Chapter 9
Note. Here is a proof for E[X1 + +XN |N = n] = E[X1 + +Xn ] = nE(X1 ) : We prove this
for E[S|N = 2] = 2E(X1 ). First of all let be the common support of the random variables
{X1 , ..., Xn } (these variables are identically distributed and so have identical density function
and so have identical support).
Next Step. Note that on the event {N = 2} the random variable S becomes X1 + X2 so it
takes on all values s = x1 + x2 where x1 and x2 . So the support of the conditional
random variable S|{N = 2} is the set = {s = x1 + x2 : x1 , x2 }.
Next Step. For s we have
P {S = s | N = 2} = P {S = x1 + x2 | N = 2} = P {X1 + X2 = x1 + x2 | N = 2}
P (X1 + X2 = x1 + x2 , N = 2) P (X1 + X2 = x1 + x2 ) P (N = 2)
= = = P (X1 +X2 = x1 +x2 )
P (N = 2) P (N = 2)
Next Step.
E(S|N = 2) = s P (S = s|N = 2) = (x1 + x2 )P (X1 + X2 = x1 + x2 )
s x1 x2
Theorem. Let {X1 , X2 , ...} be an independent and identically distributed sample from a claim.
If the number of claims N is independent of the claim amount , then
Proof.
E(S) = E( N j=1 Xj )
N
= k=0 P (N = k) E( j=1 Xj | N = k)
k
= k=0 P (N = k) E( j=1 Xj ) independence is used here - check it out
= k=0 P (N = k) k E(X1 )
( )
= k=0 k P (N = k) E(X1 ) = E(N ) E(X1 )
1
Theorem. Let {X1 , ..., Xn } be an independent and identically distributed sample from a claim.
If the number of claims N is independent of the claim amount , then
N
(i) E( i=1 Xi |N ) = N E(X1 )
N
(ii) Var( i=1 Xi |N ) = N Var(X1 )
N
(iii) Var( i=1 Xi ) = E(N ) Var(X1 ) + Var(N ) E(X1 )2 compound variance formula
N N n
E( Xi |N )(n) = E( Xi |N = n) = E( Xi ) = n E(X1 )
i=1 i=1 i=1
Since this is true for all n we have proved the claim of part (i) .
N N n
Var( Xi |N )(n) = Var( Xi |N = n) = Var( Xi ) = n Var(X1 )
i=1 i=1 i=1
Since this is true for all n we have proved the claim of part (ii) .
Proof of (iii).
Var(S) = E[Var(S|N )] + Var[E(S|N )]
N
= E[Var( N
i=1 |N )] + Var(E[ i=1 |N ])
Note. When the primary distribution is Poisson(), then the compound variance formula re-
duces to
2
Example. Let N Poisson(2) and let {X1 , X2 , ...} be an independent identically distributed
N
sequence with common distribution N (15 , 2 = 5) . Calculate E( N
i=1 Xi ) and Var( i=1 Xi )
Solution. We have
E(N ) = 2 , Var(N ) = 2
N
E( Xi ) = E(N ) E(X1 ) = (2)(15) = 30
i=1
N
Var( Xi ) = E(N ) Var(X1 ) + Var(N ) E(X1 )2 = (2)(5) + (2)(15)2 = 460
i=1
3
Individual risk model
Aggregate loss: total amount of loss in one period.
Collective risk model
x fX (x)
100 0.8
500 0.16
1000 0.04
Calculate the probability that aggregate claims will be exactly 600.
Solution.
