Am 1 Notes

ACTUARIAL MATHEMATICS I LECTURE
NOTES
JANE ADUDA
November 4, 2021
Contents
LIST OF FIGURES iv
LIST OF TABLES v
COURSE OUTLINE v
1 Introduction to Life Insurance 1
1.1 Traditional Insurance contracts . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Term insurance . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.2 Whole life insurance . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.3 Endowment insurance . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.4 With-profit insurance . . . . . . . . . . . . . . . . . . . . . . . . 2

i
Actuarial Mathematics I Jane A. Aduda
1.2 Modern insurance contracts . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Universal life insurance . . . . . . . . . . . . . . . . . . . . . . 3
1.2.2 Unitized with-profit . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.3 Equity-linked insurance . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Underwriting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Premiums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Annuities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.6 Other insurance contracts . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.7 Pension benefits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.7.1 Defined benefit pensions . . . . . . . . . . . . . . . . . . . . . . 8
1.7.2 Defined contribution pensions . . . . . . . . . . . . . . . . . . . 8
2 Survival models and the life tables 9
2.1 Survival models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.1 The future lifetime random variable . . . . . . . . . . . . . . . 9
2.1.2 The force of mortality . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.3 Actuarial notation . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.1.4 The probability Density Function of Tx . . . . . . . . . . . . . . 21
2.1.5 Mean and standard deviation of Tx . . . . . . . . . . . . . . . . 22
2.1.6 Curtate future lifetime . . . . . . . . . . . . . . . . . . . . . . . 26
2.1.7 The complete and curtate expected future lifetimes, e◦x and ex 28
2.1.8 Initial and Central rates of Mortality . . . . . . . . . . . . . . 29
2.2 The Life Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.3 The select mortality table . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3.1 Select, ultimate and aggregate mortality rates . . . . . . . . . 36
ii
2.4 Construction of Life Tables . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.5 Multiple Decrement Tables . . . . . . . . . . . . . . . . . . . . . . . . 40
2.5.1 Conventional notation: . . . . . . . . . . . . . . . . . . . . . . . 41
2.5.2 Fractional Age Assumptions . . . . . . . . . . . . . . . . . . . . 43
2.5.3 Uniform Distribution of Deaths (UDD) . . . . . . . . . . . . . . 43
2.5.4 Constant Force of Mortality . . . . . . . . . . . . . . . . . . . . 45
2.6 Some Analytical Laws of Mortality . . . . . . . . . . . . . . . . . . . . 46
2.6.1 The uniform or De Moivre’s Law . . . . . . . . . . . . . . . . . 46
2.6.2 Gompertz’ Law . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.6.3 Makeham’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.7 Problem set 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3 Mortality Data 53
3.1 Determination of Exposure to Risk by the Census Method . . . . . . 53
3.1.1 Assumptions made in estimating risk time . . . . . . . . . . . 53
3.1.2 The Concept of mx . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.1.3 Derivation of rates from an actual population . . . . . . . . . . 56
3.1.4 An Alternative method of calculating Exposed to Risk . . . . . 57
3.1.5 The Census Method . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.2 Graduation Methods and Applications . . . . . . . . . . . . . . . . . . 59
3.2.1 Graduation of observed mortality rates . . . . . . . . . . . . . 59
3.2.2 Criteria for a good graduation . . . . . . . . . . . . . . . . . . . 62
3.2.3 Methods of Graduation . . . . . . . . . . . . . . . . . . . . . . . 64
3.2.4 Testing the smoothness of a good graduation . . . . . . . . . . 66
4 Monetary Functions 69
iii
4.1 Life Insurance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.1.1 Payable at the end of the year of death . . . . . . . . . . . . . . 70
4.1.2 The variance of the present value of benefits . . . . . . . . . . 71
4.1.3 Payable at the moment of death . . . . . . . . . . . . . . . . . . 73
4.1.4 The variance of the present value of benefits . . . . . . . . . . 75
4.1.5 The relationship between Ax and Āx . . . . . . . . . . . . . . . 76
4.2 Life Annuities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.3 Benefit Premiums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.4 Provisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5 Joint Life Benefits 84
List of Figures
1 Examples of over graduated and under graduated values . . . . . . . 63
List of Tables
1 Abbreviated decennial life table for U.S. Males. . . . . . . . . . . . . . 32
2 Life table example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3 Life table example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4 Extract 1 of A1967-70 life Table . . . . . . . . . . . . . . . . . . . . . . 38
5 Extract 2 of A1967-70 life Table . . . . . . . . . . . . . . . . . . . . . . 39
6 Multiple Decrement Table 1 . . . . . . . . . . . . . . . . . . . . . . . . 40
iv
9 Period of investigation 1 January 1970 to 31 December 1974 . . . . . 56
10 aggregate experience. Life year. Central exposed to risk . . . . . . . . 57
11 population aged 55 last birthday on 1 January . . . . . . . . . . . . . 59
12 Graduation A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
13 Graduation B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
14 Graduation A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
15 Graduation B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
v
STA 2291: ACTUARIAL MATHEMATICS I
AIM: This course introduces the life table, the oldest actuarial model for con-
tingencies (e.g. survival, death, sickness) relating to human life and combines this
with financial mathematics in order to calculate values for annuities and life as-
surance policies for one life.
SYLLABUS: Survival models and the life table including select mortality func-
tions. Population theory: complete and curtate expectation of life, central death
rates, and stationary population including average ages. Determination of expo-
sure to risk by the census method, graduation methods and applications; Formulae
for means and variances of assurances and annuities; commutation functions; pre-
miums; net premiums; provisions; joint life benefits.
PRE-REQUISITES: FINANCIAL MATH I
CO-REQUISITE: PROB & STATS II
COURSE OUTLINE:
Week 1-4
• Survival models and the life tables
(i) The mortality table

(ii) Probability for the Age-at-Death and time-Until-Death for a Person Age
X
(iii) Force of Mortality
(iv) Other Life Table Characteristics
(v) Assumptions for Fractional Ages
(vi) Some Analytical Laws of Mortality
(vii) Select Ultimate and Aggregate Mortality Tables
• Population theory
(i) Curtate-Future-Lifetime
(ii) Complete and curtate expectation of life
(iii) Central death rates
(iv) Stationary population including average ages
Week 5-10
vi
• CAT I
• Determination of exposure to risk by the census method
• Graduation methods and applications
• Life Insurance
(i) Insurances Payable at the Moment of Death -level & varying benefits
(ii) Insurances Payable at the End of the Year of Death -level & varying
benefits
(iii)Relationships between Insurances Payable at Moment of Death and at
End of Year of Death
• Life Annuities
(i) Continuous Life Annuities

(ii) Discrete Life Annuities
(iii) Life Annuities with m-thly Payments
Week 11-14
• CAT II
• Benefit Premiums
(i) Fully Continuous Premiums

(ii)Fully Discrete Premiums
(iii) Premiums paid m-thly
(iv) Office premiums and net premiums
• Provisions
• Joint life benefits
• Revision
REFERENCES:
1. NL Bowers, HU Gerber, JC Hickman et al. Actuarial mathematics. 2nd ed.

Society of Actuaries, 1997. 753 pages.
2. WF Scott Life Assurance Mathematics, Heriot-Watt University, 1999.

vii
3. Neill, A. Heinemann Life contingencies., 1977.
4. Parmenter. Theory of Interest and Life Contingencies.
5. HU Gerber Life insurance mathematics. 3rd ed. Springer; Swiss Association

of Actuaries, 1997. 217 pages.
6. Batten & London. Life Contingencies. A Logical Approach to Actuarial Math-

ematics. 2005
7. Formulae & Tables for Actuarial Examinations.(Institute of Actuaries)
viii
1 Introduction to Life Insurance
Actuaries apply scientific principles and techniques from a range of other disci-
plines to problems involving risk, uncertainty and finance.
1.1 Traditional Insurance contracts
Term, whole life and endowment insurance are the traditional products, provid-
ing cash benefits on death or maturity, usually with predetermined premium and
benefit amounts.
1.1.1 Term insurance
This pays a lump sum benefit on the death of the policyholder, provided death
occurs before the end of a specified term. Term insurance allows a policyholder
to provide a fixed sum for his or her dependants in the event of the policyholder’s
death.
• Level term insurance indicates a level sum insured and regular, level pre-
miums.
• Decreasing term insurance indicates that the sum insured and (usually)
premiums decrease over the term of the contract. This type of contract is
popular in the UK where it is used in conjunction with a home mortgage; if
the policyholder dies, the remaining mortgage is paid from the term insurance
proceeds.
• Renewable term insurance offers the policyholder the option of renewing

the policy at the end of the original term, without further evidence of the
policyholder’s health status.
• Convertible term insurance offers the policyholder the option to convert to

a whole life or endowment insurance at the end of the original term, without
further evidence of the policyholder’s health status.
1
1.1.2 Whole life insurance
This contract pays a lump sum benefit on the death of the policyholder whenever
it occurs. For regular premium contracts, the premium is often payable only up to
some maximum age, such as 80. This avoids the problem that older lives may be
less able to pay the premiums.
1.1.3 Endowment insurance
This contract offers a lump sum benefit paid either on the death of the policyholder
or at the end of a specified term, whichever occurs first. It is a combination of a
term insurance benefit and a savings element. If the policyholder dies, the sum
insured is paid just as under term insurance; if the policyholder survives, the sum
insured is treated as a maturing investment.
Endowment insurance policies are becoming popular in developing nations, par-

ticularly for ‘micro-insurance’ where the amounts involved are small. It is hard for
small investors to achieve good rates of return on investments, because of heavy
expense charges. By pooling the death and survival benefits under the endow-
ment contract, the policyholder gains on the investment side from the resulting
economies of scale, and from the investment expertise of the insurer.
1.1.4 With-profit insurance
Under with-profit arrangements, the profits earned on the invested premiums are
shared with the policyholders. These often takes the form of cash dividends or
reduced premiums. They could also be used to increase the sum insured, through
bonuses called ‘reversionary bonuses’ (awarded during the term of the contract;)
and ‘terminal bonuses’ (awarded when the policy matures, either through the death
of the insured, or when an endowment policy reaches the end of the term). Once a
reversionary bonus is awarded it is guaranteed.
Reversionary Bonuses may be expressed as a percentage of the total of the pre-

vious sum insured plus bonus, or as a percentage of the original sum insured plus
2
a different percentage of the previously declared bonuses. Reversionary and termi-

nal bonuses are determined by the insurer based on the investment performance
of the invested premiums.
1.2 Modern insurance contracts
In recent years insurers have provided more flexible products that combine the
death benefit coverage with a significant investment element, as a way of compet-
ing for policyholders savings with other institutions, for example, banks or open-
ended investment companies (e.g. mutual funds or unit trusts). Additional flexi-
bility also allows policyholders to purchase less insurance when their finances are
tight, and then increase the insurance coverage when they have more money avail-
able.
1.2.1 Universal life insurance
It combines investment and life insurance. The policyholder determines a premium

and a level of life insurance cover. Some of the premium is used to fund the life
insurance; the remainder is paid into an investment fund. Premiums are flexible,
as long as they are sufficient to pay for the designated sum insured under the term
insurance part of the contract. Under variable universal life, there is a range of
funds available for the policyholder to select from.
1.2.2 Unitized with-profit
This is an evolution from the conventional with-profit policy, designed to be more

transparent than the original. Premiums are used to purchase units (shares) of
an investment fund, called the with-profit fund. As the fund earns investment
return, the shares increase in value (or more shares are issued), increasing the
benefit entitlement as reversionary bonus. The shares will not decrease in value.
On death or maturity, a further terminal bonus may be payable depending on the
performance of the with-profit fund.
3
1.2.3 Equity-linked insurance
It has a benefit linked to the performance of an investment fund. There are two
different forms. The first is where the policyholder’s premiums are invested in
an open-ended investment company style account; at maturity, the benefit is the
accumulated value of the premiums. There is a guaranteed minimum death benefit
payable if the policyholder dies before the contract matures. In some cases, there
is also a guaranteed minimum maturity benefit payable.
1.3 Underwriting
This is the process of collecting and evaluating information on rating factors ( their
age, gender, smoking habits, occupation, any dangerous hobbies, and personal and
family health history.) which basically give an indication of the policyholder’s risk
level.
The purpose of underwriting is, first, to classify potential policyholders into

broadly homogeneous risk categories, and secondly to assess what additional pre-
mium would be appropriate for applicants whose risk factors indicate that stan-
dard premium rates would be too low. On the basis of the application and sup-
porting medical information, potential life insurance policyholders will generally
be categorized into one of the following groups:
• Preferred lives who have very low mortality risk based on the standard
information. The preferred applicant would have no recent record of smoking;
no evidence of drug or alcohol abuse; no high-risk hobbies or occupations;
no family history of disease known to have a strong genetic component; no
adverse medical indicators such as high blood pressure or cholesterol level or
body mass index.
• Normal lives who may have some higher rated risk factors than preferred
lives (where this category exists), but are still insurable at standard rates.
Most applicants fall into this category.
4
• Rated lives who have one or more risk factors at raised levels and so are not
acceptable at standard premium rates. However, they can be insured for a
higher premium. An example might be someone having a family history of
heart disease. These lives might be individually assessed for the appropriate
additional premium to be charged. This category would also include lives with
hazardous jobs or hobbies which put them at increased risk.
• Uninsurable lives who have such significant risk that the insurer will not
enter an insurance contract at any price.
The rigour of the underwriting process will depend on the type of insurance being
purchased, on the sum insured and on the distribution process (weather through
brokers or by direct marketing) of the insurance company. Term insurance is gen-
erally more strictly underwritten than whole life insurance, as the risk taken by
the insurer is greater. Under whole life insurance, the payment of the sum insured
is certain, the uncertainty is in the timing. Under, say, 10-year term insurance, it
is assumed that the majority of contracts will expire with no death benefit paid. If
the underwriting is not strict there is a risk of adverse selection by policyholders,
that is, that very high-risk individuals will buy insurance in disproportionate num-
bers, leading to excessive losses. Since high sum insured contracts carry more risk
than low sum insured, high sums insured would generally trigger more rigorous
underwriting.
1.4 Premiums
A life insurance policy may involve a single premium, payable at the outset of the
contract, or a regular series of premiums payable provided the policyholder sur-
vives, perhaps with a fixed end date. In traditional contracts the regular premium
is generally a level amount throughout the term of the contract; in more modern
contracts the premium might be variable, at the policyholder’s discretion for in-
vestment products such as equity-linked insurance, or at the insurer’s discretion
for certain types of term insurance. Regular premiums may be paid annually, semi-
annually, quarterly, monthly or weekly. An important feature of all premiums is
that they are paid at the start of each period.
5
1.5 Annuities
Annuity contracts offer a regular series of payments. When an annuity depends on

