Chapter 2: Life Tables

Chapter 2 Life Tables

1. To apply life tables

2. To understand two assumptions for fractional ages: uniform

distribution of death and constant force of mortality

3. To calculate moments for future lifetime random variables

4. To understand and model the effect of selection

Actuaries use spreadsheets extensively in practice. It would be very helpful if we could express
survival distributions in a tabular form. Such tables, which are known as life tables, are the
focus of this chapter.

- 2. 1 Life Table Functions

Below is an excerpt of a (hypothetical) life table. In what follows, we are going to define the
functions I, and d,, and explain how they are applied.

x 1,, d,
0 1000 16
1 984 7
2 977 12
3 965 75
4 890 144

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

In this hypothetical life table, the value of lo is 1,000. This starting value is called the radix of
the life table. For x = 1, 2, ... ., the function Ix stands for the expected number of persons who
can survive to age x. Given an assumed value of 10 , we can express any survival function S 0(x) in
a tabular form by using the relation

ix= loSo(x).

In the other way around, given the life table function l,, we can easily obtain values of S 0(x) for
integral values of x using the relation

S0 (x) = l,.
I,
Furthermore, we have

=S (t)=S0 (x+t)
,p, ' S,(x)

which means that we can calculate iPx for all integral values of t and x from the life table
function I_,.

The difference ix - l,+1 is the expected number of deaths over the age interval of [x, x + t). We

denote this by ,d,. It immediately follows that ,d, = !, - lx+1·

We can then calculate ,q, and mlnq_, by the following two relations:

1- IX+! n d x+m l.r:+m - [ x+m+11

l , mlnqx = - / - = [
x x ·'

When t = I, we can omit the subscript I and write 1dx as d,. By definition, we have

,d, = d, + dx+I + ... + dx+l-1·

Graphically,
d, + dx+I + dx+2 + dx+3 + + dx+1-I = ,d,= fx -fx+I

i i i i i age
x x+ 1 x+2 x+3 x+4 x+t-1 x+t

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

Chapter 2: Life Tables

Also, when I= 1, we have the following relations:

fx+I
dx= Ix- Ix+[, Px=- J, ·
-,an d qx=-
1x 1x

Summing up, with the life table functions Ix and dx, we can recover survival probabilities iPx and
death probabilities 1 qx for all integral values oft and x easily.

:% FORMULA%:
'/'. Life Table Functions /'.'

·%
:% lx+I (2 I) %
%;
:%
,Px=-
1
. ·

t dx= I, - Ix+t= dx+ dx+I + ... + dx+t-1 (2.2) %

:%
%
%
,d,
Ix -1,+I
,q, =-,-= I 1_ t.,.,
/
(2.3)
•% x x x

Exam questions are often based on the Illustrative Life Table, which is, of course, provided in
the examination. To obtain a copy of this table, download the most updated Exam MLC syllabus.
On the last page of the syllabus, you will find a link to Exam MLC Tables, which encompass
the Illustrative Life Table. You may also download the table directly from
http://www.soa.org/files/edu/edu-2008-spring-mlc-tables.pdf.

The Illustrative Life Table contains a lot of information. For now, you only need to know and
use the first three columns: x, I,, and IOOOq,. For example, to obtain the value q6o, simply use
the column labeled lOOOq,. You should obtain 1000q60 = 13.76, which means q6o = 0.01376. It
is also possible, but more tedious, to calculate q60 using the column labeled!,; we have q6o = I -

'61 I l6o = 0.01376.

To get values of 1p, and ,q, for t > I, you should always use the column labeled !,. For example,
we have sP6o = '6 5 I !60 = 7533964 I 8188074 = 0.9201 I and sq6o = I - sP6o = I - 0.92011 =

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

 Chapter 2: Life Tables

0.07989. Here, you should not base your calculations on the column labeled 1OOOqx, partly
because that would be a lot more tedious, and partly because that may lead to a huge rounding
error.

Example 2.1

You are given the following excerpt ofa life table:

x !_, dx
20 96178.01 99.0569
21 96078.95 102.0149
22 95976.93 105.2582
23 95871.68 108.8135
24 95762.86 112.7102
25 95650.15 116.9802
Calculate the following:
(a) sP20
(b) q24

(c) 4;1q20

- Solution

(a) 95650.15 =0.994512.

'Pw /20 96178.0 I

112.7102
(b) q24 = 95762.86
0.001177.
24

,d,4 112.7102
(c) 411 q20 = !;; = 96178.01 0.001172.

