2 Life Insurance

2.1 Introduction
In this section, we shall study insurance models built to reduce the financial
impact caused by an unexpected death. For a given policy, the present value
of the benefit depends on the moment of death and on investment earnings
between the issue of the policy and the moment of death.
For simplicity, we ignore the financial uncertainty, and we assume a
deterministic model for interest rates. However, we will use a probabilistic
model for survival, and consequently, the present value of the benefit is a
random variable.

Example 2.1. A fully discrete 1-year term life insurance policy of 100 is
issued to (x). The insurer pays 100 at the end of the year if the policyholder
dies in the first year. Suppose the insurance company issues this policy to
1000 policyholders. You are given that qx = 0.002 and i = 10%.

2.2 Insurances Payable at the Moment of Death

The amount and the time of payment of a life insurance benefit depend only
on the length of the interval from the issue of the insurance to the death
of the insured. The actuarial present value (i.e. the value of the benefit
payment at the policy issue), denoted by the random variable Z, can be
decomposed in two parts:

• the benefit function, bt ;

Rt
• the discount function vt (e.g. vt = e− 0 δs ds
),

where t represents the length of the interval from issue to payment.

Remark
Although the interest rates are known, they may vary with time. However,
we assume by default that the force of interest is constant. That is vt = v t
The random variable Z is then the product

Z = bT v T ,

where T is the realization of T (x) (i.e. the value taken by T (x)).

For each policy, we will define Z (i.e. bT ) and will calculate the premium
from the characteristics of the distribution of Z.

1
2.2.1 Level Benefit Insurance
n-year term life insurance
The insurer pays 1 at the moment of death if the insured (x) dies before age
x + n. The death benefit function is then given by

1, t < n
bt = .
0, t ≥ n

The present value random variable Z is

 T
v , T <n
Z= .
0, n≤T <ω−x

The actuarial present value (APV) of the insurance is given by the ex-
pectation of the random variable Z.
Remark
Special symbols are used to represent different types of policies.
j
b k| Ax:
1
n|

• b represents the benefit (= 1 by default)

• A is the sign for an insurance policy

• If a horizontal bar is present over the A (i.e. A), it means that the
benefit is payable continuously (at the moment of death)

• The subscript x is the age at issue (for instance, in the case of an

insurance for (40), x can be displayed as 40, [40], [20] + 20, etc., de-
pending on the type of mortality table used (select, aggregate, etc.)
and depending on the age of selection if we use a select table)

• The subscript n in the angle (i.e. n|) represents the policy maturity.

• The subscript k on the left of A represents the deferred period of the

insurance.

• If a 1 over the subscript x is present, this means that the insurance is

contingent on death (i.e. this is a n-year term life insurance).

• The superscript j indicates by which factor we have to multiply the

force of interest (= 1 by default)

2
 The actuarial present value of Z
Z n
E [Z] = v t t px µ (x + t) dt
0
= A1 .
x: n|

The variance of Z

V ar [Z] = E Z 2 − E [Z]2
 
 2
= 2A1 − A1 .
x: n| x: n|

More generally, we can find the j th moment of the distribution of Z

Z ∞
(zt )j fT (t) dt
 j
E Z =
Z0 n
= v jt t px µ (x + t) dt
0
Z n
= e−(δj)t t px µ (x + t) dt.
0
j
= A1 .
x: n|

Remark
We must be careful when
 j  computing higher moments of Z. If bt takes the
j
values 0 or 1, then E Z = A 1 . However, if the benefit function takes
x: n|
values other than 0 or 1, then we must also take the j th power of the benefit.
We can graph Z and obtain its cumulative distribution function.

