infinite Actuary, LLC Page 71

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

(b) Writing the result from part (a) in integral form we have:
Z t Z t
dXs = θf (s)Xs dWs
0 0
Z t
Xt − X0 = θf (s)Xs dWs
0

Taking expectations, we have:

Z t 
E[Xt ] − E[X0 ] = E θf (s)Xs dWs
0
hR i
t
But we know that X0 = 1 and E 0 θf (s)Xs dW s = 0, so:

E[Xt ] = 1
1 2
Rt
f (s)2 ds
E[eθMt − 2 θ 0 ]=1
1 2 t f (s)2 ds
R
E[eθMt ] = e 2 θ 0

This
R t is 2the moment generating function for a normal distribution with mean 0 and variance
f (s) ds. Hence20 :
0

 Z t 
2
Mt ∼ N 0, f (s) ds
0

20
It’s worth pausing for a moment and noting that the result form part (b) relates to the property that the Ito
integral of a deterministic function is normal.

© 2023 The Infinite Actuary, LLC Page 72

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.5. Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener process.
Suppose Xt follows the following generalized SDE:

dXt = µ(Xt , t)dt + σ(Xt , t)dWt

Where µ and σ are functions of Xt and t

If µ(Xt , t) = 0, show that Xt satisfies the key martingale property E(Xt |Fs ) = Xs .

© 2023 The Infinite Actuary, LLC Page 73

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 5

Let’s take the integrals of the SDE:

Z t Z t Z t
dXv = µ(Xv , v)dv + σ(Xv , v)dWv
s s s
Z t Z t
Xt − Xs = µ(Xv , v)dv + σ(Xv , v)dWv
s s

Taking expectations under filtration Fs , we have:

Z t  Z t 
E(Xt − Xs |Fs ) = E µ(Xv , v)dv Fs +E σ(Xv , v)dWv Fs
s s

 
Rt
It is clear that E s σ(Xv , v)dWv Fs = 0, so:

Z t 
E(Xt |Fs ) = Xs + E µ(Xv , v)dv Fs
s

Now if we know that µ(Xt , t) = 0 then:

E(Xt |Fs ) = Xs

So we have shown that Xt is a martingale

Note: Pause and take a minute to reflect on what we just proved. We have basically shown that if
there is no drift in the SDE, we have a martingale. This is a very common trick to use to prove
we have a martingale! Also check out the supplemental FAQ DSG section on martingales for more
info.

© 2023 The Infinite Actuary, LLC Page 74

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.6. (Bachelier Model – Arithmetic Brownian Motion). Let (Ω, F , P) be a probability space
and {Wt : t ≥ 0 } be a standard Wiener process. Suppose Xt follows an arithmetic Brownian
motion with SDE:

dXt = µdt + σdWt

Where µ and σ are constant. Let t < T .

(a) Show that:

XT = Xt + µ(T − t) + σWT −t

Where21 WT −t = WT − Wt ∼ N(0, T − t)

(b) Deduce that XT given Xt = x follows a normal distribution with the following parameters:

E(XT |Xt = x) = x + µ(T − t)

V(XT |Xt = x) = σ 2 (T − t)

21
As a technical footnote, when we write WT −t = WT −Wt we are saying they have the same distribution N(0, T −t)

© 2023 The Infinite Actuary, LLC Page 75

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 6

This one is pretty straightforward, which I think is very appreciated in this tough problem set!

(a) Start by taking the integral of the given equation:

Z T Z T Z T
dXs = µds + σdWs
t t t
XT − Xt = µ(T − t) + σ(WT − Wt )
XT = Xt + µ(T − t) + σWT −t

(b) Since Xt , µ and σ are deterministic (filtered on Ft ), then XT clearly follows a normal distri-
bution as follows:

XT |Xt = x ∼ N[x + µ(T − t), σ 2 (T − t)]

From the above equation, the result clearly follows:

E(XT |Xt = x) = x + µ(T − t)

V(XT |Xt = x) = σ 2 (T − t)

© 2023 The Infinite Actuary, LLC Page 76

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.7. (Black-Scholes Model – Geometric Brownian Motion). Let (Ω, F , P) be a probability

space and {Wt : t ≥ 0 } be a standard Wiener process. Suppose Xt follows the following geometric
Brownian motion with SDE:

dXt = µXt dt + σXt dWt

Where µ and σ are constants.

(a) If we define Yt = log(Xt ), then show that:

 
1
dYt = µ − σ 2 dt + σdWt
2

(b) Show that:

1 2
XT = Xt e(µ− 2 σ )(T −t)+σWT −t

Where WT −t ∼ N(0, T − t)

(c) Given Xt = x, deduce that XT follows a lognormal distribution with the following parameters:

E(XT |Xt = x) = xeµ(T −t)

 2 
V(XT |Xt = x) = x2 eσ (T −t) − 1 e2µ(T −t)

© 2023 The Infinite Actuary, LLC Page 77

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 7

(a) If we set Yt = log(Xt ) and use the Taylor expansion22 , we have:

∂Yt 1 ∂ 2 Yt
dYt = dXt + (dXt )2
∂Xt 2 ∂Xt2
1 1
= (µXt dt + σXt dWt ) − (σ 2 Xt2 dt)
Xt 2Xt2
 
1 2
= µ − σ dt + σdWt
2

(b) Taking the integrals, we have:

Z T Z T  Z T
1 2
d(logXu ) = µ − σ du + σdWu
t t 2 t
 
1 2
log XT − log Xt = µ − σ (T − t) + σ(WT − Wt )
2
 
1 2
log XT = log Xt + µ − σ (T − t) + σ(WT − Wt )
2
1 2
XT = Xt e(µ− 2 σ )(T −t)+σWT −t

Where WT −t ∼ N(0, T − t)
(c) From the result in (b), it is clear that XT |Xt = x is lognormal, since the exponential of a
normal distribution is lognormal.
Recall that if random variable L ∼ Lognormal(µ∗ , σ∗2 ) then:

1 2
E(L) = eµ∗ + 2 σ∗
h 2 i 2
V (L) = eσ∗ − 1 e2µ∗ +σ∗

From the result in (b), we can see that our parameters are:
◦ µ∗ = µ − 12 σ 2 (T − t)

◦ σ∗2 = σ 2 (T − t)
Using this23 gives the desired result:

E(XT |Xt = x) = xeµ(T −t)

 2 
V(XT |Xt = x) = x2 eσ (T −t) − 1 e2µ(T −t)
22
Note the phrase “Taylor expansion” is used quite a bit in the Chin textbook. You can think of this as similar to
applying (1b) of Ito’s Lemma.
23
And, of course, x is a constant so we can use the properties E(kZ) = kE(Z) and V (kZ) = k2 V (Z) for a random
variable Z and constant k

© 2023 The Infinite Actuary, LLC Page 78

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.8. (Generalized Geometric Brownian Motion). Let (Ω, F , P) be a probability space and
{Wt : t ≥ 0 } be a standard Wiener process. Suppose Xt follows the generalized geometric
Brownian motion with SDE:

dXt = µt Xt dt + σt Xt dWt

Where µt and σt are time dependent.

(a) If we define Yt = log(Xt ), then show that:

 
1
dYt = µt − σt2 dt + σt dWt
2

(b) Show that:

RT RT
XT = Xt e[ t (µs − 12 σs2 )ds+ t σs dWs ]

(c) Given Xt = x, deduce that XT follows a lognormal distribution with the following parameters:
RT
E(XT |Xt = x) = xe t µs ds
 RT 2  RT
V(XT |Xt = x) = x2 e t σs ds − 1 e2 t µs ds

© 2023 The Infinite Actuary, LLC Page 79

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 8

This problem is very similar to the prior question, except that we can’t evaluate the integrals by
treating µ and σ as constant.

(a) If we set Yt = log(Xt ) and use the Taylor expansion, we have:

∂Yt 1 ∂ 2 Yt
dYt = dXt + (dXt )2
∂Xt 2 ∂Xt2
1 1
= (µt Xt dt + σt Xt dWt ) − (σ 2 X 2 dt)
Xt 2Xt2 t t
 
1 2
= µt − σt dt + σt dWt
2

(b) Taking the integrals, we have:

Z T Z T  Z T
1 2
d(log Xs ) = µs − σs ds + σs dWs
t t 2 t
Z T  Z T
1
log XT − log Xt = µs − σs2 ds + σs dWs
t 2 t
Z T  Z T
1 2
log XT = log Xt + µs − σs ds + σs dWs
t 2 t
RT RT
XT = Xt e[ t (µs − 12 σs2 )ds+ t σs dWs ]

(c) From the result in (b), it is clear that XT |Xt = x is lognormal, since the exponential of a
normal distribution is lognormal.
Recall that if random variable L ∼ Lognormal(µ∗ , σ∗2 ) then:

1 2
E(L) = eµ∗ + 2 σ∗
h 2 i 2
V (L) = eσ∗ − 1 e2µ∗ +σ∗

From the result in (b), we can see that our parameters are:
RT
◦ µ∗ = t µs − 12 σs2 ds

RT
◦ σ∗2 = t σs2 ds
Using this gives the desired result:
RT
E(XT |Xt = x) = xe t µs ds
 RT 2  RT
V(XT |Xt = x) = x2 e t σs ds − 1 e2 t µs ds

© 2023 The Infinite Actuary, LLC Page 80

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.9. Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener process.
Suppose Xt follows the following geometric Brownian motion with SDE:

dXt = µXt dt + σXt dWt

Where µ and σ are constants.

(a) Show that if Yt = Xtn for a constant n, then Yt follows the geometric Brownian motion with
SDE:
 
1
dYt = n µ + (n − 1)σ 2 Yt dt + nσYt dWt
2

(b) Given Yt , deduce that YT follows a lognormal distribution of the form:

1 2
YT = Yt en(µ− 2 σ )(T −t)+nσWT −t WT −t ∼ N(0, T − t)

(c) Show that the mean and variance are given by:

1 2
E(YT |Yt = y) = yen(µ+ 2 (n−1)σ )(T −t)
2 σ 2 (T −t) 1 2
V(YT |Yt = y) = y 2 (en − 1)e2n(µ+ 2 (n−1)σ )(T −t)

© 2023 The Infinite Actuary, LLC Page 81

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 9

(a) Use Yt = Xtn and a Taylor expansion to get:

∂Yt 1 ∂ 2 Yt
dYt = dXt + (dXt )2
∂Xt 2 ∂Xt2
1
= nXtn−1 dXt + n(n − 1)Xtn−2 (dXt )2
2
1
= nXtn−1 (µXt dt + σXt dWt ) + n(n − 1)Xtn−2 (σ 2 Xt2 dt)
  2
1
= n µ + (n − 1)σ 2 Xtn dt + nσXtn dWt
2
 
1 2
= n µ + (n − 1)σ Yt dt + nσYt dWt
2

(b) Start with the result from part (a):

 
1 2
dYt = n µ + (n − 1)σ Yt dt + nσYt dWt
2

Applying Ito’s Lemma Trick (from the supplemental FAQ DSG) gives:
 
1 2
d ln(Yt ) = n µ − σ dt + nσdWt
2

Integrating gives:
 
1 2
ln(YT ) − ln(Yt ) = n µ − σ (T − t) + nσWT −t
2
 
1 2
ln(YT ) = ln(Yt ) + n µ − σ (T − t) + nσWT −t
2

Thus, YT can be written as:

1 2
YT = Yt en(µ− 2 σ )(T −t)+nσWT −t

It is clear that YT |Yt = y is lognormal, since the exponential of a normal distribution is

lognormal.

© 2023 The Infinite Actuary, LLC Page 82

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

(c) Recall that if random variable L ∼ Lognormal(µ∗ , σ∗2 ) then:

1 2
E(L) = eµ∗ + 2 σ∗
h 2 i 2
V (L) = eσ∗ − 1 e2µ∗ +σ∗

From the result in (b), we can see that our parameters are:
◦ µ∗ = n µ − 12 σ 2 (T − t)

◦ σ∗2 = n2 σ 2 (T − t)
Thus, the mean and variance are given by:

1 2
E(YT |Yt = y) = yen(µ+ 2 (n−1)σ )(T −t)
2 σ 2 (T −t) 1 2
V(YT |Yt = y) = y 2 (en − 1)e2n(µ+ 2 (n−1)σ )(T −t)

© 2023 The Infinite Actuary, LLC Page 83

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.10. (Ornstein-Uhlenbeck Process). Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 }

be a standard Wiener process. Suppose Xt follows the Ornstein-Uhlenbeck process with SDE:

dXt = κ(θ − Xt )dt + σdWt

Where κ, θ and σ are constants.

(a) Using Yt = eκt Xt , show that for t < T :

Z T
XT = Xt e−κ(T −t) + θ 1 − e−κ(T −t) + σe−κ(T −s) dWs
 
t

(b) Deduce that XT given Xt = x follows a normal distribution

(c) Find the mean and variance of XT given Xt = x

© 2023 The Infinite Actuary, LLC Page 84

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 10

(a) Expanding Yt = eκt Xt and using stochastic product rule gives:

d(eκt Xt ) = deκt · Xt + eκt · dXt

= κeκt Xt dt + eκt dXt
= κeκt Xt dt + eκt (κ(θ − Xt )dt + σdWt )
= κθeκt dt + σeκt dWt

Integrating the expression above, we have:

Z T Z T Z T
d(eκs Xs ) = κθeκs ds + σeκs dWs
t t t
Z T Z T
eκT XT − eκt Xt = κθ eκs ds + σ eκs dWs
t t
Z T
eκT XT = eκt Xt + θ eκT − eκt + σ eκs dWs


t
Lastly, multiply both sides by e−κT to get the desired result:
Z T
−κ(T −t)
+ θ 1 − e−κ(T −t) + σe−κ(T −s) dWs
 
XT = Xt e
t

(b) Given the answer to (a), we can see that XT given Xt = x is written as a constant term plus
an integral. Thus, we need to show that the integral is normally distributed.
The intuition for proving this is as follows. Note that the Ito integral of a deterministic
function is normal. This is easy to see because the integral is a sum of increments which are
independent and normally distributed. Thus, the Ito integral of a deterministic function is
normal, which means the integral above is normally distributed. Thus, XT given Xt = x is
normally distributed.
We can also write out the argument stated verbally above mathematically. To prove we have
a normal distribution, note that we can write the final term in the expression for XT as:

Z T n−1
X  
σe−κ(T −s) dWs = lim σe−κ(T −ti ) Wti+1 − Wti
t n→∞
i=0

Where
 
T −t
ti = t + i , t = t0 < t1 < t2 < ... < tn−1 < tn = T
n
Due to the stationary increment of a standard Wiener process, we know that each Wiener
increment is normally distributed and each term in the sum is simply a normally distributed
variable multiplied by a deterministic exponential term. Thus, the summation term is also
normally distributed and it follows that XT given Xt = x is normally distributed.

© 2023 The Infinite Actuary, LLC Page 85

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

(c) Let’s try obtain the mean and variance by first focusing on the integral term:
Z T 
−κ(T −s)
E σe dWs = 0
t
Using Ito’s Isometry and simplifying gives:
"Z
T 2 # T
σ2 
Z 
−κ(T −s) 2 −2κ(T −s)
1 − e−2κ(T −t)

E σe dWs =E σ e ds =
t t 2κ

Given that every other term in our definition of XT given Xt = x in part (a) is deterministic,
we have:

E(XT |Xt = x) = xe−κ(T −t) + θ 1 − e−κ(T −t)

 

σ2 
1 − e−2κ(T −t)

V(XT |Xt = x) =
2κ

Thus, we have shown that:

 2
 
−κ(T −t) −κ(T −t) σ −2κ(T −t)
 
XT |Xt = x ∼ N xe +θ 1−e , 1−e
2κ

© 2023 The Infinite Actuary, LLC Page 86

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.11. (Geometric Mean-Reverting Process).

Note: As a reminder, we stated in the beginning of this problem set that the Chin textbook has
substantial errors for this question. We have re-written this question to show how to correctly walk
through these computations. The Chin textbook asks you to prove parts (c) and (g). However,
to make this easier to work through, we have added in the extra parts to more easily split the
computations into step-by-step parts. Ultimately, in part (g) we derive the correct variance and
Chin gives an incorrect variance formula.

Suppose that Xt follows the SDE:

dXt = κ (θ − log Xt ) Xt dt + σXt dWt

Let Yt = log Xt and Zt = eκt · Yt

Recall that if random variable L ∼ Lognormal(µ∗ , σ∗2 ) then:

1 2
E(L) = eµ∗ + 2 σ∗
h 2 i 2
V (L) = eσ∗ − 1 e2µ∗ +σ∗

(a) Calculate dYt

(b) Calculate dZt

(c) Show that for t < T :

T
σ2
    Z
−κ(T −t) −κ(T −t)
YT = Yt e + θ− · 1−e + σe−κ(T −s) dWs
2κ t

Or, equivalently,

σ2
   Z T
log XT = log Xt e−κ(T −t) + θ − · 1 − e−κ(T −t) + σe−κ(T −s) dWs
2κ t
Z T
(d) You are given that S = σe−κ(T −s) dWs where S is a normally distributed random variable.
t
Calculate the expected value and variance of S.

(e) State the name for the distributions of XT and YT conditioned on the information set at time
t.

(f) Calculate the following:

(i) E(YT |Yt = y)
(ii) V(YT |Yt = y)

© 2023 The Infinite Actuary, LLC Page 87

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

(g) Calculate the following:

(i) E(XT |Xt = x)
(ii) V(XT |Xt = x)

© 2023 The Infinite Actuary, LLC Page 88

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION

(a) Calculate d(Yt )

Note: The solution below shows just one possible approach to this question using Ito’s Lemma.
Keep in mind you could also solve this through other approaches, such as a Taylor expansion.
Yt = Ft = log Xt
From the SDE for Xt give in the question, we know that:
◦ at = κ (θ − log Xt ) Xt
◦ st = σXt
Partial Derivatives:
∂F
◦ ∂t =0
∂F 1
◦ ∂Xt = Xt
∂2F
◦ ∂Xt2
= − X12
t

∂F ∂F 2
dFt = [at ∂Xt
+ ∂t + 21 s2t ∂X
∂ F
2 ]dt +
∂F
∂Xt st dWt
t

d log Xt = [(κ (θ − log Xt ) Xt ) X1t + 12 (σXt )2 · (− X12 )]dt + 1

Xt σXt dWt
t

1
dYt = [κ (θ − log Xt ) − σ 2 ]dt + σdWt
2
(b) Calculate d(Zt )
Note: The solution below shows just one possible approach to this question using Ito’s Lemma.
Keep in mind you could also solve this through other approaches, such as a Taylor expansion.
Zt = Ft = eκt · log Xt
From the SDE for Xt give in the question, we know that:
◦ at = κ (θ − log Xt ) Xt
◦ st = σXt
Partial Derivatives:
∂F
◦ ∂t = κeκt · log Xt
∂F eκt
◦ ∂Xt = Xt
∂2F e κt
◦ ∂Xt2
= −X 2
t

∂F ∂F 2
dFt = [at ∂Xt
+ ∂t + 21 s2t ∂X
∂ F
2 ]dt +
∂F
∂Xt st dWt
t
κt
 κt 
eκt
dFt = [κ (θ − log Xt ) Xt eXt + κeκt · log Xt + 21 (σXt )2 · − X
e
2 ]dt + Xt σXt dWt
t
1 2 κt
dFt = [κ (θ − log Xt ) eκt + κeκt · log Xt − 2 σ e ]dt + σeκt dW t

1
dZt = [κθeκt − σ 2 eκt ]dt + σeκt dWt
2

© 2023 The Infinite Actuary, LLC Page 89

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

(c) Show that for t < T :


σ2
   Z T
−κ(T −t) −κ(T −t)
YT = Yt e + θ− · 1−e + σe−κ(T −s) dWs
2κ t

For part (c), the key is to start with part (b) and integrate both sides from s = t to s = T .
From there, it is just simple algebra.
From part (b):
dZt = d(eκt Yt ) = [κθeκt − 12 σ 2 eκt ]dt + σeκt dWt
Next, integrate both sides from s = t to s = T :
Z T Z T Z T
κs κs 1 2 κs
d(e Yt ) = [κθe − σ e ]ds + σeκs dWs
t t 2 t
Z T Z T
eκT YT − eκt Yt = κθ − 12 σ 2 eκs ds + σeκs dWs


t t
   Z T
2 eκT κt
eκT YT − eκt Y t = κθ − σ
2 κ − e
κ + σeκs dWs
t
  Z T
σ2
eκT YT eκt Y eκT eκt σeκs dWs


− t = θ− 2κ − +
t
  Z T
σ2
eκT YT eκt Yt eκT eκt σeκs dWs


= + θ− 2κ − +
t
  Z T
σ2
e−κ(T −t) e−κ(T −t) σe−κ(T −s) dWs


YT = Yt + θ− 2κ · 1− +
t
Lastly, since Yt = log Xt we also have that:

σ2
   Z T
−κ(T −t) −κ(T −t)
log XT = log Xt e + θ− · 1−e + σe−κ(T −s) dWs
2κ t
Z T
(d) You are given that S = σe−κ(T −s) dWs where S is a normally distributed random variable.
t
Calculate the expected value and variance of S.
S is written in the form of an Ito integral, so E(S) = 0 and therefore V ar(S) = E(S 2 ).
Therefore, the variance calculation boils down to applying Ito’s Isometry to calculate E(S 2 )
Z T 2 Z Th i2 Z T
−κ(T −s) −κ(T −s)
2
Var(S) = E(S ) = E( σe dWs ) = E( σe ds) = E( σ 2 e−2κ(T −s) ds)
Z T t t t
2
2 e−2κ(T −s) ds) = σ 2 · 2κ
1
e−2κ(T −T ) − e−2κ(T −t) = σ2κ 1 − e−2κ(T −t)



= σ E(
t

σ2  
Thus, Var(S) = 1 − e−2κ(T −t)
2κ
(e) State the name for the distributions of XT and YT conditioned on the information set at time
t.
From parts (c) and (d), we know that YT can be written as a normally distributed variable
S plus a constant.

