Professional Documents
Culture Documents
LIFE CONTINGENCIES 1
1
1 SURVIVAL DISTRIBUTIONS AND LIFE
TABLES
1.1 Some Special Random Variables and the Survival
Function
Definition 1.1. Let (x) denote a person aged x and X be a random vari-
able representing the age-at-death of (0). We will call X as the age-at-
death (of a newborn) random variable.
Since X is a random variable, we can talk about its cdf as well as the
complement of its cdf. The cdf of X,
FX (x) = P r[X ≤ x]
gives the probability that (0) will die before attaining age x. Furthermore,
the complement of the cdf of X, which we will denote by s(x), i.e.,
gives the probability that a newborn will survive to age x. Because of this,
we will call s(x) as the survival function.
Definition 1.2. Let T (x) (or simply T ) denote a random variable repre-
senting the future lifetime of (x). We will call T , the time-until-failure
(of a person aged x) random variable.
The cdf of T , which is G(t) = P r[T ≤ t] gives the probability that (x)
will die before attaining age x + t. This probability statement is given a
special symbol, t qx . The complement of G(t) gives the probability that (x)
will survive to age x + t. The notation we will use to denote this probability
statement is t px . Obviously, t px + t qx = 1.
We note that if x = 0, then T (0) = X and x p0 = s(x), x ≥ 0.
Example 1.2. Express the following probability statements in terms of
the survival function of T and/or its complement.
1. The probability that (x) will survive t years and die within the fol-
lowing u years.
2. The probability that (x) will survive to age x + t then die before
attaining age x + t + 1.
2
Definition 1.3. Let K(x) (or simply K) be a discrete random variable
representing the completed future years lived by x, i.e., K(x) = [T (x)]. We
will call K as the curtate future lifetime of (x). The pmf of K is
where k = 0, 1, 2, . . .
Recall the survival function, s(x), we mentioned earlier. The survival
function is a continuous function in x, defined on the interval 0 ≤ x ≤ ω,
where ω is called the limiting age. Moreover, s(x) is a decreasing function
with s(0) = 1 and s(ω) = 0. We can express t px and t qx in terms of s(x)
as what the following shows: We first note that x+t p0 = (x p0 )(t px ). Then,
x+t p0 s(x + t)
t px = = ,
x p0 s(x)
and
s(x + t) s(x) − s(x + t)
t qx = 1 − t px = 1 − = .
s(x) s(x)
x
Example 1.3. Let s(x) = 1 − 105 , where 0 ≤ x ≤ 105. Find the following
probabilities.
1. The probability that (0) will survive to age 15.
2. The probability that (0) will die between ages 15 and 42.
3. The probability that a life aged 15 will survive to age 42.
4. The probability that a life aged 15 will die before attaining age 42.
0 FX (x + ∆x) − FX (x)
FX (x) = lim = fX (x).
∆x→0 ∆x
Then,
FX (x + ∆x) − FX (x)
P r[x < X ≤ x + ∆x|X > x] =
1 − FX (x)
FX (x + ∆x) − FX (x) ∆x
=
∆x 1 − FX (x)
∆x
≈ fX (x)
1 − FX (x)
fX (x)
= ∆x
1 − FX (x)
fX (x)
The expression 1−F X (x)
gives the conditional pdf of X at exact age x
given survival to x. This conditional pdf is called the force of mortality and
3
is denoted by µ(x). The force of mortality is also called the failure rate or
the hazard rate.
Consider the following,
d d d
s(x) = [1 − FX (x)] = − FX (x) = −fX (x).
dx dx dx
Thus,
fX (x) s0 (x)
µ(x) = =− .
1 − FX (x) s(x)
Moreover, since s(x) is a decreasing function, then s0 (x) < 0, thus the force
of mortality is a nonnegative function.
We now find an expression for the pdf of X in terms of the force of
mortality. We know that
s0 (y) d
µ(y) = − = − log s(y),
s(y) dy
or equivalently,
−µ(y)dy = d log s(y).
Integrating from x to x + n,
Z x+n Z x+n
s(x + n)
− µ(y)dy = − d log s(y) = log s(x+n)−log s(x) = log = log n px .
x x s(x)
d d d h Rx i d Rx
f (x) = FX (x) = [1 − s(x)] = 1 − e− 0 µx (s)ds = − e− 0 µx (s)ds
dx dx Z x dx dx
− 0x µx (s)ds
R d
= e µx (s)ds = x p0 µ(x).
dx 0
4
d d d s(x + t)
gT (t) = GT (t) = t qx = 1−
dt dt dt s(x)
0
1 0 s(x + t) s (x + t)
= − s (x + t) = − = t px µx (t).
s(x) s(x) s(x + t)
Thus, the pdf of T is
gT (t) = t px µx (t).
Clearly, t px µx (t) denotes the probability that (x) dies between ages x+t
and x + t + dt and Z ∞
t px µx (t)dt = 1.
0
5
We will use the notation n dx for E[n Dx ], thus n dx denotes the expected
number of deaths between ages x and x + n among the l0 newborns.
Now, since s(x) = llx0 ,
d d
s(x) lx
µ(x) = − dx = − dx ,
s(x) lx
or equivalently,
d
lx = −lx µ(x).
dx
The expression lx µ(x) is the expected density of deaths in the age interval
(x, x + dx). Furthermore,
µ(x)lx dx = −dlx .
lim t(−t px ) = 0.
t→∞
Therefore,
Z ∞
eox = E[T (x)] = t px dt.
0
Thus,
6
Z ∞
V ar[T (x)] = E[T (x)2 ] − (E[T (x)])2 = 2 tt px dt − (eox )2 .
0
We note that we assumed above that E[T (x)] and E[T (x)2 ] exist. More-
over, we can also find the median and the mode of the distribution of T (x).
