General Probability

Outline

PROBABILITY FOR ACTUARIAL SCIENCE

1 General Probability
Nguyễn Văn Thìn
2 RANDOM VARIABLES
BỘ MÔN XÁC SUẤT THỐNG KÊ
KHOA TOÁN - TIN HỌC
TRƯỜNG ĐẠI HỌC KHOA HỌC TỰ NHIÊN [Link]

Ngày 25 tháng 2 năm 2026

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 1 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 2 / 112

RANDOM VARIABLES RANDOM VARIABLES

Outline 2. RANDOM VARIABLES

1 General Probability

2 RANDOM VARIABLES
Let S be a sample space with a collection of events1 M. Any
2.1 Cumulative Distribution and Probability Density Functions (measurable2 ) function X : S → R is called a random variable.
2.2 Applications in Insurance and Actuarial Science
2.3 Transformations of a Random Variable
2.4 Expectation of Random Variables
2.5 Percentiles – Mode – Standardized Random Variable
2.6 Frequently Discrete Distribution
2.7 Frequently Continuous Distribution

1
In fact, M must be a σ-algebra on S.
2
The measurability of X implies that (X ≤ x) ∈ M so that P (X ≤ x) is defined,
for all x ∈ R.
Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 3 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 4 / 112
 RANDOM VARIABLES Cumulative Distribution and Probability Density Functions RANDOM VARIABLES Cumulative Distribution and Probability Density Functions

Cumulative Distribution and Probability Density Functions Cumulative Distribution and Probability Density Functions
Let P be a probability on M and let X be a random variable (on S). The
probability of the event (X ≤ x) = {s ∈ S : X (s) ≤ x} defines the
cumulative distribution function FX (x), cdf in brief, of X , (iii) lim FX (x) = 0 and lim FX (x) = 1.
x→−∞ x→∞
(iv) The left-hand side limit of FX exists at each point a ∈ R,
FX (x) = P (X ≤ x) ≡ P ({s ∈ S : X (s) ≤ x}) .
FX (a−) ≡ limFX (x) ≤ FX (a) .
We have the following properties of the cumulative distribution function. x↑a

(i) FX : R → R is an increasing function, i.e.

(v) For all a, b ∈ R,
FX (x) ≤ FX (y ) whenever x ≤ y P (a < X ≤ b) ≡ P ({x ∈ S : a < X (s) ≤ b}) = FX (b) − FX (a) and
P (X = a) = P ({x ∈ S : X (s) = a}) = FX (a) − FX (a−).
. In some actuarial model, the function
(ii) FX is continuous from the right-hand side on R, i.e. sX (x) = P (X > x) ≡ P ({s ∈ S : X (s) > x}) = 1 − FX (x) is called the
survival function of X.
limFX (x) = FX (a) , for all a ∈ R.
x↓a

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 5 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 6 / 112

RANDOM VARIABLES Applications in Insurance and Actuarial Science RANDOM VARIABLES Applications in Insurance and Actuarial Science

Random Variables in Insurance and Actuarial Science Random Variables in Insurance and Actuarial Science
Uncertainty and Potential Loss Role of Insurance – Actuarial Science

Over a given period, individuals face the risk of unpredictable losses Insurance allows individuals to:
Exchange an uncertain loss X
Let X denote the potential loss, modeled as a random variable
For a fixed and known cost called the premium
The exact amount of loss is unknown in advance This transfer of risk provides financial stability
The potential loss X may arise from: Definition
Automobile accidents Actuarial science is the discipline that studies uncertain events and their
Fire, theft, storms, or hurricanes at home economic consequences.
Premature death of the head of household
Hospitalization due to illness Actuaries assess the fair price of insurance contracts
Probability and statistics are essential tools to:
Common Feature
Measure risk
All these losses are uncertain in both occurrence and magnitude.
Manage uncertainty

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 7 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 8 / 112
 RANDOM VARIABLES Applications in Insurance and Actuarial Science RANDOM VARIABLES Applications in Insurance and Actuarial Science

Cumulative Distribution and Probability Density Functions Cumulative Distribution and Probability Density Functions

X a1 a2 ··· an ···
Pr p1 p2 ··· pn ···
We pay attention on the following cases.
Case 1. X is a discrete random variable, i.e. X (S) is an at most countable We have P
set. In other word, X can take finite values a1 < a2 < · · · < an or (i) 0 ≤ pk ≤ 1, for all k, and k pk = 1. P
countable values a1 < a2 < · · · < an < · · ·. In this case, the function (ii) For all D ⊂ R, P (X ∈ D) ≡ P ({s ∈ S : X (s) ∈ D}) = an ∈D pn . In
fX : R → R, defined by fX (x) = P (X = x), is called the probability particular, X X
density function, pdf in brief, or probability mass function, pmf in brief, of FX (x) = pn ≡ fX (t) ,
the random variable X. In this case, let pn = P (X = an ), we have an ≤x t≤x

 which shows that FX is a step function,

pn , x = an , for some n,
fX (x) =
0, otherwise.

 0,Pk
x < a1
FX (x) = i=1 pi , ak ≤ x < ak+1 , for some k,
Hence, the pdf of X can be identified by the following probability table
1, x > an , for all n.


Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 9 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 10 / 112

RANDOM VARIABLES Applications in Insurance and Actuarial Science RANDOM VARIABLES Applications in Insurance and Actuarial Science

Cumulative Distribution and Probability Density Functions Cumulative Distribution and Probability Density Functions

Solution:
Example 1 (Discrete Random Variable (Claim Count)) ................................... ...................................
.......................................................................
Returning to our auto insurance company, we now model the number of
.......................................................................
claims filed by a policyholder during a policy period. Let N denote the
number of claims. .......................................................................
Based on historical data for a standard policyholder (Question 544, SOA .......................................................................
Sample Questions), the probability distribution is: .......................................................................
n 0 1 2 3 .......................................................................
P(N = n) 0.4 0.3 0.2 0.1 .......................................................................
.......................................................................
Objective: Find and interpret the cumulative distribution function FN (n).
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 11 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 12 / 112
 RANDOM VARIABLES Applications in Insurance and Actuarial Science RANDOM VARIABLES Applications in Insurance and Actuarial Science

Cumulative Distribution and Probability Density Functions Cumulative Distribution and Probability Density Functions

Example 2
Case 2. X is a continuous random variable, i.e. X (S) is an interval of R
and FX is a continuous function on R. In this case, we have When an accident occurs, the loss amount X (in thousands) is modeled.
P (X = a) = 0, for all a ∈ R, and there exists a function fX : R → R, For partial damage to a car (Question 51, SOA Sample Questions), the
called the probability density function, pdf in brief, of X such that density function is:
(
0.5003e −x/2 , 0 < x < 15
Z
P (X ∈ D) ≡ P ({s ∈ S : X (s) ∈ D}) = fX (x) dx, fX (x) =
D 0, otherwise
for all (measurable) subset D of R.R ∞We have (Note: 0.5003 ≈ 1/(2(1 − e −7.5 )) for proper normalization).
(i) fX (x) ≥ 0, for all x ∈ R, and −∞ fX (x) dx = 1. Objective: Find the probability that a loss is between $2,000 and $4,000
Rx
(ii) FX (x) = −∞ fX (t) dt, for all x ∈ R. In particular, FX′ (x) = fX (x), (i.e., 2 < x < 4).
for all point x ∈ R so that fX is continuous at x.

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 13 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 14 / 112

RANDOM VARIABLES Applications in Insurance and Actuarial Science RANDOM VARIABLES Applications in Insurance and Actuarial Science

Cumulative Distribution and Probability Density Functions Cumulative Distribution and Probability Density Functions

Solution:
................................... ...................................
....................................................................... Case 3. X is a mixed random variable, i.e. it is a mixture of a discrete and
....................................................................... continuous distribution. In this case, its cdf is a piecewise continuous
....................................................................... function on R. Formally, there exist a discrete pdf fX1 (x), a continuous
pdf fX2 (x), and 0 < p < 1 such that fX (x) = pfX1 (x) + (1 − p) fX2 (x),
....................................................................... for all x ∈ R. It is noted that FX is discontinuous at each point x ∈ R so
....................................................................... that P (X1 = x) > 0 and in this case, the “jump” of FX at x is given by
.......................................................................
FX (x) − FX (x−) = p · P (X1 = x) .
.......................................................................
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 15 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 16 / 112
 RANDOM VARIABLES Applications in Insurance and Actuarial Science RANDOM VARIABLES Transformations of a Random Variable

Cumulative Distribution and Probability Density Functions Transformations of a Random Variable

Example 3 (Frequency–Severity Model)

Let X be a random variable with pdf fX (x) and let g : R → R be a
( (measurable) function. We consider the random variable Y = g (X ).
0, if Ā occurs, If X is discrete, we have
Y =
X , if A occurs.
fY (y ) = P (Y = y ) = P (g (X ) = y ) = P X ∈ g −1 (y )


X
A: event of an accident with p probability, = fX (x)
X : the loss amount in the event of an accident, x∈g −1 (y )
Y : the total claim payment.
Objective: Find the cdf of Y . In the case that X is continuous, we have
(i) If g is strictly increasing, then
Solution:
FY (y ) = P (Y ≤ y ) = P (g (X ) ≤ y ) = P X ≤ g −1 (y ) = FX g −1 (y ) ,



................................... ...................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 17 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 18 / 112

RANDOM VARIABLES Transformations of a Random Variable RANDOM VARIABLES Expectation of Random Variables

Transformations of a Random Variable Expectation of Random Variables

Let X be a random variable with pdf fX (x) and g : R → R be a

(measurable) function. The expectation of the random variable
and if g is strictly decreasing, then
g ◦ X ≡ g (X ) is defined by
FY (y ) = P (Y ≤ y ) = P (g (X ) ≤ y ) = P X ≥ g −1 (y ) = 1−FX g −1 (y ) = sX g −1 (y ) .




X
E [g (X )] = g (x) fX (x)
(ii) If g is an one-to-one differentiable function, then x

1 for discrete random variable X,

fY (y ) = fX g −1 (y ) ·


.
|g ′ (g −1 (y ))| Z ∞
E [g (X )] = g (x) fX (x) dx
−∞

for continuous random variable X, and

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 19 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 20 / 112
 RANDOM VARIABLES Expectation of Random Variables RANDOM VARIABLES Expectation of Random Variables

Expectation of Random Variables Expectation of Random Variables

for mixed random variable with fX (x) = pfX1 (x) + (1 − p) fX2 (x), where
fX1 (x) is a discrete pdf, fX2 (x) is a continuous pdf, and 0 < p < 1,

X Z ∞
E [g (X )] = p g (x) fX1 (x) + (1 − p) g (x) fX2 (x) dx.
x −∞ Example 4 (Expectation (Mean Loss))
Using the loss severity distribution from Example 2: fX (x) = 0.5003e −x/2
In particular, with g (x) = x, E (X ) is called the expectation of X, which is
for 0 < x < 15 (in thousands).
also called the mean of X, denoted by µX . In general, with g (x) = x n ,
Objective: Calculate the expected loss E (X ).
n ∈ N, the expectation of g (X ) = X n , E (X n ), is called the nth-moment
of X.

Specially, if X ≥ 0, then
Z ∞ Z ∞
n
E (X ) = sX (x)dx, E (X ) = nx n−1 sX (x)dx.
0 0

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 21 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 22 / 112

RANDOM VARIABLES Expectation of Random Variables RANDOM VARIABLES Expectation of Random Variables

Expectation of Random Variables Expectation of Random Variables

Solution:
................................... ................................... Letting g (x) = (x − µX )2 when µX exists, the expectation of
g (X ) = (X − µX )2 is called the variance of X, denoted by Var (X ),
.......................................................................
.......................................................................
h i
Var (X ) = E (X − µX )2 .
.......................................................................
....................................................................... The square root of the variance of X is called the standard deviation of X,
....................................................................... denoted by σX , p
σX = Var (X ).
.......................................................................
....................................................................... Coefficient of variation of X is σX /µX .
....................................................................... In general, with g (x) = (x − µX )n , n ∈ N, when µX exists, the
expectation of g (X ) = (X − µX )n , E ((X − µX )n ), is called the
....................................................................... nth-central moment of X.