the survival of the recipient, it is called a ‘life annuity’. The recipient is called an
annuitant. If the annuity continues until the death of the annuitant, it is called
a whole life annuity. If the annuity is paid for some maximum period, provided
the annuitant survives that period, it is called a term life annuity. Annuities are
often purchased by older lives to provide income in retirement. Buying a whole life
annuity guarantees that the income will not run out before the annuitant dies.
• Single Premium Deferred Annuity (SPDA) Under an SPDA contract, the

policyholder pays a single premium in return for an annuity which com-
mences payment at some future, specified date. The annuity is ‘life contin-
gent’, meaning the annuity is paid only if the policyholder survives to the
payment dates. If the policyholder dies before the annuity commences, there
may be a death benefit due. If the policyholder dies soon after the annuity
commences, there may be some minimum payment period, called the guaran-
tee period, and the balance would be paid to the policyholder’s estate.
• Single Premium Immediate Annuity (SPIA) This contract is the same

as the SPDA, except that the annuity commences as soon as the contract is
effected. This might, for example, be used to convert a lump sum retirement
benefit into a life annuity to supplement a pension. As with the SPDA, there
may be a guarantee period applying in the event of the early death of the
annuitant.
• Regular Premium Deferred Annuity (RPDA) The RPDA offers a deferred

life annuity with premiums paid through the deferred period. It is otherwise
the same as the SPDA.
• Joint life annuity. A joint life annuity is issued on two lives, typically a
married couple. The annuity (which may be single premium or regular pre-
mium, immediate or deferred) continues while both lives survive, and ceases
on the first death of the couple.
6
• Last survivor annuity. A last survivor annuity is similar to the joint life an-
nuity, except that payment continues while at least one of the lives survives,
and ceases on the second death of the couple.
• Reversionary annuity. A reversionary annuity is contingent on two lives,

usually a couple. One is designated as the annuitant, and one the insured
(counter). No annuity benefit is paid while the insured life survives. On the
death of the insured life, if the annuitant is still alive, the annuitant receives
an annuity for the remainder of his or her life.
1.6 Other insurance contracts
The insurance and annuity contracts described above are all contingent on death or
survival. There are other life contingent risks, in particular involving short-term
or long-term disability. These are known as morbidity risks.
• Income protection insurance When a person becomes sick and cannot

work, their income will, eventually, be affected. For someone in regular em-
ployment, the employer may cover salary for a period, but if the sickness
continues the salary will be decreased, and ultimately will stop being paid at
all. For someone who is self-employed, the effects of sickness on income will
be immediate. Income protection policies replace at least some income during
periods of sickness. They usually cease at retirement age.
• Critical illness insurance Some serious illnesses can cause significant ex-
pense at the onset of the illness. The patient may have to leave employment,
or alter their home, or incur severe medical expenses. Critical illness insur-
ance pays a benefit on diagnosis of one of a number of severe conditions, such
as certain cancers or heart disease. The benefit is usually in the form of a
lump sum.
• Long-term care insurance This is purchased to cover the costs of care in

old age, when the insured life is unable to continue living independently. The
benefit would be in the form of the long-term care costs, so is an annuity
benefit.
7
1.7 Pension benefits
Many actuaries work in the area of pension plan design, valuation and risk man-
agement. The pension plan is usually sponsored by an employer. Pension plans
typically offer employees (also called pension plan members) either lump sums or
annuity benefits or both on retirement, or deferred lump sum or annuity benefits
(or both) on earlier withdrawal. Some offer a lump sum benefit if the employee
dies while still employed. The benefits therefore depend on the survival and em-
ployment status of the member, and are quite similar in nature to life insurance
benefits i.e., they involve investment of contributions long into the future to pay
for future life contingent benefits.
1.7.1 Defined benefit pensions
Defined Benefit (DB) pensions offer retirement income based on service and salary
with an employer, using a defined formula to determine the pension. The defined
benefit is funded by contributions paid by the employer and (usually) the employee
over the working lifetime of the employee. The contributions are invested, and the
accumulated contributions must be enough, on average, to pay the pensions when
they become due.
1.7.2 Defined contribution pensions
Defined Contribution (DC) pensions work more like a bank account. The em-
ployee and employer pay a predetermined contribution (usually a fixed percentage
of salary) into a fund, and the fund earns interest. When the employee leaves or
retires, the proceeds are available to provide income throughout retirement.
8
2 Survival models and the life tables
Many insurance policies provide a benefit on the death of the policyholder. When
an insurance company issues such a policy, the policyholder’s date of death is un-
known, so the insurer does not know exactly when the death benefit will be payable.
In order to estimate the time at which a death benefit is payable, the insurer needs
a model of human mortality, from which probabilities of death at particular ages
can be calculated.
2.1 Survival models
We will model the future lifetime of an individual as a continuous random variable.

Though we will study this in the context of human mortality, the theory can equally
be applied in other problems such as:-
-Analysing the lengths of time that surviving individuals hold insurance policies.
Here, mortality is replaced by “withdrawal”
-Analysing the lengths of time that surviving individuals remain healthy. Here,
mortality is replaced by “sickness”
2.1.1 The future lifetime random variable
The starting point of a simple mathematical model of survival is the observation

that the future lifetime of a person (called “a life” in actuarial work) is not known
in advance.
Let (x) denote a life aged x, where x ≥ 0. The death of (x) can occur at any
age greater than x, and we model the future lifetime of (x) by a continuous random
variable which we denote by Tx . This means that x+Tx represents the age-at-death
random variable for (x) .
We can also model the future lifetime of a new-born as a random variable de-
noted by T = T0 , which is continuously distributed in the interval [0, ω] where
9
0 < ω < ∞. The maximum age ω is called the limiting age. It is the youngest age
to which no one survives. typical values of ω for practical work are in the range
100 − 120.The possibility of surviving beyond age ω is excluded by the model for
convenience and simplicity.
Let Fx be the distribution function of Tx , so that
Fx (t) = pr [Tx ≤ t]
Then Fx (t) represents the probability that (x) does not survive beyond age x + t,
and we refer to Fx as the lifetime distribution from age x.
In many life insurance problems we are interested in the probability of survival

rather than death, and so we define Sx as
Sx (t) = 1 − Fx (t) = pr [Tx > t]
Thus, Sx (t) represents the probability that (x) survives for at least t years, and Sx
is known as the survival function.
The distribution function of the random variable Tx , 0 ≤ x ≤ ω must satisfy the

following very important relationship.
Fx (t) = pr [Tx ≤ t] = pr [T ≤ x + t|T > x] (2.1)
Now, recall from probability theory that for two events A and B
pr [A and B]
pr [A|B] =
pr [B]
Given this, we can now rearrange the right side of equation (2.1) to give
pr [x < T ≤ x + t]
pr [Tx ≤ t] =
pr [T > x]
which can be written as
F0 (x + t) − F0 (x)
Fx (t) = (2.2)
S0 (x)
10
Since, using the laws of probability, we can make probability statements about the
age-at-death in terms of either the survival function or the distribution function.
The probability that a new-born dies between ages x and x + t is given by
pr [x < T ≤ x + t] = F0 (x + t) − F0 (x)
and also using the result Sx (t) = 1 − Fx (t), we get the results
pr [x < T ≤ x + t] = S0 (x) − S0 (x + t) (2.3)
and from equation (2.2), we see that
S0 (x + t)
Sx (t) = (2.4)
S0 (x)
Since equation (2.2) can also be written as
S0 (x) − S0 (x + t) S (x + t)
Fx (t) = =1− 0
S0 (x) S0 (x)
and Sx (t) = 1 − Fx (t), hence the result in equation (2.4) which can also yield the
result
S0 (x + t) = S0 (x)Sx (t) (2.5)
This is a very important result. It shows that we can interpret the probability of
survival from age x to age x + t as the product of
1. the probability of survival to age x from birth, and
2. the probability, having survived to age x, of further surviving to age x + t.
Similarly, any survival probability for (x) , for, say, t + u years can be split into
the probability of surviving the first t years, and then, given survival to age x + t,
11
subsequently surviving another u years. That is,
S0 (x + t + u)
Sx (t + u) =
S0 (x)
S (x + t) S0 (x + t + u) (2.6)
= 0
S0 (x) S0 (x + t)
= Sx (t)Sx+t (u)
Any survival function for a lifetime distribution must satisfy the following condi-
tions to be valid.
1. Sx (0) = 1; that is, the probability that a life currently aged x survives 0 years
is 1. This is the same as saying Fx (0) = pr [Tx ≤ 0] = 0. In any case Sx (0) =
1 − Fx (0)
2. limt→∞ Sx (t) = 0; that is, all lives eventually die.
3. The survival function must be a non-increasing function of t; it cannot be

more likely that (x) survives, say 10.5 years than 10 years, because in order
to survive 10.5 years, (x) must first survive 10 years.
For the distributions used in this course, we make three additional assumptions:
1. Sx (t) is differentiable for all t > 0. Note that coupled with the condition that
the survival function must be a non-increasing function of t, this means that
d
S (t) ≤ 0 for all t > 0.
dt x
2. limt→∞ tSx (t) = 0
3. limt→∞ t2 Sx (t) = 0
These last two assumptions ensure that the mean and variance of the distribution
of Tx exist. These are not particularly restrictive constraints, we do not need to
worry about distributions with infinite mean or variance in the context of individ-
uals’ future lifetimes. These three extra assumptions are valid for all distributions
that are feasible for human lifetime modelling.
12
Example 2.1. A survival distribution is defined by
S0 (x) = ax2 + b 0 ≤ x ≤ ω.
Determine a and b.
Solution:
For x = 0, S0 (0) = 1 = b
For x = ω, S0 (ω) = 0 = aω 2 + 1
aω 2 = −1
1 x2
a=− so that S 0 (x) = − +1
ω2 ω2
Example 2.2. Let

61
t
F0 (t) = 1 − 1 −
120
Calculate the probability that
(a) a newborn life survives beyond age 30,
(b) a life aged 30 dies before age 50, and
(c) a life aged 40 survives beyond age 65.
Solution:
(a) The required probability is

16
30
S0 (30) = 1 − F0 (30) = 1 − 1 − = 0.9532.
120
(b) The required probability is
F0 (50) − F0 (30)
pr (30 < x ≤ 50|x > 50) = F30 (20) = = 0.0410
1 − F0 (30)
13
(c) The required probability is
S0 (65)
S40 (25) = = 0.9395.
S0 (40)
We remark that in the above example, S0 (120) = 0, which means that under this
model, survival beyond age 120 is not possible. In this case we refer to 120 as the
limiting age of the model. In general, if there is a limiting age, we use the Greek
letter ω to denote it. In models where there is no limiting age, it is often practical
to introduce a limiting age in calculations.
Example 2.3. Let

F0 (t) = 1 − e−0.008x , x≥0
Find the probability that a newborn baby dies between age 60 and age 70.
Solution: The required probability is
pr (60 < x ≤ 70) = F0 (70) − F0 (60) = e−0.48 − e−0.56 = 0.04757
Example 2.4. The survival distribution function for an individual is determined

to be
75 − x
S0 (x) = 0 ≤ x ≤ 75
75
(a) Find the probability that the person dies before reaching the age of 18.
(b) Find the probability that the person lives more than 55 years.
(c) Find the probability that the person dies between the ages of 25 and 70.
Solution:
(a) The probability that the person dies before reaching the age of 18.

75 − 18
F0 (18) = 1 − S0 (18) = 1 − = 1 − 0.76 = 0.24
75
(b) The probability that the person lives more than 55 years.
75 − 55
S0 (55) = = 0.26667
75
14
(c) The probability that the person dies between the ages of 25 and 70.
75 − 25 75 − 70
S0 (25) − S0 (70) = − = 0.6667 − 0.0667 = 0.6
75 75
2.1.2 The force of mortality
The force of mortality is an important and fundamental concept in modelling future

lifetime. In statistics it is referred to as The hazard rate. We denote the force of
mortality at age x by µx and define it as
1
µx = lim+ × pr [T ≤ x + h|T > x] (2.7)
h→0 h
Note: We will always assume the limit exists.

We can define the force of mortality in two ways:
1. Either by thinking in terms of a new-born
1
µx+t = lim+ × pr [T ≤ x + t + h|T > x + t]
h→0 h
or
2. by thinking in terms of a person who has already attained the age in question
1
µx+t = lim+ × pr [Tx ≤ t + h|Tx > t]
h→0 h
We will often use µx+t for a fixed age x and 0 ≤ t ≤ ω − x

From equation (2.7) we see that an equivalent way of defining µx is
1
µx = lim+ × pr [Tx ≤ h] (2.8)
h→0 h
which can be written in terms of the survival function Sx as
1
µx = lim+ × [1 − Sx (h)] (2.9)
h→0 h
15
Note that the force of mortality depends, numerically, on the unit of time; if we are
measuring time in years, then µx is measured per year. The force of mortality is
best understood by noting that for very small h, formula (2.7) gives the approxima-
tion
µx h ≈ pr [T ≤ x + h|T > x] (2.10)
Thus, for very small h, we can interpret µx h as the probability that a life who has
attained age x dies before attaining age x + h. For example, suppose we have a
male aged exactly 50 and that the force of mortality at age 50 is 0.0044 per year.
A small value of h might be a single day, or 0.00274 years. Then the approximate
probability that (50) dies on his birthday is 0.0044 × 0.0027 = 1.2 × 10−5 .
We can also relate the force of mortality to the survival function from birth, S0 as
S0 (x + h)
Sx (h) =
S0 (x)
formula (2.7) gives
1 S (x) − S0 (x + h)
µx = lim+ 0
S0 (x) h→0 h

1 d
= − S0 (x)
S0 (x) dx
Thus,
−1 d
µx = S (x) (2.11)
S0 (x) dx 0
From standard results in probability theory, we know that the probability density
function for the random variable Tx , which we denote fx (t), is related to the distri-
bution function Fx (t) and the survival function Sx (t) by
d d
fx (t) = Fx (t) = − Sx (t)
dt dt
So, it follows from equation (2.11) that
f0 (x)
µx =
S0 (x)
16
We can also relate the force of mortality function at any age x + t, t > 0, to the
lifetime distribution of Tx . Assume x is fixed and t is variable. Then d(x + t) = dt
and so
1 d
µx+t = − S (x + t)
S0 (x + t) d(x + t) 0
1 d
=− S (x + t)
S0 (x + t) dt 0
1 d
=− S (x)Sx (t)
S0 (x + t) dt 0
S0 (x) d
=− S (t)
S0 (x + t) dt x
−1 d
= S (t)
Sx (t) dt x
Hence
fx (t)
µx+t = (2.12)
Sx (t)
This relationship gives a way of finding µx+t given Sx (t). We can also use equation
(2.11) to develop a formula for Sx (t) in terms of the force of mortality function. We
use the fact that for a function g whose derivative exists,
d 1 d
log g(x) = g(x)
dx g(x) dx
so from equation (2.11) we have
d
µx = − log S0 (x)
dx
and integrating this identity over (0, y) yields

Z y
µx dx = −(log S0 (y) − log S0 (0))
0
As log S0 (0) = log Pr [T0 > 0] = log 1 = 0, we obtain

Z y
S0 (y) = exp − µx dx
0
17
from which it follows that

Z x+t Z t
S0 (x + t)
Sx (t) = = exp − µr dr = exp − µx+s ds (2.13)
S0 (x) x 0
This means that if we know µx for all x > 0, then we can calculate all the survival
probabilities Sx (t), for any x and t. In other words, the force of mortality function
fully describes the lifetime distribution, just as the function S0 does. In fact, it
is often more convenient to describe the lifetime distribution using the force of
mortality function than the survival function.
Example 2.5. As in Example 2.2, let

61
t
F0 (t) = 1 − 1 −
120
for 0 ≤ x ≤ 120. Derive an expression for µx .

