0% found this document useful (0 votes)
58 views26 pagesRT Exam
Uploaded by
Parita PanchalCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
58 views26 pagesRT Exam
Uploaded by
Parita PanchalCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 26
20.8
‘An insurance company offers accident insurance for employees. A total of 650 policies have been
issued split between two categories of employees. The first category contains 400 policies, and
claims occur on each policy according to a Poisson process at a rate of one claim per 20 years, on
average. In this category all claim amounts are £3,000. In the second category, claims occur on
each policy according to a Poisson process at arate of one claim per 10 years, on average. In this
category, the claim amount
\er £2,000 or £3,000 with probabilities 0.4 and 0.6, respectively.
All policies are assumed to be independent. Let 5 denote the aggregate annual claims from the
portfolio.
“
(i)
(ii)
Calculate the mean, variance and coefficient of skewness of S. [4]
Using the normal distribution as an approximation to the distribution of S, calculate ¥
such that the probability of $ exceeding Y is 10%. 1
The insurance company decides to effect reinsurance cover with aggregate retention
£100,000, so that the insurance company then pays out no more than this amount in
claims each year. In the year following the inception of this reinsurance, the numbers of
policies in each of the two groups remains the same but, because of changes in the
‘employment conditions of which the company was unaware, the probability of a ciaim in
group 2 falls to zero. Using the normal distribution as an approximation to the
distribution of 5, calculate the probability of a claim being made on the reinsurance
treaty, BI
[Total 10] 20.8
(i) Mean, variance and coefficient of skewness
We have N, ~Paisson(400x a) Poisson\20) , Xy = £3,000 , Nz ~ Poisson(250x a Poisson(25)
and:
£2,000 with probability 0.4
= Se with probability 0.6
Working in £000s, we find that:
E(X2) = (20.4) +(3*0.6)=2.6
e(x3)=(2 xo4)+(# x06)=7
e(x})=(2?<0.4)+(3?x06)=19.4 rea]
Let 5, denote the annual aggregate claims from category i. Using the assumption that the
policies are independent and the result that, for a compound Poisson random variable T, the «th
central moment of T is given by 2E(X*) , we obtain:
E(S) =£(S;) + E(S9) = (203) + (252.6) = 125 = £125,000 fu
var(5) = varl5;)+ war(Ss)
= EUG) + 2g (3)
=(20x9)+(25*7)
=355
= £7355, 000,000 [1]
skew(5)=skew(S;)+skew(S.)
= ALE(?) + EUG)
=(20%x27)+(25x19.4)
=1,025
3 (4,025«10°) (ii) List five other risk criteria that would be considered desirable by a general insurer. (i)
Other desirable features of a risk
Other desirable features are:
Individual risks should be independent of one another.
The probability that the insured event will occur should be small.
Large numbers of similar risks should be pooled in order to reduce the variance and achieve
greater certainty.
The insurer's liability should be limited.
Moral hazards should be eliminated as far as possible since these are difficult to quantify,
result in selection against the insurer and lead to unfairness in the treatment of some
policyholders. 19.8 The aggregate claim amount from a portfolio has a compound negative binomial distribution.
SABE ()
(ii)
(iii)
(iv)
Show that if 5 =X, +--+ Xy, then:
Ms(t)=My [log My (t)] BI
If N has Type 2 negative binomial distribution with k=2 and p=0.9, and X hasa
gamma distribution with « =10 and 2
0.1, determine an expression for Mg(t). [2]
(a) Calculate the mean and variance of S.
(b) Using suitable approximation, estimate the aggregate amount which will be
exceeded with probability 0.1%. [4]
The insurer in fact has 100 identical independent portfolios of this type. Let:
T= Sy 49+ S109
(a) Determine the moment generating function for T .
(b) Using a normal approximation, estimate the total aggregate claim amount from
the whole business which will be exceeded with probability 0.1%.