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 23 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 24 / 112
 RANDOM VARIABLES Expectation of Random Variables RANDOM VARIABLES Expectation of Random Variables

Expectation of Random Variables Expectation of Random Variables

For g (x) = min(x, d) and any non-negative random variable X , (iv) MX (t) may not exist for all values of t and if it exists on an open
Z dh interval containing 0, we have
i Z d
E[min(X , d)] = 1 − FX (x) dx = sX (x)dx. (1) dn
0 0 dt n MX (t) t=0
= E (X n ) , for n = 0, 1, 2, · · ·

. In particular, MX (0) = 1, MX′ (0) = E (X ), and MX′′ (0) = E X 2 .

With g (x) = e tx , t ∈ R, the expectation of g (X ) = e tX defines a function
in t, called the moment generating function MX of X, mgf in brief, (v) If MX (t) is analytic on an open interval containing 0, we have
∞
E (X n )
 
MX (t) = E e tX .
X
MX (t) = tn.
n!
n=0
For random variables X, Y, we have
(i) E (aX + bY ) = aE
(X ) + bE (Y ) for any numbers a, b ∈ R. (vi) Let X, Y be two random variables such that MX (t) = MY (t) on an
(ii) Var (X ) = E X 2 − (E (X ))2 . open interval containing 0. X and Y must have the same must have
(iii) Var (aX + b) = a2 Var (X ) for any a, b ∈ R. identical distributions.

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 25 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 26 / 112

RANDOM VARIABLES Expectation of Random Variables RANDOM VARIABLES Expectation of Random Variables

Expectation of Random Variables Expectation of Random Variables

Solution:
................................... ...................................
.......................................................................
Example 5 (Moment Generating Function)
.......................................................................
The loss amount X (given an accident) has pdf
.......................................................................
1 −x/3000 .......................................................................
fX (x) = e , x > 0.
3000 .......................................................................
Objective: Derive the moment generating function MX (t) and use it to .......................................................................
find E (X ) and E (X 2 ). .......................................................................
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 27 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 28 / 112
 RANDOM VARIABLES Expectation of Random Variables RANDOM VARIABLES Expectation of Random Variables

Expectation of RV in Insurance and Actuarial Science Expectation of RV in Insurance and Actuarial Science
Large and Potentially Unlimited Losses Loss Sharing Between Parties

Coverage modifications imply loss sharing between:

In some situations, the potential loss may be: The policyholder (insured)
Very large The insurance company (insurer)

In detail,
Even theoretically unlimited
Policyholder bears part of the loss
Full insurance coverage in such cases may be very costly Insurer bears the remaining portion
Result: reduced insurance premium
To reduce the burden of insurance cost, coverage can be modified.
Three Common Modifications Behavioral Effects
1 Deductibles Loss sharing may influence policyholder behavior
Example: With deductibles, drivers may be more careful
2 Policy limits (caps)

3 Coinsurance
Insurance contracts may combine:
Deductibles + Limits + Coinsurance

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 29 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 30 / 112

RANDOM VARIABLES Expectation of Random Variables RANDOM VARIABLES Expectation of Random Variables

Expectation of RV in Insurance and Actuarial Science Expectation of Random Variables in Insurance and
Excess-of-Loss Insurance Contract Actuarial Science
Insurer Payment
Excess-of-Loss Insurance Contract
An excess-of-loss insurance contract specifies a deductible d
The insurer reimburses losses only beyond the amount d We denote the insurer payment by XI .
The deductible d is pre-specified in the contract
XI = X − min(X , d)
Loss Sharing with Deductible
Equivalently,
Let X denote the total loss. (
0, if X ≤ d,
Policyholder pays the first d units of loss XI = (16.0.5)
X − d, if X > d.
Insurance company pays the excess, if any

Loss Allocation Remark

Policyholder payment = min(X , d) XI represents the portion of loss paid by the insurer.

Insurer payment = X − min(X , d)

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 31 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 32 / 112
 RANDOM VARIABLES Expectation of Random Variables RANDOM VARIABLES Expectation of Random Variables

Expectation of Random Variables in Insurance and Expectation of Random Variables in Insurance and
Actuarial Science Actuarial Science
Expected Insurer Payment Expected Insurer Payment

The expected payment of the insurer is

E(XI ) = E(X ) − E[min(X , d)] (16.0.6)
The quantity E[min(X , d)] is called the limited expected value.
Example 6 (Deductible (Exponential Distribution))
Limited Expected Value
An insurance policy has a deductible of 3. Losses X under the policy has
For any non-negative random variable X ,
pdf f (x) = 0.1e −0.1x , x > 0. (Question 337, SOA).
Z dh i Objective: Calculate the expected claim payment net of the deductible.
E[min(X , d)] = 1 − FX (x) dx (16.0.7)
0

Interpretation
This integral represents the expected loss up to the deductible level d.