Solution: As S0 (x) = (1 − x/120)1/6 , it follows that
x i− 56

d 1h 1
S (x) = 1− −
dx 0 6 120 120
and so
−1 d 1 h x i−1 1
µx = S0 (x) = 1− =
S0 (x) dx 720 120 720 − 6x
As an alternative, we could use the relationship

d d 1 h x i 1 1
µx = − log S0 (x) = − log 1 − = =
dx dx 6 120 720 (1 − x/120) 720 − 6x
Example 2.6. The survival distribution function for an individual is determined

to be
75 − x
S0 (x) = 0 ≤ x ≤ 75
75
Calculate
(a) F0 (x)
(b) f0 (x)
(c) P r(20 < X < 50)

18
Solution:
(a)
x′
F0 (x) = 1 − S0 (x) = x>0
75
(b)
d 1
f0 (x) = F0 (x) =
dx 75
(c)
75 − 20 75 − 50
P r(20 < X < 50) = S0 (20) − S0 (50) = − = 0.7333 − 0.3333 = 0.4
75 75
2.1.3 Actuarial notation
The notation used in the previous sections, Sx (t), Fx (t) and fx (t), is standard in
statistics. Actuarial science has developed its own notation, International Actuar-
ial Notation, that encapsulates the probabilities and functions of greatest interest
and usefulness to actuaries. The force of mortality notation, µx , comes from In-
ternational Actuarial Notation. The actuarial notation for survival and mortality
probabilities are:
1. t px is the probability that (x) survives to at least age x + t,
t px = pr [Tx > t] = Sx (t) (2.14)
2. t qx is the probability that (x) dies before age x + t,
t qx = pr [Tx ≤ t] = 1 − Sx (t) = Fx (t) (2.15)
3. qx is the probability that (x) survives u years, and then dies in the sub-
u|t
sequent t years, that is, between ages x + u and x + u + t. This is called a

deferred mortality probability, because it is the probability that death
occurs in some interval following a deferred period.
u|t
qx = pr [u < Tx ≤ u + t] = Sx (u) − Sx (u + t) (2.16)
19
We may drop the subscript t if its value is 1, so that px represents the probability
that (x) survives to at least age x + 1. Similarly, qx is the probability that (x) dies
before age x + 1. In actuarial terminology qx is called the mortality rate at age x
and the force of mortality µx is the instantaneous measure of mortality at age x. It
is the continuous measure equivalent of the discrete quantity qx .
The relationships below follow immediately from the definitions above
t px + t qx = 1
u|t
qx = u px − u+t px
Also
x+t p0
t px =
x p0
Therefore, for any age x and for u > 0 and t > 0

p
u+t px = x+u+t 0
x p0
p p0
= x+u 0 × x+u+t
(2.17)
x p0 x+u p0
=u px × t px+u
Similarly
u+t px = t px × u px+t
i.e. the probability of surviving for time (u + t) after age x is given by multiplying
the probability of surviving for time u and the probability of then surviving for a
further time t or multiplying
The probability of surviving for time t and the probability of then surviving for a
further time u.
This is referred to as the consistency condition.
20
Similarly,
1 d d
µx+t = − t px ⇒ p = −t px µx+t (2.18)
t px dt dt t x
fx (t)
µx+t = ⇒ fx (t) = t px µx+t from (2.12) (2.19)
Sx (t)
and
Z t
t px = exp − µx+s ds from (2.13). (2.20)
0
2.1.4 The probability Density Function of Tx
The distribution function of Tx is Fx (t). We also want to know it’s probability den-
sity function (p.d.f). This is denoted by fx (t) and recall that
Z t
Fx (t) = fx (s)ds
0
which can be written in actuarial notation as

Z t
t qx = s px µx+s ds (2.21)
0
This is an important formula, which can be interpreted as follows. Consider time s,

where 0 ≤ s < t. The probability that (x) is alive at time s is s px , and the probability
that (x) dies between ages x + s and x + s + ds, having survived to age x + s, is
(loosely) µx+s ds, provided that ds is very small. Thus s px µx+s ds can be interpreted
as the probability that (x) dies between ages x + s and x + s + ds. Now, we can sum
over all the possible death intervals s to s + ds -which requires integrating because
these are infinitesimal intervals -to obtain the probability of death before age x + t.
In the special case when t = 1, formula (2.20) becomes
Z 1
qx = s px µx+s ds (2.22)
0
21
When qx is small, it follows that px is close to 1, and hence s px is close to 1 for

0 ≤ s < t. Thus Z 1
qx ≈ s px µx+s ds ≈ µx+ 1
0 2
where the second relationship follows by the mid-point rule for numerical integra-
tion.

61
t
F0 (t) = 1 − 1 −
120
for 0 ≤ x ≤ 120 Calculate both qx and µx+1/2 for x = 20 and for x = 110, and comment
on these values.
Solution: We have 1/6
S0 (x + 1) 1
px = = 1−
S0 (x) 120 − x
giving q20 = 0.00167 and q110 = 0.01741, and from the solution to Example 2.2,
µ20+1/2 = 0.00168 and µ110+1/2 = 0.01754. We see that µx+1/2 is a good approxima-
tion to qx when the mortality rate is small, but is not such a good approximation,
at least in absolute terms, when the mortality rate is not close to 0.
2.1.5 Mean and standard deviation of Tx
Now, we consider the expected future lifetime of (x), E[Tx ], denoted in actuarial
notation by e◦x . We also call this the complete expectation of life. In order to
evaluate e◦x , we note from formulae (2.18) and (2.19) that
d
fx (t) = t px µx+t dt = − p (2.23)
dt t x
From the definition of an expected value, we have

Z ∞
◦
ex = tfx (t)dt
0
Z ∞
= tt px µx+t dt
0
22
We can now use (2.23) to evaluate this integral using integration by parts as
Z ∞

◦ d
ex = − t p dt
0 dt t x
Z ∞
∞
= − tt px |0 − t px dt
0
and earlier, we stated the assumption that eventually everyone dies, i.e. limt→∞ t px =
0, which gives
Z ∞
◦
ex = t px dt (2.24)
0
Similarly, for E[Tx2 ], we have

Z ∞
2
E[Tx ] = t2 fx (t)dt
0
Z ∞
2 d
=− t p dt
0 dt t x
Z ∞ (2.25)
2 ∞
= − t t px |0 − t px 2tdt
0
Z ∞
=2 tt px dt
0
So we have integral expressions for E[Tx ] and E[Tx2 ]. For some lifetime distribu-
tions we are able to integrate directly. In other cases we have to use numerical
integration techniques to evaluate the integrals in (2.24) and (2.25). The variance
of Tx can then be calculated as
2
V [Tx ] = E[Tx2 ] − e◦x

61
t
F0 (t) = 1 − 1 −
120
for 0 ≤ x ≤ 120 Calculate e◦x and V [Tx ] for (a) x = 30 and (b) x = 80.
23
Solution : As S0 (x) = (1 − x/120)1/6 , we have

1/6
S0 (x + t) t
t px = = 1−
S0 (x) 120 − x
Now recall that this formula is valid for 0 ≤ t ≤ 120 − x, since under this model
survival beyond age 120 is impossible. Technically, we have
 1/6
 1− t
120−x
for x + t ≤ 120,
t px =
0 for x + t > 120
So the upper limit of integration in equation (2.24) is 120 − x, and

Z 120−x 1/6
◦ t
ex = 1− dt
0 120 − x
Now let y = 1 − t/(120 − x), so that t = (120 − x)(1 − y), giving

Z 1
◦
ex = (120 − x) y 1/6 dy
0
6
= (120 − x)
7
Then e◦30 = 77.143 and e◦80 = 34.286 Under this model the expectation of life at any
age x is 6/7 of the time to age 120.
For the variance we require E[Tx2 ] Using equation (2.25) we have
Z 120−x
2
E[Tx ] =2 tt px dt
0
Z 120−x 1/6
t
=2 t 1− dt
0 120 − x
again lets substitute y = 1 − t/(120 − x) giving

Z 1
2 2
y 1/6 − y 7/6 dy

E[Tx ] =2 (120 − x)
0
2 6 6
=2 (120 − x)
7 13
24
Then
2
V [Tx ] =E[Tx2 ] − e◦x
2 !
6 6 6
= (120 − x)2 2 − −
7 13 7
= (120 − x)2 (0.056515)
= ((120 − x) (0.23773))2
So V [T30 ] = 21.3962 and V [T80 ] = 9.5092.

Since we know under this model that all lives will die before age 120, it makes
sense that the uncertainty in the future lifetime should be greater for younger
lives than for older lives.
Example 2.9. In a particular survival model, we have:
0.01
µx = for 0 ≤ x < 100
1 − 0.01x
Find the complete expectation of life at exact age 20.
solution First, we must find t px , the survival function for a life of exact age 20.
Z 20+t
t p20 = exp − µs ds
20
Z 20+t
0.01
= exp − ds
20 1 − 0.01s
= exp − [− ln (1 − 0.01s)]20+t

20
1 − 0.01 (20 + t)
=
1 − 0.01 (20)
0.8 − 0.01t
=
0.8
1 − 0.01t t
= =1−
0.8 80
As the limiting age in the survival model is 100, the complete future lifetime for a
life of exact age 20 must be less than 80 years.
25
Now we have:
Z 100−20
e◦20 = t p20 dt
0
Z 80
t
= 1− dt
0 80
80
t2

= t−
160 0
= 80 − 40 = 40
Thus, the complete expectation of life for a life of exact age 20 is 40 years.
2.1.6 Curtate future lifetime
In many insurance applications we are interested not only in the future lifetime of
an individual, but also in what is known as the individual’s curtate future lifetime.
The curtate future lifetime random variable is defined as the integer part of future
lifetime, and is denoted by Kx for a life aged x. If we let ⌊ ⌋ denote the floor function,
we have
Kx = ⌊Tx ⌋
In other words Kx is equal to Tx rounded down to the integer below. The curtate
future lifetime is the number of whole years lived in the future by an individual. As
an illustration of the importance of curtate future lifetime, consider the situation
where a life aged x at time 0 is entitled to payments of 1 at times 1, 2, 3, . .
. provided that (x) is alive at these times. Then the number of payments made
equals the number of complete years lived after time 0 by (x). This is the curtate
future lifetime.
For k = 0, 1, 2, . . . , Kx = k if and only if (x) dies between the ages of x + k and
26
x + k + 1. Thus for k = 0, 1, 2, . . . ,
P r [Kx = k] = P r [k ≤ Tx < k + 1]
= Fx (k + 1) − Fx (k)
= [1 − Sx (k + 1)] − [1 − Sx (k)]
= Sx (k) − Sx (k + 1)
= k px − k+1 px
= k px − k px px+k

= k px 1 − px+k
= k px qx+k
= k| qx
The expected value of Kx is denoted by ex , so that ex = E [Kx ], and is referred to as

the curtate expectation of life. So
E [Kx ] = ex
X∞
= kP r [Kx = k]
k=0
X∞
= k (k px − k+1 px ) (2.26)
k=0
= (px − 2 px ) + 2 (2 px − 3 px ) + 3 (3 px − 4 px ) + . . .
X∞
= k px
k=1
Note that the lower limit of summation is k = 1.
27
Similarly,
∞
X
Kx2 k 2 (k px − k+1 px )