Example 2.2. You are given that n = 20, µ(x) = 0.02, and δ = 5%. Find
A1 and P (0.2 < Z < 0.5).
30: 20|

Whole life insurance

The insurer pays 1 at the moment of death. The benefit function bt = 1, t >
0.
The present value random variable

Z = vT , 0 < T < ω − x

3
The actuarial present value is then
Z ω−x
E [Z] = v t t px µ (x + t) dt
0
= Ax .
Whole life insurance is the limiting case of n-year term insurance as n → ∞
(or ω − x)
Ax = lim A 1 .
n→∞ x: n|
The variance of Z
V ar [Z] = E Z 2 − E [Z]2
 

2
= 2 Ax − Ax .
We can graph Z and obtain its cumulative distribution function.
Example 2.3. A fully continuous whole life insurance policy of 100 is issued
to (70). You are given that µ(x) = (110 − x)−1 , 0 < x < 110, and δ = 5%.
Find V ar[Z], P (Z < 50), and P (Z < 10).

n-year pure endowment insurance

The insurer pays 1 at time n if and only if the insured is alive at age x + n.
The present valus random variable

0, T < n
Z= .
vn, n ≤ T < ω − x
The actuarial present value
E [Z] = v n n px = A 1 .
x: n|

Remark:
We can use two symbols to denote the actuarial present value of an n-year
pure endowment: A 1 or n Ex .
x: n|
The variance of Z
V ar [Z] = E Z 2 − E [Z]2
 
!2
2
= A 1 − A 1 .
x: n| x: n|

We can graph Z and obtain its cumulative distribution function.

4
n-year endowment insurance
The insurer pays 1 at the moment of death if x dies before age x + n or 1
at time n if x is alive at age x + n. The present value random variable Z
 T
v , T ≤n
Z= .
vn, n ≤ T < ω − x
Now,
Z n
E [Z] = v t t px µ (x + t) dt + v n n px
0
= A1 +A 1 = Ax: n| .
x: n|
x: n|

Moreover,
 2
2
V ar [Z] = Ax: n| − Ax: n| .
We can graph Z and obtain its cumulative distribution function.
Remark:
A 1 ≤ Ax ≤ Ax: n|
x: n|

Example 2.4. A fully continuous 20-year endowment insurance policy of 1

is issued to (40). You are given that µ(x) = (100 − x)−1 , 0 < x < 100, and
δ = 10%. Find E[Z] and P (Z < 0.5).

m-year deferred insurance

The insurer pays 1 at the moment of death if (x) dies after age x + m. The
present value random variable

0, 0<T <m
Z= T .
v , m≤T <ω−x
The actuarial present value is
Z ω−x
E [Z] = v t t px µ (x + t) dt
m
= m Ex Ax+m
= m| Ax .

Moreover,
2

2
V ar [Z] = m| Ax − m| Ax .
We can graph Z and obtain its cumulative distribution function.

5
Example 2.5. Suppose that ω = 100, x = 50, and δ = 0.06. Suppose also
that the mortality follows a De Moivre distribution. Find Pr [Z > E [Z]], if
E [Z] = 10| A50 .
We first have to find the expectation of Z

10| A50 = 0.166342.

We can now find the probability

Pr [Z > E [Z]] = 0.3979031.

Remark

• Recursive Relationships

• Suppose i = 0...

2.2.2 Varying Benefit Insurance

Up to now, we have considered only level benefit. However, it is possible
that the benefit varies with time. For instance, the benefit can increase or
decrease over a part of the term of the policy.
An annually increasing n-year term life insurance providing $1 at the
moment of death during the first year, $2 at the moment of death in the
second year, ..., $n at the moment of death in the nth year, and nothing after
is such that 
bt + 1c , 0 < t < n
bt = ,
0, t≥n

bT + 1c v T , 0 < T < n

Z= ,
0, n≤x<ω−x

where bxc is the floor of x, i.e. the greatest integer function.

The actuarial present value is denoted with an additional I, which stands
for increasing


IA 1 = E [Z]
x: n|
Z n
= bt + 1c v t t px µ (x + t) dt
0
= A1 + 1| Ax:
1 + ... + n−1| Ax:
1 .
x: n| n−1| 1|

6
 It is possible to compute higher moments of the distribution of Z. How-
ever, care must be exercised since the payments are not only 0 or 1. Because
of that, we cannot only  multiply the force of interest by some factor. For
2
instance, to get E Z , we have multiply the force of interest by 2, but we
also have to square the benefit function.
We just have seen the case where the benefit increases every year. It
is possible to be in the situation where the benefit increases more (or less)
frequently than once per year. For a m-thly increasing n-year term life
1
insurance, the benefit is m paid at the moment of death if the death occurs
2
during the first m-th of a year, m if death occurs during the second m-th of
nm
a year, ..., m = n if death occurs during the last m-th of the last year of
the term. For such an insurance, we then have
 btm+1c
bt = m , 0<t<n ,
0, t≥n

