Chapter 1
Survival Distributions and Life Tables
1.1 Introduction to Probability
1.1.1 Random Experiment
Any process with several possible outcomes, none of which is known in advance with certainty, e.g.: tossing a coin; winning a Lotto number; death claims during a year incurred by an insurance company.
1.1.2 Sample Space
Set of outcomes of a random experiment.
Example 1.1.1 Tossing a coin S = {H, T}
Winning Lotto number S =
{
(
446)
possible outcomes}
Death claims in a year S = {0, 1, 2 …N}
where N = total number of insured Life expectancy of a newborn S = [0, ω]
where ω = limiting age at death
1.1.3 Event
Any subset of a sample space
Example 1.1.2 S = Death claims in a year
A = {female deaths}
B = {deaths with face amounts > $250,000}
The probability of an event A is denoted by P(A) is some value in [0, 1] indicating the
likelihood of the occurrence of event A.
P(A) = 1 ⟹ A is certain to occur P(A) = 0 ⟹ A is certain not to occur
Example 1.1.3 Toss a coin 3 times: A = {≤3 heads occur}, B = {4 heads occur}. Then
⟹P(A) = 1 and P(B) = 0.
Facts:
3. If A and B are disjoint events, then P(A∪B)=P(A)+P(B)
1.1.4 Finite Sample Spaces
Let S = { s1, s2, … sn}, P
{
si}
=pi , i = 1, 2, …n. Then∑
i=1 n
pi=P{S}=1∧if A⊆S , P(A)=
∑
siϵAP{si}
Example 1.1.4 Roll a fair die: S = {1, 2, 3, 4, 5, 6}, then P{1}=P{2}=…=P{6}=1/6 A = {even numbers occurs} = {2, 4, 6}
P(A)=P{2}+P{4}+P{6}=1/2
1.1.5 Continuous Sample Spaces
The probability of an event is defined by a probability density function (p.d.f.) f(t) such that 0≤ f(t)≤1 , t ϵ S and
∫
t∈S
f(t)dt=1 . Then
P(A)=
∫
t∈Af(t)dt
Example 1.1.5 Randomly generate a number in [0, 1];
f(t) = 1, t ϵ [0, 1], A = {number≤0.25} ⟹ P(A)=
∫
0 0.25
f(t)dt=0.25
1.1.6 Independent Events
Two events A, B are independent if P(A ∩ B)=P(AB)=P(A)P(B)
Example 1.1.6 Toss a fair coin twice: S = {HH, HT, TH, TT} and P{HH} = 1 4. Using independence:
P{HH}=P
{
heads∈1sttoss∧heads∈2ndtoss}
=(
1 2)(
1
2
)
=1 4
1.1.7 Conditional Probability
P(A|B)=P(AB) P(B)
Example 1.1.7 Roll a fair die: A = {even number occurs}, B = {number > 2 occurs} Directly, P(A|B)=P¿
By formula, P
(
A|B)
=P(AB) /P(B)=P{4,6}/P{3,4,5,6}=2 6 4 6
=1
2
Similarly, P(B|A)=P(AB)/P(A)= 2 6 3 6
=2 3
Facts:
1. P
(
B|B)
=12. If A , B are independent P A|B¿=P(AB) P(B) =
P(A)P(B)
P(B) =P(A) i.e. the occurrence of B has no influence on the occurrence of A. Similarly, P
{
B|A}
=P{B}.1.1.8 Random Variable
A random variable X is a real valued function defined on a sample space S, i.e. for every outcome s ∈ S, X assigns a real number X(s).
Example 1.1.8 Toss a coin twice: S ={HH, HT, TH, TT} Let X = number of heads
obtained. Then X{HH} = 2, X{HT} = X{TH} = 1, X{TT} = 0,. i.e. X = {0, 1, 2}. If the coin is fair,
P{X=0}=P{TT}=1 4
P{X=1}=P{HT}+P{TH}=1 2
P{X=2}=P{HH}=1 4
Continuous random variable is a real valued continuous function defined on a sample
space S.
1.1.9 Probability Distribution of a Random Variable
Given a probability distribution on a sample space S, this uniquely determines a probability distribution for any random variable X defined on S, since
P{X ≤t}=P
{
s:X(s)≤t}
.1.1.10 Discrete Distributions
X takes on a finite set of values and the probability distribution f(x) = P{X=x}
Example 1.1.10 Roll a fair die, X = outcome. Then
f(x)=P{X=x}=
{
1/6for x=1,2, … ,60otherwise
Fact: f(x)≥0∧
∑
x
f(x)=¿P{S}=1¿.
1.1.11 Continuous Distributions
X is a continuous random variable and there exists a function f, s.t. 0≤ f(t)≤1, for all t,
∫
−∞
∞
f(x)dx=1and P{X ≤t}=
∫
−∞
t
f(x)dx,
then f(x) is the probability density function of x.
