Institute of Actuaries of India

Subject SP6 – Financial Derivatives Principles

November 2023 Examination

INDICATIVE SOLUTION

Introduction

The indicative solution has been written by the Examiners with the aim of helping candidates. The solutions
given are only indicative. It is realized that there could be other points as valid answers and examiner have
given credit for any alternative approach or interpretation which they consider to be reasonable.
 IAI SP6-1123
Solution 1:

i) Differentiate the Put – Call parity equation and demonstrate the equivalence. [2]

ii) “C” carries more rights than “A” since in “C” we can exercise two options separately. In
terms of price, they will be the same since the two pieces will have the same optimal
exercise time. Hence “C” is preferred to “A”. [2]

iii) Vega = s^2 * σ * T * gamma. If there is no volatility skew, then vega neutral implies that
Gamma is zero. [2]

iv) Call increases in value while the put decreases in value. [2]

v) The price remains the same. [2]

vi) The asset is drift less so when the American option pays off there is precisely a 50% chance
(in the risk neutral measure) that the European option will pay off too. So, the European
is worth half as much as the American. [2]
[12 Marks]

Solution 2:

i) If the interest rates are zero, time dependence of volatility is irrelevant. Otherwise, the
time dependence matters since the stock will drift in risk neutral measure and whether
the volatility occurs before or after the drifting will make a difference. [3]

ii) Forward rates are always drift less – so we are back in the zero – interest case. [2]
[5 Marks]

Solution 3:

i) Default can be considered as a Bernoulli process as it either happens or doesn’t happen.

So the probability of default of A is P(A) means default can be considered as a Bernoulli
distribution with expected value P(A) and standard deviation (P(A)*(1-P(A))^0.5. Same
applies to B. Since:
𝐸[𝑋𝐴 𝑋𝐵 ] − 𝐸[𝑋𝐴 ]. 𝐸[𝑋𝐵 ]
𝜌𝐴𝐵 =
𝜎𝐴 𝜎𝐵
]
𝐸[𝑋𝐴 𝑋𝐵 (𝑤ℎ𝑖𝑐ℎ 𝑖𝑠 𝑡ℎ𝑒 𝑗𝑜𝑖𝑛𝑡 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑑𝑒𝑓𝑎𝑢𝑙𝑡 𝑃𝐴𝐵 ) = 𝜌𝐴𝐵 ∗
√𝑃𝐴 (1 − 𝑃𝐴 )𝑃𝐵 (1 − 𝑃𝐵 )𝑃𝐵 + 𝑃𝐴 𝑃𝐵
[6]

ii) Probability of Company X defaulting is 3.6%

Probability of S Bank defaulting is 0.5%
Joint probability of default of X and S bank is 0.25*(0.036*(1-0.036)*.005*(1-
0.005))0.5+0.005*0.036 = 0.346% which is the probability that K bank will incur a loss.
[2]

iii) The expected loss, given a $100 million exposure with a probability of 0.346% is $100
million x 0.346% = $346,000 [1]

iv) If the default probabilities were independent, the expected loss would have been
0.5%*3.6%*100 Million = $ 18,000. The default correlation has increased the expected
loss by ~19 times. [1]
[10 Marks]
Page 2 of 10
 IAI SP6-1123
Solution 4:

i) St follows Geometric Brownian Motion. This implies

𝑆𝑡 = 𝑆0 exp(𝜇𝑡 + 𝜎𝑊𝑡 ).

Let us define Lt = ln(Zt). Then

𝐿𝑡 = 𝑙𝑛(𝐵𝑡−1 𝑆𝑡 ) = 𝑙𝑛(𝐵𝑡−1 ) + 𝑙𝑛(𝑆𝑡 ) = 𝑙𝑛(𝑒 −𝑟𝑡 ) + 𝑙𝑛(𝑆0 𝑒 𝜇𝑡+𝜎𝑊𝑡 )
= −𝑟𝑡 + 𝜇𝑡 + 𝜎𝑊𝑡 .