(END]

ct;; Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

Chapter 2: Life Tables

Example 2.2

You are given:

x
(i) So(x) = 1- - 0 ,,;x,,; JOO
JOO'

(ii) 10 = 100
(a) Find an expression for I, for 0,, x,, 100.
(b) Calculate q2 .
(c) Calculate 3q2.

- Solution

(a) lx=loSo(x)= 100-x.

I
(b) q, = 12 I-1, = 98-97
98
2
98

(c) ,q, =-, -= 98

1 -1, 98-95
2

2
3
98
[ END]

In Exam MLC, you may need to deal with a mixture of two populations. As illustrated in the
following example, the calculation is a lot more tedious when two populations are involved.

Example 2.3

For a certain population of 20 years old, you are given:

(i) 2/3 of the population are nonsmokers, and 1/3 of the population are smokers.
(ii) The future lifetime of a nonsmoker is uniformly distributed over [O, 80).
(iii) The future lifetime of a smoker is uniformly distributed over [O, 50).
Calculate sP•o for a life randomly selected from those surviving to age 40.

'\' Actcx 2013 Johnny Li and Andrew Ng I SoA Exam MLC

...

- Solution
(
The calculation of the required probability involves two steps.
First, we need to know the composition of the population at age 20.
- Suppose that there are 120 persons in the entire population initially. At time 0 (i.e., at age 20),
2 I
there are -120 nonsmokers and -/20 smokers.
3 3
- For nonsmokers, the proportion of individuals who can survive to age 40 is I - 20/80 = 3/4.
For smokers, the proportion of individuals who can survive to age 40 is I - 20/50 = 3/5. At
32 3I
age 40, there are --/20 = 0.5120 nonsmokers and --120 = 0.2120 smokers. Hence, among
43 53
those who can survive to age 40, 517 are nonsmokers and 2/7 are smokers.
Second, we need to calculate the probabilities of surviving from age 40 to age 45 for both
smokers and nonsmokers.
For a nonsmoker at age 40, the remaining lifetime is uniformly distributed over [O, 60). This
means that the probability for a nonsmoker to survive from age 40 to age 45 is I - 5160 =
11/12.
- For a smoker at age 40, the remaining lifetime is uniformly distributed over [O, 30). This
means that the probability for a smoker to survive from age 40 to age 45 is I - 5/30 = 516. (
Finally, for the whole population, we have (

5 11 2 5 25 (
sP<o = 7xJ2+7x6 = 28 · f-

[ END)

- 2. 2 Fractional Age Assumptions

We have demonstrated that given a life table, we can calculate values of 1Px and ,q,, when both t
and x are integers. But what if t and/or x are not integers? In this case, we need to make an
assumption about how the survival function behaves between two integral ages. We call such an
assumption a fractional age assumption.

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

Chapter 2: Life Tables

In Exam MLC, you are required to know two fractional age assumptions:
I. Uniform distribution of death
2. Constant force of mortality
We go through these assumptions one by one.

Assumption 1: Uniform Distribution of Death

The Uniform Distribution of Death (UDO) assumption is extensively used in the Exam MLC
syllabus. The idea behind this assumption is that we use a bridge, denoted by U, to connect the
(continuous) future lifetime random variable Tx and the (discrete) curtate future lifetime random
variable K,. The idea is illustrated diagrammatically as follows:

(
Tx 4
Death occurs

0
(Age x)

It is assumed that U follows a uniform distribution over the interval [O, l ], and that U and Kx are
independent. Then, for 0 :Sr< l and an integral value of x, we have

,qx = Pr(T, r)
= Pr(U < rnKx = 0)
= Pr(U < r) Pr(Kx = 0)
=rqx.

The second last step following from the assumption that U and K, are independent, while the
last step follows from the fact that U follows a uniform distribution over [O, l ].

:% Key Equation for the UDD Assumption i'.:.

;% ,q, = rq,, for 0 r<I (2.4)

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

 Chapter 2: Life Tables

This means that under UDO, we have, for example, o.4q 50 = 0.4q 50 . The value of q 50 can be
(-
obtained straightforwardly from the life table. To calculate ,px, for 0 Sr< 1, we use ,px = 1 - 4x

= 1 - rqx. For example, we have o 1P20 = 1 - O. lq20.

