Stat 475

Life Contingencies

Chapter 4: Life insurance

Review of (actuarial) interest theory — notation

We use i to denote an annual effective rate of interest.

The one year present value (discount) factor is denoted by

v = 1/(1 + i).

i (m) is an annual nominal rate of interest, convertible m times

per year.

The annual discount rate (a.k.a., interest rate in advance) is

denoted by d.

d (m) is an annual nominal rate of discount, convertible m

times per year.

The force of interest is denoted by δ (or δt if it varies with

time).

2
Review of (actuarial) interest theory — relationships

To accumulate for n periods, we can multiply by any of the

quantities below; to discount for n periods, we would divide by any
of them.
n Period Accumulation Factors
!mn !−nr
i (m) d (r )
n
(1 + i) = 1 + = (1 − d)−n = 1− = e δn
m r

If the force of interest varies with time, we can discount from time
n back to time 0 by multiplying by
Rn
e− 0 δt dt

3
Valuation of life insurance benefits

The timing of life insurance benefits generally depends on the

survival status of the insured individual. Since the future lifetime
of the insured individual is a random variable, the present value of
life insurance benefits will also be a random variable.

We’ll commonly denote the random variable representing the PV

of a life insurance benefit by Z .
Unless otherwise specified, assume a benefit amount of $1.

We’re often interested in various properties (e.g., mean, variance)

of Z .
The mean of Z is referred to as the expected value of the
present value, expected present value (EPV), actuarial
present value (APV), or simply actuarial value.

4
Whole life insurance — Benefits paid at moment of death

The first type of life insurance we’ll consider is whole life

insurance.
Consider the case where the benefit is paid at the moment of
death (this is sometimes referred to as the continuous case).

For this case, the present value of the benefit is Z = v Tx = e −δTx .

The corresponding EPV is denoted by Ax

EPV for Whole Life Insurance — Continuous Case
h i Z ∞
Ax = E [Z ] = E e −δTx = e −δt t px µx+t dt
0

5
Whole life insurance — Benefits paid at moment of death
We can calculate the second moment for Z similarly:
 2  Z ∞
−δTx
e −(2δ)t t px µx+t dt
 2
E Z =E e =
0

We can find this second moment by computing the expectation at

twice the force of interest, 2δ. When we calculate the expectation
at twice the force of interest, we denote it with the symbol 2 Ax .
Then for this case, we have E Z 2 = 2 Ax .
 

Then we can calculate the variance of Z :

2
V [Z ] = E Z 2 − E [Z ]2 = 2 Ax − Ax
 

We may also be interested in various percentiles or functions of Z

— all the usual rules of random variables apply.

6
Example
Assume that a particular individual, currently age x, has a future
lifetime described by a random variable with density
1
fx (t) = for 0 < t < 60
60
This person wants to purchase insurance that will provide a benefit
of $1 at the moment of death. Assuming a force of interest of
δ = 0.06:
1 Find the EPV and variance for this death benefit.
2 Find the minimum value H such that P(Z ≤ H) ≥ 0.9.
3 Find the minimum single premium H that an insurance
company must charge in order to be at least 90% certain that
this premium will be adequate to fund the death claim, should
their assets accumulate at δ = 0.06.

7
Whole life insurance —
Benefits paid at end of the year of death

Next we consider a whole life insurance in which the death benefit

is paid at the end of the year in which the insured dies (this is
sometimes called the annual case).

Kx is the time corresponding to the beginning of the year of death;

Kx + 1 is the end of the year of death.

Since the benefit is paid at the end of the year of death, the
present value of the benefit is Z = v Kx +1 .

Then this is a discrete random variable.

What does its pmf look like?

8
Whole life insurance —
Benefits paid at end of the year of death

We can find the mean and variance of this random variable:

EPV and Variance for Whole Life Insurance — Annual Case

h i X∞
E [Z ] = E v Kx +1 = v k+1 k |qx = Ax
k=0
 2  ∞
2 Kx +1
X
(k+1)
E [Z ] = E v = v2 k |qx = 2 Ax
k=0

V [Z ] = E Z 2 − E [Z ]2 = 2 Ax − (Ax )2
 

9
Whole life insurance example

Consider a $50, 000 whole life insurance policy issued to (x), with
death benefit paid at the end of the year of death. Let Z be the
present value of the death benefit RV.

