1

Chapter 3 Aggregate Loss Models

1 Introduction
Aggregate losses are the total losses paid by an insurer for a defined portfolio of insureds in one
period, say a year. They are two ways to go about adding the losses in order to obtain the total for
the year.

Definition 1 The collective risk model,

, is the aggregate losses, with

as the random variable representing the number of losses in the year.

s are independent and

identically (i.i.d.) random variables, unless otherwise specified. More formally, the independence
assumptions are
1. Conditional on , the random variables

are i.i.d. random variables.

2. Conditional on , the common distribution of the random variables

does not depend on .
3. The distribution of does not depend in any way on the values of

.

Definition 2 The individual risk model represents the aggregate loss as a sum,

, of a fixed number, , of insurance contracts. The loss amounts for the contracts are
(

), where

s are assumed to be independent but are not assumed to be identically

distributed.

The individual risk model is used in modeling the losses of a group life or health insurance policy
that covers a group of employees. Each employee can have
1. different coverage (life insurance benefit as a multiple of salary)
2. different level of loss probabilities (different ages and health status)


2 The Compound Model for Aggregate Claims
2.1 Compound Model
Definition 3

has a compound distribution if it is a combination of

representing the number of losses and the

s representing the size of the loss .

2

Some common properties of the compound distribution are:
() ()() () ()

() ()()
Proof:
1. From probability generating function of ,

() (

)
(

)( ) [

]( )

( ) [

] ( )

( )[

()]

()

()]

Here the primary distribution is the claim frequency distribution and the secondary
distribution is the claim size distribution.

()

()]

()
()

()

()]

() ()()

()

()

()] [

()]

()]

()

() [( )]

() ()[( )]

()

()

() [

()]

()

() ()()

2. Double expectation theorem
() [()] [(

)] [()] ()()

() [()] [()] [()] [()]
()

() ()()

Many of the aggregate loss problems on the exam involve identifying and severity ,
and then finding () and (), and then applying the normal approximation to to compute
probabilities when the sample size is large.
If the severity is discrete, then the aggregate loss distribution is discrete, and a continuity
correction is required. This means adding or subtracting from the bound. For example, if you
3

are asked for the probability that is greater than 100, then we would calculate the probability
that the normal variable is greater than 100.5.

Example 1 For a group of 100 insureds, the number of losses per insured follows a negative
binomial distribution with . The claim sizes follow an inverse gamma distribution
with . The number of losses is independent of claim sizes, and claim sizes are
independent of each other. Determine the mean and variance of aggregate losses.

Example 2 A claim severity distribution is exponential with mean 1000. An insurance
company will pay the amount of each claim in excess of a deductible of 100. Calculate the
variance of the amount paid by the insurance company for one claim, including the possibility
that the amount paid is 0.

Example 3 For a group insurance policy, the number of claims from a group has a binomial
distribution with mean 100 and variance 20. The size of each claim has the following distribution:
Claim size 1 2 3 4
Probability 0.5 0.35 0.10 0.05
Determine the mean and variance of aggregate losses.

Example 4 For aggregate losses

, you are given:

(i) has a Poisson distribution with mean 500.
(ii)

have mean 100 and variance 100.

(iii)

are mutually independent.

(iv) For a portfolio of insurance policies, the loss ratio is the ratio of aggregate losses to
aggregate premiums collected.
(v) The premium collected is 1.1 times the expected aggregate losses.
Using the normal approximation to the compound Poisson distribution, calculate the probability
that the loss ratio exceeds 0.95.

Example 5 The number of claims on a policy has a Poisson distribution with . Claim
sizes have a gamma distribution with . Aggregate losses for the policy are
approximated with a lognormal distribution matching the mean and variance of the aggregate
distribution. Calculate

() using this approximation.

4

Example 6 You are the producer of a television quiz show that gives cash prizes. The
number of prizes, , and the prize amounts, , have the following distributions:
( ) ( )
1 0.8 0 0.2
2 0.2 100 0.7
1000 0.1
Your budget for prizes equals the expected prizes plus the standard deviation of prizes. Calculate
your budget.

Example 7 The number of claims on a policy is uniformly distributed on the integer 1
through 5. Claim sizes have a Poisson distribution with mean 30. Calculate the probability that
the aggregate loss greater or equal to 120.