If we denote the median of T (x) by m(x), then the median of T (x) is the
solution of the equation
1 s(x + m(x)) 1
P r[T (x) > m(x)] = or = .
2 x 2
The mode of the distribution of T (x) is the value t where the pdf of T , i.e.,
the function gT (t) = t px µx (t) attains its maximum value.
The expected value of K(x) is denoted by ex , and is called the curtate-
expectation-of-life. Thus,
∞
X ∞
X ∞
X
ex = E[K(x)] = k k px qx+k = k∆(−k px ) = k(−k px )]∞
0 + k+1 px .
k=0 k=0 k=0
Using the same argument as above, then if E[K(x)] exists then lim k(−k px ) =
k→∞
0. Thus,
∞
X
ex = E[K(x)] = k px .
k=1
Also,
∞
X ∞
X ∞
X
E[K(x)2 ] = k 2 k px qx+k = ∆(−k px ) = k 2 (−k px )]∞
0 + (∆k 2 )(k+1 px )
k=0 k=0 k=0
Therefore,
∞
X
V ar[K(x)] = E[K(x)2 ] − (E[K(x)])2 = (2k − 1)k px − (ex )2 .
k=1
Another life table function is Lx , which denotes the total expected num-
ber of years lived between ages x and x + 1 by the survivors of the l0
lives.Thus,
Z 1
Lx = lx+1 + tlx+t µx (t)dt,
0
where lx+1 represents the years lived between ages x and x + 1 of those
R1
who survive to age x + 1, and 0 tlx+t µx (t)dt represents the years lived by
those who died between ages x and x + 1.
7
Our expression for Lx may be simplified in the following way:
Z 1 Z 1 Z 1
Lx = lx+1 + tlx+t µx (t)dt = lx+1 − tdt lx+t = lx+1 −tlx+t ]10 + lx+t dt,
0 0 0
or equivalently, Z 1
Lx = lx+t dt.
0
In this form, we can say that Lx is the mean value of lx over the age interval
(x, x + 1). An approximation for Lx is given by
1 1
Lx ≈ (lx+1 + lx ) = lx − dx .
2 2
Another life table function is the central death rate over the age interval
(x, x + 1), denoted by mx and defined as
R1
0
lx+t µx (t)dt −lx+t ]10 lx − lx+1 dx
mx = R1 = = = .
l dt Lx L x Lx
0 x+t
If we extend the age interval from (x, x + 1) to (x, x + n), then the
definitions of Lx and mx can be extended appropriately as follows:
Z n Z n
L
n x = nlx+n + tl µ
x+t x (t)dt = lx+t dt,
0 0
and Rn
l µ (t)dt
0 R x+t x lx − lx+n n dx
n mx = n = = .
l dt
0 x+t
L x Lx
Another life table function is Tx which denotes the total number of years
lived beyond age x by the l0 lives. Thus,
Z ∞ Z ∞ Z ∞
Tx = tlx+t µx (t)dt = − tdt lx+t = lx+t dt.
0 0 0
8
Definition 1.4. Suppose x is an integer and t is a real number such that
0 ≤ t ≤ 1. The the survival function using the given functional age as-
sumptions are as follows:
1. Uniform death distribution: s(x + t) = (1 − t)s(x) + ts(x + 1);
2. Constant force of mortality: s(x + t) = s(x)eµt , where µ = − log px ;
1 1−t t
3. Balducci: s(x+t) = s(x) + s(x+1
Example 1.5. Given s(40) = 0.7746 and s(41) = 0.7681. Find µ40 ( 14 ),
using the three fractional age assumptions.
Example 1.6. Using the UDD assumption, show that t qx = tqx , 0 ≤ t ≤ 1.
Example 1.7. Using the constant force of mortality assumption, show that
µx (t) = µ.
9
where h is the constant of proportionality. This implies that
π
tan x, 0 ≤ x ≤ 2
e−x , x > 0
1
1− 1+x , x ≥0
x
2. (Ex. 3.5) If s(x) = 1 − 100 , 0 ≤ x ≤ 100, calculate
x
p
3. (Ex. 3.7) If s(x) = 1− 100 , 0 ≤ x ≤ 100, evaluate
10
(a) 17 p19 ; (d) µ(36);
(b) 15 q36 ;
(c) 15|13 q36 ; (e) E[T (36)]
∞
X
4. (Ex. 3.8) Confirm that k| q0 = −∆s(k) and that k| q0 = 1.
k=0
x lx
0 40
10 39
70 26
calculate
(a) E[T ];
(b) V ar[T ];
(c) the median of T ;
(d) the mode of the distribution of T
11
10. (Ex. 3.19) If
µx (t) = t, t ≥ 0,
calculate
(a) t px µx (t);
(b) eox
HINT: If Z ∼ N (0, 1), then
1 2
fZ (z) = √ e−z /2 .
2π
Acx
µ(x) = ,
1 + Bcx
for x > 0.
log(log c) − log A
x0 = .
log c
3 10
15. (Ex. 3.44) If µ(x) = 100−x − 250−x for 40 < x < 100, calculate
(a) 40 p50 ;
12
2 Life Insurances
2.1 Introduction
Insurance systems were established to reduce the adverse financial impact
of some type of random events, usually loss of life or an object of some
or of great value. In this chapter we will discuss some basic types of life
insurances. There will also be two different types of classification. One is
based on whether the benefit is paid at the moment of death of the insured
or at the end of the year of death of the insured. The other classification
is one whether the benefit size is fixed or if it varies depending on the time
of death of the insured.
Under these classification we will consider the following basic types of
insurance:
1. WHOLE LIFE insurance pays the benefit upon the death of the in-
sured any time in the future.
