Home Page

Risk theory
Title Page

The insurance as a tool to mitigate risks: Contents

JJ II
the model of individual risks J I

Page 1 of 36
José Daniel López-Barrientos, PhD
Go Back

Spring 2022 Full Screen

Close

Quit
Contents

Home Page
1 Definition of the model of individual risks 3
Title Page

2 Features of the model of individual risks 14

Contents

2.1 Actuaries do it with frequency and severity! . . . . . . . . . . . . 15

JJ II

3 Three advantages of... 19 J I

3.1 The model of individual risks . . . . . . . . . . . . . . . . . . . . . 19 Page 2 of 36

3.2 The model of collective risks . . . . . . . . . . . . . . . . . . . . . 20 Go Back

Full Screen
4 An actuarial example 29
4.1 Equal ages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Close

4.2 Different ages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Quit

1. Definition of the model of individual risks
Theorem 1.1. (Central Limit Theorem. See [3, Theorem 2.5.1] or [6, Theorem Home Page

II.2].) Let X1 , X2 , ... be a sequence of independent and identically distributed random Title Page

variables with common mean µ, and finite variance σ2 > 0; and let
Contents

X
n
Sn : = Xi . JJ II
i =1
J I
Then
Page 3 of 36
  Zx
Sn − nµ 1 z2
P √ ≤x →√ e− 2 dz, for − ∞ < x < ∞ as n → ∞. (1.1) Go Back
σ n 2π −∞
Full Screen

Close

Quit
Definition 1.2. [1, Section 2.1] Let Xi be the loss associated with a claim from the i-th
policy in a portfolio with n contracts. We define the loss random variable of the portfolio

Home Page
Sn : = X1 + · · · + Xn .
Title Page

Contents

JJ II

J I

Page 4 of 36

Go Back

Full Screen

Close

Quit
Example 1.3. [1, Exercise 3.15] Consider a random survivorship group consisting of
two subgroups:

1. The survivors of a group of 1600 newborns; Home Page

Title Page
2. The survivors of a group of 540 10-year old children.
Contents
An excerpt from the appropriate mortality table for both subgroups follows.
JJ II
(x) `x
J I
0 40
Page 5 of 36
10 39
Go Back
70 26
If Y1 and Y2 are the numbers of survivors to age 70 from subgroups 1 y 2, respectively, Full Screen

estimate the number c such that P(Y1 + Y2 > c) = 0.05. Assume the lives are Close

independent. Quit
Solution. Let X := Y1 + Y2 . By Theorem 1.1, X follows a Normal distribution
with mean EY1 + EY2 , and variance varY1 + varY2 . since Y1 and Y2 follow the
26
and 540, 26



binomial law with parameters 1600, 40 39 , respectively, we have Home Page

that:
Title Page

26 26
EX = 1600 × + 540 × Contents
40 39
= 1400, JJ II
26 14 26 13 J I
varX = 1600 × × + 540 × ×
40 40 39 39
= 484. Page 6 of 36

Go Back
X−1400
Observe that Z := √ is a Normal standard random variable. We are asked
484
Full Screen
c−1400
to find c such that √
484
= 1.644853627. Then c = 1436.18678. This completes
Close
the exercise.
Quit
Example 1.4. [5, Example 9.1] A person might lose his job with a probability of 10%,
and when this happens, the loss raises to 5000$. The costumer is willing to pay a
premium of 550$ for insurance. How many contracts of this type should the Fondo Home Page

Mivivienda sell to have a 95% chance of actually making money?

Title Page

Contents

JJ II

J I

Page 7 of 36

Go Back

Full Screen

Close

Quit
Solution. Let n be the number of sold contracts, and let N be a Binomial random
variable with parameters (n, 0.1). Thus, the total claim is 5000N. This means
that EN = 0.1 × n, and varN = 0.1 × 0.9 × n. The exercise requests to find a Home Page

value of n such that P(5000N < 550n) = 0.95. Therefore,

Title Page

0.95 = P(5000N < 550n) = P( N < 0.11n

 )  √ 
Contents

N − EN 0.11n − 0.1n 0.01 n

= P √ <√ =P Z< √ . JJ II
varN 0.1 × 0.9 × n 0.09
√ J I
Theorem 1.1 tells us that we need to find n such that √ n
0.01
= 1.645. This yields
0.09
Page 8 of 36
n = 2434.989109.
Go Back

Full Screen

Close

Quit
Example 1.5. Fondo Mivivienda issues one-year credit default insurance for 1$ and
2$. The claim probabilities for the year of coverage are 0.02 and 0.1. The following table
specifies how many people nk are there in each category, depending on the benefit bk , Home Page

and the claim probability qk , for k = 1, 2, 3, 4.

