BAS 310-Risk Theory for Actuarial Science

LECTURE-NOTES

Lecturer: Mr. Saviour Chibeti

August 3, 2022

1
1 RISK MODEL 1
Objectives:

ˆ Explain the notion of individual risk model.

ˆ Look at the approximations of individual risk model

1.1 Introduction

The random variable of interest in the individual risk model is the total claims on a portfolio of
insurance contracts. We wish to determine, for example, the likelihood that a specic amount of
capital will be adequate to cover these claims or the portfolio's value-at-risk at level 95%, which
corresponds to the 95% quantile of its cumulative distribution function (cdf). The aggregate
of all claims on each individual policy, which is considered to be independent, is how the total
claims are modeled.

2 Individual Risk Model(IRM)

Using a distinct distribution for each risk to represent each risk is an alternate approach to
modeling aggregate loss. In a portfolio of risks, for instance, one risk might be modeled with a
gamma distribution and another with an exponential distribution. The total of the individual
estimated losses would represent the expected aggregate loss and is known as an individual risk
model. . This is given by
n
X
S= Xi .
i=1

where Xi 's are random variables representing the individual losses and n is the number of in-
dividual risks in the portfolio and it known.

In this model, a portfolio made up of a xed number of risks is taken into account. It is
assumed that:
• these risks are independent

• the claim amounts resulting from these risks are not (necessarily) distributed randomly

• the number of risks remains constant throughout the insurance period.

For each risk, the following assumptions are made:

• the number of claims from the j − th risk, Nj , is either 0 or 1

• the probability of a claim from the j − th risk is qj

Since a person can only pass away once in a given period of time, the model is especially suit-
able for life insurance policies because it is assumed that the number of claims from the j − th
risk, Nj , will either be 0 or 1.

Consider a portfolio of n independent policies. Suppose that the probability of a claim arising

2
on any given policy in a portfolio is q , the mean is given by µ and the variance is σ 2 , we have
that the average aggregate claim is given by

E(S) = nqµ,

and the variance of the aggregate claim is given by

V ar(S) = nqσ 2 + nq(1 − q)µ2 .

Example 2.0.1. Consider a portfolio of 1, 000 independent policies. Suppose that the probabil-
ity of a claim arising on any given policy in the portfolio is 0.004. The claim amounts follow a
Gamma distribution with parameters α = 5 and β = 0.002. Find the mean and variance of the
aggregate claim amount in Kwacha.

Solution
Remember, we learn by doing. We now compute the mean and the variance of the claim amount,
we have that the mean is given by,

α
µ=
β

5
=
0.002

= K2, 500

and the variance is given by,

α
σ2 =
β2

5
=
0.0022

= K1, 250, 000.

So the mean and variance of the aggregate claim amount are:

E(S) = nqµ

= 1, 000 × 0.004 × 2, 500

= K10, 000

3
and

V ar(S) = nqσ 2 + nq(1 − q)µ2

= 1, 000 × 0.004 × 1, 250, 000 + 1, 000 × 0.004 × 0.996 × 2, 5002

= 5, 000, 000 + 24, 900, 000

= K29, 900, 000

Task 2.0.2. Assume there is a chance of 0.2 that there is a claim. When a claim occurs the loss
is exponentially distributed with parameter λ = 0.5. Find the mean and variance of the claim
distribution. Suppose there are 500 independent policies with this loss distribution, compute the
mean and variance of their aggregate loss.

2.1 Approximations of the Individual Risk Model

The distribution of S can be approximated, which is a whole dierent strategy. By using a

normal distribution with the same mean and variance as S , we can approximate the distribu-
tion of S if we think of S as the sum of a "big" number of random variables. However, it is
challenging to dene "big" explicitly, and this approximation is frequently unsatisfactory for
the insurance industry.

Considering the mean and variance of the aggregate loss S , we may approximate S by

 
S − E(S) s − E(S)
P (S < s) = P p ≤p
V ar(S) V ar(S)

 
s − E(S)
=P Z≤ p
V ar(S)

 
s − E(S)
≃Φ p
V ar(S)

Example 2.1.1. For the aggregate loss S in a collective risk model, the number of claims
follow a poison distribution with parameter λ = 100 and the claim size follow an exponential
distribution with parameters α = 0.5. Approximate the distribution of S using the normal
distribution and compute
i The probability that the aggregate loss will exceed 180
ii The probability that the aggregate loss won't exceed 230
Solution

4
Remember, we learn by doing. We now compute the average aggregate loss which is given
by

E(S) = E(N )E(X).

We have that the average number of claims E(N ) is

E(N ) = 100

and the average claim size is

1
E(X) =
α

1
=
0.5

= 2.

Then we have that

E(S) = E(N )E(X)

= 100 × 2

= 200.

The variance of the aggregate loss is given by

V ar(S) = E(N )V ar(X) + V ar(N )[E(X)]2 .

Computing the variance of the number of claims and claim size, we have that

V ar(N ) = 100

and

5
 1
V ar(X) =
α2

1
=
0.52

= 4.

Therefore, we have that

V ar(S) = E(N )V ar(X) + V ar(N )[E(X)]2

= 100 × 4 + 100 × 22

= 800.

We can now approximate the distribution of the aggregate loss by the normal.
i To nd the probability that the aggregate loss will exceed 180, we have to compute

P (S > 180)

We now have
 
S − E(S) 180 − E(S)
P (S > 180) = P p > p
V ar(S) V ar(S)

 
180 − 200
≃1−Φ √
800

= 1 − Φ(−0.7071)

= 0.7612

i To nd the probability that the aggregate loss won't exceed 230, we have to compute

P (S > 230)

6
We now have
 
S − E(S) 230 − E(S)
P (S < 230) = P p < p
V ar(S) V ar(S)

 
230 − 200
≃Φ √
800

= Φ(1.0607)

= 0.85543