4
X1 = 100 , X2 = 500
X = 500 , X = 100
1 2
with probability
( )
e5 56
P (S = 600 , N = 6) = P (N = 6 , X1 = 100 , ... , X6 = 100) = 6! (0.8)6 = 0.0383
5
Solution. {
1
0 S=0 Prob. =
5
S = 25
Prob. = ( 35 )( 13 ) = 1
1 5
S = 150 Prob. = ( 35 )( 23 ) = 2
5
N
S = 100
Prob. = ( 15 )( 23 )( 32 ) = 4
45
2 S = 250 Prob. = 2( 51 )( 23 )( 13 ) = 4
45
S = 400 Prob. = ( 51 )( 13 )( 31 ) = 1
45
Example (Problem 9.55 of the textbook). A population has two classes of drivers. The
number of accidents per individual driver has a geometric distribution. For a driver selected
at random from Class I, the geometric distribution parameter has a uniform distribution over
the interval (0,1). Twenty-five percent of the drivers are in Class I. All drivers in Class II have
expected number of claims 0.25. For a driver selected at random from this population,
determine the probability of exactly two accidents.
Solution.
1 1
2
P(N = 2 | Class I) = P (N = 2|)(1)d = d
0 0 (1 + )3
2
(u 1)2
= du u=1+
1 u3
2( ) 2( )
2
u2 2u + 1 1 2 1 1 2 3
= du = 2+ 3 du = 2u + u du
1 u3 1 u u u 1 u
6
[ u2 ]u=2
= ln |u| + 2u1 + = 0.068
2 u=1
2 (0.25)2
P(N = 2 | Class II) = = = 0.032
(1 + )3 (1.25)3
P(N = 2) = P (N = 2|class II)P (class II) + P (N = 2|class I)P (class I) =
(0.068)(0.25) + (0.032)(0.75) = 0.041
Example. There are two groups of good and bad drivers in the city. The good drivers make
up 70% of the population of the drivers. The annual claim amount of each good driver is
uniformly distributed on (0, 3000). The annual claim for each bad driver is exponentially
distributed with mean = 5000. As many as 100 drivers are chosen at random. Find the
mean and variance of the aggregate claim.
Using the table , for Exponential with mean the k-th moment is k (k + 1). therefore:
E(X 2 |bad) = (5000)2 (3) = (25, 000, 000)(2) = 50, 000, 000
7
Some statistics of compound distribution
Note 1:
If PX (z) = E[z X ] is the PGF of a random variable X , then
[ ]
d X [ ]
PX (z) =E (z ) = E X z X1
dz
[ ]
d [ ]
PX (z) =E (X z X1
) = E X(X 1) z X2
dz
[ ]
(k)
PX (z) = E X(X 1) (X k + 1) z Xk
P (1) = E(X)
X
PX (1) = E[X(X 1)]
Note 2:
8
[ ]
PS (z) = E(z S ) = E(z 0 )P (N = 0) + E z X1 ++Xk | N = k P (N = k)
k=1
[ ]
= P (N = 0) + E z X1 ++Xk P (N = k)
k=1
[ ]k
= P (N = 0) + E z X1 P (N = k) independent and identical distribution
k=1
k
= P (N = 0) + k=1 w P (N = k) w = E(z X1 )
k
= k=0 w P (N = k)
( )
= E wN = PN (w) = PN (PX (z))
So
PS (z) = PN (PX (z))
Then :
9
E(S 2 ) = MS (0) = PN [MX (0)] (MX
(0))2 + P [M (0)] M (0)
N X X
= PN (1) (MX
(0))2 + P (1) M (0)
N X
{ }
= E(N ) E(X 2 ) E(X)2 + E(N 2 )E(X)2
Then
{ }
= E(N )Var(X) + E(N 2 )E(X)2 E(N )2 E(X)2
{ }
= E(N )Var(X) + E(N 2 ) E(N )2 E(X)2
Example (problem 9.57 of the textbook) . Aggregate losses have a compound Poisson
claim distribution with = 3 and individual claim amount distribution p(1) = 0.4, p(2) = 0.3,
p(3) = 0.2, and p(4) = 0.1. Determine the probability that aggregate losses do not exceed 3.
Solution .