This equation is equivalent to a linear interpolation between Ix and lx+l, that is,

Ix+,= (1 - r)lx + rlx+l·

Proof
,Px =1-rq, =(1-r)+rpx
1. I
..!± !:-=(1-r)+r-'± L
I, I,
Ix+.= (1-r)l_, +rl... ,
D
What if the subscript on the left-hand-side is greater than 1? In this case, we should first use
equation ( 1.6) from Chapter 1 to break down the probability into smaller portions. As an
example, we can calculate 2 5p 30 as follows:

2.sPJo = 2PJ0 x osp32 = 2PJ0 x (1 - 0.5qn).

The value of 2p 30 and qn can be obtained from the life table straightforwardly.

What if the subscript on the right-hand-side is not an integer? In this case, we should make use
of a special tr_ick, which we now demonstrate. Let us consider o 1p 5_7 (both subscripts are not
integers). The trick is that we multiply this probability with o.1p5, that is,

0.1P5 x O. IPS.7 = 0.8P5·

This gives 01
p 57 = osPs = l-O.Sq, . The value of q 5 can be obtained from the life table.
01Ps 1-0.7q,
To further illustrate this trick, let us consider 35 q4 6 • This probability can be evaluated by first
calculating J5P46:

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

Chapter 2: Life Tables

1P, p p p (1-0.lq) ,p,(1-0.lq,)

Th en, we have p4 _6 =• - - ' 'o.1 ' = ' ' ' ' and finally q4.6 = 1
3 _5
o.,P, .,p, - . q,
1 06
3.5
l-0.6q,

The values of 4p4, q 8 and q 4 can be obtained from the life table.

Let us study the following example.

Example 2.3

You are given the following excerpt ofa life table:

x I, d,
60 100000 300
61 99700 400
62 99300 500
63 98800 600
64 98200 700
65 97500 800

Assuming uniform distribution of deaths between integral ages, calculate the following:
(a) 026P61
(b) 22q60
(c) 0Jq62s

- Solution

(a) 026P61 = 1-026q61=1-0.26x400/99700 = 0.998957.

(b) 22q60= 1-22/)60= l-2P60 Xo2/)62= l -2p60x (l -0.2q62)

1 99300
= 1- "(1-0.2xd")=l- =0.008.
160 162 100000 99300

(c) Here, both subscripts are non-integers, so we need to use the trick. First, we compute o.JP62.s:

t' Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

 Chapter 2: Life Tables

Then, we have

98800 (1-0.1 x
= II p., = P620.l P63 = p,, (I - O. lq63) = 99300 98800 0.998382 .
0.3 P628 0 500
o.sP62 o.sP62 I- .8q62 J-0.8x-
99300
Hence, o1q62.& = I - 0.998382=0.001618.
[END]

Sometimes, you may be asked to calculate the density function of T, and the force of mortality
from a life table. Under UDO, we have the following equation for calculating the density
function:

fx(r) = q,, I.

d d d d
Proof f, (r) =-F,(r) = -Pr(T. < r) = - ,q. =-(rq,) = q,. D
' dr dr ·· dr · dr ·

Under UDO, we have the following equation for calculating the force of mortality:
q,
µx+r =-,--,
-rq,

Proof In general, fx(r) = ,p,µx+" Under UDO, we have fx(r) = q, and ,p, = I - rq,. The result
follows. D

Let us take a look at the following example.

Example 2.4 [Course 3 Spring 2000 #12]

For a certain mortality table, you are given:

(i) µs 0 5 = 0.0202
(ii) µsis= 0.0408
(iii) J/S25 = 0.0619
(iv) Deaths are uniformly distributed between integral ages.
Calculate the probability that a person age 80.5 will die within two years.
(A) 0.0782 (B) 0.0785 (C) 0.0790 (D) 0.0796 (E) 0.0800

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

Chapter 2: Life Tables

- Solution
The probability that a person age 80.5 will die within two years is 2q 805 . We have

ospso x 2Pso.s = 2.sPso .

This gives

2 Pso o_sPs1
2Psos
o.sPso
p 80 p 81 (l-0.5q 82 )
=
l-0.5q80
(I - q80 )(1- q,, )(1 -0.Sq.,)
I -O.Sq,0

We then need to find q 80 , q 81 and q82 from the information given in the question. Using µso.s, we

have µ805 = qso =} q80 = 0.0200. Similarly, by using µ 81 5 and µ 825 , we obtain qs1 = 0.0400
l-0.5q 80

and q82 = 0.0600.