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 33 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 34 / 112

RANDOM VARIABLES Expectation of Random Variables RANDOM VARIABLES Expectation of Random Variables

Expectation of Random Variables in Insurance and Expectation of Random Variables in Insurance and
Actuarial Science Actuarial Science
Expected Insurer Payment Expected Insurer Payment

Solution:
................................... ...................................
.......................................................................
Example 7 (Finding the Deductible)
.......................................................................
Losses under an insurance policy are uniformly distributed on the interval
.......................................................................
[0, 100]. A deductible is set so that the expected claim payment of losses
....................................................................... net of the deductible is 32 (Question 336, SOA).
....................................................................... Objective: Calculate the deductible d.
.......................................................................
.......................................................................
.......................................................................
.......................................................................
Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 35 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 36 / 112
 RANDOM VARIABLES Expectation of Random Variables RANDOM VARIABLES Expectation of Random Variables

Expectation of Random Variables in Insurance and Expectation of Random Variables in Insurance and
Actuarial Science Actuarial Science
Expected Insurer Payment Policy Limits, coinsurance

Solution: Policy Limits

............................................. .............................................
...........................................................................................
Limit L: insurer pays at most L.
...........................................................................................
........................................................................................... XI = min(X , L)
...........................................................................................
........................................................................................... Expected payment:
........................................................................................... E(XI ) = E[min(X , L)]
...........................................................................................
........................................................................................... Coinsurance
...........................................................................................
........................................................................................... Coinsurance rate k:
........................................................................................... XI = kX , 0<k ≤1
...........................................................................................
...........................................................................................
........................................................................................... E(XI ) = k E(X )
Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 37 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 38 / 112

RANDOM VARIABLES Expectation of Random Variables RANDOM VARIABLES Expectation of Random Variables

Expectation of Random Variables in Insurance and Expectation of Random Variables in Insurance and
Actuarial Science Actuarial Science
Policy Limits, coinsurance Policy Limits, coinsurance

Solution:
...................................... ......................................
.............................................................................
Example 8 (Policy Limit) .............................................................................
Losses X under a boat insurance policy has cdf .............................................................................
F (x) = 1 − e −(ln 2)x/400 , x > 0. For each loss, the claim payment is the .............................................................................
amount of the loss, up to a maximum of 1000 (Question 563, SOA). .............................................................................
Objective: Calculate the expected claim payment. .............................................................................
.............................................................................
.............................................................................
.............................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 39 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 40 / 112
 RANDOM VARIABLES Expectation of Random Variables RANDOM VARIABLES Expectation of Random Variables

Expectation of Random Variables in Insurance and Expectation of Random Variables in Insurance and
Actuarial Science Actuarial Science
Policy Limits, coinsurance Policy Limits, coinsurance

Solution:
................................... ...................................
Example 9 (Coinsurance) .......................................................................
A construction worker experiences a back injury resulting in a loss under .......................................................................
an insurance policy. Losses under the policy are uniformly distributed on .......................................................................
the interval [2, 10]. The insurer reimburses the worker for 60% of this loss
.......................................................................
(Question 504, SOA).
Objective: Calculate the 30th percentile of the construction worker’s .......................................................................
unreimbursed loss. .......................................................................
.......................................................................
.......................................................................
.......................................................................
Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 41 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 42 / 112

RANDOM VARIABLES Expectation of Random Variables RANDOM VARIABLES Expectation of Random Variables

Expectation of Random Variables in Insurance and Expectation of Random Variables in Insurance and
Actuarial Science Actuarial Science
Combination of Coverages Combination of Coverages

Deductible d, limit L, coinsurance k: Example 10 (Combination of Coverages)

 An insurance policy reimburses a loss up to a benefit limit of 10. The
0,
 X ≤d policyholder’s loss, Y , follows a distribution with density fY (y ) = 2y −3 for
XI = k(X − d), d < X ≤ L y > 1 (Question 50, SOA). Now suppose the policy is modified with:

k(L − d), X > L

A deductible of d = 2
h i A coinsurance factor of k = 0.8
E(XI ) = k E(min(X , L)) − E(min(X , d)) A policy limit of L = 8 (applied after deductible and coinsurance)
Objective: Find the expected insurer payment.

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 43 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 44 / 112
 RANDOM VARIABLES Expectation of Random Variables RANDOM VARIABLES Expectation of Random Variables

Expectation of Random Variables in Insurance and Expectation of Random Variables in Insurance and
Actuarial Science Actuarial Science
Combination of Coverages Revisit Frequency–Severity Model

Solution: Frequency–Severity Model

...................................... ......................................
(
0, if Ā occurs,
............................................................................. Y =
X , if A occurs.
.............................................................................
............................................................................. A: event of an accident with p probability,
............................................................................. X : the loss amount in the event of an accident,
............................................................................. Y : the total claim payment.
............................................................................. Cdf
............................................................................. FY (y ) = (1 − p) + pFX (y ), y ≥ 0.
.............................................................................
Mean and Variance
.............................................................................
E(Y ) = pµX , Var(Y ) = p(1 − p)µ2X + pσX2
Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 45 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 46 / 112

RANDOM VARIABLES Percentiles – Mode – Standardized Random Variable RANDOM VARIABLES Percentiles – Mode – Standardized Random Variable

Percentiles – Mode – Standardized Random Variable Percentiles – Mode – Standardized Random Variable

Example 11 (Percentiles and Median)

The 100p-th percentile xp , 0 < p < 1, satisfies both of the following An insurance policy reimburses a loss up to a benefit limit of 10. The
inequalities : P (X ≤ xp ) ≥ p and P (X ≥ xp ) ≥ 1 − p. In the case of policyholder’s loss, Y , follows a distribution with density function
continuous distribution, we have FX (xp ) = P (X ≤ xp ) = p. The 50-th (Question 50, SOA):
percentile is called the median. (
The mode is a point x at which the function fX (x) is maximized. 2y −3 , y > 1
fY (y ) =
Let X be a random variable with mean µ and standard deviation σ. The 0, otherwise.
expression X σ−µ is called the standardized version of X.
This is a continuous distribution with a heavy tail, but it is not one of the
"common" distributions (exponential, normal, uniform, etc.) covered later.
Objective: Find the median loss.

