Financial Mathematics

J2FIN: Problem Sheet 9 Solution

Dr Jia Shao
Financial Mathematics 2024

1. (a) Similarities:
ˆ Both models produce outputs given a specified set of inputs, e.g.
data and assumptions;
ˆ Both models will make assumptions about the future behaviour
of financial variables, e.g. asset returns, inflation.
Differences:
ˆ A deterministic model is based on one set of parameters, and it
can in practice be very hard to pick the ‘correct’ set;
ˆ In stochastic models, no single value is used, and variations are
allowed for by the application of probability theory. So a stochas-
tic model produces a different output every time it is run, which
will allow us to see the distribution of the results.
(b) The company might prefer a stochastic model because:
For a deterministic model, deciding which set of input variables to
use may be a challenge. And it will be hard for the company to
show it has picked the ‘right’ investment assumptions. A stochastic
model avoids this issue by assuming a range of possible outcomes
could apply. This is particularly useful for the insurer because they
hold long-term contracts which are likely to be uncertain in the na-
ture/term/currency, and the variation built into a stochastic model
is much more likely to capture this properly. Running the stochastic
model lots of times will produce a range of results with associated
probabilities.
(c) There are many reasons that could reasonably be argued. Some are:
ˆ The company is assuming that investment returns are indepen-
dent year-on-year. In practice this is not appropriate. As the
returns in the past year sometimes inform the returns in the
current year;
ˆ The company is assuming that returns are Normally distributed
each year;
ˆ The Normal distribution assigns a non-zero probability to returns
smaller than −100%;
ˆ And in general it is not appropriate to assume a greater loss
value than the initial investment outlay is possible.
(d) ˆ The company could correct for lack of dependence by assuming
the returns of each year are a function of the previous year’s
returns;

1 PS9S,J2FIN2024
 Financial Mathematics

ˆ Or it could consider changing its approach to a different type

of model. The company could change to a distribution that
does not permit returns below −100%, e.g. Like the LogNormal
distribution.
 2

2. (a) E [1 + it ] = exp µ + σ2 = 1.05




var [1 + it ] = exp 2µ + σ 2 × exp σ 2 − 1 = 0.032 = 0.0009
2

As (E [1 + it ]) = exp 2µ + σ 2






exp 2µ + σ 2 × exp σ 2 − 1 = 1.052 × exp σ 2 − 1
=> σ 2 = ln 0.0009


1.052 + 1 = 0.00081599
i.e. σ = 0.0285655
=> exp µ + 0.00081599


2 = 1.05
µ = ln(1.05) − 0.00081599
2 = 0.0483822
(b) P (0.01 < it < 0.03) = P (1.01 < 1 + it < 1.03)

P (ln(1.01) < ln (1 + it ) < ln(1.03))

 
ln(1.01) − 0.0483822 (ln (1 + it ) − 0.0483822) ln(1.03) − 0.0483822
=P √ < √ < √
0.00081599 0.00081599 0.00081599
=P(−1.34539 < Z < −0.658955) where Z ∼ N (0, 1)
=0.254962 − 0.089250 = 16.6%

(c) The probability calculated in part (b) is small. This is reasonable

as the expected return in any year is 5%, and we are being asked to
calculate the probability that the return is within a range which does
not include the expected value.
(d) If Sn represents the accumulated value of £1 invested in the fund at
the end of year n, Sn is log-normally distributed with parameters nµ
and nσ 2

P (10000S3 < 11000) = P (S3 < 1.1) = P (ln (S3 ) < ln(1.1))
 
ln(1.1) − 3 × 0.0483822
=P Z < √ where Z ∼ N (0, 1)
3 × 0.00081599
=P(Z < −1.00732) = 0.15689

 
σ2
3. (a) E [(1 + it )] = exp µ + 2 = 1.035
2



var [(1 + it )] = exp 2µ + σ exp σ 2 − 1 = 0.022
Therefore:
0.022


exp σ 2 − 1 = 1.035
 2
2 0.022
σ = ln 1.0352 + 1 = 0.00037333
σ = 0.0193
exp µ + 0.00037333


2 = 1.035
0.00037333
µ = ln(1.035) − 2 = 0.0342

2 PS9S,J2FIN2024
 Financial Mathematics

(b) Option A:



(1 + it ) ∼ Lognormal µ, σ 2 so ln (1 + it ) ∼ N µ, σ 2 .

t=15
Y t=15
X
ln (1 + it ) = ln (1 + it )
t=1 t=1

As {it } are independent:

t=15
X
ln (1 + it ) ∼ N 15µ, 15σ 2


t=1

t=15
Y
(1 + it ) ∼ Lognormal 15µ, 15σ 2


t=1

15µ = 0.513, and 15σ 2 = 0.0056.