Substituting qso, qs 1 and q82 , we obtain 2fJso s = 0.921794, and hence 2qso.s = I - 2Pso.s = 0.0782.
Hence, the answer is (A).

[ END)
Assumption 2: Constant Force of Mortality

The idea behind this assumption is that for every age x, we approximate µx+r for 0 :> r < I by a
constant, which we denote by f1x. This means

which implies Px = e-P, and f1x = -ln(px).

We are now ready to develop equations for calculating various death and survival probabilities.
First of all, for any integer-valued x, we have

rPx = (p,)', O=>r<I

:c-Acrex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

 Chapter 2: Life Tables

J - ,µn,.d11 - J' j.l,du - r I r

D
Proof ,Px=e" =e" =e-"' =(e-'·) =(px)'.

For example, oiPso = (pso) 0·3 , and o4q62 = I - o.4p62 = I - (p62 ) 0 .4. We can generalize the equation
above to obtain the following key formula.

?,

Key Equation for the Constant Force of Mortality Assumption

?,% p - (p X•
)' for 0 $ r < I and r + u $ I (2.5)
?,
rx+u- ?,
'./.::,, , , , ,, ,,,,. ,,.,, , ,, / " / f " / " I ' / / / ,,,,,.,,.,,.//"'/!'_,,.,,,..,,,,..,,.,,!'/I'////,,. ./,,,,.,,.,,,,..,.,,,,,,///!'/'""/,,,,.,,,,,,,,,,.,,
///_////,/////////'////,///////.///////////////////////////////'//////////////////'/////"
,,/,,,,.,,,,,,z

Proof , P.m = e -J;'"'""" = e -f>·" = e-µ,, = (p,)'. [The second step follows from the fact that

given 0 $ r < I, u +tis always less than or equal to I when 0 $ t $ r.] D

Notice that the key equation for the constant force of mortality assumption is based on p, while
that for the UDO assumption is based on q.

This key equation means that, for example, 0 2p 303 = (p30) 0.2. Note that the subscript u on the
right-hand-side does not appear in the result, provided that the condition r + u $ I is satisfied.
But what if r + u > I? The answer is very simple: Split the probability! To illustrate, let us
consider 0 8p 303 . (Here, r + u = 0.8 + 0.3 = I. I > I.) By using equation (1.6) from Chapter I, we
can split o. 8p 30 .3 into two parts as follows:

08P303 = 0.7P303 x O.IP31·

We intentionally consider a duration of0.7 years for the first part, because 0.3 + 0.7 =I, which
means we can apply the key equation rPx+u = (p.r)' to it. As a result, we have

ogp303= (pJo)07 x
(p31 )0.1 .

The values of p 30 and p31 can be obtained from the life table straightforwardly.

{;i Aetex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

Chapter 2: Life Tables

To further illustrate, let us consider s.6P40B· We can split it as follows:

s.6P•o.s = o.2P4o.s x s.•P41 = o.2P•o.s x sP•1 x o.4P46 = (p4o)0.2 x 5p41 x

(p 46 )0.4 .

The values of p40, sp41 and p 46 can be obtained from the life table straightforwardly.

Example 2.5 ,

You are given the following excerpt of a life table:

x Ix dx
60 100000 300
61 99700 400
62 99300 500
63 98800 600
64 98200 700
65 97500 800
Assuming constant force of mortality between integral ages, calculate the following:
(a) 026P61
(b) 22q60

(c) 0Jq62&

- Solution
026 026
(a) 026P61 = (p61) = (99300/99700) = 0.998955.

02
99300 ( 98800 ) · = 0.008002.
100000 99300

(c) First, we consider 0JP62s:

0 JP62 8 = 0.2/)62.8 x 0 IP63 = (p 62 )0.2 (p 63 )0.1 .

Hence,

0.2 ( JO.I
= 1-(p )0.2(p )0.1 = ]- 164
,,q62.8 62 63 (l l
62 J 63

= l -(98800J'·' (98200)'·' = 0.001617.

99300 98800
[ END I

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

 Chapter 2: Life Tables

Let us conclude this section with the following table, which summarizes the formulas for the
two fractional age assumptions.

UDD Constant force

1Px 1 - rqx

,-qx 1 - (p,)'

____IL,_
µx+r -ln(p,)
1-rq,

In the table, x is an integer and 0 r < 1. The shaded formulas are the key formulas that you
must remember for the examination.

