Stat 475

Life Contingencies I

Chapter 5: Life annuities

Life annuities

Here we’re going to consider the valuation of life annuities.

A life annuity is regularly (e.g., continuously, annually,
monthly, etc.) spaced series of payments, which are usually
based on the survival of the policyholder.
These annuities are important for retirement plans, pensions,
structured settlements, life insurance, and in many other
contexts.

Many of the ideas and notation used in the valuation of life

annuities borrow from annuities-certain.
As with life insurance, the life-contingent nature of the
payments means that the present value of benefits is a
random variable rather than a fixed number.

2
Review of annuities-certain
Recall the actuarial symbols for the present value of n-year
annuities-certain:
n
X 1 − vn
Annuity-immediate: a n = vk =
i
k=1

n−1
X 1 − vn
Annuity-due: ä n = vk =
d
k=0
n
1 − vn
Z
Continuous annuity: ā n = e −δt dt =
0 δ
nm
(m)
X 1 k/m 1 − vn
mth ly annuity-immediate: an = v = (m)
m i
k=1

3
Review of annuities-certain — continued

Increasing annuity-immediate:
n
X ä n − nv n
(Ia)n = k vk =
i
k=1

Decreasing annuity-immediate:
n
X n − an
(Da)n = (n + 1 − k) v k =
i
k=1

For the corresponding accumulated values, we replace a by s in

each case. For example:

a n (1 + i)n = s n

4
Whole life annuity-due

Consider an annuity that pays (x) an amount of $1 on an annual

basis for as long as (x) is alive, with the first payment occuring
immediately.
This type of life annuity is known as a whole life
annuity-due.

We’ll denote the random variable representing the PV of this

annuity benefit by Y :

Y = 1 + v + v 2 + · · · + v Kx = ä Kx +1 (1)

The pmf of Y is given by:

P Y = ä k = k−1 |qx for k = 1, 2, 3, . . .

5
Whole life annuity-due EPV

The EPV of this annuity is denoted by äx and we can use our
general strategy to find this EPV:
∞
X
äx = v k k px
k=0

We could also take the expectation of the expression given in (1):

Important Relation
1 − Ax
äx = E [Y ] = so that 1 = d äx + Ax
d

This relationship actually holds between the underlying random

variables, not just the expected values.

6
Whole life annuity-due EPV, variance, and recursion
Finally, we could calculate the EPV using the standard formula for
the expectation of a discrete RV:
∞
X
äx = ä k+1 k |qx
k=0

We can find the variance of Y in a similar manner:

2A
x − (Ax )2
V [Y ] =
d2

(Note that we cannot find the second moment of Y by computing

the first moment at double the force of interest.)

Further, we have the recursion äx = 1 + v px äx+1

7
Term annuity-due
Now consider an annuity that pays (x) an amount of $1 on an
annual basis for up to n years, so long as (x) is alive, with the first
payment occuring immediately.
This type of life annuity is known as a term annuity-due.

For this type of annuity, PV of the benefit is:


ä Kx +1 if Kx ≤ n − 1
Y = ä min(Kx +1, n) =
ä n if Kx ≥ n

In this case, the EPV is denoted by äx:n

EPV for Term Annuity-Due

n−1 n−1
1 − Ax:n X X
äx:n = = v t t px = ä k+1 k |qx + n px ä n
d
t=0 k=0

8
Example

Suppose that a person currently age 65 wants to purchase a 3-year

term annuity due with annual payments of $50, 000 each. Given
that

p65 = 0.95 p66 = 0.91 p67 = 0.87 i = 7%

calculate the expected value and variance of the present value of

this annuity benefit. [132,146.91; 503,557,608.1]

Now suppose that a life insurance company sells this type of

annuity to 100 such people, all age 65, all with the same payment
amounts. Assuming these people are all independent, calculate the
expected value and variance of the total present value of all of the
annuity benefits. [13,214,691; 50,355,760,810]

9
Whole life annuity-immediate
Now we’ll turn to consider life annuities-immediate, starting with a
whole life annuity-immediate.
This type of annuity pays (x) an amount of $1 on an annual
basis for as long as (x) is alive, with the first payment
occuring at age x + 1.