2. n-YEAR TERM insurance pays the benefit upon the death of the
insured if death occurs within n years after issue.
3. n-YEAR PURE ENDOWMENT pays the benefit at the end of n
years only if the insured survives to that time.
To denote the (net) present value of the insurance, that is, the (net)
present value of the benefit, we will use the random variable Z. Z is
dependent on several things. These are the size of the benefit, interest rate
and more importantly, the time of death of the insured, because this date
will tell us when the benefit is to be paid by the insurer to the beneficiaries
of the insured. Thus, for each type of insurance we will define the random
variable Z and then we will derive an expression for its expected value. The
expected value of Z, E[Z] is called the net single premium (NSP) of the
insurance.
13
denote the benefit function, and vT to denote our discount function, then
the random variable Z is defined as Z = bT vT .
and
bT = 1, T ≥ 0.
Thus,
vT
if T ≥ 0
Z=
0 otherwise.
Therefore, Z ∞
Āx = E[Z] = v t t px µx (t)dt.
0
Z ∞ Z ∞ Z ∞
2 2t −2δt
E[Z ] = v t px µx (t)dt = e t px µx (t)dt = e(−2δ)t t px µx (t)dt.
0 0 0
Thus,
E[Z j ] at δt is equal to E[Z] at jδt .
This is called the rule of moments which holds generally for insur-
ances paying only a unit amount, and when the force of interest is
deterministic. We will use the notation 2 Āx for E[Z 2 ]. i.e., the actu-
arial present value for a whole insurance for a unit amount calculated
at 2δ.
vT
if T ≤ n
vT =
0 otherwise.
and
bT = 1, T ≥ 0.
Thus,
vT
if T ≤ n
Z=
0 otherwise.
14
Therefore, Z n
Ā1x:n| = E[Z] = v t t px µx (t)dt.
0
vn
if T ≥ n
vT =
0 otherwise.
and
1 if T ≥ n
bT =
0 otherwise.
Thus,
vn
if T ≥ n
Z=
0 otherwise.
Therefore,
Z ∞ Z ∞
1
Ax:n| = E[Z] = v n t px µx (t)dt = v n t px µx (t)dt = v n [1−G(n)] = v n (1−n qx ) = v n n px .
n n
We can derive the E[Z] in another way. Since the size of the pay-
ment and the payment date are predetermined, the only element of
uncertainty is whether or not a claim will occur. Let
1 if (x) survives to age x + n
Y =
0 otherwise.
Then, Z = v n Y . Thus,
1
Āx:n| = E[Z] = E[v n Y ] = v n E[Y ] = v n n px .
Also,
vT
if T ≤ n
vT =
vn if T > n.
and
bT = 1, T ≥ 0.
15
Thus,
vT
if T ≤ n
Z=
vn if T > n.
Therefore,
Z n Z ∞
t
Āx:n| = E[Z] = v t px µx (t)dt + v n t px µx (t)dt = Ā1x:n| + Āx:n|
1
.
0 n
vT vT
if T ≤ n 0 if T ≤ n if T ≤ n
Z1 = ; Z2 = ; Z3 =
0 if T > n. vn if T > n. vn if T > n.
2 1
= Āx:n| + 2Āx:n|
1
− [(Ā1x:n| )2 + 2Ā1x:n| Āx:n|
1 1 2
+ (Āx:n| ) ]
2 1
= Āx:n| + 2Āx:n|
1
− [Ā1x:n| + Āx:n|
1 2
]
2 1
= Āx:n| + 2Āx:n|
1
− (Āx:n| )2
vT = v T if T ≥ 0,
and
0 if T ≤ m
bT =
1 if T > m.
Thus,
0 if T ≤ m
Z=
vT if T > m.
Therefore,
Z ∞
m| Āx = E[Z] = v t t px µx (t)dt.
m
16
Example 2.1. The pdf of T (x) is given by
1
80 if 0 ≤ T ≤ 80
gT (t) =
0 elsewhere.
At the force of interest δ, calculate for Z
1. the NSP;
2. V ar[Z];
3. the 90th percentile, ξZ0.9 .
Example 2.2. Consider a 5 year deferred whole life insurance payable at
the moment of death of (x). The individual is subject to a constant force of
mortality µ = 0.04. For the distribution of the present value of the benefit
payment at δ = 0.10, calculate
1. the expectation;
2. the variance;
3. the median ξZ0.5 .
Example 2.3. Assume that each of 100 independent lives are
i.) aged x;
ii.) subject to the constant force of mortality µ = 0.04;
iii.) insured for the death benefit amount of 10 units, payable at the mo-
ment of death.
The benefit payments are to be withdrawn from an investment fund earning
δ = 0.06. Calculate the minimum amount at t = 0 so that the probability
is approximately 0.95 that sufficient funds will be on hand to withdraw the
benefit payment at the death of each individual.
vK+1 = v K+1 if K = 0, 1, 2, . . .
and
bK+1 = 1 if K = 0, 1, 2, . . .
17
Thus,
Z = v K+1 if K = 0, 1, 2, . . . .
Therefore,
∞
X
Ax = E[Z] = v k+1 k px qx+k
k=0
and
∞
X
2 2
V ar[Z] = Ax − (Ax ) , where Ax = 2
e−2δ(k+1) k px qx+k .
k=0
v K+1
if K = 0, 1, 2, . . . , n − 1
Z=
0 if K = n, n + 1, n + 2, . . . .
and
n−1
X
A1x:n | = E[Z] = v k+1 k px qx+k .
k=0
bK+1 = 1 if K = 0, 1, 2, . . .
and
v K+1
if K = 0, 1, 2, . . . , n − 1
vK+1 =
vn if K = n, n + 1, n + 2, . . . .