-i' 2. 3 Select-and-Ultimate Tables

Insurance companies typically assess risk before they agree to insure you. They cannot stay in
business if they sell life insurance to someone who has just discovered he has only a few months
to live. A team of underwriters will usually review information about you before you are sold
insurance (although there are special insurance types called "guaranteed issue" which cannot be
underwritten). For this reason, a person who has just purchased life insurance has a lower
probability of death than a person the same age in the general population. The probability of
death for a person who has just been issued life insurance is called a select probability. In this
section, we focus on the modeling of select probabilities.

Let us define the following notation.

[x] indicates the age at selection (i.e., the age at which the underwriting was done).
[x] + t indicates a person currently age x + t and was selected at age x (i.e., the underwriting
was done at age x). This implies that the insurance contract has elapsed for I years.

i Actex 2013 I Johnny L1 and Andrew Ng I SoA Exam MLC

Chapter 2: Life Tables

For example, we have the following select probabilities:

- qlxl is the probability that a life age x now dies before age x + 1, given that the life is selected

at age x.
- q[x]+i is the probability that a life age x +I now dies before age x +I+ I, given that the life
was selected at age x.

Due to the effect of underwriting, a select death probability q1, 1+1 must be no greater than the
corresponding ordinary death probability qx+i· However, the effect of underwriting will not last
forever. The period after which the effect of underwriting is completely gone is called the select
period. Suppose that the select period is n years, we have

q[xJ+r < qx+r, fort < n,

The ordinary death probability q,+1 is called the ultimate death probabilitv. A life table that
contains both select probabilities and ultimate probabilities is called a select-and-ultimate life
table. The following is an excerpt of a (hypothetical) select-and-ultimate table with a select
period of two years.

x x+2 x+2
40 0.04 0.08 42
41 0.05 0.09 43
42 0.06 0.10 44
43 0.07 0.11 45

It is important to know how to apply such a table. Let us consider a person who is currently age
41 and is just selected. The death probabilities for this person are as follows:

Age 41: q[41] = 0.05

Age 42: ql41]+1 = 0.07
Age 43: ql41]+2 = q43 = 0.09
Age 44: q[41]+J = q44 = 0.10
Age 45: q[41}t4 = q45 = 0.11

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

 Chapter 2: Life Tables

As you see, the select-and-ultimate table is not difficult to use. We progress horizontally until
we reach the ultimate death probability, then we progress vertically as when we are using an
ordinary life table. To further illustrate, let us consider a person who is currently age 41 and was
selected at age 40. The death probabilities for this person are as follows:

Age 41: q[40]+1 = 0.06

Age 42: q[40]+2 = q42 = 0.08
Age 43: q[40]+3 = q.3 = 0.09
Age44: q[40]+4 = q•• = 0.10
Age 45: q[40]+5 = q.5 = 0.11

Even though the two persons we considered are of the same age now, their current death
probabilities are different. Because the first individual has the underwriting done more recently,
the effect of underwriting on him/her is stronger, which means he/she should have a lower death
probability than the second individual.

We may measure the effect of underwriting by the index of selection, which is defined as
follows:

I(x,k) = J- ql,J+k .
q.t+k (

(
For example, on the basis of the preceding table, 1(41, I)= 1 - q1•11+ 1/q42 = 1 - 0.07/0.08 = 0.125. \
If the effect of underwriting is strong, then q 1,1+k would be small compared to qx+k, and therefore
I(x, k) would be close to one. By contrast, if the effect of underwriting is weak, then qlxJ+k would

be close to q,+k, and therefore l(x, k) would be close to zero.

<; Actex 2013 I Johnny l.i and Andrew Ng I SoA Exam MLC
Chapter 2: Life Tables

Example 2.6 [Course 3 Fall 2001 #2)

For a select-and-ultimate mortality table with a 3-year select period:

(i)
x x+3 x+3
60 0.09 0.11 0.15 63
61 0.10 0.12 0.16 64
62 0.11 0.13 0.17 65
63 0.12 0.14 0.18 66
64 0.13 0.15 0.19 67

(ii) White was a newly selected life on 01101/2000.