(c)_ Comment on your answers to parts
ib) and (iv)(b). fa]
[Total 13] 19.8
(i) MGF of s
The MGF of S is:
ait)=Ete'S)=£[ Ete |) ]
using the standard result for conditional means from page 16 of the Tables.
Looking at the inner expression, we have:
E(e'S |N=n)= e(ete m 2 =F (eM)... £(et%)
Now each of these terms is just the MGF of the random variable x .
(q) So:
Ele
and hence:
E(e® |N)=[My (th)
So:
Ms(t) =e[ {my wh" |= f(eNPamete) My [log my (t)]
(i) Compound negative binomial distribution
We first need the individual MGFs. Using results from the Tables, we have:
ry” 10
matt)=(- 2) =(1—108)
Combining these, using the result from part (i):
2
M(t) = My [logMy “(as
[1]
(1)
(¥]
[4]
(1) (iii(a) Mean and variance of S
We could differentiate this expression to find the mean and variance of S. However, it is much
easier to use the standard compound distribution formulae:
E(S) =E(X)E(N)
and: —_var(s)=[€(X)]° var(N) + var(X}E(N)
Using the results from the Tables for the individual distributions:
e(x) = 2 = 29 ~100
z 04
a
varix)= 5 =
aL yyy)
Pp
og
oft. 2A
y 0g
= 0.24691
(1)
e Using the formulae above:
E(S) = E(X)E(N) = 22.222
and: — var(s) =[E(X)]° var(N) + var(x}E(W)=1007 0.24691 + 1,000%0.22222=2,691.358 [1]
(iii)(b) Aggregate amount
We now assume that S has an approximate normal distribution with this mean and variance. So,
standardising in the usual way, we have:
k-E(S)
P(S>k)=0.001 => r{moa)> 4) 0.001
\var(S)
\
Using the percentage points of the standard normal distribution, we find that:
0902 => k =22.222+3.0902/2, 691.358 = 182.54 [1] (iv)(a) MGF of T
The MGF of T is:
My {t) = Ele )= Efe + $100) = Ele), (e100) =[ M(t)
6 09 200
=| —————___ (4)
a }
Tio
The mean and variance of T are 100 times the corresponding results for S, ie:
E(T) =100E(S) =2,222
and: —_ var(T) =100var(s)=269, 135.8 (1]
(iv)(b) Total aggregate amount
So the corresponding figure for the aggregate amount exceeded with probability 0.001 is:
2,222 + 3.0902y269, 135.8 = 3,825.37 (1)
(iv) Comment
This is substantially less than one hundred times the corresponding answer to part (iii)(b). The
Central Limit Theorem tells us that as the number of portfolios increases, bad experience in some
of the portfolios will be offset by better experience in others, leading to a situation where the
overall variation is relatively smaller. Pooling of similar risks reduces the overall variance. We can
see this happening here. [1] 19.1 (i) State the two conditions that must hold for a risk to be insurable. 19.1 (i) Criteria for an insurable risk
The two conditions are:
. The policyholder must have an interest in the risk being insured.
. The risk must be of a financial and reasonably quantifiable nature. small insurer wishes to model its claim costs for motor insurance using a simple generalised
linear model based on the three factors:
"0, {
j=1 — for'fast' cars
FS;
for ‘young’ ol
for ‘old’ drivers
j=0 — for'slow' cars
k=1 — for'town' areas
TC, =
k=0 for ‘country’ areas
The insurer is considering three possible models for the linear predictor:
Model 1: YO+FS +TC
Model 2: YO+FS +YOFS +TC
Model 3: YO" FS*TC
(i) Write each of these models in parameterised form, stating how many non-zero parameter
values are present in each model [6]
(il) Explain why Model 1 might not be appropriate and why the insurer may wish to avoid
using Model 3. (2)
(ii) The student fitting the models has said ‘We are assuming a normal error structure and we
are using the canonical link function.’ Explain what this means. 8 (iv) The table below shows the student’s calculated values of the scaled deviance for these
three models and the constant model.