© 2023 The Infinite Actuary, LLC Page 90

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

Therefore, we know that YT is normally distributed.

This implies that XT is lognormal.

(f) Calculate the following:

(i) E(YT |Yt = y)
(ii) V(YT |Yt = y)
For part (f), we have really done all of the hard work already and this just boils down to
plugging in prior results:

T
σ2
    Z
−κ(T −t)
YT = Yt e + θ− · 1 − e−κ(T −t) + σe−κ(T −s) dWs
2κ t

Finally, calculate the moments of the equation above, plugging in the results from part (d).
σ2
   
−κ(T −t)
Thus, E(YT |Yt = y) = ye + θ− · 1 − e−κ(T −t)
2κ

σ2  
And, V(YT |Yt = y) = 1 − e−2κ(T −t)
2κ
(g) Calculate the following:
(i) E(XT |Xt = x)
(ii) V(XT |Xt = x)
For part (g), we want to leverage the following equations given in the question:

1 2
E(L) = eµ∗ + 2 σ∗
h 2 i 2
V (L) = eσ∗ − 1 e2µ∗ +σ∗

First, solve for the parameters:

 2

µ∗ = log x · e−κ(T −t) + θ − σ2κ · 1 − e−κ(T −t)

2
σ∗2 = σ2κ 1 − e−2κ(T −t)

It is a little messy, but the rest is just plugging in the parameters:

2 2
 
µ∗ + 21 σ∗2 log x·e−κ(T −t) + θ− σ2κ ·(1−e−κ(T −t) )+ σ4κ (1−e−2κ(T −t) )
E(XT |Xt = x) = e =e
h 2 i 2
V(XT |Xt = x) = eσ∗ − 1 e2µ∗ +σ∗
2 2
 2  h   i
σ
(1−e−2κ(T −t) ) 2 log x·e−κ(T −t) + θ− σ2κ ·(1−e−κ(T −t) ) + σ2κ (1−e−2κ(T −t) )
= e 2κ −1 ·e

Notice how this is different from Chin. Chin has a mistake, for example, the “-1” term in the
first line of the variance formula in Chin should not be in the set brackets of the exponential.

© 2023 The Infinite Actuary, LLC Page 91

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.12. (Cox-Ingersoll-Ross (CIR) Model). Let (Ω, F , P) be a probability space and {Wt : t ≥ 0}
be a standard Wiener process. Suppose Xt follows the Cox-Ingersoll-Ross (CIR) Model with SDE:

p
dXt = κ(θ − Xt )dt + σ Xt dWt , X0 > 0

Where κ, θ and σ are constants.

(a) Is Xt normally distributed? Lognormally distributed?

(b) Using Zt = eκt Xt , and taking the integrals from t to T , show that:

h i Z T
−κ(T −t) −κ(T −t)
σe−κ(T −s)
p
XT = Xt e +θ 1−e + Xs dWs
t

(c) Using Zt2 = e2κt Xt2 , and taking the integrals from t to T , show that:
Z T Z T
XT2 = Xt2 e−2κ(T −t) 2
+ (2κθ + σ ) e −2κ(T −s)
Xs ds + 2σ e−2κ(T −s) Xs1.5 dWs
t t

(d) Find the mean and variance of XT given Xt = x

Note: For part (d), you may find the following hints helpful:
RT
◦ t e−2κ(T −s) xe−κ(s−t) + θ 1 − e−κ(s−t) ds = κ1 (x − θ) e−κ(T −t) − e−2κ(T −t) + θ
1 − e−2κ(T −t)

 


2

 2
 2)
◦ x2 e−2κ(T −t) + 2κθ+σ (x − θ) e−κ(T −t) − e−2κ(T −t) + θ(2κθ+σ 1 − e−2κ(T −t)



κ 2κ
2
− xe−κ(T −t) + θ 1 − e−κ(T −t)
 
2 2
= xσκ e−κ(T −t) − e−2κ(T −t) + θσ −κ(T −t) + e−2κ(T −t)



2κ 1 − 2e

© 2023 The Infinite Actuary, LLC Page 92

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 12

(a) Note that X is not normally distributed and also not lognormally distributed. X has a CIR
distribution as stated in the question. One way we can tell
√ that X is not normal/lognormal is
by looking at the innovation term in its SDE. It has a Xt term, which is not characteristic
of normal/lognormal SDEs.

(b) Expanding Zt = eκt Xt using stochastic product rule gives:

d(eκt Xt ) = deκt · Xt + eκt · dXt

 p 
= κeκt Xt dt + eκt κ(θ − Xt )dt + σ Xt dWt
p
= κθeκt dt + σeκt Xt dWt

Integrating both sides of the equation, we have:

Z T Z T Z T p
κs κs
d(e Xs ) = κθe ds + σeκs Xs dWs
t t t
Z T p
XT eκT − Xt eκt = θ[eκT − eκt ] + σeκs Xs dWs
t
Z T p
XT eκT = Xt eκt + θ[eκT − eκt ] + σeκs Xs dWs
t
h i Z T
XT = Xt e−κ(T −t) + θ 1 − e−κ(T −t) + σe−κ(T −s)
p
Xs dWs
t

(c) Expanding Zt2 = e2κt Xt2 and applying stochastic product rule24 , we have that:

d(e2κt Xt2 ) = de2κt · Xt2 + e2κt · d(Xt2 )

= 2κe2κt Xt2 dt + e2κt (2Xt dXt + (dXt )2 )
h p i
= 2κe2κt Xt2 dt + 2e2κt Xt κ(θ − Xt )dt + σ Xt dWt + e2κt (σ 2 Xt dt)
= e2κt 2κθ + σ 2 Xt dt + 2σe2κt Xt1.5 dWt

Taking the integrals on both sides, we have:

Z T Z T Z T
2κs
Xs2 ) 2 2κs
e2κs Xs1.5 dWs


d(e = 2κθ + σ e Xs ds + 2σ
t t t

24
Here, we also used the fact from stochastic product rule that d(Xt2 ) = d(Xt · Xt ) = Xt dXt + Xt dXt + (dXt )2 =
2Xt dXt + (dXt )2

© 2023 The Infinite Actuary, LLC Page 93

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

Now, similar to part (b), we just need to integrate and rearrange to get the desired solution:
Z T Z T
2κT
XT2 2κt
Xt2 2 2κs
e2κs Xs1.5 dWs


e =e + 2κθ + σ e Xs ds + 2σ
t t

Z T Z T
XT2 = Xt2 e−2κ(T −t) + (2κθ + σ ) 2
e −2κ(T −s)
Xs ds + 2σ e−2κ(T −s) Xs1.5 dWs
t t

(d) Note that for this part, we don’t have a normal/lognormal random variable, so it involves a
bit more work to solve for our moments. The basic idea is we will leverage the results from
the prior two parts.
Similar to previous problems, under the expectation term, we can drop the integral relating
to dWs since this will be equal to 0 on expectation:

h i
E(XT |Xt = x) = xe−κ(T −t) + θ 1 − e−κ(T −t)

To get the variance, we must find the expectation of XT2 with t ≤ s ≤ T :

RT
E(XT2 |Xt = x) = x2 e−2κ(T −t) + (2κθ + σ 2 ) t e−2κ(T −s) E(Xs |Xt = x)ds

Next, plug in the results from part (b) to get:

RT
= x2 e−2κ(T −t) + (2κθ + σ 2 ) e−2κ(T −s) xe−κ(s−t) + θ 1 − e−κ(s−t) ds


t

Now, use the hint given in the question to simplify:

 
2κθ+σ 2 θ(2κθ+σ 2 )
= x2 e−2κ(T −t) + (x − θ) e−κ(T −t) − e−2κ(T −t) + 1 − e−2κ(T −t)



κ 2κ

Now since we know:

V(XT |Xt = x) = E(XT2 |Xt = x) − [E(XT |Xt = x)]2

Using the above and the second hint given in the question yields:

xσ 2  −κ(T −t)  θσ 2  
V(XT |Xt = x) = e − e−2κ(T −t) + 1 − 2e−κ(T −t) + e−2κ(T −t)
κ 2κ

© 2023 The Infinite Actuary, LLC Page 94

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.13. (Brownian Bridge Process). Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be
a standard Wiener process. Suppose Xt follows the Brownian bridge process with SDE:

y − Xt
dXt = dt + dWt , X1 = y
1−t

(a) Using

y − Xt
Yt =
1−t
Take the integrals and show that under initial condition, X0 = x and for 0 ≤ t ≤ 1:
 Z t 
1
Xt = yt + (1 − t) x + dWs
0 1−s

(b) Show that Xt follows a normal distribution and find the mean and variance for Xt

(c) State the value of V (Xt ) at times t = 0 and t = 1. What does this mean?

Note: The Chin textbook only asks you to do parts (a) and (b), but we at TIA added in part (c)
because we think it gives you some good practice with the intuition around Brownian bridges. Feel
free to work through it for bonus practice!

© 2023 The Infinite Actuary, LLC Page 95

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 13

(y−Xt )
(a) Expanding Yt = (1−t) using Ito’s formula and the relevant Taylor expansion terms, we have:

∂Yt ∂Yt 1 ∂ 2 Yt
d(Yt ) = dt + dXt + (dXt )2
∂t ∂Xt 2 ∂Xt2
 

y − Xt  1   y − X 
t
= dt −  dt + dWt  + 0
 
(1 − t)2  1−t 1−t 
| {z }
dXt
 
1
=− dWt
1−t

Integrating both sides of the equation, we have:

Z t Z t 
1
dYs = − dWs
0 0 1−s
Z t 
1
Yt − Y0 = − dWs
0 1−s

Now if we substitute for Yt and Y0 (using the fact that X0 = x) we get:

 Z t
y − Xt y − x 1
− =− dWs
1−t 1−0 0 1−s
Z t 
1
y − Xt − (1 − t)(y − x) = −(1 − t) dWs
0 1−s
Z t 
1
−Xt + (1 − t)x + ty = −(1 − t) dWs
0 1−s
 Z t 
1
Xt = yt + (1 − t) x + dWs
0 1−s

(b) In order to prove that Xt follows a normal distribution, we refer to 3.2.2.4, where we found
that equations of the following form follow a normal distribution:
Z t
f (s)dWs
0

Clearly, in our case, the final integral term follows this form:
Z t
1
dWs
0 1−s
Therefore, we conclude that Xt follows a normal distribution.

© 2023 The Infinite Actuary, LLC Page 96

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

Now, we can try to find the mean and variance of this variable by using the following two
facts25 :
Z t
1
E dWs = 0
0 1−s
"Z 2 #
t Z t 
1 1 t
E dWs =E 2
ds =
0 1−s 0 (1 − s) 1 − t

Since we know from (a) that:

 Z t 
1
Xt = yt + (1 − t) x + dWs
0 1−s

We can then see that the mean and variance26 of Xt are:

E(Xt |X0 = x) = yt + x(1 − t)

"Z 2 #
t
2 1
V(Xt |X0 = x) = (1 − t) × E dWs = t(1 − t)
0 1−s

(c) Plugging into the result from part (b), we see for t = 0 and t = 1:

V(Xt |X0 = x) = t(1 − t) = 0

In other words, the variance equals 0. This is not surprising, because the Brownian bridge is
fixed at these times by construction since X0 = x and X1 = y.

Warning: Please note the Chin textbook has an error and incorrectly gets a variance formula
for part (b) of:

1
V(Xt |X0 = x) =
1−t
Intuitively, we know this result makes no sense because it fails to give a variance of 0 at the
fixed endpoints of the bridge at times 0 and 1.
For reference, where they went wrong is they said:
"Z 2 #
t
1 1
E dWs =
0 1−s 1−t
t
instead of 1−t
25
R tWhere in theR last step, we can Rcompute this integral using u-substitution with u = (1 − s) ⇒ du = −ds:
1−t 1 1
1
0 (1−s)2
ds = 1 u2
· (−du) = 1−t u−2 du = − u1 |u=1 1 t−1 1
u=1−t = −1 + 1−t = 1−t + 1−t = 1−t
t

26
Note that for the variance equation, we used the fact that V (X) = E(X 2 ) − E(X)2 = E(X 2 ) where X is the
integral term and we know E(X) = 0.

© 2023 The Infinite Actuary, LLC Page 97

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.14. (Forward Curve from an Asset Price Following a Geometric Brownian Motion). Let
(Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener process. Suppose an asset
price St at time t follows a geometric Brownian motion such that:

dSt = µSt dt + σSt dWt

where µ is the drift parameter and σ is the volatility. Let’s also define the forward price F (t, T )
as an agreed-upon price set at time t to be paid or received at time T, t ≤ T and is given by the
relationship:

F (t, T ) = E(ST |Ft )

Show that the forward curve follows:

dF (t, T )
= σdWt
F (t, T )

© 2023 The Infinite Actuary, LLC Page 98

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 14

Given St follows a geometric Brownian motion, we can use the result from 3.2.2.7 to show that:

E(ST |Ft ) = St eµ(T −t) = F (t, T )

For Ito’s Lemma, we will use the following partial derivatives:

• ∂F
∂t = −µF

• ∂F
∂St = eµ(T −t)
∂2F
• ∂St2
=0

Now we can expand F (t, T ) using Ito’s lemma:

∂F ∂F 1 ∂2F
dF (t, T ) = dt + dSt + (dSt )2
∂t ∂St 2 ∂St2
= −µSt eµ(T −t) dt + eµ(T −t) [µSt dt + σSt dWt ]
= σSt eµ(T −t) dWt = σF (t, T )dWt

Thus, we have shown the desired result that:

dF (t, T )
= σdWt
F (t, T )

© 2023 The Infinite Actuary, LLC Page 99

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.15. (Forward Curve from an Asset Price Following a Geometric Mean-Reverting Process).
Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a standard Wiener process. Suppose an
asset price St at time t follows a geometric mean-reverting process such that:

dSt
= κ(θ − log St )dt + σdWt
St

where κ is the mean reversion rate, θ is the long term mean and σ is the volatility parameter. Let’s
also define the forward price F (t, T ) as an agreed-upon price set at time t to be paid or received
at time T, t ≤ T and is given by the relationship:

F (t, T ) = E(ST |Ft )

Show that the forward curve follows:

dF (t, T )
= σe−κ(T −t) dWt
F (t, T )

© 2023 The Infinite Actuary, LLC Page 100

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 15

Given St follows a geometric mean-reverting process, we can use the result from 3.2.2.11 to know
that:

2 2
 
log St ·e−κ(T −t) + θ− σ2κ ·(1−e−κ(T −t) )+ σ4κ (1−e−2κ(T −t) )
E(ST |St ) = e = F (t, T )

Next, let’s compute partial derivatives:

h   i
• ∂F = κe−κ(T −t) log S − θ − σ 2 κe−κ(T −t) − σ 2 −2κ(T −t)
∂t t 2κ 2 e F (t, T )

e−κ(T −t)
• ∂F
∂St = St F (t, T )
h −κ(T −t) −2κ(T −t) i
∂2F
• ∂St2
= −e +e
S2
F (t, T )
t

Now if we expand F (t, T ), we have:

∂F ∂F 1 ∂2F
dF (t, T ) = dt + dSt + (dSt )2
∂t ∂St 2 ∂St2
σ2 σ 2 −2κ(T −t)
   
−κ(T −t) −κ(T −t)
= κe log St − θ − κe − e F (t, T )dt
2κ 2
" #
e−κ(T −t) 1 e−κ(T −t) + e−2κ(T −t)
+ F (t, T )dSt + − F (t, T ) · σ 2 St2 dt
St 2 St2
σ2 σ 2 −2κ(T −t)
   
dF (t, T ) −κ(T −t) −κ(T −t)
= κe log St − θ − κe − e dt
F (t, T ) 2κ 2
σ2 σ2
+ e−κ(T −t) [κ(θ − log St )dt + σdWt ] − e−κ(T −t) dt + e−2κ(T −t) dt
2 2
= σe−κ(T −t) dWt

© 2023 The Infinite Actuary, LLC Page 101

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.16. (Forward-Spot Price Relationship I). Let (Ω, F , P) be a probability space and {Wt : t ≥
0} be a standard Wiener process. We define the forward curve F (t, T ) following the SDE:

dF (t, T )
= σ(t, T )dWt
F (t, T )

as an agreed-upon price of an asset with the current spot price St to be paid or received at time
T, t ≤ T where:

F (t, T ) = E(ST |Ft )

such that ST is the spot price at time T and σ(t, T ) is a time-dependent volatility

Show that the spot price has the following SDE:

 Z t Z t 
dSt ∂ log F (0, t) ∂σ(u, t) ∂σ(u, t)
= − σ(u, t) du + dWu dt + σ(t, t)dWt
St ∂t 0 ∂t 0 ∂t

© 2023 The Infinite Actuary, LLC Page 102

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 16

Let’s expand log F (t, T ) using Ito’s Lemma:

1 1
d log F (t, T ) = dF (t, T ) − dF (t, T )2
F (t, T ) 2F (t, T )2
1
= σ(t, T )dWt − σ(t, T )2 dt
2

Now if we take the integrals on both sides, we have:

t t
1 t
Z Z Z
d log F (u, T ) = σ(u, T )dWu − σ(u, T )2 du
0 0 2 0
Z t
1 t
Z
log F (t, T ) = log F (0, T ) + σ(u, T )dWu − σ(u, T )2 du
0 2 0
1
Rt Rt
σ(u,T )2 du+
F (t, T ) = F (0, T )e− 2 0 0 σ(u,T )dWu

Now if we use T = t then we can set the spot price St = F (t, t) and our formula becomes:

1
Rt Rt
σ(u,t)2 du+
St = F (0, t)e− 2 0 0 σ(u,t)dWu

Now, let’s try obtaining an equation for dSt based on Ito’s lemma:

∂St 1 ∂ 2 St
 
∂St
dSt = + 2 dt + dWt
∂t 2 ∂Wt ∂Wt

To plug into the equation above, first we need to compute partial derivatives.
∂St
Let’s start by computing ∂Wt :
Z t 
∂St ∂
= σ(u, t)dWu St = σ(t, t)St
∂Wt ∂Wt 0

∂ 2 St
Next, calculating ∂Wt2
gives:

∂ 2 St
= σ(t, t)2 St
∂Wt2

© 2023 The Infinite Actuary, LLC Page 103

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

∂St
The partial for ∂t takes the most work. Let’s denote the exponential term as A so that S = F · A:

1
Rt Rt
σ(u,t)2 du+
St = F (0, t) |e− 2 0
{z
0 σ(u,t)dWu
}
A

Here, we are temporarily dropping the timescripts for notational simplicity. Then:

∂S ∂F ∂A ∂F S ∂A S
= ·A+F · = · + ·
∂t ∂t ∂t ∂t F ∂t A

∂A
∂S ∂F S
Thus, we need to compute ∂t = ∂t · F + ∂t
A S, which gives27 :

 Z t  Z t 
∂St ∂F (0, t) −1 1∂ 2 ∂σ(u, t)
= F (0, t) St + − σ(u, t) du + dWu St
∂t ∂t 2 ∂t 0 0 ∂t
 Z t Z t 
∂F (0, t) −1 1 2 ∂σ(u, t) ∂σ(u, t)
= F (0, t) St − σ(t, t) + 2σ(u, t) du − 2 dWu St
∂t 2 0 ∂t 0 ∂t
 Z t Z t 
∂ log F (0, t) 1 2 ∂σ(u, t) ∂σ(u, t)
= − σ(t, t) − σ(u, t) du + dWu St
∂t 2 0 ∂t 0 ∂t

Substituting these values into our original equation for dSt , we have:

∂St 1 ∂ 2 St
 
∂St
dSt = + 2 dt + dWt
∂t 2 ∂Wt ∂Wt
 Z t Z t  
∂ log F (0, t) 1 2 ∂σ(u, t) ∂σ(u, t) 1 2
= − σ(t, t) − σ(u, t) du + dWu St + σ(t, t) St dt
∂t 2 0 ∂t 0 ∂t 2
+ σ(t, t)St dWt

Simplifying yields the desired result:

 Z t Z t 
dSt ∂ log F (0, t) ∂σ(u, t) ∂σ(u, t)
= − σ(u, t) du + dWu dt + σ(t, t)dWt
St ∂t 0 ∂t 0 ∂t

27
hRNote that we i are usingR the deterministic calculus fact proved in the Leibniz drill problem set that
t t
∂
∂t 0
σ(u, t) 2
du = σ(t, t)2 + 0 2σ(u, t) ∂σ(u,t)
∂t
du

© 2023 The Infinite Actuary, LLC Page 104

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.17. (Clewlow-Strickland 1-Factor Model). Let (Ω, F , P) be a probability space and {Wt :
t ≥ 0 } be a standard Wiener process. We define the forward curve F (t, T ) following the process

dF (t, T )
= σe−α(T −t) dWt
F (t, T )

where t ≤ T , α is the mean reversion parameter and σ is the volatility. We also know that

F (t, T ) = E(ST |Ft )

Where ST is the spot price at time T .