bT m+1c

m vT , 0 < T < n
Z= .
0, n≤T <ω−x

The actuarial present value is

 
E [Z] = I (m) A 1 .
x: n|

If m → ∞, we have an insurance with benefit t payable at the moment of

death. This is a continuously increasing n-year term life insurance

t, 0 < t < n
bt = ,
0, t ≥ n

T vT , 0 < T < n

Z= ,
0, n≤T <ω−x

and we denote the actuarial present value by

E [Z] = I¯Ā 1


x: n|
Z n
= t v t t px µ (x + t) dt
Z0 n
= s| A 1 ds.
0 x: n−s|

7
If we are rather considering a increasing whole life insurance, we replace n
by ∞ (or by ω − x, if there is a limit age ω), and we drop the angle (i.e. n|)
in the previous development.
We considered the case where the benefit increases with time. It is also
possible to have a decreasing benefit. In this case, we necessarily have an
n-year term life insurance. We cannot have a whole life insurance, because
we are assuming that the benefit is positive! It would be stupid to buy an
insurance and have a negative benefit at the time of death, which would
mean that the insured has to pay!
An annually decreasing n-year term life insurance provides n at the mo-
ment of death during the first year, n − 1 at the moment of death during
the second year, and so on, with coverage terminating at the end of the nth
year. This insurance is defined by

n − btc , 0 < t < n
bt = ,
0, n≤t<ω−x

(n − bT c) v T , 0 < T < n

Z= ,
0, n≤T <ω−x

and the actuarial present value is

E [Z] = DA x1 : n|
Z n
= (n − btc) v t t px µ (x + t) dt
0
= A1 + A1 + ... + A 1 .
x: 1| x: 2| x: n|

As for the increasing insurance, the decreasing insurance can have a benefit
function decreasing more (or less) frequently than once per year. The benefit
function may also decrease continuously with time.
In the case of an endowment insurance, the increasing insurance will
pay $n if the insured survives the term of the insurance, and the decreasing
1
insurance will pay $ m if the insured survives the whole term.
We have the following relations, which are easily verified with the graphs
of bT versus T .



DA 1 + IA 1 = (n + 1) A 1 ,
x: n| x: n| x: n|
 

(m)
 
(m)
 1
D A 1 + I A 1 = n+ A1 ,
x: n| x: n| m x: n|

8
 + I¯Ā



D̄Ā 1 1 = n Ā 1 .
x: n| x: n| x: n|

Assuming that k ≤ n, we may have an n-year term life insurance with

benefit increasing during the first k years, and then remaining constant at
k. This insurance is such that

 bt + 1c , t < k
bt = k, k≤t<n ,
0, t≥n


 bT + 1c v T , 0 < T < k


Z= k vT , k≤T <n ,
0, n≤T <ω−x


and the actuarial present value is denoted by

 
I k| Ā 1 = E [Z]
x: n|


= IA 1 + k k| Ax:
1 .
x: k| n−k|

Example 2.6. (35) buys a special whole life insurance of 100. The death
benefit is increasing continuously at a force of interest of 2%. Find the
variance if δ = 8% and µ(x) = 0.02.

2.3 Insurances Payable at the End of the Year of Death

In precious section, we developed models for life insurance with death benefit
payable at the moment of death. Now, we consider the case where the
payment is made at the end of the year of death. Consequently, the time of
the payment and the benefit itself only depend on K (x).
Again, the actuarial present value random variable Z can be decomposed
in two parts

• the benefit function bk+1 (i.e. the benefit amount payable);

• the discount function vk+1 , where k represents the integer number of

years from the issue of the policy until the death of the insured. This
means that the death occurs in year k + 1 of insurance. By default we
use constant force of interest, which means that vk+1 = v k+1 .