P (S = 0) = P (N = 0) = e3
10
Next :
P (S = 1) = P (N = 1 , X1 = 1) = P (N = 1)P (X1 = 1) = (e3 3)(0.4) = (1.2)e3
Next :
P (S = 2) = P (N = 1 , X1 = 2) + P (N = 2 , X1 = 1 , X2 = 1) = P (N = 1)P (X1 =
3 32
2) + P (N = 2)P (X1 = 1)P (X2 = 1) = (e3 3)(0.3) + ( e 2 )(0.4)2 = (1.62)e3
Next :
P (S = 3) = P (N = 1 , X1 = 3) + P (N = 2 , X1 = 1 , X2 = 2) + P (N = 2 , X1 = 2 , X2 =
3 32 3 33
1) + P (N = 3 , X1 = 1 , X2 = 1 , X3 = 1) = (e3 3)(0.2) + 2( e 2 )(0.4)(0.3) + ( e 3! )(0.4)3 =
(1.968)e3
Next :
Adding up the probabilities such found , we have
P (S 3) = 5.788 e3 = 0.288
Theorem. Suppose that the random variables {S1 , ..., Sn } are compound Poisson with
parameters 1 , 2 , ... , n and severity distribution functions F1 (x) , ... , Fn (x). If the Si s are
independent, then the sum S = S1 + + Sn has a compound Poisson distribution with
parameter = 1 + + n . Further we have the following:
(i): If the severity distribution functions for S1 , ... , Sn are F1 (x) , ... , Fn (x) , then the
severity distribution function for S is F (x) = nj=1 j Fj (x)
(ii): If the severity density functions for S1 , ... , Sn are f1 (x) , ... , fn (x) , then the severity
density function for S is f (x) = nj=1 j fj (x)
(iii): If the severity PGF functions for S1 , ... , Sn are P1 (z) , ... , Pn (z) , then the severity
PGF function for S is P (z) = nj=1 j Pj (z)
(iv): If the severity MGF functions for S1 , ... , Sn are M1 (t) , ... , Mn (t) , then the severity
MGF function for S is M (t) = nj=1 j Mj (t)
Proof. Let Pj (z) be the PGF of Fj (x) . Let Nj be the Poisson random variable for Sj . Then
11
And:
n
PS (z) = j=1 PSj (z)
n j (Pj (z)1)
= j=1 e
( )
n
= exp j=1 {j (Pj (z) 1)}
( n )
n
= exp j=1 j Pj (z) j=1 j
( { })
n j
= exp j=1 Pj (z) 1
12
Normal and Log-normal approximation for aggregate risk
models
Note. When the severity distribution of X is discrete , use continuity correction. The
continuity correction works this way: Suppose that a and b are two values of S such that S
does not take any value in the interval (a, b). Then the probabilities P (S > a) and P (S b)
a+b
are the same. We use P (S > 2 ) instead of these two. On the other hand, the probabilities
P (S a) and P (S < b) are equal, but we use P (S < a+b
2 ) instead of them.
Here are two examples:
Example (from the textbook). An insurable event has a 10% probability of occurring and
when it occurs results in a loss of 5,000. Market research has indicated that consumers will
pay at most 550 to purchase insurance against this event. How many policies must a company
sell in order to have a 95% chance of making money (ignoring expenses)?
Solution. Let n be the number of policies sold and let C be the number of claims. Then
C Binomial(n; 0.1). Then
E(C) = nq = (0.1)n
Var(C) = nq(1 q) = (0.1)(0.9)n
The cost for the company will be 5000C and what it receives is 550n. We want to have
13
5000C < 550n with 95% probability. So , we choose n so as to satisfy:
Since
CE(C)
N (0, 1) , the minimum number n must satisfy
Var(C)
0.11n 0.1n
= 1.96 n 3457.4 minimum n = 3458
(0.1)(0.9)n
Note. In these sort of problems where we search for the number n, do not do continuity
correction.