(a) Show that:

Z t Z t
−α(t−u) 1
σe dWu = log St − log F (0, t) + σ 2 e−2α(t−u) du
0 2 0

(b) Show that:

σ2
 
dSt ∂ log F (0, t) −2αt
= + α(log F (0, t) − log St ) + (1 − e ) dt + σdWt
St ∂t 4

(c) Show that the forward curve at time t is given by

 e−α(T −t)
St σ 2 −αT 2αt
(e −1)(e−αt −e−αT )
F (t, T ) = F (0, T ) e 4α e
F (0, t)

(d) Show that F (t, T ) given F (0, T ) follows a lognormal distribution with the mean and variance
below:

E[F (t, T )|F (0, T )] = F (0, T )

σ2 −2α(T −t) −e−2αT ]−1
V[F (t, T )|F (0, T )] = F (0, T )2 e 2α [e

© 2023 The Infinite Actuary, LLC Page 105

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 17

(a) We are given that:

dF (t, T )
= σe−α(T −t) dWt
F (t, T )
Next, let’s compute d log F (t, T ):

1 1
d log F (t, T ) = dF (t, T ) − dF (t, T )2
F (t, T ) 2F (t, T )2
1
= σe−α(T −t) dWt − σ 2 e−2α(T −t) dt
2

Now if we take the integrals on both sides, we have:

Z t Z t Z t
1 2 −2α(T −u)
−α(T −u)
d log F (u, T )du = σe dWu −
σ e du
0 0 0 2
1 t 2 −2α(T −u)
Z Z t
log F (t, T ) = log F (0, T ) − σ e du + σe−α(T −u) dWu
2 0 0
If we set T = t such that F (t, t) = St we can write:

1 t 2 −2α(t−u)
Z Z t
log St = log F (0, t) − σ e du + σe−α(t−u) dWu
2 0 0
Z t
1 t 2 −2α(t−u)
Z
−α(t−u)
σe dWu = log St − log F (0, t) + σ e du
0 2 0
Let’s keep this result in our back pocket - we will need it later in the problem.
(b) Next, we are going to leverage what we already showed in the prior problem 3.2.2.16. We will
start off by stating the result of the prior question, then plugging in our volatility function
to simplify and calculate partial derivatives. As a reminder, for 3.2.2.16 we had that if:
dF (t, T )
= σ(t, T )dWt
F (t, T )
where σ(t, T ) is a time-dependent volatility function, the corresponding spot price SDE is:
 Z t Z t 
dSt ∂ log F (0, t) ∂σ(u, t) ∂σ(u, t)
= − σ(u, t) du + dWu dt + σ(t, t)dWt
St ∂t 0 ∂t 0 ∂t

Now in our case, σ(t, T ) = σe−α(T −t) so we can calculate the partial differential with respect
to T :
∂σ(t, T )
= −ασe−α(T −t)
∂T
dSt
Thus, from our equation for St we can evaluate the integral terms:
t t t
σ2 σ2
Z Z Z
∂σ(u, t) 2 −2α(t−u)
2αe−2α(t−u) du = − 1 − e−2αt


σ(u, t) du = − ασ e du = −
0 ∂t 0 2 0 2

© 2023 The Infinite Actuary, LLC Page 106

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

and Z t Z t
∂σ(u, t)
dWu = − ασe−α(t−u) dWu
0 ∂t 0
So now we have:
Z t
∂ log F (0, t) σ 2
 
dSt −2αt

−α(t−u)
= + 1−e − ασe dWu dt + σ(t, t)dWt
St ∂t 2 0

In order to solve the integral term, we turn to our earlier Ito’s Lemma derivation of d log F (t, T )
that was mentioned to keep in our back pocket from part (a):
Z t
1 t 2 −2α(t−u)
Z
−α(t−u)
σe dWu = log St − log F (0, t) + σ e du
0 2 0

Multiplying through by α we get:

t
1 t 2 −2α(t−u)
Z Z
−α(t−u)
ασe dWu = α[log St − log F (0, t)] + ασ e du
0 2 0
σ2 t
Z
= α[log St − log F (0, t)] + 2αe−2α(t−u) du
4 0
σ2
1 − e−2αt


= α[log St − log F (0, t)] +
4
dSt
We can substitute this in to our solution for St :

∂ log F (0, t) σ 2 σ2
 
dSt −2αt

−2αt


= + 1−e − α[log St − log F (0, t)] − 1−e dt+σ(t, t)dWt
St ∂t 2 4
Recognizing that σ(t, t) = σ and simplifying yields our result:

σ2
 
dSt ∂ log F (0, t) −2αt
= + α(log F (0, t) − log St ) + (1 − e ) dt + σdWt
St ∂t 4

(c) From part (a), when we derived the Ito’s lemma expression for d log F (t, T ), we found that:
Z t Z t
1 2 −2α(T −u)
log F (t, T ) = log F (0, T ) − σ e du + σe−α(T −u) dWu
2 0 0
1
Rt Rt
σ 2 e−2α(T −u) du+ σe−α(T −u) dWu
F (t, T ) = F (0, T )e− 2 0 0

We can evaluate the first integral in the exponent quite easily:

t
σ 2 −2αT 2αt
Z
σ 2 e−2α(T −u) du =


e e −1
0 2α
For the second integral, we must first recall the identity we proved in part (a):
Z t Z t
−α(t−u) 1
σe dWu = log St − log F (0, t) + σ 2 e−2α(t−u) du
0 2 0

© 2023 The Infinite Actuary, LLC Page 107

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

With this in mind, we can try and evaluate the second integral:

Z t Z t
σe−α(T −u) dWu = e−α(T −t) σe−α(t−u) dWu
0
0
1 t 2 −2α(t−u)
Z 
−α(T −t)
=e log St − log F (0, t) + σ e du
2 0
σ2
   
−α(T −t) St −2αt


=e log + 1−e
F (0, t) 4α

Therefore, our exponent integral terms can now be reduced to:

t t
−σ 2 −2αT 2αt
Z Z
1 2 −2α(T −u)
σe−α(T −u) dWu =


− σ e du + e e −1
2 0 0 4α
σ2
   
−α(T −t) St −2αt


+e log + 1−e
F (0, t) 4α
 
St
= e−α(T −t) log
F (0, t)
2
σ  −αT +αt
− e−αT −αt − e2αt−2αT + e−2αT

+ e
4α  
−α(T −t) St
=e log
F (0, t)
2
σ −αT  αt
e − e−αt − e2αt−αT + e−αT

+ e
4α  
−α(T −t) St
=e log
F (0, t)
2
σ −αT  2αt
(e − 1)(e−αt − e−αT )

+ e
4α

Finally, substituting this into our original equation for F (t, T ) we get:

 e−α(T −t)
St σ 2 −αT 2αt
(e −1)(e−αt −e−αT )
F (t, T ) = F (0, T ) e 4α e
F (0, t)

(d) For this final part, let’s go back to our Ito’s lemma expression for d log F (t, T ):

1
d log F (t, T ) = σe−α(T −t) dWt − σ 2 e−2α(T −t) dt
2
Taking integrals, we have:

t t
σ2 σ 2 h −2α(T −t)
Z Z
1 2 −2α(T −u) i
σ e du = 2α · e−2α(T −u) du = e − e−2αT
0 2 4α 0 4α

t
σ 2 h −2α(T −t)
Z i
log F (t, T ) = log F (0, T ) + σe−α(T −u) dWu − e − e−2αT
0 4α

© 2023 The Infinite Actuary, LLC Page 108

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

Now, the first step to this is proving that log F (t, T ) follows
R t a normal distribution. This can
be shown using the result from problem 3.2.2.4. Since 0 σe−α(T −u) dWu is in the form of
Rt
0 f (u)dWu we can conclude that the integral follows a normal distribution.
So all that’s left is finding the normal parameters for log F (t, T ).
Computing the mean parameter is straightforward:
Z t 
−α(T −u)
E σe dWu = 0
0

σ 2 h −2α(T −t) i
E[log F (t, T )] = log F (0, T ) − e − e−2αT
4α
To find the variance of log F (t, T ), we need to get the expected value of the squared integral
term:
"Z 2 #
t t
σ 2 −2α(T −t)
Z 
−α(T −u) 2 −2α(T −u)
E σe dWu =E σ e du = [e − e−2αT ]
0 0 2α

So at this point, we have proved28 that log F (t, T ) follows a normal distribution and have
found the expected value and variance of log F (t, T ):

σ 2 h −2α(T −t) σ 2 −2α(T −t)

 i 
−2αT 2 −2αT
log F (t, T ) ∼ N µ∗ = log F (0, T ) − e −e , σ∗ = [e −e ]
4α 2α

This implies that F (t, T ) follows a lognormal distribution. Recall that if random variable
L ∼ Lognormal(µ∗ , σ∗2 ) then:

1 2
E(L) = eµ∗ + 2 σ∗
h 2 i 2
V (L) = eσ∗ − 1 e2µ∗ +σ∗

Plugging in our parameter values and simplifying then yields our desired result:

E[F (t, T )|F (0, T )] = F (0, T )

σ2 −2α(T −t) −e−2αT ]−1

V[F (t, T )|F (0, T )] = F (0, T )2 e 2α [e

28
Note: Throughout the solutions to part (d), you can assume we are conditioning on F (0, t).

© 2023 The Infinite Actuary, LLC Page 109

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.18. (Constant Elasticity of Variance Model). Let (Ω, F , P) be a probability space and
{Wt : t ≥ 0 } be a standard Wiener process. Suppose an asset price St > 0 follows a constant
elasticity of variance (CEV) model of the form:

dSt = rSt dt + σ(St , t)St dWt

where r is a constant and σ(St , t) is a local volatility function.

Let σ(St , t) = αStβ−1 with α > 0 and 0 < β < 1 .

(a) Using Ito’s formula, show that:

  
dσ(St , t) 1 2
= (β − 1) r + (β − 2)σ(St , t) dt + σ(St , t)dWt
σ(St , t) 2

(b) Show that for t < T , ST conditional on St is given by:

1
 Z T  1−β
ST = erT er(1−β)t St1−β + α(1 − β) e−r(1−β)µ dWu
t

Note: For (b), you may find it helpful to define Xt = e−rt St and perform a Taylor expansion
on dXt

© 2023 The Infinite Actuary, LLC Page 110

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 18

(a) This is a simple derivation using a Taylor expansion with σ(St , t) = αStβ−1 :

∂σ(St , t) 1 ∂ 2 σ(St , t)
dσ(St , t) = dSt + (dSt )2
∂St 2 ∂St2
1
= α(β − 1)Stβ−2 dSt + α(β − 1)(β − 2)Stβ−3 (dSt )2
2
1
= α(β − 1)St (rSt dt + σ(St , t)St dWt ) + α(β − 1)(β − 2)Stβ−3 (σ(St , t)2 St2 (dWt )2 )
β−2
2
1 2(β−1)
= α(β − 1)Stβ−2 (rSt dt + α(Stβ−1 )St dWt ) + α(β − 1)(β − 2)Stβ−3 ((α2 St )St2 dt)
2
1
= α(β − 1)Stβ−1 rdt + α2 (β − 1)St2β−2 dWt + α3 (β − 1)(β − 2)St3β−3 dt
  2 
1 3 2
= (β − 1) rσ(St , t) + (β − 2)σ(St , t) dt + σ(St , t) dWt
2

Therefore, we have shown that:

  
dσ(St , t) 1 2
= (β − 1) r + (β − 2)σ(St , t) dt + σ(St , t)dWt
σ(St , t) 2

(b) To find the solution of the CEV model, we let Xt = e−rt St and use Taylor’s expansion:

∂Xt ∂Xt 1 ∂ 2 Xt
dXt = dt + dSt + (dSt )2
∂t ∂St 2 ∂St2
 
= −re−rt St dt + e−rt rSt dt + αStβ dWt + 0

= αe−rt Stβ dWt

= αe−r(1−β)t Xtβ dWt

Taking the integrals on both sides29 , we get:

Z T Z T
dXu
=α e−r(1−β)u dWu
t Xuβ t
Z T
1 h
1−β 1−β
i
XT − Xt =α e−r(1−β)u dWu
1−β t
1
 Z T  1−β
1−β
XT = Xt + α(1 − β) e−r(1−β)u dWu
t
1
 Z T  1−β
ST = e rT
er(1−β)t St1−β + α(1 − β) −r(1−β)µ
e dWu
t
29 1
As a warning, the textbook incorrectly evaluates the integral on the left hand side and forgets about the 1−β
term from integration. This is corrected in this file but this mistake is in the Chin textbook.

© 2023 The Infinite Actuary, LLC Page 111

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.2.19. (Geometric Average). Let (Ω, F , P) be a probability space and {Wt : t ≥ 0 } be a

standard Wiener process. Suppose an asset price St follows the geometric Brownian motion model
of the form:

dSt = µSt dt + σSt dWt

where µ and σ are constant parameters. Consider the geometric average30 of St :

1
Rt
log(Su )du
Gt = e t 0 G0 = S0

(a) Show that Gt satisfies the following SDE:

 
dGt 1 1 2 σ
= µ − σ dt + √ dW̃t
Gt 2 6 3
where
√ Z t
3
W̃t = Wu du
t 0

(b) Is W̃t a standard Wiener process?

(c) Under what conditions is this SDE from part (a) valid31 ?

30
Warning: This is another question where the Chin textbook has a mistake. The definition of Gt should say
log(Su ) instead of just Su . This has been corrected in this file.
31
Here by “valid”, we mean that it is an SDE written in the form of an Ito process such that W̃t is a standard
Wiener process.

© 2023 The Infinite Actuary, LLC Page 112

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 123-147, Question 19

(a) We know from problem 3.2.2.7:

1 2
Su = S0 e(µ− 2 σ )u+σWu

or
 
1
log Su = log S0 + µ − σ 2 u + σWu
2

So if we take the logarithm of Gt , we get:

1 t
Z
log Gt = log Su du
t 0
Z t   
1 1 2
= log S0 + µ − σ u + σWu du
t 0 2
Z t Z t
σ t
 Z
1 1 1 2
= log S0 du + µ − σ u du + Wu du
t 0 t 0 2 t 0
σ t
  Z
1 1 2
= log S0 + µ− σ t+ Wu du
2 2 t 0

Next, from problem 3.2.1.13, we can deduce that:

Z t  3
t
Wu du ∼ N 0,
0 3

In our case, we can say

Z t  
σ 1
Wu du ∼ N 0, σ 2 t
t 0 3

We can state this slightly differently:

  √ Z t
σ 1 2 3
√ W̃t ∼ N 0, σ t where W̃t = Wu du ∼ N(0, t)
3 3 t 0

Therefore, returning to our original equation, we have:

 
1 1 2 σ
log Gt = log S0 + µ − σ t + √ W̃t
2 2 3

© 2023 The Infinite Actuary, LLC Page 113

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

Exponentiating the prior equation gives:

1
µ− 12 σ 2 )t+ √σ W̃t
Gt = S0 e 2 ( 3

We are almost done! Now we just need to do a Taylor expansion on Gt :

∂Gt ∂Gt 1 ∂ 2 Gt
dGt = dt + dW̃t + 2
(dW̃t )2
∂t ∂ W̃t 2 ∂ W̃t
σ2
 
1 1 σ
= µ − σ 2 Gt dt + √ Gt dW̃t + Gt dt
2 2 3 3

 
dGt 1 1 σ
= µ − σ 2 dt + √ dW̃t
Gt 2 6 3

(b) W̃t is not a standard Wiener process. We know that a standard Wiener process should have
increments that are normally distributed with mean 0 and variance equal to the length of the
increment. That is, W̃t+u − W̃t ∼ N(0, u). However, recall that we explored this exact same
question already in 3.2.1.13 and showed the variance is not u, so we do not have a standard
Wiener process.
u 2
Additionally, note that the variance computed in 3.2.1.13 was t+u 6 u. Note that in general
=
the variance will typically not equal u, but for the special case t = 0, the variance does
simplify to equal u.

(c) Due to the result from (b), this SDE is only valid if the geometric average starts at time
t = 0.

© 2023 The Infinite Actuary, LLC Page 114

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.3.1. Let (Ω, F , P) be a probability space and consider n assets with prices St(i) , i = 1, 2, ...n
satisfying the SDEs:

(i) (i) (i) (i)

dSt = µ(i) St dt + σ (i) St dWt

with

  
(i) (j)
dWt dWt = ρ(ij) dt

(i)
where {Wt : t ≥ 0}, i = 1, 2, ..., n are standard Wiener processes, ρ(ij) ∈ (−1, 1), i 6= j and ρ(ii) = 1

(1) (2) (n)

By considering the function f (St , St , ..., St ) and using Ito’s formula, show that:

n n n
(1) (2) (n)
X (i) ∂f 1 X X (ij) (i) (j) (i) (j) ∂2f
df (St , St , ..., St ) = µ(i) St (i)
dt + ρ σ σ St St (i) (j)
dt
∂St 2 ∂S ∂S
i=1 i=1 j=1 t t
n
X (i) ∂f (i)
+ σ (i) St (i)
dWt
i=1 ∂St

© 2023 The Infinite Actuary, LLC Page 115

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 155-158, Question 1
(1) (2) (n)
By expanding df (St , St , ..., St ) using Taylor’s formula and limiting ourselves to just the first
two terms, we have:

n n n
(1) (2) (n)
X ∂f (i) 1 XX ∂2f (i) (j)
df (St , St , ..., St ) = (i)
dS t + (i) (j)
dSt dSt
2
i=1 ∂St i=1 j=1 ∂St ∂St
n
X ∂f  (i) (i) (i) (i) (i)

= (i)
µ S t dt + σ S t dW t
i=1 ∂St
n n
1 XX ∂2f 
(i) (i) (i) (i) (i)

(j) (j) (j)

+ (i) (j)
µ S t dt + σ St dW t µ(j) St dt + σ (j) St dWt
2
i=1 j=1 ∂St ∂St

Now we recall the following identities:

• (dt)2 = 0
(i)
• (dWt )2 = dt
(i) (j)
• (dWt )(dWt ) = ρ(ij) dt
(i)
• (dWt )dt = 0

Then we can simplify the equation above to our solution:

n n n
(1) (2) (n)
X (i) ∂f 1 X X (ij) (i) (j) (i) (j) ∂2f
df (St , St , ..., St ) = µ(i) St (i)
dt + ρ σ σ St St (i) (j)
dt
∂St 2 ∂S ∂S
i=1 i=1 j=1 t t

n
X (i) ∂f (i)
+ σ (i) St (i)
dWt
i=1 ∂St

© 2023 The Infinite Actuary, LLC Page 116

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.3.2. Let (Ω, F , P) be a probability space and consider 2 assets with prices St(1) and St(2) which
satisfy the SDEs:

(1) (1) (1) (1)

dSt = µ(1) St dt + σ (1) St dWt
(2) (2) (2) (2)
dSt = µ(2) St dt + σ (2) St dWt
with

  
(1) (2)
dWt dWt = ρdt

(1) (2)
where µ(1) , µ(2) , σ (1) , σ (2) are constants and {Wt : t ≥ 0}, {Wt : t ≥ 0} are standard Wiener
processes with correlation ρ.