Let S15 be the accumulated amount after 15 years of a single invest-
ment of one:  
2
E [S15 ] = exp 15µ + 15σ = exp 0.513 + 0.00560


2 2 = 1.6750
The accumulated sum is 500 × E [S15 ] = £837.49 (or 837.67 with full



precision). Var [S15 ] = exp 30µ + 15σ 2 exp 15σ 2 − 1 = exp(2 ×
0.513 + 0.0056)[exp(0.0056) − 1] = 0.0157
p
Therefore, the standard deviation of £500 is 500 × Var [S15 ] =
£62.65 (or £62.77 with full precision)
Option B:
The accumulated sum after 10 years is:
500 × (1.04)10 = £740.12
The expected value of the accumulated sum at the end of 15 years is:
 
740.12× 0.15 × 1.0055 + 0.25 × 1.015 + 0.4 × 1.0455 + 0.2 × 1.075 =
£884.83
To find the variance of the accumulation we first find the expected
value of the square of the accumulation as follows:
 
740.122 × 0.15 × 1.00510 + 0.25 × 1.0110 + 0.4 × 1.04510 + 0.2 × 1.0710 =
793, 430.11 [Or 793,425.50 using rounded figures.]
The variance of the accumulation is: 793, 430.11−884.832 = 10500.179.
√
The standard deviation is 10500.179 = £102.47

3 PS9S,J2FIN2024
 Financial Mathematics

(c) For Option A we require

 
775
P S15 <
500
=P (S15 < 1.55)
=P (ln (S15 ) < ln(1.55)) where ln (S15 ) ∼ N 15µ, 15σ 2 = N (0.513, 0.00560)

ln(1.55) − 0.513
=P(N (0, 1)) < √
0.00560
=P(N (0, 1)) < −0.9988) = 15.9% (or 15.8% with full precision)

For Option B we examine the lowest possible pay out.

There is a 15% probability that the amount will be 500 × 1.0410 ×
1.0055 = £758.81 <£775. This is the minimum value that the poli-
cyholder will receive.
There is a 25% probability that the amount will be 500 × 1.0410 ×
1.015 = £777.87 > £775. Therefore the probability of a payment
less than £775 is 15%.
(d) Option A is riskier in term of a lower minimum possible pay out, but
Option B is riskier in terms of having a larger standard deviation.
The risk of a shortfall relative to £775 is greater for Option A.
1
4. (a) E [it ] = j = 2 (i1 + i2 )

2
Var [it ] = s2 = E i2t − E [it ]
 
 2
1 2 2

1
= i1 + i2 − (i1 + i2 )
2 2
1 2
1 2
i1 + i22 − i1 + i22 + 2i1 i2


=
2 4
 2
1 2 1 1
i1 + i22 − i1 i2 =


= (i1 − i2 )
4 2 2

(b) E [S20 ] = (1 + j)20 = 4.5/2 = 2.25

√
Therefore j = 20 2.25 − 1 = 0.0413797

20
Var [S20 ] = 1 + 2j + j 2 + s2 − (1 + j)40 = (0.75/2)2 = 0.3752
√
s2 = 20
0.3752 + 1.041379740 − 1 − 2 × 0.0413797 − 0.04137972 =

2
0.0014867 = 12 (i1 − i2 )
i1 > i2 therefore take the positive root:
√
i1 − i2 = 2 × 0.0014867 = 0.077115
i1 + i2 = 2 × j = 2 × 0.0413797 = 0.082759
So, i1 = 0.079937, and i2 = i1 − 0.077115 = 0.002822.
10 50 20
5. (a) 80 5% + 80 10% + 80 x = RM
x − 3% = 2(RM − 3%) => x = 2RM − 3%

