Chapter 1: Survival Distributions

Chapter 1 Survival Distributions

1. To define future lifetime random variables

2. To specify survival functions for future lifetime random variables

3. To define actuarial symbols for death and survival probabilities and

develop relationships between them

4. To define the force of mortality

ln Exam FM, you valued cash flows that are paid at some known future times. In Exam MLC,
by contrast, you are going to value cash flows that are paid at some unknown future times.
Specifically, the timings of the cash flows are dependent on the future lifetime of the underlying
individual. These cash flows are called life contingent cash flows, and the study of these cash
flows is called life contingencies.

It is obvious that an important part of life contingencies is the modeling of future lifetimes. In
this chapter, we are going to study how we can model future lifetimes as random variables. A
few simple probability concepts you learnt in Exam P will be used .

• 1. 1 Age-at-death Random Variable

Let us begin with the age-at-death random variable, which is denoted by T0 . The definition of To
can be easily seen from the diagram below.

'(' Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

 Chapter 1: Survival Distributions

Death occurs

Age
0 To

The age-at-death random variable can take any value within [O, oo). Sometimes, we assume that
no individual can live beyond a certain very high age. We call that age the limiting age, and

denote it by w. !fa limiting age is assumed, then To can only take a value within [O, m].

We regard To as a continuous random variable, because it can, in principle, take any value on
the interval [O, oo) if there is no limiting age or [O, w] if a limiting age is assumed. Of course, to
model T0 , we need a probability distribution. The following notation is used throughout this
study guide (and in the examination).
- F0(1) = Pr(To t) is the (cumulative) distribution function of T0 •

- fo(t) = is the probability density function of To. For a small interval !J.t, the product
dt
fo(t)M is the (approximate) probability that the age at death is in between t and t + !J.t.

In life contingencies, we often need to calculate the probability that an individual will survive to
a certain age. This motivates us to define the survival function:

S0 (t) = Pr( To > t) = l - Fo(t). f--

Note that the subscript "O" indicates that these functions are specified for the age-at-death
random variable (or equivalently, the future lifetime ofa person age 0 now).

Not all functions can be regarded as survival functions. A survival function must satisfy the
following requirements:
I. S0 (0) = l. This means every individual can live at least 0 years.
2. S0 (w) = 0 or iimS0 (t) = 0. This means that every individual must die eventually.

3. S0 (1) is monotonically decreasing. This means that, for example, the probability of surviving
to age 80 cannot be greater than that of surviving to age 70.

<(:' Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

Chapter 1: Survival Distributions

Summing up,fo(t), F 0(t) and S0 (t) are related to one another as follows.

·% . . .•. %'
:t; ·f 0 R M U L A
%
Relations between/0 (1), Fo(I) and S 0(1)
d d
/ 0 (1) =-F0 (1) = --S0 (1), (I.I)
dt dt
;% f,w f
S 0 (1)=, 0 (u)du=l-
f'
f 0 (u)du=l-F0 (1), (1.2) %
10

•/'.
;% Pr(a<To5b)= ff b
0 (u)du=F0 (b)-F0 (a)=S 0 (a)-S 0 (b). (1.3)
('.
%;
a

Note that because To is a continuous random variable, Pr(T0 = c) = 0 for any constant c. Now, let
us consider the following example.

Example 1.1

You are given that S0 (t) = 1- t/100 for 0 5 t 5 l 00.

(a) Verify that S0 (t) is a valid survival function.
(b) Find expressions for Fo(t) andfo(t).
(c) Calculate the probability that To is greater than 30 and smaller than 60.

- Solution
(a) First, we have S0(0) = 1 - 01100 = I.
Second, we have So(l 00) =I - 1001100 = 0.
Third, the first derivative of S 0(t) is -1/100, indicating that S0 (t) is non-increasing.
Hence, S0(t) is a valid survival function.

(b) We have F0(t) = 1 -S0 (t) = t/100, for 0 5 t 5 100.

d
Also, we have andfo(t) = - F0 (t) = 1/100, for 0 5 t 5 100.
dt
(c) Pr(30 < To< 60) = S0 (30) - So(60) = (1 - 30/100) - (1 - 601100) = 0.3.
[ END]

'.c• Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

 Chapter 1: Survival Distributions

- 1. 2 Future Lifetime Random Variable

Consider an individual who is age x now. Throughout this text, we use (x) to represent such an
individual. Instead of the entire lifetime of (x), we are often more interested in the future
lifetime of (x). We use Tx to denote the future lifetime random variable for (x). The definition of
Tx can be easily seen from the diagram below.