E =
k=0
= (px − 2 px ) + 4 (2 px − 3 px ) + 9 (3 px − 4 px ) + 16 (4 px − 5 px ) . . .
= px − 2 px + 42 px − 43 px + 93 px − 94 px + 164 px − 165 px + . . .
= px + 3 2 px + 5 3 p x + 7 4 px + 9 5 px + . . .
X∞ (2.27)
= (2k − 1) k px
k=1
X∞ ∞
X
=2 k k px − k px
k=1 k=1
∞
X
=2 k k p x − ex
k=1
2.1.7 The complete and curtate expected future lifetimes, e◦x and ex
As the curtate future lifetime is the integer part of future lifetime, we can obtain
an approximate relationship between e◦x and ex by writing
Z ∞ ∞ Z
X j+1
e◦x = t px dt = t px dt
0 j=0 j
And approximating each integral using the trapezium rule for numerical integra-
tion, we obtain Z j+1
1
t px dt ≈ (j px + j+1 px )
j 2
and hence ∞ ∞
X 1 1 X
e◦x ≈ (j px + j+1 px ) = + j px
j=0
2 2 j=1
giving the frequently used approximation
1
e◦x ≈ ex + (2.28)
2
28
Example 2.10. Show algebraically that
ex = px (1 + ex+1 )
solution
∞
X
ex = k px
k=1
∞
X
= px + k px
k=2
X∞
= px + px .j px+1 from the consistency condition
j=1
" ∞
#
X
= px 1 + j px+1
j=1
= px [1 + ex+1 ]
Intuitively, this is saying that the life expectancy for a person now aged x is one
year more than their expectancy when they reach age x + 1, provided that they do
survive to age x + 1.
2.1.8 Initial and Central rates of Mortality
qx is called the initial rate of mortality because it is the probability that a life, alive
at age x (the initial time) dies before age x+1. An alternative , often used (especially
in demography) is the central rate of mortality, denoted mx . By definition,
qx
mx = R 1
p dt
0 t x
It represents the probability that a life, live between ages x and x + 1 dies. The
R1
denominator 0 t px dt is interpreted as the amount of time spent alive between ages
x and x + 1 by a life alive at age x. qx can also be represented by the integral
29
R1
p
0 t x
µx+t dt implying that
R1
p µx+t dt
0 t x
mx = R1
p dt
0 t x
Note that mx is a measure of the rate of mortality over the year from exact age x to
exact age x+1, whereas the force of mortality, µx , is a measure of the instantaneous
rate of mortality at exact age x. mx is estimated by statisticians in the form
Number of deaths
mx =
Total time spent alive and at risk
called “occurrence or exposure rates”
2.2 The Life Table
A life table is a bookkeeping system used to keep track of mortality. We could

track a cohort of lives and at the end of each year, we keep track of the number of
survivors. In actuarial science and demography, a life table (also called a mortality
table or actuarial table) is a table which shows, for each age, what the probability is
that a person of that age will die before his or her next birthday. From this starting
point, a number of inferences can be derived such as:-
• the probability of surviving any particular year of age
• remaining life expectancy for people at different ages
A life table is in effect a means of presenting a set of values of qx usually for all
integral ages running from a certain minimum age (which might be 0 or some
other age).
Definitions:
1. qx = the probability that a life aged x will die before reaching age x + 1
2. px = the probability that a life aged x survives to age x+1. Obviously px +qx = 1
30
3. n qx = the probability that a life aged x will die before reaching age x + n
4. n px = the probability that a life aged x survives to age x + n. similarly n px +

n qx = 1
5. ω is called the limiting age or the maximum age of a population. It is the

youngest age to which no one survives or the first age at which the value of lx
becomes negligible.
6. m|n
qx is the probability that (x) survives m years, and then dies in the sub-
sequent n years, that is, between ages x + m and x + m + m. This is called
a deferred mortality probability, because it is the probability that death
occurs in some interval following a deferred period. For n = 1, m|n qx is written
as m| qx
Life tables are usually constructed separately for men and for women because
of their substantially different mortality rates. Other characteristics can also be
used to distinguish different risks, such as smoking status, occupation, and socio-
economic class.
Consider a large group, or “cohort”, of males, for example, who were born on the
same day. If we could follow the cohort from birth until all members died, we could
record the number of individuals alive at each birthday (say x) and the number
dying during the following year. The ratio of these is the probability of dying at age
x, usually denoted by qx . It turns out that once the qx ’s are all known the life table
is completely determined. Consider table 1
The columns of the table, from left to right, are:
x: age
lx , “the survivor-ship function”: the number of persons alive at age x. For exam-
ple of the original 100,000 U.S. males in the hypothetical cohort, l50 = 89,867 (or
89.867%) live to age 50. l0 , is called the “radix” of the table, arbitrarily set to
100,000. It is an arbitrary positive number. lω = 0 or
for x ≥ ω, lω = 0
31
x lx dx qx mx Lx Tx ex
0 100000 1039 0.01039 0.01044 99052 7182893 71.8
1 98961 77 0.00078 0.00078 98922 7083841 71.6
2 98883 53 0.00054 0.00054 98857 6984919 70.6
3 98830 41 0.00042 0.00042 98809 6886062 69.7
4 98789 34 0.00035 0.00035 98771 6787252 68.7
5 98754 30 0.00031 0.00031 98739 6688481 67.7
6 98723 27 0.00028 0.00028 98710 6589742 66.7
7 98696 25 0.00026 0.00026 98683 6491032 65.8
8 98670 22 0.00023 0.00023 98659 6392348 64.8
9 98647 19 0.00020 0.00020 98637 6293689 63.8
10 98628 16 0.00017 0.00017 98619 6195051 62.8
20 97855 151 0.00155 0.00155 97779 5211251 53.3
50 89867 566 0.00630 0.00632 89584 2370098 26.4

51 89301 615 0.00689 0.00691 88993 2280513 25.5
75 51387 2822 0.05492 0.05649 49976 482656 9.4

76 48565 2886 0.05943 0.06127 47121 432679 8.9
80 36750 3044 0.08283 0.08646 35228 261838 7.1
90 9878 1823 0.18460 0.20408 8966 38380 3.9
100 528 177 0.33505 0.40804 439 1190 2.3
Table 1: Abbreviated decennial life table for U.S. Males.
since life tables are assumed to obey a mathematical formula implying that lx con-
verges asymptotically to zero.
dx : number of deaths in the interval (x, x + 1) for persons alive at age x, computed
as dx = lx − lx+1 . For example, of the l50 = 89, 867 persons alive at age 50, d50 =
l50 − l51 = 89867 − 89301 = 566 died prior to age 51. n dx denotes the number of people
which died between ages x and x + n, i.e., n dx = lx − lx+n .
ω−1
X
for x < ω, lx = dy , (dω = 0 of course)
y=x
qx : probability of dying at age x. Also known as the (age-specific) risk of death. By

32
dx d 566
construction, qx is also equal to . Thus, for example, q5 = 50 = = 0.00630.
lx l50 89, 867
mx : the mortality rate or the central death rate at age x. Generally these quantities
are estimated from the data, and are the sole input to the life table. That is, all
other quantities are determined once the mx ’s are specified. By construction, mx =
dx
, the number of deaths at age x divided by the number of person-years at risk
Lx
at age x. Note that the mortality rate, mx , and the probability of death, qx , are not
identical. For a one year interval they will be close in value, but mx will always be
larger. It is also known as the central death rate.
Lx : total number of person-years lived by the cohort from age x to x + 1. This

is the sum of the years lived by the lx+1 persons who survive the interval, and
the dx persons who die during the interval. The former contribute exactly 1 year
each, while the latter contribute, on average, approximately half a year, so that
Lx = lx+1 + 0.5 × dx . This approximation assumes that deaths occur, on average, half
way in the age interval x to x +1. Such is satisfactory except at age 0 and the oldest
R1
age, where other approximations are often used. Lx is also given as Lx = 0 lx+t dt
Tx : total number of person-years lived by the cohort from age x until all members
of the cohort have died. This is the sum of numbers in the Lx column from age x to
R∞
the last row in the table. Also Tx = x lx+t dt
Tx
ex : the (remaining) life expectancy of persons alive at age x, computed as ex = .
lx
T50 2, 370, 099
For example, at age 50, the life expectancy is e50 = = = 26.4.
l50 89, 867
Notes
1. Life expectancy is not the same as median survival time, the latter being the
time at which only 50% of a cohort are still alive. For example, of the 100,000
persons alive at age 0, 51,387 are alive at age 75, and 48,565 are alive at
age 76. The median survival time at birth (age 0) is thus between 75 and 76
additional years (and can be shown to be 75.5), while the life expectancy at
birth is e0 = 71.8 additional years.
2. The calculation of life expectancy for a person should not be confused with
33
predicting their survival time. While newborn U.S. males have a life ex-
pectancy of 71.8 years, any given U.S. male may die tomorrow or live to age
100. One need not predict actual survival times in order to compute life ex-
pectancy (the average survival time).
If we consider the life table, we can see that
lx+n
n px =
lx
and
lx+m − lx+m+n
m|n
qx = = m px − m+n px
lx
Example 2.11. Complete the entries in the following table:
Agex lx dx px qx
0 100000
1 99523
2 89123
3 87174
4 86234
5 85346
Table 2: Life table example 1
Age lx Age lx Age lx Age lx

0 100000 25 98246 50 93735 75 66605
5 99202 30 97776 55 91357 80 53925
10 99129 35 97250 60 88038 85 38329
15 99036 40 96517 65 83114 90 22219
20 98709 45 95406 70 76191 95 9419
Table 3: Life table example 2
Example 2.12. Using table 3, and given that l36 = 97126 find:
(i) l10
(ii) d35
(iii) 5 d10
34
(iv) The probability that a newborn will die before reaching 50 years.
(v) The probability that a newborn will live more than 60 years.
(vi) The probability that a newborn will die when his age is between 45 years and
65 years old.
(vii) The probability that a 25 year old will die before reaching 50 years.
(viii) The probability that a 25 year old will live more than 60 years.
(ix) The probability that a 25 year old will die when his age is between 50 years
and 65 years old.
(x) 5 p20 , 5 p40 , 5 p60 , 5 p80 ; check whether 5 p20 > 5 p40 > 5 p60 > 5 p80
2.3 The select mortality table
Before being accepted for life assurance cover, potential policyholders are often
required to undergo a medical examination to satisfy the insurer that they are in
a ‘reasonable’ level of health. Lives who fail to satisfy the requirements laid down
by the insurance company will often be refused cover (or required to pay a higher
premium for the same level of cover). As a result of this filtering, lives who have
recently been accepted for cover can be expected to be in better health (and, thus,
experience lighter mortality) than the general population at the same age. This
effect is known as selection (i.e. the process of choosing lives for membership of a
defined group, rather than random sampling).
However, as the duration since selection increases, the extent of the lighter mortal-
ity experienced by the select group of lives can be expected to reduce (as previously
healthy individuals are exposed to the same medical conditions as the general pop-
ulation). In practice, select lives are often assumed to experience lighter mortality
for a period of, say, s years (known as the select period). However, once the duration
since selection exceeds the select period, the lives are assumed to experience the
ultimate mortality rates appropriate for the general population at the same age.
35
Thus, we now consider the construction and application of a select life table, where
mortality varies by age and duration since selection.
Examples of selection include:
(a)temporary initial selection: Which is exercised by a life assurance company in

deciding whether or not to accept a person for life assurance cover. Selection
takes place by producing satisfactory medical evidence and it is known as
underwriting
(b)self selection: Which is exercised by lives when choosing to purchase an annuity

(i.e. exchanging a capital sum for the receipt of an income for life)
These are examples of positive selection, where the select lives are likely to expe-
rience lower mortality rates than the general (or ultimate) population of the same
age for a specified duration since selection only. However, a life retiring early on
grounds of ill-health is likely to experience higher mortality than the ultimate pop-
ulation of the same age. This is an example of negative selection.
2.3.1 Select, ultimate and aggregate mortality rates
Most select life tables are constructed to explore the effect of temporary initial
selection (i.e. where selected lives experience lighter mortality than the general
population studied for a specified duration since selection).
Suppose that the select period is s years. Consider a life who is currently of exact
age (x + r) and who was selected at age x. Thus, the duration since selection is r
years. Now, if r < s then we expect the life to experience lower mortality than the
ultimate population at the same age and we define the select mortality rate at
age x + r as follows:
q[x]+r =pr[a life now aged (x + r) who joined the select group at age x dies before
attaining age (x + r + 1)]
Note that [x] is used to denote the age at selection and r is the duration since
selection, so that the current age of the life is (x + r). Thus, as the life is expected
36
to experience lower mortality than an ultimate life of the same age, we have:
q[x]+r < qx+r for r<s
Similarly, consider another life who is also currently of exact age (x + r) , but who
was selected at age (x + 1). Thus, in this case, the duration since selection is (r − 1)
years. We define the select mortality rate at age (x + r) for this life as follows:
q[x+1]+(r−1) =pr[a life now aged (x + r) who joined the select group at age (x + 1)
dies before attaining age (x + r + 1)]
Note that, in this case, [x + 1] is used to denote the age at selection and (r − 1)
is the duration since selection, so that the current age of the life is also (x + r).
However, as this life has been selected more recently, we would expect this life
to experience lighter mortality over the year of age (x + r) to (x + r + 1) than the
life selected at age x. Thus, we have:
q[x+1]+(r−1) < q[x]+r for r < s
However, if r ≥ s, then we expect lives of the same age who were selected s or more
years previously have the same rates of mortality, regardless of age at selection. In
this case, all lives selected s or more years previously will experience the rates of
mortality of the ultimate population at the same age.
Then, for lives of age (x + 2) and select durations of 2 years or more, we have:
q[x]+2 = q[x−1]+3 = q[x−2]+4 = . . . = qx+2
However, for select durations of less than two years, we have:
q[x+1]+1 < qx+2 and q[x+2] < q[x+1]+1 < qx+2
consider table 4 (an extract of A1967-70 table) The convention is that each row
represents how mortality rates change as duration since selection increases. Thus,
for a life selected at age 60, denoted by [60], the rate of mortality in the year of
age 60 to 61 is q[60] and the rate of mortality in the year of age 61 to 62 is q[60]+1 .
37
age [x] q[x] q[x]+1 qx+2 x+2

60 0.00669904 0.00970168 0.01774972 62
61 0.00723057 0.01055365 0.01965464 63
62 0.00779397 0.01146756 0.02174310 64
63 0.00839065 0.01244719 0.02403101 65
64 0.00902209 0.01349653 0.02653550 66
Table 4: Extract 1 of A1967-70 life Table
However, two years after selection, the lighter mortality experienced as a result of
selection is assumed to wear off, and the rate of mortality experienced in the year
of age 62 to 63 is simply that of the ultimate population at the same age, q62
Thereafter, the life is assumed to be an ultimate life and so, for any duration
since selection r ≥ 2, the rate of mortality experienced in the year of age (x + r) to
(x + r + 1) is qx+r .
Also, the rates displayed on the upwards diagonal represent the rate of mortal-
ity experienced by lives of the same age but with a different duration since selec-
tion.
2.4 Construction of Life Tables
Choose a starting age for the table, denoted by α, and an arbitrary radix, denoted
by lα . α can be zero or any other value.
Then, for all ages x > α, we calculate recursively the values of lx using lx+1 =
lx × (1 − qx ) and determine the values of dx using dx = lx − lx+1 . When completed,
this gives the ultimate portion of the table.
Suppose that the select period is s years. Using a deterministic interpretation of

the life table, we use the l[x]+r for r < s to denote the number of lives who are alive
at age (x + r)from an initial group of l[x] lives selected at age [x]. Then, we calculate
38
the values of l[x] , l[x]+1 , l[x]+2 , . . . , l[x]+(s−1) recursively using:
l[x]+s lx+s lx+s

p[x]+(s−1) = ≡ ⇒ l[x]+(s−1) =
l[x]+(s−1) l[x]+(s−1) 1 − q[x]+(s−1)
l[x]+(s−1) l[x]+(s−1)
p[x]+(s−2) = ⇒ l[x]+(s−2) =
l[x]+(s−2) 1 − q[x]+(s−2)
..
.
l[x]+1 l[x]+1
p[x] = ⇒ l[x] =
l[x] 1 − q[x]
And, then we calculate d[x]+r for r = 0, 1, 2, . . . , (s − 1) using d[x]+r = l[x]+r − l[x]+r+1 .