9
 The present value, at policy issue of the benefit payment is

zk+1 = bk+1 v k+1 ,

and the random variable Z is then

Z = bK+1 v K+1 ,

where K is the realization of K (x) . In general, we have

w−x−1
X
E [Z] = bk+1 v k+1 k| qx .
k=0

n-year term life insurance

An n-year term life insurance provides $1 at the end of the year of death if
(x) dies before age x + n, we have

1, k = 0, 1, ..., n − 1
bk+1 = ,
0, k = n, n + 1, ...
 K+1
v , K = 0, 1, ..., n − 1
Z= .
0, K = n, n + 1, ..., ω − x − 1
The actuarial present value for this insurance is

A1 = E [Z]
x: n|
n−1
X
= v k+1 k| qx
k=0
n−1
X
= v k+1 k px qx+k .
k=0

Note that we removed the bar over the A since the benefit is not payable
continuously (i.e. at the moment of death), but rather discretely (i.e. at the
end of the year of death).
We can obtain higher moments of Z. Again, if the benefit is 0 or 1,
we find the jth moment by simply multiplying the force of interest by the
factor j. However, if the benefit takes on values other than 0 or 1, it is still
possible to find the jth moment of Z, but we have to remember to take
the jth power of the benefit function as well. All of this works well if the
moments of Z exist, of course.

10
 In our case, we have
 2
2
V ar [Z] = A1 − A1 ,
x: n| x: n|

where
n−1
X
2
A1 = v 2(k+1) k| qx
x: n|
k=0
n−1
X
= e−2δ(k+1) k| qx .
k=0

We can graph and give the probability function of Z.

More generally, the jth moment of the distribution of Z is
n−1
X j
E Zj = v (k+1)
 
k| qx
k=0
j
= A1 .
x: n|

It is then pretty simple to find moments of Z for the case of a level benefit
other than 0 or 1 as well. In the case where the benefit varies from year to
year, however, we must be more careful.
We can develop a recurrence relation for A 1
x n|

Example 2.7. Suppose that i = 5%, and that the mortality follows De
Moivre with ω = 100. Find A 1 and P [Z < E[Z]].
40: 15|

15−1
X
A1 = v k+1 k| q40
40: 15|
k=0
15−1
X
= v k+1 k p40 q40+k
k=0
14 
1 k+1
   
X k 1
= 1−
1.05 100 − 40 100 − 40 − k
k=0
14 
1 k+1 1
X   
=
1.05 60
k=0

11
  14 
X k
1 1
=
63 1.05
k=0
1 15

!
1 − 1.05
 
1
= 1
63 1 − 1.05
= 0.17299.

A whole life insurance

A whole life insurance pays 1 at the end of the year of death i.e.

bk+1 = 1, k = 0, 1, 2, ... ,

Z = v K+1 , K = 0, 1, 2, ..., ω − x − 1 .
The actuarial present value for this insurance is
∞
X
Ax = v k+1 k| qx ,
k=0

and if there is a limit age ω

ω−x−1
X
Ax = v k+1 k| qx .
k=0

In our case, we would then have

V ar [Z] = 2
Ax − (Ax )2 ,

where
ω−x−1
X
2
Ax = v 2(k+1) k| qx
k=0
ω−x−1
X
= e−2δ(k+1) k| qx .
k=0

We can graph and give the probability function of Z.

We can derive a recurrence relation for Ax
ω−x−1
X
Ax = v k+1 k| qx
k=0

12
 ω−x−1
X
= vqx + v k+1 k| qx
k=1
ω−x−2
X
= vqx + vpx v j+1 j| qx+1
j=0
= vqx + vpx Ax+1 .

Example 2.8. Suppose that i = 5%, and that K (x) ∼ Geometric (0.04),
what is Ax and P [0.1 < Z < 0.7]?
∞
X
Ax = v k+1 k| qx
k=0
X∞
= v k+1 k px qx
k=0
∞ 
1 k+1
X 
= (1 − 0.04)k (0.04)
1.05
k=0
 ∞ 
0.04 X 0.96 k
 
=
1.05 1.05
k=0
  !
0.04 1
= 0.96
1.05 1 − 1.05
= 0.4444.

n-year endowment insurance

The insurer pays 1 at the end of the year of death if the death occurs during
the first n years, or pays 1 at time n otherwise. The present value random
variable  K+1
v , K = 0, 1, ..., n − 1
Z= .
vn, K = n, n + 1, ..., ω − x − 1
The actuarial present value is