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 47 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 48 / 112
 RANDOM VARIABLES Percentiles – Mode – Standardized Random Variable RANDOM VARIABLES Percentiles – Mode – Standardized Random Variable

Percentiles – Mode – Standardized Random Variable Percentiles – Mode – Standardized Random Variable

Solution:
................................... ...................................
.......................................................................
....................................................................... Example 12 (Percentiles and Policy Limits)
....................................................................... Continuing with the loss variable Y from Example 11, which has pdf
....................................................................... fY (y ) = 2y −3 , y > 1. The policy has a benefit limit of 10.
....................................................................... Objective: Find the 90th percentile of the reimbursement paid, given that
....................................................................... a loss occurs.
.......................................................................
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 49 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 50 / 112

RANDOM VARIABLES Percentiles – Mode – Standardized Random Variable RANDOM VARIABLES Percentiles – Mode – Standardized Random Variable

Percentiles – Mode – Standardized Random Variable Percentiles – Mode – Standardized Random Variable

Solution:
................................... ................................... Example 13 (Mode of a Distribution)
.......................................................................
Damages to a car in a crash are modeled by a random variable with
....................................................................... density function (Question 119, SOA):
....................................................................... 
....................................................................... c(4 − x),
 2<x <4
....................................................................... f (x) = c(x − 4), 4<x <6

0, otherwise

.......................................................................
....................................................................... where c is a constant. This is a triangular distribution symmetric about
....................................................................... x = 4, not one of the standard distributions covered later.
....................................................................... Objective: Find the mode of this distribution.

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 51 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 52 / 112
 RANDOM VARIABLES Percentiles – Mode – Standardized Random Variable RANDOM VARIABLES Percentiles – Mode – Standardized Random Variable

Percentiles – Mode – Standardized Random Variable Percentiles – Mode – Standardized Random Variable

Solution:
................................... ...................................
.......................................................................
....................................................................... Example 14 (Mode and Percentiles)
....................................................................... Continuing with the damage distribution from Example 13. A particular
....................................................................... car is insured with a deductible of 2. This car was involved in a crash with
resulting damages in excess of the deductible.
....................................................................... Objective: Find the median damage given that the damage exceeds the
....................................................................... deductible (2).
.......................................................................
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 53 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 54 / 112

RANDOM VARIABLES Percentiles – Mode – Standardized Random Variable RANDOM VARIABLES Percentiles – Mode – Standardized Random Variable

Percentiles – Mode – Standardized Random Variable Percentiles – Mode – Standardized Random Variable

Solution:
................................... ...................................
.......................................................................
Example 15 (Standardized Random Variable)
.......................................................................
....................................................................... Return to the discrete claim count distribution from Example 1:
....................................................................... P(N = 0) = 0.4, P(N = 1) = 0.3, P(N = 2) = 0.2, P(N = 3) = 0.1.
.......................................................................
Objective: Compute the mean µ, standard deviation σ, and the
.......................................................................
standardized value for a policyholder who files 3 claims.
.......................................................................
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 55 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 56 / 112
 RANDOM VARIABLES Percentiles – Mode – Standardized Random Variable RANDOM VARIABLES Percentiles – Mode – Standardized Random Variable

Percentiles – Mode – Standardized Random Variable Percentiles – Mode – Standardized Random Variable
Solution:
...................................... ...................................... Example 16 (Standardization and Percentiles)
.............................................................................
For the loss distribution Y with pdf fY (y ) = 2y −3 , y > 1, we previously
.............................................................................
computed:
.............................................................................
Z ∞ Z ∞
............................................................................. E (Y ) = y · 2y −3 dy = 2y −2 dy = 2,
............................................................................. Z1 ∞ Z1 ∞
............................................................................. 2
E (Y ) = 2 −3
y · 2y dy = 2y −1 dy = ∞.
............................................................................. 1 1
............................................................................. Thus, Y has a mean of 2 but an infinite variance!
............................................................................. Objective: Explain why the standardized version Y σ−µ is not defined, and
what this implies.

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 57 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 58 / 112

RANDOM VARIABLES Percentiles – Mode – Standardized Random Variable RANDOM VARIABLES Frequently Discrete Distribution

Percentiles – Mode – Standardized Random Variable Frequently Discrete Distribution

Solution: 1. Bernoulli Distribution.

................................... ...................................
X = 0, 1,
.......................................................................
....................................................................... fX (1) = P (X = 1) = p and fX (0) = P (X = 0) = 1 − p.
.......................................................................
We have E (X ) = p, Var (X ) = p (1 − p), and
.......................................................................
....................................................................... MX (t) = (1 − p) + pe t .
.......................................................................
....................................................................... A typical Bernoulli distribution comes from Bernoulli trial in which a
random experiment is realized. Given an event A with P (A) = p. Let
.......................................................................
X = 1 for the “success” of the trial, i.e. A occurs, and let X = 0 for
....................................................................... the “failure”, i.e. A does not occur.

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 59 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 60 / 112
 RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

Solution:
................................... ...................................
Example 17 (Bernoulli Distribution (Accident Occurrence))
.......................................................................
Returning to our auto insurance company, consider a single policyholder in
a given year. From the company’s data (Question 18, SOA), the .......................................................................
probability of an accident for a driver aged 31-65 is 0.02. .......................................................................
Define the random variable: .......................................................................
.......................................................................
(
1, if the policyholder has at least one accident
X = .......................................................................
0, otherwise
.......................................................................
Objective: Identify the distribution of X and compute its mean and .......................................................................
variance.
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 61 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 62 / 112

RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

2. Uniform Distribution.

X = 1, 2, · · · , n,
Example 18 (Discrete Uniform Distribution)
1
fX (x) = n, for x = 1, 2, · · · , n. An auditor at the insurance company randomly selects one claim file from
a batch of 80 claims for review. The claims are numbered 1 through 80.
We have E (X ) = n+1 n2 −1 Each claim is equally likely to be selected.
2 , Var (X ) = 12 , and
Let X be the claim number selected.
et e nt −1
Objective: Identify the distribution of X and find the probability that the
MX (t) = n · e t −1 . selected claim number is between 20 and 30 (inclusive).

A typical uniform distribution comes from random experiment that all

outcomes have the same probability of realization.

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 63 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 64 / 112
 RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

Solution: 3. Binomial Distribution.