Thus,
v K+1
if K = 0, 1, 2, . . . , n − 1
Z=
vn if K = n, n + 1, n + 2, . . . .
Therefore,
18
n−1
X ∞
X n−1
X
Ax:n| = v k+1 k px qx+k + v k+1 k px qx+k = v k+1 k px qx+k +v n n px = A1x:n| +Ax:n|
1
,
k=0 k=n k=0
since
∞
X ∞
X n−1
X
v k+1 k px qx+k = v n ( k px qx+k − k px qx+k ) = v n (1−n qx ) = v n n px .
k=n k=0 k=0
19
We now consider an m year deferred insurances payable at the moment
of death and at the end of the year of death. We note that for an m year
deferred whole life insurance, although T is continuous, Z is mixed with
pmf at 0 because {Z = 0} corresponds to {T ≤ m}. Thus,
FZ (z) = P r[Z ≤ z]
= P r[Z = 0] + P r[0 < Z ≤ z]
= P r[T ≤ m] + P r[0 < v T ≤ z]
log z
= P r[T ≤ m] + P r[T ≥ ]
log v
log z
= FT (m) + 1 − FT ( )
log m
bT = [T + 1], vT = v T .
So,
Z = [T + 1]v T ,
where T ≥ 0. Therefore,
Z ∞
(I Ā)x = E[Z] = [t + 1]v t t px µx (t)dt.
0
20
1
2. An mth ly increasing whole life insurance providing a payment of m
th 2
at the moment of death during the first m of the year; m at the mo-
ment of death during the second mth of a year, and so on, increasing
1
by m at mth ly intervals thoughout the term of the insurance. Then,
when T > 0,
[T m + 1]
bT = , vT = v T .
m
So, for T ≥ 0,
v T [T m + 1]
Z= .
m
Therefore,
Z ∞ t
v [tm + 1]
(I (m) Ā)x = E[Z] = t px µx (t)dt.
0 m
vT = v T , T > 0.
Thus,
V T (n − [T ])
if T < n
Z= .
0 of T ≥ n
Therefore,
Z n
(DĀ)1x:n| = E[Z] = v t (n − [t])t px µx (t)dt.
0
21
1. An increasing whole life insurance paying K + 1 units at the end of
K + 1 years provided the insured dies after K complete years. Then,
for K = 0, 1, 2, . . .
Thus, whenever K = 0, 1, 2, . . .
Z = (K + 1)v K+1 ,
and
∞
X
(IA)x = E[Z] = (k + 1)v k+1 k px qx+k .
k=0
and
vK+1 = v K+1 , K = 0, 1, 2, . . . .
Thus,
(K + 1)v K+1
if K = 0, 1, 2, . . . , n − 1
Z= .
0 of K = n, n + 1, n + 1, . . .
Therefore,
n−1
X
(IA)1x:n| = E[Z] = (k + 1)v k+1 k px qx+k .
k=0
22
Thus,
(n − K)v K+1
if K = 0, 1, 2, . . . , n − 1
Z= .
0 of K = n, n + 1, n + 1, . . .
Therefore,
n−1
X
(DA)1x:n| = E[Z] = (n − k)v k+1 k px qx+k .
k=0
Z n
Ā1x:n| = v t t px µx (t)dt
0
n−1
X Z k+1
= v t t px µx (t)dt
k=0 k
n−1
XZ 1
= v k+s k+s px µx+k (s)ds since t = k + s
k=0 0
n−1
X Z 1
= v k+1 k px v s−1 s px+k µx+k (s)ds
k=0 0
n−1
X Z 1
= v k+1 k px qx+k v s−1 ds since under UDD s px+k µx+k (s) = qx+k
k=0 0
n−1
X Z 1 Z 1
k+1 s−1
= v k px qx+k s̄1| since v dt = (1 + i) v s ds = (1 + i)ā1| = s̄1|
k=0 0 0
1
(1 + i)t 1
Z
i i
= A since s̄1| = (1 + i)t dt = |0 =
δ x:n| 0 log(1 + i)t δ
Moreover,
i i
Āx = lim Ā1x:n| = lim A = Ax .
n→∞ n→∞ δ x:n| δ
Alternatively, we recall that T = K + S, where S represents the frac-
tional part of the year lived in the year of death. We know that under
UDD, K and S are independent then K + 1 and 1 − S are also indepen-
dent. Moreover, S is uniformly distributed in (0, 1). This implies that 1−S
is also uniformly distributed in (0, 1). Thus,
i
Āx = E[v T ] = E[v K+1 (1 + i)1−S ] = E[v K+1 ]E[(1 + i)1−s ] = Ax .
δ
23
This is true because 1 − S is uniformly distributed in (0, 1), then
Z 1
i
E[(1 + i)1−S ] = (1 + i)1−s ds = .
0 δ
Consider an n-year term increasing insurance payable at the moment of
death. The present value random variable is
[T + 1]v T
if T < n
Z= .
0 of T ≥ n
But, [T + 1] = K + 1 and since T = K + S, then
(K + 1)v K+1
if K = 0, 1, 2, . . . , n − 1
W = .
0 of K = n, n + 1, n + 1, . . .
Then,
Z = W v S−1 = W (1 + i)1−S .
Since W is a function of K + 1 and K + 1 and 1 − S are independent under
UDD, we have
i
E[Z] = E[W (1 + i)1−S ] = E[W ]E[(1 + i)1−S ] = (IA)1x:n| .
δ
Example 2.6. Calculate the NSP and variance for a 10000, 30-year en-
dowment insurance providing the death benefit at the moment of death for
a male aged 35 at issue policy. Use the illustrative life table, the UDD
assumptions and i = 0.06.
Example 2.7. Calculate for a life aged 50, the APV for an annually de-
creasing 5-year term insurance paying 5000 at the moment of death in the
first year, 4000 in the second year and so on. Use the illustrative life table,
the UDD assumption and i = 0.06.