Title Page

k qk bk nk Contents

1 0.02 1 500 JJ II
2 0.02 2 500
J I
3 0.1 1 300
Page 9 of 36
4 0.1 2 500
Go Back

Fondo Mivivienda wishes to charge an amount which equals the 95th percentile of
Full Screen
the distribution of aggregate claims for this population of 1800 people. Compute such
Close
percentile.
Quit
Solution. Let X j be the claim of the j-th person. We need to find c such that
P(S ≤ c) = 0.95, where S = X1 + X2 + · · · + X1800 . We compute ES and varS:

Mean Variance Home Page

k qk bk bk q k bk2 qk (1 − qk ) nk Title Page

1 0.02 1 0.02 0.0196 500 Contents

2 0.02 2 0.04 0.0784 500

JJ II
3 0.1 1 0.1 0.09 300
J I
4 0.1 2 0.2 0.36 500
Page 10 of 36
Then
Go Back
X
1800 X
4
ES = EX j = nk bk qk = 160, Full Screen
j=1 k=1

X
1800 X
4 Close

varS = varX j = nk bk2 qk (1 − qk ) = 256.

Quit
j=1 k=1
To use Theorem 1.1, we standardize and find:
c − 160
√ = 1.645. (1.2)
256
Home Page
√
This yields c = 1.645 × 256 + 160 = 186.32.
Title Page

Example 1.6. (Example 1.5 continued.) The central administration bureau of Fondo Contents

Mivivienda orders that the premium paid by the j-th individual is (1 + θ ) · EX j , where
JJ II
X j stands for the claim of the j-th person, and θ > 0. Calculate θ.
J I
Solution. From the former answer we know that ES = 160 and varS = 256. On
Page 11 of 36

the other hand, the new information tells us that c = (1 + θ )ES. Substituting
Go Back
in (1.2) and solving for θ gives us θ = 0.1645. This means that a person from
Full Screen
the first group will pay 1 × 0.02 × 1.1645 = 0.0239, one from the second group
Close
will end up paying 2 × 0.02 × 1.1645 = 0.04658, one from the third group pays
1 × 0.1 × 1.1645 = 0.11645, and a person from the fourth group pays 0.2329. Quit
Exercises and extensions 1.7. a. How many contracts are to be sold in Example 1.4
if Fondo Mivivienda wishes to have a 97.5% chance of making a profit? How
many for a 99% chance? Home Page

b. The density function of an individual loss Y is given by Title Page


0.02 1 − y
if 0 < y < 100,
Contents

100
f (y) =
 0 in other case.
JJ II

J I
The amount Z that Fondo Mivivienda will pay in the case of a claim is of only
Page 12 of 36
80% of the part of the claim that overtakes a deductible of 10$.
Go Back

(b1) What is the expression that defines Z in terms of Y?

Full Screen

(b2) Find the expected value of Z.

Close

Quit
c. Fondo Mivivienda has followed for months the claim process of life insurance in
Peru, and estimates that the expected value and the standard deviation of the
claim number N are 6.7 and 2.3, respectively. Analogously, the individual claims Home Page

have an expected value of 179,247, and a standard deviation of 52,141.

Title Page

(c1) Determine the expected value and the variance of the aggregate claims Contents

X
N JJ II
S= Xi .
1
J I

Assume that ES = EX · EN, and varS = (EX )2 varN + EN · varX. Page 13 of 36

Go Back
(c2) Calculate the probability that the aggregate claims exceed in 40% its ex-
Full Screen
pected value. Use only a Normal approximation.
Close
d. Read [6] and solve the exercises therein.
Quit
2. Features of the model of individual risks
• It focuses on the distribution function of the total claim amount S for the Home Page

portfolio of an insurer. Title Page

• It assumes that the risks in a portfolio are independent random variables Contents

with finite mean an variance. JJ II

J I
• It considers S as the sum of a large number of random variables, hence the
use of the Central Limit Theorem 1.1 is natural. Page 14 of 36

Go Back
• An alternative to the MIR is the model of collective risks (see [8]).
Full Screen

Close

Quit
2.1. Actuaries do it with frequency and severity!

