Semester Two Examinations, 2023 MATH4090

Venue ____________________
Seat Number __________
Student Number |__|__|__|__|__|__|__|__|
Family Name ____________________
This exam paper must not be removed from the venue
First Name ____________________

School of Mathematics & Physics

Semester Two Examinations, 2023
MATH4090 Computation in Financial Mathematics
This paper is for St Lucia Campus students.

Examination Duration: 120 minutes For Examiner Use Only

Planning Time: 10 minutes Question Mark

Exam Conditions:

•This is a Closed Book examination - no written materials permitted

•Casio FX82 series or UQ approved and labelled calculator only
•During Planning Time - Students are encouraged to review and plan
responses to the exam questions
•This examination paper will be released to the Library

Materials Permitted in the Exam Venue:

(No electronic aids are permitted e.g. laptops, phones)

None

Materials to be supplied to Students:

Additional exam materials (e.g. answer booklets, rough paper) will
be provided upon request.

1 x 14-Page Answer Booklet

Instructions to Students:
 If you believe there is missing or incorrect information
impacting your ability to answer any question, please state
this when writing your answer.
 There are four (4) questions, totalling 40 marks, thus 40% of
the course mark.
 Answer all of the questions

Total _________

Page 1 of 7
Semester Two Examinations, 2023 MATH4090 Computation in Financial Mathematics

In all questions, unless otherwise stated, assume the Black-Scholes model under
the risk-neutral probability measure. Specifically, the price of the underlying asset
is assumed to follow the Geometric Brownian Motion (GBM) dynamics given by

dSt = rSt dt + σSt dWt , S0 > 0 is given.

Here, r > 0 and σ > 0 respectively are the constant risk-free interest rate and
the instantaneous volatility, and {Wt } is a Brownian motion under the risk neutral
measure.

You can assume the existence of a function which returns a random variable
Z ∼ Normal(0, 1).

You can also assume the existence of the function C BS (St , t, K, T, r, σ, “call”/“put”)
which returns the time-t price of a T -maturity and K-strike European option under
the Black-Scholes model.

When you are asked to describe a numerical method, follow a structure similar to
those discussed in the lectures, assignments and tutorials. Do not give Matlab code
as it will not be marked.

We use the notation x+ = max(x, 0).

Page 2 of 7
Semester Two Examinations, 2023 MATH4090 Computation in Financial Mathematics

No-arbitrage binomial trees

Let N be the number of time periods of the tree, where N is a positive integer. Also
let tn = n∆t, where ∆t = T /N , n = 0, . . . , N , be the timesteps of the tree. We
denote by Snj , 0 ≤ n ≤ N , −n ≤ j ≤ n, the price of the underlying asset at node
(n, j) of the tree. With probability pu ∈ (0, 1), the asset price Snj , 0 ≤ n ≤ N − 1,
j+1
−n ≤ j ≤ n, can move up to Sn+1 , whereas, with probability (1 − pu ), Snj can move
j−1
down to Sn+1 . Here, the probability pu is the same for all nodes (n, j) of the tree,
and is given by √
r∆t −σ ∆t
e −e
pu = √ √ . (1)
eσ ∆t − e−σ ∆t
An example a binomial tree with N = 3 is given in Figure 1.
WLPHW WLPHW WLPHW WLPHW

6W

6W

6W 6W

6W 6W

6W 6W

6W

6W

Figure 1: An example of 3-period binomial tree. For simplicity, Snj = Stjn .

Page 3 of 7
Semester Two Examinations, 2023 MATH4090 Computation in Financial Mathematics

Question 1. (10 marks) Consider a T -maturity European-style financial prod-

uct whose payoff is given by
K2
f (ST , K1 , K2 ) = (K2 − ST )+ − (K1 − ST )+ ,
K1
where K2 > K1 > 0. Here, ST denotes the price of the underlying asset at
maturity.

To price this financial product, we use the no-arbitrage binomial tree method
discussed in class. We let N be the number of time periods of the tree, where
N is a positive integer. A description of a binomial tree is given on Page 3 of
the exam booklet.

Recall that Snj , 0 ≤ n ≤ N , −n ≤ j ≤ n, is the price of the underlying asset

at node (n, j) of the tree. We denote by Pnj , 0 ≤ n ≤ N − 1, −n ≤ j ≤ n, the
numerical price of the aforementioned financial product at node (n, j) obtained
by the no-arbitrage binomial method. In particular, P00 is the no-arbitrage
price of the financial product at (S0 , t0 ).

(a) (5 marks) Describe a binomial method to approximate P00 .