The department must buy a maintenance contract to cover repairs if there is at least a 10%
probability that aggregate maintenance costs in a given year will exceed 120% of the expected
costs.
Using the normal approximation for the distribution of the aggregate maintenance costs,
calculate the minimum number of computers needed to avoid purchasing a maintenance
contract.
Solution. Let N be the random variable representing the number of maintenance calls for a
single machine. We are given that
14
Let Y be the random variable representing the maintenance cost upon each call. Let X be the
aggregate maintenance cost for each machine ; so N is the primary random variable for X ,
and Y is its secondary distribution. Then
E(X) = E(N ) E(Y ) = (30)(80) = 240
Var(X) = E(N ) Var(Y ) + Var(N ) E(Y )2 = (3)(200)2 + (3)(80)2 = 139, 200
Let n be the number of machines. Let S be the aggregate maintenance cost. Then
The company will avoid purchasing the contracts if P (S > 1.2E(S)) < 0.1. Now:
( )
0.1 > P (S > 1.2E(S)) = P
SE(S)
>
0.2E(S)
Var(S) Var(S)
( )
(0.2)(240n)
= P Z>
139200 n
( )
= P Z> 48 n
139200 n
48 n
> (0.9) = 1.282 n > 99.3
139200 n
number of computers needed 100
Note. We recall that a X lognormal(, 2 ) is the one with support (0, ) and the density
function
(ln x)2
f (x) = 1 e 2 2
x 2 2
In fact , we have ln(X) N ( , 2 ). Some people take this as the definition of a lognormal
random variable: it is a r.v. whose logarithm is normal. One can easily calculate the raw
moments :
1
E(X k ) = exp(k + k 2 2 )
2
Especially
E(X) = exp( + 1 2 )
2
E(X 2 ) = exp(2 + 2 2 )
15
These two equalities are usually used to find and 2 ; see the example below.
Example . You are asked to price a Workers Compensation policy for a large employer.
The employer wants to buy a policy from your company with an aggregate limit of 150% of
total expected loss. You know the distribution for aggregate claims is Lognormal. You are
also provided with the following:
Calculate the probability that the aggregate loss will exceed the aggregate limit.
Solution.
E(S) = E(N )E(X) = (50)(4500) = 225, 000
Var(S) = E(N )Var(X) + Var(N )E(X)2 = (50)(3000)2 + (12)2 (4500) = 3, 366, 000, 000
Now we find the parameters of the lognormal distribution:
( )
exp + 12 2 = E(S) = 225, 000 + 12 2 = ln(225000) = 12.32
exp(2 + 2 2 ) = E(S 2 ) = Var(S) + E(S)2 = 3366000000 + (225000)2 2 + 2 2 =
( )
ln 3366000000 + (225000)2 = 24.71
Solving
these two equations gives us:
= 12.29
2 = 0.064 = 0.254
Then:
P (S > 1.5E(S)) = P (S > (1.5)(225000)) = P (S > 337500) = P (ln(S) > ln(337500)) =
( )
P ln(S)12.29
0.254 > ln(337500)12.29
0.254 = P (Z > 1.72) = 1 (1.72) = 0.043
Example (from the Finans guide - page 269) . You own a fancy light bulb factory.
Your workforce is a bit clumsy they keep dropping boxes of light bulbs. The boxes have
varying numbers of light bulbs in them, and when dropped, the entire box is destroyed. You
are given:
16
(i) Expected number of boxes dropped per month : 50
(ii) Variance of the number of boxes dropped per month: 100
(iii) Expected value per box: 200
(iv) Variance of the value per box: 400
You pay your employees a bonus if the value of light bulbs destroyed in a month is less than
8000.
Assuming independence and using the normal approximation, calculate the probability that
you will pay your employees a bonus next month.
Solution. Let
S = X1 + + XN
where N is the number of boxes destroyed in a month, and Xi is the value of the i-th box
destroyed.