Scaled
Model " Degrees of freedom
Deviance
1 50 7
YO+FS+TC 10
YO+FS+YO.FS+TC 5
YO*FS*TC 0
Complete the table by filling in the missing entries in the degrees of freedom column and
carry out the calculations necessary to determine which model would be the most
appropriate.
[5]
[Total 16] 12.9
(i) Parameterised form
In parameterised form, the linear predictors are (with i, j and k corresponding to the levels of
YO, FS and TC respectively):
Model 1: @;+ Aj +7 (4 parameters)
There is one parameter to set the base level for the combination ¥Oo,FS9,TCo and one additional
parameter for each of the higher levels of the three factors.
Model 2: ay + (5 parameters)
There are four parameters for the 2x2 combinations of YO and FS (assuming TCp) and one
additional parameter for the higher level of TC.
Model 3: ip (8 parameters)
There are eight parameters for the 2x22 combinations of ¥O, FS and TC.
[2 for each model]
(ii) Problems with Model 1 and Model 3
Model 1 does not allow for the possibility that there may be interactions (correlations) between
some of the factors. For example, it may be the case that young drivers tend to drive fast cars and
to live in towns. w
With Model 3, which is a saturated model, it would be possible to fit the average values for each
group exactly ie there are no degrees of freedom left. This defeats the purpose of applying a
statistical model, as it would not ‘smooth’ out any anomalous results. (
The problem referred ta with Model 3 corresponds to the idea of undergraduation in Subject C52.
(ii) Explaining ‘normal error structure’ and ‘canonical link function’
Normal error structure means that the randomness present in the observed values in each
category (eg young/fast/town) is assumed to follow a normal distribution. iba]
‘The link function is the function applied to the linear estimator to obtain the predicted values.
Associated with each type of error structure isa ‘canonical’ or ‘natural’ link function, In the case
of a normal error structure, the canonical fink function is the identity function 2 (iv) Compare models
The completed table, together with the differences in the scaled deviance and degrees of
freedom, is shown below.
Scaled A Scaled
Model DF . ADF
Deviance Deviance
Constant: 1 50 7
Model 1: YO+FS+TC 10 4 40 3
Model 2: YO+FS+YOFS+TC 5 3 5 1
Model 3: YO*FS*TC Qo o 3 3
BI
Comparing the constant model and Model 1
The difference in the scated deviances is 40.
This is greater than 7.815, the upper 5% critical value of a 3 distribution.
So Model 1 is a significant improvement over the constant model. 04)
Alternatively, using the AIC to compare models, since A(deviance}>2x A(parameters) Model 1 is a
significant improvement over the constant model.
Comparing Model 1 and Model 2
The difference in the scaled deviances is 5.
This is greater than 3.841, the upper 5% critical value of a yf distribution.
So Model 2 is a significant improvement over Model 1. [4]
Alternatively, using the AIC to compare models, since A(deviance)>2x A(parameters) Model 2 is a
significant improvement over Model 1.
Comparing Model 2 and Model 3
The difference in the scaled deviances is 5.
This is less than 7.815, the upper 5% critical value of a 3 distribution.
So Model 3 is not a significant improvement over Model 2. pl
Alternatively, using the AIC to compare models, since A(deviance) ¥ 2x A{parameters) Model 3is
not a significant improvement over Model 2.
So Model 2 would be the mast appropriate in this case. fl 21.7 The cumulative claims paid each year under a certain cohort of insurance policies are recorded in
mag table below, for accident years 2010, 2011, 2012 and 2013.
0)
(ii)
(iii)
Development Year
Accident Year 0 1 2 3
2010 2,457 4,196 4,969 5,010
2011 2,648 4,715 5,561
2012 3,084 5,315
2013 3,341
Calculate the development factors under the basic chain ladder technique and state the
assumptions underlying the use of this method. [4]
The rate of claims inflation over these years, measured over the 12 months to the middle
of each year, is given in the table below.