Definition 2.1. Let Xi be the loss associated with the i-th claim from a portfolio with
Home Page
n contracts. The loss random variable of the portfolio is defined as
Title Page

S : = X1 + X2 + · · · + X N , Contents

where N is a random variable supported on non-negative integers, and N is independent JJ II

of the sequence of i.i.d. random variables ( X j : j = 0, ..., N ). J I

Page 15 of 36

Go Back

Full Screen

Close

Quit
Theorem 2.2. The hypotheses of the model of collective risks give that

ES = EXEN.
Home Page

Proof. The theorem of complete probability (see [7, Theorem I.3.29]) yields that Title Page

ES = E[E(S|N )]. Now,

Contents

X
E[E(S|N )] = E [S|N = n] p(n) JJ II
n
X X
" N # J I
= E Xi |N = n p(n)
Page 16 of 36
n 1
X X
n X
" #
Go Back
= E Xi p ( n ) = E[nX ] p(n)
n 1 n
X X Full Screen

= EXnp(n) = EX np(n)
Close
n n

If follows that ES = EXEN. Quit

The very same result ([7, Theorem I.3.29]) implies that
∞
X
P( S ≤ s ) = P(S ≤ s|N = n) p(n)
Home Page
n=0
X∞ X
n ∞
X
!
= P Xk ≤ s p(n) = P( X1 + · · · + Xn ≤ s ) p ( n ). Title Page

n=0 k=1 n=0 Contents

JJ II

J I

Page 17 of 36

Go Back

Full Screen

Close

Quit
Example 2.3. a. Show how to use the model of collective risks to solve Example 1.4.

b. What model would you choose for each of the following situations?
Home Page
(i) A group of workers with different gender, age and death benefit.
Title Page

(ii) A reinsurance contract paying medical malpractices exceeding certain amount.

Contents

(iii) A dental plan for a person who goes to the dentist twice a year. Each contract
JJ II
covers families of different sizes at the same price.
J I
Solution. a. S = X1 + · · · + X N , where P( X = 5000) = 1, and N is Bin(n, 0.1).
Page 18 of 36

b. (i) Model of individual risks. Go Back

(ii) Model of collective risks. Full Screen

(iii) Model of collective risks. Close

Quit
3. Three advantages of...

Home Page
3.1. The model of individual risks
Title Page
• Its hypotheses are general enough.
Contents

• Its use is straightforward and can be implemented with fairly reasonable

JJ II
resources.
J I

• It amounts for the features of a heterogeneous group with common finite Page 19 of 36

mean and variance. Go Back

Full Screen

Close

Quit
3.2. The model of collective risks

It is easy to model losses from data with many policy features.

Home Page

Example 3.1. [4, Example 6.3] A house can suffer a random number of accidents in Title Page

a period. The insurer pays 80% of the losses that exceed a 1,000$ deductible with a
Contents

maximum payment of 100,000$. All the claims in excess of 50,000$ are re-insured.
JJ II

(a) Describe the total loss of the person before any payment takes place. J I

Page 20 of 36
(b) Describe the total loss for the insurer before the payment from the reinsurance.
Go Back
(c) Describe the aggregate loss for the re-insurer.
Full Screen

(d) Describe the aggregate loss for the insurer after the payment of the reinsurance. Close

