0% found this document useful (0 votes)

13 views121 pages

TIA MFD Solutions Spring-2023 V2

Uploaded by

alexyjq866

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

13 views121 pages

TIA MFD Solutions Spring-2023 V2

Uploaded by

alexyjq866

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 121

The Infinite Actuary’s

Practice Questions from the

Mathematics of Financial Derivatives (MFD)

Textbook

Zak Fischer, FSA CERA

TIA QFI Quant Online Seminar 2023

Please read this page as background before starting the problem set!!!

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.

Since the beginning of the QFI track, the MFD textbook has been the first reading on the QFI
syllabus. There are many end-of-chapter practice problems in the 3rd edition of the MFD textbook.
However, there is no published solution manual to the 3rd edition practice problems1 . In the past,
this created frustration for students because the MFD book was a large source of practice questions,
but there were no solutions.

Therefore, I’ve made this problem set! This includes the top 50-60 (or so) problems I recommend
working from the MFD textbook with solutions!

One frequent question student’s ask is when to work a practice problem set. I think it is best
to work through this problem set as you go through Section 1 of the seminar. You can either go
chapter-by-chapter, or finish the entire textbook and then work this problem set afterwards. I do
think the quantiative material of Section 1 is best solified through practice, so I highly recommend
spending time working through the practice tab of the TIA seminar.

One other recommendation – if you are bogged down by the number of practice problems, feel free
to only work through the questions marked as “recommended problems” on your first pass. You
can always work on the remaining questions closer to the exam date if you have time.

Lastly, some additional information:

• Use this to supplement your understanding, but keep in mind this is not a substitute for
going through the DSG/videos for the entire syllabus!

• I highly recommend going through the review videos in the seminar and the supplemental
FAQ DSG. These cover the common tips and tricks to know.

• 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 clarifica-
tions. Sometimes multiple textbook questions are blended into one combined drill problem.
Sometimes I have added additional parts, etc. So keep in mind these questions are definitely
motivated from the text, but there are some intentional differences!

• For all questions in this problem set, you may assume you only need to validate the key
martingale property, and the other two technical conditions can be ignored and are assumed
to be satisified. This textbook is not very rigorous about always checking the two technical
properties; however, on exam day (especially if the question is worth a high number of points)
I recommend confirming the two additional martingale technical properties. For more info,
see the “Martingales” section of the supplemental FAQ DSG!

1
If anyone ever finds a 3rd edition solution manual, please email me. But I do not believe it exists at the time I’m
writing this. I have only ever seen a solutions manual up to the 2nd edition of the MFD textbook, which does have
somewhat different end-of-chapter problems.

© 2023 The Infinite Actuary, LLC Page 1

TIA QFI Quant Online Seminar 2023

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 for the Spring 2023 exam was posted on 11/9/2022
◦ V2 was posted on 3/19/2023. We fixed MFD Ch 6 Q2 which had some textbook mistakes
which should now be fixed

Good luck and enjoy! ,

© 2023 The Infinite Actuary, LLC Page 2

QFI Quant Online Seminar – MFD Practice Questions 2023

• Start date is in 12 months

• Maturity is in 24 months

• Payments are made each 6 months

• Floating rate is the USD LIBOR rate

• Swap rate = 5% per annum

(a) State the total pay-fixed swap cash flow at each relevant timestep

(b) Describe how one could create the synthetic equivalent of this interest rate swap using Forward
Rate Agreements (FRAs). Use the convention that an FRA buyer receives float and pays
fixed

For part (c), you are given the following additional information:

• An interest rate caplet is a call option on interest rates, and will have a payoff when the
floating LIBOR rate exceeds the fixed rate

• An interest rate floorlet is a put option on interest rates, and will have a payoff when the
fixed rate exceeds the floating LIBOR rate

(c) Describe how one could create the synthetic equivalent of this interest rate swap using interest
rate options. The available interest rate options are caplets and floorlets.

© 2023 The Infinite Actuary, LLC Page 3

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 1, Question 2

(a) The table below shows the exchange of cashflows for a plain vanilla swap. Note that time is
given in years:

Swap Leg T=0 T=1 T=1.5 T=2

Floating Leg 0 0 N × L212 N × L218
Fixed Leg 0 0 −N × 5% 2 −N × 5% 2
N N
Pay-Fixed Swap CF 0 0 2 × (L12 − 5%) 2 × (L18 − 5%)

Note that the relevant times when cash flows occur is at times T = 1.5 and T = 2

(b) The investor can go long on two Forward Rate Agreements to pay a fixed rate of 2.5% on a
notional of N with maturity of 18 and 24 months. The payoff profile is shown below:

FRA Payoff T=0 T=1 T=1.5 T=2

Investor Receives 0 0 N × L212 N × L218
Investor Pays 0 0 −N × 2.5% −N × 2.5%
L12 L18

CF for Long FRA Investor 0 0 N× 2 − 2.5% N× 2 − 2.5%

Note this is identical to the payoff for the swap in part (a)

(c) The investor can take the following four options all beginning in 12 months:
◦ Long position in a 6 month interest rate caplet at 2.5% with notional N
◦ Short position in a 6 month interest rate floorlet at 2.5% with notional N
◦ Long position in a 12 month interest rate caplet at 2.5% with notional N
◦ Short position in a 12 month interest rate floorlet at 2.5% with notional N

The payoff profile for the first two positions at T = 1.5 is shown below. Note that regardless
of the LIBOR rate, the payoff at this timestep is identical to that of the swap.

L12 L12
Case > 2.5%
2 2 <= 2.5%
N × L212 − 2.5%

Long IR Caplet 0
L12

Short IR Floorlet 0 −N × 2.5% − 2
N × L212 − 2.5% L12

Total CF for Investor N× 2 − 2.5%

The payoff profile at T = 2 is essentially the same as the above table, simply replace L12 with
L18 . Therefore, these four interest rate options replicate a synthetic interest rate swap.

© 2023 The Infinite Actuary, LLC Page 4

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 1: Question 3).

You are given the following:

• The 1 year, arbitrage-free wheat futures price equals Ft

• The annual cost of storage and insurance is denoted by, $c and $s respectively

• The annual effective interest rate is r

Using the information above, answer the questions below:

(a) Come up with the theoretical bounds for the lowest and highest value for Ft

(b) Let Ft = $1, 500, r = 5%, s = $100, c = $150 and the spot price of wheat, St = $1, 470.

Form an arbitrage portfolio and determine the profit one could earn at expiry

© 2023 The Infinite Actuary, LLC Page 5

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter 1, Question 3

(a) From the onset it must hold that, at the very least, Ft must be floored by the current price
of wheat with interest added:

Ft ≥ St (1 + r)

If this condition was violated, then one could simply short sell wheat, invest the proceeds in
a risk free account and buy a future contract on wheat. At expiry, one could guarantee a
profit of St (1 + r) − Ft as shown below:

Position Payoff at t Payoff at T

Buy futures $0 ST − Ft
Invest St -St +St (1 + r)
Sell wheat +St -ST
Total $0 St (1 + r) − Ft

On the other side, we can try to find the upper limit by buying the wheat and selling the
futures. When buying wheat, we would need to pay $c and $s since those are part of the
costs of holding the wheat from time t to T .

Position Payoff at t Payoff at T

Short futures $0 Ft − ST
Borrow St + c + s +St + c + s -(St + c + s)(1 + r)
Buy wheat and pay costs -(St + c + s) ST
Total $0 Ft − (St + c + s)(1 + r)

If the payoff at T , is greater than 0, there exists an arbitrage opportunity. Therefore, we have
a cap for the value of Ft :

Ft ≤ (St + c + s)(1 + r)

Putting the two identities together, we have:

St (1 + r) ≤ Ft ≤ (St + c + s)(1 + r)

Note that the lower bound doesn’t include storage and insurance costs because when we short
the physical wheat, we aren’t responsible for the storage costs and they don’t factor into the
arbitrage calculation.

© 2023 The Infinite Actuary, LLC Page 6

QFI Quant Online Seminar – MFD Practice Questions 2023

(b) In this case, the St (1 + r) = $1, 543.50 which is greater than the current futures price.
Therefore, the first inequality is violated and we can make an arbitrage profit as follows:

Position Payoff at t Payoff at T

Buy futures $0 ST − $1, 500
Invest St -$1,470 $1,543.50
Sell wheat $1,470 -ST
Total $0 $43.50

Therefore, the profit that can be earned at expiry with this strategy is $43.50

© 2023 The Infinite Actuary, LLC Page 7

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 1: Question 4). Consider the following market conditions:

• A stock currently trades at St = 100

• An at-the-money call option on the stock trades for $3

• The corresponding at-the-money put trades at $3.5

• The stock doesn’t pay any dividends and there are no transaction costs

• Traders can borrow and lend at an annual effective interest rate of 5% per year

• A forward contract on the delivery of the stock in 12 months is available at a price of Ft

Given the information above, answer the following questions:

(a) What is the value of Ft ?

(b) If the market quotes a price of Ft = $101, construct arbitrage portfolios by:
(i) Combining a forward and synthetic forward position
(ii) Using put-call parity to identify an arbitrage portfolio that gives a riskless profit of $5.26
today

For each arbitrage strategy in part (b), make sure to identify the relevant instruments used and also
the payoffs at times t and t + 1

© 2023 The Infinite Actuary, LLC Page 8

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter 1, Question 4

(a) Remember that a forward can be synthetically recreated by borrowing St and purchasing
the stock today. Under this approach, at time t + 1, the investor would have to pay back
St × (1 + r) and would have the stock.
In an arbitrage-free environment, the costs of both approaches must be identical and so:
Ft = St × (1 + r)
Ft = $100 × (1.05) = $105

(b) (i) Arbitrage Portfolio 1

As we saw, the fair price for a forward is $105 making the market quote undervalued. The
investor should go long a forward and go short a synthetic forward:

Position Payoff at t Payoff at t + 1

Long forward $0 St+1 − $101
Short stock +$100 −St+1
Invest at risk-free rate -$100 +$105
Total $0 $4

Therefore, a riskless profit of $4 can be earned in a year under current market conditions

© 2023 The Infinite Actuary, LLC Page 9

QFI Quant Online Seminar – MFD Practice Questions 2023

(ii) Arbitrage Portfolio 2

Recall that using put-call parity, the following equation should hold:
K
Ct − P t = S t − (1+r)(T −t)
Where
◦ St is the current price of the underlying stock
◦ Ct is the price of a call on the stock with strike price K expiring at T
◦ Pt is the price of a put on the stock with strike price K expiring at T

Under current market conditions, the left side of the equation is:

Ct − Pt = $3 − $3.5 = −$0.5
The right side of the equation is:
K $100
St − (1+r)(T −t)
= $100 − (1.05)1
= $4.76

Clearly, there is an arbitrage opportunity. The investor should go long the left side of the
equation and short the right side. The payoff profile is shown below:

Position Payoff at t Payoff at t + 1

Long Call -$3 max(St+1 − $100, $0)
Short put +$3.5 min(St+1 − $100, $0)
Short stock +$100 −St+1
Invest at risk-free rate − $100
1.05
+$100
Total $5.26 $0

Therefore a riskless profit of $5.26 can be made immediately under current market conditions
using the above strategy.

© 2023 The Infinite Actuary, LLC Page 10

QFI Quant Online Seminar – MFD Practice Questions 2023

(c) Is this option value independent of the choice of martingale measure?

© 2023 The Infinite Actuary, LLC Page 11

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 2, MFD Question: 1

(a) We can find the risk neutral probability of an up move, p∗u using the following:

260
1+r−d 1 + (0.05 × 0.25) −
p∗u = = 320 260
280
= 0.3917
u−d 280 − 280

Of course, the risk neutral probability of a down move p∗d = 1 − p∗u = 0.6083

(b) Given this is a 3 month option the price can be given by:

1
Ct = EQ [CT ]
1 + 0.25 × r
The expected price of CT is simply the probability weighted average of the payoff under each
scenario:

EQ [CT ] = [p∗u × max(STu − St , 0)] + [p∗d × max(STd − St , 0)]

EQ [CT ] = [0.3917 × $40] + [0.6083 × 0] = $15.67

Therefore,

1
Ct = × $15.67 = $15.47
1 + 0.25 × 0.05
(c) Yes, the fair market value of the call option is independent of the procedure used to obtain
the adjusted probabilities

© 2023 The Infinite Actuary, LLC Page 12

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 2: Question 4). In a conventional 4-period binomial environment where each period
is 1 month, we have the following conditions:

1
S0 = $50 u = 1.15 d= r = 5%
u

The stock does not pay dividends, and is expected to grow at an annual rate of 15%.

You are also given that the risk-neutral probability of an up move is given by:

1 + r∆ − d
p∗ =
u−d

Given the information above, answer the following questions:

(a) What is the annual volatility (σ) of St if the stock return is known to have a log-normal
distribution?

(b) Plot the binomial tree and calculate the price C0 of an at-the-money call option under this
framework

© 2023 The Infinite Actuary, LLC Page 13

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 2, MFD Question: 4

(a) We can use the following equation to deduce our value for sigma:
q
1
σ
u = 1.15 = e 12

σ = 48.415%

(b) Note that the call option only pays off if the stock price at t = 4 is greater than that at t = 0.
This only occurs at the top 2 of the 5 ending nodes.
The risk-neutral probability of an up move, p∗ is given by the following formula:

1 1
∗ 1 + r∆ − d 1 + 0.05 × 12 − 1.15
p = = 1 = 0.48
u−d 1.15 − 1.15

The call option price at any time node, Ct , (t < 4), can be given by the following formula:

1 h i
Ct = (p∗ Ct+1
u
) + (1 − p∗ )Ct+1
d
1 + r∆
The binomial tree can be drawn out as shown below:

St = 87.45 Ct = 37.45

St = 76.04 Ct = 26.25

St = 66.13 Ct = 16.54 St = 66.13 Ct = 16.13

St = 57.50 Ct = 9.82 St = 57.50 Ct = 7.71

St = 50 Ct = 5.60 St = 50 Ct = 3.69 St = 50 Ct = 0

St = 43.48 Ct = 1.76 St = 43.48 Ct = 0

St = 37.80 Ct = 0 St = 37.80 Ct = 0

St = 32.88 Ct = 0

St = 28.59 Ct = 0

Therefore, the price is C0 = 5.60

© 2023 The Infinite Actuary, LLC Page 14

QFI Quant Online Seminar – MFD Practice Questions 2023

You are given the following:

• St = $102

• Annual volatility = 30%

• r = 5% which is constant over the next 3 months

• The stock process can be approximated by a binomial process with ∆ = 1 month

You are also given that the risk-neutral probability of an up move is given by:

1 + r∆ − d
p∗ =
u−d

Given the information above, answer the following questions:

(a) Using the binomial model, price a 3 month European call option with a strike K = $120.