Throughout this question, you can assume that Vt is normally distributed and written in the
following form:

(1) (2)
σ (1) Wt − σ (2) Wt
Vt = p
(σ (1) )2 + (σ (2) )2 − 2ρσ (1) σ (2)

(1)
St
(a) By letting Ut = (2) show that Ut follows the SDE:
St

dUt = µUt dt + σUt dVt

Where

µ = µ(1) − µ(2) + (σ (2) )2 − ρσ (1) σ (2)

q
σ = (σ (1) )2 + (σ (2) )2 − 2ρσ (1) σ (2)

(b) Show that for s ≥ 0:

   
(1) (2) (2) (1)
E Wt+s Wt = E Wt+s Wt = ρt

(c) Show that {Vt : t ≥ 0} is a standard Wiener process

© 2023 The Infinite Actuary, LLC Page 117

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 155-158, Question 2

(1)
St
(a) By using Taylor’s expansion and remembering that Ut = (2) :
St

∂Ut (1) ∂Ut (2) 1 ∂ 2 Ut (1) 2 1 ∂ 2 Ut (2) 2 ∂2U (1) (2)

dUt = (1)
dSt + (2)
dSt + (1) 2
(dS t ) + (2) 2
(dS t ) + (1) (2)
dSt dSt
∂St ∂St 2 ∂(St ) 2 ∂(St ) ∂St ∂St
(1) (1)
1 (1) St (2) St 
(2) 2
 1 (1) (2)
= (2) dSt − (2) dSt + 0 + (2) · σ (2) St dt − (2) σ (1) St σ (2) St ρdt
St (St ) 2 (St ) 3 (St ) 2
 
(1) (1)
= µ(1) Ut dt + σ (1) Ut dWt − µ(2) Ut dt + σ (2) Ut dWt + (σ (2) )2 Ut dt − ρσ (1) σ (2) Ut dt
   
(1) (2)
= µ(1) − µ(2) + (σ (2) )2 − ρσ (1) σ (2) Ut dt + σ (1) dWt − σ (2) dWt Ut
  q
= µ(1) − µ(2) + (σ (2) )2 − ρσ (1) σ (2) Ut dt + (σ (1) )2 + (σ (2) )2 − 2ρσ (1) σ (2) Ut dVt
| {z }
σ

Where in the last step, we plugged in the given substitution for Vt that was stated in the
question stem.
Therefore, we have proved that:

dUt = µUt dt + σUt dVt

Where

µ = µ(1) − µ(2) + (σ (2) )2 − ρσ (1) σ (2)

q
σ = (σ (1) )2 + (σ (2) )2 − 2ρσ (1) σ (2)
(1) (2)
σ (1) Wt − σ (2) Wt
Vt = p
(σ (1) )2 + (σ (2) )2 − 2ρσ (1) σ (2)

(b) We can re-write with increments:

     
(1) (2) (1) (1) (1) (2) (1) (2)
E Wt+s Wt = E (Wt+s − Wt + Wt )Wt = E Wt Wt = ρt

Where in the last step we used that:

(1) (2)
Cov(Wt , Wt ) (1) (2)

(1) (2)

ρ= √ √ ⇒ Cov(Wt , Wt ) = ρt ⇒ E Wt Wt = ρt
t× t
Thus, by symmetry,
   
(1) (2) (2) (1)
E Wt+s Wt = E Wt+s Wt = ρt

© 2023 The Infinite Actuary, LLC Page 118

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

(c) Recall that there are 3 conditions to satisfy for Vt to be a standard Wiener process:
◦ V0 = 0 and Vt has a continuous path for t ≥ 0
◦ Vt+s − Vt ∼ N(0, s)
◦ Increments are independent of one another
(1) (2)
For the first condition, we know that W0 = W0 = 0, which implies V0 = 0. Additionally,
paths of W are continuous, and thus so are paths of Vt .
For the second condition, we know the following32 :

(1) (1)
Wt+s − Wt ∼ N(0, s)
(2) (2)
Wt+s − Wt ∼ N(0, s)
 
(1) (1) (2) (2)
Cov Wt+s − Wt , Wt+s − Wt = ρs

Also note that we are given Vt is normally distributed, and it follows that its increments are
as well.
So now let’s try to find the expected value and variance of Vt+s and Vt :

(1) (1) (2) (2)

σ (1) (Wt+s − Wt ) − σ (2) (Wt+s − Wt )
Vt+s − Vt = p
(σ (1) )2 + (σ (2) )2 − 2ρσ (1) σ (2)
   
(1) (1) (2) (2)
σ (1) E Wt+s − Wt − σ (2) E Wt+s − Wt
E (Vt+s − Vt ) = p = 0
(σ (1) )2 + (σ (2) )2 − 2ρσ (1) σ (2)
   
 (1) (1) (2) (2) 
(σ (1) )2 V Wt+s −Wt +(σ (2) )2 V Wt+s −Wt
 
 −2σ(1) σ(2) Cov W (1) −W (1) ,W (2) −W (2) 
t+s t t+s t
V (Vt+s − Vt ) =  = s
 
 (σ (1) )2 + (σ (2) )2 − 2ρσ (1) σ (2) 

Therefore, we have proved that Vt+s − Vt ∼ N(0, s) and the second condition is satisfied.

The third condition will be proved on the next page.

32 Cov(X,Y )
If you are having any trouble seeing that the covariance is ρs, try recalling the formula ρ = σ X σY
. And here,
√
σ of both processes is s. Thus, the covariance is ρs

© 2023 The Infinite Actuary, LLC Page 119

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

For the third condition, in this case it is sufficient to prove that:

E[(Vt+s − Vt )Vt ] = 0

Let’s get right into it, noting that the min{t, t + s} terms are a result of using 2.2.1.4 and
also leveraging the results from part (b):

E[(Vt+s − Vt )Vt ] = E(Vt+s Vt ) − E(Vt2 )

(1) (2) (1) (2)
" #
(σ (1) Wt+s − σ (2) Wt+s )(σ (1) Wt − σ (2) Wt )
=E −t
(σ (1) )2 + (σ (2) )2 − 2ρσ (1) σ (2)
"     #
(1) (1) (2) (2)
(σ (1) )2 E Wt+s Wt +(σ (2) )2 E Wt+s Wt
   
(1) (2) (2) (1)
−σ (1) σ (2) E Wt+s Wt −σ (1) σ (2) E Wt+s Wt
= −t
(σ (1) )2 + (σ (2) )2 − 2ρσ (1) σ (2)
(σ (1) ) min{t,t+s}+(σ ) min{t,t+s}
2 (2) 2
h i
−σ (1) σ (2) ρt−σ (1) σ (2) ρt
= −t
(σ (1) )2 + (σ (2) )2 − 2ρσ (1) σ (2)
=t−t
= 0

Thus we have proved all 3 conditions and have shown that Vt is a standard Wiener process

© 2023 The Infinite Actuary, LLC Page 120

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.3.3. Let (Ω, F , P) be a probability space and consider 2 assets with prices St(1) and St(2) which
satisfy the SDEs:

(1) (1) (1) (1)

dSt = µ(1) St dt + σ (1) St dWt
(2) (2) (2) (2)
dSt = µ(2) St dt + σ (2) St dWt
with

  
(1) (2)
dWt dWt = ρdt

(1) (2)
where µ(1) , µ(2) , σ (1) , σ (2) are constants and {Wt : t ≥ 0}, {Wt : t ≥ 0} are standard Wiener
processes with correlation ρ

(1) (2)
(a) By letting Ut = St St show that Ut follows the SDE:

dUt = µUt dt + σUt dVt

Where

µ = µ(1) + µ(2) + ρσ (1) σ (2)

q
σ = (σ (1) )2 + (σ (2) )2 + 2ρσ (1) σ (2)
(1) (2)
σ (1) Wt + σ (2) Wt
Vt = p
(σ (1) )2 + (σ (2) )2 + 2ρσ (1) σ (2)

(b) Show that {Vt : t ≥ 0} is a standard Wiener process

© 2023 The Infinite Actuary, LLC Page 121

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 155-158, Question 3

(1) (2)
(a) This part involves doing a Taylor expansion and simplifying, remembering Ut = St St :

∂Ut (1) ∂Ut (2) 1 ∂ 2 Ut (1) 2 1 ∂ 2 Ut (2) 2 ∂2U (1) (2)

dUt = (1)
dSt + (2)
dSt + (dSt ) + (dSt ) + dSt dSt
∂St ∂St 2 ∂(S (1) )2 2 ∂(S (2) )2 (1)
∂St ∂St
(2)
t t
(2) (1) (1) (2) (1) (2)
= St dSt + St dSt + 0 + 0 + (dSt )(dSt )
   
(2) (1) (1) (1) (1) (2) (2) (2) (1) (2)
= St µ(1) St dt + σ (1) St dWt + St µ(2) St dt
+ σ (2) St dWt + ρσ (1) σ (2) St St dt
   
(1) (2)
= µ(1) + µ(2) + ρσ (1) σ (2) Ut dt + Ut σ (1) dWt + σ (2) dWt
  q
= µ(1) + µ(2) + ρσ (1) σ (2) Ut dt + (σ (1) )2 + (σ (2) )2 + 2ρσ (1) σ (2) Ut dVt
| {z }
σ

Therefore, we have proved that:

dUt = µUt dt + σUt dVt

Where

µ = µ(1) + µ(2) + ρσ (1) σ (2)

q
σ = (σ (1) )2 + (σ (2) )2 + 2ρσ (1) σ (2)
(1) (2)
σ (1) Wt + σ (2) Wt
Vt = p
(σ (1) )2 + (σ (2) )2 + 2ρσ (1) σ (2)

(b) This is simply a variation of the previous problem. We can make the same arguments as the
solution in the prior problem, except replace −σ (2) → σ (2) to reach the desired conclusion:

(1) (2)
σ (1) Wt + σ (2) Wt
Vt = p ∼ N(0, t)
(σ (1) )2 + (σ (2) )2 + 2ρσ (1) σ (2)

is also a standard Wiener process

© 2023 The Infinite Actuary, LLC Page 122

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

3.2.3.10. (Heston Model). Let (Ω, F , P) be a probability space and let {WtS : t ≥ 0}, {Wtσ : t ≥ 0}
be two standard Wiener processes with correlation ρ ∈ (−1, 1). Suppose the asset price St takes
the form

dSt = µSt dt + σSt dWtS , S0 > 0

dσt2 = κ(θ − σt2 )dt + ασt dWtσ , σ0 > 0
dWtS dWtσ = ρdt

where µ, κ, θ, α are constants and σt is the stochastic volatility of St

(a) If we define {Bt : t ≥ 0} as a standard Wiener process where Bt is independent of Wtσ , show
that:
p
(i) ρWtσ + 1 − ρ2 Bt ∼ N(0, t)
p
(ii) d(ρWtσ + 1 − ρ2 Bt ) · dWtσ = ρdt
p
(iii) Briefly explain why proving the two prior parts shows that WtS = ρWtσ + 1 − ρ2 Bt

(b) Using the above relation, show that for 0 ≤ t ≤ T we can write:
  Z T Z T
ST /εT 1 p
log = µ(T − t) − (1 − p2 ) σu2 du + 1−ρ2 σu dBu
St /εt 2 t t
Rs Rs
σu dWuσ − 21 ρ2 2 du
where εs = eρ 0 0 σu

(c) Prove that the relation between εT and εt can be expressed as:
Z T
εT = εt + ρ σu εu dWuσ
t

(d) Conditional on Ft and {σu : t ≤ u ≤ T } show that:

   
ST /εT 1 2
log Ft , {σu : t ≤ u ≤ T } ∼ N µ − σRM S (T − t), σRM S (T − t)
2
St /εt 2

where
s
T
1 − p2
Z
σRM S = σu2 du
T −t t

© 2023 The Infinite Actuary, LLC Page 123

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 3, Pages 175-178, Question 10

(a) (i) Let’s start with the first identity. Given that both Wtσ and Bt are normal random variables,
we know that a linear combination of independent normals is also going to be normal. Taking
the expected value and variance we get the moments:
p p
E(ρWtσ + 1 − ρ2 Bt ) = ρE(Wtσ ) + 1 − ρ2 E(Bt ) = 0
p
V(ρWtσ + 1 − ρ2 Bt ) = ρ2 V(Wtσ ) + (1 − ρ2 )V(Bt ) = ρ2 t + (1 − ρ2 )t = t

Therefore:
p
ρWtσ + 1 − ρ2 Bt ∼ N(0, t)

(ii) Now we attempt to prove the second identity:

p p
d(ρWtσ + 1 − ρ2 Bt ) · dWtσ = (ρdWtσ + 1 − ρ2 dBt ) · dWtσ
p
= ρ(dWtσ )2 + 1 − ρ2 dBt · dWtσ
= ρdt + 0
= ρdt

(iii) To summarize, we have shown that:

p
ρWtσ + 1 − ρ2 Bt ∼ N(0, t)
p
d(ρWtσ + 1 − ρ2 Bt ) · dWtσ = ρdt

From the givens in the question stem, we also know that:

WtS ∼ N(0, t)
dWtS dWtσ = ρdt
p
Pause and briefly compare these equations. We have shown that ρWtσ + 1 − ρ2 Bt and
WtS have some similar properties. First, they are both N(0, t). Second, they have the same
correlation behavior with dWtσ .
Now, note that bivariate normal random variables are fully specified by their mean, variance
and correlation. Using this property, since we have the same mean, variance and correlation,
we know that:

p
WtS = ρWtσ + 1 − ρ 2 Bt

© 2023 The Infinite Actuary, LLC Page 124

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

(b) Let’s start by writing the SDE of St in terms of dWtσ and dBt :
p
dSt = µSt dt + σSt dWtS = µSt dt + σt St (ρdWtσ + 1 − p2 dBt )

Now let’s use Ito’s lemma for d log St :

 2
1 1 1
d log St = dSt − (dSt )2
St 2 St
1 2 2 2
 
1
= µdt + σdWtS −
σ St dt
St 2
p 1
= µdt + σt (ρdWtσ + 1 − p2 dBt ) − σt2 dt
  2
1 2 p
= µ − σt dt + ρσt dWtσ + 1 − ρ2 σt dBt
2

Taking the integrals:

Z T Z T   Z T Z T
1 2 σ
p
2
d log Su = µ − σu du + ρ σu dWu + 1 − ρ σu dBu
t t 2 t t
Z T Z T Z T
1 p
log ST = log St + µ(T − t) − σu2 du +ρ σu dWuσ + 1−ρ2 σu dBu
2 t t t

Now if we define:
Rs Rs
σu dWuσ − 12 ρ2 2 du
εs = e ρ 0 0 σu

Then:
Z s Z s
1
log(εs ) = ρ σu dWuσ − ρ2 σu2 du
0 2 0

So:
Z T Z T
1
log(εT ) − log(εt ) = ρ σu dWuσ − ρ2 σu2 du
t 2 t

Combining prior equations gives:

Z T Z T
1 p
log ST = log St + µ(T − t) − (1 − ρ2 ) σu2 du + 1−ρ2 σu dBu + log(εT ) − log(εt )
2 t t
 
ST /εT
Lastly, we can re-write the S and  terms as log ST − log T − (log St − log(εt )) = log St /εt
Thus,

  Z T Z T
ST /εT 1 p
log = µ(T − t) − (1 − p2 ) σu2 du + 1−ρ2 σu dBu
St /εt 2 t t

© 2023 The Infinite Actuary, LLC Page 125

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

Rt Rt
σu dWuσ − 12 ρ2 2 du
(c) Let’s use Ito’s lemma on dεt , remembering εt = eρ 0 0 σu
:

∂εt ∂εt σ 1 ∂ 2 εt
dεt = dt + dW t + (dWtσ )2
∂t ∂Wtσ 2 ∂(Wtσ )2
1 1
= − ρ2 σt2 εt dt + ρσt εt dWtσ + ρ2 σt2 εt dt
2 2
= ρσt εt dWtσ

Taking the integrals we have:

Z T Z T
dεu = ρσu εu dWuσ
t t
Z T
εT = εt + ρ σu εu dWuσ
t

(d) Now from our answer in part (b), we know that

  Z T Z T
ST /εT 1 p
log = µ(T − t) − (1 − p2 ) σu2 du + 1−ρ2 σu dBu
St /εt 2 t t
RT
Since t σu dBu is normally distributed33 , we can conclude that the entire random variable is
normally distributed. Now we just need to find the expected value and variance conditioned
on Ft and {σu : t ≤ u ≤ T }

  Z T
ST /εT 1
E log Ft , {σu : t ≤ u ≤ T } = µ(T − t) − (1 − p )
2
σu2 du
St /εt 2 t
 
1 2
= µ − σRM S (T − t)
2

where
s
T
1 − p2
Z
σRM S = σu2 du
T −t t

33
Remember we are conditioning on Ft and {σu : t ≤ u ≤ T }, so σu can be treated as a constant. Thus, we are
integrating a constant times dBu which we know must be normally distributed.

© 2023 The Infinite Actuary, LLC Page 126

QFI Quant Online Seminar – Chin Practice Questions Chapter 3 - Practice Problems

For the variance, we have:

  Z T 
ST /εT
V log Ft , {σu : t ≤ u ≤ T } = (1 − p )V
2
σu dBu Ft , {σu : t ≤ u ≤ T }
St /εt
"tZ 2 #
T
= (1 − p2 )E σu dBu Ft , {σu : t ≤ u ≤ T }
t
Z T 
2
= (1 − p )E σu2 du Ft , {σu : t ≤ u ≤ T }
t
Z T
2
= (1 − p ) σu2 du
t
2
= σRM S (T − t)

Therefore, we can conclude that:

   
ST /εT 1 2
log Ft , {σu : t ≤ u ≤ T } ∼ N µ − σRM S (T − t), σRM S (T − t)
2
St /εt 2

© 2023 The Infinite Actuary, LLC Page 127

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

CHIN CHAPTER 4 - PRACTICE PROBLEMS

© 2023 The Infinite Actuary, LLC Page 128

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.1.1. Let {Wt : t ≥ 0 } be a P-standard Wiener process and let X be a real-valued random
variable on the probability space (Ω, F , P) such that EP (|X|2 ) < ∞

For the following process:

Mt = EP (X|Ft )

(a) Show that

EP (Mt2 ) < ∞

(b) Prove that Mt is a P-martingale with respect to the filtration Ft generated by Wt

© 2023 The Infinite Actuary, LLC Page 129

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 192-194, Question 1

(a) Recall we are given that Mt = EP (X|Ft ). We can prove the expectation is finite by noting
that:
h i h i
EP (Mt2 ) = EP EP (X|Ft )2 ≤ EP EP (X 2 |Ft ) = EP (X 2 ) = EP (|X|2 ) < ∞

See notes below for how we derived the above:

◦ The first equation is simply using the definition given in the question
◦ The inequality is derived using Jensen’s inequality34
◦ Then we use tower property (double expectation) and simplify
Thus,
EP (Mt2 ) < ∞

(b) To show that Mt is a P-martingale, we note that:

(i) Mt is clearly Ft -adapted since it is written in the form Mt = EP (X|Ft ), so we can see
it is written as an expectation conditioned on Ft and thus it is non-anticipating.
(ii) Under filtration Fs , 0 ≤ s ≤ t we find:

h i
EP (Mt |Fs ) = EP EP (X|Ft ) Fs
= EP (X|Fs )
= Ms

(iii) We can deduce:

h i h i
EP (|Mt |) = EP EP (X|Ft ) ≤ EP EP (|X||Ft ) = EP (|X|)

Where above, we again used Jensen’s inequality, this time with the convex function
ϕ(X) = |X|. Thus, now we just need to show EP (|X|) < ∞ which will give us the
desired result that EP (|Mt |) < ∞.
Noting we are given that EP (|X|2 ) < ∞ and applying Jensen’s inequality one last time,
in this case again using convex function ϕ(X) = X 2 , we get that:
h i2 q
EP (|X|) ≤ EP (|X|2 ) < ∞ ⇒ EP (|X|) ≤ EP (|X|2 ) < ∞

And thus, EP (|X|) < ∞, so we have that EP (|Mt |) < ∞.

From the three results above, we have shown that Mt is a P-martingale with respect to Ft

34
Recall from Chin Ch 1 that Jensen’s inequality for convex functions states that if ϕ : R 7→ R is a convex function,
then ϕ[E(X|G)] ≤ E[ϕ(X)|G]. For this question, we use convex function ϕ(X) = X 2

© 2023 The Infinite Actuary, LLC Page 130

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.1.2. Let {Wt : t ≥ 0 } be a P-standard Wiener process and let X be a real-valued random
variable on the probability space (Ω, F , P) such that EP (|X|2 ) < ∞

Given that the following process:

Mt = EP (X|Ft )

is a P-martingale with respect to filtration Ft generated by Wt then there exists an adapted process
γu , 0 ≤ u ≤ T such that:

Z t
Mt = M0 + γu dWu , 0≤t≤T
0

Show that for 0 ≤ t ≤ T , we have that:

(a) If X = WT then γt = 1

(b) If X = WT2 then γt = 2Wt

(c) If X = WT3 then γt = 3(Wt2 + T − t)

1 2 (T −t)
(d) If X = eσWT then γt = σeσWt + 2 σ

© 2023 The Infinite Actuary, LLC Page 131

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 192-194, Question 2
∂Mt
To solve these parts, it is helpful to recognize that dMt = γt dWt , thus γt = ∂Wt

(a) If X = WT then

Mt = EP (WT |Ft ) = Wt
∂Mt
γt = =1
∂Wt

(b) If X = WT2 then in order to solve for Mt , recall from problem 2.2.3.2 that Wt2 − t is a
P-martingale:

EP (WT2 − T |Ft ) = Wt2 − t

Therefore, we can write:

Mt = EP (WT2 |Ft ) = Wt2 + T − t

∂Mt
γt = = 2Wt
∂Wt

(c) If X = WT3 then in order to solve for Mt , recall from problem 2.2.3.4 that Wt3 − 3tWt is a
P-martingale:
EP (WT3 − 3T WT |Ft ) = Wt3 − 3tWt
Therefore, we can write:

Mt = EP (WT3 |Ft ) = Wt3 + 3(T − t)Wt

∂Mt
γt = = 3(Wt2 + T − t)
∂Wt
1 2t
(d) If X = eσWT then recall from problem 2.2.3.3 that eσWt − 2 σ is a P-martingale:
 1 2
 1 2
EP eσWT − 2 σ T Ft = eσWt − 2 σ t

Therefore, we can write:

1 2 (T −t)
Mt = EP (eσWT |Ft ) = eσWt + 2 σ
∂Mt 1 2
γt = = σeσWt + 2 σ (T −t)
∂Wt

© 2023 The Infinite Actuary, LLC Page 132

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.2.1. (Novikov’s Condition I). Let {Wt : t ≥ 0 } be a P-standard Wiener process on the
probability space (Ω, F , P) and let θt be an adapted process35 , 0 ≤ t ≤ T . Consider the identity:

Rt Rt
θs dWs − 12 θs2 ds
Zt = e − 0 0

If we are given:

 1 RT 2 
EP e 2 0 θt dt < ∞

Then show that Zt is a positive P-martingale for 0 ≤ t ≤ T

35
You can treat θt as deterministic for this question. The Chin solution appears to assume that you have a
deterministic θt (for example, when factoring out the first exponential term). This is a minor technical note, but I
did just want to just briefly mention that here.