................................... ...................................
X = 0, 1, · · · , n,
.......................................................................
....................................................................... 
n

....................................................................... fX (x) = p x (1 − p)n−x , for x = 0, 1, · · · , n,
x
.......................................................................
....................................................................... E (X ) = np, Var (X ) = np (1 − p), and
.......................................................................
MX (t) = ((1 − p) + pe t )n .
.......................................................................
....................................................................... A typical Binomial distribution happens when we repeat a Bernoulli
....................................................................... trial n times independently and let X be the number of successes.

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 65 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 66 / 112

RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

Solution:
................................... ...................................
.......................................................................
Example 19 (Binomial Distribution)
.......................................................................
The insurance company writes four new policies for adults earning their
.......................................................................
first driver’s license. From the company’s pool, 20% are high-risk drivers
(Question 31, SOA). Let X be the number of high-risk drivers among .......................................................................
these four new policyholders. .......................................................................
Objective: Find the probability that exactly two of these four are high-risk .......................................................................
drivers. .......................................................................
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 67 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 68 / 112
 RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

4. Geometric Distribution. A typical Geometric distribution arises when we repeat independently
a Bernoulli trial and let X be the number of trials to get the first
success. If Y is the number of failures until the first success, then
X = 1, 2, · · ·, Y = X − 1, E (Y ) = 1−p 1−p p
p , Var (Y ) = p 2 , and MY (t) = 1−(1−p)e t .

fX (x) = (1 − p)x−1 p, for x = 1, 2, · · ·,

1−p An additional important property of the random variables X, Y is that

E (X ) = p1 , Var (X ) = p2
, and
it is memoryless (or it has the lack of memory property) : it does not
remember how many times you had tried before when you try the
pe t
next time. In mathematical notation,
MX (t) = 1−(1−p)e t .
P ( X = n + k| X > n) = P (X = k) .

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 69 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 70 / 112

RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

Solution:
................................... ...................................
.......................................................................
Example 20 (Geometric Distribution)
.......................................................................
An actuary examines low-risk auto policies one by one, continuing until a
.......................................................................
policy with a claim is found. Each low-risk policy has a 10% probability of
having a claim (Question 106, SOA). Let X be the number of policies .......................................................................
examined until the first claim is found. .......................................................................
Objective: Find the probability that the first claim is found on the third .......................................................................
policy examined.
.......................................................................
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 71 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 72 / 112
 RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

Solution:
................................... ...................................
.......................................................................
Example 21 (Memoryless Property) .......................................................................
Continuing with the actuary examining low-risk policies. Suppose the .......................................................................
actuary has already examined 5 policies and none had a claim. .......................................................................
Objective: Find the probability that the actuary will need to examine at .......................................................................
least 3 more policies to find the first claim. .......................................................................
.......................................................................
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 73 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 74 / 112

RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

5. Negative Binomial Distribution.

X = 0, 1, 2, · · ·,
   
r +x −1 x r +x −1 Example 22 (Negative Binomial Distribution)
fX (x) = r
p (1 − p) = p r (1 − p)x , for
x r −1 A high-risk driver has a probability of 0.20 of having an accident in any
x = 0, 1, 2, · · ·, given year (based on Question 18, SOA). Years are independent. Let Y be
E (X ) = r (1−p)
, Var (X ) = r (1−p)
, and the number of years (failures) without an accident before the driver has
p p2
their second accident.

p
r Objective: Find the probability that the driver has exactly 3 accident-free
MX (t) = 1−(1−p)e t .
years before the second accident.
A typical negative binomial distribution arises when we repeat
independently a Bernoulli trial and let X be the number of failures in
order to achieve r successes.

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 75 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 76 / 112
 RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

Solution: 6. Hypergeometric Distribution.

................................... ................................... X = 0, 1, · · · , n,
.......................................................................
  

K 
N −K 
....................................................................... x n−x
fX (x) =   , for x = 0, 1, · · · , n, in which
....................................................................... N
 
....................................................................... n
 
....................................................................... n
= 0 whenever k < 0 or k > n,
k
.......................................................................
nK nK K

N−n
....................................................................... E (X ) = N , Var (X ) = N 1− N N−1 .

....................................................................... A typical Hypergeometric distribution arises when we choose n

....................................................................... objects without replacement from a group of N objects in which there
are K objects of a specific type and let X be the number of objects of
this type in these n objects.
Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 77 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 78 / 112

RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

Solution:
................................... ...................................
Example 23 (Hypergeometric Distribution) .......................................................................
The insurance company has 80 claim files in its inventory. 30 of these .......................................................................
claims are from Source A (high-risk region) and 50 are from Source B .......................................................................
(low-risk region). An auditor randomly selects 5 claims without .......................................................................
replacement for a detailed review (adapted from Question 90, SOA). Let
.......................................................................
X be the number of claims from Source A in the sample.
Objective: Find the probability that exactly 2 of the 5 selected claims are .......................................................................
from Source A. .......................................................................
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 79 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 80 / 112
 RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

Solution:
................................... ...................................
.......................................................................
Example 24 (Binomial vs. Hypergeometric)
.......................................................................
Consider the same population of 80 claims with 30 from Source A.
Compare the probability of selecting exactly 2 from Source A under: .......................................................................
Sampling without replacement (Hypergeometric, as in Example 23): .......................................................................
≈ 0.3547. .......................................................................
Sampling with replacement (Binomial, n = 5, p = 30/80 = 0.375): .......................................................................
.......................................................................
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 81 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 82 / 112

RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

7. Poisson Distribution.

X = 0, 1, 2, · · ·,
e −λ λx Example 25 (Poisson Distribution)
fX (x) = x! , for x = 0, 1, 2, · · ·,
An insurance company determines that the number of claims received in a
E (X ) = Var (X ) = λ, and week, N, follows a Poisson distribution with mean λ = 1 (based on
MX (t) = e λ(e
t −1)
. Question 15, SOA, where P[N = n] = 1/2n+1 corresponds to
λ = ln 2 ≈ 0.693, but we adjust for illustration).
Since fXfX(n+1) λ
(n) = n+1 , the maximum of fX (x) is obtained for the
Objective: Find the probability that exactly 2 claims are received in a
largest x = n + 1 so that λ > n + 1, i.e. x = [[λ]]. When λ is an given week.
integer, the mode is any number between λ − 1 and λ.