We’ll denote the random variable representing the PV of this

annuity benefit by Y ∗ :

Y ∗ = v + v 2 + · · · + v Kx = a Kx

with pmf

P Y ∗ = a k = k |qx


for k = 0, 1, 2, 3, . . .

10
Whole life annuity-immediate (continued)

We have the relationship Y ∗ = Y − 1, where Y is the PV of the

whole life annuity-due and Y ∗ is the corresponding PV of the
corresponding whole life annuity-immediate.
The only difference between these two annuities is the
payment at time 0.
From this relationship, we can obtain expressions for the EPV
and variance of Y ∗
2A
x − (Ax )2
ax = äx − 1 V [Y ∗ ] = V [Y ] =
d2

We also have the recursion relation: ax = v px + v px ax+1

11
Term annuity-immediate
Now consider an annuity that pays (x) an amount of $1 on an
annual basis for up to n years, so long as (x) is alive, with the first
payment occuring at age x + 1.
This type of life annuity is known as a term
annuity-immediate.

For this type of annuity, PV of the benefit is:


∗ a Kx if Kx ≤ n
Y = a min(Kx , n) =
an if Kx > n

In this case, the EPV is denoted by ax:n

EPV for Term Annuity-Due
n
X
ax:n = v t t px = äx:n + n Ex − 1
t=1

12
Example

Assume that 1
20 Ex = 0.35, 20 qx = 0.3, and Ax:20 = 0.2.

Find:

1 Ax:20 [0.55]

2 äx:20 [13.21]

3 ax:20 [12.56]

13
SOA Practice Problem #25

Your company currently offers a whole life annuity product that

pays the annuitant 12,000 at the beginning of each year. A
member of your product development team suggests enhancing the
product by adding a death benefit that will be paid at the end of
the year of death.

Using a discount rate, d, of 8%, calculate the death benefit that

minimizes the variance of the present value random variable of the
new product. [150,000]

14
Whole life continuous annuity
Next we’ll consider continuous life annuities, starting with a whole
life continuous annuity.
This type of annuity pays (x) continuously at a rate of $1 per
year for as long as (x) is alive, starting now.

Letting Y denote the random variable representing the PV of this

annuity benefit, we have: Y = a Tx

The EPV for this annuity is denoted ax . . .

EPV for Whole Life Continuous Annuity
Z ∞
1 − Ax
ax = e −δt t px dt =
0 δ
2A − A 2


x x
. . . and the variance is V [Y ] =
δ2

15
Term continuous annuity
We can also have a continuous annuity that pays for a maximum
of n years, so long as (x) is alive.
This type of life annuity is known as an n-year term
continuous annuity.

For this type of annuity, PV of the benefit is:


a Tx if Tx ≤ n
Y = a min(Tx , n) =
an if Tx > n

In this case, the EPV is denoted by āx:n

EPV for n-year Term Continuous Annuity
Z n
1 − Āx:n
āx:n = e −δt t px dt =
0 δ

16
Example

Suppose that the survival model for (x) is described by the

exponential model with µx = µ ∀ x.

Let Y be the PV of a continuous whole life annuity payable at a

rate of 1 per year to (x).

d h
1
i
1 Find ax − (δ+µ)2
dµ
2 Assuming that δ = 0.08 and µ = 0.04, find the expected
value and variance of Y . [8.333; 13.889]

3 Find the probability that Y < 5. [0.2254]

17
SOA Practice Problem #67

For a continuous whole life annuity of 1 on (x):

Tx is the future lifetime random variable for (x).
The forces of interest and mortality are constant and equal.
āx = 12.50
Calculate the standard deviation of āTx . [7.217]

18
SOA Practice Problem #79

For a group of individuals all age x, you are given:

30% are smokers and 70% are nonsmokers.
The constant force of mortality for smokers is 0.06 at all ages.
The constant force of mortality for nonsmokers is 0.03 at all
ages.
δ = 0.08  
Calculate Var āTx for an individual chosen at random from this
group. [14.1]

19
mth ly life annuities
We can also analyze life annuities that pay on an mth ly basis.