© 2023 The Infinite Actuary, LLC Page 133

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 194-197, Question 1

Recall that in order to prove that Zt is a martingale, we must prove the following:

• EP (Zt |Fs ) = Zu

• EP (|Zt |) < ∞

• Zt is Ft -adapted

Let’s start with the first requirement (see footnote below for more details on how to evaluate the
blue expectation term)36 :
 R 
− 0t θs dWs − 21 0t θs2 ds
R
P P
E (Zt |Fu ) = E e Fu
 R 
− 21 0t θs2 ds − 0t θs dWs
R
=e P
·E e Fu
   R 
− 12 0t θs2 ds − 0u θs dWs − ut θs dWs
R R
=e P
·E e Fu · E e
P
Fu
1
Rt Ru 1
Rt
θs2 ds 2
= e− 2 0 · e− 0 θs dWs
· e2 u θs ds

Ru Ru
θs dWs − 12 θs2 ds
= e− 0 0 = Zu

For the second requirement:

 Rt Rt

P P − θs dWs − 12 θs2 ds
E (|Zt |) = E e 0 0

 Rt 1 t 2
R 
= EP e− 0 θs dWs − 2 0 θs ds
1 t 2
R  Rt 
= e− 2 0 θs ds · EP e− 0 θs dWs
1
Rt 1
Rt
θs2 ds θs2 ds
= e− 2 0 · e2 0

= 1<∞

Rt 1
Rt 2
Finally, Zt is written in the form Zt = e− 0 θs dWs − 2 0 θs ds . We are given θt is adapted, and can also
see that Zt only depends on Ws for s ≤ t. Therefore, Zt is non-anticipating and thus Ft -adapted.

Therefore, we have shown that Zt is a P-martingale.

Rt  R 
36 t
Note that the blue expectation can be evaluated by noting from 3.2.2.4 that u
θs dWs ∼ N 0, u θs2 ds . Thus,
1 2 1 Rt 2
since it is normal, the exponential of it is lognormal with expected value eµ∗ + 2 σ∗ = e 2 u θs ds

© 2023 The Infinite Actuary, LLC Page 134

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.2.2. (Novikov’s Condition II). Let {Wt : t ≥ 0 } be a P-standard Wiener process on the
probability space (Ω, F , P) and let θt be an adapted process, 0 ≤ t ≤ T . Consider the identity:

Rt Rt
θs dWs − 12 θs2 ds
Zt = e − 0 0

If we are given:

 1 RT 2 
EP e 2 0 θt dt < ∞

Use Ito’s Lemma to show that Zt is a positive P-martingale for 0 ≤ t ≤ T

© 2023 The Infinite Actuary, LLC Page 135

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 194-197, Question 2

Note that this is simply a repetition of the prior question, but we will prove that the function is a
martingale a different way. Let’s set:

Zt = f (Xt ) = eXt

where

Z t Z t
1
Xt = − θs dWs − θs2 ds
0 2 0
1
dXt = −θt dWt − θt2 dt
2

To use Ito’s Lemma, it will help to have the following partial derivatives:

∂f ∂2f ∂f
= = eXt and =0
∂Xt ∂Xt2 ∂t

Now plug into Ito Lemma (1b) to compute dZt :

∂f ∂f 1 ∂2f
dZt = dXt + dt + θt2 dt
∂Xt ∂t 2 ∂Xt2
 
Xt 1 2 1
=e −θt dWt − θt dt + 0 + θt2 eXt dt
2 2
= −θt Zt dWt

Notice how we have written in terms of an SDE with no drift term. Thus, using the results of
3.2.2.5, we have shown that the key martingale property is satisfied EP [Zt |Fu ] = Zu .

Lastly, from problem 4.2.2.1, we have already shown in that problem that we have the remaining
two conditions for a martingale so, again, we have shown that Zt is a P-martingale.

© 2023 The Infinite Actuary, LLC Page 136

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.2.3. Let (Ω, F , P) be a probability space and Q be another probability measure (Ω, F , Q)
such that Q is absolutely continuous with respect to P on F

Using the Radon-Nikodym theorem, show that if:

dQ
= Zt
dP Ft

for all 0 ≤ t ≤ T then Zt is a positive P-martingale

Note: I find it a bit odd the SOA chose to put this question on the QFI QF syllabus. This question
uses terminology like “absolutely continuous” and “Radon-Nikodym theorem”, which are covered
in Definition 4.3 and Theorem 4.4 (off syllabus). So this is basically a question testing tools that
aren’t directly on the syllabus. However, much of this problem can be solved by using the knowledge
we have about the Radon-Nikodym derivative from MFD. My general advice here is to not spend
much time spinning your wheels here, and feel free to skim through the solution to get a general
idea of the steps. The TIA solution includes a footnote to summarize the technical parts that you
should take as a given.

© 2023 The Infinite Actuary, LLC Page 137

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 194-197, Question 3

As usual, to prove a martingale we will show each of the three required properties. Before jumping
into these, it will be helpful to write out the equation below37 , which will be helpful for the derivation
of the second property:

  Z Z
P dQ dQ
E = · dP = dQ = 1
dP Ω dP Ω

Keeping the above equation in our back pocket (we’ll use it to make the step in blue below), now
let’s attempt to prove our usual 3 conditions for a martingale.

Under filtration, Ft , we define the Radon-Nikodym derivative as:

 
dQ
Zt = E P
Ft
dP

Taking the expectation and using the tower property, we can prove the first condition:

     
dQ dQ
P
E (Zt |Ft ) = E P
E P
Ft Fs = E P
F s = Zs
dP dP

dQ
Next move on to the second property, noting from the footnote that dP > 0:

   
dQ dQ
|Zt | = Ft = Ft
dP dP

    
dQ dQ
P
E (|Zt |) = E P
E P
Ft =E P
=1<∞
dP dP

Thus, we have shown the second property that EP (|Zt |) < ∞.

 
Finally, since Zt is written in the form Zt = E dP Ft , we can see it is written as an expectation
P dQ

conditioned on the information at time t, and thus clearly Ft -adapted.

Therefore, we have shown that Zt is a P-martingale

37
As a technical footnote, the division
here is valid because we are given that Q is absolutely continuous with
respect to P, so we know that P dQ dP
> 0 = 1. Stated another way, with almost certainty, the Radon-Nikodym
derivative of Q with respect to P is positive, so we have dQ
dP
>0

© 2023 The Infinite Actuary, LLC Page 138

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.1. (Geometric Brownian Motion). Consider an economy consisting of a risk-free asset and a
stock (risky asset). At time t the risk-free asset, Bt and stock price St have the following diffusion
processes:

dBt = rt Bt dt
dSt = µt St dt + σt St dWt

where rt is the risk-free rate, µt is the stock price drift rate, σt is the stock price volatility and
{Wt : 0 ≤ t ≤ T } is a P-standard Wiener process on the probability space (Ω, F , P)

(a) From the following discounted stock price process:

Rt
Xt = e− 0 ru du
St

and using Girsanov’s theorem, show that by changing the measure P to an equivalent risk-
neutral measure Q, Xt is a Q-martingale.

(b) Show that38 :

Z t
Xt = X0 + σu Xu dW̃u , 0≤t≤T
0

Where:
Z t
W̃t = Wt + λu du
0

µt −rt
and W̃t is a Q-standard Wiener process with λt = σt is the market price of risk

(c) Show that under Q, the stock price follows the following process:

dSt = rt St dt + σt St dW̃t

Please note that for many of the Section 4 Chin problems, they do not worry about proving
all three martingale properties and focus on just proving the main martingale property that
the expected future value equals the current value, or equivalently, the drift has no SDE as
seen in 3.2.2.5. We will follow that same convention in this section of the problem set. Also
see the supplemental FAQ DSG on martingales for more info.

38
The proposed solution Chin gives to part (b) uses Theorem 4.2 (One-Dimensional Martingale Representation
Theorem) which is off-syllabus. However, we’ve made our own solution to part (b) which I think is intuitive to derive
without directly knowing that off-syllabus theorem. Note that this happens for part (b) in this problem, as well as a
few other spots in Chapter 4 including in part (b) of the next problem.

© 2023 The Infinite Actuary, LLC Page 139

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 1

(a) We are given that:

Rt
Xt = e− 0 ru du
St

Which gives the following partial derivatives

∂X
= −rt Xt
∂t
∂X Rt
= e− 0 ru du
∂St
∂2X
=0
∂St2

Now, let’s apply a Taylor expansion to get the dXt term:

∂X ∂X 1 ∂2X
dXt = dt + dSt + (dSt )2
∂t ∂St 2 ∂St2
Rt
= −rt Xt dt + e− 0 ru du
dSt
= (µt − rt )Xt dt + σt Xt dWt
  
µt − rt
= σt Xt dt + dWt
σt
= σt Xt dW̃t

where:

t
µt − rt
Z
W̃t = Wt + λu du λt =
0 σt

From Girsanov’s theorem, there exists an equivalent martingale or risk-neutral measure on

the filtration Fs defined by the Radon-Nikodym derivative 39 :
Rs Rs
λ2u du− 21
Zs = e− 0 0 λu dWu

so that W̃t is a Q-standard Wiener process.

Now given that our term for dXt had no drift term, then Xt is a Q-martingale

39
Note that this is just the formula for the Radon-Nikodym derivative stated in Chin Theorem 4.6. Also, be
forewarned that if you do read the Chin solution to this question, they write the wrong formula for Zs and accidentally
swap the λ2u and λu terms. They do this a few times in Chapter 4 in fact, so just make sure you use the formula we
have above which correctly has the λ2u on the deterministic integral and λu on the stochastic integral.

© 2023 The Infinite Actuary, LLC Page 140

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

(b) From part (a), recall that we have already showed:

dXt = σt Xt dW̃t

Integrating gives the desired result:

Z t
Xt = X0 + σu Xu dW̃u 0≤t≤T
0

(c) By substituting dWt = dW̃ − λt dt into the equation for dSt , we get:

dSt = µt St dt + σt St dWt
   
µt − rt
= µt St dt + σt St dW̃t − dt
σt
= µt St dt + σt St dW̃t − St (µt − rt )dt
= rt St dt + σSt dW̃t

© 2023 The Infinite Actuary, LLC Page 141

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.2. (Arithmetic Brownian Motion). Consider an economy consisting of a risk-free asset and a
stock (risky asset). At time t the risk-free asset, Bt and stock price St have the following diffusion
processes:

dBt = rt Bt dt
dSt = µt dt + σt dWt

where rt is the risk-free rate, µt is the stock price drift rate, σt is the stock price volatility and
{Wt : 0 ≤ t ≤ T } is a P-standard Wiener process on the probability space (Ω, F , P)

(a) From the following discounted stock price process:

Rt
Xt = e− 0 ru du
St

and using Girsanov’s theorem, show that by changing the measure P to an equivalent risk-
neutral measure Q, Xt is a Q-martingale

(b) Show that:

Z t
Xt = X0 + σu Bu−1 dW̃u , 0≤t≤T
0

Where:
Z t
W̃t = Wt + λu du
0

µt −rt St
and W̃t is a Q-standard Wiener process with λt = σt is the market price of risk

(c) Show that under Q, the stock price follows the following process:

dSt = rt St dt + σt dW̃t

© 2023 The Infinite Actuary, LLC Page 142

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 2

(a) Just like in the previous problem, we are given that:

Rt
Xt = e− 0 ru du
St

Which gives the following partial derivatives:

∂X
= −rt Xt
∂t
∂X Rt
= e− 0 ru du
∂St
∂2X
=0
∂St2

Now, let’s apply a Taylor expansion to get the dXt term:

∂X ∂X ∂2X
dXt = dt + dSt + (dSt )2
∂t ∂St ∂St2
Rt
= −rt Xt dt + e− 0 ru du
dSt
Rt Rt
= −rt e− 0 ru du
St dt + e− 0 ru du
(µt dt + σt dWt )
Rt Rt
= e− 0 (µt − rt St )dt + σt e− 0 ru du dWt
ru du
  
− 0t ru du
R µt − rt St
= σt e dt + dWt
σt
= σt Bt−1 dW̃t

where:

t
µt − rt St
Rt Z
ru du
Bt = e 0 W̃t = Wt + λu du λt =
0 σt

From Girsanov’s theorem there exists an equivalent martingale or risk-neutral measure on

the filtration Fs defined by the Radon-Nikodym derivative:
Rs Rs
λ2u du− 21
Zs = e− 0 0 λu dWu

so that W̃t is a Q-standard Wiener process.

Now given that our term for dXt had no drift term, then Xt is a Q-martingale

© 2023 The Infinite Actuary, LLC Page 143

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

(b) Start with our result from (a):

dXt = σt Bt−1 dW̃t

If we integrate, then we get the desired result:

Z t
Xt = X0 + σu Bu−1 dW̃u 0≤t≤T
0

(c) By substituting dWt = dW̃ − λt dt into the equation for dSt , we get:

dSt = µt dt + σt dWt
   
µt − rt St
= µt dt + σt dW̃t − dt
σt
= µt dt + σt dW̃t − (µt − rt St )dt
= rt St dt + σdW̃t

© 2023 The Infinite Actuary, LLC Page 144

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.3. (Discounted Portfolio). Consider an economy consisting of a risk-free asset and a stock
(risky asset). At time t the risk-free asset, Bt and stock price St have the following diffusion
processes:

dBt = rt Bt dt
dSt = µt St dt + σt St dWt

where rt is the risk-free rate, µt is the stock price drift rate, σt is the stock price volatility and
{Wt : 0 ≤ t ≤ T } is a P-standard Wiener process on the probability space (Ω, F , P)

Consider a trader whose portfolio at time t is Πt consisting of φt shares of stock and ψt units
invested in the risk-free assets. The discounted portfolio value is given as:

Rt
Yt = e− 0 ru du
Πt

By using Girsanov’s theorem, change the measure P to an equivalent risk-neutral measure Q and
show that the discounted portfolio, Yt is a Q-martingale

© 2023 The Infinite Actuary, LLC Page 145

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 3

The time t value of the portfolio Πt is given by:

Πt = φt St + ψt Bt

Next, compute the dΠt term:

dΠt = φt dSt + ψt dBt

= φt (µt St dt + σt St dWt ) + rt ψt Bt dt
= (φt µt St + rt ψt Bt )dt + φt σt St dWt
= [(φt St + ψt Bt )(µt + rt ) − µt ψt Bt − rt φt St ]dt + φt σt St dWt
= [Πt (µt + rt ) − µt ψt Bt − rt φt St ]dt + φt σt St dWt
= [rt Πt + µt (φt St + ψt Bt ) − µt ψt Bt − rt φt St ]dt + φt σt St dWt
= [rt Πt + (µt − rt )φt St ]dt + φt σt St dWt
= rt Πt dt + φt σt St (λt dt + dWt )

where:

µt − rt
λt =
σt
Rt
Now we can try expanding the term dYt using Ito’s lemma where Yt = e− 0 ru du
Πt and first
calculating partial derivatives:

∂Yt Rt
= −rt e− 0 ru du Πt
∂t
∂Yt Rt
= e− 0 ru du
∂Πt
∂ 2 Yt
=0
∂Π2t

Now we will use Ito’s Lemma:

∂Yt ∂Yt ∂ 2 Yt
dYt = dt + dΠt + (dΠt )2
∂t ∂Πt ∂Π2t
Rt
= −rt Yt dt + e− 0 ru du
dΠt
Rt Rt
= −rt e− 0 ru du
Πt dt + e− 0 ru du
[rt Πt dt + φt σt St (λt dt + dWt )]
Rt
= e− 0 ru du
φt σt St (λt dt + dWt )
Rt
= e− 0 ru du
φt σt St dW̃t

© 2023 The Infinite Actuary, LLC Page 146

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

where

Z t
W̃t = Wt + λu du
0

From Girsanov’s theorem there exists an equivalent martingale or risk-neutral measure on the
filtration Fs defined by the Radon-Nikodym derivative:

Rs Rs
λ2u du− 12
Zs = e − 0 0 λu dWu

so that W̃t is a Q-standard Wiener process. Given that under the risk neutral measure Q, the
discounted stock price diffusion process:

Rt
dYt = e− 0 ru du
φt σt St dW̃t

has no dt term, then the discounted portfolio Yt is a Q-martingale

© 2023 The Infinite Actuary, LLC Page 147

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.4. (Self-Financing Trading Strategy). Consider an economy consisting of a risk-free asset

and a stock (risky asset). At time t the risk-free asset, Bt and stock price St have the following
diffusion processes:

dBt = rt Bt dt
dSt = µt St dt + σt St dWt

where rt is the risk-free rate, µt is the stock price drift rate, σt is the stock price volatility and
{Wt : 0 ≤ t ≤ T } is a P-standard Wiener process on the probability space (Ω, F , P)

Consider a trader whose portfolio at time t is Πt consisting of φt shares of stock and ψt units
invested in the risk-free assets. The discounted portfolio value is given as:

Rt
Yt = e− 0 ru du
Πt

Under the risk-neutral measure Q, the discounted portfolio is a Q-martingale.

Show that the portfolio (φt , ψt ) trading strategy has the values:

γt Bt
φt =
σt St
σt Πt − γt Bt
ψt =
σt Bt

Where γt , 0 ≤ t ≤ T is an adapted process such that:

Z t
Yt = Y0 + γv dW̃v 0≤t≤T
0

© 2023 The Infinite Actuary, LLC Page 148

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 4

Since Yt is a Q-martingale with respect to the filtration Ft and, from the previous problem, we
know that:

Rt
dYt = e− 0 ru du
φt σt St dW̃t

where

Z t
W̃t = Wt + λu du
0
µt − rt
λt =
σt

Note we are also given that:

Z t
Yt = Y0 + γv dW̃v 0≤t≤T
0

So,

dYt = γt dW̃t

Thus, for 0 ≤ t ≤ T , we can see that:

Rt
γt = φt σ t e − 0 ru du
St

Rearranging, we have:

Rt
γt e 0 ru du γt Bt
φt = =
σt St σt St

Rt
ru du
where Bt = e 0

Since Πt = ψt Bt + φt St ⇒ σt Πt = σt ψt Bt + γt Bt , we can conclude:

σ t Πt − γ t B t
ψt =
σt Bt

© 2023 The Infinite Actuary, LLC Page 149

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.5. (Self-Financing Portfolio). Consider an economy consisting of a risk-free asset and a

stock (risky asset). At time t, the risk-free asset Bt and stock price St have the following diffusion
processes:

dBt = rt Bt dt
dSt = µt St dt + σt St dWt

where rt is the risk-free rate, µt is the stock price drift rate, σt is the stock price volatility and
{Wt : 0 ≤ t ≤ T } is a P-standard Wiener process on the probability space (Ω, F , P)

Consider a trader whose portfolio at time t is Πt consisting of φt shares of stock and ψt units
invested in the risk-free assets.