2011 2.1%
2012 10.5%
2013 3.2%
Calculate the development factors under the inflation-adjusted chain ladder technique
and state the assumptions underlying the use of this method. (6)
Based on the development factors calculated in parts (i) and (ii), calculate the fitted values
under these two models and comment on how these compare with the actual
values. (7)
[Total 17] 21.7. (i) BCL development factors and assumptions
The development factors are:
4,196 +4,715 +5,315
2,457 +2, 648+ 3,084
= 1.737208
_ 4,969+5,561
© 4,196+4,715
181686
008251 (2)
The assumptions underlying the method are:
. Claim amounts are fully run off by the end of Development Year 3 tw
. Payments from each origin year will develop in the same way in monetary terms. I
. Weighted average past inflation will be repeated in the future. (a)
(ii) Inflation-adjusted chain ladder development factors and assumptions
First we need to disaccumulate the figures to obtain incremental figures. If we do this we obtain
the following results:
Development year
Accident year 0 1 2 3
2010 2457 1,739 7713 41
2011 2,648 2,067 846
2012 3,084 2,231
2013 3,341 We now inflate the incremental figures to get real values (in mid-2001 terms):
Development year
Accident year a 1 2 3
2010 2860.70 1983.09 T9774 41
2011 3019.67 2133.14 46
2012 3182.69 2231
2013 3341
al
We can naw find the cumulative figures in real terms:
Development year
Accident year o 1 2 &
2010 2860.70 4843.79 5641.53 5682.53
2011 3019.67 5152.81 5998.81
2012 3182.69 5413.69
2013 3341
So the development factors are now:
&= 843.79 +5,152.81+5,413.69 =1.700340
2,860.70 + 3,019.67 + 3,182.69
oy SELSSHEL
= eT 007268 Qi
The assumptions underlying the inflation adjusted chain ladder method are:
. the claim amounts are fully run off by the end of Development Year 3 (41
. Payments from each origin year will develop in the same way in real terms. [4]
. Rates of past and future claims inflation are appropriate. fu (iii) Fitted values
We now use the development factors to find the fitted claim amounts for past years. Using the
development factors for the basic chain ladder method we get:
Auuidert Yeur
2010
2011
2012
2013
So the fitted incremental payments are:
Accident Year
2010
2011
2012
2013
2,457
2,648
3,084
3,341
2,a57
2,648
3,084
3341
Development Year
1
4,268.32
4,600.13
5,357.52
2
5,043.81
5,435.90
Development Year
1
1,811
1,952
2,274
2
776
836
3
5,085.43
4]
3
a2
(1)
If we compare the fitted and the actual incremental payments, we get the following:
Accident Year
Fitted 2010
Actual 2010
A-F
Fitted 2011
Actual 2011
A-F
2,457
2,457
2,648
2,698
Development Year
1 2
1,811.32 775.49
1,739 773
72.32 =2.49
1,952.13 (835.78
2,067 846
114.87 10.22
41.62
41
0.62 Fitted 2012
Actual 2012
A-F
Fitted 2013
Actual 2013
A-F
3,084
3,084
3,341
3,341
2,273.55
2231
42.55
[2]
We can now repeat the process with the inflation-adjusted chain ladder method. Using the
inflation adjusted development factors on the real data amounts for Development Year 0, we get:
‘Accident Year
2010
2011
2012
2013
a
2,860.70
3,019.67
3,182.69
3,341
Development Year
z
4,864.17
5,134.47
5,411.65
2
5,663.98
5,978.73
3
5,705.15
fa]
if we now disaccumulate and compare with the actual incremental (real) values:
Accident Year
Fitted 2010
Actual 2010
A-F
Fitted 2011
Actual 2011
A-F
0
2,860.70
2,860.70
3,019.67
3,019.67
Development Year
1
2,003.47
1,983.09
—20.38
2,114.80
2,133.14
18.34
2
799.81
797.74
2.08
844.26
B45
1,74
41.16
aL
-0.16 Fitted 2012
Actual 2012
A-F
Fitted 2013
Actual 2013
A-F
3,182.69
3,182.69
3,341
3341
2,228.96
2231
2.04
The inflation adjusted method appears to fit the original data better, given that the residuals are
smaller in this case. However, there are not many categories of data, and i
either method will provide satisfactory results in the future.