For example, another extract from the A1967-70 select life table is shown below in
table 5.
age [x] l[x] l[x]+1 lx+2 x+2

60 29615.936 29417.538 29132.138 62
61 29130.898 28920.265 28615.051 63
62 28600.975 28378.059 28052.632 64
63 28023.708 27788.571 27442.681 65
64 27396.808 27149.632 26783.206 66
Table 5: Extract 2 of A1967-70 life Table
Example 2.13. Suppose a mortality table is represented by the function

√
lx = 1, 000 100 − x
find
(i) The probability of a newborn surviving to age 19
(ii) The probability of a life aged 36 dying before attaining age 51
Example 2.14. Suppose that the life table function is given by lx = 10, 000(100−x)2
for 0 ≤ x ≤ 100
(i) Compute the survival function for newborn lives.

39
(ii) Compute the survival function for lives currently aged 20.
(iii) State the limiting age for the population
Example 2.15. Given that 2 q52 = 0.07508 and considering the life table below, de-
termine d51
x lx dx
50 1000 20
51
52 35
53 37
2.5 Multiple Decrement Tables
Like a life table, a multiple decrement table (MDT) represents the progress of a
model population as it advances from age to age. Multiple decrement models are
extensions of standard mortality models whereby there is simultaneous operation
of several causes of decrement. The population is subject to more than one ‘mode’
of decrement. A life fails because of one of these decrements. For example we could
have a MDT representing the progress of a population of bachelors which decrease
because of both death and marriage as shown in the double decrement table 6.
Age x No. of bachelors Deaths Marriages

20 100,000 488 4,988
21 94,524 479 4,998
22 89,047 468 4,974
23 83,605 455 4,920
Table 6: Multiple Decrement Table 1
MDT’s with more than two modes of decrement can also be met. Examples of modes
of decrement include:
• life insurance contract is terminated because of death/survival or withdrawal

(lapse).
40
• an insurance contract provides coverage for disability and death, which are
considered distinct claims.
• life insurance contract pays a different benefit for different causes of death
(e.g. accidental death benefits are doubled).
• pension plan provides benefit for death, disability, employment termination

and retirement among others.
Consider a company where the population of male employees might suffer losses
because of death, withdrawal (i.e. resignation) or transfer to an overseas branch as
shown in table 7.
Age No. of employees Deaths Withdrawals Overseas Transfers

(τ )
x (al)x or lx (ad)dx or d(1)
x
(ad)w
x
or d(2)
x
(ad)tx or d(3)
x
20 100,000 452 5,517 2,569
21 91,462 433 4,780 2,431
22 83,818 414 4,136 2,302
23 76,966 402 4,076 2,264
2.5.1 Conventional notation:
• lx(τ ) represents the surviving population present at exact age x.
• d(j)
x
represents the number of lives exiting from the population between ages
x and x + 1 due to decrement j.
• d(τ
x
)
denote the total number of exits by all modes between ages x and x + 1.
Thus m
X
(τ )
dx = d(j)
x
j=1
where m is the total number of possible decrements, and therefore,
d(τ
x
)
= lx(τ ) − lx+1
(τ )
41
• The probability that a life (x) will leave the group within one year as a result
of decrement j is given by:
d(j)
qx(j) = (τx )
lx
• The probability that (x) will leave the group (regardless of decrement) is given
by:
m m
d(τ ) X d(j) X
qx(τ ) = (τx ) = x
(τ )
= qx(j)
lx l
j=1 x j=1
• The probability that (x) will remain in the group for at least one year is given
by:
(τ )
(τ ) (τ )
lx+1 lx(τ ) − d(τ )
px = 1 − qx = (τ ) = x
lx lx(τ )
• We also have the probability of remaining in the group after n years
(τ )
(τ )
lx+n
n px = (τ ) = p(τ
x
)
× p(τ )
× p(τ )
× . . . × p(τ )
lx x+1 x+2 x+n−1
and the complement

n qx(τ ) = 1 −n p(τ
x
)
• The number of failures due to decrement j over the interval [x, x+n] is
n−1
X
(j)
n dx = d(j)
x+t
t=0
These relationships should be straightforward to follow:
n d(j)
x
= lx(τ ) ×n qx(j)
and
n d(τ
x
)
= lx(τ ) ×n qx(τ )
The MDT in table 7 can be expanded to look like table 8 below.
Example 2.16. Using the previously given multiple decrement table, compute and
interpret the following:
42
x lx(τ ) d(1)
x
d(2)
x
d(3)
x
qx(1) qx(2) qx(3) qx(τ ) p(τ
x
)
20 100,000 452 5,517 2,569 0.00452 0.05517 0.02569 0.08538 0.91462

21 91,462 433 4,780 2,431 0.00473 0.05226 0.02658 0.08358 0.91642
22 83,818 414 4,136 2,302 0.00494 0.04935 0.02746 0.08175 0.91825
23 76,966 402 4,076 2,264 0.00522 0.05296 0.02942 0.08760 0.91240
(i) 2 d(3)
21
(ii) 3 p(τ
20
)
(1)
(iii) 2 q11
(2)
(iv) 1|2 q20
2.5.2 Fractional Age Assumptions
Life table functions such as lx , px or µx are usually tabulated at integer ages only,
but sometimes e may need to compute probabilities involving non-integer ages
or durations, such as 2.5 p37.5 . Such probabilities can be approximated using two
methods:-
(i) The assumption of uniform distribution of deaths.
(ii) The assumption of constant force of mortality.
2.5.3 Uniform Distribution of Deaths (UDD)
This method is based on the assumption that, for integer x, and 0 ≤ t ≤ 1, the
function t px µx+t is constant. Since this is the density (pdf) of time to death from
age (x), this assumption is equivalent to a uniform distribution of time to death.
In other words for an individual aged exactly (x), the probability of dying on one
particular day over the next year is the same as that of dying on any other day over
the nest year.
43
The UDD assumption implies that µx+t is an increasing function of t, since it

assumes that t px µx+t is constant and yet we know that t px is a decreasing function
of t. Therefore µx increases over the year of age.
Rs
Since s qx = 0 t px µx+t dt, by putting s = 1, we mist have
t px µx+t = qx for (0 ≤ t ≤ 1) (2.29)
And if t px µx+t is constant throughout the year, then

Z s
s qx = qx dt = s.qx (2.30)
0
Since qx can be obtained from the table, we can use this approximation for any s qx
or s px for (0 ≤ s ≤ 1)
Note: We must have an integer x in the formula s qx = s.qx so we can use this
formula on problems such as 0.5 p58 . For problems such as 0.5 p55.5 , this approximation
is not applicable. We use the formula
(t − s)qx
t−s qx+s = for (0 ≤ s ≤ t ≤ 1) (2.31)
1 − s.qx
The equation 2.31 can be proved as below:-
t−s qx = 1 − t−s px+s

p
= 1 − t x since by the principle of consistency t px = s px .t−s qx+s
s px
1 − t qx 1 − t.qx
=1− =1−
1 − s qx 1 − s.qx
(1 − s.qx ) − (1 − t.qx )
=
1 − s.qx
(t − s)qx
=
1 − s.qx
Example 2.17. Show that under the UDD assumption over each year of age, for
x = 0, 1, 2, . . . , ω − 1 and for 0 ≤ t < 1
(i) lx+t = lx − t.dx

44
(ii) lx+t = (1 − t)lx + t.lx+1
Solution
lx+t = lx .t px
= lx (1 − t qx )
= lx (1 − t.qx )
= lx − t.(lx qx )
= lx − t.dx for (i) and
= lx − t(lx − lx+1 )
= (1 − t)lx + t.lx+1 for (ii)
2.5.4 Constant Force of Mortality
For integer x and 0 ≤ t ≤ 1, we suppose that µx+t = µ a constant. Then using the
formula
Z t
t px = exp − µx+s ds = e−tµ (2.32)
0
we can find the required probabilities.

Note that px = e−µ so µ = − log px , which we can find from the tables. Hence, for
0 ≤ s ≤ t < 1, we have
Z t
t−s px+s = exp − µx+r dr = e(t−s)µ (2.33)
s
from which we can calculate any required probabilities.

Example 2.18. Calculate 3 p62 1 based on PFA92C20 table using
2
(i) the UDD assumption
(ii) the constant force of mortality assumption.
Solution
We split up the probability as follows:
45
3p62 1 = 1 p62 1 ∗ 2 p63 ∗ 1 p65 .

2 2 2 2
From the PFA92C20 table,
l65 9, 703.708
2 p63 = = = 0.992617
l63 9, 775.88
(i) Under the UDD assumption
p62 1 − q62 1 − 0.002885

1 p62 1 = = 1 = = 0.998555
2 2
1 p62 1 − 2 .q62 1 − 0.5 ∗ 0.002885
2
1 p65 = 1 − 1 q65 = 1 − 0.5 ∗ q65 = 1 − 0.5 ∗ 0.004681 = 0.997660

2 2
which gives 3 p62 1 = 0.99855 ∗ 0.992617 ∗ 0.997660 = 0.988863

2
(ii) Under the constant force of mortality assumption

1 12 1 1
1 p62 1 = exp − µ = e−µ = (p62 ) 2 = (1 − 0.002885) 2 = 0.998556
2 2 2
1 1
1 p65 = (p65 ) 6 2 = (1 − 0.004681)6 2 = 0.997657
2
which gives 3 p62 1 = 0.998556 ∗ 0.992617 ∗ 0.997657 = 0.988861

2
2.6 Some Analytical Laws of Mortality
It is often useful to be able to express qx or µx as some simple mathematical function

of age x. Some of the simplest such laws are the following
2.6.1 The uniform or De Moivre’s Law
Let X be a uniform random variable on the interval [a, b]. The PDF of this random
variable is 
1

b−a
for a ≤ x ≤ b and
f (x) =
 0 elswhere
46
This is a two-parameter model with parameters a and b. If X is the age-at-death

random variable, we take a = 0a and b = ω then the PDF is

1

ω
for a ≤ x ≤ b and
f (x) =
 0 elswhere
In the actuarial context, this survival model is known as De Moivre’s Law which
was proposed in 1729 and states that, for all ages x such that 0 ≤ x < ω, we have:
1
µx = (2.34)
ω−x
Thus, as expected, the force of mortality is an increasing function of age. Then, we

can derive the survival function as follows:
Z x+t
t px = exp − µs ds
x
Z x+t
1
= exp − ds
x ω−s
= exp [ln(ω − s)]x+t

x
ω − (x + t)
=
ω−x
Example 2.19. Consider the uniform distribution model as defined above. Find
F0 (x), S0 (x) and µx .
solution
Zx Zx
1 h s ix x
F0 (x) = f0 (s)ds = ds = =
ω ω 0 ω
0 0
ω−x
S0 (x) = 1 − F0 (x) =
ω
f0 (x) 1 ω−x 1
µx = = / =
s0 (x) ω ω ω−x
47
2.6.2 Gompertz’ Law
Gompertz’ Law was proposed in 1829 and was based on the observation that, over
a large range of ages, the function µx is log-linear. Thus, for all ages , we have:
µx = BC x (2.35)
Then, assuming that the underlying force of mortality follows Gompertz’ Law, the
parameter values and can be determined given the value of the force of mortality
at any two ages. To ensure that the force of mortality is a non-negative increasing
function of age, we require that the parameter values and are such that B > 0 and
c > 1. We can derive the survival function as follows:
Z t
t px = exp − µx+s ds
0
Z t Z t
x+s x s
= exp − Bc ds = exp − Bc c ds
0 0
Z t
x s ln c
= exp − Bc e ds
0

−B x s ln c t
= exp c e 0
ln c

−B x t
= exp c c −1
ln c
−B

Now, if we define an auxiliary parameter g such that g = exp ln c
then we can
express the survival function as:
x t
px = exp ln(g)cx ct − 1 = g c (c −1)

t (2.36)
In practice, Gompertz’ Law is often found to be a reasonable approximation for the

force of mortality at older ages.
Example 2.20. For a force of mortality µx that is known to follow Gompertz’ law,
calculate the parameters B and c given that µ50 = 0.017609 and µ55 = 0.028359
Solution:
1/6
Bc55

µ55 0.028359
= = c5 =
µ50 Bc50 0.017609
48
producing c = 1.1 and B = 0.00015
Example 2.21. Show that if mortality experience conforms to Gompertz’ law, then:

log c
− log(− log px ) = log − x log c
B(c − 1)
Suggest how this property could be used.

Solution
We can start from the formula
Z 1
px = exp − µx+t dt
0
taking logs and changing signs gives

Z 1 Z 1 Z 1
x+t x
− log px = µx+t dt = Bc dt = Bc ct dt
0 0 0
integrating by writing the integrand as an exponential function gives
1 1
et log c Bcx (c − 1)
Z
x t log c x
− log px = Bc e dt = Bc =
0 log c 0 log c
taking logs and changing signs again gives

log c
− log(− log px ) = − log B − x log c − log(c − 1) + log log c = log − x log c
B(c − 1)
The R.H.S is a linear function of x. This means that we can estimate the param-
eters B and c by fitting a straight line to the graph of − log(− log px ) plotted as a
function of x. The slope of the line willh be − log
i c from which c can be estimated.
log c
The intercept on the y-axis will be log B(c−1) from which B can be estimated.
2.6.3 Makeham’s Law
Makeham’s Law was proposed in 1860, and incorporated the addition of a constant
term in the Gompertz’ expression for the force of mortality. The rationale behind
this is that an age-independent allowance is required for the incidence of accidental
49
deaths. Thus, for all ages x ≥ 0, we have
µx = A + Bcx (2.37)
Then, assuming that the underlying force of mortality follows Makeham’s Law, the
parameter values A, B and c can be determined given the value of the force of
mortality at any three ages. To ensure that the force of mortality is a non-negative
increasing function of age, we require that the parameter values A, B and c are
such that A ≥ −B, B > 0 and c > 1. We can derive the survival function using the
same approach adopted above for Gompertz’ Law to obtain:
x (ct −1)
t px = st g c (2.38)
−B

where s = exp(−A) and g = exp ln c
.
50
2.7 Problem set 1
1. In a certain population, the force of mortality equals 0.025 at all ages (assume
no upper age). Calculate
(a) The probability that a new born will survive to age 5.