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

13
Recall that a pure endowment pays the benefit only if the insured is still
alive at the end of the term

A 1 =A 1 =n Ex = v n n px .
x:n| x: n|

So again,
Ax: n| = A 1 +A 1 .
x: n|
x: n|

The variance of Z is
 2
2
V ar [Z] = Ax: n| − Ax: n| ,

where
n−1
X
2
Ax: n| = v 2(k+1) k| qx + v 2n n px
k=0
n−1
X
= e−2δ(k+1) k| qx + e2δn n px .
k=0

We can graph and give the probability function of Z.

m-year deferred insurance

For a deferred insurance with a benefit of $1 payable at the end of the year
of death, the insured will receive a payment only if the death occurs after
age x + m 
0, k = 0, 1, ..., m − 1
bk+1 = ,
1, k = m, m + 1, ..., ω − x − 1

0, K = 0, 1, ..., m − 1
Z= .
v K+1 , K = m, m + 1, ..., ω − x − 1
The actuarial present value for this insurance is
ω−x−1
X
m| Ax = v k+1 k| qx
k=m
= vm m px Ax+m.

The variance of Z is
2

2
V ar [Z] = m| Ax − m| Ax ,

14
where
ω−x−1
X
2
m| Ax = v 2(k+1) k| qx
k=m
ω−x−1
X
= e−2δ(k+1) k| qx .
k=m
We can graph and give the probability function of Z.

2.3.1 Varying Benefit Insurance

We now study increasing and decreasing insurances with benefit paid at the
end of the year of death. Since the payment is made at the end of the year,
the benefit can only change at the beginning of year.
The increasing whole life insurance pays a benefit of $k at the end of the
year of death if the death occurs during the k th year.
bk+1 = k + 1, k = 0, 1, 2, ... ,

Z = (K + 1) v K+1 , K = 0, 1, 2, ... .
The actuarial present value for this insurance is
(IA)x = E [Z]
ω−x−1
X
= (k + 1) v k+1 k| qx .
k=0
If the benefit of a whole life insurance increases only during the first n years,
a benefit of $n is paid if death occurs after the first n year.

(k + 1) , k = 0, 1, ..., n − 1
bk+1 = ,
n, k = n, n + 1, ...

(K + 1) v K+1 , K = 0, 1, ..., n − 1

Z= .
n v K+1 , K = n, n + 1, ...
The actuarial present value is given by
 
I n| A = E [Z]
x
n−1
X ω−x−1
X
= (k + 1) v k+1 k| qx + n v k+1 k| qx
k=0 k=n
= (IA) 1 +n n| Ax .
x: n|

15
 An annually decreasing n-year term insurance provides for a payment of
$ (n − k + 1) if death occurs during the k th year of the n-year term.

n − k, k = 0, 1, ..., n − 1
bk+1 = ,
0, k = n, n + 1, ...

(n − K) v K+1 , K = 0, 1, ..., n − 1

Z= .
0, K = n, n + 1, ...

The actuarial present value for this insurance is

(DA) 1 = E [Z]
x: n|
n−1
X
= (n − k) v k+1 k| qx
k=0
n−1
X
= (n − k) v k+1 k px qx+k
k=0
n−1
X
= (n − k) v k k px v qx+k
k=0
n−1
X
= (n − k) k Ex v qx+k
k=0
n−1
X
= (n − k) k Ex A 1
x+k: 1|
k=0
n−1
X
= (n − k) k| Ax:
1 .
1|
k=0

Also,
n−1
X
(DA) 1 = (n − k) v k+1 k| qx
x: n|
k=0
n−1
X n−k−1
X
= (1) v k+1 k| qx
k=0 j=0

X n−j−1
n−1 X
= v k+1 k| qx
j=0 k=0

16
 n−1
X
= A1
x: n−j|
j=0
Xn
= A1
x: m|
m=1

Remark 2.9. We uses the notation k|n Ax instead of k| Ax:

1 . The disad-
n|
vantage is that we cannot represent a deferred endowment insurance by only
one symbol

m| Ax: n| = m| Ax:
1 + m| A 1
n|
x: n|
= m|n Ax + m+n Ex .