(b) Using the stock, St and a risk-free borrowing, Bt construct a portfolio to replicate the option

© 2023 The Infinite Actuary, LLC Page 15

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 2, MFD Question 5

(a) Using the formulae:

u = eσ∆t
1
d=
u
(1 + r∆ − d)
p=
(u − d)
up
St+1 = St · u
down
St+1 = St · d
Cterminal = Max(St − K, 0)
up down )
(p · Ct+1 + (1 − p) · Ct+1
Ct =
(1 + r)∆t
We can deduce the call price as shown below. For example, for the up-up node, we can derive
Ct by using:
12.26 · p
Ct = 1 = 6.13
1.05 12

u = 1.0904, d = 0.9170, p = 0.5024

St = 132.26 Ct = 12.26

St = 121.28 Ct = 6.13

St = 111.22 Ct = 3.07 St = 111.22 Ct = 0

St = 102 Ct = 1.54 St = 102 Ct = 0

St = 93.53 Ct = 0 St = 93.53 Ct = 0

St = 85.77 Ct = 0

St = 78.66 Ct = 0

Therefore, the call should be priced at Ct = $1.54

© 2023 The Infinite Actuary, LLC Page 16

QFI Quant Online Seminar – MFD Practice Questions 2023

(b) At each timestep, we want to replicate the option’s payoffs by forming a portfolio that will
expire in 1 month. We will tackle these one timestep at a time.
As a first step, we need to calculate the call Delta at each node since this represents the
number of shares of the stock we will need to purchase. Since the call is only valuable in the
“up” nodes, we can ignore the Delta calculations for the “down” nodes. Delta is given by the
following formula:

(Cu,t+1 − Cd,t+1 )
∆t =
(Su,t+1 − Sd,t+1 )

T=0

3.07
∆0 = = 0.174
(111.22 − 93.53)

Therefore, we will need to purchase 0.174 shares of the stock and fund part of this by bor-
rowing:

B0 = (0.1735 ∗ $102) − $1.54 = $16.16

Note that we subtract $1.54 so that the initial cash outlay is the same as the price of the call.
This portfolio will result in a perfect replication of the call payoff as shown:

Position Initial Value Uptick Value Downtick Value

Long 0.174 shares of Stock -$17.70 +$19.30 +$16.23
Borrowing @ 5% +$16.16 -$16.23 -$16.23
Portfolio Value -$1.54 +$3.07 $0

As we can see, the total payoff matches the value of the option at inception as well as the up
and down nodes so we have created a portfolio that has replicated the option perfectly for
the first month.

T=1
If the preceeding move was a downtick, there is no need to create a portfolio since the call
option is worthless in all future time nodes.
However, if the preceding move was an uptick, we have:

6.13
∆u1 = = 0.318
(121.29 − 102)

Since we already have 0.174 shares of stock, we need to purchase an additional 0.145 shares.
This will be funded by borrowing an additional:

B1u = (0.145 ∗ $111.23) = $16.08

© 2023 The Infinite Actuary, LLC Page 17

QFI Quant Online Seminar – MFD Practice Questions 2023

The payoff profile of the rebalanced position is shown below:

Position Initial Value Uptick Value Downtick Value

Long 0.318 shares of Stock -$35.37 +$38.57 +$32.44
Borrowing @ 5% +$32.30 -$32.44 -$32.44
Portfolio Value -$3.07 +$6.13 $0

Again, this matches the call values perfectly.

T=2
Continuing in the same manner as we did for the previous timestep, we see that we need
to buy another 0.265 units of stock and borrow an additional $32.13. This results in the
following payoff profile:

Position Initial Value Uptick Value Downtick Value

Long 0.583 shares of Stock -$70.70 +$77.10 +$64.84
Borrowing @ 5% +$64.57 -$64.84 -$64.84
Portfolio Value -$6.13 +$12.26 $0

(c) As we saw, the arbitrage-free price of this 3 month call is $1.54. Therefore, the market call
is overpriced so an arbitrageur can sell the $5 call and go long a synthetic call using the
underlying stock and borrowing amounts as described in part (b) above. This will result in
an immediate riskless profit of $3.46 .

© 2023 The Infinite Actuary, LLC Page 18

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 2: Question 6). You are given the following:

• An annual effective risk-free rate of r = 6%

• The stock price follows: dSt = St − St−1 = µSt−1 + σ · St−1 ·

◦ Here, we have that = +1 with probability p and = −1 with probability 1 − p

• Volatility is 12% per year

• The stock pays no dividends

• S0 = 100

Given the information above, answer the questions below:

(a) If µ = r and St is arbitrage-free, what is the value of p?

1
(b) Is p = 3 consistent with an arbitrage-free St ?

(c) What do p and represent under these conditions?

© 2023 The Infinite Actuary, LLC Page 19

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 2, MFD Question: 6

(a) We are given that St − St−1 = µSt−1 + σ · St−1 ·

If µ = r then St = St−1 + rSt−1 + σ · St−1 · t

Taking the expected value of the above equation gives that:

E P [St |St−1 ] = St−1 (1 + r + σ)p + St−1 (1 + r − σ)(1 − p)

We also know that the expected return for the stock in this case is r. Under the probability
distribution specified, p must satisfy:

E P [St |St−1 ] = St−1 (1 + r)

Combining the prior two equations, we have:

(1 + r + σ)p + (1 + r − σ)(1 − p) = (1 + r)

1 + r − σ + 2σp = (1 + r)

2σp = σ

1
p=
2

Recall that in order for St to be arbitrage-free, the discounted stock price must be a martin-
gale. As shown, p = 21 is the only value that allows this condition to hold
1
(b) No this would not hold because, as shown above, only p = 2 allows St to be arbitrage-free

(c) represents the underlying random process that drives the “unexpected” movement in the
stock price each period. At every timestep, the “unexpected” portion moves by either +σS
or −σS depending on the value of
p represents the probability of positive while 1 − p represents the probability of a negative
. When µ 6= r, there is a risk premium in the economy. To summarize, p represents the true
probability of an up move.

© 2023 The Infinite Actuary, LLC Page 20

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 3: Question 2). If it exists, find the limit of the following sequences for n = 1, 2, 3...

(a) Xn = (−1)n

(b) Xn = sin nπ
3

© 2023 The Infinite Actuary, LLC Page 21

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 3, Question 2

(a) In this case, we have a split function:

(
1 for even numbers
Xn =
−1 for odd numbers

Therefore we have no convergence.

(b) This is another split function with a slightly more irregular pattern:

0
 √
n = 3, 6, 9...
Xn = − 23 n = 4, 5, 10, 11...
 √3

2 n = 1, 2, 7, 8...

Therefore, there is no convergence.

(c) This function clearly oscillates and the gap only gets bigger as n gets bigger since we are
multiplying by a factor of n. Therefore, there is no convergence here either.
–

In other words, the answer for all three parts is:

© 2023 The Infinite Actuary, LLC Page 22

QFI Quant Online Seminar – MFD Practice Questions 2023

Determine whether the following partial sum is convergent:

n
X 1
k!
k=1

If it is convergent, determine the value it converges to. If it divergent, explain why.

© 2023 The Infinite Actuary, LLC Page 23

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 3, MFD Question: 4

Recall that:

∞
X 1
= e1
k!
k=0

Using the fact that 0! = 1 and changing our lower bound from k = 1 to k = 0, we can see that our
sum converges to:

n
" #
X 1
lim − 1 = e − 1 ≈ 1.718282
n→∞ k!
k=0

© 2023 The Infinite Actuary, LLC Page 24

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 3: Questions 8-9). Consider the following two functions:

f (x) = x3

(
x · sin( πx ) 0<x<1
f (x) =
0 x=0

R1
(a) For the first function, calculate 0 f (x)dx

Additionally, you are given that part (a) can be calculated for the second function as follows:

Z 1
f (x)dx ≈ −0.18
0

(b) Consider splitting the interval [0,1] into 4 equal sub-intervals and calculate the following sums:
P4
◦ i=1 f (xi )(xi − xi−1 )
P4
◦ i=1 f (xi−1 )(xi − xi−1 )

© 2023 The Infinite Actuary, LLC Page 25

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 3, Questions: 8-9

(a)
1
x4
Z
1
x3 dx = = 0.25
0 4 0

(b) For both functions, x0 = 0, x1 = 0.25, x2 = 0.5, x3 = 0.75, x4 = 1 and xi − xi−1 = 0.25
Therefore, for the first function:

4
X 1 1 + 8 + 27 + 64
f (xi )(xi − xi−1 ) = = 0.390625
4 64
i=1
4
X 1 0 + 1 + 8 + 27
f (xi−1 )(xi − xi−1 ) = = 0.140625
4 64
i=1

For the second function (note that the only point that generates a√ non-zero
√
value is when we
3 4π 3 3 3 3
evaluate f and x3 = .75, in which case f (x3 ) = 4 sin( 3 ) = 4 · − 2 = − 8 ):

4
" √ #
X 1 3 3
f (xi )(xi − xi−1 ) = 0+0− + 0 = −0.162380
4 8
i=1
4
" √ #
X 1 3 3
f (xi−1 )(xi − xi−1 ) = 0+0+0− = −0.162380
4 8
i=1

(c) For the first function, as expected, the “true” integral falls between the two values we cal-
culated. We also note that using either extreme as a proxy for the height of the rectangle
doesn’t yield accurate results. As the number of rectangles increases and (xi − xi−1 ) → 0, we
approach the true area given by the integral.
For the second function, since the function oscillates so rapidly, the sums do not approximate
the integral very well. Therefore, the rectangle approach is not useful in this case.

© 2023 The Infinite Actuary, LLC Page 26

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 3: Question 10).

Take the partial derivative of the following function with respect to x, y, z respectively:

x+y+z
f (x, y, z) =
(1 + x)(1 + y)(1 + z)
That is, compute fX , fY and fZ .

© 2023 The Infinite Actuary, LLC Page 27

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 3, Question 10
∂f (x,y,z)
We start by considering ∂x . First recognize that this function can be written as:

1
f (x, y, z) = (x + y + z)(1 + x)−1 ·
(1 + y)(1 + z)

Where the final term can be treated as a constant. Differentiating, we have:

∂f (x, y, z) 1 x+y+z 1
= − 2
·
∂x (1 + x) (1 + x) (1 + y)(1 + z)

∂f (x, y, z) 1 x+y+z
= −
∂x (1 + x)(1 + y)(1 + z) (1 + x)2 (1 + y)(1 + z)

∂f (x, y, z) 1 x+y+z
= 1−
∂x (1 + x)(1 + y)(1 + z) (1 + x)

∂f (x, y, z) 1−y−z
=
∂x (1 + x)2 (1 + y)(1 + z)

Similarly, we can find the partial derivatives for y and z:

∂f (x, y, z) 1−x−z
=
∂y (1 + x)(1 + y)2 (1 + z)

∂f (x, y, z) 1−x−y
=
∂z (1 + x)(1 + y)(1 + z)2

© 2023 The Infinite Actuary, LLC Page 28

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 4: Questions 1-2). Suppose you can bet on a local city council election where the
market offers you the following payoffs:

(
$1, 000 if Candidate A wins
R=
−$1, 500 if Candidate A loses

Let p be the probability of Candidate A winning

(a) Calculate the expected gain on the bet when p = 0.6 and when p = 0.9

Now, consider that you have a friend who offers you the following separate bet, R∗

(
$1, 500 if Candidate A wins
R∗ =
−$1, 000 if Candidate A loses

(b) Using R and R∗, construct a portfolio of bets such that you receive a guaranteed risk-free
return

(c) How would your strategy change if p = 0.2?

© 2023 The Infinite Actuary, LLC Page 29

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 4, Questions: 1-2

(a) The expected gain is given by weighting the payoffs by the probability that they occur. When
p = 0.6:

$1, 000 × 0.6 − $1, 500 × 0.4 = $0

When p = 0.9:

$1, 000 × 0.9 − $1, 500 × 0.1 = $750

(b) If one goes long R∗, and short R.

Position Candidate A Wins Candidate A Loses

Long R∗ Payoff $1,500 -$1,000
Short R Payoff -$1,000 $1,500
Total Payoff $500 $500

Therefore, there will be a guaranteed $500 risk-free profit.

(c) Clearly, the payoff function above doesn’t depend on the value of p at all so the same strategy
can be employed regardless of the probability of Candidate A winning.

© 2023 The Infinite Actuary, LLC Page 30

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 4: Question 3). Consider a game of roulette with the following features:

• Only two outcomes are possible: red or black

• A bet of any amount may be placed on either color

• If the correct color is chosen, $1 is paid per unit bet. Choosing incorrectly pays $0

• The outcome of red and black are equally likely

(a) How much should an investor be willing to pay (per unit bet) for a chance to pay this game?

(b) Does the answer change if the investor is allowed to play the game infinitely many times?

© 2023 The Infinite Actuary, LLC Page 31

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 4, Question 3

(a) The expected payoff for this game if the customer picked red is:

Ered = 0.5 × $1 + 0.5 × $0 = $0.5

This is the same as the expected payoff if a customer picks black.

Therefore, one should only pay $0.5 to play this game.

(b) The number of times one plays the game doesn’t change how much should be charged for the
game.

© 2023 The Infinite Actuary, LLC Page 32

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 5: Question 1). You are given two discrete random variables that can only assume
values of 0 and 1.

Their joint distribution is shown below:

P(Y = 1) P(Y = 0)
P(X = 1) 0.2 0.4
P(X = 0) 0.15 0.25

(a) What are the marginal distributions for X and Y ?

(b) Are X and Y independent?