(b) The probability that a life aged exactly 10 will die before age 12.
(c) The probability that a life aged exactly 5 will die between ages 10 and
12.
(d) The complete expectation of life for a new born.
(e) The curtate expectation of life for a new born.
2. A mortality table, which obeys the Gompertz’ law approximately for older
ages has l50 = 90, 000 l70 = 70, 000 and l90 = 20, 000.
Estimate the values of l60 and l80 and hence estimate the probability that a
life aged 60 will survive 20 years.
3. Which of the following groups has the highest level of mortality (assumed
constant)?
(a) A population with a force of mortality of 0.4% p.a.

(b) A population with an initial rate of mortality of 0.4%
(c) A population maintained at a constant of 10,000 in which there were 20
deaths during a 6-month period.
(d) A closed group consisting initially of 10,000 individuals in which there
were 20 deaths during a 6-month period.
4. If µx = 0.01908 + 0.001(x − 70) for x ≥ 55, calculate 5 q60
5. If the distribution of Tx follows De Moivre’s law, find e◦0 and var(Tx )
6. Mortality of a gropu of lives is assumed to follow Gompertz’law. Calculate µx

for a 30-year old and a 70 year old, given that µx is 0.003 for a 50 year old and
0.01 for a 60 year old.
7. On a particular mortality table, e45 = 40.20 and e46 = 39.27, calculate q45 .
51
8. Express q30 , e30 and 5 p35 in terms of probabilities of the random variable k30 ,
which represents the curtate future life time of a life aged exactly 30.
9. Calculate the complete and curtate expectation of life for an animal subject
to a constant force of mortality of 0.05 p.a.
10. You are given that p80 = 0.988. Estimate 0.5 p80 assuming:
(a) a uniform distribution of deaths between integer ages

(b) a constant force of mortality between integer ages.
52
3 Mortality Data
3.1 Determination of Exposure to Risk by the Census Method
A rate might be defined as ‘the relative number in an aggregate who acquire the
characteristic x during the interval of time h’, but this is not precise, since the size
of the given aggregate may change during the time interval h by the operation of
other factors than that being measured. In calculating a death rate (x is death) we
generally find that not only is the population under investigation subject to decre-
ment other than by death but that it is also subject to increment. For example,
in the investigation of mortality by life offices there are new policies as well as
surrenders, lapses and maturities of existing policies; and the investigation of the
general mortality of the population of a country there are births, emigrants and
immigrants. In order to obtain a measure of the death rate for a population or
aggregate of this type it is necessary to approximate, and it is in this way that the
concept of ‘risk time’ and ‘exposed to risk’ arises.
In order to construct a life table, a set of rates qx for all integral ages x is re-
quired,and, if the life table is to serve any useful purpose, the set of rates qx must
be derived from measurement of an actual population.
3.1.1 Assumptions made in estimating risk time
suppose a count has been conducted and let
• lx′ = the number attaining exact age x during the investigation period.
• nx+r = the number of new entrants at exact age x + r during the period,
0 < r < 1.
• wx+r = the number of withdrawals at exact age x + r during the period, 0 <
r < 1.
53
• θx = the number of deaths occurring out of these groups of lives between

exact age x and exact age x + 1 while they are under observation, i.e. within
the investigation [period; deaths of new entrants after entry are included but
deaths of withdrawals after withdrawal are not included.
• qx′ = the observed rate of mortality between exact age x and exact age x + 1.
′
Then, if 1−r qx+r is the rate of mortality at age x + r, for the time period 1 − r, i.e.
′ ′
′ lx+r − lx+1
1−r qx+r = ′
lx+r
then
X X
θx = lx′ qx′ + ′
nx+r .1−r qx+r − ′
wx+r .1−r qx+r
r r
′
In order to solve this equation , we assume that 1−r qx+r = (1 − r)qx′ , which implies
that
1 1 1
= r. + (1 − r) .
lx+r lx+1 lx
We may then write
" #
X X
θx = qx′ lx′ + (1 − r)nx+r − (1 − r)wx+r
r r
So
X
(1 − r)nx+r
r
is the sum of the fractional parts of the year of age during which the new entrants
were at risk. similarly the deduction of
X
(1 − r)wx+r
r
means that the withdrawals are also dealt with in a similar way since they will
have been included in either lx′ or r nx+r and have to be removed from the risk
P
for those parts of the year of age x and x + 1 after the dates of withdrawals. what
is left in the brackets in the equation above is the sum of risk time contributed by
54
lx′ +
P
r nx+r lives during the interval of age x to age x + 1. We can also see that
θ
qx′ = P P .
lx′ + r (1 − r)nx+r − r (1 − r)wx+r
the denominator is denoted by Ex and is called the initial exposed to risk. the
word ‘initial’ indicates that Ex is equivalent to the number of ‘starters’ at exact age
x that would generate θx deaths on the assumption that there were no withdrawals
or new entrants between exact ages x and x + 1
3.1.2 The Concept of mx
Alternatively, one can consider the central death rate. This is the average mortality
over the age interval x to x + 1. In life table notation this is given by
dx dx
mx = R 1 =
lx+t dt Lx
0
If this notion is now transferred to a real population rather that a life table popu-
lation, the we can write
θx
m′x = P P P
lx′ + r (1 − r)nx+r − r (1 − r)wx+r − r (1 − r)θx+r
where the deaths are dealt with the same way as the withdrawals. This denom-
inator is referred to as the central exposed to risk and is denoted by Exc . The
relationship between the initial exposed to risk Ex and the central exposed to risk
Exc is
θx
Ex ≈ Exc +
2
55
3.1.3 Derivation of rates from an actual population
Suppose that a set of values of qx′ is required based on mortality experienced during
1 January 1970 to 31 December 1974, and that the population comprises the policy
holders of a life office. suppose that each of the following information is given on a
card for each life that was at risk during the period of investigation.
• Date of birth
• Date of entry into assurance
• Date of withdrawal
• Date of death.
The dates of withdrawal and death are given if these occurred between 1 Jan-
uary 1970 and 31 December 1974. The information from the cards is summarized
in the table below. For simplicity we assume that the population is limited to four-
teen lives
Case Date of Birth Date of entry Date and mode of exit

to assurance from observation
A 1 April 1918 1 July 1954 1 Oct 1973 -Withdrawal
B 27 Jan 1919 25 June 1959 31 Dec 1974 -Existing
C 25 Dec 1915 26 Dec 1971 25 June 1972 -Withdrawal
D 1 July 1919 1 May 1971 1 Nov 1974 -Death
E 1 Dec 1916 2 Feb 1953 31 Dec 1974 -Existing
F 7 June 1920 2 Jan 1970 31 Dec 1974 -Existing
G 9 Sept 1916 10 Nov 1960 26 Jan 1972 -Death
H 1 Mar 1917 7 April 1969 10 Mar 1970 -Withdrawal
I 21 Oct 1917 28 Aug 1973 6 Feb 1974 -Death
J 18 Feb 1913 10 May 1964 31 Dec 1974 -Existing
K 10 Oct 1918 17 Nov 1973 31 Dec 1974 -Existing
L 15 April 1917 16 April 1970 10 Nov 1974 -withdrawal
M 25 Jan 1915 4 Oct 1969 3 April 1970 -Withdrawal
N 20 June 1919 22 July 1974 31 Dec 1974 -Existing
Table 9: Period of investigation 1 January 1970 to 31 December 1974
Take life A for example, born on 1 April 1918 and entering the assurance on ! July
56
1954 and finally withdrawing on 1 October 1973. This particular life was at risk
for 3.75 years, i.e. 1 January 1970 to 1 October 1973. those still existing at the end
of the study are also called the ‘enders’ and can be denoted by ex . similarly, those
under observation as on 1 January 1970 (the beginning of the investigation period)
are called ‘beginners’ and denoted by bx . the new entrants are denoted by nx and
those who withdraw are denoted by wx where x is the age at classification. Taking
care of all the fractional ages is a difficult task especially in reality where we are
dealing with a really large population so we have to make some assumptions.
Assuming uniform distribution this data can be classified as follows:-
Case Classification 49 50 51 52 53 54 55 56 57 58 59 60 61
1 1
A b51 w55 2
1 1 1 2
1 1
B b50 e55 2
1 1 1 1 2
C n56 w56 0
1 1
D n51 θ55 2
1 1 1 2
1 1
E b53 e58 2
1 1 1 1 2
1 1
F n49 e54 2
1 1 1 1 2
1 1
G b53 θ55 2
1 2
1 1
H b52 w53 2 2
1 1
I n55 θ56 2 2
1 1
J b56 e61 2
1 1 1 1 2
1 1
K n55 e56 2 2
1 1
L n53 w57 2
1 1 1 2
1 1
M b54 w55 2 2
N n55 e55 0
1
Totals Exc 2
1 12 3 4 12 6 7 5 21 3 12 2 21 1 21 1 1 1
2
Table 10: aggregate experience. Life year. Central exposed to risk
the total exposures add up to 38.
3.1.4 An Alternative method of calculating Exposed to Risk
Assuming that lx′ is continuous just as lx is a continuous function, we can have:-

Let f (x + r, t)dr denote the number of lives between age x + r and x + r + dr at time
57
t. Then Z 1
f (x + r, t)dr
0
will be the number of lives aged x last birthday at time t. This can be denoted by
Px (t) such that Z 1
Px (t) = f (x + r, t)dr.
0
Consider a period of investigation 0 to T and assume that Exc is required with re-
spect of this period. at any time t, the population at risk for the age interval x to
x + 1 is that population such that if a member died at time t he would be included
in θx , the deaths during 0 to T at age x last birthday. At time t each life in this
population will contribute an element of risk time dt to Exc , and hence the total risk
from age x to age x + 1 during the period 0 to T is
Z T Z T Z 1
Px (t)dt = f (x + r, t)drdt
0 0 0
Notice however that Z T

′
f (x + r, t)dt = lx+r .
0
Hence Z 1
′
Exc = lx+r dr
0
which can be given approximately as
1
Exc = (lx′ + lx+1
′
)
2
3.1.5 The Census Method
Recall that Z T
Exc = Px (t)dt
0
If Px (t) is approximately linear over t each calender year it follows that
T −1
1 X 1
Exc = Px (0) + Px (t) + Px (T )
2 1
2
58
Example 3.1. Estimate Exc by census method based on the following data:-
Calender year population

1998 46, 233
1999 42, 399
2000 42, 618
2001 42, 020
Table 11: population aged 55 last birthday on 1 January
Solution
Using the census approximation, the central exposed to risk for the period 1.1.98
to 1.1.01 is
c 1 1 1
E55 = [P55 (1.1.98) + P55 (1.1.99)] + [P55 (1.1.99) + P55 (1.1.00)] + [P55 (1.1.00) + P55 (1.1.01)]
2 2 2
1 1
= P55 (1.1.98) + P55 (1.1.99) + P55 (1.1.00) + P55 (1.1.01)
2 2
1 1
= (46, 233) + (42, 399) + (42, 618) + (42, 020)
2 2
= 129, 143.5
3.2 Graduation Methods and Applications
3.2.1 Graduation of observed mortality rates
The crude mortality rates derived from a mortality investigation will not be the
final rates that are published for use in actuarial calculations. These rates will
have to pass through a further process called graduation.
Definition 3.1 (Graduation). Graduation refers to the process of using statistical

techniques to improve the estimates provided by the crude rates.
Most age distributions are usually corrected by methods of graduation or smooth-

ing. The aims of graduation are:-
• To produce a smooth set of rates that are suitable for a particular purpose.
59
• To reduce the random sampling errors as much as possible.
• To use information available from adjacent ages- to improve reliability of the

estimates.
Therefore graduation results in a “smoothing” of the crude rates. However, gradu-

ated age distributions normally provide only plausible patterns or rough approxi-
mations of the expected age distribution for two reasons:-
• Genuine irregularities in the age composition, which may exist before the
graduation, may be obliterated in the graduated figures.
• The choice of the technique for graduation is largely arbitrary.
Graduation also cannot remove any bias arising from faulty data collection or oth-
erwise.
Suppose that a number of individuals, N , are each subject to the possibility that
a certain event will occur within a given time-period. Assume that the probability
that this event will occur in any given case is q, and that the random variables
associated with the events are independent. Let the number of instances in which
the event occurs be denoted by θ. Under the above assumptions, we have
θ ∼ binomial(N, q)
The ratio
θ
q̂ =
N
is referred to as the crude incidence rate (or crude occurrence rate) of the
event under consideration. N is referred to as the exposed to risk (i.e. the number
of individuals exposed to the risk that the event being considered will happen), and
is often denoted by E. If the “true” incidence rate q depends on some parameter x
(say age), then we write Ex , θx , qx , and q̂x for E, θ, q, and q̂ respectively. Thus
θx ∼ binomial(Ex , qx )
60
and
θx
q̂x =
Ex
Suppose further that a mortality investigation has yielded the following data set:
{θx , Ex : x = x1 , x2 , . . . , xr }
where
θx =number of deaths between ages x and x + 1

Ex =exposed to risk between ages x and x + 1
and x1 , x2 , . . . , xr are consecutive integers. The crude q-type rates are therefore
θx
q̂x = (x = x1 , x2 , . . . , xr )
Ex
We shall now consider how to construct a table of estimates of qx from the data.
Graduation is the process of finding suitable estimates and their properties. Let a
set of graduated values of {q̂x } be denoted by {qx◦ }. Note that {qx◦ } are not the same
as the true values {qx }, but they are generally closer to the true values that the
crude rates, {q̂x }.