First consider an annuity that pays an amount of 1 per year, paid

in installments of amount m1 at the beginning of each mth of a
year, so long as (x) lives.
(m) 1
(m) 1 − v Kx +m
The PV of this annuity is Y = ä =
Kx
(m) 1
+m d (m)

The EPV of Y is given by

(m) ∞
(m) 1 − Ax X 1 k
E [Y ] = äx = = vm k px
d (m) m
k=0
m

(m) (m) 1
We can also consider the “immediate” version: ax = äx −
m

20
mth ly term annuities
Now consider an annuity that pays an amount of 1 per year, paid
in installments of amount m1 at the beginning of each mth of a
year, so long as (x) lives, but for a maximum of n years.

The PV of this annuity is

 
(m) 1
min Kx + m ,n
(m) 1−v
Y = ä =
d (m)
 
(m) 1
min Kx + m ,n

The EPV of Y is given by

(m) mn−1
(m) 1 − Ax:n X 1 k
E [Y ] = äx:n = = vm k px
d (m) k=0
m m

We can also consider the “immediate” version:

(m) (m) 1 1
ax:n = äx:n + n Ex −
m m

21
Deferred annuities

As we did with life insurance, we can consider the notion of

deferring annuity benefits.
In this case, the first annuity payment would occur at some
point in the future (rather than at the beginning or end of the
first year).

We denote this deferrment in the “usual” manner, so that, for

example, the EPV of a whole life annuity due for (x), deferred u
years, would be denoted by
∞
X
u |äx = v k k px
k=u

The same idea also applies to the continuous, immediate, and

mth ly life annuity forms.

22
Deferred annuities (continued)

As we did with life insurance, we can use the idea of deferrment to

construct or deconstruct various types of annuities.

This leads to various relationships among the annuities and their

EPVs:

u |äx = äx − äx:u u |äx = u Ex äx+u

n−1
X
äx:n = äx − n Ex äx+n äx:n = u |äx:1
u=0

Again, there are analogous relationships for the other annuity

payment forms.

23
Example

You are given that:

ä30:10 = 7.79064 ä40:10 = 7.78144 ä50 = 15.1511

10 p30 = 0.99611 10 p40 = 0.99233 i = 0.06

Find:
1
10 E30 [0.5562]

2 ä30:20 [12.1187]

3 ä30 [16.7884]

24
SOA Practice Problem #55

For a 20-year deferred whole life annuity-due of 1 per year on (45),

you are given:
`x = 10(105 − x), 0 ≤ x ≤ 105
i =0
Calculate the probability that the sum of the annuity payments
actually made will exceed the EPV at issue of the annuity. [0.450]

25
Guaranteed annuities
We can also consider annuities in which some of the payments are
certain, rather than being contingent on the policyholder being
alive at the time of payment.
That is, some of the payments may be guaranteed rather
than life contingent.
This creates a sort of annuity that’s a hybrid between the ones
we’ve studied in this chapter and annuities-certain.
Any payments made after the annuitant dies would go to a
beneficiary.
This reduces the risk of a very poor return, but there’s a cost
associated with it...

For example, consider a whole life annuity due that will pay for a
minimum of n years, even if death occurs within the first n years.
This is sometimes called a “life and n-year certain” annuity.

26
Guaranteed annuities (continued)

The PV of this annuity benefit would be


ä n if Kx ≤ n − 1
Y =
ä Kx +1 if Kx ≥ n

The EPV of an whole life annuity due that guarantees the first n
payments is denoted by ä x:n

We can relate this EPV to other annuity EPVs:

ä x:n = ä n + n Ex äx+n

27
Example

A person age 65 has accumulated a sum of $100, 000 in her

retirement account. She wishes to use this money to purchase an
level payment annuity, with the first payment occurring today.

