******FINAL EXAM

CHAPTER 3
QUANTIFYING
RISKS

Mahdzan & Boey 2015 1

Learning Outcomes
Review and apply the concepts from probability
and statistics to risk management.
Demonstrate how pooling independent risks
lowers risk.
Explain how correlated losses increase risk.

Mahdzan & Boey 2015 2

 1.0 Basic Concepts from
Probability and Statistics
Probability -The chances of an uncertain event
happening, in view that there are more than one
possible outcome.
A random variable (X): The outcome is unknown and is
a result of a random occurence.
Classified as:
a) Discrete random variables black or white discrete

b) Continuous random variables height, weight

possible - no mitigation, no protection

probable - some prevention, protective measure that can reduce possibility of losses

Mahdzan & Boey 2015 3

a) Discrete random variable
Takes the form of a specific value, such as 0,1,2,3,... and so
on.
E.g.: A die tossed can produce a finite number of outcomes,
that is, six possible outcomes (1,2,3,4,5,or 6).

Mahdzan & Boey 2015 4

b) A continuous random variable
Take on an infinite (unlimited) number of possible values.
E.g.:
1. weight and height of people,
2. amount of rainfall in a month,
3. income of the population

Mahdzan & Boey 2015 5

A probability distribution: lists all possible outcomes and
the probability that each outcome will occur.
E.g.: die tossing - each outcome has an equal chance of
occuring, hence, a one sixth (1/6) probability. The total
probability of all outcomes occuring is 1

Mahdzan & Boey 2015 6

The probability distribution is shown in Table
1: each outcome has an equal probability of
occuring.
outcome
X Probability of heads appearing
1 1/6 = 0.166..
2 1/6 = 0.166..
3 1/6 = 0.166..
4 1/6 = 0.166..
5 1/6 = 0.166..
6 1/6 = 0.166..
Total 1.00

Table 1: Possible outcomes (X) for a die tossed

Mahdzan & Boey 2015 7

Example for non-equal probability of outcomes:
The chances of accidental losses to a motorcyle.

X Probability
RM0 0.50
RM500 0.30
RM1,000 0.10
RM2,500 0.05
RM5,000 0.03
RM10,000 0.02
Total 1.00

Table 2: Possible outcomes (X) for a motorcycle accidental losses

Mahdzan & Boey 2015 8

• Maximum possible loss: the maximum dollar amount of
losses in the worst case scenario.

 Maximum probable loss: the estimated loss that occur

when there are protective or safeguard measures in place.

Mahdzan & Boey 2015 9

 A normal probability distribution:

 The cummulative probabilities of a continuous random

variable.
 Shaped like a bell curve. like normal distribution

 The law of large numbers: when an experiment is performed a

large number of times, the average results will approach the
expected value.
mean

Fig 1: A normal probability distribution

less risky, that will be very

much random event that

would cost the insurance

standard deviation

Mahdzan & Boey 2015 10

 Skewness: The lack of asymmetry of a probability distribution.
 If skewness :
i. Zero - normal distribution
ii. Negative - to the left
iii. Positive - to the right

Fig 2: Negative and positive skewness

Mahdzan & Boey 2015 11

Characteristics of Probability
distributions
Expected value mean

The expected value of a real-valued random

variable indicates the central distribution of the
variable.
To compute the expected value:
Multiply each possible outcome by its
probability, and add up the results.

Mahdzan & Boey 2015 12

 i ii iii
E.g.: Possible Outcome Probability Exp. Loss (RM)
𝑛
(RM) (pi)
𝑥𝑖 𝑝𝑖
(xi)
𝑖=1

0 0.50 0 x 0.50 =0
500 0.30 500 x 0.30 = 150
1,000 0.10 1,000 x 0.1 =100
2,500 0.05 2,500 x 0.05 =125
5,000 0.03 5,000 x 0.05 = 250
10,000 0.02 10,000 x 0.02 = 200
Total 1.0 825

Table 3: Probability distribution and expected loss for a motorcycle accident

The mathematical function of the expected value can be expressed as:

[Link] = x1p1 + x2p2 + x3p3 +........ + xnpn
𝑛
= 𝑖=1 𝑥𝑖 𝑝𝑖

Exp. Value of X = (0)(0.5) + (500 x 0.3) + (1000)(0.1) + (2500) (0.05) + (5000)(0.05)

+ (10000) (0.02)
= RM825
Mahdzan & Boey 2015 13
Variance: the spread of outcomes around the expected value.
Used to measure risk – a low variance shows low risk while a high
variance shows high risk.

Standard deviation: square root of variance and is more

commonly used.

Mahdzan & Boey 2015 14

 𝑁
Variance = 𝑖=1 𝑝𝑖 (𝑥𝑖 − 𝜇)2

Where,
N = the total number of outcomes
pi= the probability of the outcomes
xi = the possible outcomes
𝜇= the expected value.

𝑁
Standard deviation = 𝑖=1 𝑝𝑖 (𝑥𝑖 − 𝜇)2

Mahdzan & Boey 2015 15

Example 1:
Consider a loss with three possible outcomes: RN500, RM1,000
and RM1,500 with probability of 0.25, 0.50 and 0.25
respectively. First, calculate the expected value, 𝜇 (also shown
in column iii in Table 4).