3 LIFE ANNUITIES
In the theory of interest we discussed annuities. We said that an annuity
is a sequence of periodic payments. To determine the present value as well
as the accumulated value of an annuity only three things were considered,
the size of the periodic payments, the term of the annuity and the interest
rate used. Moreover the term of the annuity is fixed or pre-defined. In this
chapter we will consider another type of annuity - the life annuity. Since
this is an annuity, it will be a sequence of periodic payments. However, the
term of payment is also contingent on the survival of the annuitant. This
means that the payments are made for as long as the annuitant is alive and
24
the payments terminates upon the death of the annuitant. Since the time of
death of an individual is not pre-determined, the term of the life annuity is
uncertain. Thus, to determine the present value of the annuity, we consider
the interest rate as well as the time until death of the annuitant.
There are two methods in finding the present value of a life annuity.
These are the aggregate payment technique and the current payment tech-
nique. The steps in finding the present value using these techniques are:
1. Aggregate payment technique
(a) Record interest only present value of all payments to be made
by the annuity if death occurs at time t;
(b) Multiply the present value, found in (a.) by the probability or
the pdf of death at time t;
(c) Add/Integrate over all times of death t.
2. Current payment technique
25
As it was with Theory of Interest, there is another classification of life
annuities depending on the time when the annuity payment is made. We
have the continuous life annuity wherein the annual payments is payable
continuously within the year. We also have the life annuity-due wherein
the annual payments are made at the beginning of each year; and the life
annuity-immediate, wherein the annual payments are made at the end of
each year.
In the aggregate payment technique, we will denote by Y , the random
variable representing the present value of all payments considering interest
alone. Thus, Y is a function of T , when the annuity payments are made
continuous and is a function of K when the annuity payments are made
either at the beginning or at the end of each year.
26
Then, the pdf of Y is
d d log(1 − δy)
fY (y) = FY (y) = FT [− ]
dy dy δ
1 log(1 − δy) 1
= fT [− ] for 0<y< .
1 − δy δ δ
We can see that the cdf and pdf of Y are dependent on the cdf and the
pdf of T . To show the relationship of Āx and āx , we have
Z ∞
Āx = v t t px µx (t)dt.
0
1 − vT 1−Z
Y = āT | = = ,
δ δ
where Z = v T , for T > 0. Then,
1 − E[Z] 1 − Āx
āx = E[Y ] = = .
δ δ
Furthermore,
1−Z 1 1
V ar[Y ] = V ar[āx ] = V ar[ ] = 2 V ar[v T ] = 2 [ 2Āx − (Āx )2 ].
δ δ δ
Consider now an n-year temporary life annuity. Then under the aggre-
gate payment technique, the random variable Y is defined as
(
āT | if T < n
Y = .
ān| of T ≥ n
Thus,
Z n Z ∞
āx:n| = E[Y ] = āt| t px µx (t) + ān| t px µx (t)
0 n
Z n
= āt| t px µx (t) + ān| t px ,
0
since Z ∞
t px µx (t)dt = t qx |∞
n = 1 − n qx = n px .
n
27
Under the current payment technique, we have
Z n
āx:n| = v t t px dt.
0
To show the relationship between Āx:n| and āx:n| , we have
(
āT | if T < n
Y = .
ān| of T ≥ n
is equivalent to
vT
1−Z if T < n
Y = where Z= .
δ vn if T ≥ n
Thus,
1−Z 1 − Āx:n|
āx:n| = E[Y ] = E[ ]= .
δ δ
Also,
1−Z 1 1
V ar[Y ] = V ar[ ] = 2 V ar[Z] = 2 [ 2Āx:n| − (Āx:n| )2 ].
δ δ δ
We note that the actuarial accumulated value value at the end of the
term of an n-year temporary life annuity of 1 per year payable continuously
while (x) survives is denoted by s̄x:n| . Thus,
Z n t Z n Z n
1 v t px t Ex 1
s̄x:n| = āx:n| = dt = dt = dt.
n Ex 0 n Ex 0 n Ex 0 n−t Ex+t
We now consider the m-year deferred whole life annuity. Under the
aggregate payment technique, we have
0 if T < n
Y =
v n āT −n| if T ≥ n
or equivalently,
0 if T < n
Y =
āT | − ān| if T ≥ n
Thus,
Z ∞
n| āx = E[Y ] = (āt| − ān| )t px µx (t)dt
n
Z ∞
= v n āt−n| t px µx (t)dt
n
Z ∞
n
= v ās| n+s px µx+n (s)ds let t − n = s
0
Z ∞
= v n n px ās| s px+n µx+n (s)ds
0
= n Ex āx+n
28
Under the current payment technique,
Z ∞
n| āx = v t t px dt.
n
Moreover,
Z ∞ Z ∞
v t t px dt = v n+s n+s px ds let t=n+s
n 0
Z ∞
n
= v n px v s s px+n ds
0
= n Ex āx+n .
Also,
Z ∞ Z ∞ Z n
t t
n| āx = v t px dt = v t px dt − v t t px dt
n 0 0
1 − Āx 1 − Āx:n|
= āx − āx:n| = −
δ δ
Āx:n| − Āx
=
δ
2. V ar[āT | ];
Y = äK+1| for K = 0, 1, 2, . . . .
Thus,
∞
X
äx = E[Y ] = E[ äK+1| ] = äK+1| k px qx+k .