Age
0 x x+ T,

Now
t t
Death
occurs
i ...
Time from now
0 Tx

[Note: For brevity, we may only display the portion starting from age x (i.e., time 0) in future
illustrations.]

If there is no limiting age, Tx can take any value within [O, oo ). If a limiting age is assumed, then
Tx can only 'take a value within [O, OJ- x]. We have to subtract x because the individual has
attained age x at time 0 already.

We let S,(t) be the survival function for the future lifetime random variable. The subscript "x"
here indicates that the survival function is defined for a life who is age x now. It is important to
understand that when modeling the future lifetime of (x), we always know that the individual is
alive at age x. Thus, we may evaluate Sx(t) as a conditional probability:

S,(t)=Pr(T, >l)=Pr(T, >x+tlT, >x)

Pr(T0 > x+I nT0 > x) Pr(T, >x+I) S0 (x+t)
=
Pr(T, > x) Pr(T, > x) S 0 (x)

The third step above follows from the equation Pr(AIB) = Pr(AnB) , w h'1c h you learnt in
Pr(B)
Exam P.

\:' Actex 2013 I Johnny Li and Andrew Ng J SoA Exam MLC

Chapter 1: Survival Distributions

:r/////////////,'/'//////////////,;/'//////////////,'/'//////////////,';."/////////////1:

Survival Function for the Future Lifetime Random Variable

:%
:% S(
S (I)= o x+t
)
(1.4)
%:%:
:% ' S0 (x) t.;:

With Sx(t), we can obtain Fx(I) andf,(t) by using

d
F,(t) =I - S,(t) and fx(t) = dt F, (t),

respectively.

Example 1.2

You are given that S0 (t) =I -t/100 for 0 $ t $JOO.

(a) Find expressions for S1o(t), F1o(t) and.fio(t).
(b) Calculate the probability that an individual age 10 now can survive to age 25.
(c) Calculate the probability that an individual age 10 now will die within 15 years.

- Solution

(a) In this part, we are asked to calculate functions for an individual age 10 now (i.e., x = 10).
Here, w = l 00 and therefore these functions are defined for 0 $ I $ 90 only.

. h S () S (JO+t) = --'------'-----
l-(JO+t)/100 t
First we ave 10 t = 0 1--, for 0 st s 90.
' s, (l 0) 1-101100 90

Second, we have F 10 (1) = I - S 10 (t) = t/90, for 0 $ t $ 90.

Finally, we have ft 0 (t) = F, 0 (1) = _l .

dt 90
(b) The probability that an individual age 10 now can survive to age 25 is given by
15 5
Pr(T10 > 15) = S10(15) = I - - = - .
90 6
(c) The probability that an individual age 10 now will die within 15 years is given by
1
Pr(T10 S 15) = F10(15) = I - S10(15) = - .
6
[ END]

'.(: Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

 Chapter 1: Survival Distributions

• 1. 3 Actuarial Notation

For convenience, we have designated actuarial notation for various types of death and survival
probabilities.

Notation 1: 1Px
We use iPx to denote the probability that a life age x now survives to t years from now. By
definition, we have
1Px = Pr(Tx > t) = Sx(t).
When I = 1, we can omit the subscript on the left-hand-side; that is, we write 1Px as Px·

Notation 2: 1 qx
We use 1qx to denote the probability that a life age x now dies before attaining age x + t. By
definition, we have
,qx = Pr(Tx $I) = Fx(t).
When I= 1, we can omit the subscript on the left-hand-side; that is, we write 1qx as qx.

Notation 3: 11uqx
We use 11 uq, to denote the probability that a life age x now dies between ages x + t and x + t + u.
By definition, we have 1-

11uqx = Pr(t < Tx '.'> t + u) = Fx(t + u) - Fx(/) = Sx(I) - Sx(t + u).

When u = 1, we can omit the subscript u; that is, we write / p qx as 11qx.

Note that when we describe survival distributions, "p" always means a survival probability,
while "q" always means a death probability. The "I" between I and u means that the death
probability is deferred by t years. We read "t I u" as "I deferred u". It is important to remember
the meanings of these three actuarial symbols. Let us study the following example.

D'· Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

Chapter 1: Survival Distributions

Example 1.3

Express the probabilities associated with the following events in actuarial notation.
(a) A new born infant dies no later than age 45.
(b) A person age 20 now survives to age 38.
(c) A person age 57 now survives to age 60 but dies before attaining age 65.
Assuming that S0(t) = e-00125 ' fort<: 0, evaluate the probabilities.