s not clear whether
2] 20.14 Claims occur on a portfolio of insurance policies according to a Poisson process with Poisson
agg rometer 2. Claim amounts, X,,X2,..., are assumed to be identically distributed with moment
ig factor @(>0). The
insurer's adjustment coefficient, R , is defined to be the smallest positive root of the equation:
generating function My(t). The insurer calculates premiums using a loa
A+¢cr=AMx(r)
where c is the insurer's premium income rate.
(i) Using the above equation for R , or otherwise, show that, provided R is small, an
approximation to R is R, where:
ga2ie/a-m)
eae
where 4=E[X;] and o” =var[X)]. [4]
(ii) Describe how the adjustment coefficient can be used to assess reinsurance arrangements
on the basis of security, (3]
(iii) The Poisson parameter, 2, for this portfolio is 20 and all individual claims are for a fixed
amount of £5,000. The insurer's premium loading factor, 0, is 0.15 and proportional
reinsurance can be purchased from a reinsurer who calculates premiums using a loading
factor of 0.25.
Calculate the maximum proportion of each claim that could be reinsured so that the
insurer's security, measured by A, is greater than the insurer's security without
reinsurance. (9]
(Total 16] 20.14 (i) Approximation to R
The MGF of X can be written:
2 2
myter=cte™ y= etn 104 OO yma ste EO)
2
So: yl R)=1+ REX) +E) +
Assuming that R is small enough for terms in R® and higher powers to be neglected, we can
write:
p
My(R)=1+Ru + (1? +0?)
(a
(1) Substituting this into the defining equation for R :
e x
Ate AL Rut Ue +07)
Subtracting 4 from both sides and rearranging, we get:
re
(c—au)R =sue +0") if)
Dividing through by R and simplifying, we get:
Ac-Au) _Ac/A-w)
aueso*) oF tye
fl
This is the required expression.
(il) Assessment of reinsurance
The adjustment coefficient can be used to assess the effectiveness of different reinsurance
arrangements, using Lundberg’s inequality to find an upper bound for the probability of ruin for
the insurer under different reinsurance arrangements. An arrangement that produces a lower
upper bound for the probability of ruin is in some sense more secure for the insurer than an
arrangement that has a higher upper bound for the probability of ruin. 3]
Note however that the adjustment coefficient cannot tell us anything about the relative
profitability of different reinsurance arrangements. This will need to be assessed using other
means.
(iil) Maximum reinsurance
First we consider the insurer's security without reinsurance. The equation for the adjustment
coefficient is:
1+ (1+ O)myr=My(r)
Substituting in @=0.15, m, =5000 and My(r)=e"", we get:
50007
1+5750r=e (a
The rate of premium income is:
c=(1+0)im, =1.15x20x5000=115, 000
So, using the approximation derived for the adjustment coefficient in part (i), we have:
2{115000 /20- S000)
rx 21215000 /20- 5000) aot al
50007 Now assume that a proportion k of each risk is reinsured. The net premium income is now:
[1.15 5000 —1.25x kx 5000] 2 =(5750—6250k) A [1]
The MGF of the insurer’s net claim payments is now:
Mylt)= (25000(1-K)R ty
So the equation for the insurer's adjustment coefficient is now:
1+(5750-6250k)R = e°O-*" fa)
Using the same approximation to R as before, we have:
21(5750~6250k)-SO00(1-k)]___ 35k
(]
50007 (1—k)* 50000(1 — k)?
If we want the insurer's security with reinsurance to be greater than without reinsurance, we
want the adjustment coefficient with reinsurance to be larger, ie:
3-5k Z
——,>6x10* (]
50000(1 - ky
Rearranging this inequality, we get:
3k? —k<0 (1]
The solution of this is 0