2.2 Convolution Method
An alternative approach to find the distribution of the sum of random variables is the method of
convolution.
The random sum,

(where has a counting distribution) has

distribution function

() ( )

( )

()

where

() ( ) is the common distribution function of the

s,

( ), and

() is the -fold convolution of the cdf of . It can be obtained as

() {



and

()

()
( )

()

()

()
( )

()

For =1, the equation reduces to

()

(). By differentiating, the pdf is

()

()
( )

()

()

()
( )

()

Let

() and

() for . Then, has a pf

() ( )

()

() ( )

()

Example 8 Let

[ ] for , and let

() ( )

() ( ) , then

() and

() can be calculated as follows.

Example 9 For an insurance coverage, the number of claims has a negative binomial
distribution with parameters and . Claim size is distributed as follows:
Claim size 1 2 3
Probability 0.5 0.4 0.1
Calculate

().

Example 10 The number of claims on an insurance coverage follows a binomial distribution
with parameters . The size of each claim has the following distribution:
Claim size 0 1 2
Probability 0.5 0.35 0.15
Calculate the probability of aggregate claims of 3 or more.

Example 11 The number of claims on an insurance coverage follows a Poisson distribution
with mean 4. The size of each claim has the following distribution:
Claim size 1 2 3 4
Probability 0.7 0.2 0.05 0.05
Calculate the probability of aggregate claims of 4 or more.

Example 12 Customers arrive in a store at a Poisson rate of 0.5 per minute. The amount of
profit the store makes on each customer is randomly distributed as follows:
Profit 0 1 2 3
Probability 0.7 0.1 0.1 0.1
Determine the probability of making 2 profits in 10 minutes.
6

2.3 The Recursive Method
Suppose that the severity distribution

() is define on representing multiples of

some convenient monetary unit. The number represents the largest possible payment and could
be infinite. Further suppose that the frequency distribution,

, is a member of the ( ) class

and therefore satisfies

Then, the following result holds.
Theorem 1 For the ( ) class,

()

( )

Corollary: For the ( ) class,

()

The starting value of the recursive is

() ( )

()]

]

Example 13 The number of claims in a period has a geometric distribution with mean 4. The
amount of each claim X follows ( ) . The number of claims and claim
amounts are independent. is the aggregate claim amount in the period. Calculate

().

Example 14 The number of claims on an insurance coverage follows a binomial distribution
with parameters . The size of each claim has the following distribution:
Claim size 0 1 2
Probability 0.5 0.35 0.15
Calculate the probability of aggregate claims of 3 or more.

Example 15 Taxis arrive at an airport in a Poisson process at a rate of 4 per minute. Each taxi
picks up 1 to 4 passengers, with the following probabilities:
Number of passengers 1 2 3 4
Probability 0.7 0.2 0.05 0.05
Calculate the probability that in one minute 4 or more passengers leave the airport by taxi.
7

Example 16 For an insurance coverage, the number of claims has a zero-modified negative
binomial distribution with parameters , and

. Claim size is distributed as

follows:
Claim size 1 2 3
Probability 0.5 0.4 0.1
Calculate

().


3 Analytic Results
For most choices of distributions of and

s, the compound distributional values can only be

obtained numerically, for example the recursive method. But, for certain combination of choices,
simple analytic results are available, thus reducing the computational problems considerably.
For Example, compound geometric-exponential, compound negative binomial-
exponential, and exponential severities.

1. Compound negative binomial-exponential
( ) ()
The moment generating function (mgf) of is

()

()]

[( )

] { [( )

]}

{[ ( )]

})

()]
where

( )]

the pgf of the binomial distribution with parameter and ( ) , and

()
[ ( )]

is the mgf of the exponential distribution with mean ( ).

The distribution function is

() (

) (

( )

()

When , has a compound geometric-exponential distribution, then
8

()

()
]


This is a two-point mixture of a degenerate distribution with probability 1 at zero with
weight

and an exponential distribution with mean ( ) with weight

.


Example 17 Claim counts follow a geometric distribution with mean 0.2. Claim sizes follow
an exponential distribution with mean 1000.
(a) Determine the probability that aggregate losses will be greater than 5000.
(b) A stop loss reinsurance contract pays all losses with an aggregate deductible of 5000,
determine the expected losses paid under the contract.

