(Oct. 13, Part 1) Individual Risk Model

Nonlife Actuarial Models
Chapter 3
Aggregate-Loss Models
Learning Objectives
• Individual risk model
• Collective risk model
• De Pril recursion
• Compound process for collective risk
• Approximation methods
• Stop-loss reinsurance
2
3.1 Individual Risk and Collective Risk Models
(1) Individual risk model:
• The number of policies in the block is n. We assume the loss of

each policy, denoted by Xi , for i = 1, · · · , n, to be independently
and identically distributed as X. The aggregate loss of the block of
policies, denoted by S, is then given by
S = X1 + · · · + Xn . (3.1)
• X follows a mixed distribution with a probability mass at point zero.
3
(2) Collective risk model:
• Let N be the number of losses in the block of policies, and Xi be

the amount of the ith loss, for i = 1, · · · , N.
• The aggregate loss S is given by
S = X1 + · · · + XN . (3.2)
• N is a random variable, and N and X are assumed to be indepen-

dent.
• S has a compound distribution.
4
Remarks:
• There are advantages in modeling the claim frequency and claim

severity separately, and then combine them to obtain the aggregate-
loss distribution.
• For example, expansion of insurance business may have impacts on

the claim frequency but not the claim severity.
• Cost control (or general cost increase) and innovation in technology

may a ect the claim severity with no e ects on the claim frequency.
5
3.2 Individual Risk Model
• The basic equation of the individual risk model is
S = X1 + · · · + Xn . (3.1)
• As n is fixed, the mean and variance of S are given by
E(S) = n E(X) and Var(S) = n Var(X). (3.3)
• X is a mixed-type distribution with prob mass at point 0.
• Let the probability of a loss be and the probability of no loss be

1 .
• When there is a loss, the loss amount is Y , which is a positive

continuous random variable with mean Y and variance Y2 .
6
• Thus, X = Y with probability , and X = 0 with probability 1 .
We can write X as
X = IY, (3.4)
where I is a Bernoulli random variable distributed independently of
Y , so that
(
0, with probability 1 ,
I= (3.5)
1, with probability .
• Thus, the mean of X is
E(X) = E(I)E(Y ) = Y, (3.6)
• Using equation (A.118) in Appendix A.11, we have
7
Var(X) = Var(IY )
= [E(Y )]2 Var(I) + E(I 2 )Var(Y )
2 2
= Y (1 )+ Y. (3.7)
• Equations (3.6) and (3.7) can be plugged into equation (3.3) to ob-
tain the mean and variance of S.
• Example 3.1: Assume there is a chance of 0.2 that there is a

claim. When a claim occurs the loss is exponentially distributed
with parameter = 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.
8
• Solution: The mean and variance of the loss in a loss event is
1 1
Y = = = 2,
0.5
and
2 1 1
Y =2
= = 4.
0.52
Thus, the mean and variance of the loss incurred by a random policy
are
E(X) = (0.2)(2) = 0.4,
and
Var(X) = (2)2 (0.2)(1 0.2) + (0.2)(4) = 1.44.
The mean and variance of the aggregate loss are
E(S) = (500)(0.4) = 200,
9
and
Var(S) = (500)(1.44) = 720.
3.2.1 Exact distribution of S using convolution:
• We first consider the distribution of X1 + X2 , where X1 and X2 are

both continuous.
• The pdf of X1 + X2 is given by the 2-fold convolution

Z x Z x
2
f (x) = fX1 +X2 (x) = f1 (x y)f2 (y) dy = f2 (x y)f1 (y) dy.
0 0
(3.8)
• The pdf of X1 + · · · + Xn can be calculated recursively. Suppose

the pdf of X1 + · · · + Xn 1 is given by the (n 1)-fold convolution
f (n 1) (x), then the pdf of X1 + · · · + Xn is the n-fold convolution
10
given by
Z x
n (n 1)
f (x) = fX1 + ··· + Xn (x) = f (x y)fn (y) dy
0
Z x
(n 1)
= fn (x y)f (y) dy. (3.9)
0
• Now we consider the case where Xi are mixed-type random variables.
• We assume that the pf-pdf of Xi is given by