This value was found by using the norm.dist function of Excel; in fact the command
= norm.dist(0.9975 , 0 , 1 , T RU E)
17
Using convolution to find the distribution of aggregate loss
from section 9.3
and fS (x).
n (n1)
(x k) fX (k)
FX (x) = FX
k=0
n (n1)
f
X (x) = fX (x) fX (k)
k=0
18
x
n (n1)
F (x) = FX (x y) fX (y) dy
X
0
x
(n1)
fX
n
(x) = fX (x y) fX (y) dy
0
Example for the discrete case (from the Finans guide - page 265). An insurance
portfolio produces N claims, where
n P (N = n)
0 0.5
1 0.2
2 0.2
3 0.1
x fX (x)
1 0.9
2 0.1
Individual claim amounts and N are mutually independent. Complete the following table:
x 0 (x)
fX 1 (x)
fX 2 (x)
fX 3 (x)
fX fS (x)
0
1
2
3
4
P (N = n)
Solution.
19
x 0 (x)
fX 1 (x)
fX 2 (x)
fX 3 (x)
fX fS (x)
0 1 0 0 0 0.5
1 0 0.9 0 0 0.18
2 0 0.1 0.81 0 0.182
3 0 0 0.18 0.729 0.0909
4 0 0 0.01 0.162 0.0182
P (N = n) 0.5 0.2 0.2 0.1
Example for the discrete case (from the Finans guide - page 266) . The number of
claims in a period has a geometric distribution with mean 4. The amount of each claim X
follows
P (X = x) = 0.25 ; x = 1 , 2 , 3 , 4
The number of claims and the claim amounts are independent. S is the aggregate claim
amount in the period. Calculate FS (3) .
Solution.
x 0 (x)
fX 1 (x)
fX 2 (x)
fX 3 (x)
fX fS (x)
0 1 0 0 0 0.2
1 0 0.25 0 0 0.04
2 0 0.25 0.0625 0 0.048
3 0 0.25 0.125 0.0156 0.0576
P (N = n) 0.2 0.16 0.128 0.1024
n 4n
P (N = n) = =
(1 + )n+1 5n+1
20
Now by looking at the last column we get:
Example for the continuous case (from the Finans guide - page 267). Severities
have a uniform distribution on [0 , 100]: The frequency distribution is given by
n Probability
0 0.60
1 0.30
2 0.10
Solution of (i).
x x
1 1 x
(just the ordinary distribution) FX (x) = f (t) dt = dt =
0 0 100 100
[ ]t=x
2
x
1
x
xt 1 1 1 1
FX (x) = FX (x t)f (t) dt = dt = (x t)2
= x2
0 0 100 100 10000 2 t=0 20000
Solution of (ii).
0 1
FS (x) = P (N = 0)FX (x) + P (N = 1)FX (x) + P (N = 2)FX2 (x)
1 1 3x x2
= 0.60 + (0.30)( x) + (0.10)( x2 ) = 0.60 + +
100 20000 1000 200000
Example (from the Finans guide - page 270) . The number of claims, N , made on an
insurance portfolio follows the following distribution:
21
n P (N = n)
0 0.7
2 0.2
3 0.1
If a claim occurs, the benefit is 0 or 10 with probability 0.8 and 0.2, respectively. The number
of claims and the benefit for each claim are independent. Calculate the probability that
aggregate benefits will exceed expected benefits by more than 2 standard deviations.
Solution.
Std(S) = 16.04 = 4
22
= P (S = 0 | N = 0)P (N = 0) + P (S = 0 | N = 2)P (N = 2) + P (S = 0 | N = 3)P (N = 3)
Then
Example (from the Finans guide - page 271) . For a collective risk model the number
of losses, N , has a Poisson distribution with = 20. The common distribution of the
individual losses has the following characteristics:
(i) E(X) = 70
(ii) E(X 30) = 25
(iii) P (X > 30) = 0.75
(iv) E(X 2 | X > 30) = 9000
An insurance covers aggregate losses subject to an ordinary deductible of 30 per loss.