2.3.2 Recursive Relations

We now develop some recurrence relations. As before, the important thing
is to understand the recurrence relation and to be able to explain it with a
graph.
In this section, we saw previously a recurrence formula for A 1 and for
x: n|
Ax . The same recurrence relation exists for an endowment insurance
n−1
X
Ax: n| = v k+1 k| qx + v n n px
k=0
n−1
X
= v qx + v k+1 k| qx + v px v n−1 n−1 px+1
k=1
n−2
X
= v qx + v j+2 j+1| qx + v px v n−1 n−1 px+1
j=0
n−2
X
= v qx + v px v j+1 j| qx+1 + v px v n−1 n−1 px+1
j=0
 
n−2
X
= v qx + v px  v j+1 j| qx+1 + v n−1 n−1 px+1 
j=0
= v qx + v px Ax+1: n−1| ,

where
Ax+n: 0| = 1,

17
and
A 1 = 0.
x+n: 0|

For a deferred insurance, we have

m| Ax:
1 = vm m px Ax+m
n|
= v px v m−1


m−1 px+1 Ax+m
= v px m−1| A 1 .
x+1: n|

Note that
0| A 1 =A 1 .
x+m: n| x+m: n|

For the increasing insurance we have

 
(IA) 1 = v qx + v px A 1 + (IA) 1 ,
x: n| x+1: n−1| x+1: n−1|

where
(IA) 1 = 0.
x+n: 0|

For the decreasing insurance,

(DA) 1 = n v qx + v px (DA) 1 ,
x: n| x+1: n−1|

where
(DA) 1 = 0.
x+n: 0|

2.3.3 Upper (m)

Up to now, we have seen the cases where the benefit is payable at the moment
of death, or at the end of the year of death. We can generalize and consider
1
the case where the payment is made at the end of the m -th of the year of
death. We denote this by the superscript (m). For instance, for a whole life
insurance,
ω−x−1
X m−1X h (1+j)
i
k+
A(m)
x = v m
j 1 qx
k+ m |m
k=0 j=0
m(ω−x)−1
X 1+j
= v m j 1
| qx .
m m
j=0

18
Also
n−1
X m−1
X h (1+j)
i
(m) k+
A1 = v m
j 1 qx
k+ m |m
x: n|
k=0 j=0
mn−1
X 1+j
= v m j 1
| qx ,
m m
j=0

and
ω−x−1
X m−1X  h i

(m)
(m) (1 + j) k+
(1+j)
I A = k+ v m
j 1 qx
k+ m |m
x m
k=0 j=0
m(ω−x)−1  
X 1+j 1+j
= v m j 1
| qx .
m m m
j=0

2.4 Relationships between Insurances Payable at the Mo-

ment of Death and the End of the Year of Death
We want to go from the continuous case to the discrete case. Unless oth-
erwise specified, we will assume a uniform distribution of deaths over each
year of age (i.e. U DD assumption).
We will show that
i
Ax = Ax ,
δ
is derived using two different methods. We will present a third one
Z ω−x
Ax = v t t px µ (x + t) dt
0
ω−x−1
X Z k+1
= v t t px µ (x + t) dt
k=0 k
ω−x−1
X Z 1
= v k+s k+s px µ (x + k + s) ds
k=0 0
ω−x−1
X Z 1
= v k k px v s s px+k µ (x + k + s) ds
k=0 0
ω−x−1
X Z 1
= v k k px v s qx+k ds
k=0 0

19
 ω−x−1
X Z 1
k+1
= v k px qx+k v s−1 ds
k=0 0
Z 1
= Ax v s−1 ds
0
1
!
v s−1
= Ax
ln v 0
!
1
1− v
= Ax
ln v
 
1 − (1 + i)
= Ax
−δ
i
= Ax .
δ
In the fifth equality, we use the U DD assumption to replace s px+k µx+k (s)
by qx+k . By a similar reasoning, we can obtain the relation
i
A(m)
x = Ax .
i(m)
By expressing any other insurance product in terms of Ax , we can find
it is equivalent in discrete time.
Recall that
A 1 =A 1.
x: n| x: n|

Consequently, we must always isolate the part ”pure endowment” of the

insurance product, if there is one, before converting in discrete time.
For an n-year term life insurance, we have