- Solution
(a) The probability that a new born infant dies no later than age 45 can be expressed as 4sqo.
[Here we have "q" for a death probability, x = 0 and I= 45.]

Further, 4sqo = F 0(45) =I -So(45) = 0.4302.

(b) The probability that a person age 20 now survives to age 38 can be expressed as 18p 2o. [Here
we have "p" for a survival probability, x = 20 and I= 38 - 20 = 18.]

Further, we have 1sp20 =S20 (18) =

s0 (38)
0.7985.
S 0 (20)

(c) The probability that a person age 57 now survives to age 60 but dies before attaining age 65
can be expressed as 315 q 57 . [Here, we have "q" for a (deferred) death probability, x = 60, t =
60 - 57 = 3, and u = 65 - 60 = 5.]

Further, we have 315q57 = S57(3)-Ss1(8) =

s0 (60) -
S0 (65)
= 0.058357.
s0 (57) s (57)
0

[ END]

Other than their meanings, you also need to know how these symbols are related to one another.
Here are four equations that you will find very useful.

Equation 1: rPx + ,q, = I

This equation arises from the fact that there are only two possible outcomes: dying within t
years or surviving to I years from now.

'<:' Actex 2013 I Johnny Li and Andrew Ng SoA Exam MLC

 Chapter 1: Survival Distributions

Equation 2: 1+uPx = 1Px X uPx+1

The meaning of this equation can be seen from the following diagram.

Survive from time t to t + u:

Survive from time 0 tot: probability= uPx+ 1
probability= 1Px

0 I t+ u
(Age x) (Age x + t)

Survive from time 0 to I+ u:\

probability= 1+uPx

Mathematically, we can prove this equation as follows:

=S(t+u)=S0 (x+t+u)=S0 (x+t)S0 (x+t+u)=S(t)S (u)= x .

•+uPx x S (x) S (x) S (x+t) ' ,., ,p, ,,p,.,
0 0 0

This equation arises naturally from the definition of 11 ,,q,.

We have 11uqx= Pr(/< Tx $I+ u) = Fx(I + u) - F,(t) = 11uqx = 1+uqx- ,q,.
Also, 11uqx = Pr(I < Tx $I+ u) = Sx(t) - S,(t + u) = 1Px - i+uPx·
I
(
Equation 4: 11uqx = 1Px
-
X ,,qx+1
The reasoning behind this equation can be understood from the following diagram:

Survive from time 0 to time I:

probability= 1Px

Time from now

0 t t+u
(Age x) (Age x +I)

l:' Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

Chapter 1: Survival Distributions

Mathematically, we can prove this equation as follows:

(from Equation 3)

= rPx - rPx x .,px+r (from Equation 2)

= rPx (I - .,px+r)

= 1Px X 11qx+t (from Equation I)

Here is a summary of the equations that we just introduced.

:% Relations between tPx. 1qx and ,,.qx ;.;;.

" i
X
·I'. 1Px + ,q,= I , (1.5) lj:
%:
t+uPx = 1Px X uPx+1, (J .6)
:;:; 1j't,qx = t+uqx - I q, = 1Px - /+uPx = tPx X uqx+I• (1.7) %

Let us go through the following example to see how these equations are applied.

Exampl_e 1.4

You are given:

(i) Px = 0.99
(ii) Px+I = 0.985
(iii) JPx+I = 0.95
{iv) qx+J = 0.02
Calculate the following:

(a) Px+l
{b) 2[lx

(c) 2Px+I
(d) iPx

(e) 1;2q,

t Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

 Chapter 1: Survival Distributions

- Solution

(a) Px+l = 1- qx+J

= I - 0.02 = 0.98

(b) 2Px = Px X Px+l

= 0.99 x 0.985 = 0.97515

(c) Consider JP.r+I = 2Px+1X Px+J

=> 0.95 = 2Px+t x 0.98

=> 2Px+1 = 0.9694

( d) lPx = Px X 2Px+l
= 0.99 x 0.9694 = 0.9597

(e) q2qx = Px X 2q.r+1

= Px (1 - 2Px+1)

= 0.99 (1 - 0.9694) = 0.0303

[END]

- 1. 4 Curtate Future Lifetime Random Variable

In practice, actuaries use Excel extensively, so a discrete version of the future lifetime random
variable would be easier to work with. We define

Kx =LT, J,
where LY J means the integral part of y. For example, LI J= 1, L4.3 J = 4 and Li 0.99 J = 10. We

call Kx the curtate future lifetime random variable.