Similar considerations hold for the central death rates, (mx ) or µx+ 1 , in which
2
case the data set is (θx , Exc : x = x1 , x2 , . . . , xr ), where Exc is the central exposed to
risk between ages x and x + 1. Exc ∼ = mid-year population aged x last birthday.
The crude death rates are
θx
m̂x or µ̂x+ 1 = (x = x1 , x2 , . . . , xr )
2 Exc
n o
and the set of graduated rates is written as {m◦x } ◦
or µx+ 1 .
2
n o
The crude estimates {q̂x } or µ̂x+ 1 will progress more or less roughly, i.e. it is
2
unlikely that the crude estimates will progress smoothly. In large part, this is be-
cause they have each been estimated independently and hence suffer independent
61
sampling errors. The smaller the sample size, the less smoothly the crude esti-
mates are likely to progress. For several reasons we would prefer to work with {qx }
or {µx }, which are smoothed functions of age. Therefore we graduate
n or smooth
o the
crude estimates, to produce a set of graduated estimates {qx◦ } and µ◦x+ 1 , that do
2
progress smoothly with age.
A purely practical reason for smoothing mortality data is that we will use the
life table to compute financial quantities such as premiums for life assurance con-
tracts. It is very desirable that such quantities progress smoothly with age, since
irregularities (jumps or other anomalies) are hard to justify in practice. We could
calculate these quantities using our crude mortality rates, and then smooth the
premium rates etc. directly, but it’s much more convenient to have smoothed mor-
tality rates to begin with.
3.2.2 Criteria for a good graduation
These can be summarized as:-
• smoothness
• adherence to data
• suitability for the purpose at hand.
(i)Smoothness Vs Adherence to data

smoothness and adherence to data are usually conflicting requirements. Per-
fect smoothness (extreme example: a straight line) pays no attention to the
data, while perfect adherence to data means no smoothing at all.
If the graduation results in values that are smooth but show no adherence to
the data, then we say the data may be over graduated. on the other hand,
where insufficient smoothing has been carried out, it results to under grad-
uation.
62
Figure 1: Examples of over graduated and under graduated values
n o
The set of the graduated values, {qx◦ } or µ◦x+ 1 should progress reasonably
2
smoothly as the age x increases. This is because they are expected to be near
the true rates {qx } or {µx }, which are likely to be smooth as age increases
for biological reasons. (Exceptions may occur: for example at the ages when
young people (especially men) are first allowed to drive motor vehicles, there
may be a large rise in the death rates because of the effects of motor vehicle
accidents). Another reason for requiring a smooth set of graduated rates is
that, in commercial practice, premiums
n and
o reserves are expected to progress
◦ ◦
smoothly with age, and if {qx } or µx+ 1 were not smooth this requirement
2
might not be met.
(ii) Suitability for the purpose at hand

The suitability of a graduation for practical purposes depends very much on
what that work is, and can only be addresses in particular cases. However,
two very important observations are:-
(a) In life insurance work, losses result from premature deaths (benefits are paid
sooner than expected), so we must not underestimate mortality.
(b) in pensions or annuity work, losses result from delayed deaths (benefits are
paid for longer than expected -this means a higher income for individuals
with a lower expected life time), so we must not overestimate mortality.
However, the graduation should “fit the data”.

63
3.2.3 Methods of Graduation
There are many methods that can be applied. The following three methods are
examples of methods that can be employed.
(i) The Graphical Method

This involves fitting a curve near the points {q̂x }. The graphical method may
be employed when data is very scanty, and allowances may be made for spe-
cial features of the data (e.g. suspected inaccuracies due to data errors at
some ages). It is a simple exercise but results depend greatly on the skill and
experience of the investigator (and are possibly subject to bias). The results
of a graphical graduation are likely to be fairly smooth, although the values
of {qx◦ } are likely to be available to only a small number of decimal places, and
this may lead to slight irregularities in {qx◦ } or its differences {∆qx◦ }.
Advantages:
(a) May be applied when data is very scanty.
(b) May be applied quickly without the need for a computer.
(c) Special features (e.g. suspected data errors or other assumptions such as the
general shape of life table values) may be allowed for.
Disadvantages:
(a)Subject to bias of investigator.
(b)Skill and experience are required for good results.
(c)Smoothness may not be ideal.
(ii) Graduation by a mathematical formula

This is the most important method of graduation for standard tables. Con-
sider the graduation of {q̂x }. the idea is to find a function qx (α) of age x and
an s-vector of parameters α which is such that
qx = qx (α◦ ) (x = x1 , x2 , . . . , xr )
64
where α◦ is unknown. The vector α◦ is estimated by the method of maximum

likelihood, which gives the vector α̂. The graduated rates are defined as
qx◦ = qx (α̂) (x = x1 , x2 , . . . , xr ).
There is generally no need to test for smoothness, since {qx (α̂)} will almost
certainly progress smoothly as x increases, because of the form of the function
qx (α) (for a given α).
Advantages:
(a) Allows for a wide choice of formulae.
(b) Parameters are chosen “efficiently” in the statistical sense (by MLE etc.).
(c) Smoothness is automatic

Disadvantages:
(a) Suitable only for large data sets e.g. when constructing new standard tables.
(b) It may be difficult to find a family of curves which fit well.
(c) The method requires computers etc. and may take some time.
(iii) Graduation by reference to a standard table

This is a special case of “graduation
n o by a mathematical formula”. A standard
T
table with values qx or µTx+ 1 is chosen, and it is postulated that (for
2
example)
qx (α) =α1 qxT

.
or qx (α) =(α1 + α2 x)qxT
qx (α) is linked to qxT by means of simple functions of α (the true value of which
α◦ is unknown). It is not necessary to test for smoothness as it may be as-

sumed that the standard table qxT is smooth.
Advantages:
(a) Its easier to carry out than a new graduation.
(b) Parameters are may be found “efficiently”, and usually quite easily.
65
(c) Smoothness is guaranteed (since the standard table will be smooth).
(d) May be used when data is fairly scanty.

Disadvantages:
(a) There must be a suitable standard table.
(b) The choice of the formula linking the new table to the standard table may be
open to debate.
3.2.4 Testing the smoothness of a good graduation
The tests for smoothness will be used so as to check for under graduation. It is
possible to fit a higher order polynomial to any set of observed data. The fitted
polynomial will be smooth in mathematical sense if it is differentiable many times,
but it may not progress smoothly from age to age. Smoothness is measured in
terms of differentiability. Consequently, we seek a more rough-and-ready measure
of smoothness having regard to the scale with which we work (usually the year of
age).
Smoothness is usually
n o assessed by looking at third differences of the graduated
◦ ◦
quantities {qx } or µx+ 1 , i.e. by the measure
2
r −3
xX
|∆3 qx◦ |.
x=x1
For example:-
The first difference: ◦

∆qx◦ = qx+1 − qx◦
◦ ◦
The second difference: ∆2 qx+1 = ∆qx+1 − ∆qx◦
The third difference: ∆3 qx◦ = ∆2 qx+1
◦
− ∆2 qx◦
The third difference measures the change in curvature. The criterion of smooth-
◦
ness o used is such that the third differences of the graduated quantities {qx }
n usually
or µ◦x+ 1 should:-
2
66
(a) be small in magnitude compared with the quantities themselves; and
(b) progress regularly.
Example: Compare the smoothness of the rates in graduations A and B (that were
used to produce the graphs above) over the range 40 to 45.
Age (x) Exposed to risk (Ex ) No. of deaths (dx ) Crude rates (q̂x ) graduated rates (qx◦ )
40 67, 300 62 0.000921 0.001056
41 65, 368 63 0.000964 0.001167
42 65, 391 84 0.001285 0.001290
43 62, 917 86 0.001367 0.001426
44 66, 537 120 0.001804 0.001576
45 62, 302 121 0.001942 0.001742
Table 12: Graduation A
Age (x) Exposed to risk (Ex ) No. of deaths (dx ) Crude rates (q̂x ) graduated rates (qx◦ )
40 67, 300 62 0.000921 0.000867
41 65, 368 63 0.000964 0.001018
42 65, 391 84 0.001285 0.001215
43 62, 917 86 0.001367 0.001494
44 66, 537 120 0.001804 0.001701
45 62, 302 121 0.001942 0.001945
Table 13: Graduation B
67
Solution: We find the third differences by calculating successive differences i.e.

(∆qx◦ = qx+1
◦
− qx◦ , ∆2 qx◦ = ∆qx+1
◦
− ∆qx◦ , ∆3 qx◦ = ∆2 qx+1
◦
− ∆2 qx◦ ), gives:-
(x) (qx◦ ) ∆qx◦ ∆2 qx◦ ∆3 qx◦

40 0.001056 0.000111 0.000012 0.000001
41 0.001167 0.000123 0.000013 0.000001
42 0.001290 0.000136 0.000014 0.000002
43 0.001426 0.000150 0.000016
44 0.001576 0.000166
45 0.001742
Table 14: Graduation A
For graduation A, the third differences are very small (in fact as small as rounding
permits) which indicates that the graduated rates are very smooth.
(x) (qx◦ ) ∆qx◦ ∆2 qx◦ ∆3 qx◦

40 0.000867 0.000151 0.000046 0.000036
41 0.001018 0.000197 0.000082 −0.000154
42 0.001215 0.000279 −0.000072 0.000109
43 0.001494 0.000207 0.000037
44 0.001701 0.000244
45 0.001945
Table 15: Graduation B
For graduation B, the third differences are relatively large (with a total magnitude
of 0.000299), which indicates the rates are irregular.
68
4 Monetary Functions
Insurance systems are established to reduce the adverse financial impact of some
types of random events. Within these systems, individuals and organizations adopt
utility models to represent preferences, stochastic models to represent uncertain fi-
nancial impact and economic principles to guide pricing.
Due to the long term nature of insurances, the amount of investment earnings up
to the time of payment provides a significant element of uncertainty.
The models for these monetary functions depend on the time to death of the in-
sured. The models are built in terms of functions of T , where T is a random vari-
able representing the insured’s future life (the time until death).
4.1 Life Insurance
A life assurance is a contract or policy which promises to pay a specified sum, say S
on the death of a given life (the life assured) at any future time, or between certain
specified dates. The death benefit S is called the sum assured. The benefit may
be level of varying (increasing or decreasing) in a manner specified in the contract.
Policies under which payments depend on the death or survival of more than one
life may also be issued.
(i) Whole life assurances

A whole life assurance is a policy providing a certain sum assured, S on the
death of (x) at any future date. The amount and time of payment of a life
insurance benefit will depend only on the length of the time interval from the
issue of the insurance to the death of the insured. The elapsed time from pol-
icy issue to the death of the insured is the insured’s future life time random
variable T = T (x)
Suppose that, when calculating the value of benefits, the life office assumes
that it’s funds will earn interest at a constant rate, i per annum. The corre-
sponding force of interest per annum is
δ = loge (1 + i)
69
and the present value of 1 due at time t years is
v t = (1 + i)−t = e−δt
4.1.1 Payable at the end of the year of death
An assurance payable at the end of the year of death pays the benefit S at the
end of the year of death (years being measured from the date of issue of the
policy). Though the actual practice is to pay the death benefit as soon as the
death claim is made, as soon as legal formalities are completed, it’s simpler
to first consider a case where the benefit is paid at the end of the policy year
of death i.e. on the anniversary of effecting the policy, after death.
Suppose that a death benefit of S is payable at the end of the year of death
and T represents the future life time of (x). Recall that
t qx = prob {T ≤ t}
which is F (t), i.e. the distribution function of T . Therefore, f (t), the probabil-
ity distribution function (p.d.f) of T , i.e. P r {T = t} =t |qx (t = 0, 1, 2, . . .)
and the present value of the assurance is say Z such that
Z = g(T ) = Sv T +1 .
The mean (expected) present value of the whole life assurance benefit is
given by
∞
X
E(Z) = g(t)f (t)dt
t=0
where f (t) is the p.d.f of T (for t > 0) and g(t) is the value of the benefit; hence
∞
X
E [g(t)] = Sv t+1 .t |qx
t=0
=SAx
The present value of a benefit of 1 payable at the end of the year of death (no
matter when the death occurs) of a life aged x is denoted by the symbol Ax
70
For life insurance,the expectation of the present value random variable Z is

also known as the net single premium.
Therefore
Ax =vqx + v 2 1 |qx + v 3 2 |qx + . . . + v ω−x ω−x−1 |qx

∞
X
= v t+1 .t px .qx+t
t=0
∞
X
= v t+1 .t |qx
t=0
∞
1 X
= v t+1 .dx+t
lx t=0
We define commutation functions. Commutation functions reduce the compu-

tations in calculations involving both interest and mortality functions. Com-
mutation functions are tabulated in the Actuarial tables.
Dx = v x lx
Cx = v x+1 .dx
and ∞
X
Mx = Cx+t .
t=0
Then multiplying Ax by v x /v x , we obtain;

∞
1 X x+t+1
Ax = x v .dx+t
v lx t=0
∞
1 X
= Cx+t
Dx t=0
Mx
=
Dx
4.1.2 The variance of the present value of benefits
Recall that Z = g(t) = Sv t+1 is the p.v. of S due at the end of the year of death
of (x) and the mean of Z = SAx .
71
Therefore
var(Z) = E(Z 2 ) − [E(Z)]2

∞
X
= S 2 v 2(t+1) t |qx − (SAx )2
t=0
" ∞
#
X
∗ t+1
=S 2
(v ) t |qx − (Ax ) 2
(where v ∗ = v 2 )
t=0
2
A∗x − (Ax )2

=S
where the rate of interest i∗ is such that
1 1
v∗ = ∗
= v2 =
1+i (1 + i)2
i.e.
i∗ = 2i + i2 .
Note The force of interest δ ∗ , corresponding to i∗ is such that
∗
v ∗ = e−δ = v 2 = e−2δ
i.e.
δ ∗ = 2δ.
The standard deviation of Z is found by getting the square root of var(Z).