(iii) White's age on 01/01/2001 is 61.
(iv) Pis the probability on 01/01/2001 that White will be alive on 01/01/2006.
Calculate P.
(A)O s P < 0.43

(B) 0.43 s P < 0.45

(C) 0.45 s p < 0.47
(D)0.47 s P < 0.49
(E) 0.49 s P < 1.00

- Solution
White is now age 61 and was selected at age 60. So the probability that White will be alive 5

years from now can be expressed as P = sP[60J+l · We have

P = sP[60J+l

= Pi60J+l x Pi60J+2 x Pi60J+3 x P[60J+4 x P160J+s

= P160J+1 x P160J+2 x P63 x P64 x P6s
=(I - q160J+1)(1 - q!60J+2)(l - q63)(1 - q64)(1 - q6s)
=(l-O.ll)(J-0.13)(1-0.15)(1-0.16)(1-0.l7)

= 0.4589.
Hence, the answer is (C).
[ END]

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

 (c2-rn)
________..,_"--'"'.,,..______________________________________________________________________
2: Life Tables
....
Chapter

Sometimes, you may be given a select-and-ultimate table that contains the life table function/,.
In this case, you can calculate survival and death probabilities by using the following equations:

/[x]+1+.1·
s P[x]+I =- - - '
1[xj+I

l /[x]+l+s
1·q[x]+1 = --- - .
1[xJ+t

Let us study the following two examples.

Example 2.7

You are given the following select-and-ultimate table with a 2-year select period:
x lrxJ l :x +l lx+2 x+2
30 9907 9905 9901 32
31 9903 9900 9897 33
32 9899 9896 9892 34
33 9894 9891 9887 35

Calculate the following:

(a) zqp1J

(b) ZP[JO]+l

(c) 112qp1J+l

- Solution

11 9897
(a) 2
q1,, 1 = 1- " 1• 2 = 1- /33 = 1- = 0.000606.
/[31) /[JI] 9903

(b) - l[JOJ+l+2 = = 9897 = 0 999192

2P[3o]+I - · · ·
/[JO]+! /[JOJ+l 9905

l[JIJ+l+I - l[Ji]+l+l+2 I,, -1,, 9897 -9887 00 I 0 I 0

(c) ll2q[JIJ+I = l = 0. ·
[31]+1 /[JI]+! 9900

( END ]

'" Acrex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

Chapter 2: Life Tables

Example 2.8 [Course 3 Fall 2000 #10]

You are given the following extract from a select-and-ultimate mortality table with a 2-year
select period:

x fix] Ix +I lx+2 x+2

60 80625 79954 78839 62
61 79137 78402 77252 63
62 77575 76770 75578 64

Assume that deaths are uniformly distributed between integral ages.

Calculate o.9q[60J+06·
(A) 0.0 I 02 (B) 0.0 I 03 (C) 0.0104 (D) 0.0105 (E) 0.0106

- Solution
Recall that when UDO is assumed and the subscript on the right-hand-side is not an integer, we
will need to use the trick. We first calculate 09p16o]+-06· Using the trick, we have

o.6Pl60J x o.9P[60J+O 6 = 1.sP[60J·

Then, we have

1,p[60J P160J osP[60J+I

o.9 Pr6oJ+o.6 = =
o.6 Pr6oJ o6Pr6oJ
P16010 - 0.5q1601+1)
l-0.6q[60)

/[60)+1 [1- 0.5 (1- /[60]+2

/[60) 1[60)+1
JJ
1-0.6 (1-
11601 1
+
/[60)
J
79954 (1-0.5 (1- 78839 ))
80625 79954
- 1-0.6(1- 79954)
80625
= 0.989705.

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

 Chapter 2: Life Tables

As a result, o9q[60J+0.6 = I - 0.989705 = 0.0103. Hence, the answer is (B). (-

Alternatively, you can make use of the fact that the UDD assumption is equivalent to a linear
interpolation between/., and lx+i ·This means that
l[60J+o6 = 0.6 x 79954 + 0.4 x 80625 = 80222.4,
and that
/[60]+15 = 0.5 x 79954 + 0.5 x 78839 = 79396.5.
As a result,
80222.4- 79396.5
0.9 q[60]+0.6 = 0.0103.
80222.4

[END)

• 2. 4 Moments of Future Lifetime Random Variables

In Exam P, you learnt how to calculate the moments of a random variable.

- If Wis a discrete random variable, then E(W) = Lw Pr(W = w) .

w

- If Wis a continuous random variable, then E(W) = [ wfw(w)dw, where fw(w) 1s the

density function for W.

2
- To calculate variance, we can always use the identity Var(W) = E(W ) - [E(W)]2.