We’re given:

qx = 0.01 qx+1 = 0.02 qx+2 = 0.03 qx+3 = 0.04 i = 10%

Find Pr [36, 000 ≤ Z ≤ 42, 000]

10
Whole life insurance — Benefits paid mth ly
Now we consider the case where the whole life death benefit is paid
at the end of the 1/m period of a year in which the insured dies.
(m) 1
For this case, the present value of the benefit is Z = v Kx +m
.

Then for this discrete random variable we have:

EPV and Variance for Whole Life Insurance — mth ly Case
h (m) 1 i X ∞
k+1 (m)
E [Z ] = E v Kx + m = v m k | 1 qx = Ax
m m
k=0
 2  ∞
2
(m)
Kx 1
+m
X
( k+1
m ) (m)
E [Z ] = E v = v2 k | 1 qx = 2 Ax
m m
k=0

(m) 2
 
(m)
V [Z ] = E Z 2 − E [Z ]2 = 2 Ax − Ax
 

11
General strategy for calculating EPV

We can always find the EPV of any life-contingent payment (not

just life insurance benefits) by summing over all possible payment
times the product of:
1 The amount of the payment
2 An appropriate present value (discount) factor
3 The probability that the payment will be made

All of the EPV formulas for life insurance benefits we’ve seen are
specific cases of this principle.

Note that this formula only works for calculating EPVs.

12
Recursion formulas
Using the summations given above, we can derive various
recursion formulas for the expected present value of life insurance
benefits, i.e., formulas relating successive values of the EPV.
We’ll encounter these types of formulas in many contexts.
These formulas are useful for a number of reasons.

For the annual case, we have

EPV of Whole Life Insurance — Annual Case
Ax = v qx + v px Ax+1

For the mth ly case, we have

EPV of Whole Life Insurance — mth ly Case
(m) (m)
Ax = v 1/m 1 qx + v 1/m 1 px Ax+ 1
m m m

13
Relating the whole life EPV values
Note that in order to calculate Ax , we only need the information in
a life table.

However, in order to calculate Ax , we need the full survival model.

If we’re not given this information (i.e., if we only have a life
table), then we’ll have to make some sort of fractional age
assumption as before.

Making the UDD assumption gives the relationships:

Relationships Between Whole Life EPV Values under UDD
UDD i (m) UDD i
Ax = Ax Ax = (m) Ax
δ i
These relationships are often used as approximations, but are only
exact under UDD.

14
Numerical EPV values for whole life insurance

(12) Note the pattern

x Ax Ax Ax
between the values for
20 4, 922 5, 033 5, 043 the three cases, at each
40 12, 106 12, 379 12, 404 given age.
60 29, 028 29, 683 29, 743
Using the UDD
80 59, 293 60, 641 60, 764 approximations, we can
100 87, 068 89, 158 89, 341 calculate approximate
(12)
values for Ax and Ax .
Table 4.3 from Dickson et al.: Then we can compare
EPV values for a whole life in- them to the actual values
surance with a death benefit shown in the table.
of 100, 000, using Makeham’s
mortality model and i = 5%.

15
SULT Whole Life Example

Let Z be the PV of a $100,000 whole life insurance (annual case)

issued to (45). Let i = 5% and mortality be given by the Standard
Ultimate Life Table (SULT).

(a) Calculate E [Z ]
(b) Calculate the standard deviation of Z .
(c) Recalculate E [Z ] for the monthly case, i.e., if the death
benefit was payable at the end of the month of death. Use the
UDD fractional age assumption.

16
Term life insurance — Benefits paid at moment of death
The next type of life insurance we’ll consider is (n-year) term life
insurance.
First consider the continuous case, where the benefit is paid
at the moment of death.

