SCHOOL OF BUSINESS, ECONOMICS AND MANAGEMENT

BAS380 - LOSS DISTRIBUTION AND CREDIBILITY THEORY

MID-SEMESTER EXAMINATION

TUESDAY 8TH OCTOBER, 2024

09:00 - 12:00 HOURS

Time allowed: 3 HOURS Plus 5 Minutes Reading Time

Instructions to Candidates:
(1) Read the instructions very carefully.
(2) Check that you have the correct examination paper in front of you.
(3) There are FIVE questions in this paper. Answer ANY FOUR(4) questions.
(4) All questions must be answered in the answer booklet provided only.
(5) No books, files or mechanical/ electronic aids are permitted in the examination
room. Students are permitted to use a non-programmable electronic calcula-
tor. Tables for Actuarial Examinations (2002) shall be provided.
(6) Begin answering each question on a new page.
(7) Write down the number of questions that you have answered on the cover of
the examination answer booklet provided.
(8) There shall be NO communication among students during the examination.
Any students caught doing this will be disqualified.

DO NOT TURN THIS PAGE UNTIL YOU ARE TOLD TO DO SO.

1
QUESTION ONE

(a) Explain and/or state the following terminologies.

(i) Extreme events.

(ii) Fisher-Tippet Theorem.
(iii) Peak over threshold approached.
(iv) Block maxima approach.

[4 marks]

(b) The number of claims from the named insurance company under the motor car in-
surance policy is modelled using the Generalized Extreme Value (GEV) distribution.
The goal of this analysis is to evaluate the likelihood of exceeding the highest claims
amount observed in previous years. The analysis provides the following results:

Parameter Estimate Standard Error

Location 263,425.54 29,833.23
Scale 99,630.76 30,421.89
Shape -0.3645 0.3924

Table 1: Estimated parameters of the GEV distribution.

Return Period (years) 1.5-Year Level 2-Year Level 2.5-Year Level

Amount(ZMW) 253,893.0 297,607.3 322,786.5

Table 2: Estimated return levels for different periods.

(i) State what approach was used to seclude the maximum claim amount under
this analysis.
[2 marks]
(ii) Using the parameters estimated from Table 1, state what type of generalized
extreme value distribution(GEV) was used and provide a reason for your an-
swer.
[2 marks]
(iii) Interpret the outputs given in Table 2.
[3 marks]
(iv) Compute annual exceedance probability for each year level in Table 2 and in-
terpret.

Page 2 of 7
 [3 marks]

(c) The Burr distribution is a flexible probability distribution used in economics, finance,
and reliability analysis to model various data distributions, including heavy tails, and
is commonly used in risk management and actuarial science for predicting extreme
events. Show that the expressions of mean and variance of the Burr distribution
when α = 3 and γ = 2 are given by

3π √ 9π 2
 
E[X] = λ and Var[X] = λ 1 −
16 128

[5 marks]

(d) Claim amounts on a portfolio of insurance policies are modeled using Weibull dis-
tribution. A quarter of losses are below 15 and a quarter of losses are above 80.
Estimate the parameters c, γ of the Weibull distribution that fit this data. Hence use
the results and determine whether or not this Weibull distribution has a heavier tail
than that of the exponential distribution with parameter c.

[6 marks]

[Total: 25 marks]

Page 3 of 7
QUESTION TWO

(a) For three years, an insurance company has insured buildings in three different towns
against the risk of fire damage. Aggregate claims in the jth year from the ith town
are denoted by Xij for i = 1, 2, 3 and j = 1, 2, 3. The data is given in the table below.

Town i Year j
Year 1 Year 2 Year 3
Town 1 8,130 9,210 8,870
Town 2 7,420 6,980 8,130
Town 3 9,070 8,550 7,730

Table 3: Aggregate claims (in ZMW) from three towns over three years

Calculate the expected claims from each town for the next year using the assump-
tions of the Empirical Bayes Credibility Theory model 1.

[12 marks]

(b) The number of claims from the life insurance policy is modelled by the random vari-
able X with density
f (x) = θxe−θx , x ≥ 0
The parameter θ is assumed to follow a prior gamma distribution with parameters α
and λ.