For each of the following choices of φt , find the corresponding ψt so that the trading strategy at
time t, (φt , ψt ) is self-financing:

(a) φt = α where α is a constant

(b) φt = Stn , n > 0

Rt
(c) φt = 0 Sun , n > 0

© 2023 The Infinite Actuary, LLC Page 150

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 5

By definition, the portfolio is self-financing if:

dΠt = φt dSt + ψt dBt

so this is the condition we will be trying to satisfy in this question

Before we start on the specific cases, we take note of the following identities:

Rt
ru du
Bt = e 0 satisfies dBt = rt Bt dt
(dSt )2 = σt2 St2 dt
(dSt )v = 0 for v ≥ 3

(a) For φt = α, we have:

Πt = αSt + ψt Bt

Since this is a linear combination, we can ignore any second order terms when applying Ito’s
lemma to Πt :

∂Πt ∂Πt ∂Πt ∂Πt

dΠt = dSt + dBt + dψt + dα
∂St ∂Bt ∂ψt ∂α
= αdSt + ψt dBt + Bt dψt

In order for this portfolio to be self-financing, we need the final term to be 0. That is,
Bt dψt = 0. Therefore, the solution is:

ψt = β where β is a constant

© 2023 The Infinite Actuary, LLC Page 151

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

(b) For φt = Stn , n > 0, we have:

Πt = φt St + ψt Bt = Stn+1 + ψt Bt

Applying Ito’s lemma, we have:

∂Πt ∂Πt ∂Πt 1 ∂ 2 Πt

dΠt = dSt + dBt + dψt + (dSt )2
∂St ∂Bt ∂ψt 2 ∂St2
1
= (n + 1)Stn dSt + ψt dBt + Bt dψt + (n)(n + 1)Stn−1 (dSt )2
2

Now since we want the portfolio to be self-financing, we need the following condition to be
true:

dΠt = Stn dSt + ψt dBt

Setting this equal to the expression we derived, we have:

1
Stn dSt + ψt dBt = (n + 1)Stn dSt + ψt dBt + Bt dψt + (n)(n + 1)Stn−1 (dSt )2
2
−nStn dSt − 21 (n)(n + 1)Stn−1 (dSt )2
dψt =
Bt

or40 :

t t
Sun (n)(n + 1)σu2 Sun+1
Z Z
1
ψt = −n dSu − du
0 Bu 2 0 Bu
Z t Z t
Ru 1 Ru
ψt = −n e− 0 rv dv
Sun dSu − e− 0 rv dv
(n)(n + 1)σu2 Sun+1 du
0 2 0

40
Note: For those following along in the Chin textbook, they do make a couple algebraic mistakes for part (b).
The Chin textbook accidentally drops off the n and 12 terms. I think I hopefully have those all cleaned up and fixed
in this TIA solution.

© 2023 The Infinite Actuary, LLC Page 152

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

Rt
(c) For φt = 0 Sun , n > 0 we have:
Z t 
Πt = Sun St + ψt Bt
0

Applying Ito’s lemma, we have:

∂Πt ∂Πt ∂Πt ∂Πt

dΠt = dSt + dBt + dψt + dt
∂St ∂Bt ∂ψt ∂t
Z t 
= Sun dSt + ψt dBt + Bt dψt + Stn+1 dt
0

Now since we want the portfolio to be self-financing, we need the following condition to be
true:
Z t 
dΠt = Sun dSt + ψt dBt
0

Setting this equal to the expression we derived, we have:

Z t  Z t 
Sun dSt + ψt dBt = Sun dSt + ψt dBt + Bt dψt + Stn+1 dt
0 0

Bt dψt + Stn+1 dt = 0

or

t t
Sun+1
Z Z Ru
ψt = − du = − e− 0 rv dv
Sun+1 du
0 Bu 0

© 2023 The Infinite Actuary, LLC Page 153

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.6. (Stock Price with Continuous Dividend Yield – Geometric Brownian Motion). Consider
an economy consisting of a risk-free asset and a stock (risky asset). At time t the risk-free asset,
Bt and stock price St have the following diffusion processes:

dBt = rt Bt dt
dSt = (µt − Dt )St dt + σt St dWt

where rt is the risk-free rate, µt is the stock price drift rate, Dt is the continuous dividend yield, σt
is the stock price volatility and {Wt : 0 ≤ t ≤ T } is a P-standard Wiener process on the probability
space (Ω, F , P)

Consider a trader whose portfolio at time t is Πt consisting of φt shares of stock and ψt units
invested in the risk-free assets. The discounted portfolio value is given as:

Rt
Yt = e− 0 ru du
Πt

Using Girsanov’s theorem, show that by moving to risk-neutral measure Q, the discounted portfolio,
Yt is a Q-martingale and that under Q, the stock price follows:

dSt = (rt − Dt )St dt + σt St dW̃t

where

Z t
W̃t = Wt + λu du is a Q-standard Wiener process
0

and
µt − rt
λt = is the market price of risk
σt

© 2023 The Infinite Actuary, LLC Page 154

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 6

At time t, the portfolio Πt is valued at:

Πt = φt St + ψt Bt

Since the trader will receive Dt St dt for every stock held, our incremental gain/loss on the portfolio
can be defined as:

dΠt = φt dSt + φt Dt St dt + ψt dBt

= φt [(µt − Dt )St dt + σt St dWt ] + φt Dt St dt + ψt dBt
= φt [µt St dt + σt St dWt ] + ψt rt Bt dt
= φt µt St dt + ψt rt Bt dt + φt σt St dWt
= rt (φt St + ψt Bt )dt + φt (µt − rt )St dt + φt σt St dWt
= rt Πt dt + φt (µt − rt )St dt + φt σt St dWt
= rt Πt dt + φt σt St (λt dt + dWt )

where

µt − rt
λt =
σt

Using the result from problem 4.2.3.3 for finding dYt , we find that:

Rt
dYt = e− 0 ru du
φt σt St dW̃t

Since this doesn’t have a dt term, by applying Girsanov’s theorem, the discounted portfolio Yt is a Q-martingale

We can try to substitute dWt = dW̃t − λt dt into our given expression for dSt :

dSt = (µt − Dt )St dt + σt St dWt

 
µt − rt
= µt St dt − Dt St dt + σt St dW̃t − σt St dt
σt
= (rt − Dt )St dt + σt St dW̃t

© 2023 The Infinite Actuary, LLC Page 155

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.7. (Stock Price with Continuous Dividend Yield – Arithmetic Brownian Motion).

Warning: This problem has significant issues with it. We are simply presenting this one “as is”
how it is explained in the Chin textbook. I would not spend much time or get bogged down on this
question as it has many issues with it. The solution on the next page does include a couple footnotes
that we added to warn you about some of the issues with this question.

Consider an economy consisting of a risk-free asset and a dividend-paying stock price (risky asset).
At time t, the risk-free asset Bt and the stock price St have the following diffusion processes

dBt = rt Bt dt

dSt = (µt − Dt ) dt + σt dWt

such that rt is the risk-free rate, µt is the stock price drift rate, Dt is the continuous dividend yield,
σt is the stock price volatility (which are all time dependent) and {Wt : 0 ≤ t ≤ T } is a P-standard
Wiener process on the probability space (Ω, F , P). At time t, we consider a trader who has a
portfolio valued at Πt holding φt shares of stock and ψt units being invested in a risk-free asset.
From the following discounted portfolio value
Rt
Yt = e− 0 ru du
Πt

Show, using Girsanov’s theorem, that by changing the measure P to an equivalent risk-neutral
measure Q, then the discounted portfolio Yt is a Q-martingale. Show also that under the Q-measure
the stock price follows
dSt = (rt − Dt ) St dt + σt dW
ft
ft = Wt + t λu du is a Q-standard Wiener process such that
R
where W 0

µt − rt St
λt =
σt
is defined as the market price of risk.

© 2023 The Infinite Actuary, LLC Page 156

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 7

At time t, the portfolio Πt is valued as

Πt = φt St + ψt Bt

and since the trader will receive Dt dt for every stock held, then in differential form

dΠt = φt (dSt + Dt dt) + ψt dBt

= φt [µt dt + σt dWt ] + ψt rt Bt dt
= rt Πt dt + φt (µt − rt St ) dt + φt σt dWt
= rt Πt dt + φt σt (λt dt + dWt )
µt −rt St
where λt = σt . Following problem 4.2.3.3, the discounted portfolio becomes
 Rt  Rt
dYt = d e− 0 ru du Πt = φt σt e− 0 ru du dW̃t

such that Z t
W̃t = Wt + λu du
0
By applying Girsanov’s theorem to change the measure P to an equivalent risk-neutral measure Q,
under which W ft is a Q-standard Wiener process, the discounted portfolio Yt is a Q-martingale. By
substituting 4142 dWt = dW̃t + λt dt into dSt = (µt − Dt ) dt + σt dWt , the diffusion process under the
risk-neutral measure becomes
dSt = (rt − Dt ) St dt + σt dW̃t

41
As a reminder, this problem has significant mistakes. One issue is the reading translated the −Dt dt in the dWt
equation to −Dt St dt in the dW̃t equation after applying the Girsanov shift. However, there’s no algebraic reason you
can “tack on” the St like they do in the solution. I reached out to the textbook author and he confirmed this is an
error.
42
The textbook solution writes dWt = dW̃t +λt dt but to derive the last equation, you’ll want to use the substitution
dWt = dW̃t − λt dt instead.

© 2023 The Infinite Actuary, LLC Page 157

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.8. (Commodity Price with Cost of Carry). Consider an economy consisting of a risk-free
asset and a commodity (risky asset). At time t the risk-free asset, Bt and commodity price St have
the following diffusion processes:

dBt = rt Bt dt
dSt = µt St dt + Ct dt + σt St dWt

where rt is the risk-free rate, µt is the stock price drift rate, Ct is the cost of carry for storage per
unit of time, σt is the stock price volatility and {Wt : 0 ≤ t ≤ T } is a P-standard Wiener process
on the probability space (Ω, F , P)

Consider a trader whose portfolio at time t is Πt consisting of φt shares of stock and ψt units
invested in the risk-free assets. The discounted portfolio value is given as:

Rt
Yt = e− 0 ru du
Πt

Using Girsanov’s theorem, show that by moving to risk-neutral measure Q, the discounted portfolio,
Yt is a Q-martingale and that under Q, the stock price follows:

dSt = rt St dt + Ct dt + σt St dW̃t

where

Z t
W̃t = Wt + λu du is a Q-standard Wiener process
0

and
µt − rt
λt = is the market price of risk
σt

© 2023 The Infinite Actuary, LLC Page 158

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 8

At time t, the portfolio Πt is valued at:

Πt = φt St + ψt Bt

Since the trader will pay Ct dt for storage for every unit of the risky asset they hold, our incremental
gain/loss on the portfolio can be defined as:

dΠt = φt dSt − φt Ct dt + ψt dBt

= φt [µt St dt + Ct dt + σt St dWt ] − φt Ct dt + ψt dBt
= φt µt St dt + ψt rt Bt + φt σt St dWt
= rt (φt St + ψt Bt )dt + φt (µt − rt )St dt + φt σt St dWt
= rt Πt dt + φt (µt − rt )St dt + φt σt St dWt
= rt Πt dt + φt σt St (λt dt + dWt )

where

µt − rt
λt =
σt

Using the result from problem 4.2.3.3 for finding dYt , we find that:

Rt
dYt = e− 0 ru du
φt σt St dW̃t

Since this doesn’t have a dt term, by applying Girsanov’s theorem, the discounted portfolio Yt is a Q-martingale

We can try to substitute dWt = dW̃t − λt dt into our given expression for dSt :

dSt = µt St dt + Ct dt + σt St dWt
 
µt − rt
= µt St dt + Ct dt + σt St dW̃t − σt St dt
σt
= rt St dt + Ct dt + σt St dW̃t

© 2023 The Infinite Actuary, LLC Page 159

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.9. (Pricing a Security Derivative). Consider an economy consisting of a risk-free asset and
a stock (risky asset). At time t the risk-free asset, Bt and stock price St have the following diffusion
processes:

dBt = rt Bt dt
dSt = (µt − Dt )St dt + σt St dWt

where rt is the risk-free rate, µt is the stock price drift rate, Dt is the continuous dividend yield, σt
is the stock price volatility and {Wt : 0 ≤ t ≤ T } is a P-standard Wiener process on the probability
space (Ω, F , P)

Consider a trader whose self-financing portfolio at time t is Πt consisting of φt shares of stock and
ψt units invested in the risk-free assets.

Let Ψ(ST ) represent the payoff of a derivative security at time T such that Ψ(ST ) is FT -measurable.
Assuming the trader begins with initial capital Π0 and trading strategy (φt , ψt ) such that the
portfolio value at time T is:

ΠT = Ψ(ST ) almost surely

Show that under the risk-neutral measure Q:

 R 
− tT ru du
Q
Ψ(St ) = E e Ψ(ST ) Ft 0≤t≤T

© 2023 The Infinite Actuary, LLC Page 160

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 9

At time t, in order to calculate the price of a derivative security, Ψ(St ) with payoff Ψ(ST ) at time
T , we note from 4.2.3.6 that since the discounted portfolio value is a Q martingale:

Rt
 RT
  R 
− − − 0T ru du
e 0 ru du Q
Πt = E e 0 ru du
ΠT F t Q
=E e Ψ(ST ) Ft

Thus,

 R 
− tT ru du
Πt = E e Q
Ψ(ST ) Ft

Since the derivative security Ψ and portfolio Π have identical values at T and assuming that the
trader begins with initial capital Π0 , we can safely say that the value of the portfolio is the value
of the derivative at t:

Πt = Ψ(St )

Therefore, we find that:

 R 
− tT ru du
Ψ(St ) = E e Q
Ψ(ST ) Ft 0≤t≤T

© 2023 The Infinite Actuary, LLC Page 161

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.10. (First Fundamental Theorem of Asset Pricing). Consider an economy consisting of a

risk-free asset and a stock (risky asset). At time t the risk-free asset, Bt and stock price St have
the following diffusion processes:

dBt = rt Bt dt
dSt = (µt − Dt )St dt + σt St dWt

where rt is the risk-free rate, µt is the stock price drift rate, Dt is the continuous dividend yield, σt
is the stock price volatility and {Wt : 0 ≤ t ≤ T } is a P-standard Wiener process on the probability
space (Ω, F , P)

Consider a trader whose self-financing portfolio at time t is valued at Πt . Let us define an arbitrage
strategy such that the following 3 criteria are satisfied:

(i) Π0 = 0

(ii) PrP (ΠT ≥ 0) = 1

(iii) PrP (ΠT > 0) = 0

Prove that if the trading strategy has a risk-neutral probability measure Q, then it does not admit
any arbitrage opportunities.

© 2023 The Infinite Actuary, LLC Page 162

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 10

We will prove this result by contradiction43 . Let’s suppose that we are operating in an environment
Q that does allow arbitrage opportunities. We will hope this leads to a contradiction, which will
then complete the proof.

Since Q allows for arbitrage opportunities, from the definition given in the question we know that:

PrQ (ΠT ≥ 0) = 1

In addition44 , since ΠT ≥ 0 then the discounted value of ΠT should also be greater than or equal
to 0:

 RT  Z ∞  RT  Z ∞  RT 
− ru du − ru du
Q
E e 0 ΠT = Q
Pr e 0 ΠT ≥ x dx = PrQ ΠT ≥ xe 0 ru du dx = 0
| {z } 0 0
0

which implies that:

PrQ (ΠT ≥ 0) = 0
Thus, we have reached a contradiction. We have said both PrQ (ΠT ≥ 0) = 0 and PrQ (ΠT ≥ 0) = 1,
which is a contradiction.

Therefore, we have proved by contradiction that the risk-neutral probability measure Q does not
admit any arbitrage opportunities.

43
A proof by contradiction starts by assuming what we are trying to prove is false. The statement is then proven
if this leads to a contradiction.
44
Here, we are using the property from the prelims that when you have a random
R ∞variable X that has a support that
is non-negative, the expected value can be computed from the formula E(X) = 0 S(x)dx where S(x) = P r(X > x)
is the survival function.

© 2023 The Infinite Actuary, LLC Page 163

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.11. (Second Fundamental Theorem of Asset Pricing). Consider an economy consisting of a

risk-free asset and a stock (risky asset). At time t the risk-free asset, Bt and stock price St have
the following diffusion processes:

dBt = rt Bt dt
dSt = (µt − Dt )St dt + σt St dWt

where rt is the risk-free rate, µt is the stock price drift rate, Dt is the continuous dividend yield, σt
is the stock price volatility and {Wt : 0 ≤ t ≤ T } is a P-standard Wiener process on the probability
space (Ω, F , P)

Consider a trader whose portfolio at time t is Πt consisting of φt shares of stock and ψt units
invested in the risk-free assets.

Prove that if the market is complete (every derivative security can be hedged) then the risk-neutral
measure is unique.

© 2023 The Infinite Actuary, LLC Page 164

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 11

The basic idea of this proof is we will assume that two risk-neutral measures Q1 and Q2 exist. We
will then prove that they must be equal to each other, which means the risk-neutral measure is
unique.

Let A ∈ FT , and we consider the payoff of an arbitrary derivative security

( RT
e 0 ru du A ∈ FT
Ψ(ST ) =
0 A∈/ FT

Since the market model is complete then there is a portfolio value process Πt , 0 ≤ t ≤ T with the
same initial condition Π0 that satisfies ΠT = Ψ(ST )

Since Q1 and Q2 are both risk-neutral measures, the discounted portfolio value is a martingale
under both Q1 and Q2 :

h RT i h RT i
Q1 (A) = EQ1 e− 0 ru du Ψ(ST ) = EQ1 e− 0 ru du ΠT = Π0

h RT i h RT i
Q2 (A) = EQ2 e− 0 ru du Ψ(ST ) = EQ2 e− 0 ru du ΠT = Π0

Now since A is an arbitrary set, this implies that:

Q1 = Q2

Therefore, we have proved that the risk-neutral measure is unique in a complete market.

© 2023 The Infinite Actuary, LLC Page 165

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.12. (Change of Numéraire). Consider an economy consisting of a risk-free asset and a

stock (risky asset). At time t the risk-free asset, Bt and stock price St have the following diffusion
processes:

dBt = rt Bt dt
dSt = (µt − Dt )St dt + σt St dWt

where rt is the risk-free rate, µt is the stock price drift rate, Dt is the continuous dividend yield, σt
is the stock price volatility and {Wt : 0 ≤ t ≤ T } is a P-standard Wiener process on the probability
space (Ω, F , P)

Let Ψ(ST ) represent the payoff of a derivative security at time T such that Ψ(ST ) is F -measurable.
(i)
Under risk-neutral measure Q, let numéraire Nt , i = 1, 2 be a strictly positive process for a non-
dividend-paying asset with the following diffusion process:

(i) (i) (i) (i) (i)

dNt = rt Nt dt + vt Nt dW̃t

(i) (i)
where vt is the volatility and W̃t is a Q-standard Wiener process.

Show that for 0 ≤ t ≤ T :

! !
(1) Q(1) Ψ(ST ) (2) Q(2) Ψ(ST )
Nt E (1)
Ft = Nt E (2)
Ft
NT NT

and

(1) (2)
dQ(1) Nt Nt
= ÷
dQ(2) Ft N0
(1)
N0
(2)

where N (i) is a numéraire and Q(i) is the measure under which the stock price discounted by N (i)
is a martingale

© 2023 The Infinite Actuary, LLC Page 166

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 12
Rt
(i) (i) (i) (i) (i)
We are given that dNt = rt Nt dt + vt Nt dW̃t . Since Bt−1 = e− 0 ru du
, we see that using
stochastic product rule:
(i) (i) (i) (i) (i) (i) (i) (i) (i) (i) (i)
d(Bt−1 Nt ) = d(Bt−1 )Nt +Bt−1 d(Nt ) = −rt Bt−1 Nt dt+rt Nt Bt−1 dt+vt Nt Bt−1 dW̃t = Bt−1 Nt vt dW̃t

Since there is no dt term, we deduce that this is a Q-martingale (as seen in 3.2.2.5). Taking the
log of this and applying Ito’s formula we have:

!2
(i) 1 (i) 1 1 
(i) 2

d(log(Bt−1 Nt )) = (i)
· d(Bt−1 Nt ) − (i)
· d(Bt−1 Nt )
Bt−1 Nt 2 Bt−1 Nt
!2
(i) (i) 1 1 (i) (i)
= vi dW̃t −− (i)
· (Bt−1 Nt vt )2 dt
2 Bt−1 Nt
(i) (i) 1 (i)
= vi dW̃t − (vt )2 dt
2

Taking integrals, we have:

    Z t Z t
1 (i) 2
(i) (i)
log Bt−1 Nt − log B0−1 N0 = vu(i) dW̃u(i) − (v ) du
0 0 2 u

!
(i)
Bt−1 Nt t t
Z Z
1 (i) 2
log = vu(i) dW̃u(i) − (v ) du
(i)
B0−1 N0 0 0 2 u

Since B0 = 1 we can write:

Rt (i) (i) Rt 1 (i) 2

(i) (i)
Bt−1 Nt = N0 e 0 vu dW̃u − 0 2 (vu ) du

Using Girsanov’s theorem to change from the Q measure to an equivalent Q(i) measure, the Radon-
Nikodym derivative is:

(i)
dQ(i) Rt (i) (i) R t 1 (i) 2 Nt
=e 0 (vu )dW̃u − 0 2 (vu ) du =
dQ (i)
Ft N0 Bt

© 2023 The Infinite Actuary, LLC Page 167

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

Such that:

Z t
(i) (i)
Wt = W̃t − vu(i) du
0

is a Q(i) -standard Wiener process.

Under the risk-neutral measure Q we know from previous problems that the stock price follows:

dSt = (rt − Dt )St dt + σt St dW̃t

where

Z t 
µu − ru
W̃t = Wt + du
0 σu

is a Q-standard Wiener process.