(c) Calculate the conditional expectation E[X|Y = 1]

© 2023 The Infinite Actuary, LLC Page 33

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 5, Question 1

(a) The marginal distributions in this case are simply the sum of the probabilities of every instance
where the variable takes on a particular value:

P(X = 1) = P(X = 1|Y = 1) + P(X = 1|Y = 0) = 0.2 + 0.4 = 0.6

P(X = 0) = P(X = 0|Y = 1) + P(X = 0|Y = 0) = 0.15 + 0.25 = 0.4
P(Y = 1) = P(Y = 1|X = 1) + P(Y = 1|X = 0) = 0.2 + 0.15 = 0.35
P(Y = 0) = P(Y = 0|X = 1) + P(Y = 0|X = 0) = 0.4 + 0.25 = 0.65

(b) In order for X and Y to be independent of one another, the following relationship must hold:

E[XY ] = E[X] · E[Y ]

Calculating E[XY ] is easy because there is only one quadrant where both values are nonzero
and, thus, contributing to the expectation. Therefore, E[XY ] = 0.2.
Similarly, E[X] = P (X = 1) = 0.6 and E[Y ] = P (Y = 1) = 0.35. Taking the product, we
have E[X] · E[Y ] = 0.6 × 0.35 = 0.21 6= 0.2
This implies that X and Y are not independent.

(c) The conditional expectation, E[X|Y = 1] is given by:

E[X|Y = 1] = (1) · P(X = 1|Y = 1) + (0) · P(X = 0|Y = 1)

= P(X = 1|Y = 1)
P(X = 1, Y = 1)
=
P(Y = 1)
0.2
=
0.35
= 0.5714

© 2023 The Infinite Actuary, LLC Page 34

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 5: Question 2). Let the random variable Xn be a binomial process:

n
X
Xn = Bi
i=1

Where successive instances of Bi are independent and Bi is distributed according to:

(
1 with probability p
Bi =
0 with probability 1 − p

(a) Calculate P[X4 > 0] and P[X4 > 2]

(b) Calculate the expected value and variance of X3

© 2023 The Infinite Actuary, LLC Page 35

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 5, Question 2

(a) There is one path for X4 = 0 which is for all the Bi to be equal to 0. The probability of this
happening is (1 − p)4 . Therefore, the probability of X4 being greater than 0 is:

P[X4 > 0] = 1 − (1 − p)4

To calculate P[X4 > 2] we can map out all the cases where X4 = 3 and X4 = 4:

P[X4 = 4] = P[B1 = 1] × P[B2 = 1] × P[B3 = 1] × P[B4 = 1] = p4

To calculate the probability that X4 = 3 we can list out all the combinations of Bi that will
get us there:

P[X4 = 3] = P[Bi = 1, 1, 1, 0]+P[Bi = 1, 1, 0, 1]+P[Bi = 1, 0, 1, 1]+P[Bi = 0, 1, 1, 1] = 4p3 (1−p)

Thus,
P[X4 > 2] = 4p3 (1 − p) + p4

(b) For this, we can simply leverage our knowledge of the binomial model. We know that the
expected value and variance are given by:

E[Xn ] = np ∴ E[X3 ] = 3p

V[Xn ] = np(1 − p) ∴ V[X3 ] = 3p(1 − p)

© 2023 The Infinite Actuary, LLC Page 36

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 5: Question 3). Let Z be a random variable that is exponentially distributed with
λ > 0 such that:

P(Z < z) = 1 − e−λz

Also let S be a random variable that is the sum of two independent exponential variables, Z1 and
Z2 with the same parameter, λ:

S = Z1 + Z2

(a) Calculate E[Z] and V[Z]

(b) Calculate E[S] and V[S]

© 2023 The Infinite Actuary, LLC Page 37

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 5, Question 3

(a) Note: For this part, you may just know these answers off the top of your head from the
prelims, school, etc. That is perfectly fine! If you want to see more information, you can read
through the work below. Just make sure you get the same final answer.
The density function of the variable is given by:

∂P(Z < z)
f (z) = = λe−λz
∂z
Therefore, the expected value of Z (using integration by parts) is given by:

Z ∞ ∞ Z ∞
−λz 1
E[Z] = λ λz
ze dz = −ze + e−λz dz =
0 0 0 λ

Similarly for the variance, we need to calculate the expected value of Z 2 . Using deterministic
calculus (assumed as background knowledge for the QFI exams) gives:
Z ∞
2
2
E[Z ] = λ z 2 e−λz dz =
0 λ2
2
2 1 1
V[Z] = − = 2
λ2 λ λ

(b) Since the two distributions share a mean, the sum of the two exponential distributions is a
gamma distribution with parameters α = 2 and θ = λ1 . We can simply obtain the mean and
variance from the following identities:

2
E[S] = αθ =
λ
2
V [S] = αθ2 =
λ2

© 2023 The Infinite Actuary, LLC Page 38

QFI Quant Online Seminar – MFD Practice Questions 2023

A random variable Z has a Poisson distribution:

λk e−λ
P(Z = k) =
k!

for k = 0, 1, 2...

(a) Use the following expansion:

λ2 λ3
eλ = 1 + λ + + + ...
2! 3!
to validate that

∞
X
P(Z = k) = 1
k=0

(b) Suppose we have two independent random Poisson variables: Z1 and Z2 with parameters λ1
and λ2 respectively. What is the distribution of Z1 + Z2 ? State the conclusion verbally, but
also make sure to mathematically justify your response.

© 2023 The Infinite Actuary, LLC Page 39

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 5, Questions 4-5

(a) The expression for e that was provided can be rewritten as:

λ0 λ1 λ2 λ3
eλ = + + + + ...
0! 1! 2! 3!
∞
X λk
=
k!
k=0

Therefore, we have:

∞ ∞
X X λk e−λ
P(Z = k) =
k!
k=0 k=0
∞
X λk
= e−λ
k!
k=0
−λ λ
=e e
= 1

(b) The distribution we are looking for is:

P(Z1 + Z2 = k)

for k = 0, 1, 2...
For any given Z1 = i, Z2 must be equal to k − i in order for the variables to sum to k.
Therefore, the expression above can be rewritten:

k
X
P(Z1 + Z2 = k) = P(Z1 = k − i, Z2 = i)
i=0
k
X e−λ1 λk−i
1 e−λ2 λi2
=
(k − i)! i!
i=0
k
X λk−i
1 λ2
i
= e−(λ1 +λ2 )
i!(k − i)!
i=0
k
e−(λ1 +λ2 ) X k!
= λk−i λi2
k! i!(k − i)! 1
i=0
k
e−(λ1 +λ2 ) X k k−i i
= λ λ2
k! i 1
i=0

e−(λ1 +λ2 )
= (λ1 + λ2 )k
k!

© 2023 The Infinite Actuary, LLC Page 40

QFI Quant Online Seminar – MFD Practice Questions 2023

We recognize this as the probability function for a Poisson random variable with parame-
ter (λ1 + λ2 ). Therefore, the sum of two independent Poisson random variables
is, itself, a Poisson random variable with a parameter equal to the sum of the
parameters from the variables being added.

Z = Z1 + Z2 ∼ Pois(λ1 + λ2 )
Note that we relied on the following two relationships in order to derive this proof:

n
n
X n i n−i
(a + b) = ab
i
i=0

n n!
=
i i!(n − i)!

© 2023 The Infinite Actuary, LLC Page 41

QFI Quant Online Seminar – MFD Practice Questions 2023

Let Y be a random variable with E[Y ] < ∞

(a) Show that the following function is a martingale:

Mt = E[Y |It ]

Note: For all questions in this problem set, you may assume you only need to validate the key
martingale property, and the other two technical conditions can be ignored and are assumed
to be satisified.

(b) Given the ever increasing sequence of information sets (I0 ⊆ I1 ⊆ ...IT ⊆), is every conditional
expectation a martingale?

© 2023 The Infinite Actuary, LLC Page 42

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 6, Question 1

(a) Note that notationally, It represents all available information at time t.

In order for a function to be a martingale, the expectation of the future value of this function
given all available information should be equal to its current value. Let us see if this is the
case for our function for an arbitrary future time t + s:

E(Mt+s |It ) = E[E(Y |It+s )|It ]

= E(Y |It )
= Mt

Which means that Mt is a martingale!

(b) Yes. The above result shows that every conditional expectation is a martingale provided the
conditioning is with respect to the same filtration.

© 2023 The Infinite Actuary, LLC Page 43

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 6: Question 2).

Consider the random variable:

n
X
Xn = Bi
i=1

Where Bi is based on the result of a fair coin flip:

(
+1 Head
Bi =
−1 Tail

Since the coin is fair, it has a 50% probability of landing heads and 50% probability of landing
tails. Assume for parts (a) - (d) that the coin is fair.

(a) Calculate the following:

E[X4 |I1 ]
E[X4 |I2 ]
E[X4 |I4 ]

(b) What does this tell us about the following function:

Zi = E[X4 |Ii ] where i = 1, 2, 3, 4

(c) Now consider the function2 :

√
Vi = Xi + i

Is Vi a martingale?
(d) Can Vi be converted into a martingale by a transformation, but under the same probability
measure? Provide such a conversion or explain why it is not possible.
(e) For part (e), you can select any rigged coin with a constant probability of heads p. Can Vi be
converted into a martingale by converting to a different probability measure? Provide such a
conversion or explain why it is not possible.
Note: In other words, for part (e), if a rigged coin with any constant probability p of heads
is selected, is it possible to find an appropriate p such that Vi is converted to a martingale?
Due to the physical constraints of the problem, assume that you can find a rigged coin with
any probability of heads p, but that this probability is constant over time.
2
√ √
Note: The textbook defines Vi = Bi + i but I think they meant to say Vi = Xi + i, so that’s been corrected
here. For reference, the reason why I think they meant to use Xi is they write E(Bi+k |Bi ) = Bi in their solution -
which doesn’t make sense since that expected value would be 0. But if they swapped in X then it makes more sense.
You can simply just use this PDF which has this corrected for you.

© 2023 The Infinite Actuary, LLC Page 44

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 6, Question 2

(a) Given that a coin has a 50% chance of being heads and 50% chance of getting tails, the
unconditional expected value of Bn and Xn is 0. If we flip our first coin, we now have I1
which contains the value of X1 . Looking forward from this point, our expectation value of all
future Bi = 0. Therefore:

E(X4 − X1 ) = 0

This means that our expectation of X4 is simply our current value of X1 :

E(X4 |I1 ) = X1

Similarly:

E(X4 |I2 ) = X2

E(X4 |I4 ) = X4

(b) Clearly, based on the above results, we can conclude that Zi is a martingale. We can see this
reinforced from the results of MFD Ch6, Q1.

(c) Is Vi a martingale? Justify your answer.

◦ No, V is not a martingale
√
◦ √
We know that Vi = Xi + i. The process Xi is a martingale, but then adding the
i causes Vi to not be a martingale because it adds positive drift. In other words, Vi
is expected to grow as i increases which is not a characteristic of processes that are
martingales. Therefore, Vi is not a martingale.
◦ The above bullet point gives us the verbal intuition for why we know we don’t have a
martingale. We can also see this mathematically:
√ √
EtP [VT ] = Vt + T − t 6= Vt

(d) Can Vi be converted into a martingale by a transformation, but under the same probability
measure? Provide such a conversion or explain why it is not possible.
√
◦ Yes, V̄i = Vi − i is a martingale (this removes the drift piece discussed in c).

(e) Can Vi be converted into a martingale by converting to a different probability measure?

Provide such a conversion or explain why it is not possible.
◦ No. The drift component subtracted for (d) is not constant over i. Therefore, a simple
change of probability measure cannot convert Vi to a martingale.

© 2023 The Infinite Actuary, LLC Page 45

QFI Quant Online Seminar – MFD Practice Questions 2023

© 2023 The Infinite Actuary, LLC Page 46

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 6, Question 3

We test for a martingale by checking the future expectation of Xt+s given the current information
set, It

(a)

E(Xt+s |It ) = E(2Wt+s + (t + s)|It )

= 2Wt + (t + s) 6= 2Wt + t

Therefore, Xt is not a martingale

(b) Note: If you are having trouble following the steps below, I highly recommend watching the
“Increments” review video!

Therefore, Xt is not a martingale

(c) As we can see above, the incremental time piece, s prevents Xt = Wt2 from being a martingale.
Therefore, transforming the previous problem by subtracting this piece makes it a martingale.
This is the case for Xt = Wt2 − t. Therefore, this function is a martingale

© 2023 The Infinite Actuary, LLC Page 47

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 6: Question 4).

You are given the following representation that holds true given a sequence of information sets It :

Z T
MT (Xt ) = M0 (X0 ) + g(t, Xt )dWt
0

where Xt follows the SDE:

dXt = µdt + σdWt

Given the following cases of M (·), determine the g(·) term:

(a) MT (XT ) = WT

(b) MT (XT ) = WT2 − T

Note that as a hint for (b), you are given the following equation:

Z T
1 2
Wt dWt = WT − T
0 2

© 2023 The Infinite Actuary, LLC Page 48

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 6, Question 4

Given the representation:

Z T
MT (Xt ) = M0 (X0 ) + g(t, Xt )dWt
0

dXt = µdt + σdWt

The right hand side is always a martingale since g(t, Xt ) is adapted to filtration generated by
Wt . Our work will involve some guesswork and seeing how the chosen term for g(·) fits given the
equation.

(a) MT (XT ) = WT
RT
In this case, it looks like the entire term M0 (X0 ) + 0 g(t, Xt )dWt simply needs to reduce to
WT . This is the case when g(t, Xt ) = 1 , because we see in this case:
Z T
MT (Xt ) = M0 (X0 ) + dWt = W0 + WT = WT
0

(b) MT (XT ) = WT2 − T

The right hand side of the equation should look familiar given the hint:
Z T
1 2
Wt dWt = WT − T
0 2

Therefore, if we let g(t, Xt ) = 2Wt we get:

Z T
MT (Xt ) = M0 (X0 ) + 2Wt dWt = (W02 − 0) + WT2 − T = WT2 − T
0

© 2023 The Infinite Actuary, LLC Page 49

QFI Quant Online Seminar – MFD Practice Questions 2023

Show that Y (t), (t ≥ 0) is a martingale when:

c2 t

Y (t) = exp cW (t) −
2

Where c is an arbitrary constant.

Note: Remember that for all questions in this problem set, you may assume you only need to
validate the key martingale property, and the other two technical conditions can be ignored and are
assumed to be satisified. This textbook is not very rigorous about always checking the two technical
properties; however, on exam day (especially if the question is worth a high number of points) I
recommend confirming the two additional martingale technical properties. For more info, see the
“Martingales” section of the supplemental FAQ DSG!