For this case, the present value of the benefit is

 T
v x if Tx < n
Z=
0 if Tx ≥ n

1
The corresponding EPV is denoted by Āx:n

EPV for n-year Term Life Insurance — Continuous Case

Z n
1
Āx:n = E [Z ] = e −δt t px µx+t dt
0

17
Term life insurance — Benefits paid at moment of death

We can also calculate the second moment for Z :

Z n
e −2δt t px µx+t dt = 2 Āx:n
 2 1
E Z =
0

Some term life insurance example problems:

1 How would you expect the EPV for a term insurance to vary
as n increases?
2 Redo the whole life example, using the same survival model as
before, but this time assuming that the person wishes to
purchase a 20-year term insurance (with benefit payable at
the moment of death) rather than a whole life insurance
policy. Compare the answers for the two insurances.

18
Term life insurance —
Benefits paid at end of the year of death

Next we consider the annual case for an n-year term life insurance.

For this case, the present value of the benefit is

 K +1
v x if Kx ≤ n − 1
Z=
0 if Kx ≥ n

1
Then the EPV is denoted by Ax:n

EPV for Term Life Insurance — Annual Case

n−1
X
1
Ax:n = E [Z ] = v k+1 k |qx
k=0

19
Term life insurance — Benefits paid mth ly
Now we consider the case where the term life death benefit is paid
at the end of the 1/m period of a year in which the insured dies.

For this case, the present value of the benefit is

( (m) 1 (m)
v Kx + m if Kx ≤ n − 1
m
Z= (m)
0 if Kx ≥ n

Then the EPV is denoted by A(m)1x:n

EPV for Term Life Insurance — mth ly Case

nm−1
X
A(m)1x:n = E [Z ] = v (k+1)/m k | 1 qx
m m
k=0

20
Term life insurance example and relationships
A person age x wishes to purchase a 3-year term life insurance
policy with benefit amount $400, 000 payable at the end of the
year of death.

Find the EPV of this benefit, assuming that

px = 0.97 px+1 = 0.96 px+2 = 0.94 i = 0.10

It turns out that the relationships between EPV values derived for
whole life insurance also work for term life insurance.
They’re exact under UDD and approximations otherwise.

Relationships Between Term Life EPV Values under UDD

1 UDD i 1 UDD i
Āx:n = Ax:n 1
A(m)1x:n = (m) Ax:n
δ i

21
Pure endowment

A pure endowment is a type of contract that pays a benefit at

the end of a fixed time period (e.g., n years) if the policyholder is
still alive at that time.

This type of policy is not typically sold by itself, but is nonetheless

important:
It can be combined with other types of insurance.
It can be used to find the EPV of life contingent payments.

For this case, the present value of the benefit is


0 if Tx < n
Z= n
v if Tx ≥ n

22
Pure endowment

The EPV of an n-year pure endowment is denoted by Ax:n1

Note that there are not separate continuous and mth ly cases
for a pure endowment.
The alternate (more convenient) notation n Ex is also used to
denote the EPV of an n-year pure endowment.

We will commonly use n Ex as a sort of general “life-contingent

discount factor”.

EPV for Pure Endowment

Ax:n1 = n Ex = E [Z ] = v n n px

23
SULT Pure Endowment Example

Let Z by the PV of a $100,000 10-year pure endowment issued to

(45). Let i = 5% and mortality be given by the Standard Ultimate
Life Table (SULT).

(a) Calculate E [Z ]
(b) Redo part (a) using i = 9%

24
Endowment insurance
Endowment insurance combines term insurance with a pure
endowment.
An n-year endowment insurance pays a benefit if the insured
dies within n years.
It also pays a benefit (of the same amount) at the end of n
years if the person is alive at that point.