Poisson distribution is usually used as an approximation for Binomial

distribution in the case of large value of n and small probability p,
np < 5 for example.
Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 83 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 84 / 112
 RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

Solution:
................................... ...................................
.......................................................................
....................................................................... Example 26 (Poisson Approximation to Binomial)
....................................................................... An insurance company has found that 1% of all applicants for life
....................................................................... insurance have diabetes (Question 260, SOA). The company randomly
selects 200 applicants. Let X be the number of applicants with diabetes.
.......................................................................
Objective: Approximate P(X ≤ 1) using the Poisson distribution.
.......................................................................
.......................................................................
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 85 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 86 / 112

RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

Remark.
Solution:
................................... ................................... With Binomial, negative Binomial, and Poisson distribution, we have
....................................................................... the recursive relationship fXfX(k−1)
(k)
= a + kb , k = 1, 2, 3, · · ·, in which
.......................................................................
p
a = − 1−p , b = (n+1)p
1−p , for Binomial distribution; a = 1 − p,
....................................................................... b = (r − 1) (1 − p), for negative Binomial distribution; and a = 0,
b = λ, for Poisson distribution.
.......................................................................
....................................................................... Note that the variance of Binomial distribution is smaller than its
....................................................................... mean, the variance of Poisson distribution is equal to its mean, and
....................................................................... the variance of negative Binomial distribution is greater than its mean.
....................................................................... These three distributions are frequently used in modeling losing in
....................................................................... property of casualty insurance.

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 87 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 88 / 112
 RANDOM VARIABLES Frequently Discrete Distribution RANDOM VARIABLES Frequently Discrete Distribution

Frequently Discrete Distribution Frequently Discrete Distribution

Solution:
................................... ...................................
Example 27 (Overdispersion in Claim Counts)
.......................................................................
Consider three different portfolios of auto insurance policies:
.......................................................................
1 Binomial: 10 policies, each with accident probability 0.2. Mean = 2,
Variance = 1.6 (variance < mean). .......................................................................
2 Poisson: Average of 2 claims per week. Mean = 2, Variance = 2 .......................................................................
(variance = mean). .......................................................................
3 Negative Binomial: Wait for r = 2 successes with p = 0.5. Number .......................................................................
of failures Y has mean = 2, Variance = 4 (variance > mean).
.......................................................................
Objective: Interpret the variance-to-mean ratio for each distribution.
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 89 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 90 / 112

RANDOM VARIABLES Frequently Continuous Distribution RANDOM VARIABLES Frequently Continuous Distribution

Frequently Continuous Distribution Frequently Continuous Distribution

1. Uniform distribution

X ∼ Unif (a, b). a ≤ X ≤ b with −∞ < a < b < ∞,

Example 28 (Uniform Distribution (Repair Costs))
1 An automobile insurance policy has a deductible of 250. In the event that
fX (x) = b−a ,
the automobile is damaged, repair costs can be modeled by a uniform
a+b (b−a)2
random variable on the interval (0, 1500) (Question 59, SOA). Let X
E (X ) = 2 (which is also its median), Var (X ) = 12 , and denote the repair cost.
Objective: Find the probability that the repair cost exceeds the
e bt −e at deductible, and the expected claim payment.
MX (t) = (b−a)t .

b n+1 −an+1
We also have E (X n ) = (n+1)(b−a) .

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 91 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 92 / 112
 RANDOM VARIABLES Frequently Continuous Distribution RANDOM VARIABLES Frequently Continuous Distribution

Frequently Continuous Distribution Frequently Continuous Distribution

2. Normal distribution
Solution:
................................... ...................................
X ∼ N µ, σ 2 . −∞ < X < ∞,


.......................................................................
.......................................................................
....................................................................... 2
 
fX (x) = √1
σ 2π
exp − (x−µ)
2σ 2 ,
.......................................................................
.......................................................................
....................................................................... E (X ) = µ, Var (X ) = σ 2 , and
.......................................................................
 
....................................................................... MX (t) = exp µt + σ2 t 2
.
2
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 93 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 94 / 112

RANDOM VARIABLES Frequently Continuous Distribution RANDOM VARIABLES Frequently Continuous Distribution

Frequently Continuous Distribution Frequently Continuous Distribution

It is noted that if X ∼ N µ, σ 2 then Z = X σ−µ ∼ N (0, 1) which is

called standard normal distribution with pdf

2
φ (z) ≡ fZ (z) = √12π e −z /2 , its cdf FZ (z) is usually denoted by Φ (z)
or N (z).
Example 29 (Normal Distribution (Claim Amounts))
Claims filed under auto insurance policies follow a normal distribution with
Conversely, if Z ∼ N (0, 1) and X = µ + σZ then X ∼ N µ, σ 2 .


mean 19,400 and standard deviation 5,000 (Question 66, SOA). Let X be
the amount of a randomly selected claim.
A random variable X such that ln X ∼ N µ, σ 2 is called to follow

Objective: Find the probability that a randomly selected claim exceeds
the lognormal distribution (alsowritten as log-normal) and 20,000.
2

fX (x) = exp − (ln 2σ
1
√
xσ 2π
x−µ)
2 for x > 0, 0 elsewhere,
2
  2  2
E (X ) = exp µ + σ2 , and Var (X ) = e σ − 1 e 2µ+σ .