Example 1.1.11 Generate a random variable from [0, 1], X = outcome. Then
f(x)=
{
1for0≤ x ≤10otherwise
Example 1.1.12 X = standard normal random variable. Then
f(x)= 1
√
2πe−x2
2 ,−∞ ≤ x ≤ ∞
1.1.12 Distribution Function of a Random Variable
F(x)=P{X ≤ x}=
{
∑
t ≤ xf(t)if x is discrete
∫
−∞
x
By the fundamental Theorem of Calculus: f(x)=∂ F(x) ∂ x .
1.1.13 Expected Value of a Random Variable
E(X)=Expected value of X=
{
∑
xx P{X=x}if x is discrete
∫
−∞
∞
t f(t)dt if x is continuous
Interpretation: The expected value is the weighted average of the values of X where the weights are based on the probability of occurrence of these values.
Example 1.1.13 Consider an insurance policy which pays out $100,000 if an individual
dies during the year. Let X = payout. SupposeP{individual dies}=0.01. Then
{
P[X=$100,000]=0.01P[X=0]=0.99
E(X)=0.01(100,000)+0.99(0)=1,000.
1.1.14 Variance of a Random Variable
Discrete Case:
Var(X)=variance of X=
∑
x(
x−E(X))
2P{X=x}¿
(
∑
x
x2P{X=x}
)
−(
E(X))
2¿E
(
X2)
−(
E(X))
2Continuous Case:
Var(X)=
∫
−∞
∞
(
x−E(X))
2f(x)dx¿
∫
−∞
∞ x2f
(x)dx−
(
E(X))
2¿E
(
X2)
−(
E(X))
2Interpretation:
Var(X) = probability weighted measure of spread of the values of X.
Var(X)=(100,000−1,000)20.01+(0−1,000)20.99
¿98,010,000+990,000=99,000,000
Alternatively,
Var(X)=E
(
X2)
−(
E(X))
2=(100,000)20.01+(0)20.99−10002¿100,000,000−1,000,000=99,000,000
Standard deviation of X SD(X)=
√
Var(X)In the previous example, SD(X) =
√
99,000,000≈9,9501.1.15 Survival Distribution of a Random Variable
Define the random variable X = age at death of a newborn F(x) = distribution function of X
= P{X ≤ x}
¿P{newborn diesby age x}
S(x) = 1 – F(x)
= survival function of X
¿P{X>x}
¿P{newborn dies after age x}
¿P{newborn survives¿age x}
Note
1) S(0)=1
2) 0≤ S(x)≤1, x ≥0
3) S(x)>S(y), y>x
i.e.S(x) is a strictly decreasing function of x.
4) limx→ ωS(x)=0, where ω=limitingage at death
Any function possessing the above four properties qualifies for a survival function.
Example 1.1.15
S(x)=e−x, x ≥0
S(x)=1−x/100,0≤ x ≤100
Any probability function of interest can be expressed in terms of survival functions.
Example 1.1.16 P{newborn dies between30∧40}=P{30<X<40}
¿S(30)−S(40)
S(30)
______________________________________________________________ 0 30 40
S(40)
P{X ≤30}=1−S(30)etc .
Example 1.1.17 S(x)= 1
10
√
100−x ,0≤ x ≤100Find the probability that a life aged 0 will die between age 19 and age 36.
S(19)
______________________________________________________________ 0 19 36
S(36)
Probability = S(19)−S(36)
¿ 1
10
[
√
81−√
64]
¿ 1
10
1.2 Future Life time of Individual Age x
T(x) = future lifetime of an individual age (x)= time until death for a person age (x)
1.2.1 Distribution function of T(x)
P[T(x)≤ t]=tqx
________________ tqx ___________________
__________________________________________________________________ x x+t
tpx=1−tqx
¿P[T(x)>t]
1.2.2 Relationship Between T(x) and x
X = time until death of a newborn T(x) = time until death for a person age x
1.X=T(0)
2.tqx=P
{
(x)dies within t years}
¿P
{
newborn dies between x∧x+t|newborn survived¿age x}
¿P{newborn dies between x∧x+t}P{newborn survived¿age x}
¿S(x)−S(x+t)
s(x)
3. tpx=1−tqx
¿S(x+t)
S(x)
¿P
{
newborn survives¿x+t|newborn survived¿age x}
1.2.3 Other Probability Functions of T(x)
1qx=qx
¿P
{
(x)dies within a year}
¿S(x)−S(x+1)
S(x)
________________ qx ___________________
__________________________________________________________________ x x+1
1px=px
¿1−qx
¿s(x+1)
s(x)
t|uqx=P
{
(x)dies between x+t∧x+t+u}
¿S(x+t)−S(x+t+u)
¿tpx−t+upx=tpx uqx+t
________ t|uqx ___________
__________________________________________________________________ x x+t x+t+u
t|1qx=t|qx=P[(x)diesbetween x+t∧x+t+1]
¿S(x+t)−S(x+t+1)
S(x)
¿tpx−t+1px=tpxqx+t
1.2.4 Curtate Future Lifetime of x
K(x) = curtate future lifetime of (x) P[K(x)=k]=P[k<T(x)≤ k+1]
¿P
{
(x)dies between x+k∧x+k+1}
¿k|qx
Note
∑
k=0 ∞
P[K(x)=k]=
∑
k=0 ∞k|qx=1
where 0|qx=qx.
i.ek|qxisthe probability distributionfor the discrete random variable K(x).