In the case where the death benefit portion is paid at the moment
of death, the present value of the entire benefit (assuming as usual
a benefit amount of $1) is
 T
v x if Tx < n
Z=
vn if Tx ≥ n
= v min(Tx ,n)

25
Endowment insurance
For the continuous case, the EPV is denoted Āx:n , which we can
write in terms of other EPV symbols:
EPV of Endowment Insurance — Continuous Case
1
Āx:n = E [Z ] = Āx:n + Ax:n1

For the annual case, the PV of the benefit is

 K +1
v x if Kx ≤ n − 1
Z= n
v if Kx ≥ n
= v min(Kx +1,n)

Then the EPV for this case is denoted by Ax:n

EPV of Endowment Insurance — Annual Case
1
Ax:n = E [Z ] = Ax:n + Ax:n1

26
Endowment insurance

For the mth ly case, the PV of the benefit is

( (m) 1 (m)
v Kx + m if Kx ≤ n − 1
m
Z= (m)
vn if Kx ≥ n
 
(m) 1
min Kx + m ,n
=v

(m)
Then the EPV for the mth ly case is denoted by Ax:n

EPV of Endowment Insurance — mth ly Case

(m)
Ax:n = E [Z ] = A(m)1x:n + Ax:n1

27
Relationships for endowment insurance

It’s important to note that the UDD relationships we developed for

the whole life and term insurance work only for death benefits, not
for endowment benefits.

Therefore, in order to apply the UDD approximation, we must first

split the term insurance benefit from the endowment portion:

Relationships Between Term Life EPV Values under UDD

UDD i (m) UDD i
Āx:n = A1 + Ax:n1 1
Ax:n = (m) Ax:n + Ax:n1
δ x:n i

As before, these relationships are exact under the UDD

assumption, and approximate otherwise.

28
SULT Term Example

Let Z by the PV of a $100,000 20-year (annual case) term

insurance issued to (45). Let i = 5% and mortality be given by the
Standard Ultimate Life Table (SULT).

(a) Calculate E [Z ]
(b) Calculate P(Z > 0)
(c) Recalculate E [Z ] for the continuous case, i.e., if the death
benefit was payable at the moment of death. Use the UDD
fractional age assumption.

29
Deferred insurance benefits
In all of the types of insurances we’ve discussed thus far, the death
benefit period starts immediately at the time of purchase.
It’s also possible to defer the coverage until some future time.

For example, consider the continuous case of a whole life

insurance. Suppose we wanted to defer this insurance for u years.
The PV of the benefit would be

0 if Tx < u
Z=
v Tx if Tx ≥ u

We denote this deferment of benefits in much the same way as we

did for a deferred mortality probability.
EPV for Deferred Whole Life Insurances
(m) (m)
u |Āx = u Ex Āx+u u |Ax = u Ex Ax+u u |Ax = u Ex Ax+u

30
Deferred insurance benefits — term insurance

Similarly, we can also consider deferred term life insurances. If we

were to defer an n year continuous term insurance by u years, the
PV of the benefit would be

0 if Tx < u or Tx ≥ u + n
Z=
v Tx if u ≤ Tx < u + n

1
The corresponding EPV is denoted by u |Āx:n

EPV for Deferred Term Life Insurance — Continuous Case

1 1
u |Āx:n = u Ex Āx+u:n

The annual and mth ly cases work similarly.

31
Deferred insurance benefits — relationships

Using the principle of benefit deferment allows us to develop some

useful relationships among the various insurance EPV values:
n−1
X
1 1 1 1 1
u |Ax:n = Ax:u+n − Ax:u Ax:n = r |Ax:1
r =0
∞
X
1 1
Ax = r |Ax:1 Ax = Ax:n + n |Ax
r =0

There are analogous continuous and mth ly versions of these

relationships as well.

32
Another SULT Term Example

Let Z by the PV of a $100,000 17-year (annual case) term

insurance issued to (45). Let i = 5% and mortality be given by the
Standard Ultimate Life Table (SULT).

(a) Calculate E [Z ]
(b) Calculate the standard deviation of Z
(c) Redo part (a) for a 30-year term product, leaving everything
else the same.