𝑛
[Link] (𝜇 )= 𝑖=1 𝑥𝑖 𝑝𝑖
= x1p1 + x2p2 + x3p3 +........ + xnpn
= 500 (0.25) + 1000 (0.50) + 1500 (0.25)
= 125 + 500 + 375
= 1000
Mahdzan & Boey 2015 16
Next, compute the variance and standard deviation
(workings for variance is shown in column iv – vi of Table
4).

Table 4: Computation of variance and standard deviation (Example 1)

i ii iii iv v vi
Possible Probability
Outcomes 𝒙𝒊 𝒑𝒊 𝒙𝒊 − 𝝁 (𝒙𝒊 − 𝝁)𝟐 𝒑𝒊 (𝒙𝒊 − 𝝁)𝟐
xi pi
RM500 0.25 125 500-1,000 = (-500) 250,000 0.25 x 25,000 = 62,500
RM1,000 0.50 1,000 – 1,000 = 0 0 0.50 x 0 = 0
500
RM1,500 0.25 375 1,500 -1,000 = 500 250,000 0.25 x 25,000 = 62,500

1,000 125,000

average losses that have paid by the insurance company to

𝑛
Exp. Value = 𝑖=1 𝑥𝑖 𝑝𝑖 = 1000 policyholder

𝑁
Variance = 𝑖=1 𝑝𝑖 (𝑥𝑖 − 𝜇)2 = 125,000

𝑁 distribution of risk, risk proxy

Std Deviation = 𝑖=1 𝑝𝑖 (𝑥𝑖 − 𝜇)2 = 353.55

Mahdzan & Boey 2015 17

 2.0 Pooling risks
Risk pooling - Used by insurance companies to spread risk over a
large group of individuals facing similar risks
The concept of risk pooling:
 The cost of individuals with higher risk (the less skillful
swimmers) are offset by those with lower risk (the more skillful
ones).
 The combined risks of all individuals in a pool creates a lower
probability that an adverse situation will occur.

Mahdzan & Boey 2015 18

Example 1:
Two men, Billy and Bully, are exposed to the risk of accident.
Assume that the probability of an accident occuring is 10% with a
loss of RM5,000, and that the probability of no accident occuring
is 90%. The chance of losses for these two men are uncorrelated.
Under normal circumstances where they each bear the
consequences of their own loss, the expected loss is shown in
Table 5.
Possible Outcome Possible Outcome (RM) Probability
xi pi
No accident RM0 0.9
Accident RM5,000 0.1

Table 5: Probability distribution of Billy and Bully independently

Mahdzan & Boey 2015 19

We first compute expected loss, variance and standard deviation of
Billy and Bully independently.
Table 6: Computation of expected loss, variance and standard deviation (Example 1)

i ii iii iv v vi
Possible Probability
Outcomes
(RM) pi 𝒙𝒊 𝒑𝒊 𝒙𝒊 − 𝝁 (𝒙𝒊 − 𝝁)𝟐 𝒑𝒊 (𝒙𝒊 − 𝝁)𝟐
xi
0 0.9 0 0 – 500 = (-500) 250,000 0.9 x 250,000 = 225,000
0.1 x 20,750,000 =
5,000 0.1 500 5,000 – 500 = 4,500 20,250,000
2,075,000

500 2,250,000

𝑛
Exp. loss = 𝑖=1 𝑥𝑖 𝑝𝑖 = 500
𝑁
Variance = 𝑖=1 𝑝𝑖 (𝑥𝑖 − 𝜇)2 = 2,250,000

𝑁
Std. Deviation = 𝑖=1 𝑝𝑖 (𝑥𝑖 − 𝜇)2 = 1,500

Mahdzan & Boey 2015 20

 bear the risk/loss tgt

Now, Say that Billy and Bully decide to pool their risks. The
probability distribution changes, as shown in Table 7.

Table 7: Probability distribution of Billy and Bully after pooling risks

i ii iii iv v vi vii viii

Possible Pooled Loss born Pooled Exp. Loss

outcomes Loss by each Probability
(RM) person (RM) 𝒙𝒊 − 𝝁 (𝒙𝒊 − 𝝁)𝟐 𝒑𝒊 (𝒙𝒊 − 𝝁)𝟐
xipi
xi pi
No accident - Billy
No accident - Bully
0 0 (0.9)(0.9) = 0.81 0 -500 250000 202500
Accident – Billy
No accident – Bully
5000 2500 (0.1)(0.9) = 0.09 225 2000 4000000 360000
No accident – Billy
Accident – Bully
5000 2500 (0.1)(0.9) = 0.09 225 2000 4000000 360000
Accident – Billy
Accident – Bully
10000 5000 (0.1)(0.1) = 0.01 50 4500 20250000 202500
500 1125000

𝑛
Exp. Loss= 𝑖=1 𝑥𝑖 𝑝𝑖 = 500

𝑁
Variance = 𝑖=1 𝑝𝑖 (𝑥𝑖 − 𝜇)2 = 1125000

𝑁
Std. Deviation = 𝑖=1 𝑝𝑖 (𝑥𝑖 − 𝜇)2 = 1060.66

Mahdzan & Boey 2015 21

 Pooling risks that are correlated

Correlated risks: A single event that result in a simulatenous

occurence of a high frequency of large losses.
Correlated losses imply that the chances of extreme losses occuring
to all participants in a pool is greater than if losses were
uncorrelated.
A perfectly positive correlation: if one of the individuals experience
losses, the other two will also experience loss.

Mahdzan & Boey 2015 22