Calculate the variance of the aggregate payments of the insurance.
Solution. We have
S = (X1 30)+ + + (XN 30)+
However:
E[(X 30)2+ ] = E[(X 30)2+ ] | X > 30]P (X > 30) + E[(X 30)2+ ] | X 30]P (X 30)
= 0.75E[(X 30)2+ ] | X > 30] + (0.25)(0) = 0.75 E[(X 30)2 ] | X > 30]
23
{ }
= 0.75E(X 2 60X + 900 | X > 30) = 0.75 E(X 2 | X > 30) 60 E(X | X > 30) + 900
{ }
= 0.75 E(X 2 | X > 30) 60 E(X 30 | X > 30) 1800 + 900
24
Stop Loss Insurance
from section 9.3
is called stop-loss insurance. The expected value E[(S d)+ ] is called net stop-loss
premium. As usual, we have
[1 FS (x)]dx = d (x d)fS (x) dx
d
E[(S d)+ ] =
FS (x)] =
d)fS (x)
x>d [1 x>d (x
Note. Although this equality is true in theory, but it is not so practical to use it because it is
dicult to determine the support of the random variable S and so we might not know the
upper limit of this integration (analyse this diculty for the next example). A practical way
of using this premium is to use the equality
n Probability
0 0.60
1 0.30
2 0.10
Solution.
25
E(X) = 50 E(N ) = (0)(60) + (1)(0.30) + (2)(0.10) = 0.5
3x x2
We already calculated that FS (x) = 0.60 + 1000 + 200000 . By dierentiating we have
3 x
fS (x) = 1000 + 100000 . Therefore
5( ) [ ]5
3 x 3 x2
E(S 5) = + dx + 5[1 FS (5)] = x+ + 5[1 FS (5)]
0 1000 100000 1000 200000 0
[ ( )]
3, 025 35 52
= + 5 1 0.60 + + = 1.9395
200, 000 1000 200000
Important Note. When X has continuous distribution we use the technique used in this
example, but when X has discrete distribution we use the recursive formula in part (i) of the
following theorem ; see the example following the theorem.
d
Proof. E[(S d)+ ] = [1 FS (x)]dx = [1 FS (x)]dx [1 FS (x)]dx
d a a
d
= [1 FS (x)]dx [1 FS (a)]dx
a a
26
So:
E[(Sa)+ ]E[(Sb)+ ]
1 FS (a) = ba
bd da
= E[(S a)+ ] + E[(S b)+ ]
ba ba
Proof. Take the two points a and b as in the previous theorem. The equality
E[(S b)+ ] = E[(S a)+ ] (b a)[1 FS (a)] shows that E[(S b)+ ] < E[(S a)+ ] for
a < b. Furthermore, for every a < d < b, the value of E[(S d)+ ] is found by taking the linear
interpolation of E[(S a)+ ] and E[(S b)+ ] , therefore the function f (t) = E[(S t)+ is
decreasing over the interval t [a, b].
Note. If X is any loss random variable, then by taking S = X, we notice the above theorems
hold for a single loss variable too. Specially, the function f (t) = E[(X t)+ ] is decreasing.
Example (from the Finans guide - page 288) . For a collective risk model:
(i) The number of losses has a Poisson distribution with = 2.
(ii) The common distribution of the individual losses is:
27
x f (x)
1 0.6
2 0.4
Solution.
Next Step:
FS (0) = P (S = 0) = P (N = 0) = e2
( )
e2 2
fS (1) = P (S = 1) = P (N = 1 , X = 1) = P (N = 1) P (X = 1) = (0.6) = (1.2)e2
1!
fS (2) = P (S = 2) = P (N = 1 , X2 = 2) + P (N = 2 , X1 = 1 , X2 = 1)
e2 2 e2 22
= (0.4) + (0.6)2 = e2 (0.8) + 2e2 (0.6)2
1! 2!