4 PS9S,J2FIN2024
 Financial Mathematics

Substitute into first equation and get => RM = (10 × 5% + 50 ×

10% − 20 × 3%)/40 = 12.25%
OR
(10% − 3%)/ (RM − 3%) = 1
=> RM = 10%
(b) x − 3% = 2(12.25% − 3%)
=> x = 21.5%
OR
x − 3% = 2(10% − 3%)
=> x = 17%
(c) 5% − 3% = β(12.25% − 3%)
=> β = 0.216
OR
5% − 3% = β(10% − 3%)
=> β = 0.286
(d) There are basic problems in testing the model since, in theory, ac-
count has to be taken of the entire investment universe open to in-
vestors, not just capital markets. An important asset of most in-
vestors, for example, is their human capital (i.e. the value of their
future earnings).
Models have been developed which allow for decisions over multiple
periods and for the optimization of consumption over time to take
account of this. Other versions of the basic CAPM have been pro-
duced which allow for taxes, inflation, and also for a situation where
there is no riskless asset.
In the international situation there is no asset that is riskless for all
investors (due to currency risks) so a model has been developed which
allows for groups of investors in different countries, each of which con-
siders their domestic currency to be risk-free There is a discrepancy
between the values obtained for the expected return on the market
using the security market line and the weightings by market capital-
ization Some other assumptions of CAPM are unrealistic, e.g. ev-
eryone has the same estimates for the means/variances/covariances,
everyone has the same single time horizon.
Cov(Ri ,RM )
6. (a) i. βi = Var(RM ) .
ii. ri − r0 = βi (rM − r0 )
where r0 is the return on the risk-free asset.
iii. Since RM = Σπi Ri , it follows that
Var (RM ) = Σπi Cov (Ri , RM )
and so Σπi βi = Σπi Cov (Ri , RM ) / Var (RM )
= Var (RM ) / Var (RM ) = 1

5 PS9S,J2FIN2024
 Financial Mathematics

(b) i. The proportions are given by proportions of market capitalisa-

tion so that
π1 = 2/5, π2 = π3 = π4 = 1/5.
ii. Cov (R1 , RM ) = 2/5 × 4 + 1/5 × 1 + 1/5 × 1 + 1/5 × 1 = 11/5,
so β1 = (11/5)/(8/5) = 11/8.
Similarly, Cov (R2 , RM ) = 2/5 × 1 + 1/5 × 3 + 1/5 × 1 + 1/5 × 1 =
7/5, so β2 = 7/8
and Cov (R3 , RM ) = 2/5 × 1 + 1/5 × 1 + 1/5 × 2 + 1/5 × 1 = 6/5,
so β3 = 6/8.
Now it follows from (c)(iii) (i.e. Σπi βi = 1 ) that β4 = 5/8.
iii. We conclude that, since ri − r0 = βi (rM − r0 ),
11% = 11/8 × (rM − r0 ) so that rM = 11%
Then:
Asset number 1 2 3 4 Market
Expected return 14% 10% 9% 8% 11%
2
7. (a) i. E [A3 ] = A0 eµt+0.5σ t
= 100e0.05×3+0.5×0.2×3 = £123.37
µt+0.5σ 2 t 2
ii. E [B3 ] = B0 e = 100e0.08×3+0.5×0.3 ×3 = £145.50
q

q


(b) i. SD [A3 ] = A20 e2µt+σ2 t eσ2 t − 1 = 1002 e2∗0.05∗3+0.22 ∗3 e0.22 ∗3 − 1 =
£44.05 q


ii. SD [B3 ] = B02 e2µt+σ2 t eσ2 t − 1 = £81.01
(c) E [P3 ] = 0.5E [A3 ] + 0.5E [B3 ] = £134.44
(d) V [P3 ] = 0.52 V [A3 ] + 0.52 V [B3 ] + 2Corr × 0.5 × 0.5 SD [A3 ] SD [B3 ] =
0.25 × 44.052 + 0.25 × 81.012 + 2 × 0.3 × 0.5 × 0.5 × 44.05 × 81.01 =
2, 661.03
=> SD [P3 ] = £51.59
(e) The expected return of the portfolio falls halfway between the ex-
pected return on each of the one-stock investment strategies.
But the standard deviation is well below halfway between the two
one-stock strategies.
The price of risk for stock A is 23.37/44.05 = 0.53
The price of risk for stock B is 45.5/81.01 = 0.56
But the price of risk for the portfolio is 34.44/51.59 = 0.67
So the portfolio delivers a better expected return per unit of risk
This is because the assets are not fully correlated, which shows the
benefit of diversification.

6 PS9S,J2FIN2024