If ω=limiting age at death, then
∑
k=0 ω−x−1P
{
K(x)=k}
=∑
k=0 ω−x−1k|qx=1
Example 1.2.1 S(x)= 1
10
√
100−x ,0≤ x ≤100a) Find the probability that a life age 19 will die before age 36.
b) Find the probability that a life age 19 dies between ages 36 and 75.
Answer:
________ 17q19 __________
__________________________________________________________________
0 19 36
Probability=17q19=S(19)−S(36) S(19) =
1
10
[
√
81−√
64]
1
10
√
81=1 9
b¿ ________ 17|39q19 __________ __________________________________________________________________
0 19 36 75
Probability=17|39q19=S(36)−S(75) S(19)
¿
1
10
[√
64−√
25]
110
√
81¿1
3
Alternatively, Probability=17|39q19=17p19.39q36
¿ S(36)
S(19).
S(36)−S(75) S(36) =
1 3
1.2.5 Force of Mortality
μx = Force of Mortality at age x
¿
−∂ ∂ x S(x)
S(x)
¿
−∂
∂ x
[
1−F(x)]
1−F(x)
¿ f(x)
Where F(x) = distribution function of X, and X = age of a newborn, f(x) = probability density function of X.
Interpretation: Since ∂
∂ x g(x)=∆ x →lim0
g(x+∆ x)−g(x)
∆ x ,
μx=lim ∆ x→0
S(x)−S(x+∆ x)
∆ x S(x)
¿ lim
∆ x→0
P{x dies between x∧x+∆ x} ∆ x
¿instataneous rate of deathof x .
1.2.6 Relationship to Other Mortality Functions
Since ∂
∂ x log
[
g(x)]
=g '(x)/g(x) μx=−∂
∂ xlog
[
S(x)]
.i.e.
∫
x x+nμydy=
∫
x x+n−∂
∂ y log
[
S(y)]
dy=log[
S(x)]
−log[
S(x+n)]
¿log
[
S(x)S(x+n)
]
=¿log(npx)¿
⟹ npx=e
−∫
x x+n
μydy
¿particular:
∫
0 x
μydy=−logxp0=−log[S(x)]
⟹S(x)=e
−∫
0
x μydy
T(x) = time until death for a person age x
tqx = distribution function of T(x) = P
{
T(x)≤ t}
=G(t)∂
∂ ttqx = g(t)
= probability density function of T(x)
¿ ∂
∂ t
S(x)−S(x+t)
S(x)
¿−S '
(x+t)
¿− S'(x
+t)
S(x+t)∗S(x+t)
S(x)
¿tpxμx+t
i.e. tpxμx+t= p.d.f. of T(x). Then
∫
0 ∞
tpxμx+tdt=1
∫
n n+m
tpxμx+tdt=n|mqx
∫
0 r
tpxμx+tdt=rqx
∫
r ∞
tpxμx+tdt=rpx
In particular, for a newborn,
∫
0 ∞S(t)μtdt=
∫
0 ∞[
−∂∂ t S(t)]
dt=1∫
x ∞
S(t)μtdt=S(x)=e
−∫
0
x μtdt
1.2.7 Requirements for a Force of Mortality
For μxto be an acceptable force of mortality function, S(x)=e−∫
0
x
μtdt must be an acceptable survival function. i.e.
0≤ S(x)≤1, S(0)=1
y>x⟹S(x)>S(y)
lim
x→ ∞S(x)=¿0¿
Note
lim
x→ ∞S(x)=¿0⟹x → ∞lim
∫
0 xμydy=∞¿
Example 1.2.2 Which of the following functions can serve as a force of mortality?
1) BCx B > 0 , 0 < C < 1 , x ≥ 0
3) k(x+1)n k > 0 , n ≥ 0 , x ≥ 0
Answer:
1.¿S(x)=e
−∫
0
x μtdt
∫
0 x
B Cydy= BC y
logC¿
⟹S(x)=exp
[
−BlogC(C
x−1)
]
Check
S(x)≥0, S(0)=1
lim
x→ ∞S(x)=
exp(B)
logC ≠0, since0≤C ≤1 Hence μx=BC
x is not a force of mortality function.
2.¿
∫
0 x
B(y+1)−0.5dy=B
√
y+10.5
|
0x=2B[
√
x+1−1]
,i.e.S(x)=exp(−2B
[
√
x+1−1]
) CheckS(x)≥0, S(0)=1
lim
x→ ∞S(x)=0,
S(x) is strictly decreasing.