Additionally, simplify the following expression:

A = E0 (ekWt )

© 2023 The Infinite Actuary, LLC Page 50

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 7, Question 2

Remember that in order for a variable to be a martingale, the future expected value based on the
presently available information set should be equal to the current value:

E[Y (t + s)|It ] = Et [Y (t + s)] = Y (t)

The main idea of the steps below is to split into the information known at time t, and the increment
from time t to time t + s (for more info, check out the Increments review video!):
h i
c2 (t+s)
Et [Y (t + s)] = Et exp(c · W (t + s) − 2 )
h i
c2
= Et exp(c[W (t) + W (t + s) − W (t)] − 2 [t + s])

c2
h i
c2 t
= exp cW (t) − 2 · Et exp(c[W (t + s) − W (t)] − s)
2
| {z }
1
h i
c2 t
= exp cW (t) − 2 = Y (t)

Thus, we have validated the key martingale property that the expected future value equals the
current value.

Note that from the property below, it is clear that:

c2

Et exp(c[W (t + s) − W (t)] − s) = 1
2

For the last question, we know this property from the moment generating function review video:

2t
A = E0 (ekWt ) = e.5k

Note: The equation above is a very important equation you want to have memorized for exam day!
It is explained further in the moment generating function review video.

© 2023 The Infinite Actuary, LLC Page 51

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 8: Question 1). Determine the limit as n → ∞ for each of the quantities below:

1 2 k−1

(a) 1 1 − n 1− n ... 1 − n
λ n

(b) 1 − n

λ k

(c) 1 − n

© 2023 The Infinite Actuary, LLC Page 52

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 8, MFD Question: 1

(a) Focusing purely on the second term, we can see that that as n → ∞, 1 − n1 → 1. This is

true of every term in the sequence since for any k, the ratio of nk will tend to 0 as n grows
larger. Therefore,

1 2 k−1
lim 1 1 − 1− ... 1 − = 1
n→∞ n n n
λ n

(b) If we let Fn = 1 − n , we can try to derive a meaningful limit:

n−λ
ln[Fn ] = n ln
n
n−λ

ln n
ln[Fn ] = 1
n

Then, by applying L’Hopital’s rule3 and the Quotient rule4 , we have that:
h i
1 n−(n−λ)
n−λ · n2
n
→
− n12

n λ
→ · 2 · −n2
n−λ n
−λn
→ → −λ
n−λ
Summarizing, this shows that

lim ln[Fn ] = −λ
n→∞

Thus,

Fn → e−λ

λ k

(c) lim 1 − = 1k = 1
n→∞ n

3
According to L’Hopital’s rule, if the limit of a numerator and denominator are both 0, then the following
f (x) f 0 (x)
relationship holds: lim = lim 0
n→∞ g(x) n→∞ g (x)
0
(x)g 0 (x)
4
If two functions, f (x) and g(x) are differentiable then ( fg(x)
(x) 0
) = f (x)g(x)−f
(g(x))2

© 2023 The Infinite Actuary, LLC Page 53

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 8: Question 2). Consider the random variable:

n
X
Xn = Bi
i=1

Where Bi is independent and distributed according to:

(
+1 with probability p
Bi =
0 with probability 1 − p

(a) Suppose now, n → ∞ while p → 0 such that λ = np. What is the probability P (Xn = k) in
terms of λ, n and k?

(b) Let n → ∞. Show that this formula reduces to the Poisson distribution:

λk e−λ
P(Xn = k) =
k!
(c) Verbally interpret using the limits n → ∞ and p → 0. Can you relate this to insurance
applications?

© 2023 The Infinite Actuary, LLC Page 54

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 8, MFD Question: 2

(a) This is the binomial distribution which we are familiar with:

n k
P(Xn = k) = p (1 − p)n−k
k

Substituting λ = np

k
λ n−k

n λ
P(Xn = k) = 1−
k n n

(b) Using the solution above and simplifying, we have:

k
λ n−k λk λ n−k

n λ n!
1− = 1−
k n n k! (n − k)!nk n
n
λk n(n − 1)(n − 2)...(n − k + 1) 1 − nλ
= · ·
k! nk 1 − nλ
k
n
λk 1 − nλ

1 2 k−1
= 1 1− 1− ... 1 − k
k! n n n 1− λ n

But remember that as n → ∞:

1 2 k−1
1 1− 1− ... 1 − →1
n n n

λ k

1− →1
n
λ n

1− → e−λ
n

So our expression reduces to:

λk −λ
lim P(Xn = k) = e
n→∞ k!

(c) The probability of an event occuring decreases but the number of trials from which an event
can occur increases. This is similar to insurance – where many people are insured and often
the probability of any one individual having an accident or claim is relatively small. This
relates closely to the rare events framework and the Poisson model.

© 2023 The Infinite Actuary, LLC Page 55

QFI Quant Online Seminar – MFD Practice Questions 2023

(a) Calculate P r(Wt ≥ a|Ta > t)

(b) Calculate P r(Wt ≥ a|Ta ≤ t)

(c) Determine the CDF of Ta

© 2023 The Infinite Actuary, LLC Page 56

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 8, MFD Question: 3

(a) This part should be very easy if you understand how the notation is set up.
If Ta > t, this means the process Wt first hits barrier a after time t. This means that Wt
must be less than a.
P r(Wt ≥ a|Ta > t) = 0

(b) First notice that WTa = a. That is, at time Ta , the Brownian motion has value a. We want
to see whether the value Wt is above or below a.
Since the distribution of a Brownian motion process is symmetric, we have that:
1
P r(Wt ≥ a|Ta ≤ t) = 2
1
P r(Wt ≤ a|Ta ≤ t) = 2
The first equation is the desired result:
1
P r(Wt ≥ a|Ta ≤ t) =
2
Note: If you struggled on part (b), think of a simple example to help you see this logic.
Suppose W3 = 5. Determine P r(W6 ≥ 5|W3 = 5). By symmetry, the probability is equal to
1 1
2 . Conditioned on W3 = 5, the probability that W6 is above 5 is 2 and the probability that
1
W6 is below 5 is 2 . In other words, Wt − WTa = W6 − W3 is normally distributed with mean
0, and thus is symmetric.

(c) First note that:

P r(Wt ≥ a) = P r(Wt ≥ a|Ta ≤ t)P (Ta ≤ t) + P r(Wt ≥ a|Ta > t)P (Ta > t)
= 12 P (Ta ≤ t) + 0 [we get this step by plugging in (a) and (b)]
Thus, P (Ta ≤ t) = 2 · P r(Wt ≥ a)
Luckily, Wt is a very familiar process, so P r(Wt ≥ a) is easy to compute.
Wt ∼ N (0, t). Therefore, P r(Wt < a) = Φ( √at )
P r(Wt ≥ a) = 1 − P r(Wt < a) = 1 − Φ( √at )
Translating this back to the process for T gives:

a
P (Ta ≤ t) = 2 · P r(Wt ≥ a) = 2 · 1 − Φ √
t

© 2023 The Infinite Actuary, LLC Page 57

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 9: Question 1). Let Ws be a Wiener process defined over [0, t]. Consider the
integral: Z t
Ws2 dWs
0

(a) Write an approximation of the above integral using the following three different Riemann
sums and evaluate the expected value of each:
n
X
2
(i) L = Wi−1 (Wi − Wi−1 )
i=1
n
X
(ii) R = Wi2 (Wi − Wi−1 )
i=1
n
X Wi2 + Wi−1
2
(iii) M = (Wi − Wi−1 )
2
i=1

(b) Write the integral in discrete time using an Ito sum and calculate the expectation of this sum

Note: This problem closely relates to the “Non-Anticipating” Review Video.

© 2023 The Infinite Actuary, LLC Page 58

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 9, MFD Question: 1

(a) The only difference between the three Reimann sums will be the time point at which the
integrand is evaluated (the height of the rectangle).
(i) Using the leftmost point in the partition:

n
" #
X
2
Expected Riemann sum = E Wi−1 (Wi − Wi−1 )
i=1

Based on our knowledge of the Brownian motion term, Wt , we know that Wt follows a
normal distribution with an expected value of 0 and a variance of t. That means the
expected value of future increments of Wt is 0. Therefore, the expected value of the
second term in the sum is 0:

n n
" #
X X
2 5
E Wi−1 (Wi − Wi−1 ) = ti−1 · 0 = 0
i=1 i=1

(ii) Using the rightmost point in the partition:

" n #
X
Expected Riemann sum = E Wi2 (Wi − Wi−1 )
i=1

We can expand the term as follows:

n n
" #
X X
E Wi2 (Wi − Wi−1 ) = (E[Wi3 ] − E[Wi2 Wi−1 ])
i=1 i=1

Since Wt is normally distributed, all odd moments are equal to 0 and the first term can
be eliminated:
n
X
=− E[Wi2 Wi−1 ]
i=1
n
X
=− E[E[Wi2 Wi−1 |Ii−1 ]]
i=1
n
X
=− E[Wi−1 E[Wi2 |Ii−1 ]]
i=1

The last term in the equation can be separated out into the known value of Wt−1 plus
the variance of (Wt − Wt−1 ):
n
X
2
=− E[Wi−1 (Wi−1 + (ti − ti−1 ))]
i=1
5
Note that because V ar(Wt ) = E[Wt2 ] − E[Wt ]2 , and E[Wt ] = 0, we show that V ar(Wt ) = E[Wt2 ] = t

© 2023 The Infinite Actuary, LLC Page 59

QFI Quant Online Seminar – MFD Practice Questions 2023

n
X n
X
3
=− E[Wi−1 ] − (ti − ti−1 )E[Wi−1 ]
i=1 i=1

= 0

(iii) Since this is the average of the previous two integrals, it also has an expected value of
0.

(b) The Ito sum is represented by the using the leftmost point in the partition:

n
X
2
Expected Ito sum = Wi−1 (Wi − Wi−1 )
i=1

This is identical to the first sum in part (a) so we know that it also has an expected value of
0.

© 2023 The Infinite Actuary, LLC Page 60

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 9: Bonus Question). Let Ws be a Wiener process defined over [0, t]. Consider the
integral: Z t
Ws dWs
0

Note: In the previous problem, we saw we got the same expected value for all three of the Riemann
sums. However, I wanted to add this bonus question to emphasize that stochastic integrals are very
sensitive to their integral definition, and you can indeed get different results, as will be seen in this
question.

(a) Write an approximation of the above integral using the following three different Riemann
sums and evaluate the expected value of each:
n
X
(i) L = Wi−1 (Wi − Wi−1 )
i=1
n
X
(ii) R = Wi (Wi − Wi−1 )
i=1
n
X Wi + Wi−1
(iii) M = (Wi − Wi−1 )
2
i=1

(b) Write the integral in discrete time using an Ito sum and calculate the expectation of this sum

© 2023 The Infinite Actuary, LLC Page 61

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 9

(a) (i) Using the leftmost point in the partition, we have independent increments so the expec-
tation evaluates to zero:

n n
" # " #
X X
Expected Riemann sum = E Wi−1 (Wi − Wi−1 ) = E(Wi−1 )E(Wi − Wi−1 ) = 0
i=1 i=1

(ii) To evaluate the right point integral, note that:

Wi · (Wi − Wi−1 ) = (Wi − Wi−1 )2 + Wi−1 (Wi − Wi−1 )

Thus, we have that the expected Riemann sum is calculated as follows:

E [ ni=1 Wi (Wi − Wi−1 )]
P
Pn 2

=E i=1 (Wi − Wi−1 ) + Wi−1 (Wi − Wi−1 )
Pn 2

=E i=1 (Wi − Wi−1 )

= t
(iii) Since this is the average of the previous two integrals, we get .5t .

(b) The Ito sum is represented by the using the leftmost point in the partition:

n
X
Expected Ito sum = Wi−1 (Wi − Wi−1 )
i=1

This is identical to the first sum in part (a) so we know that it also has an expected value of
0.

(c) This is false! Notice how we got different results between parts (i) - (iii) in this question!! The
stochastic integral evaluates to a different answer depending on how we define the stochastic
integral. For more information on this, make sure to check out the non-anticipating review
video.

© 2023 The Infinite Actuary, LLC Page 62

QFI Quant Online Seminar – MFD Practice Questions 2023

(a) Simplify the below quantities:

n
X n
X n
X
[tj Wtj − tj−1 Wtj−1 ] = [tj (Wtj − Wtj−1 )] + [(tj − tj−1 )Wtj−1 ]
j=1 j=1 j=1

(b) Use this information to show that:

Z t Z t
sdWs = tWt − Ws ds
0 0

(c) Derive the equation in (b) using Ito’s Lemma

(d) In the equation in (b), state which integral is defined in the sense of Ito only

Note: For part (c) and beyond, I recommend first going through the DSG and videos for MFD
Chapter 10.

© 2023 The Infinite Actuary, LLC Page 63

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 9, MFD Questions 2-4

(a) The sum on the left hand side of the equation is simply a telescoping sum and can be reduced:

n
X
[tj Wtj − tj−1 Wtj−1 ] = t1 Wt1 − t0 Wt0 + t2 Wt2 − t1 Wt1 + ... + tn Wtn − tn−1 Wtn−1
j=1

= tn Wtn − t0 Wt0
= tWt

The last simplification occurs because Wt0 = W0 = 0 and tn = t

To summarize, both sides of the equation give in (a) can be simplified greatly to tWt

(b) We have shown that our original equation can be rewritten as:

n
X n
X
tWt = [tj (Wtj − Wtj−1 )] + [(tj − tj−1 )Wtj−1 ]
j=1 j=1

In the limit, as the partition width goes to 0, the two terms on the right hand side can be
rewritten as an integral:

n
X Z t
[tj (Wtj − Wtj−1 )] → sdWs
j=1 0

n
X Z t
[(tj − tj−1 )Wtj−1 ] → Ws ds
j=1 0
Z t Z t
∴ tWt = sdWs + Ws ds
0 0
Z t Z t
sdWs = tWt − Ws ds
0 0
Rt
Note that the integral 0 sdWs is defined in the sense of the Ito integral only

© 2023 The Infinite Actuary, LLC Page 64

QFI Quant Online Seminar – MFD Practice Questions 2023

(c) Apply Ito’s Lemma with F (Wt , t) = tWt . Then a = 0 and σ = 1.