A1 = Ax − n Ex Ax+n
x: n|
i i
= Ax − n Ex Ax+n
δ δ
i
= (Ax − n Ex Ax+n )
δ
i
= A1 .
δ x: n|
Similarly,

m| Ax:
1 = m Ex A 1
n−m| x+m: n−m|

20
 i
= m Ex A 1
δ x+m: n−m|
i
= m| Ax:
1 ,
δ n−m|

and


IA 1 = Ax1 : n| + 1| Ax1 : n−1| + 2| Ax1 : n−2| + ... + n−1| Ax1 : 1|
x: n|
i i i i
= A 1 + 1| Ax1 : n−1| + 2| Ax1 : n−2| + ... + n−1| Ax1 : 1|
δ x : n| δ δ δ
i
= (IA)x1 : n| .
δ
We obtained three cases in which we can directly multiply by the factor δi .
In fact, under the following two conditions, this will be true if bt = bk+1 , i.e.
what is paid at the moment of death
depends only on the year during which
occurs the death. For instance, IA x does not satisfy this condition.
If this condition is not satisfied, we cannot apply the factor δi directly,
but we can start from the basic principles to find a relation.

Example 2.10.
Z ω−x
t v t t px


IA x = µ (x + t) dt
0
ω−x−1
X Z k+1
= t v t t px µ (x + t) dt
k=0 k
ω−x−1
X Z 1
= (k + s) v k+s k+s px µ (x + k + s) ds
k=0 0
ω−x−1
X Z 1
= (k + 1) v k+1 k px v s−1 s px+k µ (x + k + s) ds
k=0 0
ω−x−1
X Z 1
− (1 − s) v k+1 k px v s−1 s px+k µ (x + k + s) ds
k=0 0
ω−x−1
X Z 1
= (k + 1) v k+1 k px qx+k v s−1 ds
k=0 0
ω−x−1
X Z 1
− v k+1 k px qx+k (1 − s) v s−1 ds
k=0 0

21
 Z 1
i
= (IA)x − Ax u (1 + i)u du
δ 0
Z 1
i
= (IA)x − Ax u eδu du,
δ 0

where we used the U DD assumption for the fifth equality. Now, using inte-
gration by parts to solve the integral, we get
( !)
1 δu 1


i u δu 1
IA x = (IA)x − Ax e − e
δ δ 0 δ2 0
  
i 1 δ 1  
= (IA)x − Ax e − eδ − 1
δ δ δ2
 
i 1 i
= (IA)x − Ax (1 + i) − 2
δ δ δ
 
i 1 i i
= (IA)x − Ax + − 2
δ δ δ δ
   
i i 1 1
= (IA)x − Ax +1−
δ δ i δ
   
i i 1 1
= (IA)x − Ax − .
δ δ d δ


Now, we can derive a formula for IA x1 : n|



h i
IA x1 : n|
= IA x
− n Ex
IA x+n + n Ax+n
   
i i 1 1
= (IA)x − Ax −
δ δ d δ
       
i i 1 1 i
− n Ex (IA)x+n − Ax+n − +n Ax+n .
δ δ d δ δ
In the second equality, we just plugged in the formula previously found.
We now know how to go from the continuous case to the discrete case,
and vice versa. But previously, we talked about higher moments for the
distribution of Z. What happens to the factor δi , for the higher moments?
In fact, the denominator becomes jδ, and the numerator becomes (1 + i)j −1.
Example 2.11.
Z ω−x
2
Ax = v 2t t px µ (x + t) dt
0

22
 ω−x−1
X Z k+1
= v 2t t px µ (x + t) dt
k=0 k
ω−x−1
X Z 1
= v 2(k+s) k+s px µ (x + k + s) ds
k=0 0
ω−x−1
X Z 1
2k
= v k px v 2s s px+k µ (x + k + s) ds
k=0 0
ω−x−1
X Z 1
= v 2k k px v 2s qx+k ds
k=0 0
ω−x−1
X Z 1
2(k+1)
= v k px qx+k v 2(s−1) ds
k=0 0
Z 1
2
= Ax v 2(s−1) ds
0
 
1
v 2(s−1)
2
= Ax  
ln v 2
0
!
1
2 1 − v2
= Ax
2 ln v
!
2 1 − (1 + i)2
= Ax
−2δ
!
(1 + i)2 − 1 2
= Ax
2δ
 2 
i + 2i 2
= Ax .
2δ

2.5 Percentile and Normal Approximation

We need to know the cumulative distribution func ion of Z to find its per-
centile. For example, the 90% percentile is given by ξ such that

Pr [Z ≤ ξ] = 0.9.