Assume that:

i = 6% 10 p65 = 0.71623 ä65 = 9.89693 ä75 = 7.21702

1 What payment amount could she afford if she purchases a

whole life annuity due? [10,104.14]

2 What payment amount could she afford if she purchases a

whole life annuity due with the first 10 years guaranteed?
[9,356.23]

28
SOA Practice Problem #88

At interest rate i:
äx = 5.6
The EPV of a 2-year certain and life annuity-due of 1 on (x)
is äx:2 = 5.6459
ex = 8.83
ex+1 = 8.29
Calculate i. [0.07886]

29
Actuarial accumulated annuity values

Thus far, we’ve mostly considered find the actuarial value of life
annuities at the time of purchase (age x or time 0).

We might also consider valuing life annuities at other points in

time.

For example, the actuarial accumulated value of an n year term

(life) annuity-due, one year after the final payment, would be

äx:n
s̈x:n = so that äx:n = n Ex s̈x:n
n Ex

We can also write analogous relationships for the other payment

forms (continuous, immediate, mth ly).

30
Annuities with variable benefits

We can also consider annuities with non-level payment patterns.

For example, we can use our general EPV strategy to find the EPV
of an arithmetically increasing n-year term life annuity-due:
n−1
X
(I ä)x:n = v t (t + 1) t px
t=0

The continuous version would be derived analogously:

Z n
¯ t e −δt t px dt


I ā x:n =
0

31
Example

Consider a n-year term annuity due that pays amount n at time 0,

n − 1 at time 1, . . . , and 1 and time n − 1. (Each payment is made
if and only if the policyholder is alive at the time of the payment.)

1 Give the pmf for the present value of benefits for this annuity.
2 Write the EPV of the annuity as a summation.
Note that this EPV would be denoted (D ä)x:n

3 Write this EPV in terms of increasing and level annuity EPVs.

32
Fractional age assumptions
By using the fractional age relationships we developed for life
insurances, we can find corresponding formulas for annuities. For
example, under UDD we have:
(m) UDD
äx = α(m) äx − β(m)
where
id i − i (m)
α(m) = (m)
and β(m) =
i d (m) i (m) d (m)
   
UDD id i −δ
Letting m → ∞ yields: ax = äx −
δ2 δ2

And for an n-year term annuity, we have

(m) UDD
äx:n = α(m) äx:n − β(m) (1 − n Ex )

33
Woolhouse’s Formula
Another approach to approximating EPVs of mth ly and continuous
life annuities is by using Woolhouse’s formula. For an mth ly
whole life annuity, the EPV approximation is given by

(m) m − 1 m2 − 1
äx ≈ äx − − (δ + µx )
2m 12m2
1 1
Letting m → ∞ yields: ax ≈ äx − − (δ + µx )
2 12

And for an n-year term life annuity, we have

(m) m−1 m2 − 1
äx:n ≈ äx:n − (1 − v n n px )− (δ + µx − v n n px (δ + µx+n ))
2m 12m2
and
1 1
ax:n ≈ äx:n − (1 − v n n px ) − (δ + µx − v n n px (δ + µx+n ))
2 12

34
SOA Practice Problem #7

For an annuity payable semiannually, you are given:

Deaths are uniformly distributed over each year of age.
q69 = 0.03
i = 0.06
1000Ā70 = 530
(2)
Calculate ä69 . [8.5957]

35
SOA Practice Problem #11

For a group of individuals all age x, 30% are smokers (s), 70% are
nonsmokers (ns),
δ = 0.10
Āsx = 0.444
Āns
x = 0.286
Tx is the future lifetime of (x)
h i
Var āTs = 8.818
h xi
Var āTns = 8.503
x
h i
Calculate Var āTx for an individual chosen at random from this
group. [9.1217]

36