∂F ∂F 1 ∂2F ∂F
◦ dFt = [at + + σt2 2 ]dt + σt dWt
∂St |{z}
∂t 2 ∂St ∂St
| {z } | {z } | {z }
0 Wt 0 t

◦ dFt = Wt dt + tdWt
Rt Rt Rt Rt
◦ Thus, 0 dFs = 0 d[sWs ] = 0 Ws ds + 0 sdWs
Rt
◦ Also, 0 d[sWs ] = tWt
Z t Z t
◦ Therefore, sdWs = tWt − Ws ds
0 0
Z t Z t
◦ sdWs = tWt − Ws ds
0 0
Note that on exam day, approach (c) tends to be quicker than (b) for most people.
Rt
(d) 0 sdWs is an Ito integral. We have a dW increment, and the integrand is a non-anticipating
function.

© 2023 The Infinite Actuary, LLC Page 65

QFI Quant Online Seminar – MFD Practice Questions 2023

© 2023 The Infinite Actuary, LLC Page 66

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 9, MFD Question: 10

• Define F = .5Wt2 . This is a function that we will hope gives us something helpful. We will
guess and see if it yields something useful.

• Next, we will apply Baby Ito’s Lemma

1 ∂2F
• dFt = [ ∂F
∂t + 2 ∂Wt2 ]dt + ∂F
∂Wt dWt = .5dt + Wt dWt

• Then, d(.5Wt2 ) = Wt dWt + .5dt

Z t
• Integrating and rearranging yields Ws dWs = .5Wt2 − .5W02 − .5t = .5Wt2 − .5t
0

© 2023 The Infinite Actuary, LLC Page 67

QFI Quant Online Seminar – MFD Practice Questions 2023

Calculate the differential dF for each of the following:

(a) f (Wt ) = Wt2

√
(b) f (Wt ) = Wt
2
(c) f (Wt ) = eWt

(d) f (Wt ) = eσWt

σ2 t
(e) f (Wt ) = eσWt − 2

© 2023 The Infinite Actuary, LLC Page 68

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 10, MFD Question: 1

Remember that using Ito’s Lemma for differentiation in a stochastic environment, we have the
following two differentiation rules:

1
df (Wt ) = f 0 (Wt )dWt + f 00 (Wt )dt
2
1 00
df (Wt , t) = ft0 (Wt , t)dt + fW
0
t
(Wt , t)dWt + fW (Wt , t)dt
2 t

With that in mind, we can tackle each of the problems:

(a) f 0 (Wt ) = 2Wt

f 00 (Wt ) = 2
∴ d(Wt2 ) = 2Wt dWt + dt

1
(b) f 0 (Wt ) = 12 W − 2
3
f 00 (Wt ) = − 14 W − 2
√ 1 1 1 3
∴ d( W ) = W − 2 dWt − W − 2 dt
2 8

2
(c) f 0 (Wt ) = 2Wt eWt
2 2
f 00 (Wt ) = 2eWt + 4Wt2 eWt
2 2 2 2
∴ d(eWt ) = 2Wt eWt dWt + [eWt + 2Wt2 eWt ]dt

(d) f 0 (Wt ) = σeσWt

f 00 (Wt ) = σ 2 eσWt

∴ d(eσWt ) = σeσWt dWt + .5σ 2 eσWt dt

2 σ2 t σ2 t
(e) ft0 (Wt , t) = − σ2 eσWt − 2 0 (W , t) = σeσWt −
fW t t 2

σ2 t
00 (W , t) = σ 2 eσWt −
fW 2
t t
σ2 t σ2 t σ2 t 2
2 σ 2 σWt − σ2 t
∴ d(eσWt − 2 ) = − σ2 eσWt − 2 dt + σeσWt − 2 dWt + 2 e dt
σ2 t σ2 t
∴ d(eσWt − 2 ) = σeσWt − 2 dWt
Note that this is a martingale since there is no drift term.

© 2023 The Infinite Actuary, LLC Page 69

QFI Quant Online Seminar – MFD Practice Questions 2023

(d) Xt = (Wt1 + Wt2 )2

© 2023 The Infinite Actuary, LLC Page 70

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 10, MFD Question: 2

As we know, the differentiation rule for a stochastic environment are:

∂X 1 ∂2X ∂X
d(X(Wt , t)) = dWt + 2
dt + dt
∂Wt 2 ∂ Wt ∂t

Using these, we can proceed:

(a) Since there is no t term we have :

∂X 1 ∂2X ∂X
= 4Wt31 = 6Wt21 =0
∂Wt 2 ∂ 2 Wt ∂t

∴ d(X(Wt , t)) = 4Wt31 dWt1 + 6Wt21 dt

(b) This equation does involve a t term so we have:

∂X 1 ∂2X eWt1 ∂X
= eWt1 = = 2t
∂Wt 2 ∂ 2 Wt 2 ∂t

eWt1

∴ d(X(Wt , t)) = eWt1 dWt1 + + 2t dt
2

(c) Here, we have that:

∂X 2 1 ∂2X Xt ∂X
= et +Wt1 = Xt 2
= = 2tXt
∂Wt 2 ∂ Wt 2 ∂t

1
∴ d(X(Wt , t)) = Xt dWt1 + + 2t Xt dt
2

(d) Since this involves two independent Brownian motion terms, we don’t have to worry about
the correlation between Wt1 and Wt2 . There is no t term so we can ignore the final term of
our original equation.

∂X 1 ∂2X ∂X 1 ∂2X
d(X(Wt1 , Wt2 )) = dWt1 + dt + dW t2 + dt
∂Wt1 2 ∂ 2 Wt1 ∂Wt2 2 ∂ 2 Wt2

∂X ∂X 1 ∂2X 1 ∂2X
= = 2(Wt1 + Wt2 ) = =1
∂Wt1 ∂Wt2 2 ∂ 2 Wt1 2 ∂ 2 Wt2

∴∴ d(X(Wt , t)) = 2(Wt1 + Wt2 )dWt1 + 2(Wt1 + Wt2 )dWt2 + 2dt

© 2023 The Infinite Actuary, LLC Page 71

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 10: Question 3). Let Wt be a Wiener process. Consider the geometric process:

1 2 )t+σW
St = S0 e(µ− 2 σ t

Given the information above, answer the following:

(a) Calculate dSt

(b) What is the expected rate of change of St ?

© 2023 The Infinite Actuary, LLC Page 72

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 10, MFD Question: 3

(a) Using the following formula, we can derive the desired identity, dSt

∂S(Wt , t) ∂S(Wt , t) 1 ∂ 2 S(Wt , t)

dS(Wt , t) = dt + dWt + dt
∂t ∂Wt 2 ∂Wt2

Calculating the partial derivatives, we have:

∂S(Wt , t) σ2
= µS(Wt , t) − S(Wt , t)
∂t 2
∂S(Wt , t)
= σS(Wt , t)
∂Wt
∂ 2 S(Wt , t)
= σ 2 S(Wt , t)
∂Wt2

Therefore, we have:

σ2 σ2

dS(Wt , t) = S(Wt , t) µ − + dt + σdWt
2 2

Thus,

dSt = µSt dt + σSt dWt

(b) The expected rate of change of St is µ since the expected movement of the dWt term is 0.
Note that since the question asked for the rate of change, the correct answer is µ and should
not involve the entire drift term.

© 2023 The Infinite Actuary, LLC Page 73

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 10: Question 4). Prove that E0 (Wt4 ) = 3t2 using the moment generating function
1 2
of a standard normal random variable Z given by MZ (t) = e 2 t

© 2023 The Infinite Actuary, LLC Page 74

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 10, MFD Question: 4

• E0 (Z 4 ) = 3 by the kurtosis of a standard normal RV∗

• W
√t
t
= Z ∼ N (0, 1)
√
• E0 (Wt4 ) = E0 ((Z t)4 ) = E0 (Z 4 t2 ) = t2 E0 (Z 4 ) = 3t2

• Proof of (*):
Can calculate fourth moment of Z using the MGF
1 2
MZ (t) = e 2 t
d4
E0 (Z 4 ) = dt4
MZ (t)|t=0
d4 12 t2
= dt4
e |t=0
1 2
= e 2 t (t4 + 6t2 + 3)|t=0
=3
Thus, E0 (Z 4 ) = 3

© 2023 The Infinite Actuary, LLC Page 75

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 11: Question 2). Consider the geometric SDE:

dSt = rSt dt + σSt dWt

Suppose a coin toss is used to approximate the dWt term based on the following:
( √
+ ∆ if Heads with probability 0.5
∆Wt = √
− ∆ if Tails with probability 0.5

(a) Describe how you could use a coin toss to generate random errors that will approximate the
dWt term. Will the limiting mean and variance of this process match that of dWt as t → 0?

(b) Assuming a current stock value of S(0) = $940, annual volatility of 15%, a constant, contin-
uous interest rate of 5%, time intervals of 2 years and the four coin flips turn out {H, T, T,
H}, simulate the stock price at the following times S(2), S(4), S(6) and S(8).

© 2023 The Infinite Actuary, LLC Page 76

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 11, MFD Question: 2

(a) If we approximate dWt ≈ ∆Wt , we have:

( √
+ ∆ if Heads with probability 0.5
∆Wt = √
− ∆ if Tails with probability 0.5

Then ∆Wt has a mean of 0 and a variance of ∆. Yes, this is identical to the distribution of
dWt

(b) Based on the geometric SDE, for all future time steps, we have the following equation for
future values of S(t):

σ2
S(t + 2) = S(t)e2[r− 2
]+σ∆W

Using this formula, we have:

0.152
√
S(2) = S(0)e2[0.05− 2
]+0.15 2
= $1, 255.78
0.152
√
S(4) = S(2)e2[0.05− 2
]−0.15 2
= $1, 097.60
0.152
√
S(6) = S(4)e2[0.05− 2
]−0.15 2
= $959.34
0.152
√
S(8) = S(6)e2[0.05− 2
]+0.15 2
= $1, 281.62

© 2023 The Infinite Actuary, LLC Page 77

QFI Quant Online Seminar – MFD Practice Questions 2023

fxx + fyy + fzz = 0

• Do the following equations satisfy Laplace’s equation?

(i) f (x, y, z) = 4z 2 y − x2 y − y 3
(ii) f (x, y) = x2 − y 2
(iii) f (x, y) = x3 − 3xy
x
(iv) f (x, y, z) = y+z

© 2023 The Infinite Actuary, LLC Page 78

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 12, MFD Question: 1

• We break down the partial derivatives, fxx , fyy , fzz for each equation as follows:
(i)
fxx = −2y fyy = −6y fzz = 8y
∴ fxx + fyy + fzz = 0

Yes, this is a Laplace equation

(ii)
fxx = 2 fyy = −2
∴ fxx + fyy = 0

Yes, this is a Laplace equation

(iii)
fxx = 6x fyy = 0
∴ fxx + fyy 6= 0

No, this is not a Laplace equation

(iv)
2x 2x
fxx = 0 fyy = fzz =
(y + z)3 (y + z)3
4x
∴ fxx + fyy + fzz = 6= 0
(y + z)3
No, this is not a Laplace equation

© 2023 The Infinite Actuary, LLC Page 79

QFI Quant Online Seminar – MFD Practice Questions 2023

ft = a2 (fxx + fyy + fzz )

Where a is a constant. Do the following functions satisfy the heat equation?

2 π 2 t+π(3x+2y+4z)]
(a) f (x, y, z, t) = e[29a

(b) f (x, y, z, t) = 3x2 + 3y 2 − 6z 2 + x + y − 9z − 3

© 2023 The Infinite Actuary, LLC Page 80

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 12, MFD Question: 2

(a) We break down the partial derivatives, fxx , fyy , fzz , ft for each equation as follows:

ft = 29a2 π 2 f (x, y, z, t)
fxx = 9π 2 f (x, y, z, t)
fyy = 4π 2 f (x, y, z, t)
fzz = 16π 2 f (x, y, z, t)
∴ a2 (fxx + fyy + fzz ) = a2 (9 + 4 + 16)π 2 f (x, y, z) = 29a2 π 2 f (x, y, z) = ft

Therefore, this PDE is a heat equation

(b) Since there is no dependence on t, ft = 0 and so we must satisfy fxx + fyy + fzz = 0. Let us
see if this is the case:

fxx = 6 fyy = 6 fzz = −12

∴ fxx + fyy + fzz = 6 + 6 − 12 = 0

Therefore, this PDE is a heat equation

© 2023 The Infinite Actuary, LLC Page 81

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 12: Question 3). Consider the following partial differential equation:

fx + 0.2fy = 0 X ∈ [0, 1] Y ∈ [0, 1]

(a) How many functions f (x, y) will satisfy the condition?

(b) Suppose we are given the boundary equation f (0, Y ) = 1. Can we find a unique solution to
the equation?

© 2023 The Infinite Actuary, LLC Page 82

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 12, MFD Question: 3

(a) There are infinite solutions to this equation including anything of the form f (x, y) = k where
k is any constant.

(b) Now that we have this boundary condition, we can solve for the unique solution:

f (x, y) = 1

This is the only solution that satisfies our original equation and the boundary condition

© 2023 The Infinite Actuary, LLC Page 83

QFI Quant Online Seminar – MFD Practice Questions 2023

Consider Xt = eYt where:

Yt ∼ N(µt, σ 2 t)

In this case, Xt is lognormally distributed.

(a) Calculate the following expectation:

E[Xt |Xs , s < t]

(b) Determine the case when e−rt Xt is a martingale. Your answer should be an equation involving
r, µ and σ.

© 2023 The Infinite Actuary, LLC Page 84

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 13, MFD Question: 1-2

(a) First note that:

E[eYt |Ys , s < t] = E[e(Yt −Ys )+Ys |Ys , s < t]

E[eYt |Ys , s < t] = eYs E[e(Yt −Ys ) |Ys , s < t]

The eYs term can be removed since Ys is what we are conditioning on. The remaining term
in the exponent has a normal distribution:

Yt − Ys ∼ N(µ(t − s), σ 2 (t − s))

Leveraging the moments of a lognormal distribution, we get:

σ2
E[eYt |Ys , s < t] = eYs e(µ+ 2
)(t−s)

Thus,

σ2
E[Xt |Xs , s < t] = Xs e(µ+ 2
)(t−s)

(b) From the result in part (a), we get:

σ2
E[e−rt Xt |Xs , s < t] = e−rt Xs e(µ+ 2
)(t−s)

σ2
E[e−rt Xt |Xs , s < t] = e−rs e−r(t−s) Xs e(µ+ 2
)(t−s)

In order for this to be a martingale, the right hand side of the equation needs to be equal to
e−rs Xs :

σ2
e−rs e−r(t−s) Xs e(µ+ 2
)(t−s)
= e−rs Xs
σ2
e−r(t−s) e(µ+ 2
)(t−s)
=1
2
(µ−r+ σ2 )(t−s)
e =1
σ2
∴µ=r−
2

σ2
Therefore, e−rt Xt is a martingale when µ = r −
2

© 2023 The Infinite Actuary, LLC Page 85

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 13: Question 3). Consider

Zt = e−rt Xt

Where Xt is an exponential Wiener process (i.e. lognormal) defined as follows:

Xt = eWt

(a) Determine the SDE for Zt

(b) Calculate the expected value of Zt

(c) Modify the definition of Xt to make Zt is a martingale

© 2023 The Infinite Actuary, LLC Page 86

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 13, MFD Question: 3

(a) We can calculate dZt by using Ito’s Lemma:

∂Zt ∂Zt 1 ∂ 2 Zt
dZt = dt + dWt + dt
∂t ∂Wt 2 ∂ 2 Wt

∂Zt ∂Zt ∂ 2 Zt
= −rZt = Zt = Zt
∂t ∂Wt ∂ 2 Wt

1
∴ dZt = − r Zt dt + Zt dWt
2

(b) Since we know that Wt ∼ N (0, t), the mean of this lognormal distribution is:

t
E[eWt ] = e 2

To calculate the expected value, we have:

1
E[Zt ] = e−rt E[eWt ] = e( 2 −r)t

(c) Given the results from (a) above, we know that we would have a martingale if the dt term is
1 ∂ 2 Zt
equal to 0. There are two dt terms in the Ito’s Lemma: ∂Z
∂t dt and 2 ∂ 2 Wt dt. Since we know
t

that:

∂Zt
= −rZt
∂t
The equation we need to solve is:

1 ∂ 2 Zt
−rZt + =0
2 ∂ 2 Wt
√
A solution to this would be e 2rWt .

√
Therefore, Zt = e−rt e 2rWt
is a martingale

QFI Quant Online Seminar – MFD Practice Questions 2023

Consider the following random variable:


1 with probability P1 = 0.3

Y = −0.5 with probability P2 = 0.2

0.2 with probability P3 = 0.5


(a) Calculate the mean and variance of Y

(b) Change the mean of this random variable to 0.05 by subtracting the appropriate constant
from Y . Calculate the new variance

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 14, MFD Question: 1

(a) Hopefully this one was pretty easy:

E[Y ] = [(1 × 0.3) + (−0.5 × 0.2) + (0.2 × 0.5)] = 0.3

E[Y 2 ] = [(12 × 0.3) + ((−0.5)2 × 0.2) + (0.22 × 0.5)] = 0.37
V ar(Y ) = E[Y 2 ] − [E[Y ]2 ] = 0.37 − 0.32 = 0.28

(b) In order to transform the variable, we simply subtract the difference between the current
mean and the desired mean:

Z = Y − 0.25

The mean of Z is 0.05 as we intended. We could go through and calculate the variance of Z
as we did for Y or we could simply take advantage of the following identity from Exam P:

V ar(X + c) = V ar(X) If c is a constant

Therefore, the variance of Z is 0.28

(c) The two constraints we have are on the new mean (0.05), variance (0.28) and that the sum
of the probabilities must add up to 1:

P1 + P2 + P3 = 1
E[Z] = 0.05 = [(1 × P1 ) + (−0.5 × P2 ) + (0.2 × P3 )]
E(Y ) = V ar(Y ) + (E[Y ])2 = 0.2825 = [(1 × P1 ) + ((−0.5)2 × P2 ) + (0.22 × P3 )]
2

This involves three equations and three unknowns. Solving these, we have:

P1 = 0.1646 P2 = 0.4024 P3 = 0.4330

Thus we have transformed the variable to have the desired mean and maintained its variance
simply by changing the underlying probability measure used.

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 14: Question 2). Suppose a stock follows a lognormal distribution such that the
return follows:
log(Rt ) ∼ N(µ = 0.17, σ 2 = 0.09)

(a) Find a function, a(Rt ), to transform the density function of the stock’s return, f (Rt ), such
that under the transformed density, a(Rt )f (Rt ), log(Rt ) has a mean equal to the risk free
rate of 0.05

(b) Find a(Rt ) such that Rt has a mean of 0

(c) Under which probability is it easier to calculate E[Rt2 ]?

(d) Does the choice of transformation impact the mean/variance?

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 14, MFD Question: 2

(a) The density function for log(Rt ) is:

1 −(x−µ)2
f (x) = p e 2σ2
(2πσ 2 )

If our goal is to replace the µ with r, we can use the following function:

[−(Rt −r)2 +(Rt −µ)2 ] [−(Rt −5%)2 +(Rt −17%)2 ]

a(Rt ) = e 2σ 2 =e .18

(b) Similar to above, we can replace the Rt − r term above with Rt :

[−Rt2 +(Rt −.17)2 ]

a(Rt ) = e .18

(d) Both transformations adjusted the mean but not the variance.

QFI Quant Online Seminar – MFD Practice Questions 2023

H(St , t) = max[C(St , t), P (St , t)]

where the time t call and put option values are:

C(St , t) = e−r(T −t) E[max(ST − K, 0)|It ]

P (St , t) = e−r(T −t) E[max(K − ST , 0)|It ]

(a) Using the above two equations, show that:

C(St , t) − P (St , t) = St − Ke−r(T −t)

(b) Consequently, show that the option price at time zero is given by:

σ2 t
h i
H(S0 , 0) = C(S0 , 0) + e−rT E max[K − S0 erT eσWt − 2 , 0]

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 15, MFD Question: 1

(a) Starting with the definitions of C(St , t) and P (St , t) provided and simplifying them, we have:

C(St , t) − P (St , t) = e−r(T −t) E[max(ST − K, 0)|It ] − e−r(T −t) E[max(K − ST , 0)|It ]

= e−r(T −t) E[max(ST − K, 0) − max(K − ST , 0)|It ]

= e−r(T −t) E[ST − K|It ]

Thus6 , we have that C(St , t) − P (St , t) = e−r(T −t) E[ST − K|It ]

Under the risk neutral measure, the discounted stock price process is a martingale, so this
simplifies to:
C(St , t) − P (St , t) = St − e−r(T −t) K

This is, of course, put-call parity!

(b) Using our proof above, we can rewrite the value of the chooser option as:

H(St , t) = max[C(St , t), C(St , t) − St + e−r(T −t) K]

H(St , t) = C(St , t) + max[e−r(T −t) K − St , 0]

Therefore:
h i
H(S0 , 0) = e−rt E C(St , t) + max[e−r(T −t) K − St , 0]
h i
H(S0 , 0) = C(S0 , 0) + e−rt e−r(T −t) E max[K − er(T −t) St , 0]
1 2 )t+σW
Since we know that St = S0 e(r− 2 σ t
:
h 1 2
i
H(S0 , 0) = C(S0 , 0) + e−rT E max[K − S0 erT eσWt − 2 σ t , 0]

6
Note: If you have any trouble following the steps above, think about the two cases St ≥ K and K > St

QFI Quant Online Seminar – MFD Practice Questions 2023

(c) The second term on the right hand side of the equation can be rewritten as follows:
h i h i
e−rT E max[K − er(T −t) St , 0] = e−rt E max[Ke−r(T −t) − St , 0]

We recognize this as the payoff for a put option with strike Ke−r(T −t) expiring at t. Therefore,

H(S0 , 0) = C(S0 , 0, K, T ) + P (S0 , 0, Ke−r(T −t) , t)

We can evaluate these option values using the Black-Scholes formula:

H(S0 , 0) = [S0 N (d1 ) − Ke−rT N (d2 )] + [Ke−r(T −t) e−rt N (−d¯2 ) − S0 N (−d
¯ 1 )]

¯ 1 )] − Ke−rT [N (d2 ) − N (−d¯2 )]

H(S0 , 0) = S0 [N (d1 ) − N (−d

Where d1 and d2 are calculated based on strike K and maturity T . d¯1 and d¯2 are calculated
based on strike Ke−r(T −t) and maturity t

QFI Quant Online Seminar – MFD Practice Questions 2023

dSt = (r − f )St dt + σSt dWt

1 2 )t+σW
(a) Show that St = S0 e(r−f − 2 σ t

1
St ef t 2t
(b) Is the process S0 ert = eσWt − 2 σ a martingale under P ? Justify your answer.

(c) Let P̄ be the probability measure

Z
1 2T
P̄ (A) = eσWT − 2 σ dP
A

What does Girsanov’s theorem imply about the process Wt − σt under P̄ ?

1
(d) Find an expression for dZt where Zt = St .

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 15, MFD Question: 2

(a) Textbook Solution

◦ This is a fairly straightforward application of Ito’s Lemma with St = F (t, Wt ) =
1 2
S0 e(r−f − 2 σ )t+σWt
1 ∂2F
◦ dSt = [ ∂F
∂t + 2 ∂Wt2 ]dt + ∂F
∂Wt dWt
∂F
◦ ∂t = (r − f − 21 σ 2 )St
∂F
◦ ∂Wt = σSt
∂2F
◦ ∂Wt2
= σ 2 St

◦ Thus, dSt = [(r − f − 12 σ 2 )St + 21 σ 2 St ]dt + σSt dWt = (r − f )St dt + σSt dWt

◦ Therefore, dSt = (r − f )St dt + σSt dWt

Alternate Solution
Note: The solution above is the recommended textbook solution for part a. This solution would
be accepted on exam day because it follow the logic set forth in the textbook, but in my opinion
it is a bit backwards. It essentially starts with what you are trying to show and then gets the
SDE. An ideal solution would actually start with the SDE given in the question stem and then
show the equation you are given in part a. This is seen in the alternate solution below.
dSt = (r − f )St dt + σSt dWt
dSt
St = (r − f )dt + σdWt
By Ito’s Lemma Trick #1(a), we have that:
d ln St = (r − f − .5σ 2 )dt + σdWt
Integrating gives:
ln St − ln S0 = (r − f − .5σ 2 )t + σWt
Rearranging gives the desired result:
1 2 )t+σW
St = S0 e(r−f − 2 σ t

1
St ef t 2t
(b) Is the process S0 ert = eσWt − 2 σ a martingale under P ? Justify your answer.
◦ Yes. Using the MGF of a normal distribution, we can evaluate the following conditional
expectation to show that the process is a martingale under P .
1 2 (t+s) 1 2 (t+s) 1 2 (t+s) 1 2s 1 2 t+σW
◦ Et [eσWt+s − 2 σ ] = eσWt − 2 σ Et [eσ(Wt+s −Wt ) ] = eσWt − 2 σ e2σ = e− 2 σ t

1 2 (t+s) 1 2t
◦ This shows that Et [eσWt+s − 2 σ ] = eσWt − 2 σ → EtP [ST ] = St for all t < T .
◦ Technical points (in general, you should mention these but can do so briefly):
∗ The information set It contains all information up to time t
∗ St is known given It (St is It -adapted)
· We know this because It ⇒ Wt ⇒ St
∗ E P |St | < ∞

QFI Quant Online Seminar – MFD Practice Questions 2023

(c) Let P̄ be the probability measure

Z
1 2T
P̄ (A) = eσWT − 2 σ dP
A

What does Girsanov’s theorem imply about the process Wt − σt under P̄ ?

◦ Note that Girsanov’s theorem can be applied with Xu = σ where
Z t
◦ W̄t = Wt − Xu du = Wt − σt
0
◦ Then W̄t is a martingale under P̄ with
Z
◦ P̄ (A) = (WT )dP
A
Z T Z T
1
Xu dWu − Xu2 du 1 2T
◦ (WT ) = e 0 2 0 = eσWT − 2 σ
◦ Note: To get full credit for this question, all you need to state is that Wt − σt
is a martingale under P̄
1
(d) Find an expression for dZt where Zt = St .
1 2 −r)t−σW
◦ From (a), it should be obvious that Zt = Z0 e(f + 2 σ t

1 ∂2F
◦ dZt = [ ∂F
∂t + 2 ∂Wt2 ]dt + ∂F
∂Wt dWt
∂F
◦ ∂t = (f + 12 σ 2 − r)Zt
∂F
◦ ∂Wt = −σZt
∂2F
◦ ∂Wt2
= σ 2 Zt

◦ Therefore, dZt = Zt ((f − r + σ 2 )dt − σdWt )

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 16: Question 1). Plot the payoff diagrams for the following instruments:

(a) A caplet with a cap rate Rcap = 6.75% written on the 3-month LIBOR Lt that is about to
expire

(b) A 2 year forward contract written on a bond which expires in 3 months with a price of $89.5

(d) A 2 year fixed payer interest rate swap with swap rate κ = 7.5%. The swap has 1.5 years to
maturity and receives the 6 month LIBOR

(e) A swaption that expires in 6 months on a 2 year fixed-payer swap with swap rate κ = 6%

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 16, MFD Question: 1

(a) A caplet won’t have any payoff unless the underlying LIBOR rate is greater than the strike:

Caplet payoff per $1 of notional

Caplet Payoff ($) 8

0
0 2 4 6 8 10 12 14
3 month Libor Rate (%)

(b) Since a forward represents an obligation to buy the bond, the holder is essentially long the
bond (note that we intersect the axis at the price of 89.5):

Payoff from Forward Contract

150
Forward Contract Payoff ($)

100

−50

−100
0 50 100 150 200
Bond Price in 3 Months ($)

QFI Quant Online Seminar – MFD Practice Questions 2023

(c) A pay-fixed FRA breaks even at the fixed rate: 7.5% in this case. If market rates are below
this amount, the holder receives less than 7.5% and loses the difference. Conversely, if rates
are higher, the holder receives more than he pays. Essentially, they are long the rate and the
net payoff is centered around 7.5%

FRA payoff per $100 of Notional

10
FRA Payoff

−10
0 5 10 15 20
Libor at the end of 3 month period (%)

(d) Since we know an Interest Rate Swap is simply a series of FRAs, the payoff at each payment
date of the swap resembles the FRA above:

Fixer Payer IR Swap Payoff per $100 of Notional

20
Fixed Payer Swap Payoff

−10
0 5 10 15 20
Libor at the end of 3 month period (%)

QFI Quant Online Seminar – MFD Practice Questions 2023

(e) A swaption on the fixed leg of a swap will be exercised and pay off when rates exceed the
strike rate since the option buyer will receive the higher floating rate. If rates are below the
strike rate, the option will not be exercised and no payoff is generated:

Swaption payoff per $1 of notional

8
Swaption payoff ($)
6

0
0 2 4 6 8 10 12 14
Swap rates at swaption maturity (%)

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 17: Question 1). You are given the following dynamics for a spot rate, rt which has
a current value of 6%, an annual drift of µ = 1% and an annual volatility of σ = 12%:

drt = µrt dt + σrt dWt

Given the information above, answer the questions below:

(a) Break up the year into 5 periods and calculate the implied magnitude of the up and down
moves for a binomial tree for each period.

(b) Draw out this tree for the spot rate rt .

(c) What are the “up” probabilities implied by the tree?

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 17, MFD Question: 1

(a) In our case, since we have annual drift and volatility, we can work with ∆ = 15 . Since ∆ and
σ are fixed constants, we can use our familiar binomial tree formulas:

√ q
σ ∆ 0.12∗ 15
u=e =e = 1.0551
√ q
−0.12∗ 15
d = e−σ ∆
=e = 0.9477

(b) The binomial tree is shown below:

t=0 t=1 t=2 t=3 t=4 t=5

rt = 7.85%

rt = 7.44%

rt = 7.05% rt = 7.05%

rt = 6.68% rt = 6.68%

rt = 6.33% rt = 6.33% rt = 6.33%

rt = 6% rt = 6% rt = 6%

rt = 5.69% rt = 5.69% rt = 5.69%

rt = 5.39% rt = 5.39%

rt = 5.11% rt = 5.11%

rt = 4.84%

rt = 4.59%

QFI Quant Online Seminar – MFD Practice Questions 2023

(c) We can obtain the probabilities using the original SDE:

rt+1 = rt + µ∆rt + σrt ∆W

It then follows that:

E[rt+1 ] = pu (t)rt+1
u
+ (1 − pu (t))rt+1
d
= rt + µ∆rt
d
rt + µ∆rt − rt+1
∴ pu (t) = u − rd
rt+1 t+1

The “up” probabilities are shown below. For example, the t = 0 entry is calculated as:

d
rt + µ∆rt − rt+1 0.06 + 0.01 · 0.2 · 0.06 − 0.0569
put = u d
= = .5031
rt+1 − rt+1 0.0633 − 0.0569

t=0 t=1 t=2 t=3 t=4

put = 0.5061

put = 0.5318

put = 0.5047 put = 0.5047

put = 0.5039 put = 0.5039

put = 0.5031 put = 0.5031 put = 0.5031

put = 0.5105 put = 0.5105

put = 0.5013 put = 0.5013

put = 0.5095

put = 0.4994

Of course, pdt = 1 − put so those can be easily calculated as well.

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 17: Question 2). Suppose that you are given the following zero coupon bond prices:

P (0, 1) = 0.94 P (0, 2) = 0.92 P (0, 3) = 0.87 P (0, 4) = 0.80

(a) Calculate the 1-year spot rates implied by these bond prices. Assume that rates are quoted
using a convention of annual effective interest rates.

(b) Calculate the 1-year forward rates implied by these bond prices. Assume that rates are quoted
using a convention of annual effective interest rates.

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 17, MFD Question: 2

(a) First we calculate the spot rates implied by the bond prices:

1
P (0, 1) = ⇒ r1 = 6.383%
(1 + r1 )
1
P (0, 2) = ⇒ r2 = 4.257%
(1 + r2 )2
1
P (0, 3) = ⇒ r3 = 4.751%
(1 + r3 )3
1
P (0, 4) = ⇒ r4 = 5.737%
(1 + r4 )4

(b) Each of the 1-year forward rates, f0,1 , f1,2 , f2,3 can be bootstrapped from the spot rates. We
use the principle that the 1-year forward rate can ‘bridge the gap’ between time nodes on the
term structure.
f0,1 is the same as r1 , the one year spot rate. This is calculated as follows:

f0,1 = r1 = 6.38%

To calculate f1,2 we note that there should be no difference between buying a 2 year bond
compared to buying a 1 year bond today and then another one year bond when the first
expires. Mathematically:

(1 + r2 )2 = (1 + r1 )(1 + f1,2 )

(1.04257)2 = (1.06383)(1 + f1,2 ) ⇒ f1,2 = 2.17%

Similarly, we can calculate f2,3 from P (0, 2) and P (0, 3):

(1 + r3 )3 = (1 + r2 )2 (1 + f2,3 )

(1.04751)3 = (1.04257)2 (1 + f2,3 ) ⇒ f2,3 = 5.75%

And lastly, we have f3,4 :

(1 + r4 )4 = (1 + r3 )3 (1 + f3,4 )

(1.05737)4 = (1.04751)3 (1 + f3,4 ) ⇒ f3,4 = 8.75%

QFI Quant Online Seminar – MFD Practice Questions 2023

Consider the Vasicek spot rate model:

drt = α(µ − rt )dt + σdWt

You are also given that α is a strictly positive constant.

(a) Show that:

σ2
E[rs |rt ] = µ + (rt − µ)e−α(s−t) and V[rs |rt ] = 1 − e−2α(s−t) ,

2α t<s

(b) For fixed t and for s → ∞, what do these two equations imply for the conditional mean and
variance of the spot rate?

(c) Now suppose we add in a parameter to capture the market price of interest rate risk which is
constant at λ which adjusts the drift term by λσ. In this case, the spot rate process is given
by:
drt = α(µ − rt )dt + σλdt + σdWt
Derive an expression for the price of a bond expiring at time s.

(d) Using the following identities:

µB B = rt B + σ|Br |λ
σ B B = σ|Br |

Where Br is the partial derivative of B with respect to r and |Br | denotes the absolute value
of the partial derivative.

Show that the drift and diffusion parameters for this bond are given as:

σλ
µB = rt + 1 − e−α(s−t)
α
σ
σB = 1 − e−α(s−t)
α
µB −rt
Additionally, verify that λ = σB

(e) Comment on what happens to the drift and diffusion parameters as the bond approaches
maturity. This is the case when t → s.

(f) Comment on what happens to the drift and diffusion parameters when the bond has a very
long maturity. This is the case when s → ∞.

(g) Is it expected that the diffusion parameter σ B of the bond is independent of the market price
of risk?

(h) Is the risk premium proportional to volatility? Is the market price of risk proportional to the
risk premium? What is the importance?

QFI Quant Online Seminar – MFD Practice Questions 2023

(i) What is the sign of the partial derivative Br ? Justify your answer both mathematically and
verbally.

For part (j), you are additionally given that the default-free discount bond closed-form pricing
formula is given by:
1 −α(T −t) )(Q−r)−T Q− σ 2 (1−e−α(T −t) )2
B(t, T ) = e α (1−e 4α3

(j) Prove that:

ln(B(t, T ))
lim − =Q
T →∞ T
What does Q represent?

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 18, MFD Question: 1

(a) Remember that under the Vasicek model µ represents the long term average spot rate and α
represents the speed of mean reversion.

drt = α(µ − rt )dt + σdWt

drt + αrt dt = αµdt + σdWt (multiply both sides by eαt )
eαt [drt + αrt dt] = eαt [αµdt + σdWt ]

We can re-write this as:

d[eαt rt ] = eαt [αµdt + σdWt ] (Let’s integrate from t to s)

Z s Z s
eαs rs − eαt rt = αµ eαx dx + eαx σdWx
t t
Z s
eαs rs − eαt rt = µeαs − µeαt + σ eαx dWx (Now let’s multiply by e−αs )
t
Z s
−α(s−t) −α(s−t) −αs
rs = e rt + µ − µe + σe eαx dWx
t
Z s
rs = µ + (rt − µ)e−α(s−t) + σe−αs eαx dWx
t

Now we can try to formulate expressions for the expected value and variance of rs given rt :

E[rs |rt ] = µ + (rt − µ)e−α(s−t)

This holds true because:

Z s
E σe−αs eαx dWx = 0
t

To get the variance, we notice that only the integral affects the variance so:
Z s
−αs αx
V[rs |rt ] = V σe e dWx
t

Applying Ito’s Isometry gives:

s
e2αs − e2αt
Z
2 −2αs 2αx 2 −2αs
V[rs |rt ] = σ e e dx = σ e
t 2α
Thus, we get the desired result:

σ2
V[rs |rt ] = [1 − e−2α(s−t) ]
2α

QFI Quant Online Seminar – MFD Practice Questions 2023

(b) For a fixed t and s → ∞, we can see that the mean converges to µ and the variance converges
σ2
to
2α
(c) The interest rate dynamics now include an addition to the drift term:
drt = α(µ − rt )dt + σ[λdt + dWt ]

We know the price of a bond is given by:

h Rs i
B(t, s) = Et e− t rv dv

From part (a), we know that:

Z v
−α(v−t) −αv
rv = µ + (rt − µ)e + σe eαx [dWx + λdx]
t

Collecting our results gives that:

Rs Rv
(µ+(rt −µ)e−α(v−t) +σe−αv eαx (dWx +λdx)dv
B(t, s) = Et [e− t t ]
Rs
e−α(v−t) dv− ts (µ−µe−α(v−t) +σe−αv tv eαx (dWx +λdx))dv
R R
B(t, s) = Et [e−rt t ]

(d) We can start by calculating the partial derivative using the result from part (c):
R s −α(v−t) −α(s−t)
Br = ∂B(t,s)
∂r(t) = − t e dv B(t, s) = − 1−e α B(t, s)
Summarizing, we have that:
1−e−α(s−t)
|Br | = α B
Therefore, using the identity we were given:

µB B = rt B + σ|Br |λ
" #
1 − e−α(s−t)
µB B = rt B + σλ B
α

σλ
µB = rt + [1 − e−α(s−t) ]
α
For the diffusion term:

σ B B = σ|Br |
σ
σB = [1 − e−α(s−t) ]
α
Lastly, note that:

σλ −α(s−t) ]
µB − rt rt + α [1 − e − rt µB − rt
= σ =λ⇒λ=
σB α [1 − e
−α(s−t) ] σB

QFI Quant Online Seminar – MFD Practice Questions 2023

(e) This is the case where the bond approaches maturity (s − t → 0)

σλ
µB = rt + [1 − e−α(0) ] → rt
α
This is expected since once the bond approaches maturity, its value converges to $1 and the
return becomes the return of holding $1 which is exactly the short term rate
For the diffusion term:

σ
σB = [1 − e−α(0) ] → 0
α
Which is also expected since a bond approaching maturity loses all volatility since the price
converges to $1

(f) For long maturity bonds, we simply assume s → ∞ which means the drift and diffusion
parameters become:

σλ
µB = rt +
α
σ
σB =
α

(g) Yes, it is expected that the diffusion term is independent of the λ parameter. As we’ve
previously seen, when transforming variables, our mean may change but our variance remains
the same. Introducing the market price of risk, λ by Girsanov’s theorem, therefore, affects
the mean but not the diffusion component of an SDE

(h) The answers to the questions are no and yes. The risk premium µ − r does not depend on the
volatility of the bond. However, the market price of risk is proportional to the risk premium.
The market price of risk is simply the risk premium standardized by the bond volatility.
These relationships are important as they allow Girsanov’s theorem to alter the drift (through
the risk premium) but not the volatility of the bond price SDE under a change of measure.

(i) The partial derivative is negative. We can see this mathematically through the equation
−α(s−t)
Br = − 1−e α B ≤ 0. We know this verbally as well, because bond prices and interest
rates are inversely related.

(j) The yield of a bond, R(t, T ) is given by the following formula:

B(t, T ) = e−T R(t,T )

1
∴ R(t, T ) = − lnB(t, T )
T
From the bond price formula that is provided:

1 −α(T −t) )(Q−r)−T Q− σ 2 (1−e−α(T −t) )2

B(t, T ) = e α (1−e 4α3

ln(B(t, T )) 1 TQ σ2
− =− (1 − e−α(T −t) )(Q − r) + + (1 − e−α(T −t) )2
T αT T 4T α3

QFI Quant Online Seminar – MFD Practice Questions 2023

If we take the limit as T → ∞, all we are left with is:

ln(B(t, T ))
lim − =Q
T →∞ T

Therefore, Q represents the yield of a consul/perpetual bond (i.e. paying perpetual

coupons) with infinite maturity

QFI Quant Online Seminar – MFD Practice Questions 2023

There are only two assets to invest in:

• Risk-free borrowing and lending at ri , i = 0, 1

• A two-period bond with price B0 paying $1 at t = 2

(a) Set up a matrix equation with state prices ψ that gives the arbitrage-free prices of the two
assets as a function of ψ,r0 and r1

(b) Show how one can get the risk-neutral probabilities in this setting given ψ,r0 and r1

(c) If we adopt a normalization using the savings account, show that the bond price is:

1
B = B0 = EQ
0
(1 + r0 )(1 + r1 )

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 18, MFD Question: 2

(a) There are 4 possible states of the world at the end of t = 2: ψ uu , ψ ud , ψ du , ψ dd . For simplicity,
we can use a numerical denomination and set up the matrix as follows:

 
ψ1
1 (1 + r0 )(1 + r1 ) (1 + r0 )(1 + r1 ) (1 + r0 )(1 + r1 ) (1 + r0 )(1 + r1 ) ψ2 
=  
B0 1 1 1 1 ψ3 
ψ4

This implies two equations:

1 = (1 + r0 )(1 + r1 )ψ1 + (1 + r0 )(1 + r1 )ψ2 + (1 + r0 )(1 + r1 )ψ3 + (1 + r0 )(1 + r1 )ψ4

B0 = ψ1 + ψ2 + ψ3 + ψ4

(b) Risk-neutral probabilities can be obtained by setting pi = (1 + r0 )(1 + r1 )ψi . We can see from
the first equation in part (a) that the sum of these terms add up to one which ensures that
a probability measure has been obtained.

(c) Part (b) implies that:

pi
ψi =
(1 + r0 )(1 + r1 )

This definition of ψi normalizes the state prices using the savings account.
Using this and B0 = ψ1 + ψ2 + ψ3 + ψ4 from part (a), we have:

p1 p2 p3 p4
B0 = + + +
(1 + r0 )(1 + r1 ) (1 + r0 )(1 + r1 ) (1 + r0 )(1 + r1 ) (1 + r0 )(1 + r1 )

Which we recognize as a probability weighted average which is simply an expectation:

1
B0 = EQ
0
(1 + r0 )(1 + r1 )

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 19: Question 2).