Example 2.12. Let the force of interest be such that

1
δt = t+4 , t≥0 ,

23
and the mortality follows the De Moivre distribution with ω = 100. The ben-
efit amount is $1. Find the actuarial present value of a whole life insurance
for (40), V ar [Z], and the 90th percentile.
First, since the mortality is De Moivre, we know that T (40) ∼ U nif orm(0, 100−
1 1
40), and then fT (40) (t) = 100−40 = 60 .
First, the present value factor
 Z t 
ds
vt = exp −
0 s+4
= exp − ln (s + 4)|t0

4
= ,
t+4
for t ≥ 0. We then have for t ≥ 0,

Z = zT
= bT vT
4
= .
T +4
The actuarial p.v. of a whole life insurance for (40) is given by
Z 100−40   
4 1
A40 = dt
0 t+4 100 − 40
Z 60   
4 1
= dt
0 t+4 60
1
= ln (t + 4)|60
0
15
1
= (ln 64 − ln 4)
15
= 0.1848.

The second moment of Z is

2
A40 = E Z 2
 
Z 60  2  
4 1
= dt
0 t+4 60
Z 60
4
= (t + 4)−2 dt
0 15
!
4 1 60
= −
15 t + 4 0

24
  
4 1 1
= − −
15 64 4
1
= ,
16
and hence
2

2
V ar [Z] = A40 − A40
1
= − (0.1848)2
16
= 0.02833.

The 90th percentile is given by ξ such that

Pr [Z ≤ ξ] = 0.9.

 
4
Pr ≤ξ = 0.9
T +4
⇔ Pr [4 ≤ (T + 4) ξ] = 0.9
⇔ Pr [4 (1 − ξ) ≤ T ξ] = 0.9
 
(1 − ξ)
⇔ Pr T ≥ 4 = 0.9
ξ
4 (1−ξ)
ξ
⇔ 1− = 0.9
60
(1 − ξ)
⇔ 1− = 0.9
15ξ
(1 − ξ)
⇔ = 1.5
ξ
⇔ 1 − ξ = 1.5ξ
⇔ 1 = 2.5ξ
⇔ ξ = 0.4,

which corresponds to T = 4 (1−0.4)

0.4 = 6.
To find a percentile of the distribution of a portfolio of policies, we shall
use the normal approximation. Let Zi denote the present value random vari-
able for the policyholder j (j = 1, ...., l). We suppose that the policyholders
are independent. Let S denote the aggregate present value random variable
for all policies
S = Z1 + ... + Zl ,

25
The normal approximation is
S − E[S]
p →W
V ar [S]
when l → ∞ and W ∼ N (0, 1).
We can use this approximation to estimate a probability or percentile.
Example 2.13. Now, suppose we have 100 identical and independent con-
tracts. If we consider the same settings as in the previous examples, find ξ
using the normal approximation.
Since the expectation is a linear operator, we simply have to add the
expectation of each individual contract to find E [S]
E [S] = E [Z1 + ... + Z100 ]
= E [Z1 ] + ... + E [Z100 ]
= 100 E [Z1 ]
= 100 A40
= 100 (0.1848)
= 18.48.
Since the contracts are independent, we can also simply add the variances
of each individual contract to find V ar [S]
V ar [S] = V ar [Z1 + ... + Z100 ]
= V ar [Z1 ] + ... + V ar [Z100 ]
= 100 V ar [Z1 ]
= 100 (0.02833)
= 2.833.
To find the 90th percentile of S, we will use the normal approximation.
" #
S − E [S] ξ − E [S]
Pr [S ≤ ξ] = 0.9 ⇔ Pr p ≤p = 0.9
V ar [S] V ar [S]
 
S − 18.48 ξ − 18.48
⇔ Pr √ ≤ √ = 0.9
2.833 2.833
 
ξ − 18.48
⇔Φ √ = 0.9
2.833
ξ − 18.48
⇔ √ = 1.282
2.833
⇔ ξ = 20.64.

26
 Generally, if
l
X
S= Zj ,
j=1

where the Zj ’s are independent and identically distributed (iid), we get

E [S] = lE [Z]
V ar [S] = lV ar [Z] .

27