(i)
By changing the Q measure to an equivalent Q(i) measure, the discounted stock price (Nt )−1 St
is a Q(i) -martingale. Therefore, to calculate the derivative payoff we have:
!  !−1 
dQ (i) dQ(i)
(i)
EQ (Ψ(ST )|Ft ) = EQ Ψ(ST ) Ft 
dQ Ft dQ Ft
!
(i) (i)
Nt (i) N B T
= (i) EQ Ψ(ST ) 0 (i) Ft
N0 Bt NT
!
(i)
BT Nt Q(i) Ψ(ST )
= E (i)
Ft
Bt N T

Thus, plugging in i = 1 and i = 2 gives:

! !
(1) (2)
BT Nt (1) Ψ(ST ) BT Nt (2) Ψ(ST )
EQ (1)
Ft = EQ (2)
Ft
Bt NT Bt NT

Therefore, we can conclude the following, which is the first statement we needed to show:

! !
(1) (1) Ψ(ST ) (2) (2) Ψ(ST )
Nt EQ (1)
Ft = Nt EQ (2)
Ft
NT NT

© 2023 The Infinite Actuary, LLC Page 168

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

Note that this implies that:

! !
(1) (2)
Q(1) N0 Ψ(St ) Q(2) N0 Ψ(St )
E (1)
F0 =E (2)
F0
Nt Nt

If we had an arbitrary Ft -measurable random variable, Xt , we could show that the change-of-
measure relationship in going from Q(1) to Q(2) is:

" !#
Q(1) Q(2) dQ(1)
E (Xt ) = E Xt
dQ(2) Ft

Therefore, we can deduce:

!
(1) (2)
N0 Ψ(St ) dQ(1) N0 Ψ(St )
(1)
=
Nt dQ(2) Ft Nt
(2)

Rearranging, we finally have the second desired result that:

(1) (2)
dQ(1) Nt Nt
= ÷
dQ(2) Ft N0
(1) (2)
N0

© 2023 The Infinite Actuary, LLC Page 169

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.13. (Black-Scholes Equation). Consider an economy consisting of a risk-free asset and a

stock (risky asset). At time t the risk-free asset, Bt and stock price St have the following diffusion
processes:

dBt = rBt dt
dSt = µSt dt + σSt dWt

where r is the risk-free rate, µ is the stock price drift rate, σ is the stock price volatility (all
constants) and {Wt : 0 ≤ t ≤ T } is a P-standard Wiener process on the probability space (Ω, F , P)

Consider a trader whose portfolio at time t is Πt consisting of φt shares of stock and ψt units
invested in the risk-free assets:

Πt = φt St + ψt Bt

Show that the portfolio (φt , ψt ) is self-financing if and only if Πt satisfies the Black-Scholes equation:

∂Πt 1 2 2 ∂ 2 Πt ∂Πt
+ σ St 2 + rSt − rΠt = 0
∂t 2 ∂St ∂St

© 2023 The Infinite Actuary, LLC Page 170

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 13

From (1b) of Ito’s Lemma, we know that:

∂Πt 1 2 2 ∂ 2 Πt
 
∂Πt
dΠt = dSt + + σ St dt
∂St ∂t 2 ∂St2

Now, we know that the portfolio is self-financing if and only if:

dΠt = φt dSt + ψt dBt

= φt dSt + ψt rBt dt

Equating dt terms from our equations above gives:

∂Πt 1 2 2 ∂ 2 Πt
rBt ψt = + σ St
∂t 2 ∂St2

And similarly comparing the dSt terms gives:

∂Πt
φt =
∂St

Substituting into our original equation for Πt , we get:

Πt = φt St + ψt Bt
∂Πt 2
∂Πt ∂t + 12 σ 2 St2 ∂∂SΠ2t
t
Πt = St +
∂St r

Simplifying yields:

∂Πt 1 2 2 ∂ 2 Πt ∂Πt
+ σ St 2 + rSt − rΠt = 0
∂t 2 ∂St ∂St

This proves our result!

The textbook adds one additional fact here for reference. Note that the Black-Scholes equation does
not depend on the growth parameter µ. The only parameter in the SDE dSt = µSt dt + σSt dWt that
matters for a self-financing portfolio in the PDE above is the volatility parameter σ.

© 2023 The Infinite Actuary, LLC Page 171

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.14. (Black-Scholes Equation with Stock Paying Continuous Dividend Yield). Consider an
economy consisting of a risk-free asset and a stock (risky asset). At time t the risk-free asset, Bt
and stock price St have the following diffusion processes:

dBt = rBt dt
dSt = (µ − D)St dt + σSt dWt

where r is the risk-free rate, µ is the stock price drift rate, D is the continuous dividend yield, σ is
the stock price volatility (all constants) and {Wt : 0 ≤ t ≤ T } is a P-standard Wiener process on
the probability space (Ω, F , P)

Consider a trader whose portfolio at time t is Πt consisting of φt shares of stock and ψt units
invested in the risk-free assets:

Πt = φt St + ψt Bt

Show that the portfolio (φt , ψt ) is self-financing if and only if Πt satisfies the Black-Scholes equation:

∂Πt 1 2 2 ∂ 2 Πt ∂Πt
+ σ St + (r − D)St − rΠt = 0
∂t 2 ∂St2 ∂St

© 2023 The Infinite Actuary, LLC Page 172

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 14

Just as in the prior question, start off with form (1b) of Ito’s Lemma:

∂Πt 1 2 2 ∂ 2 Πt
 
∂Πt
dΠt = dSt + + σ St dt
∂St ∂t 2 ∂St2

Now since the trader will receive DSt dt for every unit of stock held per unit time, the portfolio is
self-financing if and only if:

dΠt = φt dSt + φt DSt dt + ψt dBt

= φt dSt + (ψt rBt + φt DSt )dt

Equating the dt terms yields the first equation below, and similarly equating the dSt terms gives
the second equation below:

∂Πt 1 2 2 ∂ 2 Πt
ψt rBt + φt DSt = + σ St
∂t 2 ∂St2
∂Πt
φt =
∂St

Substituting into our original equation for Πt , we get:

Πt = φt St + ψt Bt
∂Πt 2
∂Πt ∂t + 12 σ 2 St2 ∂∂SΠ2t − ∂Πt
∂St DSt
t
Πt = St +
∂St r
∂Πt 1 2 2 ∂ 2 Πt ∂Πt
+ σ St 2 + (r − D)St − rΠt = 0
∂t 2 ∂St ∂St

© 2023 The Infinite Actuary, LLC Page 173

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.15. (Foreign Exchange Rate under Domestic Risk-Neutral Measure). Consider a foreign
exchange (FX) market consisting of a foreign-to-domestic FX spot rate Xt , a risk-free asset in
domestic currency Btd and a risk-free asset in foreign currency Btf . In this case Xt Btf denotes the
foreign risk-free asset quoted in domestic currency. Assume that the evolution of these values has
the following diffusion processes:

dXt = µt Xt dt + σt Xt dWtx
dBtd = rtd Btd dt
dBtf = rtf Btf dt

where µt is the drift parameter, σt is the volatility parameter, rtd is the domestic risk-free rate and
rtf is the foreign risk-free rate (all time dependent) and {Wtx : 0 ≤ t ≤ T } is a P-standard Wiener
process on the probability space (Ω, F , P)

(a) From the discounted foreign risk-free asset in the domestic currency

Xt Btf
X̃t =
Btd

show, using Girsanov’s theorem, that by changing the measure P to an equivalent domestic
risk-neutral measure Qd then X̃t is a Qd -martingale

(b) Show that under the Qd measure, the FX rate follows:

dXt = (rtd − rtf )Xt + σt Xt dW̃td

where
Z t
W̃td = Wtx + λu du
0

is a Qd -standard Wiener process with

µt + rtf − rtd
λt =
σt

© 2023 The Infinite Actuary, LLC Page 174

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 15

Xt Btf
(a) Let’s start with a Taylor expansion on dX̃t , noting that X̃t = Btd
:
We have the following partial derivatives:
∂ X̃t Btf ∂ X̃t Xt ∂ X̃t X Bf ∂ 2 X̃t
= Btd
and = and = − (Btd )t2 and ∂Xt2
=0
∂Xt ∂Btf Btd ∂Btd t

Now do the Taylor expansion:

∂ X̃t ∂ X̃t f ∂ X̃t d 1 ∂ 2 X̃t 2

dX̃t = dXt + f
dBt + d
dB t + 2 (dXt )
∂Xt ∂Bt ∂B t 2 ∂X t
! !
f   f
Bt X t X B
t t
= (µt Xt dt + σt Xt dWtx ) + Btf rtf dt − rtd Btd dt + 0
Btd Btd (Btd )2
 
= µt + rtf − rtd X̃t dt + σt X̃t dWtx
" ! #
µt + rtf − rtd x
= σt X̃t dt + dWt
σt
= σt X̃t dW̃td

where
Z t
W̃td = Wtx + λu du
0

and

µt + rtf − rtd
λt =
σt

From Girsanov’s theorem there exists an equivalent domestic risk-neutral measure Qd on the
filtration Fs defined by the Radon-Nikodym derivative:
Rs Rs
λ2u du− 12 λu dWux
Zs = e − 0 0

so that W̃td is a Qd -standard Wiener process.

Since our solution didn’t have a dt term, we can conclude that X̃t is a Qd -martingale

© 2023 The Infinite Actuary, LLC Page 175

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

(b) By substituting:
!
µt + rtf − rtd
dWtx = dW̃td − dt
σt

into our original dXt equation, we have:

dXt = µt Xt dt + σt Xt dWtx
" ! #
µt + rtf − rtd
= µt Xt dt + σt Xt dW̃td − dt
σt
= µt Xt dt + σt Xt dW̃td − µt Xt dt − rtf Xt dt + rtd Xt dt

= (rtd − rtf )Xt + σt Xt dW̃td

© 2023 The Infinite Actuary, LLC Page 176

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.16. (Foreign Exchange Rate under Foreign Risk-Neutral Measure). Consider a foreign
exchange (FX) market consisting of a foreign-to-domestic FX spot rate Xt , a risk-free asset in
Bd
domestic currency Btd and a risk-free asset in foreign currency Btf . In this case Xtt denotes the
domestic risk-free asset quoted in foreign currency. Assume that the evolution of these values has
the following diffusion processes:

dXt = µt Xt dt + σt Xt dWtx
dBtd = rtd Btd dt
dBtf = rtf Btf dt

where µt is the drift parameter, σt is the volatility parameter, rtd is the domestic risk-free rate and
rtf is the foreign risk-free rate (all time dependent) and {Wtx : 0 ≤ t ≤ T } is a P-standard Wiener
process on the probability space (Ω, F , P)

(a) From the discounted domestic risk-free asset in the foreign currency

Btd
X̃t =
Xt Btf

show, using Girsanov’s theorem, that by changing the measure P to an equivalent foreign
risk-neutral measure Qf then X̃t is a Qf -martingale

(b) Show that under the Qf measure, the FX rate follows:

dXt = (rtd − rtf + σt2 )Xt + σt Xt dW̃tf

where
Z t
W̃tf = Wtx + λu du
0

is a Qf -standard Wiener process with

µt + rtf − rtd − σt2

λt =
σt

© 2023 The Infinite Actuary, LLC Page 177

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 16

Btd
(a) Let’s start with a Taylor expansion on dX̃t , noting that X̃t = :
Xt Btf
We have the following partial derivatives:
∂ X̃t Btd ∂ X̃t Btd ∂ X̃t 1 ∂ 2 X̃t 2Btd
=− and =− and = and ∂Xt2
=
∂Xt Xt2 Btf ∂Btf Xt (Btf )2 ∂Btd Xt Btf Xt3 Btf

Now do the Taylor expansion:

∂ X̃t ∂ X̃t f ∂ X̃t d 1 ∂ 2 X̃t

dX̃t = dXt + dB + dB + (dXt )2
∂Xt ∂Btf
t
∂Btd t 2 ∂Xt2
! ! ! !
−Btd B d 1 Btd
= (µt Xt dt + σt Xt dWtx ) − t
Btf rtf dt + rtd Btd dt + σt2 Xt2 dt
Xt2 Btf Xt (Btf )2 Xt Btf Xt3 Btf
 
= rtd − rtf + σt2 − µt X̃t dt − σt X̃t dWtx
" ! #
µt − rtd + rtf − σt2
= −σt X̃t dt + dWtx
σt
= −σt X̃t dW̃tf

where
Z t
W̃tf = Wtx + λu du
0

and

µt + rtf − rtd − σt2

λt =
σt

From Girsanov’s theorem there exists an equivalent domestic risk-neutral measure Qf on the
filtration Fs defined by the Radon-Nikodym derivative:
Rs Rs
λ2u du− 12 λu dWux
Zs = e − 0 0

so that W̃tf is a Qd -standard Wiener process.

Since our solution didn’t have a dt term, we can conclude that X̃t is a Qf -martingale

© 2023 The Infinite Actuary, LLC Page 178

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

(b) By substituting:
!
µt + rtf − rtd − σt2
dWtx = dW̃tf − dt
σt

into our original dXt equation, we have:

dXt = µt Xt dt + σt Xt dWtx
" ! #
µt + rtf − rtd − σt2
= µt Xt dt + σt Xt dW̃tf − dt
σt
= µt Xt dt + σt Xt dW̃tf − µt Xt dt − rtf Xt dt + rtd Xt dt + σt2 Xt dt

= (rtd − rtf + σt2 )Xt + σt Xt dW̃tf

© 2023 The Infinite Actuary, LLC Page 179

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

4.2.3.17. (Foreign-Denominated Stock Price under Domestic Risk-Neutral Measure). Let (Ω, F , P)
be a probability space and let {Wts : 0 ≤ t ≤ T } and {Wtx : 0 ≤ t ≤ T } be P-standard Wiener
processes on filtration Ft . Let St and Xt denote the stock price quoted in foreign currency and the
foreign-to-domestic exchange rate, respectively. Each follows the following SDEs:

dSt = (µs − Ds )St dt + σs St dWts

dXt = µx Xt dt + σx Xt dWtx
dWts · dWtx = ρdt ρ ∈ (−1, 1)

where µs , Ds and σs are the stock price drift, continuous dividend yield and volatility, respectively.
µx and σx are the exchange rate drift and volatility parameters, respectively. We also assume that
ρ is the correlation coefficient between the two Wiener processes, Wx and Ws .

In addition, we also have risk-free assets in foreign and domestic currencies, Btf and Btd , respectively.
They follow the following differential equations:

dBtf = rf Btf dt
dBtd = rd Btd dt

where rf and rd are the foreign and domestic risk-free rates.

(a) Show that the stock price denominated in the domestic currency, Xt St follows the diffusion
process:

d(Xt St ) p
= (µs + µx + ρσs σx − Ds )dt + σs2 + 2ρσs σx + σx2 dWtxs
Xt St
where

σs Wts + σx Wtx
Wtxs = p
σs2 + 2ρσs σx + σx2
is a P-standard Wiener process
(b) Using Girsanov’s theorem, show that under the domestic risk-neutral measure, Qd , the stock
price denominated in domestic currency has the diffusion process:

d(Xt St ) p
= (rd − Ds )dt + σs2 + 2ρσs σx + σx2 dW̃txs
Xt St
where
!
µs + µx + ρσs σx − rd
W̃txs = Wtxs + p t
σs2 + 2ρσs σx + σx2

is a Qd -standard Wiener process

© 2023 The Infinite Actuary, LLC Page 180

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 4, Pages 221-242, Question 17

(a) Let us set Ut = Xt St . From stochastic product rule, we see that:

dUt = d(Xt St )
= St dXt + Xt dSt + dSt dXt
= St (µx Xt dt + σx Xt dWtx ) + Xt [(µs − Ds )St dt + σs St dWts ] + ρσs σx Xt St dt
= (µs + µx − Ds + ρσs σx )Xt St dt + σs Xt St dWts + σx Xt St dWtx

Thus,

d(Xt St )
= (µs + µx + ρσs σx − Ds )dt + σs dWts + σx dWtx
Xt St
Now, rewriting the innovation term as seen in 3.2.3.3 gives:

d(Xt St ) p
= (µs + µx + ρσs σx − Ds )dt + σs2 + 2ρσs σx + σx2 dWtxs
Xt St

where

σs Wts + σx Wtx
Wtxs = p
σs2 + 2ρσs σx + σx2

is a P-standard Wiener process

(b) Now if we set the portfolio Πt to be valued as:

Πt = φt Ut + ψt Btd

where φt and ψt are the units invested in Ut = Xt St and the risk-free asset Btd , respectively
Taking note that the holder of the portfolio will receive Ds Ut dt for every stock held, then
we see that (noting the step from the second to third equation is described in the footnote
below):45

dΠt = φt (dUt + Ds Ut dt) + ψt rd Btd dt

h p i
= φt (µs + µx + ρσs σx )Ut dt + σs2 + 2ρσs σx + σx2 Ut dWtxs + ψt rd Btd dt
h p i
= rd Πt dt + φt (µs + µx + ρσs σx − rd )Ut dt + σs2 + 2ρσs σx + σx2 Ut dWtxs
p
= rd Πt dt + φt σs2 + 2ρσs σx + σx2 Ut dW̃txs

45
Note that we can use the equation Πt = φt Ut + ψt Btd , and then multiply by rd dt and rearrange to write the
equation ψt rd Btd dt = rd Πt dt − φt rd Ut dt, which will help us make the step with the blue equals sign.

© 2023 The Infinite Actuary, LLC Page 181

QFI Quant Online Seminar – Chin Practice Questions Chapter 4 - Practice Problems

where

W̃txs = λt + Wtxs
µs + µx + ρσs σx − rd
λ= p
σs2 + 2ρσs σx + σx2

By applying Girsanov’s theorem to change the measure P to an equivalent risk-neutral measure

Qd , under which W̃txs is a Qd -standard Wiener process, then the discounted portfolio e−rd t Πt
is a Qd -martingale

Now if we substitute dWtxs = dW̃txs − λdt into our equation for d(Xt St ) from part (a):

d(Xt St ) p
= (µs + µx + ρσs σx − Ds )dt + σs2 + 2ρσs σx + σx2 dWtxs
Xt St

then the stock price diffusion process under the risk-neutral measure Qd becomes:

d(Xt St ) p
= (rd − Ds )dt + σs2 + 2ρσs σx + σx2 dW̃txs
Xt St

where
!
µs + µx + ρσs σx − rd
W̃txs = Wtxs + p t
σs2 + 2ρσs σx + σx2

© 2023 The Infinite Actuary, LLC Page 182

QFI Quant Online Seminar – Chin Practice Questions Chapter 5 - Practice Problems

CHIN CHAPTER 5 - PRACTICE PROBLEMS

© 2023 The Infinite Actuary, LLC Page 183

QFI Quant Online Seminar – Chin Practice Questions Chapter 5 - Practice Problems

5.2.1.9. (Compensated Poisson Process). Let {Nt : t ≥ 0 } be a Poisson process with intensity
λ > 0 defined on the probability space (Ω, F , P) with respect to the filtration Ft . Define the
compensated Poisson process, N̂ as:

N̂t = Nt − λt

Show that N̂t is a martingale

© 2023 The Infinite Actuary, LLC Page 184

QFI Quant Online Seminar – Chin Practice Questions Chapter 5 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 5, Pages 262-264, Question 9

To show that N̂ is a martingale for 0 ≤ s < t we note the following:

(i)
   
E N̂t |Fs = E N̂t − N̂s + N̂s |Fs
   
= E N̂t − N̂s |Fs + E N̂s |Fs
= E (Nt − Ns − λ(t − s)|Fs ) + N̂s
= E (Nt − Ns |Fs ) − λ(t − s) + N̂s
= λ(t − s) − λ(t − s) + N̂s
= N̂s

Above, we used the facts that N̂t = Nt − λt and also the increment Nt − Ns is independent
of Fs and has an expected value of λ(t − s)

(ii) We take the expected value of the absolute variable and apply triangle inequality:

E(|N̂t |) = E(|Nt − λt|) ≤ E(|Nt |) + |λt| = E(Nt ) + λt = 2λt < ∞

Note that this holds since Nt ≥ 0, λ > 0, and t ≥ 0, so there is no difference between the
absolute value of the variable and the variable itself

(iii) The process N̂t = Nt − λt is clearly Ft -adapted. That is, given the information set at time
t, we know value of Nt and therefore also the value of N̂t .

From the 3 observations above, we have proved N̂ is a martingale

© 2023 The Infinite Actuary, LLC Page 185

QFI Quant Online Seminar – Chin Practice Questions Chapter 5 - Practice Problems

5.2.1.10. Let {Nt : t ≥ 0 } be a Poisson process with intensity λ > 0 defined on the probability
space (Ω, F , P) with respect to the filtration Ft . Define the compensated Poisson process, N̂ as:

N̂t = Nt − λt

Show that N̂t2 − λt is a martingale

© 2023 The Infinite Actuary, LLC Page 186

QFI Quant Online Seminar – Chin Practice Questions Chapter 5 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 5, Pages 262-264, Question 10

Note that we are given N̂t = Nt − λt. Also recall that for a Poisson process, E(Nt ) = V(Nt ) = λt.