0
29
1−v K+1
Also, since äK+1| = d , we have
1 − v K+1 1 − Ax
äx = E[ äK+1| ] = E[ ]= .
d d
Moreover,
1 − v K+1 1 1
V ar[Y ] = V ar[äK+1| ] = V ar[ ] = 2 V ar[v K=1 ] = 2 [ 2Ax −(Ax )2 ].
d d d
Under the current payment technique,
∞
X ∞
X
äx = v k k px = k Ex .
k=0 k=0
For the n-year temporary life annuity due of 1 per year, we have under
the aggregate payment technique
äK+1| if K = 0, 1, 2, . . . n − 1
Y =
än| if K = n, n + 1, . . .
Thus,
n−1
X ∞
X
äx:n| = E[Y ] = äk+1| k px qx+k + äk+1| k px qx+k
k=0 k=n
n−1
X
= äk+1| k px qx+k + än| n px ,
k=0
v K+1
1−Z if K = 0, 1, 2, . . . n − 1
Y = where Y = .
d vn if K = n, n + 1, . . .
Then,
1−Z 1 − Ax:n|
äx:n| = E[Y ] = E[ ]= ,
d d
and
1−Z 1 1
V ar[Y ] = V ar[ ] = 2 V ar[Z] = 2 [ 2Ax:n| − (Ax:n| )2 ].
d d d
We now consider an n-year deferred whole life annuity-due of 1. Under
the aggregate payment technique,
0 if K = 0, 1, 2, . . . n − 1
Y =
v n äK−n| if K = n, n + 1, . . . .
30
Thus,
∞
X
n| äx = E[Y ] = v n äk−n| k px qx+k .
k=n
Furthermore,
1 − Ax 1 − Ax:n| Ax:n| − Ax
n| äx = äx − äx:n| = − = .
d d d
Finally,
∞
X ∞
X
n| äx = v k k px qx+k = v n n px v k k px+n = n Ex äx+n .
k=n k=0
Y = aK| where K = 1, 2, 3, . . . .
It follows that
1 − vK 1 − (1 + i)v K+1 1 − (1 + i)Ax
ax = E[Y ] = E[aK| ] = E[ ] = E[ ]= .
i i i
For the n-year temporary life annuity immediate of 1 per year, the apv
is
n
X
ax:n| = v k k px = äx:n| − 1 + n Ex .
k=1
For an n-year deferred whole life annuity immediate of 1 per year, the
apv is
X∞
a
n| x = v k k px .
k=n+1
31
It is very easy to show that
Therefore,
Ax = väx − ax .
Analogously, it can be shown that
1
Ax:n| = väx:n| − ax:n| .
Also,
1
Ax:n| = Ax:n| + n Ex = väx:n| − ax:n| + n Ex = väx:n| − ax:n−1|
Then,
J+1
1 − v K+ m 1 − A(m)
ä(m)
x = E[Y ] = E[ (m)
]= .
d d(m)
Under current payment technique, we have
∞
1 X h
ä(m)
x = vm h px .
m m
h=0
(m)
We show a relationship between äx and äx . We know that
32
or equivalently,
d 1
ä(m)
x = äx − (A(m)
x − Ax ).
d(m) d(m)
(m) i (m) (m)
Applying UDD, Ax = A
i(m) x
=s Ax and since d(m) ä = d, then,
1| 1|
(m)
s −1
d (m) id 1| i−i(m)
where α(m) = d(m)
s = i(m) d(m)
and β(m) = d(m) = i(m) d(m)
.
1|
(m)
A simpler relationship is äx = äx − m−12m .
Example 3.3. On the basis of the illustrative life table with interest at
i = 0.06, calculate the APV of a whole life annuity due of 1000 per month
for a retiree aged 65.
1
2. (Ex. 4.2) Let µ(x) = 1+x , ∀x > 0.
d
(b) Use (a) to show that dx Āx < 0, ∀x > 0.
dĀx
= −v(I¯Ā)x .
di
33
(a) Calculate Ā140:25| ;
(b) Determine the APV for a 25-year term insurance with benefit
amount for death at time t, bt = e0.05t , for (40) at policy issue.
5. (Ex. 4.9)
7. (Ex. 4.11) The r. v. Z, is the PV r.v. for the whole life insurance
of unit amount payable at the moment of death and issued to (x). If
δ = 0.05 and µx (t) = 0.01,
1
(a) Ax:20| ; (b) A1x:20|
(a) Describe the benefit of the insurance with APV given by the
symbol (IA)x:m| .
(b) Express the APV of (a) in terms of the symbols in Tables 4.2.1
and 4.3.1.
11. (Ex. 4.20) Show that under the constant force of mortality assump-
tion between integral ages
∞
X i + qx+k
Āx = v k+1 k px µx (k) ,
δ + µx (k)
k=0
34
12. (Ex. 4.21)
(a) Show that Z ∞
Āx = v t t px µx (t)dt
0
can be rewritten as
Z ∞
1
Āx = x
v y y p0 µy dy, x ≥ 0.
x p0 v x
4 BENEFIT PREMIUMS
4.1 Introduction
In practice life insurance is purchased by a life annuity of contract(gross)
premiums. These contract premiums provide not only for the benefit to
paid but also for the expenses in initiating and maintaining the insurance
and margins for profits for the insurance company. In contrast, benefit
premiums provide only for the benefit. In this chapter we will discuss
benefit premiums. The benefit premiums we will consider are those which
satisfy the equivalence principle. This principle satisfies,
E[L] = 0,
where L is called the insurer’s loss defined as the random variable of the
present value of the benefits to be paid by the insurer less the present value
of the benefit premiums to be paid by the insured. In other words,
35
5 Fully Continuous Premiums
In general the random variable L is defined as
L = Z − P̄ Y,
where Z is the present value of the benefit as defined in Chapter 2, Y is the
present value of the life annuity as defined in Chapter 3 and P̄ is the annual
premium payable continuously. If the equivalence principle is applied then,
E[L] = E[Z − P̄ Y ] = E[Z] − P̄ E[Y ] = 0.