(
1 i, for x = 0,
fXi (x) = (3.10)
i fYi (x), for x > 0,
in which fYi (·) are well defined pdf of positive continuous random
variables.
• The df of X1 + X2 is given by the 2-fold convolution in the Stieltjes-
11
integral form, i.e.,
Z x
2
F (x) = FX1 +X2 (x) = FX1 (x y) dFX2 (y)
0
Z x
= FX2 (x y) dFX1 (y). (3.11)
0
• The df of X1 + · · · + Xn can be calculated recursively.
• Example 3.2: For the block of insurance policies defined in Ex-

ample 3.1, approximate the loss distribution by a suitable discrete
distribution. Compute the df FS (s) of the aggregate loss of the
portfolio for s from 110 through 300 in steps of 10, based on the
discretized distribution.
12
• Solution: We approximate the exponential loss distribution by a
discrete distribution taking values 0, 1, · · · , 10. As the df of E( ) is
FX (x) = 1 exp( x), we approximate the pf by
fX (x) = (0.2) {exp [ (x 0.5)] exp [ (x + 0.5)]} , for x = 1, · · · , 9,
with
fX (0) = 0.8 + (0.2) {1 exp [ 0.5 ]} ,
and
fX (10) = (0.2) exp [ 9.5 ] .
The discretized approximate pf of the loss is
13
Table 3.1: Discretized probabilities
x fX (x) x fX (x)
0 0.8442 6 0.0050
1 0.0613 7 0.0031
2 0.0372 8 0.0019
3 0.0225 9 0.0011
4 0.0137 10 0.0017
5 0.0083
• Using the convolution method, the df of the aggregate loss S for

selected values of s is given in Table 3.2.
14
Table 3.2: The df of S by convolution
s FS (s) s FS (s)
110 0.0001 210 0.7074
120 0.0008 220 0.8181
130 0.0035 230 0.8968
140 0.0121 240 0.9465
150 0.0345 250 0.9746
160 0.0810 260 0.9890
170 0.1613 270 0.9956
180 0.2772 280 0.9984
190 0.4194 290 0.9994
200 0.5697 300 0.9998
15
3.2.3 Approximation of the Individual Risk Model:
• When n is large, by virtue of the Central Limit Theorem, S is ap-
proximately normal.
• The (exact) mean and variance of S are given in equations (3.3),

(3.6) and (3.7), which can be used to compute the approximate
distribution of S.
• Thus,
Ã !
S E(S) s E(S)
Pr(S s) = Pr p p
Var(S) Var(S)
Ã !
s E(S)
' Pr Z p
Var(S)
Ã !
s E(S)
= p . (3.21)
Var(S)
16

(Oct. 13, Part 1) Individual Risk Model

Uploaded by

Copyright:

Available Formats

(Oct. 13, Part 1) Individual Risk Model

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Oct. 13, Part 1) Individual Risk Model

Uploaded by

Copyright:

Available Formats

Nonlife Actuarial Models

• Individual risk model

• Collective risk model

• Compound process for collective risk

(1) Individual risk model:

• The number of policies in the block is n. We assume the loss of

• X follows a mixed distribution with a probability mass at point zero.

• Let N be the number of losses in the block of policies, and Xi be

• The aggregate loss S is given by

• N is a random variable, and N and X are assumed to be indepen-

• S has a compound distribution.

• There are advantages in modeling the claim frequency and claim

• For example, expansion of insurance business may have impacts on

• Cost control (or general cost increase) and innovation in technology

• The basic equation of the individual risk model is

• As n is ﬁxed, the mean and variance of S are given by

E(S) = n E(X) and Var(S) = n Var(X). (3.3)

• X is a mixed-type distribution with prob mass at point 0.

• Let the probability of a loss be and the probability of no loss be

• When there is a loss, the loss amount is Y , which is a positive

• Thus, the mean of X is

E(X) = E(I)E(Y ) = Y, (3.6)

• Using equation (A.118) in Appendix A.11, we have

• Example 3.1: Assume there is a chance of 0.2 that there is a

E(S) = (500)(0.4) = 200,

3.2.1 Exact distribution of S using convolution:

• We ﬁrst consider the distribution of X1 + X2 , where X1 and X2 are

• The pdf of X1 + X2 is given by the 2-fold convolution

• The pdf of X1 + · · · + Xn can be calculated recursively. Suppose

• Now we consider the case where Xi are mixed-type random variables.

• We assume that the pf-pdf of Xi is given by

• The df of X1 + X2 is given by the 2-fold convolution in the Stieltjes-

• The df of X1 + · · · + Xn can be calculated recursively.

• Example 3.2: For the block of insurance policies deﬁned in Ex-

fX (x) = (0.2) {exp [ (x 0.5)] exp [ (x + 0.5)]} , for x = 1, · · · , 9,

• Using the convolution method, the df of the aggregate loss S for

• The (exact) mean and variance of S are given in equations (3.3),

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.