Example 18 Claim size on a coverage follow an exponential distribution with mean 500. 100
lives are covered under the contract. Claim counts for each life follow a negative binomial
distribution with mean 0.1 and variance 1.1. Claim counts and claims are independent. A stop-
loss reinsurance contract is available at 150% of expected claim cost. You are willing to pay 1500
for the contract. Determine the aggregate deductible needed to make the cost of the contract 1500.


2. Gamma severities
If

( ), and

, then
( )
The distribution function is

()

where ( )

substituting, we get

()

This is still an infinite sum, but if the frequency distribution is bounded, for example
binomial frequency distribution, it is a finite sum.
9

() (

( )

where

, and

( )

.

Example 19 Claim counts have a binomial distribution with . Claim sizes are
exponential with mean 1000. Calculate the probability that aggregate claims are less than their
mean.

Example 20 Claim sizes follow a gamma distribution with parameters and .
Claim counts are independent of claim sizes, and have the following distribution:
0 1 2

0.6 0.3 0.1

Calculate the probability that aggregate claims are less than 120.


4 Discretizing
The recursive and convolution methods for calculating the aggregate distribution require a
discrete distribution. Usually the severity distribution is continuous. There are two methods for
discretizing the distribution:
1. Method of rounding and
2. Method of local moment matching.
In both methods, we pick a span, , the distance between the points that will have a positive
probability in the discretized distribution.

4.1 Method of Rounding
Let

denote the probability placed at . Then set

)

Note:

( ) indicates that discrete probability at x should not be included. For continuous

distributions, this will make no difference.
10

Example 21 Loss sizes follow a Pareto distribution with . The distribution will
be discretized by the method of rounding with a span of 4. Calculate the resulting probabilities of
0, 4, 8, and 12;

, and

.

Example 22 has an exponential distribution with mean 1. Calculate

of the distribution
discretized using the method of rounding with a span of 1.


4.2 Stop-Loss Insurance
Definition 4 Insurance on the aggregate losses, subject to a deductible, is called stop-loss
insurance. The expected cost of this insurance is called the net stop-loss premium and can be
computed as [( )

], where is the deductible and the notation ( )

means to use the value

in parentheses if it is positive but use zero otherwise.

For any aggregate distribution,
[( )

] [

()]

( )

()

( )

()

Note that [( )

] () ( ), since () ()(), we only have to deal with

( ).

For a discrete distribution in which the only possible values are multiples of , it becomes
( )

( )

Example 23 Claim counts follow a Poisson distribution with mean 3. Claim sizes follow an
exponential distribution with . This severity distribution is discretized uisng the method of
rounding with span 1. Claim counts and claim sizes are independent. A stop-loss reinsurance
contract has a deductible of 1.6. Calculate expected losses paid by the reinsurance contract.
11

Example 24 On an insurance coverage, the number of claims has a geometric distribution
with mean 4. The distribution of claim sizes is as follows:
2 4 6 8
( ) 0.45 0.25 0.20 0.10
Calculate [( )

].

Example 25 The number of claims on an insurance coverage had a negative binomial
distribution with mean 2 and variance 6. The claim size distribution is binomial with parameters
. A reinsurance contract pays aggregate claims over an aggregate deductible
of 2. Determine the expected aggregate loss paid by reinsurance.

Example 26 A company provides insurance to a concert hall for losses due to power failure.
You are given:
(i) The number of power failures in a year has a Poisson distribution with mean 1.
(ii) The distribution of ground up losses due to a single power failure is:
10 20 50
( ) 0.30 0.30 0.40
(iii) The number of power failures and the amounts of losses are independent.
(iv) There is an annual deductible of 30.
Calculate the expected amount of claims paid by the insurer in one year.

Example 27 You are given:
The number of claims follows a binomial distribution with .
Claim sizes follow the following distribution:
Claim size 0 1 2 3
Claim probability 0.20 0.50 0.20 0.10
A reinsurance policy has an aggregate deductible of 6.
Determine the expected aggregate amount paid by the reinsurer.

Example 28 You are the producer of a television quiz show that gives cash prizes. The
number of prizes, , and the prize amounts, have the following distributions:
( ) ( )
1 0.8 0 0.2
2 0.2 100 0.7
1000 0.1
You buy a stop-loss insurance for prizes with deductible of 200. The cost of insurance is 275% of
the expected claim cost. Calculate the cost of the insurance.