Thus,
E[Z]
.
P̄ =
E[Y ]
Consider a fully continuous level annual benefit premium for a unit
amount whole life insurance payable at the moment of death of (x), then,
L = v T − P̄ āT | where T > 0.
1 − vT
V ar[L] = V ar[v T − P̄ āT | ] = V ar[v T − P̄ ]
δ
P̄ P̄ P̄
= V ar[v T (1 + ) − ] = (1 + )2 V ar[v T ]
δ δ δ
P̄ 2 2
= (1 + ) [ Āx − (Āx ) ]
δ
Āx
Since P̄ = āx , then
36
Then, the pdf of L is
d 1 δu + P̄ 1
fL (u) = FL (u) = {fT (− log( ))}( ),
du δ δ + P̄ δu + P̄
37
Then,
L = Z − P Y = v K+1 − P äK+1| .
This implies that using the equivalence principle
Ax
Px = P = .
äx
It can be shown easily that
2
Ax − (Ax )2
V ar[L] = .
(däx )2
0.04
Example 6.1. Suppose k| qx = c(0.96)k+1 , where k = 0, 1, 2 . . ., c = 0.96
and i = 0.06, find Px .
Example 6.2. Consider a 10,000 fully discrete whole life insurance. Let π
denote an annual premium for this policy and L(π) denote the loss-at-issue
random variable for one such policy on the basis of the Illustrative Life
Table, 6% interest and issue age 35.
1. Determine the premium πa , such that the loss of L(πa ) has mean 0.
Calculate the variance of L(πa ).
2. Approximate the smallest nonnegative premium πb , such that the
probability is less than 0.5 that the loss L(πb ) is positive. Find the
variance of L(πb ).
3. Determine the premium πc , such that the probability of a positive
total loss on 100 such independent policies is 0.05 by the normal
approximation.
38
We note that the actual amount of each fractional premium is P (m) /m.
The payment term h is in years. Furthermore, we express the fractional
premium as a multiple of the corresponding annual premium. Thus, for
(m)
instance we express h P as
x:n|
(m)
äx:h|
hP = h Px:n| .
x:n| (m)
ä
x:h|
Example 6.3. Calculate the level annual benefit premium payable in semi-
annual installments for a 10,000, 20-year endowment insurance with pro-
ceeds paid at the end of the policy year of death (discrete) issued to (50),
on the basis of the Illustrative Life Table with i = 0.06.
2. (Ex. 5.2) Using the values obtained in Exercise 5.1, calculate the
standard deviation and the coefficient of variation σµ , of the following
present value random variables.
(a) Individual annuities issued at ages 20, 50, 80 with life incomes
of 1000 per year payable continuously;
(b) A group of 100 annuities, each issued at age 50, with life incomes
of 1000 per year payable continuously.
3. (Ex. 5.6) Assume µx (t) = µ and δt = δ, t ≥ 0. Given the following
definition for Y , display its distribution function.
(
(a) Y = āT | , T ≥ 0 0 0≤T <n
(c) Y =
āT | − ān| T ≥ n
( (
āT | 0 ≤ T < n ān| 0 ≤ T < n
(b) Y = (d) Y =
ān| T ≥ n āT | T ≥ n
39
Ax:n| −Ax
(a) ax:n| = 1 Ex äx+1:n| (b) n| ax = − n Ex
d
40
7 BENEFIT PREMIUMS
7.1 Introduction
In practice life insurance is purchased by a life annuity of contract(gross)
premiums. These contract premiums provide not only for the benefit to
paid but also for the expenses in initiating and maintaining the insurance
and margins for profits for the insurance company. In contrast, benefit
premiums provide only for the benefit. In this chapter we will discuss
benefit premiums. The benefit premiums we will consider are those which
satisfy the equivalence principle. This principle satisfies,
E[L] = 0,
where L is called the insurer’s loss defined as the random variable of the
present value of the benefits to be paid by the insurer less the present value
of the benefit premiums to be paid by the insured. In other words,
L = Z − P̄ Y,
Thus,
E[Z]
P̄ = .
E[Y ]
Consider a fully continuous level annual benefit premium for a unit
amount whole life insurance payable at the moment of death of (x), then,
E[v T ] Āx
P̄ (Āx ) = P̄ = = .
E[āT | ] āx
41
1 − vT
V ar[L] = V ar[v T − P̄ āT | ] = V ar[v T − P̄ ]
δ
P̄ P̄ P̄
= V ar[v T (1 + ) − ] = (1 + )2 V ar[v T ]
δ δ δ
P̄ 2 2
= (1 + ) [ Āx − (Āx ) ]
δ
Āx
Since P̄ = āx , then
42
Example 7.1. Under the constant force of mortality µ)x(t) = µ. Given
the force of interest δ, find an expression for P̄ (Āx ).
SOLUTION:
Z ∞ Z ∞ Z ∞
µ
Āx = v t t px µx (t)dt = e−δt e−µt µdt = µ e−(µ+δ) dt = .
0 0 0 µ+δ
Z ∞ Z ∞ Z ∞
1
āx = v t t px dt = e−δt e−µt dt = e−(µ+δ) dt = .
0 0 0 µ+δ
Therefore,
µ
Āx µ+δ
P̄ (Āx ) = = 1 = µ.
āx µ+δ
Example 7.2. Show that the variance of the loss L, associated with an
n-year endowment insurance can be expressed as:
2
Āx:n| − (Āx:n| )2
V ar(L) = .
(δāx:n| )2
SOLUTION:
Define Z3 = Z1 + Z2 , where
T
v if T ≤ n 0 if T ≤ n
Z1 = and Z2 = .