{ } { }
= 1.8 + 3e2 + 2e2 (1.2) + e2 (0.8) + 2e2 (0.6)2 3
28
= 1.2 + 6.92 e2
Example. The number of claims in a period has a geometric distribution with mean 4. The
amount of each claim X follows
P (X = x) = 0.25 ; x = 1 , 2 , 3 , 4
The number of claims and the claim amounts are independent. S is the aggregate claim
amount in the period. Calculate E[(S 2)+ ] and E[(S 1.6)+ ] .
Solution. We had:
x 0 (x)
fX 1 (x)
fX 2 (x)
fX 3 (x)
fX fS (x)
0 1 0 0 0 0.2
1 0 0.25 0 0 0.04
2 0 0.25 0.0625 0 0.048
3 0 0.25 0.125 0.0156 0.0576
P (N = n) 0.2 0.16 0.128 0.1024
E[(S 2)+ ] = E[(S 1)+ ] [1 FS (1)] = E(S) [1 FS (1)] = 9.2 [1 0.24] = 8.44
29
Note. When S is discrete and its support consists of jumps of step h, the following theorem
gives a formula for E[(S a)+ ] for all a being in the support, and then using the above
interpolation formula one can find all other values of E[(S a)+ ] for a not being in the
support.
Theorem. If the support of S consists of the values {0 , h , 2h , 3h , ...} then for every jh in
the support set we have
{
}
E[(S a)+ ] = h 1 FS ((n + j)h) = h sum of all survival values starting from jh
n=0
Note. This theorem is not of practical use and it might be used for theoretical discussions.
30
Panjer Recursive Formulas
section 9.6
Calculating the convolutions to find fS (x) or FS (x) is a dicult matter. The Panjers
recursive formula is an ecient way of calculating the density of S.
Theorem. Consider a collective risk model in which the severity random variable has support
in the set of non-negative integer values
X {0 , 1 , 2 , ...}
Theorem. If in the above theorem the frequency distribution is in (a, b, 0), then we have
p1 (a + b)p0 = 0 , and therefore the recursive formula reduces to
n ( )
k=1 a + n k fX (k) fS (n k)
b
fS (n) =
1 a fX (0)
n
fS (n) = kfX (k) fS (n k)
n
k=1
Note. If in the above two theorems we have X {0, 1, ..., m}, then the corresponding
formulas reduce to
(
min{n,m} )
{p1 (a + b)p0 }fX (n) + a + nb k fX (k) fS (n k)
k=1
fS (n) =
1 a fX (0)
31
where in the case of (a, b, 0) we have p1 (a + b)p0 = 0 .
x P (X = x)
1 0.5
2 0.3
3 0.2
(iii) The probabilities for certain values of the aggregate claims, S, are:
x P (S = x)
2 0.1438
3 0.1198
5 0.0294
Determine P (S = 4).
Solution. We use the recursive formula although it can be done directly too.
0.8
n n
fS (n) = kfX (k) fS (n k) = kfX (k) fS (n k)
n n
k=1 k=1
P (S = 0) = P (N = 0) = e0.8 = 0.4493
Example (Problem 9.47 of the textbook) . Aggregate claims are compound Poisson
with = 2, fX (1) = 14 , and fX (2) = 43 . For a premium of 6, an insurer covers aggregate
claims and agrees to pay a dividend (a refund of premium) equal to the excess, if any, of 75%
32
of the premium over 100% of the claims. Determine the excess of premium over expected
claims and dividends. Use the recursive formulas.
So
D {4.5 , 3.5 , 2.5 , 1.5 , 0.5}
2
n n
fS (n) = kfX (k) fS (n k) = kfX (k) fS (n k)
n n
k=1 k=1
fS (0) = P (N = 0) = e2
33