It is obvious that Kx is a discrete random variable, since it can only take non-negative integral
values (i.e., 0, !, 2, ... ).The probability mass function for Kx can be derived as follows:

Pr(Kx = 0) = Pr(O s; Tx < 1) = q,,

Pr(K, =I)= Pr(l s; Tx < 2) = 1pq,,
Pr(Kx = 2) = Pr(2 s; Tx < 3) = 2pq,, ...

<(;' Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

Chapter 1: Survival Distributions

Inductively, we have

'////////////////.1?'/////////////////////////////%1?'//////////////.'7/////////////,-;:
.

Probability Mass Function for Kx

;9 --
Pr(Kx-k)-kllqx, -
k-0, 1,2, ... (1.8) /.,

The cumulative distribution function can be derived as follows:

Pr(K, k) = Pr(Tx < k + I)= k+iq,, fork= 0, 1, 2, ....

It is just that simple! Now, let us study the following example, which is taken from a previous
SoA Exam.

Example 1.5 [Course 3 Fall 2003 #28]

For (x):
(i) K is the curtate future lifetime random variable.
(ii) qx+k = 0.1 (k + 1), k=O, 1, 2, ... , 9
Calculate Var(K A 3).
(A) I. I (B) 1.2 (C) 1.3 (D) 1.4 (E) 1.5

- Solution

The notation A means "minimum". So here KA 3 means rnin(K, 3). For convenience, we let
W = min(K, 3). Our job is to calculate Var(W). Note that the only possible values that W can
take are 0, 1, 2, and 3.

To accomplish our goal, we need the probability function of W, which is related to that of K.
The probability function of Wis derived as follows:
Pr(W= 0) = Pr(K = 0) = q, = 0.1
Pr(W= l)=Pr(K= I)= 11qx

=pxxqx+I

{J Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

 Chapter 1: Survival Distributions

= (I - qx)qx+I
= (1-0.1) x 0.2 = 0.18
Pr(W = 2) = Pr(K = 2) = 21qx

= 2Px X qx+2 = Px X Px+I X qx+2

= (1- qx)( 1 - qx+I) qx+2

= 0.9 x 0.8 x 0.3 = 0.216
Pr(W = 3) = 3) = 1 - Pr(K = 0)- Pr(K = 1)- Pr(K = 2) = 0.504.

From the probability function for W, we obtain E(W) and E(W2) as follows:
E(W)=Ox 0.1+1x0.18+2 x 0.216+3 x 0.504=2.124
E(W2 ) = 02 x 0.1 + 12 x 0.18 + 2 2 x 0.216 + 32 x 0.504 = 5.58
This gives Var(W) = E(W2 ) - [E(W)] 2 = 5.58 - 2.1242 = 1.07. Hence, the answer is (A).
[ END]

1. 5 Force of Mortality

In Exam FM, you learnt a concept called the force of interest, which measures the amount of
interest credited in a very small time interval. By using this concept, you valued, for example,
(
annuities that make payouts continuously. In this exam, you will encounter continuous life
r
contingent cash flows. To value such cash flows, you need a function that measures the
probability of death over a very small time interval. This function is called the force of mortality.

Consider an individual age x now. The force of mortality for this individual I years from now is
denoted by Px+i or Px(I). At time I, the (approximate) probability that this individual dies within a
very small period of time M is Px+1 M. The definition of Px+i can be seen from the following
diagram.

·t; Actex 2013 J Johnny Li and Andrew Ng I SoA Exam MLC

Chapter 1: Survival Distributions

Survive from time 0 to time t: Death occurs

Prob. = Sx(t) during t to t + f'..t:
Prob."" µx+1f'..t

Time from now

0 t t+ 1'.t

Death between time t and t + 1'.t:

Prob. (measured at time 0) "'fx(t)f'..t

From the diagram, we can also tell thatfx(t)f'..t = Sx(t)µx+11'.t. It follows that

fx(t) = S,(t)µx+f = 1Px µ,+ f ·

This is an extremely important relation, which will be used throughout this study manual.
Recall that fx(I) = F;(t) = -s;(I). This yields the following equation:

(1)
µ,., = - Sx(t)'

which allows us to find the force of mortality when the survival function is known.
dlnx I . dlng(x) g'(x) .
Recall that - - = - , and that by cham rule, = - - for a real-valued function g.
x gW
We can rewrite the previous equation as follows:
s;ct)
µx+f = - Sx(t)_
d[ln Sx (t)]
µr+f = - dt
- µx.,dt = d[ln S, (t)].