Hence,
𝑑𝐿𝑡 = (𝜇 − 𝑟)𝑑𝑡 + 𝜎𝑑𝑊𝑡 .

Now as Zt = f (Lt) = exp(Lt), one can apply Ito’s formula to get:

1
𝑑𝑍𝑡 = [((𝜇 − 𝑟)𝑓 ′ (𝐿𝑡 ) + 𝜎 2 𝑓 ′′ (𝐿𝑡 )) 𝑑𝑡 + (𝜎𝑓 ′ (𝐿𝑡 ))𝑑𝑊𝑡 ]
2

Substituting f’(Lt) = f’’(Lt) = exp(Lt) = Zt, one gets

𝜎2
𝑑𝑍𝑡 = (𝜇 − 𝑟 + ) 𝑍𝑡 𝑑𝑡 + 𝜎𝑍𝑡 𝑑𝑊𝑡 .
2
[6]

ii) A portfolio is self-financing if and only if changes in its value depend only on changes in
the prices of the assets constituting the portfolio.

Mathematically: If Vt denotes the value of the portfolio (t, t), then the portfolio is self-
financing if and only if
𝑑𝑉𝑡 = 𝜙𝑡 𝑑𝑆𝑡 + 𝜓𝑡 𝑑𝐵𝑡

A replicating strategy for X is a strategy which involves investing in specifiable quantities (t,
t) of stock and risk free bonds, such that the portfolio of (t, t) of stocks and bonds will be
self-financing the portfolio (t, t) and will have terminal value equal to the magnitude of the
claim; i.e. VT = TST + TBT = X.

This means that the portfolio’s cash flows at the claim exercise date match the cash flows
under the claim.

When the underlying stock follows a continuous geometric Brownian motion process, there
is an additional technical constraint for the strategy to work; namely:
𝑇
∫ 𝜙𝑡2 𝜎 2 𝑑𝑡 < ∞.
0
[6]

iii) We have from (i)

𝜎2
𝑍𝑡 = (𝜇 − 𝑟 + ) 𝑍𝑡 𝑑𝑡 + 𝜎𝑍𝑡 𝑑𝑊𝑡 .
2
The drift of this process is
𝜎2
(𝜇 − 𝑟 + ) ≥ 0.
2
Since, the process is not driftless, it will not be a Martingale.
We can thus use the Cameron – Martin – Girsanov (CMG) theorem to convert the process into
a Martingale. The CMG theorem states that there exists a probability measure Q, equivalent
to the measure P (defined by the probability distribution of Wt ), such that Z is a Martingale.
Page 3 of 10
IAI SP6-1123
Thus, to apply the theorem we set
𝜎2
𝛾𝑡 = (𝜇 − 𝑟 + )
2
and then verify that t is a pre-visible process and that
1 𝑇
𝐸𝑃 [𝑒𝑥𝑝 ( ∫ 𝛾𝑡2 𝑑𝑡)] < ∞.
2 0

Using the CMG theorem, the measure Q, that is equivalent to P is such that
𝑑𝑄 1 𝑇 𝑇
= 𝑒𝑥𝑝 (− ∫ 𝛾𝑡2 𝑑𝑡 − ∫ 𝛾𝑡 𝑑𝑊𝑡 )
𝑑𝑃 2 0 0

and the Brownian motion

𝑡
̃𝑡 = 𝑊𝑡 + ∫ 𝛾𝑠 𝑑𝑠
𝑊
0
is a Q-measure Brownian motion.

Alternately, this can also be written as

𝑑𝑊̃𝑡 = 𝛾𝑑𝑡 + 𝑑𝑊𝑡 .

If one substitutes dWt into the SDE for Z, one would get,
𝑑𝑍𝑡 = 𝜎𝑍𝑡 𝑑𝑊̃𝑡.

This is the SDE for a driftless process under measure Q. Hence, Z is a Martingale under measure
Q.

The next step in the construction of the replication strategy is to form the discounted expected
claim process
𝐸𝑡 = 𝐸𝑄 (𝐵𝑡−1 𝑋|𝑭𝑡 )
and show that this is a Q-measure Martingale as well. This can be done in a manner similar to
the way Zt is shown to be a Q-Martingale.

As both Zt and Et are Q-Martingales, the Martingale Representation Theorem (MRT) gives us a
pre-visible process t such that dEt = t dZt .

It may be noted that in order to to apply the MRT we need to show that both Zt and Et are Q-
Martingales and that the volatility of Zt is non zero with probability 1.

The replication strategy then consists of holding t units of stock and t risk free bonds, where
t is given by
𝜓𝑡 = 𝐸𝑡 − 𝜙𝑡 𝑍𝑡
[6]

iv) The portfolio replicates the claim because the portfolio is self – financing and, at time T,
when the claim falls due, the portfolios proceeds are

𝜙 𝑇 𝑆𝑇 + 𝜓 𝑇 𝐵𝑇 = 𝜙 𝑇 𝑆𝑇 + (𝐸𝑇 − 𝜙 𝑇 𝑍𝑇 )𝐵𝑇 = 𝜙 𝑇 𝑆𝑇 + (𝐸𝑇 − 𝜙 𝑇 𝐵𝑇−1 𝑆𝑇 )𝐵𝑇 = 𝐸𝑇 𝐵𝑇

= 𝐸[𝐵𝑇−1 𝑋|𝐹𝑇 ]𝐵𝑇 = 𝑋

i.e. the portfolio s proceeds will match the claim amount.

The value of the portfolio at any time t can be written

Page 4 of 10
 IAI SP6-1123
𝑉𝑡 = 𝜙𝑡 𝑆𝑇𝑡 + 𝜓𝑡 𝐵𝑡 = 𝐵𝑡 𝐸𝑡.
The SDE for Vt is given by
𝑑𝑉𝑡 = 𝑑(𝐵𝑡 𝐸𝑡 ) = 𝐵𝑡 𝑑𝐸𝑡 + 𝐸𝑡 𝑑𝐵𝑡 .

To show that the portfolio is self-financing, one needs to show that

𝑑𝑉𝑡 = 𝜙𝑡 𝑆𝑡 + 𝜓𝑡 𝐵𝑡

As the portfolio replicates the claim, the arbitrage-free condition requires that the value of the
claim equals the value of the replicating strategy. Therefore, either of the above two SDEs gives
the stochastic differential equation for the value of the claim.
[4]
[22 Marks]

Solution 5:
i) For each property, the notional amount of the property derivative contract should be set
such that the change in the index value corresponds to the change in the property value.

Given the formula: Payoff = (Initial Index - Final Index) x Notional Amount

When the index drops by 1 point, the loss in property value should be equal to the payoff
from the derivative. Therefore, the notional amount is calculated as: Notional Amount =
Property Value / Initial Index

• City centre building: Notional Amount = ₹10 million / 1500 = ₹6,667 per point
• Shopping mall: Notional Amount = ₹15 million / 1200 = ₹12,500 per point
• Office complex: Notional Amount = ₹20 million / 2000 = ₹10,000 per point
[4]

ii) Using the formula for payoff and applying the correction factor: Adjusted Payoff = (Initial
Index - Final Index) x Notional Amount x Correction Factor

• City center building: Adjusted Payoff = (1500 - 1450) x ₹6,667 x 1.05 = ₹350,000
• Shopping mall: Adjusted Payoff = (1200 - 1250) x ₹12,500 x 0.95 = ₹-593,750
• Office complex: Adjusted Payoff = (2000 - 1950) x ₹10,000 x 1.10 = ₹550,000
[4]

iii)
• City centre building: Change in Value = ₹10 million x 4% = -₹400,000
• Shopping mall: Change in Value = ₹15 million x 3% = ₹450,000
• Office complex: Change in Value = ₹20 million x 5% = -₹1 million

Interest on the loan: Interest = ₹5 million x 5% = ₹250,000

Net gain or loss: Net Gain/Loss = Total Adjusted Payoff + Change in Property Values -
Interest Net Gain/Loss = (₹350,000 - ₹593,750 + ₹550,000) + (-₹400,000 + ₹450,000 -
₹1 million) - ₹250,000 Net Gain/Loss = ₹-943,750

With the correction factors applied, CityScape has a net marked to market loss of
₹943,750 after considering all factors.
[5]
[13 Marks]

Solution 6:

i) Delta Δ for a European call option:

𝛥 = 𝑒 𝑞𝑡 𝑁(𝑑1 )
Page 5 of 10
 IAI SP6-1123
𝑠 𝜎2
ln ( ) + (𝑘 − 𝑞 + 2 ) × 𝑡
𝑘
𝑑1 =
𝜎√𝑡
And Gamma

𝑒 −𝑞𝑡 𝑛(𝑑1 )
𝛤=
𝑆𝜎𝑗√𝑡
−𝑑21
n(d1) = 𝑒 2 /√2𝜋
S 10000
K 10500
t 0.5
r 5%
sigma 22%
dividend 2.50%
d1 -0.155500265
D2 -0.311063757
N(D1) 0.44
N(d2) 0.38
Delta 0.432769897

n(d1) 0.394148042
Gamma 0.000250221

Given the underlying stock price change is 3% of ₹ 10,000, which is:

Now, using the Taylor expansion to approximate the change in option price:

Change of option price = Δ×ΔS + ½ x Γ×(ΔS)2

Percent change 3%
S 10000
Change in price 300
Overall change 141
[3]

ii) Change in price due volatility can be estimated using vega.

𝑉 = 𝑆𝑒 −𝑞𝑡 𝑛(𝑑1 )√𝑇

Volatility change 1.50%

Vega 27.52 per % change in implied volatility
Impact 27.52*1.5=41
[3]
[6 Marks]
Solution 7:

i) The LIBOR market model, also known as the Brace-Gatarek-Musiela (BGM) model, is
a financial model used to model the evolution of forward interest rates. It's significant
in pricing interest rate derivatives as it captures the dynamics of the entire forward
rate curve, allowing for a more accurate and flexible pricing mechanism. [2]

ii) Multi-Currency LIBOR Market Model (LMM):

The multi-currency LIBOR market model is an extension of the single-currency LIBOR

market model. It is used to model the evolution of a set of LIBOR rates for multiple
Page 6 of 10
IAI SP6-1123
currencies simultaneously. The significance of the multi-currency LMM lies in its ability
to price and hedge interest rate derivatives that have exposure to multiple currencies.
It is able to allow for the correlation between the entities. [2]

iii) Black’s formula is given by

C=D(0,T)×L×[F(0,T) x N(d1)−K×N(d2)]

C is the caplet price.

D(0,T) is the discount factor from time 0 to time T.

L is the notional amount.

N(d) is the cumulative distribution function of the standard normal distribution.

K is the strike rate.

d1 and d2 are given by:

𝐹(0, 𝑇) 𝜎2
ln ( 𝐾 ) + ( 2)×𝑇
𝑑1 =
𝜎 √𝑡

𝑑2 = 𝑑1 − 𝜎√𝑡
For USD:

Given:

• F(0,T) (forward LIBOR rate) = 3% or 0.03

• L((notional amount) = 1 million
• K (strike rate) = 4% or 0.04
• σ (volatility) = 20% or 0.20
• T (time to maturity) = 1 year

F 3%
K 4%
sigma 20%
time 1
Discount 0.97
d1 -1.34
d2 -1.54
N(d1) 0.09
N(d2) 0.06
Notional 1000000
Value 225.60
For EUR

• F(0,T) (forward LIBOR rate) = 2% or 0.02

• r (discount rate) = 2% or 0.02
• L(0,T) (notional amount) = 1 million
• K (strike rate) = 3.5% or 0.035
• σ (volatility) = 18% or 0.18
• T (time to maturity) = 1 year

F 2%
K 3.50%

Page 7 of 10
 IAI SP6-1123
sigma 18%
time 1
Discount 0.980198673
d1 -3.01898
d2 -3.1989766
N(d1) 0.001268151
N(d2) 0.000689582
Notional 1000000
Value 1.20
[5]

iv) This would result increase in price of caplet. [1]

v) The multi-currency LMM can be extended to price cross-currency swaptions by

incorporating the correlation between the interest rates of the two currencies and the
exchange rate dynamics. Key considerations include:
• Correlation: The correlation between the interest rates of the two currencies
plays a crucial role in determining the swaption's value.
• Quanto Adjustments: These are adjustments made to account for the risk of
changes in the exchange rate between the two currencies.
• Model Calibration: The model needs to be calibrated to market data to ensure
accurate pricing.
• Complexity: Pricing cross-currency swaptions is more complex than single-
currency swaptions due to the additional factors involved.
[3]
[13 Marks]

Solution 8:

i) Annual Premium Payment:

Company BRONCO: Premium = CDS spread × Notional principal = 0.02 × ₹100 million
= ₹2 million

Company LFA LEX: Premium = 0.025 × ₹50 million = ₹1.25 million [3]

ii) Payout in Case of Credit Event:

Company BRONCO:

Payout = (1 - Recovery rate) × Notional principal

= (1 - 0.40) × ₹100 million = ₹60 million

Company LFA LEX:

Payout = (1 - 0.40) × ₹50 million = ₹30 million [3]

iii) As company has already entered the contract widening of CDS spread would not
have any impact on the payment and CDS payment would base on the initial CDS
spread. We would have updated recovery which is given below.

Page 8 of 10
 IAI SP6-1123

Spread Notional Payment (Million)

CDS Base spread LFA LEX 2.00% 100 2
CDS Base spread BRONCO 2.50% 50 1.25

Bond recovery rate

CDS Y1 spread LFA LEX 2.50%
CDS Y1spread BRONCO 1.50%
Recovery Rate Y1 LFA LEX 40%
Recovery Rate Y1 BRONCO 40%

Recovery rate after change Default CDS payment (INR

Default payout in CDS loss Million)
Bond 1 32% 68% 68
Bond 2 67% 33% 16.66667

Year 1 Year 2 Recovery

Total payment 3.25 3.25 84.7
Total transaction INR 78 m
[5]
[11 Marks]

Solution 9:

i)
• Platykurtic distribution is a distribution with high kurtosis.
• Using a platykurtic distribution ensures that sufficient probability is allocated to the
tails of the distribution in order to prevent underestimation of the effects of large price
movements.
• Thus the approach can be quite useful when pricing out of the money options that
become valuable when extreme price events take place in the underlying.
[3]

ii) Advantages
• By choosing an appropriate α value for the gamma distribution we can use a model
that has a kurtosis higher than the corresponding lognormal model.
• A major drawback of Lognormal distribution is that their kurtosis is too low compared
with observed movements. This drawback can be overcome by using the approach
suggested by Prof Gull.
• The Gamma distribution spans all positive values which matches the range required
for an asset price.

Disadvantages

• Gamma distribution is a less familiar model for practitioners to whom its properties
may be unfamiliar. It may therefore be harder to establish the correct pricing formulae
or to incorporate the model in software.
• Using the Gamma distribution can lead to more conservative price estimates which
may not be in line with the generally accepted market practices, making it difficult to
participate in the market.
• The Gamma model breaks the link with Geometric Brownian Motion, the underlying
model from which lognormal distribution of asset prices is derived.

Page 9 of 10
IAI SP6-1123
• The volatility parameter σ is a part of the SDE defining geometric Brownian motion,
i.e. dSt=St(µdt+σdBt). There is no obvious corresponding SDE that generates the
Gamma Model. So it will not be clear how to measure or interpret the volatility of the
asset price using the gamma model.
[5]
[8 Marks]

*****************************

Page 10 of 10