33
Variable benefits — Arithmetically Increasing
We can find the EPV for benefits with various patterns.

One pattern that deserves special mention is the case of an

arithmetically increasing benefit, one in which the benefit
increases by a constant amount each year:
Z n
¯ t e −δt t px µx+t dt

1
I Ā x:n =
0
Z ∞
I¯Ā x = t e −δt t px µx+t dt


0
n−1
X
1
(IA) x:n = (k + 1) v k+1 k |qx
k=0
∞
X
(IA)x = (k + 1) v k+1 k |qx
k=0

34
Variable benefits — Geometrically Increasing

Another pattern that could be useful is that of a geometrically

increasing benefit, one in which the benefit increases by a constant
percentage each year.

Example: Consider a 20-year term insurance issued to (x) in

which the death benefit is paid at the end of the year of death.
The amount of the death benefit is $100, 000 if the insured dies in
the first year, and rises by 3% each subsequent year. Find an
expression for the expected present value of this benefit.

We could use similar logic to find the EPV of a benefit that was
arithmetically / geometrically decreasing.

35
Valuing Insurance Benefits under a Select-and-Ultimate
Mortality Model
We can calculate present values and EPVs for all of the insurances
we’ve seen using a select-and-ultimate mortality model; we simply
have to use the correct mortality values.

For example, under a 2-year select-and-ultimate mortality model,

the EPV of a whole life insurance issued to [x] would be

A[x] = v q[x] + v 2 p[x] q[x]+1 + v 3 p[x] p[x]+1 qx+2

+ v 4 p[x] p[x]+1 px+2 qx+3 = · · ·

and we can also write

1
A[x] = A[x]:n + n E[x] Ax+n
(assuming n ≥ 2 here)

36
Example (AMLCR Exercise 4.1)

x `x Ax Assuming i = 0.06, calculate

35 100, 000.00 0.151375 1
5 E35

36 99, 737.15 0.158245 2 1

A35:5
37 99, 455.91 0.165386
3
5| A35
38 99, 154.72 0.172804
39 98, 831.91 0.180505 4 Ā35:5 assuming UDD
40 98, 485.68 0.188492

37
SOA Example Multiple Choice #3

For a special whole life insurance on (x), payable at the moment of

death:
µx+t = 0.05, t > 0
δ = 0.08
The death benefit at time t is bt = e 0.06t , t > 0.
Z is the present value random variable for this insurance at
issue.
Calculate Var [Z ]. [0.04535]

38
SOA Example Multiple Choice #4

For a group of individuals all age x, 25% are smokers (s), 75% are
nonsmokers (ns), i = 0.02

k s
qx+k ns
qx+k
0 0.10 0.05
1 0.20 0.10
2 0.30 0.15
1
Calculate 10, 000Ax:2 for an individual chosen at random from this
group. [0.1730]

39
SOA Example Multiple Choice #17

For a whole life insurance of 1 on (41) with death benefit payable

at the end of year of death, you are given:
i = 0.05
p40 = 0.9972
A41 − A40 = 0.00822
2A − 2 A40 = 0.00433
41
Z is the present value random variable for this insurance.
Calculate Var [Z ]. [0.02544]

40
SOA Example Multiple Choice #34

Assuming i = 0.03, you are given the following select and ultimate
mortality table with a select period of three years.

x q[x] q[x]+1 q[x]+2 qx+3 x +3

60 0.09 0.11 0.13 0.15 63
61 0.10 0.12 0.14 0.16 64
62 0.11 0.13 0.15 0.17 65
63 0.12 0.14 0.16 0.18 66
64 0.13 0.15 0.17 0.19 67
1
Calculate 2| A[60]:2 . [0.19]

41
SOA Example Written Answer #7
For a special deferred term insurance on (40) with death benefits
payable at the end of the year of death, you are given:
The death benefit is 0 in years 1-10; 1000 in years 11-20; 2000
in years 21-30; 0 thereafter.
Mortality follows the Illustrative Life Table.
i = 0.06.
The random variable Z is the present value, at age 40, of the
death benefits.
E [Z ] = 107.
1 Write an expression for Z in terms of K40 .
2 Calculate Pr(Z = 0). [0.75]
3 Calculate Pr(Z > 400). [0.1392]
4 Calculate Var [Z ]. [36,046]

42