(e) Describe the aggregate loss of the insured. Quit

Solution. (a) The aggregate loss for the person any payment is given by

S a = X1 + · · · + X N .
Home Page

(b) The aggregate payments to the insured are given by Title Page

Contents
Sb = Y1 + · · · + YN ,
JJ II
where: J I


 0 if X j ≤ 1000,

 Page 21 of 36

Yj := 100000
0.8( X j − 1000) if 1000 < X j ≤ + 1000 = 126000,

 0.8 Go Back

 100000 if X j > 126000.
Full Screen

Close

Quit
(c) The aggregate payments made by the re-insurer are represented by

Sc = Z1 + · · · + ZN , where


 0 if X j ≤ 50000
+ 1000 = 63500,
Home Page

 0.8
Title Page
Zj := 0.8( X j − 63500) if 63500 < X j ≤ 126000,



 Contents
50000 if X j > 126000.
JJ II

J I

Page 22 of 36

Go Back

Full Screen

Close

Quit
(d) The aggregate cost for the insurer after the payment of the re-insurer is
given by Sd = W1 + · · · + WN , where


 0 if X j ≤ 1000,
Home Page


Wj := 0.8( X j − 1000) if 1000 < X j ≤ 63500, Title Page



 50000 if X j > 63500. Contents

JJ II

J I

Page 23 of 36

Go Back

Full Screen

Close

Quit
(e) The aggregate cost for the insured is given by Se = V1 + · · · + VN , where:


 Xj if X j ≤ 1000,


Vj := X j − 0.8X j + 800 = 800 + 0.2X j if 1000 < X j ≤ 126000, Home Page



 X − 100000
j if X j > 126000. Title Page

Contents

JJ II

J I

Page 24 of 36

Go Back

Full Screen

Close

Quit
 The models developed for covered and uncovered losses of policyholders,
insurers, and re-insurers are mutually consistent. This is useful for an insurer
studying the effect of changing deductibles (liability limits, coinsurance, etc.) Home Page

Example 3.2. [4, Exercise 6.1.1.1] Show that in Example 3.1, the costs for the policy- Title Page

holder (e), the insurer (d), and the re-insurer (c) add-up to the original loss (a). Contents

JJ II
Solution. Adding-up the results from (c)-(e) in Example 3.1 yields:
 J I

 Xj if X j ≤ 1000,



 Page 25 of 36
 0.8( X j − 1000) + 800 + 0.2X j if 1000 < X j ≤ 63500,
X j :=

 0.8( X j − 63500) + 50000 + 800 + 0.2X j if 63500 < X j ≤ 126000,
Go Back




 50000 + 50000 + X − 100000 if X > 126000.
Full Screen
j j
Close

Quit
 Home Page

Title Page

Contents

JJ II

Now, according to Example 3.1(a), J I

Page 26 of 36
S a = X1 + · · · + X N .
Go Back

This completes the result.

Full Screen

Close

Quit
 It is straightforward to study the impact of changing deductibles, limits of
responsibility or coinsurance.

Home Page
Example 3.3. [5, Example 3.4] Use the model of collective risks to describe the total
amount paid for the damages caused to a house. Consider a 250$ deductible. Title Page

Contents
Solution 1. Let N L be the total number of accidents (including those that do not
JJ II
exceed the deductible of 250$). Define the left-censored severity random variable:
J I

 Page 27 of 36
 0 if X ≤ 250,
Y L := Go Back
X − 250 if X > 250.
Full Screen

Close
The random variable of aggregate loss is S1 = Y1L + Y2L + ... + YNL L .
Quit
Solution 2. Let N P be the number of payments made after the claims. Define
the excess of loss severity random variable:

Home Page


Title Page

 Unknown if X ≤ 250,
Y P :=
Contents

 X − 250 if X > 250.

JJ II

J I

The random variable of aggregate loss is S2 = Y1P + Y2P + ... + YNP P . Page 28 of 36

Go Back

Full Screen

Close

Quit
4. An actuarial example

Home Page
4.1. Equal ages
Title Page
You are given a portfolio with 60 whole-life insurance contracts for 40-year old
Contents
people with a benefit of 1000$. Use the inverse transform and Monte Carlo
simulation techniques (recall [2, Algoritmo 2] and take a look here) to find the JJ II

least k such that the probability of ruin is less than 5% if the company charges J I
√
A 1x:ω−x + k varZ to each costumer. Use table M MEX 82-89(M), and assume a Page 29 of 36

discount rate of 3%. Go Back

Full Screen

Close

Quit
 Observe that, by the results in [6], it is reasonable to approximate the answer
of this exercise by assuming that the sum of the claims has a Normal distribu-
tion. We strongly recommend the reader to compare such an approximation Home Page

with the one we obtain by simulating each person’s moment of failure (see Fig-
Title Page

ures 1-2). Which of these approaches do you prefer? What are the advantages
Contents
and disadvantages of each one of them?
JJ II