(i) Prove that the posterior distribution has the following form
nX̄
f (x|θ) = θα+n−1 e−θ( 2
+λ)

Hence, state the posterior distribution of θ and its parameters.

[8 marks]
(iii) The number of claims of the ten customers recorded in the space of 2 years
are given in the table below and the prior knowledge is determined with α = 4,
λ = 1.5.

3 2 3 2 6 4 4 3 2 1
Calculate a point estimate of the number of claims using Bayesian estimation
under all-or-nothing loss.

[5 marks]

[Total: 25 marks]

Page 4 of 7
QUESTION THREE

(a) An insurance company provides coverage for individuals against the costs of med-
ical treatment. The annual cost of claims under a policy is assumed to follow a
normal distribution with a mean µ, which varies depending on the individual, and a
known variance of 25. The mean cost of claims µ is modelled as a linear function of
the person’s age x, such that
µ = α + βx
where α and β are constants to be estimated, and x represents the age of the
insured person.
You are given the following data for 50 policies, where yi is the cost of claims last
year for the ith policy and xi is the age of the corresponding person:

50
X 50
X 50
X 50
X
xi = 63.7, yi = 549.2, xi yi = 7453.2, x2i = 831.2
i=1 i=1 i=1 i=1
Determine the maximum likelihood estimates of α and β.

[10 marks]

(b) The Empirical Bayes Credibility Theory (EBCT) Model 1 is a generalization of the
normal-normal model. This model can be used to estimate the future premium or
the number of claims in the coming year. The credibility factor under EBCT Model 1
is formulated as
n
Z= E[S 2 (θ)]
n + V ar[m(θ)]
State the meaning of the following component of the formula and how they affect the
credibility factor Z.
(i) n
[2 marks]
(ii) E[S 2 (θ)]
[3 marks]
(ii) V ar[m(θ)]
[3 marks]
(c) Claim numbers from individual policies in a portfolio have a Bin(n, p) distribution.
The parameter p varies over the portfolio with a Beta(α, β) distribution. Find the
mixture distribution.

[7 marks]

[Total: 25 marks]

Page 5 of 7
QUESTION FOUR

(a) State three uses of loss distributions in general insurance.

[3 marks]

(b) Define the following terms

(i) Credibility.
(ii) Collateral data.
(iii) Prior Distribution.
(iv) Credibility premium formula.

[4 marks]

(c) Namuunza, a thrid-year actuarial science student, is doing a research project on

Bayesian credibility. She believes the total annual claim amount on a particular
insurance policy follows a normal distribution with unknown mean θ and variance 2.
Prior beliefs about θ are described by a normal distribution with a mean of 16 and a
variance of 8. Claim amounts x1 , x2 , . . . , xn are observed over n years.

(i) Derive the posterior distribution of θ and clearly state its parameters .
[12 marks]
(ii) Show that the mean of the posterior distribution of θ can be written in the form
of a credibility estimate and write down the credibility factor expression.
[1 mark]
(iii) Now suppose that total claims over the five years were 170. Calculate the
posterior probability that θ is greater than 30.
[5 marks]

[Total: 25 marks]

Page 6 of 7
QUESTION FIVE

(a) An insurance analyst believes that the number of claims made by people during a
given period follows a binomial distribution with parameters n = 3 and p. A random
sample of 120 people showed the following results:

Number of Claims Made 0 1 2 3

Number of People 40 60 15 5

(i) Show that the maximum likelihood estimate for p is p̂ = 0.2917.

[7 marks]
(iii) Perform a goodness of fit test for the binomial model Bin(3, p) at a significance
level of 5%.
[8 marks]

(b) An insurance company has a portfolio of insurance policies. Claims arise according
to a Poisson process, and claim amounts have a probability distribution with the
parameter θ.

(i) State one assumption the insurance company is likely to make when modelling
aggregate claim amounts.
[2 marks]
(ii) Explain what the Maximum Likelihood Estimate (MLE) of θ represents.
[2 marks]
(iii) State three alternative to using the MLE.
[3 marks]
(iv) Suggest three complications that may arise for the insurance company when it
uses past claims data to determine the MLE of θ.
[3 marks]

[Total: 25 marks]

END OF EXAMINATION PAPER

Page 7 of 7