It will also be handy to recall the definition of variance from the prelims V(X) = E[(X − E(X))2 ].
Thus, E[Nt − E(Nt ))2 ] = λt. Keep this property in mind, as we use it in the blue highlighted parts
below.

To show that N̂t2 − λt is a martingale for 0 ≤ s < t we prove the three martingale properties as
usual:

(i)
  h i
E N̂t2 − λt|Fs = E (N̂t − N̂s + N̂s )2 |Fs − λt
h i h i  
= E (N̂t − N̂s )2 |Fs + 2E N̂s (N̂t − N̂s )|Fs + E N̂s2 |Fs − λt
= E ((Nt − Ns ) − λ(t − s))2 |Fs + 2(Ns − λs) E [(Nt − Ns ) − λ(t − s)|Fs ] +N̂s2 − λt
 
| {z }
0
= λ(t − s) + 0 + N̂s2 − λt
= N̂s2 − λs

Thus, we have shown the first property.

(ii) The show the second property, use triangle inequality:

 
E |N̂t2 − λt| = E |(Nt − λt)2 − λt| ≤ E |(Nt − λt)2 | +|λt| = E (Nt − λt)2 +λt = 2λt < ∞




| {z }
λt

(iii) The process N̂t2 − λt = (Nt − λt)2 − λt is clearly Ft -adapted. That is, given the information
set at time t, we know value of Nt and therefore also the value of N̂t .

From the 3 observations above, we have proved N̂t2 − λt is a martingale

© 2023 The Infinite Actuary, LLC Page 187

QFI Quant Online Seminar – Chin Practice Questions Chapter 5 - Practice Problems

5.2.1.11. (Exponential Martingale Process). Let {Nt : t ≥ 0 } be a Poisson process with intensity
λ > 0 defined on the probability space (Ω, F , P) with respect to the filtration Ft . Define for u ∈ R

u −1)
Xt = euNt −λt(e
u
(a) Derive the formula E eu(Nt −Ns ) = eλ(t−s)(e −1)

(b) Show that Xt is a martingale

© 2023 The Infinite Actuary, LLC Page 188

QFI Quant Online Seminar – Chin Practice Questions Chapter 5 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 5, Pages 262-264, Question 11

(a) First note that for a Poisson random variable with mean λt, we have that:
∞ ∞ ∞
X X e−λt (λt)x X (λt)x
E(g(X)) = g(x)f (x) = g(x) = e−λt g(x)
x! x!
x=0 x=0 x=0

Given that Nt ∼ Poisson(λt) then46 for u ∈ R, we get:

∞ ∞
X (λt)x X (λteu )x u u
E(euNt ) = e−λt eux = e−λt = e−λt · eλte = eλt(e −1)
x! x!
x=0 x=0

Thus, we have shown that:

u −1)
E(euNt ) = eλt(e
Similarly, for 0 ≤ s < t and because Nt − Ns ∼ Poisson(λ(t − s))
  u
E eu(Nt −Ns ) = eλ(t−s)(e −1)

u −1)
(b) To show that Xt = euNt −λt(e is a martingale we go through our three properties as usual:

(i)
 u

E (Xt |Fs ) = E euNt −λt(e −1) |Fs
u
= e−λt(e −1) E euNt |Fs


u −1)
= e−λt(e E euNt −uNs +uNs |Fs

u
 
= e−λt(e −1) · euNs E eu(Nt −Ns ) |Fs
u −1) u −1)
= e−λt(e · euNs · eλ(t−s)(e
u −1)
= euNs −λs(e
= Xs

(ii) Proving the second property is actually quite straightforward in this case because the expo-
nential function is always positive:

u −1) u −1) u −1) u −1)

E(|Xt |) = E(euNt −λt(e ) = e−λt(e E(euNt ) = e−λt(e · eλt(e =1<∞
u
(iii) The process Xt = euNt −λt(e −1) is clearly Ft -adapted. That is, given the information set at
time t, we know value of Nt and therefore also the value of Xt .
u −1)
From the 3 observations above, we have proved Xt = euNt −λt(e is a martingale

46
Note
P that for the blue equality, this is using the Taylor series of an exponential function which is given by
kx
ek = ∞ x=0 x! for k = λte
u

© 2023 The Infinite Actuary, LLC Page 189

QFI Quant Online Seminar – Chin Practice Questions Chapter 5 - Practice Problems

5.2.2.1. (Pure Jump Process). Let {Nt : t ≥ 0 } be a Poisson process with intensity λ > 0 defined
on the probability space (Ω, F , P) with respect to the filtration Ft . Suppose St follows a pure jump
process:

dSt
= (Jt − 1)dNt
St−

where Jt is the jump size variable if the process Nt jumps at time t, Jt ⊥⊥ Nt (that is, Jt and Nt
are independent) and

(
1 with probability λdt
dNt =
0 with probability 1 − λdt

(a) Explain why the term in dNt is Jt − 1 and not Jt

(b) Show that the differential equation can also be written as:

dSt
= dMt
St−

where

Nt
X
Mt = (Ji − 1)
i=1

which is a compound Poisson process such that Ji , i = 1, 2... is a sequence of independent and
identically distributed random variables which are also independent of Nt

© 2023 The Infinite Actuary, LLC Page 190

QFI Quant Online Seminar – Chin Practice Questions Chapter 5 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 5, Pages 281-285, Question 1

(a) Let St− be the value of St just before a jump and assume there occurs an instantaneous jump
(dNt = 1) in which St changes from St− to Jt St− where Jt is the jump size. Thus

dSt = Jt St− − St− = (Jt − 1)St−

or
dSt
= (Jt − 1)dNt
St−

In other words, if there is a jump, the process increases from St− to Jt St− , and thus the
change in the process value is dSt = Jt St− − St− = (Jt − 1)St− . Thus, the “-1” term exists
because we are looking at the change in the process dSt, and so we need to have the −1St−
to subtract for the initial value before the jump.

Therefore, we have shown why the jump term has a multiplier of (Jt − 1) rather than Jt

(b) The question gives us the below formula:

Nt
X
Mt = (Ji − 1)
i=1

Since the formula we are trying to prove has a dMt term on the right-hand side, we will look
at dMt and see how it behaves.

Let Mt− be the value of Mt just before a jump event. Note that if there is a jump at time t,
then dNt = 1 and:

dMt = Mt − Mt− = Jt − 1

If there is no jump at time t, then dNt = 0 and dMt = 0

Thus, in general, we can write:

dMt = (Jt − 1)dNt

Keep in mind we were also given that:

dSt
= (Jt − 1)dNt
St−

Combining the two above equations above implies that the pure jump process can also be
expressed as:

dSt
= dMt
St−

© 2023 The Infinite Actuary, LLC Page 191

QFI Quant Online Seminar – Chin Practice Questions Chapter 5 - Practice Problems

5.2.2.2. Let {Nt : t ≥ 0 } be a Poisson process with intensity λ > 0 defined on the probability
space (Ω, F , P) with respect to the filtration Ft . Suppose St follows a pure jump process:

dSt
= (Jt − 1)dNt
St−

where Jt is the jump size variable if the process Nt jumps at time t, Jt ⊥⊥ Nt and

(
1 with probability λdt
dNt =
0 with probability 1 − λdt

Assume Jt follows a lognormal distribution such that log Jt ∼ N(µJ , σJ2 ) and Jt is also independent
of Nt .

(a) By applying Ito’s formula on log St and taking integrals, show that for t < T

NT −t
Y
ST = St Ji
i=1

provided47 Ji ∈ (0, 2], i = 1, 2... is a sequence of independent and identically distributed jump
size random variables which are independent of Nt

Note: You will find it helpful to recall the Taylor series for the natural log48 :

∞
X xn
log(1 + x) = (−1)n−1 for |x| < 1
n
n=1

(b) Given St and NT −t = n show that ST follows a lognormal distribution with mean

1 2
E(ST |St , NT −t = n) = St en(µJ + 2 σJ )

and variance

2 2
V(ST |St , NT −t = n) = St2 (enσJ − 1)en(2µJ +σJ )

(c) Given only St show that:

µJ + 1 2
2 σJ −1)(T −t)
E(ST |St ) = St eλ(e
"  #
1 2
2 2λ(T −t) eµJ + 2 σJ −1
λ(T −t)(e2(µJ +σJ ) −1)
V(ST |St ) = St2 e −e

47
The constraint Ji ∈ (0, 2] is just to ensure that the Taylor series converges.
48
As a reminder, the Chin textbook uses the notation “log” but is really referring to the natural log (ln).

© 2023 The Infinite Actuary, LLC Page 192

QFI Quant Online Seminar – Chin Practice Questions Chapter 5 - Practice Problems

SOLUTION
Problems and Solutions in Mathematical Finance, Chapter 5, Pages 281-285, Question 2

(a) Recall the Taylor series for natural log:

∞
X xn x2 x3 x4
log(1 + x) = (−1)n−1 =x− + − + . . . for |x| < 1
n 2 3 4
n=1

Let St− be the value of the St just before a jump event. Let’s expand d(log St ) using this
Taylor’s series and taking note49 that dNt · dNt = dNt

dSt 2 1 dSt 3 1 dSt 4

     
dSt 1
d(log St ) = − + − + ...
St− 2 St− 3 St− 4 St−
1 1 1
= [(Jt − 1) − (Jt − 1)2 + (Jt − 1)3 − (Jt − 1)4 ]dNt
2 3 4
= log Jt dNt

provided −1 < Jt − 1 ≤ 1 or 0 < Jt ≤ 2

By taking integrals, we have:
Z T Z T
d(log Su ) = log Ju dNu
t t
NT −t
X
log ST − log St = log Ji
i=1

NT −t
X
log ST = log St + log Ji
i=1

Lastly, exponentiate both sides and simplify to get the desired result:

NT −t
Y
ST = St Ji
i=1

where Ji ∈ (0, 2] is the jump size occurring at time ti and NT −t = NT −Nt is the total number
of jumps in the time interval (t, T ]

49
This should be easy to see, since the only possible values of dNt are 0 or 1. Thus, dNt · dNt = dNt since 02 = 0
and 12 = 1. Therefore, the powers of dNt are very easy to compute and just all equal dNt , so that’s why it can be
pulled out to the right as a common multiplier to every term of the Taylor series.

© 2023 The Infinite Actuary, LLC Page 193

QFI Quant Online Seminar – Chin Practice Questions Chapter 5 - Practice Problems

(b) Recall the equation below from our part (a) solution:

NT −t
X
log ST = log St + log Ji
i=1

Since log Ji ∼ N(uJ , σJ2 ), i = 1, 2, ..., NT −t are independent and identically distributed and
conditional on St and NT −t = n then

n
X
log ST = log St + log Ji
i=1

follows a normal distribution.

From the above equation, we can see the ST |St , NT −t = n ∼ Lognormal(nµJ , nσJ2 )
Recall that in general, for lognormal random variable L ∼ Lognormal(µ∗ , σ∗2 ) then:

1 2
E(L) = eµ∗ + 2 σ∗
2
E(L2 ) = e2µ∗ +2σ∗
h 2 i 2
V (L) = eσ∗ − 1 e2µ∗ +σ∗

Specifically, here we have that µ∗ = nµJ and σ∗2 = nσJ2

Therefore, the conditional mean and variance of ST are:

1 2
E(ST |St , NT −t = n) = St en(µJ + 2 σJ )

2
E(ST2 |St , NT −t = n) = St2 e2n(µJ +σJ )

Using that V (X) = E(X 2 ) − E(X)2 and simplifying gives our desired result:

2 2
V(ST |St , NT −t = n) = St2 (enσJ − 1)en(2µJ +σJ )

© 2023 The Infinite Actuary, LLC Page 194

QFI Quant Online Seminar – Chin Practice Questions Chapter 5 - Practice Problems

(c) Conditional on only St , by definition we have:

 
NT −t
Y
E(ST |St ) = E St Ji St 
i=1
 
NT −t
Y
= St E  Ji St 
i=1
 
NT −t
Y
= St E  Ji 
i=1
 PNT −t

log Ji
= St E e i=1

By applying the tower property, we have:

  PN 
T −t
log Ji
= St E E e i=1 NT −t
log Jt )−1]
= St eλ(T −t)[ E(e
= St eλ(T −t)[ E(Jt )−1]
1 σ2
µJ + 2 J −1)(T −t)
= St eλ(e

1 2
Where in the last step, we used that log Jt ∼ N(µJ , σJ2 ). Thus, E(Jt ) = eµJ + 2 σJ

© 2023 The Infinite Actuary, LLC Page 195

QFI Quant Online Seminar – Chin Practice Questions Chapter 5 - Practice Problems

Solving for the variance is quite similar, but more algebraically cumbersome. So, make sure
you have the expectation calculation we just walked through down pat first before you walk
through this one.

For the case of variance of ST conditional on St , the calculations are:

 
NT −t
Y
V(ST |St ) = V St Ji St 
i=1

Recall if we bring a constant outside of the variance we square it:

 
NT −t
Y
= St2 V  Ji 
i=1
 PN 
T −t
2 log Ji
= St V e i=1

Re-write the variance in terms of expectations using V (X) = E(X 2 ) − E(X)2 to get:
"    PN 2 #
PNT −t T −t
= St2 E e2 i=1 log Ji − E e i=1 log Ji
"    PN 2 #
PNT −t 2 T −t
= St2 E e i=1 log Ji − E e i=1 log Ji

Now apply the tower property to each expectation to get:

  PN      PN 2
T −t 2 T −t
2 log J 2 log J
= St E E e i=1 i NT −t − St E E e i=1 i
NT −t
2
= St2 eλ(T −t) [E(Jt )−1] − St2 e2λ(T −t) [E(Jt )−1]
"  #
1 2
2)
2(µJ +σJ 2λ(T −t) eµJ + 2 σJ −1
= St2 eλ(T −t)(e −1)
−e

2
Where in the last step, we used that log Jt ∼ N(µJ , σJ2 ). Thus, E(Jt2 ) = e2µJ +2σJ

© 2023 The Infinite Actuary, LLC Page 196

QFI Quant Online Seminar – Chin Practice Questions Hints

Appendix - Proof of Hints

© 2023 The Infinite Actuary, LLC Page 197

QFI Quant Online Seminar – Chin Practice Questions Hints

Hints
To wrap up the problem set, we will include proofs of the given hints in the problem set. Overall,
the steps of these are not too important so feel free to skip this part - we are including it to be
thorough for those who want to see these steps for reference. For the most part, these are not very
exciting and are just cumbersome algebraic / deterministic calculus steps.

T x2 −2txy+ty 2 y2 (x− yt )2
• (2.2.1.5) Hint 1: t(T −t) − T = T
t(T −t)
T

T x2 −2txy+ty 2 x2 (y−x)2
• (2.2.1.5) Hint 2: t(T −t) − t = (T −t)
Z ∞ x2
• (2.2.3.1) Hint 3: For σ > 0, xe− 2σ2 dx = σ 2 .
0
Z ∞ x2
• (2.2.3.4) Hint 4: For σ > 0, x3 e− 2σ2 dx = 2σ 4 .
0

• (3.2.2.12) Hint 5:
R T −2κ(T −s)  −κ(s−t)
+ θ 1 − e−κ(s−t) ds = 1
(x − θ) e−κ(T −t) − e−2κ(T −t) + θ
1 − e−2κ(T −t)

 


t e xe κ 2

• (3.2.2.12) Hint 6:
 2
 2)
x2 e−2κ(T −t) + 2κθ+σ (x − θ) e−κ(T −t) − e−2κ(T −t) + θ(2κθ+σ 1 − e−2κ(T −t)



κ 2κ
2
− xe−κ(T −t) + θ 1 − e−κ(T −t)
 

2 2
= xσκ e−κ(T −t) − e−2κ(T −t) + θσ −κ(T −t) + e−2κ(T −t)



2κ 1 − 2e

© 2023 The Infinite Actuary, LLC Page 198

QFI Quant Online Seminar – Chin Practice Questions Hints

Proof of Hint 1
T x2 −2txy+ty 2 y2
t(T −t) − T

2
T x2 −2txy+ty 2 t(T −t) yT
= t(T −t) − t(T −t)

2
T x2 −2txy+ty 2 −t(T −t) yT
= t(T −t)

2 2
T x2 −2txy+ y Tt
= t(T −t)

2 t2
x2 − 2xyt +y
T T2
= t(T −t)
T

(x− yt )2
= T
t(T −t)
T

Proof of Hint 2
T x2 −2txy+ty 2 x2
t(T −t) − t

T x2 −2txy+ty 2 x2 (T −t)
= t(T −t) − t(T −t)

tx2 −2txy+ty 2
= t(T −t)

x2 −2xy+y 2
= T −t

(y−x)2
= (T −t)

Proof of Hint 3
Z ∞
x2
xe− 2σ2 dx
0

x 2 −x
Apply u-substitution with u = − 2σ 2 , thus du = σ2
dx
∞
−x
Z
x2
= −σ 2 e− 2σ2 · dx
0 σ2
Z −∞
= −σ 2 eu du
0
Z 0
= σ2 eu du
−∞

= σ2

© 2023 The Infinite Actuary, LLC Page 199

QFI Quant Online Seminar – Chin Practice Questions Hints

Proof of Hint 4
x 2 −x
Apply u-substitution with u = − 2σ 2 , thus du = σ2
dx
Z ∞ x2
x3 e− 2σ2 dx
0

∞
−x
Z
x2
= −x2 · σ 2 e− 2σ2 · dx
0 σ2
Z −∞
= u(2σ 2 ) · σ 2 eu du
0
Z −∞
= 2σ 4 ueu du
0

= 2σ 4 [ueu − eu ]|u→−∞
u→0

= 2σ 4

Proof of Hint 5
RT
e−2κ(T −s) xe−κ(s−t) + θ 1 − e−κ(s−t) ds


t
RT
= e−2κT e2κs xe−κ(s−t) + θ 1 − e−κ(s−t) ds


t
RT 
= e−2κT xe−κ(−s−t) + θ e2κs − e−κ(−s−t)


t ds
RT 
= e−2κT xeκ(s+t) + θ e2κs − eκ(s+t)


t ds
RT RT RT
= xe−2κT t eκ(s+t) ds + θe−2κT t e2κs ds − θe−2κT t eκ(s+t) ds
RT RT RT
= xe−2κT eκt t eκs ds + θe−2κT t e2κs ds − θe−2κT eκt t eκs ds
     
eκT −eκt e2κT −e2κt eκT −eκt
= xe−2κT eκt · κ + θe−2κT · 2κ − θe−2κT eκt · κ

   
eκT −eκt e2κT −e2κt
= (x − θ) · e−2κT eκt · κ + θe−2κT · 2κ

1
(x − θ) e−κ(T −t) − e−2κ(T −t) + θ
1 − e−2κ(T −t)



= κ 2

© 2023 The Infinite Actuary, LLC Page 200

QFI Quant Online Seminar – Chin Practice Questions Hints

Proof of Hint 6

To simplify the notation here, let’s denote A = e−κ(T −t)

 2
 2)
x2 A2 + 2κθ+σ (x − θ) A − A2 + θ(2κθ+σ 1 − A2 − [xA + θ [1 − A]]2



κ 2κ

 
2κθ+σ 2 θ(2κθ+σ 2 )
(x − θ) A − A2 + 1 − A2 − 2xθ(A − A2 ) + θ2 (1 − A)2


 
= κ 2κ

 
σ2 θ(2κθ+σ 2 )
(x − θ) A − A2 + 1 − A2 − 2xθ(A − A2 ) + θ2 (1 − A)2


 
= 2θ + κ 2κ

 
2 2 θ(2κθ+σ 2 )
= −2θ2 + x σκ + 2xθ − θ σκ A − A2 + 1 − A2 − 2xθ(A − A2 ) + θ2 (1 − A)2


 
2κ

 2 2

θσ 2
= −2θ2 + x σκ − θ σκ A − A2 + (θ2 + 1 − A2 − θ2 (1 − 2A + A2 )



2κ )

 2 2

θσ 2
= −2θ2 + x σκ − θ σκ A − A2 + 1 − A2 + θ2 (2A − 2A2 )



2κ

xσ 2

 θσ 2

θσ 2
A − A2 + −2θ2 − A − A2 + 1 − A2 + θ2 (2A − 2A2 )



= κ κ 2κ

xσ 2

 2

θσ 2
A − A2 + − θσ 2A − 2A2 + 1 − A2



= κ 2κ 2κ

xσ 2 θσ 2
A − A2 + −2A + 2A2 + 1 − A2



= κ 2κ

xσ 2 θσ 2
A − A2 + 1 − 2A + A2



= κ 2κ

And, finally, plugging in our substitution A = e−κ(T −t) proves the result:
 2
 2
x2 e−2κ(T −t) + 2κθ+σ −κ(T −t) − e−2κ(T −t) + θ(2κθ+σ ) 1 − e−2κ(T −t)



κ (x − θ) e 2κ

2
− xe−κ(T −t) + θ 1 − e−κ(T −t)
 

xσ 2 θσ 2
e−κ(T −t) − e−2κ(T −t) + 1 − 2e−κ(T −t) + e−2κ(T −t)



= κ 2κ

© 2023 The Infinite Actuary, LLC Page 201