0 if T > n vn if T > n
Therefore,
2
P̄ P̄ P̄
V ar[L] = V ar Z3 1 + − = 1+ V ar[Z3 ]
δ δ δ
2
P̄
2 Āx:n| − (Āx:n| )2
= 1+ [2 Āx:n| − (Āx:n| )2 ] =
δ (δāx:n| )2
43
7.3 Fully Discrete Premiums
To find the benefit discrete premium P , we follow the same procedure as
in finding P̄ . Consider level annual premiums payable at the beginning of
each year for a unit whole life insurance payable at the end of the year of
death. Then, for K = 0, 1, 2, . . ..
Z = v K+1 and Y = äK+1| .
Then,
L = Z − P Y = v K+1 − P äK+1| .
This implies that using the equivalence principle
Ax
Px = P = .
äx
It can be shown easily that
2
Ax − (Ax )2
V ar[L] = .
(däx )2
0.04
Example 7.3. Suppose k| qx = c(0.96)k+1 , where k = 0, 1, 2 . . ., c = 0.96
and i = 0.06, find Px .
SOLUTION: P∞ a
We know that the sum of an infinite geometric series is k=0 ark = 1−r
and d = 1 − (1 + i)−1 .
∞ ∞ ∞ k+1
X
k+1
X
−(k+1) k+1
X 0.96
Ax = v k| qx =c (1 − 06) (0.96) =c
1.06
k=0 k=0 k=0
0.96
0.04 1.06
= = 0.4
0.96 1 − 0.96
1.06
1 − Ax 1 − 0.4
äx = = = 10.60.
d 1 − (1.06)−1
Therefore,
Ax 0.4
= Px = = 0.0377.
äx 10.6
To find V ar[L], we first determine 2 Ax .
∞ ∞ ∞ k+1
X X X 0.96
2
Ax = v 2(k+1) k| qx = c (1 − 06)−2(k+1) (0.96)k+1 = c
(1.06)2
k=0 k=0 k=0
0.96
!
0.04 (1.06)2
= 0.96 = 0.2445.
0.96 1 − (1.06) 2
Therefore,
2
Ax − (Ax )2 0.2445 − (0.4)2
V ar[L] = 2
= = 0.2347.
(däx ) [(0.0566)(10.60)]2
44
7.4 True m-thly Payment Premiums
Most of the time, the premium paid for an insurance policy is made several
times in a year, instead of once a year. The premium paid m times per
year, with no adjustment in the death benefit is called the true fractional
(m)
premiums. The notations Px and P (m) (Āx ) represents the true level
annual benefit premium, payable in mthly installments, for a a unit whole
life insurance payable at the end of the year of death and at the moment
of death respectively.
We note that the actual amount of each fractional premium is P (m) /m.
The payment term h is in years. Furthermore, we express the fractional
premium as a multiple of the corresponding annual premium. Thus, for
(m)
instance we express h P as
x:n|
(m)
äx:h|
hP = h Px:n| .
x:n| (m)
ä
x:h|
Example 7.4. Calculate the level annual benefit premium payable in semi-
annual installments for a 10,000, 20-year endowment insurance with pro-
ceeds paid at the end of the policy year of death (discrete) issued to (50),
on the basis of the Illustrative Life Table with i = 0.06.
SOLUTION: From the Illustrative Life Table, we get
19
X
ä50:20| = v k k p50 = 11.291832.
k=0
45
Therefore,
(2)
A50:20| 0.036083889
10000P = 10000 (2)
= 10000 = 325.19.
50:20|
ä 11.096159
50:20|
Exercise 7.1. The following problems were lifted from the textbook “Ac-
tuarial Mathematics” (2nd edition) by Bowers, Gerber, Hickman, Jones
and Nesbitt. These problems are exercises from Chapter 6 found on pp.
197-202.
1. (Ex. 6.4) A fully continuous whole life insurance with unit benefit
has a level premium. The r.v T (x), has an exponential distribution
with E[T (x)] = 50 and the force of interest is δ = 0.06. If the principle
of equivalence is used, find the benefit premium rate.
2. (Ex. 6.6) Derive an expression for
2
Āx − (Āx )2
,
(δāx )2
46
8 Benefit Reserves
We note the following:
• At the issuance of an insurance or an annuity, the APV of the pay-
ment(s) to be made by the insurer is exactly equal to the APV of
the net premiums to be made by the insured. This is due to the
equivalence principle.
• As time goes on, the APV of the premiums paid by the insured gener-
ally decreases and for the insurer, if it is an annuity the APV generally
decreases and if it is an insurance the APV generally increases.
• This implies that at time t 6= 0, the value of the difference between
these two values is not zero.
• Thus, it is important that we have a measure of this difference because
the insurer needs to have enough funds when payments for benefits
are required to be given sometime at time t 6= 0.
• The difference of the value of the benefit at time over the value of
the benefit premiums at time t 6= 0. This is what we use to define
reserves.
tL = v u − P̄ āu|
tL = v u − P̄ (Āx )āu| ,
then the benefit reserve is E[tL]. Let E[t L] = t (V̄ Āx ). Then,
47
1. P̄ (Ā35 );
2. t (V̄ Ā35 );
We consider reserves related to benefit annual premiums payable at the
beginning of each year and insurance benefit payable at the end of the year
of death.
• Let J denote the discrete RV representing the curt ate future lifetime
of (x + k) with pmf
kL = v J+1 − P äJ+1| .
kL = v J+1 − Px äJ+1| .
k Vx = E[k L]
= E[v J+1 ] − Px E[äJ+1| ]
= Ax+k − Px äx+k
0.04
Example 8.2. If k qx = c(0.96)k+1 , k = 0, 1, 2, . . . where c = 0.96 and
i = 0.06, calculate
1. Px ;
2. k Vx
48