(b) (5 marks) Mathematically prove that

Pnj ≥ 0, 0 ≤ n ≤ N, −n ≤ j ≤ n.

You can assume that the up probability pu as in (1) satisfies 0 < pu < 1.

Question 2 (8 marks) Suppose that the share price of a company follows a Ge-
ometric Brownian Motion (GBM) with risk-neural dynamics given by

dSt = rSt dt + σSt dWt , S0 > 0 given. (2)

A convertible (zero coupon-paying) bond with a fixed face value B maturing

at time T is a financial derivative that provides its holder with the right, but
not the obligation, to convert the bond into shares of the company at any
time t < T . Upon conversion, the bondholder will receive shares equivalent to

Page 4 of 7
Semester Two Examinations, 2023 MATH4090 Computation in Financial Mathematics

R × St . Here, R is a predetermined conversion ratio representing the number

of shares that the bondholder receives upon exercising the conversion, and St
is the share price prevailing at conversion time t < T . Should the bondholder
decide against converting and retains the bond until its maturity T , the holder
will be paid the face value B.

(a) (6 marks) Describe a stable trinomial tree pricing method to approx-

imate the time-0 price of the above convertible bond. To achieve full
marks, you need to provide all necessary details of the method, including
parameters of the tree.

(b) (2 marks) Comment on the sensitivity of the convertible bond’s price to

changes in volatility σ in (2).

Question 3. (12 marks) Consider the transport Partial Differential Equation

(PDE)
∂C ∂C
+c = 0, (x, t) ∈ R × (0, T ], (3)
∂t ∂x
where C = C(x, t) and c > 0. Assume also that C satisfies the initial condition

C(x, 0) = f (x), for all x ∈ R.

To numerically approximate the solution to the PDE (3), we truncate the

infinite domain R × (0, T ] to [xmin , xmax ] × (0, T ], where xmin < 0 < xmax , with
|xmin | and xmax being sufficiently large.
−xmin
Let tn = n∆t, where ∆t = T /N , n = 0, . . . , N , xj = x0 + j∆x, ∆x = xmax2J ,
j = −J, . . . , J, We denote by C(xj , tn ) the true solution and by Cnj a finite
difference solution at the grid point (xj , tn ).

Assume that Cnj , n = 0, . . . , N − 1 and j = −J, . . . , J, are available, and

j
we want to use these values to compute Cn+1 , j = −J + 1, . . . , 0, . . . , J − 1.
In addition, you can assume that appropriate boundary conditions, i.e. when
j = −J, or j = J, are given, details of which are not relevant in this question.

Page 5 of 7
Semester Two Examinations, 2023 MATH4090 Computation in Financial Mathematics

(a) (3 marks) Derive an explicit finite difference (FD) equation at the grid
point (xj , tn+1 ), j = −J + 1, . . . , 0, . . . J − 1. Use a central FD approxima-
tion to approximate ∂C ∂x .

(b) (4 marks) Mathematically justify whether your explicit FD scheme is

consistent. If it is, what is the order of consistency in t and x?

(c) (5 marks) Is your explicit scheme unstable, unconditionally stable or

conditionally stable? Mathematically justify your answer. If it is con-
ditionally stable, clearly state the condition(s) that need too be satisfied,
and discuss its (their) implications in a computational setting. Any correct
approach is acceptable.

Question 4. (10 marks) Suppose that you are given the risk-neutral dynamics

dSt = rSt dt + σSt dWt , S0 : given, (4)

where S0 , r, and σ are positive constants.

Let S t = max0≤u≤t Su , for t ≥ 0, be the running maximum of S and define

St
Dt = − 1, t ≥ 0.
St
This is called the relative drawdown of S at time t. Let E[·] denote the
expectation operator under the associated risk-neutral measure.

(a) (4 marks) Consider a call-type payoff that pays at time T > 0

Y = e−rT (ST − K)+ I{DT <l} ,

where l and K are positive constants.

Describe an ordinary Monte Carlo method to estimate E [Y ].
In your algorithm, use M as the number of samples. If a timestepping
method is employed, use N as the number of time intervals, with time
points tn = n∆t, n = 0, 1, . . . , N , and ∆t = T /N .

Page 6 of 7
Semester Two Examinations, 2023 MATH4090 Computation in Financial Mathematics

(b) (2 marks) Now consider another payoff

Z = e−rT (ST − K)+ I{max0≤t≤T Dt <l} ,

Similar to (a), describe an ordinary Monte Carlo method to estimate E [Z].

(c) (4 marks) For E[Z], describe a control variate Monte Carlo pricing algo-
rithm. Clearly desribe the control you use.

END OF EXAMINATION

Page 7 of 7