If we consider a block of n identical whole life policies on independent lives
aged x, the total present value of the assurance benefits is
Z = Z1 + Z2 + . . . + Zn
where, for i = 1, 2, . . . , n,
Zi = present value of benefit under ith policy

(i)
= Sv T
72
where T (i) = the future life time of ith life. Since the variables {T i } are as-
sumed to be independent, the variance of Z is
n
X
Var(Z) = Var(Zi )
i=1
Xn
S 2 A∗x − (Ax )2

=
i=1
= nS 2 A∗x − (Ax )2

and so the standard deviation of the total present value is

√ p
S n [A∗x − (Ax )2 ].
Also the mean present value of a group of policies is the sum of their seperate
M.P.V’s, even if the lives are not independent.
4.1.3 Payable at the moment of death
Suppose the death benefit S is payable at the moment of death of a life aged
x. If T is the future life time of (x), then the present value of this benefit, say
Z is
Z = g(T ) = Sv T .
The mean of this variable, E(Z), is called the mean (expected) present
value (M.P.V) of the whole life assurance benefit, and is
Z ∞
E(Z) = g(t)f (t)dt
0
where f (t) is the p.d.f of T (for t > 0) and g(t) is the value of the benefit; hence
Z ∞
E(Z) = S v t .t px .µx+t dt.
0
If the value of the benefit S = 1, then we have

Z ∞
E(Z) = v t .t px .µx+t dt = Āx .
0
73
Where Āx is the present value for a whole life assurance of amount 1 payable
at the moment of death.
The probability of (x) dying between ages x + t and x + t + δt tends to t px .µx+t
as δt tends to 0, and discounting this by the term v t , we obtain
Z ∞
Āx = v t .t px .µx+t dt.
0
Alternatively, recall that
t qx = prob[T (x) ≤ t]; t ≥ 0
t px = 1 −t qx = prob[T (x) > t]; t ≥ 0.
Therefore t qx is the distribution function of T (x) and recall that t px is the

survival function of(x).
Note:
X
F (x) = P (X ≤ x) = f (x) for − ∞ < x < ∞
t≤x
where f (x) is the probability distribution of X at t.

f (x) = P (X = x)= the probability distribution of X.
Therefore
t qx = pr[T (x) ≤ t] is the distribution of T (x)
and the p.d.f, say
f (t) =F ′ (t)
d
= qx
dt t
d
= [1 −t px ]
dt
d S(x + t)
= 1−
dt S(x)
where S(x) = P r[T (x) > t] for t ≥ 0. Therefore
S ′ (x + t)
f (t) = −
S(x)
74
multiplying by S(x + t)/S(x + t), we obtain
S ′ (x + t)

S(x + t)
f (t) = −
S(x) S(x + t) .
=t px µx+t t ≥ 0
R∞
Note 0 t px .µx+t dt = 1.
Therefore Z ∞
Āx = v t .t px .µx+t dt.
0
and for a sum assured of S, the M.P.V = S Āx

We now define commutation functions
Z 1 Z 1
x+t
C̄x = v lx+t µx+t dt = Dx+t µx+t dt
0 0
and ∞
X
M̄x = C̄x+t
t=0
so multiplying Āx by v x /v x to achieve symmetry, we obtain

R∞
v x+t lx+t µx+t
0
Āx = dt
v x lx
C̄x + C̄x+1 + C̄x+2 + . . .
=
Dx
M̄x
=
Dx
4.1.4 The variance of the present value of benefits
Recall that Z = g(t) = Sv t is the p.v. of S due at the moment of death of (x)
and the mean of Z = S Āx .
75
Therefore
var(Z) = E(Z 2 ) − [E(Z)]2

Z ∞
= S 2 v 2t t px µx+t dt − (S Āx )2
0
Z ∞
∗ t
=S 2
(v ) t px µx+t dt − (Āx ) 2
(where v ∗ = v 2 )
0
2
Ā∗x − (Āx )2

=S
where ∗ indicates at the rate of interest 2i + i2 .
4.1.5 The relationship between Ax and Āx
Assuming that deaths occur on average midway through each policy year (i.e.
the year between two consecutive policy anniversaries), benefits payable at
the end of the year of death will be received, on average, 6 months later than
those immediately payable on death. This thus gives the approximate rela-
tionship
1
Āx ≈ (1 + i) 2 Ax
Theorem 4.1. If uniform distribution of deaths is assumed between the ages

x + k and x + k + 1 (for k = 0,1,2,. . . ), we have
i
Āx = Ax
δ
Proof : Since
∞
1 X
Āx = C̄x+k
Dx k=0
and ∞
1 X
Ax = Cx+k ,
Dx k=0
we can show that
i
C̄y = Cy for y = x + k(k = 0, 1, 2, . . .).
δ
76
If we assume UDD between ages y and y + 1, then
ly+t µy+t = dy (0 ≤ t < 1).
Therefore
Z 1
C̄y = v y+t ly+t µy+t dt
0
Z 1
y
= dy v v t dt
0
y 1−v .
= dy v
δ

y+1 i
= dy v
δ
i
= Cy
δ
From financial mathematics, if i is small we have
i 1 1
≈ 1 + i ≈ (1 + i) 2 .
δ 2
So we can also make the following approximations for small i
1
Āx = (1 + i)Ax
2
1
C̄x = (1 + i)Cx
2
1
M̄x = (1 + i)Mx
2
(ii) Assurances payable at the end of the m1 of a year of death.

Suppose that a benefit of 1 is payable at the end of the m1 year (measured
from the issue date) following the death of (x); for example if m = 12, the sum
assured is payable at the end of the month of death.
The mean present value of this benefit is
∞
r+1
X
A(m)
x = v m r 1
| qx .
m m
r=0
77
(m) (m)
As m → ∞, we see that Ax → Āx as expected. However the function Ax is
not often used in practice.
(iii) Temporary and deferred Assurances

A term or temporary assurance contract pays the death benefit on the
death of (x), within the term of the policy, usually written as n years.
If the sum assured is payable at the end of the year of death (if this occurs
within the n year term), then we have the present value given as:-

Sv T +1 if T < n
Z = g(T ) =
0 if T ≥ n
The value of a temporary assurance of amount 1 payable if (x) dies within n

years is given by the symbol
n−1
X
1
n Ax = Ax:n = v t+1 t |qx
t=0
n−1
1 X
= v t+1 .dx+t
lx t=0 .
n−1
1 X
= Cx+t
Dx t=0
Mx − Mx+n
=
Dx
The stress on the x signifies the fact that the life aged x must expire before
the benefit is paid. the status of x must fail before the status of n i.e. the fixed
period for the benefit to be paid.
If the benefit is payable at the moment of death, the present value as a ran-
dom variable is:- 
Sv T if T ≤ n
Z = g(T ) =
0 if T > n
78
The Mean present value of this assurance is denoted by the symbol

Z n
1
Āx:n = v t t px µx+t dt.
0
Using commutation functions, this is given by
n−1
1 1 X M̄x − M̄x+n
Āx:n = C̄x+t = .
Dx t=0 Dx
A deferred assurance contract pays a death benefit on the death of (x)

if this occurs after a certain period called the period of deferment, which is
usually written as n years.
If the benefit is payable at the end of the year of death (if this occurs after the
deferred period has elapsed), the present value is:-

Sv T +1 if T ≥ n
Z = g(T ) =
0 if T < n
The mean present value is then given as
n |Ax = v n n px Ax+n .
So
∞
X
n |Ax = v t+1 t |qx
t=n
∞
1 X
= Cx+t .
Dx t=n
Mx+n
=
Dx
If the benefit is payable at the moment of death (if this occurs after the defer-
79
ment period has elapsed), the present value is:-


Sv T if T > n
Z = g(T ) =
0 if T < n
and the ,mean present value is then given as
n |Āx = v n n px Āx+n .
So
Z ∞
n |Āx = v t t px µx+t dt
n
∞
1 X
= C̄x+t .
Dx t=n
M̄x+n
=
Dx
(iv) Pure endowments and endowment assurances

A pure endowment contract is one where a payment made on a certain date
in the future to a life aged x if the life is then alive. This contract only has a
survival benefit. The present value of this benefit is

v n if T ≥ n
Z=
0 if T < n
and hence the M.P.V. is given as
v n n px .
Since
P r {T ≥ n} = n px ,
The M.P.V. is written as
n Ex = Ax:n1 = Āx:n1 = v n n px .
80
Therefore
lx+n v x+n lx+n Dx+n
Ax:n1 = v n .n px = v n = x
=
lx v lx Dx
An endowment assurance is a contract that has both a survival and a
death benefit. The benefit is payable if (x) dies within n years or on maturity
of the contract at time n years, whichever comes first. An endowment assur-
ance is therefore a combination of a term assurance and a pure endowment
assurance (of the same term).
If the benefit is payable at the end of the year of death, the present value is

Sv T +1 if T < n
Z = g(T ) =
Sv n if T ≥ n
and the mean present value of a benefit of amount 1 is:-
1
Ax:n =Ax:n + Ax:n1
Mx − Mx+n + Dx+n .
=
Dx
If the benefit is payable at the moment of death, the benefit is


Sv T if t < n
Z = g(T ) =
Sv n if t ≥ n
and the mean present value is therefore
1
Āx:n =Āx:n + Āx:n1
M̄x − M̄x+n + Dx+n .
=
Dx
However, the relationship between the benefit payable at the moment of death
and that payable at the end of the year of death IS NOT GIVEN BY
1
Āx:n ≈ (1 + i) 2 Ax:n ,
81
but rather the correct relationship is
1
1
Āx:n ≈ (1 + i) 2 Ax:n + Ax:n1
1
The factor (1+i) 2 is also called the acceleration factor. It accelerates payments
from the end of the year of death to the moment of death.
(iv) Varying assurances
1. Consider a case under a whole life contract where a death benefit of β(t)
in year t is payable at the end of the year of death. The present value is
given by
Z = g(T ) = β(T + 1)v T +1 (T = 0, 1, 2, . . .)
and hence the M.P.V is

∞
X
β(t + 1)v t+1 t |qx .
t=0
When β(t) = t for t = 1, 2, . . . , the M.P.V is written as (IA)x . Therefore we

have ∞
X
(IA)x = (t + 1)v t+1 t |qx .
t=0
In terms of commutation functions we have

∞
1 X
(IA)x = (t + 1)Cx+t
Dx t=0
Mx + Mx+1 + . . .
=
Dx
Rx
=
Dx
where ∞
X
Rx = Mx+t
t=0
2. Suppose the contract provides a sum of β(t) immediately on thr death of
82
(x) at time t years. The present value of the benefit is given by
Z = g(T ) = β(T )v T (T > 0).
If β(t) = t for all t > 0, then the M.P.V is written as (I¯Ā)x . This is the
mean present valus of a whole life assurance with sum assured propor-
tional to duration so that the benefit at time t is t, i.e.
Z ∞
(I¯Ā)x = tv t t px µx+t dt
0
Z ∞
1
= tDx+t µx+t dt
Dx 0
3. If β(t) = [t] + 1, where [t] indicates the integer part of t. The M.P.V is
written as (I Āx )x . This represents a benefit payable at the moment of
death of 1 in the first year, 2 in the second year and so on.
Therefore
∞
X
(I Āx )x = t |Āx
t=0
∞
1 X
= M̄x+t
Dx t=0
R̄x
=
Dx
where ∞
X
R̄x = M̄x+t
t=0
83
4.2 Life Annuities
4.3 Benefit Premiums
4.4 Provisions
5 Joint Life Benefits
84

Am 1 Notes

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Am 1 Notes

Uploaded by

Copyright:

Available Formats

ACTUARIAL MATHEMATICS I LECTURE

1 Introduction to Life Insurance 1

1.1 Traditional Insurance contracts . . . . . . . . . . . . . . . . . . . . . . 1

1.1.1 Term insurance . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1.2 Whole life insurance . . . . . . . . . . . . . . . . . . . . . . . . 2

1.1.3 Endowment insurance . . . . . . . . . . . . . . . . . . . . . . . 2

1.1.4 With-profit insurance . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Modern insurance contracts . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2.1 Universal life insurance . . . . . . . . . . . . . . . . . . . . . . 3

1.2.2 Unitized with-profit . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2.3 Equity-linked insurance . . . . . . . . . . . . . . . . . . . . . . 4

1.6 Other insurance contracts . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.7 Pension benefits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.7.1 Defined benefit pensions . . . . . . . . . . . . . . . . . . . . . . 8

1.7.2 Defined contribution pensions . . . . . . . . . . . . . . . . . . . 8

2 Survival models and the life tables 9

2.1 Survival models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.1.1 The future lifetime random variable . . . . . . . . . . . . . . . 9

2.1.2 The force of mortality . . . . . . . . . . . . . . . . . . . . . . . . 15

2.1.3 Actuarial notation . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2.1.4 The probability Density Function of Tx . . . . . . . . . . . . . . 21

2.1.5 Mean and standard deviation of Tx . . . . . . . . . . . . . . . . 22

2.1.6 Curtate future lifetime . . . . . . . . . . . . . . . . . . . . . . . 26

2.1.8 Initial and Central rates of Mortality . . . . . . . . . . . . . . 29

2.2 The Life Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2.3 The select mortality table . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.3.1 Select, ultimate and aggregate mortality rates . . . . . . . . . 36

2.4 Construction of Life Tables . . . . . . . . . . . . . . . . . . . . . . . . . 38

2.5 Multiple Decrement Tables . . . . . . . . . . . . . . . . . . . . . . . . 40

2.5.1 Conventional notation: . . . . . . . . . . . . . . . . . . . . . . . 41

2.5.2 Fractional Age Assumptions . . . . . . . . . . . . . . . . . . . . 43

2.5.3 Uniform Distribution of Deaths (UDD) . . . . . . . . . . . . . . 43

2.5.4 Constant Force of Mortality . . . . . . . . . . . . . . . . . . . . 45

2.6 Some Analytical Laws of Mortality . . . . . . . . . . . . . . . . . . . . 46

2.6.1 The uniform or De Moivre’s Law . . . . . . . . . . . . . . . . . 46

2.6.2 Gompertz’ Law . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

2.6.3 Makeham’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

2.7 Problem set 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3.1 Determination of Exposure to Risk by the Census Method . . . . . . 53

3.1.1 Assumptions made in estimating risk time . . . . . . . . . . . 53

3.1.2 The Concept of mx . . . . . . . . . . . . . . . . . . . . . . . . . . 55

3.1.3 Derivation of rates from an actual population . . . . . . . . . . 56

3.1.4 An Alternative method of calculating Exposed to Risk . . . . . 57

3.1.5 The Census Method . . . . . . . . . . . . . . . . . . . . . . . . . 58

3.2 Graduation Methods and Applications . . . . . . . . . . . . . . . . . . 59

3.2.1 Graduation of observed mortality rates . . . . . . . . . . . . . 59

3.2.2 Criteria for a good graduation . . . . . . . . . . . . . . . . . . . 62

3.2.3 Methods of Graduation . . . . . . . . . . . . . . . . . . . . . . . 64

3.2.4 Testing the smoothness of a good graduation . . . . . . . . . . 66

4.1 Life Insurance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

4.1.1 Payable at the end of the year of death . . . . . . . . . . . . . . 70

4.1.2 The variance of the present value of benefits . . . . . . . . . . 71

4.1.3 Payable at the moment of death . . . . . . . . . . . . . . . . . . 73

4.1.4 The variance of the present value of benefits . . . . . . . . . . 75

4.1.5 The relationship between Ax and Āx . . . . . . . . . . . . . . . 76

4.2 Life Annuities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

4.3 Benefit Premiums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

5 Joint Life Benefits 84

1 Examples of over graduated and under graduated values . . . . . . . 63

1 Abbreviated decennial life table for U.S. Males. . . . . . . . . . . . . . 32

2 Life table example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3 Life table example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4 Extract 1 of A1967-70 life Table . . . . . . . . . . . . . . . . . . . . . . 38