Hence μx=B(x+1)−0.5 is a legitimate force of mortality function.
3.¿
∫
0 x
k(y+1)ndy=k(y+1) n+1
n+1
|
x0=
k
n+1
[
(x+1) n+1−1]
⟹S(x)=exp
(
−kn+1
[
(x+1) n+1−1]
)
Check
lim
x→ ∞S(x)=0, S(x)is strictly decreasing
Hence μx=k(x+1)
n is a legitimate force of mortality function.
Example 1.2.3 The mortality pattern of a certain population may be described as follows: Out of every 98 lives born together, one dies annually until there are no survivors.
a) Find S(x) for this population Answer:
Number of survivors out of 98 at the end of x years is 98 – x. i.e.
S(x)=98−x
98 ,0≤ x ≤98
b) What is the probability that a life aged 30 will survive to attain age 35?
Answer:
________ 5p30 _______ ______________________________________________
30 35 Probability = 5p30
¿S(35)
S(30)
¿98−35
98−30
¿63
68
c) What is the force of mortality at 48?
Answer:
μx= −S'(x)
S(x)
¿ 1
98−x
S(x)=98−x
98 ∧S
'
⟹μ48= 1 98−48
¿0.02
1.3 Life Tables
Notations:
l0= initial group of newborns
L(x)= number of survivors to age x
Then L(x) Bin
[
l0, S(x)]
∧E[
L(x)]
=l0S(x)=lx Dxn = number of deaths between x and x + n from the l0 lives Then nDx Bin
[
l0, S(x)−S(x+n)]
∧E[
nDx]
=l0[
S(x)−S(x+n)]
¿lx−lx+n ¿ndx Since S(x)– S(x+n)=P[newborn dies between x∧x+n]
¿P[newborn survives¿x dies between x∧x+n]
¿P[newborn survives¿x]∗P[newborn dies between x∧x+n|survival¿x]
¿S(x)nqx
⟹E
[
nDx]
=l0[
S(x)nqx]
¿lxnqx=lx−lx+n
E
[
L(x+n)]
=l0S(x)npx=lxnpx=lx+nNotation:
1¿1Dx=Dx∧E
[
1Dx]
=1dx=dx=lx−lx+1=lxqx 2¿lx+1=lxpxμx= − ∂
∂ x S(x) S(x) =
− ∂
∂ xl0S(x) l0S(x)
= − ∂
∂ xlx lx
lx=l0S(x)=l0exp
[
−∫
0 xμydy
]
=l0∫
0 xS(t)μtdt=
∫
x ∞
ltμtdt
lx+n=lxnpx=lxexp
[
−∫
x x+nμydy
]
=lx∫
n ∞px
n μx+tdt=
∫
n ∞lx+tμx+tdt
lxnqx=lx−lx+n=lx
∫
0 npx
t μx+tdt=
∫
0 nlx+tμx+tdt
A life table is a table of expected survivors and expected deaths at each age starting with an initial cohort of l0 lives (called the radix of the table).
Example 1.3.1 Consider the following life table
Age x lx dx
0 100,000 501
1 99,499 504
2 98,995 506
3 98,489 509
4 97,980 512
5 97,468 514
a . P{newborn survives¿age5}=S(5)=l0S(5) l0
=l5 l0
=0.97468
b . P{newborn diesbetween3∧5}=S(3)−S(5)=
(
l3−l5)
l0¿
(
l3−l4)
+(
l4−l5)
l0
¿
(
d3−d4)
¿(509−512)
100,000
¿0.0102 1
c . P{1year old dies between3∧5}=s(2)−s(4) s(1) =
(
l2−l4)
l1¿
(
d2−d3)
l1
¿(506−509)
99,499
¿0.0102 0
1.3.2 Moments of T(x)
Recall T(x) = time until death for a person age x, T is continuous with p.d.f. f(t) = npxμx+t
and d.f. G(t) = P
[
T ≤t]
=tqx Integration by parts: Given functions f(x), g(x)∫
0 t
f (x)
[
∂∂ xg(x)
]
dx=f(x)g(x)|
0t−∫
0 tg(x)
[
∂∂ xf(x)
]
dxFor time untildeath r . v . T , for a given function z(T)s . t . z(T)
[
1−G(t)]
→0as t → ∞.E
[
z(T)]
=∫
0 ∞z(t)f(t)dt
¿
∫
0 ∞
−z(t) ∂
∂t
[
1−G(t)]
dt¿−z(t)
[
1−G(t)]
|
∞0+
∫
0 ∞[
∂∂t z(t)
]
[
1−G(t)]
dt¿z(0)+
∫
0 ∞[
∂∂t z(t)
]
tpxdt sinceG(0)=0qx=0¿particular , for z(T)=T , z(0)=0∧E(T)=
∫
0 ∞t pt xμx+tdt=
∫
0 ∞˙
ex=thecomplete expectetion of life at age x
For z(T) = T2 , z(0) = 0, z’(T) = 2T
E
(
T2)
=∫
0 ∞t2tpxdt=2
∫
0 ∞t pt xdt
Var(T)=E
(
T2)
−[
E(T)]
2=2∫
0 ∞t pt xdt−
[
e˙x]
2Example 1.3.2 Given the survival function
S(x)=e−0.05x, x ≥0
1.¿Calculate5|10q30
Answer:
________ 5|10q30 __________
__________________________________________________________________
30 35 45
q30 5|10 =
S(35)−S(45)
S(30) =
e−1.75−e−2.25 e−1.5 =e
−0.25−e−0.75
2.¿Calculate F(30)
Answer:
F(30)=P{X ≤30}=1−s(30)=1−e−1.5
3.¿Calculatee˙30
Answer:
˙ ex=
∫
0 ∞
px t dt=
∫
0 ∞
[
S(x+t)S(x)
]
dt˙ e30=
∫
0 ∞
e−0.05dt =−
(
e−0.05
0.05
)
¿ 1
0.05
¿20
4.¿Calculate Var[T(x)]
Answer:
E
[
T2]
=2∫
0 ∞t e−0.05t
dt=2
(
[
−t e−0.05t
0.05
]
|
∞0+∫
0 ∞e−0.05t
0.05 dt
)
¿ 2
0.05
∫
0∞ e−0.05t
dt
¿2
(
10.05
)
2
¿800
Var
[
T]
=E[
T2]
−[
E(T)]
2=800−400=400i .e .:Var
[
T(30)]
=4001.3.3 Median Future Lifetime of x
Median of T(x)is m(x)such that P
[
T>m]
=1/2i .e .:S(x+m) S(x) =1/2
Example 1.3.3 For S(x)=e−0.05x
, m(x)satifies
e−0.05(x+m)
e−0.05x = 1
2⟺e
−0.05m
=1 2
⟺−0.05m=log
(
12
)
=−log 2Recall the discrete curtate future lifetime of (x) random variable K(x) where P
{
K(x)=k}
=P{
k<T(x)≤ k+1}
=k|qx, k=1,2, … w−x−1P
{
K(x)≤ k}
=k+1qxP
{
K(x)>k}
=k+1pxTheorem 1.3.1
For any given function z(K)
E
[
z(K)]
=∑
t=0 w−x−1z(t)t|qx=z(0)+
∑
t=0 w−x−2Δ z(t)t+1px, where Δ z(t)=z(t+1)−z(t)
Proof:
E
[
z(K)]
=z(0)qx+z(1)1|qx+z(2)2|qx+…+z(w−x−1)w−x−1|qx¿z(0)qx+
[
z(0)+Δ z(0)]
1|qx+[
z(0)+Δ z(0)+Δ z(1)]
2|qx+…+[
z(0)+Δ z(0)+…+Δ z(w−x−2)]
w−x−1|qx¿z(0)
[
qx+1|qx+2|qx+…+w−x−1|qx]
+Δ z(0)[
1|qx+2|qx+…+w−x−1|qx]
+Δ z(1)[
2|qx+3|qx+…+w−x−1|qx]
+…+Δ z(w−x−2)w−x−1|qx¿z(0)+Δ z(0)px+Δ z(1)2px+…+Δ z(w−x−2)w−x−1px
¿z(0)+
∑
t=0 w−x−2Δ z(t)t+1px
In particular, forz(K) = K, z(0)=0, ∆ z(t)=1
E
[
K]
=∑
t=0 w−x−2t qt| x=
∑
t=0 w−x−2px t+1 =ex
ex=thecurtate expectation of life at age x .
For z(K)=K2, z(0)=0, ∆ z(t)=1
E
[
K2]
=
∑
t=0 w−x−2
t2 q x t| =
∑
t=0 w−x−2
(2t+1)t+1px
Var(K)=E
[
K2]
−[
E(K)]
2=∑
t=0 w−x−2(2t+1)t+1px+
(
ex)
2d90=6, d91=d92=3, d93=d94=d95=d96=2, d97=1
Calculate ex=E(K)∧Var(K)
Answer:
x lx dx t+1px (2t+1)t+1px t
90 21 6 15
21
15 21
0
91 15 3 12
21
36 21
1
92 12 3 9
21
45 21
2
93 9 2 7
21
49 21
3
94 7 2 5
21
45 21
4
95 5 2 3
21
33 21
5
96 3 2 1
21
13 21
6
97 1 1 0
21
0 21
7
98 0 0 −−¿ −−¿ 8
Total52 21
236 21
E(K)=ex=
∑
t=0 6p90 t+1 =
52 21=2.48
Var(K)=E
[
K2]
−[
E(K)]
2=∑
t=0 6(2t+1)t+1px−(2.48)2
¿236
21 −(2.48)
2=11.24−6.15=5.09
1.3.4 Total Years Lived Between x and x+1
For an individual age x, let Z = number of years lived between x and x +1. P{Z=1}=px
P{Z ≤ z}=
∫
0 zpx
i .e .:E(Z)=1.px+
∫
0 1t pt xμx+tdt
For lxlives age x , total expected number of years lived between x∧x+1is
Lx=lxpx+lx
∫
0 1t pt xμx+tdt=lx+1+
∫
0 1t lx+tμx+tdt
¿lx+1+
∫
0 1t[−∂
∂ t ¿lx+t]dt , since μx+t=
[
−∂∂ t lx+t
]
/lx+t¿¿lx+1+t
(
−lx+t)
|
10+
∫
0 1lx+td t ,
¿
∫
0 1
lx+tdt
i.e. Lx=
∫
0 1lx+tdt.
1.3.5 Central-Death-Rate at age x
mx=dx Lx=
∫
0 1
lx+tμx+tdt
∫
0 1
lx+tdt
¿
∫
0 1
px t μx+tdt
∫
0 1
px t dt
¿ deaths at age[x , x+1]
exposure at age[x , x+1]
1.3.6 Total Years Lived Beyond x
Recall:e˙x=complete expectation of life
¿expected number of years lived after x for an individual age x
¿lxe˙x=lx
∫
0 ∞t pt xμx+tdt=lx
∫
0 ∞px t dt=
∫
0 ∞
lx+tdt
1.3.7 Average Years Lived After x Amongst Deaths in [x, x+1]
Total deathsbetween
[
x , x+1]
=dx=∫
0 1lx+tμx+tdt
Expected number of years lived between
[
x , x+1]
=∫
0 1t lx+tμx+tdt
i.e. average number of years lived between
[
x , x+1]
amongst dxdeathsa(x)=
∫
0 1
t lx+tμx+tdt
dx =
∫
0 1
t lx+tμx+tdt
∫
0 1
lx+tμx+tdt
=
∫
0 1
t pt xμx+tdt
∫
0 1
px t μx+tdt
Example 1.3.5 Out of 100 lives age x, four die at age x + 1/4, two die at age x + 1/2 and four die at age x + 3/4.
Lx=total years lived between
[
x , x+1]
¿90+4
(
14
)
+2(
1 2)
+4(
3
4
)
=95mx=central death rate at age x=10 95
a(x)=average years lived between
[
x , x+1]
amongst dxdeaths∈[
x , x+1]
¿4¿ ¿ ¿1
2
1.4 Assumptions for Fractional Ages
Given a lifetable with values for qx, pxat integer values of x , we need¿estimate qx
1.4.1 Uniform Distribution of Deaths
qx
t =(t)qxfor0≤ t ≤1
i .e . pt x=1−tqx=1−(t)qx
Interpretation: For0≤ t+y ≤1
lx+y−lx+y+t=number of deaths betweenages x+y∧x+y+1
¿lxypx−lxy+tpx
¿lx
[
1−(y)qx]
−lx[1−(y+t)qx]¿lxt qx ¿t dx
i .e . number of deaths
any age interval t∈
[
x , x+1]
¿t
(
number of deaths between[
x , x+1])
i .e . deaths areuniformly distributed [x , x+1]
Other Results
1.) lx+t=lx−(t)dx=lx−t
[
lx−lx+1]
2.¿μx+t= − ∂
∂ tlx+t lx+t
= dx lx+t
=qx px t
= qx
(
1−(t)qx)
3.) tpxμx+t=qx
4.¿yqx+t=lx+t−lx+t+y lx+t =
(y)dx lxtpx =
(y)qx
1−(t)qx
μx+tis an increasing function of t:
lim t →1 μx+t
=μx+1= qx
(
1−qx)
=qxpx
lim
t →1 μx+t+1=μx+1=qx+1
i .e . μx+tis discontinous at the end points between successive age intervals .
1.4.2 Balducci Assumption
qx+t
1−t =(1−t)qx
Interpretation: Based on the assumption that the function 1/lx function is linear between x and x+1.
i .e . 1 lx+t=
1 lx+t
(
1 lx+1−
1 lx
)
ˇ :lx+1
lx+t= lx+1
lx +t
(
1− lx+1lx
)
i.e. 1−1−tqx+t=1−qx+(t)qx⟺1−tqx+t=(1−t)qx
Note
The Uniform Distribution of Deaths assumption assumes that the lx function is linear between x and x+1.