J I

This function returns a simulated vector of Page 30 of 36

This function simulates the years until fail-

present values of 1$ payable at the end of the Go Back
ure.
year of failure for one person.
Full Screen

Figure 1: A couple of functions that simulate years and present values, respec- Close

tively for one person. Quit

This function returns a simulated matrix of present values of 1$ payable Home Page

at the end of the year of failure for a population of people the same age.
Title Page

Contents

JJ II

J I

Page 31 of 36

Go Back

This function computes P(ruin) by means of the Monte Carlo technique. Full Screen

Figure 2: A couple of functions that compute present values for a population, Close

and the probability of ruin for the exercise in section 4.1. Quit
4.2. Different ages

Repeat the exercise in section 4.1 relaxing the hypothesis that the ages of the
Home Page
policies in the insurance portfolio are equal. Keep the assumption on the terms
Title Page
of coverage of each insurance policy. See the figure below, and Figure 3.
Contents

JJ II

J I

Page 32 of 36

Go Back

Full Screen

Close

Quit
 Home Page

This function generates k q x for k = 0, 1, ..., ω − x. Title Page

Contents

JJ II

J I

Page 33 of 36

Go Back

This function computes P(ruin) by means of the Full Screen

Monte Carlo technique.

Close

Figure 3: A step-by-step solution to the exercise in section 4.2. Note that the Quit

ages of all the members in the population are not necessarily equal.
References
[1] Bowers, N. L., Gerber, H. U., Hickman, J. C., Jones, D. A., and Nesbitt, Home Page

C. J. Actuarial Mathematics. The Society of Actuaries, 1997. Available here. Title Page

[2] Cano-Ramos, A. E., and López-Barrientos, J. D. ¿“¡Te lo aseguro!”, o “¡te Contents

lo apuesto!”?, he ahí el dilema... Miscelánea Matemática 73 (2022). JJ II

[3] Kaas, R., Goovaerts, M., Dhaene, J., and Denuit, M. Modern Actuarial J I

Risk Theory: Using R. Springer Berlin Heidelberg, 2008. Available here. Page 34 of 36

Go Back
[4] Klugman, S., Panjer, H., and Willmot, G. Loss Models. From Data to Deci-
sions. Wiley, 2004. Available here. Full Screen

Close
[5] Klugman, S. A., Panjer, H., and Willmot, G. E. Loss Models. From Data to
Quit
Decisions. Wiley, 2012.
[6] López-Barrientos, J. D., Silva-Urrutia, E., and Lemus-Rodríguez, E.
A not so marble-ous use of the convolution formula (with a Normal
moral). In Investigación estocástica y estadística en la educación, J. D. Zacarías- Home Page

Flores, H. J. Reyes-Cervantes, F. Solanbo-Tajonar-Sanabria, V. H. Vázquez-

Title Page

Guevara, B. Juárez-Hernández, H. A. Cruz-Suárez, F. Velasco-Luna, and

Contents
G. D. Salgado-Suárez, Eds. Benemérita Universidad Autónoma de Puebla,
JJ II
Puebla, Puebla, 2021, ch. 1, pp. 1–11.
J I
[7] Mood, A. M., and Boes, F. A. G. D. C. Introduction to the theory of statistics.
Page 35 of 36
McGraw Hill, New York, 1974. Available here.
Go Back

[8] Verrall, R. The individual Risk Model: A compound distribution. Journal Full Screen

of the Institute of Actuaries 116 (1989), 101–107.

Close

Quit
Thank you for your attention!
Contact information Home Page

José Daniel López Barrientos, Ph.D. Title Page

Profesor-Investigador
Contents
Facultad de Ciencias Actuariales
JJ II
Universidad Anáhuac México Tel. 01-52-55-56-27-0210 ext. 8506
J I
[email protected]
Page 36 of 36
https://www.researchgate.net/profile/Jose_Daniel_Lopez-Barrientos
Go Back

Full Screen

Close

Quit