Replacing t by u,
- µx+,,du = d[ln Sx (u)]

- J:µx+•du = J: d[lnSx(u)]
- f:µx+,,du = lnSX(t)-lnS,(O)

S,(t) =exp(- f:µ,+,,du).

This allows us to find the survival function when the force of mortality is known.

t:• Actex 2013 I Johnny L1 and Andrew Ng I SoA Exam MLC

 Chapter 1: Survival Distributions

;:f////:////////////////:////////////////:////////////////:////////////////://///////i
:/, . . .

?, FORMULA?,
Relations between Jlx+1,fx(t) and Sx(t)
fx(t) = S,(l)/lx+1 = ,p,µ,+ ,, (1.9)

s;co
µx.,=-S,(t)'
c1.to)

Sx(l)=exp(- (I.II)

Not all functions can be used for the force of mortality. We require the force of mortality to
satisfy the following two criteria:

(i) flx+t?. 0 for all x?. 0 and t?. 0.

( ii) Joroo fix+" dU = 00,

Criterion (i) follows from the fact that µx+1 M is a measure of probability, while Criterion (ii)
follows from the fact that lim S, (I)= 0.
HOO

Note that the subscript x + I indicates the age at which death occurs. So you may use µx to
denote the force of mortality at age x. For example, µ20 refers to the force of mortality at age 20.
The two criteria above can then be written alternatively as follows: (

(i) µx?. 0 for all x?. 0. (-

(ii) r µxdx = 00.

The following two specifications of the force of mortality are often used in practice.

Gompertz' law
/ix= Bex
Makeham's law
µx=A +Bex
In the above, A, Band e are constants such that A ?. -B, B > 0 and e > I.

·n Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

Chapter 1: Survival Distributions

Let us study a few examples now.

Example 1.6 '

For a life age x now, you are given:

S(l)=(I0-1)' o,,;1< 10.

x 100

(a) Find /lx+t ·

(b) Find.fx(t).

- Solution

2(10-1)
(1) 100 2
(a) µ,., = - Sx(f) = = 10-1
100
(b) You may work directly from S.,(t), but since we have found µx+, already, it would be quicker
to find.fx(t) as follows:

I =SI = (10-t)' x-2-=10-1

.fx() ,()µ,.+, 1oo 1o-t so .
[ END]

Example 1.7 '

For a life age x now, you are given

µx+I = 0.0021, 12 0.
(a) Is µx+, a valid function for the force of mortality of (x)?
(b) Find S,(t).
(c) Find.fx(t).

- Solution

(a) First, it is obvious that JJx+1 2 0 for all x and t.

Second, r µ,..du.= r 0.002udu = 0.00 lu'I: = r$J •

((' Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

 Chapter 1: Survival Distributions

Hence, it is a valid function for the force of mortality of (x).

I
(b) Sx(t) =exp(- J:µx+,,du) =exp(- J; 0.002udu) = exp(-0.00lt 2
).

(c) fx(t) = Sx(t)µx+t = 0.002texp(-O.OO 11 2).

[ END]

Example 1.8 [Course 3 Fall 2002 #35]

You are given:

(i) R=]-exp(- J:µmdt)

(ii) S =I - exp(- J: (µ,., + k)dt)

(iii) k is a constant such that S = 0. 75R.
Determine an expression for k.
(A) ln((l-qx)f(I- 0.75qx))

(B) ln((l -0.75qx)f(l -px))

(C) ln((l-0.75px)f(I - Px))
(D) ln((l-px)f(I- 0.75qx))

(E) ln((l-0.75qx)f(l -qx))

Solution
I
First, R = I - Sx( I) = I - p, = q,.
Second,

S = I - exp(- J: (µw + k)du) = I - e-' exp(- J: µ",du) = I - e-• SJ!) = I - e-• p, .

Since S = 0.75R, we have

'.l:' Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

Chapter 1: Survival Distributions

1-e-I Px = 0.75q,
-k 1-0.75qx
e
p,
ek Px
1-0.75q,

k =In ( p, ) =In( l-q, )

1-0.75q, I-0.75q,
Hence, the answer is (A).

[ END)

f, Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC