Life Insurance Mathematics

Survival models1

Jan Dhaene

1 Based on Chapter 2 in ’Actuarial Mathematics for Life Contingent Risks’

by David C.M. Dickson, Mary R. Hardy and Howard R. Waters, Cambridge
University Press, 2020 (third edition).
1 / 35
2.1 Summary

The future lifetime r.v.2

Probabilities of death and survival.
The force-of-mortality.
Basic actuarial notations.
The curtate future lifetime r.v.

2 Nothing is certain in life except death and taxes - Benjamin Frankin.

2 / 35
2.2 The future lifetime random variable
Status (x ):
not.
(x ) = a life aged x, x 0
Future lifetime of (x ):
Tx
Assumption: Tx is a continuous r.v. on (0, +∞).
Age-at-death of (x ):
x + Tx
Lifetime distribution of (x ):

Fx ( t ) = P [ T x t]
Survival function of (x ):

Sx ( t ) = 1 Fx ( t )

3 / 35
2.2 The future lifetime random variable

"The …rst person to live for a thousand years is possibly

already alive and of those of us aged between 20-30, there will
certainly be some who reach 130 years old. This will have an
instantaneous and catastrophic e¤ect on the world
population". Richard Seymour.

About a century ago, the British monarch started sending

anniversary messages to "current citizens of [the monarch’s]
realms or UK Overseas Territories" who reached the age of
100. In 1917, King George V sent a total of 24 celebratory
messages to centenarians. By 1952 this had increased more
than 10-fold to 255, and in 2016, it has exploded to nearly
60-fold to 14500.

4 / 35
2.2 The future lifetime random variable
Consider a person (x ) with
Current future lifetime: Tx .
Future lifetime at birth: T0 .
Future lifetime at age y x, given survival until age y : Ty .
Assumption:
For any y x and t 0, we assume that

P [ Ty t ] = P [ T0 y + t j T0 > y ] (2.1)

Interpretation: Starting from the cdf of T0 , the only

additonal information used to determine survival probabilties
at age x and beyond is survival or not.
Corollary: For any t, u 0, we have that

P [ Tx + t u ] = P [ Tx t + u j Tx > t ]

5 / 35
2.2 The future lifetime random variable

Lifetime distributions Fx and F0 :

F0 ( x + t ) F0 ( x )
Fx ( t ) = (2.2)
S0 ( x )

Survival functions Sx and S0 :

S0 ( x + t ) = S0 ( x ) Sx ( t ) (2.4)

Survival functions Sx +t and Sx :

Sx ( t + u ) = Sx ( t ) Sx + t ( u ) (2.5)

6 / 35
2.2 The future lifetime random variable

Consider (x ) with continuous future lifetime Tx .

Sx (t ) is a survival function for (x ) if and only if the
following conditions are satis…ed:
Condition 1:
Sx ( 0 ) = 1
Condition 2:
lim Sx (t ) = 0
t !+∞
Condition 3:

Sx (t ) is a non-increasing continuous function of t

7 / 35
2.2 The future lifetime random variable

For all survival functions Sx (t ) in this course, we make the

following assumptions:
Assumption 1:

d
Sx (t ) exists for all t > 0
dt
Assumption 2:
lim t Sx (t ) = 0
t !+∞
Assumption 3:
lim t 2 Sx (t ) = 0
t !+∞

Assumptions 2 and 3 ensure that the mean and the variance

of the distribution of Tx exist.

8 / 35
2.2 The future lifetime random variable

Example 2.1
Assume that
1
t 6
F0 ( t ) = 1 1 for 0 t 120
120

Calculate the probability that

a newborn survives beyond age 30,
a life aged 30 dies before age 50,
a life aged 40 survives beyond age 65.

9 / 35
2.3 The force of mortality

Consider a person with survival function at birth P[T0 > t ].

The force-of-mortality at age x:

def. P[T x h ]
µx = limh !0+ h

Other expression for µx :

P [ T0 x + h j T0 > x ]
µx = lim+ (2.6)
h !0 h
Intuitive interpretation:

µx dx P [ T0 x + dx j T0 > x ] (2.8)

10 / 35
2.3 The force of mortality

µx in terms of S0 :

1 d
µx = S (x )
S 0 (x ) dx 0
(2.9)

The pdf of Tx :

d d
fx ( t ) = Fx ( t ) = Sx ( t )
dt dt
µx in terms of f0 and S0 :

f 0 (x )
µx = S 0 (x )

11 / 35
2.3 The force of mortality

Suppose that x is …xed and t is variable.

Expression for µx +t :

f x (t )
µ x +t = S x (t )
(2.10)

Intuitive interpretation:

µx +t dt P [ Tx t + dt j Tx > t ]

An expression for Sx (t ):
Z t
Sx (t ) = exp µx +s ds (2.11)
0

12 / 35
2.3 The force of mortality

Example 2.2
Suppose that
1
x 6
F0 ( x ) = P [ T0 x] = 1 1 for 0 t 120
120
Derive an expression for µx .

13 / 35
2.3 The force of mortality
2.3.1 Mortality laws

Gompertz’law of mortality:

µx = Bc x , x >0

where B and c are constants such that B > 0 and c > 1.

Makeham’s law of mortality:

µx = A + Bc x , x >0

where A, B and c are constants such that A, B > 0 and

c > 1.
Both models often provide a good …t to mortality data over
certain age ranges, particularly from middle age to early old
age.

14 / 35
2.3 The force of mortality
2.3.1 Mortality laws
Example 2.3
Gompertz’law of mortality.
Derive an expression for Sx (t ).
Solution:
n o
B x
Sx (t ) = exp ln c c (c t 1)

Makeham’s law of mortality.

Derive an expression for Sx (t ).
Solution:
n o
B x
Sx (t ) = exp At ln c c (c t 1) (2.12)

Remark: This expression is often written as

x (c t
Sx ( t ) = s t g c 1)

with s = e A and g = exp ( B/ ln c ).

15 / 35
2.3 The force of mortality
Benjamin Gompertz (1779 - 1865)
Chief actuary of Alliance Assurance Company, England.
Before Gompertz, demographers collected and compiled
mortality data, but never much considered a formal ’law of
mortality’.

16 / 35
2.3 The force of mortality
Benjamin Gompertz (1779 - 1865)
He published his law of mortality in: Gompertz, B. (1825),
Philosophical Transactions of the Royal Society of London 115:
513–585.

17 / 35
2.3 The force of mortality
Benjamin Gompertz (1779 - 1865)

Gompertz’law performs a fairly good …t from middle age (35)

to old age (90).
For very old ages , it overestimates death rates.
It does not capture infant mortality and the accident hump.
He was a Jewish religious man:
"... Neither profane history nor modern experience could
contradict the possibility of the great age of the patriarchs of
the scripture..."

18 / 35
2.3 The force of mortality
William Makeham (1826 - 1891)

Gompertz (1825):
"It is possible that death may be the consequence of two
generally co-existing causes. The one is chance, without
previous disposition to death or deterioration. The other is a
deterioration, or an increased inability to withstand
destruction..."
Makeham formalized and mathematized Gompertz’law of
mortality.
He published his law of mortality in Makeham, W.M. (1860).
"On the Law of Mortality and the Construction of Annuity
Tables". J. Inst. Actuaries and Assur. Mag. 8: 301–310.

19 / 35
2.3 The force of mortality
William Makeham (1826 - 1891)

20 / 35
2.3 The force of mortality

Example 2.4
Let
µx = Bc x , x >0
B = 0.0003 and c = 1.07
Calculate Sx (t ) and fx (t ) for x = 20, x = 50 and x = 80.

21 / 35
2.3 The force of mortality
Example 2.4

22 / 35
2.3 The force of mortality
Example 2.4

23 / 35
2.4 Actuarial notation

The International Actuarial Association (IAA) was founded

in Brussels in 1895 , at the occasion of the 1st International
Congress of Actuaries (ICA).
At the 2nd ICA, which was held in London in 1898, standard
actuarial notation was unanimously adopted.
Actuarial notation is a shorthand method to allow actuaries to
record mathematical formulas that deal with interest and
mortality rates.

24 / 35
2.4 Actuarial notation
Survival rates:
not.
t px = P [ Tx > t ] = Sx ( t ) (2.13)
Mortality rates:
not.
t qx = P [ Tx t ] = Fx ( t ) (2.14)
Deferred mortality rates:
not.
u jt qx = P [ u < Tx u + t ] = Sx ( u ) Sx ( u + t ) (2.15)
Simpli…ed notations for 1 - year probabilities:
not.
px = 1 px
not.
qx = 1 qx
not.
u j qx = u j1 qx

25 / 35
2.4 Actuarial notation
Relations between survival and death rates

Survival and mortality rate add to 1:

t px + t qx = 1

Survival rates at di¤erent ages:

t +u px = t px u px +t (2.16)

Survival rates in terms of one-year survival rates:

n px = px px +1 ... px +n 1

Deferred mortality rates:

u jt qx = u px u +t px = u px t qx +u

26 / 35
2.4 Actuarial notation
Relations between forces-of-mortality and survival rates

Force-of-mortality at age x:
h qx 1 d
µx = lim+ = x p0 (2.17)
h !0 h x p0 dx
Force-of-mortality at age x + t:
1 d
µ x +t = t px (2.18)
t x dt
p
Density function of Tx :
fx (t ) = t px µx +t (2.19)
Survival rate in terms of forces-of-mortality:
Z t
t px = exp µx +s ds (2.20)
0

27 / 35
2.4 Actuarial notation
Death rates and forces-of-mortality:
Z t
t qx = s px µx +s ds (2.21)
0
Graphical interpretation:

Approximation:
qx µx + 1 , when qx is small
2

28 / 35
2.5 Mean and standard deviation of remaining lifetime
Complete expectation of life:

def.
e̊x = E[Tx ]

Evaluating e̊x :
R∞
e̊x = p dt
0 t x
(2.23)

Second moment of Tx :
Z ∞
E[Tx2 ] = 2 t t px dt (2.24)
0

Variance of Tx :

V [Tx ] := E[Tx2 ] (e̊x )2

29 / 35
2.5 Mean and standard deviation of remaining lifetime

Example 2.6
Let
x 61
F0 ( x ) = 1 1 for 0 x 120
120
Calculate e̊x and V [Tx ] for x = 30 and for x = 80.

30 / 35
2.5 Mean and standard deviation of remaining lifetime

31 / 35
2.5 Mean and standard deviation of remaining lifetime

Term expectation of life:

def.
e̊x :n j = E[min (Tx , n)]

Evaluating e̊x :n j :
Z n
e̊x :n j = t px dt
0

32 / 35
2.6 Curtate future lifetime
2.6.1 Curtate future lifetime and curtate expectation of life
Curtate future lifetime:
def.
Kx = b T x c
Probability function of Kx :
P [Kx = k ] = k px qx +k , k = 0, 1, 2, . . .
Curtate expectation of life:
not.
ex = E [ Kx ]
Evaluating ex :
∞
ex = ∑ k px (2.25)
k =1
Second moment of Kx :
∞
E[Kx2 ] = 2 ∑ k k px ex
k =1
33 / 35
2.6 Curtate future lifetime
2.6.2 The complete and curtate expected future lifetimes
Numerical values:

Approximation for the complete future lifetime:

1
e̊x ex + (2.26)
2
34 / 35
2.7 Notes and further reading

Other names for the force-of-mortality:

Survival analysis: Hazard rate.
Reliability theory: Failure rate.

35 / 35