i .e .:lx+t=lx−t
[
lx−lx+1]
Other Results
1) Using the relationships
qx=tqx+tpx.1−tqx+t
¿1−tpx+tpx.1−tqx+t i .e . pt x
(
1−1−tqx+t)
=pxi .e . pt x= px
(
1−1−tqx+t)
= px2) tqx=1−tpx
¿1−(1−t)qx−px
1−(1−t)qx
=t qx 1−(1−t)qx
3) μx+t=−l'x+t/lx+t
Since 1 lx+t=
1 lx+t
(
1 lx+1−
1
lx
)
, differentiating implicitly we get−1
(
lx+t)
2. l'x+t= 1 lx+1−
1 lx
i .e . ,−l'x+t=
(
lx+t)
2(
lx1+1− 1 lx)
μx+t=−l '
x+t lx+t =lx+t
[
1 lx+1−
1 lx
]
Butlx+1
lx+t
=1−1−tqx+1=1−(1−t). qx
and lx+t lx
=tpx=px/ [1−(1−t). qx]
i.e. μx+t=lx+t lx+1
−lx+t lx
¿ 1
1−(1−t)qx−
px
1−(1−t)qx
¿ qx
1−(1−t)qx
4) tpxμx+t=
[
px 1−(1−t)qx
][
qx
1−(1−t)qx
]
=pxqx
[
1−(1−t)qx]
2Note
¿
(
1−[
1−t]
qx)
qx1−(1−t)qx
¿qx
Recall under UDD assumption, tpxμx+t=qx
b¿Under B alducci , μx+1−t= qx
1−t . qx
¿μx+tunder UDD
5) Using the relationship
px+t
1−t =yqx+t+ypx+t.1−t−yqx+t+y
--- ---x ---x+t ---x+t+y ---x+1
We get 1−tqx=yqx+t+
(
1−yqx+t)
+(1−t−y)qxi .e . , qy x+t
[
1−(1−t−y)qx]
=(1−t)qx−(1−t−y)qxi .e . , qy x+t= yqx
1−(1−t−y)qx
1.4.3 Constant Force of Mortality
μx+t=μ , for0≤ t ≤1
Other Results
1.¿tpx=exp
{
−∫
x x+t
μydy
}
=exp{−μt}=(
px)
t2.¿tqx=1−tpx=1−exp{−μt}=1−
(
px)
t3.¿lx+t=lxtpx=lxexp{−μt}=lx
(
px)
t4.¿yqx+t=1−ypx+t=1−exp
{
−∫
x+t x+t+yμsds
}
=1−exp{−μy}=1−(
px)
yNote
2.¿tqx=tqx+y=1−exp{−μt}=1−
(
px)
t, for0≤ t+y ≤1
¿paricular q0.5 x=0.5qx+0.5
Under UDD , q0.5 x+0.5= (0.5)qx
1−(0.5)qx>(0.5)qx=0.5qx
Under Balducci , q0.5 x= (0.5)qx
1−(0.5)qx>(0.5)qx
3.¿Since px=exp{−μ}, pt x=exp{−μt}=
(
px)
t¿general , pt x+y=exp{−μt}=
(
px)
tExample 1.4.1 There are 100 persons now age 40 of whom 19 are expected to die
before age 41. Determine 0.5q40+0.375 assuming (a) UDD (b) Balducci and (c) Constant Force.
q40=0.19, p40=0.81, Want q0.5 40+0.375.
a¿Under UDD qx+t=¿ ¿ ¿
i .e . q0.5 40+0.375= (0.5)0.19
1−(0.375)0.19=0.102
Alternatively , q0.5 40+0.375=l40+0.375−l40+0.875 l40+0.375
¿ (0.5)19
100−(0.375)19=
9.5
92.875=0.102
b¿Under Balducci qx+t= (y)qx
1−(1−t−y)qx
y
i .e . q0.5 40+0.375= (0.5).19
1−(0.125)0.19=0.0097
c¿Under Constant Force qx+t=1−ypx+t=1−y¿
(
px)
y¿i .e . q0.5 40+0.375=1−0.810.5=0.1
Example 1.4.2 You are given
i. Deaths are uniformly distributed over each year of age. ii.
28
x lx
35 100
36 99
37 96
Which of the following are true? I. 1|2q36=0.091
II. m37=0.043 III. 0.33q38.5=0.021
Answer:
I . q1|2 36=l37−l39 l36
=96−87
99 =0.091⟹I istrue
II . m37=d37 L37
under UDD Lx=
∫
0 1lx+tdt=
∫
0 1(
lx−(t)dx)
dt=lx−(1/2)dxm37=d37 L37
= 96−92
96−(1/2)4=0.043⟹II istrue
III .0.33q38.5=l38.5−l38.83 l38.5
= 0.33(5)
92−0.5(5), since lx+t=lx−(t)dx
¿0.018⟹I II is false
1.4.4 Analytical Laws of Mortality
de Moivre
μx=(ω−x)−1, for0≤ x ≤ ω
Then S(x)=exp
{
−∫
0 xμydy
}
=exp{
−∫
0 x(ω−y)−1dy
}
¿exp
{
log(ω−y)|
x0
}
¿exp
{
log(
ω−x¿ω−x
ω
lx=l0S(x)=l0ω−x
ω =l0−x l0 ω
i .e . dx=lx−lx+1=l0 ω
i .e .:de Moivr e's Law assumes a contant l0
ωdeaths each year .
Also, density function of T(x)=tpxμx+t
¿s(x+t)
s(x) μx+t
¿ω−x−t
ω−x . 1 ω−x−t
¿ 1
ω−x, which is constant
Gompertz
μx=B c x
, B>0
Then S(x)=exp
{
−∫
0 xB cydy
}
¿exp
{
−B cy
logc
|
x0}
¿exp
{
−Blogc
(
cx−1
)
}
¿exp
{
−m(
cx−1)
}
, where m= Blogc
S(0)=1∧sincelim
x→ ∞S(x)=0⇒c>1
Whenc=1, μx=B=constant forceof mortality assumption∧S(x)=exp{−Bx}
Makeham
Then S(x)=exp
{
−∫
0 xA+B cydy
}
¿exp
{
−Ay−B c ylogc
|
0x}
¿exp
{
−Ax−B(
c x−1
)
logc
}
¿exp
{
−Ax−m(
cx−1)
}
, wherem= B logcSince S(x)is strictly decreasing , S '(x)<0,
i .e . ,exp
{
−Ax−B(
c x−1)
logc
(
−A−B c x)
}
<0⟺A>−B cxwhich holdif A>−B since c>1Note:When A=0, then Makeham=Gompertz
Weibull
μx=k x n
Then S(x)=exp
{
−∫
0 xk yndy
}
=exp{
−k x n+1n+1
}
, k>0, n>0.Example 1.4.3 You are given:
i. Mortality follows de Moivre’s Law ii. Var
[
T(15)]
=675Calculate e˙25.
Answer:
Under de Moivr e's Law μ
x=(ω−x)
−1, S
(x)=ω−x ω
Then pt x=S(x+t) S(x) =
ω−x−t
ω−x , μx+t= 1
i .e . p . d . f . of T(x) U(0, ω−x)
E
[
T(x)]
= ˙ex=∫
o ω−xt pt xμx+tdt
¿ 1
ω−x t2
2
|
ω−x 0
¿ω−x
2
Var
[
T(x)]
=E[
(
T(x))
2]
–(
e˙x)
2=
∫
o ω−xt2tpxμx+tdt−
(
ω−x2
)
2
¿ 1
ω−x t3
3
|
ω−x0 −(ω−x)2
4
¿(ω−x)
2
3 –
(ω−x)2
4
¿(ω−x)
2
12
Given Var
[
T(15)]
=675⟹(ω−15) 2
12 =675⟺ω−15=
√
675(12)⟺ω=105
Wante˙25= ω−25
2 =
105−25
2 =40
1.4.5 Select and ultimate Tables
A select mortality table recognizes the fact that a newly insured age x shows better (i.e. lower) mortality than an individual age x insured t years ago. This is because of the selection or underwriting process a newly insured has to undergo.
i.e., probability of dying in the coming year is a function of one’s attained age and the duration since the issue of the policy.
q[x]+t=P{life age x+t∧newly insured at age x dies∈the coming year}
Then
q[x]≤ q[x−1]+1≤ q[x−2]+2≤…
i.e. for a given attained age x, the closer the duration since issue, the better the mortality.
The effect of selection usually wears out after a certain number of years, typically 15 for mortality. If the select period is for r durations, then
q[x]+r=qx+r=q[x−1]+r+1=q[x−2]+r+2=…
qx+r=ultimate motality at attained age x+r
¿mortality rate at attained age x+r amongst insured ’ s with policiesissued
at least r years ago.
For a select period of 3 years, a select and ultimate mortality table will look as follows: Issue
Age 1 DurationSelect
2 3
Ultimate 4+
x q[x] q[x]+1 q[x]+2 qx+3 x+1 q[x+1] q[x+1]+1 q[x+1]+2 qx+4 X+2 q[x+2] q[x+2]+1 q[x+2]+2 qx+5
Aggregate mortality at age x is the mortality rate at attained age x without regard to duration since issue, i.e., select and ultimate lives of the same attained age are grouped together.
Note
l[x]+k=number of survivorsat attained age(x+k)amongst lxlives insured at age x .
For a select period of r years,
l[x]+r=lx+r
Answer:
[30] 32 34 40
For a five year period, the lx function is of the form
l[30], l[30]+1,l[30]+2,l[30]+3,l[30]+4, l35, l36, …
Then q2|6 [30]+2=l[30]+2+2−l[30]+2+8 l[30]+2
¿l[30]+4−l40
l[30]+2
Example 1.4.5 Express by a single symbol the probability that a life now age 50, insured three years ago, will die in his 59th year i.e., between ages 58 and 59 assuming a
five year select period.
Answer:
[47] 50 58 59
Probability=8|q[47]+3
¿d[47]+3+8
l[47]+3
¿ d58