We are given the following vector Markov process, Xt for a short rate rt and a long rate Rt :

r
Xt = t
Rt

Where:

1

rt+∆ α1.1 α1,2 rt ∆Wt+∆
= + 2
Rt+∆ α2,1 α2,2 Rt ∆Wt+∆

The error term is jointly normal and serially uncorrelated. All α coefficients are strictly positive.

Given the information above, answer the following:

(a) Derive a univariate representation for the short rate, rt

(b) Is rt Markov?

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 19, MFD Question: 2

(a) If we try to extract the equation implied by the matrix equation above, we get:

1
rt+∆ = α1,1 rt + α1,2 Rt + ∆Wt+∆
2
Rt+∆ = α2,1 rt + α2,2 Rt + ∆Wt+∆

Since we are interested in the first line of the matrix equation, we can try getting some of the
other terms from the second line in the hope of being able to substitute in for the Rt terms:

Rt = α2,1 rt−∆ + α2,2 Rt−∆ + ∆Wt2

2
Rt−∆ = α2,1 rt−2∆ + α2,2 Rt−2∆ + ∆Wt−∆

In other words, we lag the R equation and plug into the r equation.
Successive substitution of past values of Rt−i∆ into the original equation for rt+∆ leads to:

∞
" #
X
i−1 i 2
rt+∆ = α1,1 rt + α1,2 (α2,1 α2,2 rt−i∆ + α2,2 ∆Wt−i∆ ) + ∆Wt2 + ∆Wt+∆
1

i=1

(b) According to this representation, rt is not a Markov process since infinitely many past values
of rt are required in order to represent rt .

QFI Quant Online Seminar – MFD Practice Questions 2023

(d) Consider the system below. Can the Bi be determined independently?

   
B1 1 1
B2u B2d 

 B2  
 ψ1
 B3  = 
  
B3u B3d  ψ2
B4 B4u B4d
(e) In the system above, can all the {fi } be determined independently?

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 19, MFD Question: 3

(a) At time i = 0, how many possible states of the world are there at i = 3?
◦ There are 23 = 8 possible states

(b) Form an arbitrage portfolio that will guarantee a positive payoff at time i = 0 and nonnegative
payoff at times i ≥ 1.
◦ There are many possible answers. The key is to justify your answer and make sure there
is a positive initial payoff and nonnegative future payoffs.
◦ Example: Long the forward f0 and long 1.08 units of B1
◦ The forward has a cash flow of 1 at time 0 and -1.08 at time 1
◦ The bond has a cash flow of −.9 · 1.08 at time 0 and 1.08 at time 1
◦ The time 0 cash flows are 1 − .9 · 1.08 = .028
◦ The time 1 cash flows are 1.08 − 1.08 = 0
◦ The payoffs after time 1 are all 0

(c) Given a default-free zero-coupon bond, Bn that matures at time t = n, and all the forward
rates {f0 , . . . , fn−1 } obtain a formula that expresses Bn as a function of fi .
n−1
Y 1
◦ Bn =
1 + fi
i=0

(d) Can the Bi be determined independently?

◦ No, they depend on the same underlying rate term structure.

(e) In the system above, can all the {fi } be determined independently?
◦ No. Given a bond price, the forward rates are found recursively. The forward rate
depends on forward rates of shorter time horizons.

QFI Quant Online Seminar – MFD Practice Questions 2023

(Chapter 20: Question 1). You are given the following SDE for the instantaneous spot rate:

drt = σrt dWt

Where Wt is a Wiener process under the real-world probability and σ is a positive constant. We
are also given that r0 = 5%

(a) Obtain a PDE for the default-free discount bond price B(t, T ) under these conditions

(b) Interpret the market price of interest rate risk and its sign

(c) Derive an expression for rt

QFI Quant Online Seminar – MFD Practice Questions 2023

SOLUTION
MFD Chapter: 20, MFD Question: 1

(a) Start with the bond pricing PDE:

1
rt B = Bt + Br [(a(rt , t) − b(rt , t)λt )] + Brr b(rt , t)2
2
Where λt is the market price of risk
In our case, b(rt , t) = σrt and a(rt , t) = 0 so we get:

1
rt B = Bt − Br λt σrt + Brr σ 2 rt2
2

(b) The market price of risk equals µ−r σ where µ and σ are the drift and volatility parameters of
the bond. Generally speaking, this is the compensation investors require in order to be lured
into taking risks. λ is positive because the expected return of the bond should be greater
than the risk free rate. If it is not, investors would simply choose to invest in a risk-free bond
instead.

(c) Start with the given SDE:

drt = σrt dWt

Applying Ito’s Lemma Trick (proof in supplemental FAQ DSG) gives:

d ln(rt ) = −.5σ 2 dt + σdWt

Integrating gives:

ln(rt ) − ln(.05) = −.5σ 2 t + σWt

Thus,

2 t+σW
rt = .05 · e−.5σ t

Study of Risk Management in Stock Broking Firm
100% (4)
Study of Risk Management in Stock Broking Firm
71 pages
AFAR Testbank
No ratings yet
AFAR Testbank
56 pages
Chapter 10 Conduct of Monetary Policy (Chapter 10) Mishkin
No ratings yet
Chapter 10 Conduct of Monetary Policy (Chapter 10) Mishkin
82 pages
Competitor Analysis - Arla Foods
100% (2)
Competitor Analysis - Arla Foods
94 pages
DTC PARTICIPANTS Listing Alphabetical List
No ratings yet
DTC PARTICIPANTS Listing Alphabetical List
14 pages
MAFS5030 hw1
No ratings yet
MAFS5030 hw1
4 pages
EX 1718 Sol
No ratings yet
EX 1718 Sol
10 pages
Examen203M2-March2024
No ratings yet
Examen203M2-March2024
22 pages
Derivative Analysis HW1
No ratings yet
Derivative Analysis HW1
3 pages
Problem Set 7: Option Pricing: Multi-Period Financial Market Model
No ratings yet
Problem Set 7: Option Pricing: Multi-Period Financial Market Model
24 pages
2023 1학기
No ratings yet
2023 1학기
17 pages
Sample Exam Paper With Answers PDF
100% (1)
Sample Exam Paper With Answers PDF
6 pages
Derivative and Risk Management
No ratings yet
Derivative and Risk Management
4 pages
interestrate2024-2
No ratings yet
interestrate2024-2
32 pages
WI4675 Exercise 5
No ratings yet
WI4675 Exercise 5
2 pages
General Framework For Pricing Derivative Securities
No ratings yet
General Framework For Pricing Derivative Securities
25 pages
Examen203M2-March2023
No ratings yet
Examen203M2-March2023
19 pages
4203 (Previous Year Questions)
No ratings yet
4203 (Previous Year Questions)
7 pages
Fin404 Derivatives: Master in Financial Engineering Spring 2019
No ratings yet
Fin404 Derivatives: Master in Financial Engineering Spring 2019
7 pages
Ancien Exo + Assign
No ratings yet
Ancien Exo + Assign
312 pages
Solution Problem of Financial Derivative
0% (1)
Solution Problem of Financial Derivative
16 pages
Mock Exam 1
No ratings yet
Mock Exam 1
11 pages
Call Put Options
100% (1)
Call Put Options
3 pages
PAK Mock Exam-QFIC-SOLUTION-TOTAL-76
No ratings yet
PAK Mock Exam-QFIC-SOLUTION-TOTAL-76
191 pages
SEMINAR 5 - Solutions
No ratings yet
SEMINAR 5 - Solutions
4 pages
QFIQ_Syllabus_2022
No ratings yet
QFIQ_Syllabus_2022
9 pages
Sibgha Ibrahim FD
No ratings yet
Sibgha Ibrahim FD
4 pages
Midterm Examination Questions Derivatives BUP
No ratings yet
Midterm Examination Questions Derivatives BUP
4 pages
Problem Set For
No ratings yet
Problem Set For
3 pages
PS1 Forward Contract Pricing 2011
0% (1)
PS1 Forward Contract Pricing 2011
3 pages
EX 1 - Forwards solutions
No ratings yet
EX 1 - Forwards solutions
4 pages
Practice Questions
No ratings yet
Practice Questions
8 pages
Problem Set 4
No ratings yet
Problem Set 4
2 pages
FM Prob Set4 Options A
No ratings yet
FM Prob Set4 Options A
6 pages
Fixed Income and Credit Risk: Prof. Michael Rockinger
No ratings yet
Fixed Income and Credit Risk: Prof. Michael Rockinger
12 pages
FM Practice Exam 1
No ratings yet
FM Practice Exam 1
5 pages
Final Exam Fall '12
No ratings yet
Final Exam Fall '12
7 pages
McDonald Chapter2
100% (1)
McDonald Chapter2
9 pages
FandI SubjCID 200304 Exampaper
No ratings yet
FandI SubjCID 200304 Exampaper
6 pages
MAFS5030 hw2
No ratings yet
MAFS5030 hw2
3 pages
Examen203M2-March2022
No ratings yet
Examen203M2-March2022
6 pages
Answers CH21
No ratings yet
Answers CH21
5 pages
Ams
No ratings yet
Ams
3 pages
FM423_2022_Exam___Solution.pdf
No ratings yet
FM423_2022_Exam___Solution.pdf
12 pages
FinDer 2023-24, Homework 1, v1, To Send
No ratings yet
FinDer 2023-24, Homework 1, v1, To Send
3 pages
2024F_RMSC2001_P4S
No ratings yet
2024F_RMSC2001_P4S
5 pages
Practice Midterm Sol
No ratings yet
Practice Midterm Sol
7 pages
Mathematical Finance Theory Review and (Pages 61-83) (2024-03-21)
No ratings yet
Mathematical Finance Theory Review and (Pages 61-83) (2024-03-21)
24 pages
WSU Derivatives 200079 FinalExam 2015sem2 Solutions
No ratings yet
WSU Derivatives 200079 FinalExam 2015sem2 Solutions
9 pages
1905 00711 PDF
No ratings yet
1905 00711 PDF
55 pages
st62005-2009 0
No ratings yet
st62005-2009 0
263 pages
Homework 2
No ratings yet
Homework 2
1 page
Sugga Estions
No ratings yet
Sugga Estions
4 pages
446A1
No ratings yet
446A1
4 pages
DMP204 Formulae
No ratings yet
DMP204 Formulae
1 page
Ps 1
No ratings yet
Ps 1
3 pages
old_exam
No ratings yet
old_exam
8 pages
Bman70381 Exercises 10
No ratings yet
Bman70381 Exercises 10
7 pages
Assignment 1&2
No ratings yet
Assignment 1&2
2 pages
Volatility Interpolation
No ratings yet
Volatility Interpolation
13 pages
8B_IRRM-Questions
No ratings yet
8B_IRRM-Questions
14 pages
Marc Hoffmann, Mauricio Labadie, Charles-Albert Lehalle, Gilles Pag' Es, Huy en Pham and Mathieu Rosenbaum
No ratings yet
Marc Hoffmann, Mauricio Labadie, Charles-Albert Lehalle, Gilles Pag' Es, Huy en Pham and Mathieu Rosenbaum
10 pages
Stochastic Analysis in Finance and Insurance
No ratings yet
Stochastic Analysis in Finance and Insurance
16 pages
Finance File
No ratings yet
Finance File
3 pages
Practice Problems in Statistics and Data Reduction
From Everand
Practice Problems in Statistics and Data Reduction
Rahul Basu
No ratings yet
Emco Industries LTD: Hailey College of Commerce University of The Punjab
100% (1)
Emco Industries LTD: Hailey College of Commerce University of The Punjab
66 pages
120 Years of Performance
No ratings yet
120 Years of Performance
35 pages
Tugas Kelompok Ke-2 (Minggu 5/ Sesi 7) Financial Modelling Laboratory
No ratings yet
Tugas Kelompok Ke-2 (Minggu 5/ Sesi 7) Financial Modelling Laboratory
13 pages
AWESOME Forex Trading Strategy (Never Lose Again) 2
100% (1)
AWESOME Forex Trading Strategy (Never Lose Again) 2
6 pages
Lead Managers Doc 1
No ratings yet
Lead Managers Doc 1
12 pages
8 PDF
No ratings yet
8 PDF
52 pages
Ppe Problems
No ratings yet
Ppe Problems
4 pages
Analysis of Automobile Industry
No ratings yet
Analysis of Automobile Industry
100 pages
Current Issues of Financial Accounting Introduce
No ratings yet
Current Issues of Financial Accounting Introduce
10 pages
Practice Qs 2
No ratings yet
Practice Qs 2
24 pages
MOF Question
No ratings yet
MOF Question
5 pages
BSA2A WrittenReports Thrift-Banks
No ratings yet
BSA2A WrittenReports Thrift-Banks
5 pages
(Almost) Everything You Wanted To Know About FX Volatility Smile (Part I) — Intro to the FX Market _ by Harel Jacobson _ Medium
No ratings yet
(Almost) Everything You Wanted To Know About FX Volatility Smile (Part I) — Intro to the FX Market _ by Harel Jacobson _ Medium
19 pages
04+AWS+Startup+Day Fundraising PDF
No ratings yet
04+AWS+Startup+Day Fundraising PDF
77 pages
Deregulation Banks Hedge Fund Derivatives Mortgages
No ratings yet
Deregulation Banks Hedge Fund Derivatives Mortgages
3 pages
Autozone Writeup
No ratings yet
Autozone Writeup
7 pages
Analysis of FS
No ratings yet
Analysis of FS
10 pages
Joint Arrangements
No ratings yet
Joint Arrangements
3 pages
Dịch 1
No ratings yet
Dịch 1
2 pages
Musings On Markets - The Zomato IPO - A Bet On Big Markets and Platforms! 2
No ratings yet
Musings On Markets - The Zomato IPO - A Bet On Big Markets and Platforms! 2
16 pages
EP04 Understanding Liquidity Pools in ICT
No ratings yet
EP04 Understanding Liquidity Pools in ICT
16 pages
Advising Slip
No ratings yet
Advising Slip
1 page
Premium Report
No ratings yet
Premium Report
58 pages
Cfa l1 Currency
No ratings yet
Cfa l1 Currency
35 pages
Cryptocurrency Investment or Scam
No ratings yet
Cryptocurrency Investment or Scam
16 pages