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 95 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 96 / 112
 RANDOM VARIABLES Frequently Continuous Distribution RANDOM VARIABLES Frequently Continuous Distribution

Frequently Continuous Distribution Frequently Continuous Distribution

Solution:
................................... ...................................
.......................................................................
....................................................................... Example 30 (Lognormal Distribution (Property Losses))
....................................................................... An insurer’s annual property losses are modeled by Y = e X , where X is
....................................................................... normally distributed with mean 16 and standard deviation 1.50 (Question
315, SOA). Thus, Y follows a lognormal distribution.
.......................................................................
Objective: Find the expected value and variance of the property losses Y .
.......................................................................
.......................................................................
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 97 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 98 / 112

RANDOM VARIABLES Frequently Continuous Distribution RANDOM VARIABLES Frequently Continuous Distribution

Frequently Continuous Distribution Frequently Continuous Distribution

3. Exponential distribution
Solution:
................................... ................................... X ∼ Exp (λ), λ > 0. X > 0,
....................................................................... fX (x) = λe −λx , for x > 0, zero elsewhere,
....................................................................... FX (x) = 1 − e −λx , for x > 0, zero elsewhere,
.......................................................................
E (X ) = λ1 , Var (X ) = 1
λ2
, and
.......................................................................
λ
.......................................................................
MX (t) = λ−t ,
for t < λ.
We also have E X k = λk!k , for k ∈ N,


.......................................................................
....................................................................... the hazard rate λX (t) = λ,
....................................................................... ln 2
the median is λ , and
.......................................................................
X has the lack of memory property (as of geometric distribution)
P ( X > t + s| X > t) = P (X > s).
Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 99 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 100 / 112
 RANDOM VARIABLES Frequently Continuous Distribution RANDOM VARIABLES Frequently Continuous Distribution

Frequently Continuous Distribution Frequently Continuous Distribution

Solution:
................................... ...................................
Example 31 (Exponential Distribution (Time to Accident))
.......................................................................
The number of days that elapse between the beginning of a calendar year .......................................................................
and the moment a high-risk driver is involved in an accident is
.......................................................................
exponentially distributed. An insurance company expects that 30% of
high-risk drivers will be involved in an accident during the first 50 days of .......................................................................
a calendar year (Question 28, SOA). Let T be the time (in days) until the .......................................................................
first accident. .......................................................................
Objective: Find the parameter λ and the probability of an accident in the .......................................................................
first 80 days.
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 101 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 102 / 112

RANDOM VARIABLES Frequently Continuous Distribution RANDOM VARIABLES Frequently Continuous Distribution

Frequently Continuous Distribution Frequently Continuous Distribution

Solution:
................................... ...................................
.......................................................................
Example 32 (Memoryless Property) .......................................................................
Continuing with the high-risk driver from Example 31. Suppose a driver .......................................................................
has already gone 50 days without an accident. .......................................................................
Objective: Find the probability that the driver will have an accident in the
.......................................................................
next 30 days.
.......................................................................
.......................................................................
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 103 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 104 / 112
 RANDOM VARIABLES Frequently Continuous Distribution RANDOM VARIABLES Frequently Continuous Distribution

Frequently Continuous Distribution Frequently Continuous Distribution

4. Gamma distribution

X ∼ Gamma (α, β), α, β > 0. X > 0,

Example 33 (Gamma Distribution (Time to Multiple Accidents))
βα α−1 e −βx , A company has two electric generators. The time until failure for each
fX (x) = Γ(α) x for x > 0, zero elsewhere,
generator follows an exponential distribution with mean 10 (i.e., λ = 0.1).
The company will begin using the second generator immediately after the
E (X ) = αβ , Var (X ) = α
β2
, and
first one fails (Question 78, SOA). Let T be the total time that the
 α generators produce electricity.
MX (t) = β
, for 0 < t < β. Objective: Identify the distribution of T and find its mean and variance.
β−t

It is noted that Exp (λ) ≡ Gamma (1, λ).

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 105 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 106 / 112

RANDOM VARIABLES Frequently Continuous Distribution RANDOM VARIABLES Frequently Continuous Distribution

Frequently Continuous Distribution Frequently Continuous Distribution

5. Beta distribution
Solution:
................................... ...................................
.......................................................................
X ∼ Beta (r , s), r , s > 0. 0 < X < 1,
.......................................................................
.......................................................................
....................................................................... Γ(r +s) r −1
fX (x) = Γ(r )Γ(s) x (1 − x)s−1 , for 0 < x < 1, zero elsewhere, and
.......................................................................
.......................................................................
....................................................................... r
.......................................................................
E (X ) = r +s .

.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 107 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 108 / 112
 RANDOM VARIABLES Frequently Continuous Distribution RANDOM VARIABLES Frequently Continuous Distribution

Frequently Continuous Distribution Frequently Continuous Distribution

Solution:
................................... ...................................
Example 34 (Beta Distribution (Proportion of Claims)) .......................................................................
Let R be the proportion of right-handed babies among all newborns in a .......................................................................
town each year. R has a beta distribution with parameters α = 13 and .......................................................................
β = 2 (Question 516, SOA). In an insurance context, we could model the
.......................................................................
proportion of claims that exceed a certain threshold using a beta
distribution. .......................................................................
Objective: Find the probability that more than 90% of newborns are .......................................................................
right-handed. .......................................................................
.......................................................................
.......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 109 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 110 / 112

RANDOM VARIABLES Frequently Continuous Distribution RANDOM VARIABLES Frequently Continuous Distribution

Frequently Continuous Distribution Frequently Continuous Distribution

Example 35 (Choosing the Right Distribution) Solution:

................................... ...................................
An actuary needs to select an appropriate continuous distribution for
.......................................................................
different insurance applications.
Objective: Match each insurance scenario to the most appropriate .......................................................................
continuous distribution from this section. .......................................................................
Scenario Distribution Reason .......................................................................
Repair costs with known min and max
.......................................................................
Aggregate claim amounts
Claim sizes with positive skew .......................................................................
Time until first accident .......................................................................
Time until k-th accident .......................................................................
Proportion of claims exceeding limit .......................................................................

Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 111 / 112 Nguyễn Văn Thìn (Khoa Toán Tin Học) PROBABILITY FOR ACTUARIAL SCIENCE Ngày 25 tháng 2 năm 2026 112 / 112