First, let us focus on the moments of the future lifetime random variable T,. We call E(T,) the

complete expectation of life at age x, and denote it by;,. We have

,
e, =Jo lfx(t)dt =Jo t,pxµx.,dt.

By rewriting the integral as - r tdS, (t) and using integration by parts, we can show that the

above formula can be simplified to.

;l;• Actex 2013 [ Johnny Li and Andrew Ng [ SoA Exam MLC

 Chapter 2: Life Tables

" = Jor· , p ,dt.

e,

Note that ifthere is a limiting age, we replace oo with m-x.

The second moment of T, can be expressed as

E(T,') = r .!,
1
2
(t)dt .

Using integration by parts, we can show that the above formula can be rewritten as

E(I.; )=2
2
r t,p,dt,

which is generally easier to apply. Again, if there is a limiting age, we replace oo with OJ- x.

In the exam, you may also be asked to calculate E(T, /\ n) = E[min(T,, n)]. This expectation is

known as the n-year temporary complete expectation of life at age x, and is denoted by It

can be shown that

The following is a summary of the formulas for the moments of T,.

Moments of Tx
:::;:';:;'.. " r·
e, =Jo ,p,dt (2.6)
%
%:
E(I.;
2
) =2 rt ,P, di (2. 7)

;I':
,%%
• = r ,P, dt (2.8) ;r,,:

1
L Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC
 (c2-22i
_______...,.."--"'_,..____________________________________________________________________ Chapter 2: Life Tables

Example 2.9

You are givenµ_,= 0.0 I for all x 0. Calculate the following:

(a) ex
(b) Var(Tx)

- Solution

(a) First of all, we have iPx = e-0-011 • Then,

0

ex= 1 00

pd!= 1 e -o.ou dt=--e

00
I [ -o01< loo=--=100.
o I
0 ' x 0 -0.01 0.01
(b) We first calculate the second moment of Tx as follows:

E(T,
2
) r
= 2 te-0 01 'dt

= - r 01
e-0 'dt)

= - 2-
0.01
r e-0.0l'dt

2
= - - , = 20000.
0.01
Then, the variance of Tx can be calculated as:

Var(TJ = E(T,') -[ E(T,) ]'

2
= 20000 -100
= 10000.

[END]

Example 2.10 [Course 3 Fall 2001 #1]

You are given:

-{0.04, 0<x<40
µx - 0.05, x>40

Calculate e 2s1.
25

(A) 14.0 (B) 14.4 (C) 14.8 (D) 15.2 (E) 15.6

r.G• Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

Chapter 2: Life Tables

- Solution
First, we need to find 1p,. Because the value of µx changes when x reaches 40, the derivation of
iPx is not as straightforward as that in the previous example.
For 0 <I< 15, µ 25+1 is always 0.04, and therefore

_ -J;o.o4d11 _ -0. 041

1P2s - e - e ·

Fort> 15, µ25+1 becomes 0.05, and therefore

-
1P2s - e
{I" 0
0.04d11+
J' O.OSd11l -
15 - e--0.04xl5-0.05(1-l5) -
- e
--0.051+0 15
.

Given the expressions for 1p25, we can calculate e 2s1 as follows:

25

rs
= .b ,p,,dt +
r25
,p,,dt
= 1s e--0.04'dt + J:s e-0.os1+0_1sdt

= e --004,
15
+eo.1s _ _[e--0.os, ]"
-0.04 0 -0.05 15

= 15.60.
Hence, the answer is (E).
[ END)

Now, we focus on the moments of the curtate future lifetime random variable K,. The first
moment of Kx is called the curtate expectation of life at age x, and is denoted by e,. The formula
for calculating ex is derived as follows:

e, =E(KJ

k,,,Q

= :Lk,1q,
k=O

= Oxqx +Ix 111q, +2x 211q, +3x 311qx + ...

= (p, - ,p.) + 2(,p, - ,p,) + 3(,p, - ,p,) + ...
=p,+,p,+,p,+ ...

?,:• Actex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

 Chapter 2: Life Tables

If there is a limiting age, we replace oo with w - x.

I

The formula for calculating the second moment of Kx can be derived as follows:

00 00

E(K;) = 2)' Pr(Kx = k) = l:k' klqx

k:O b=O

= O' xq, +I' x i11qx +2' x i11q, +3' x 'l'q, + ...

= (p, - 2 p,) + 4(,p, - ,p,) +9(,px - 4 p,) + "'
= Px + 3 zPx + 5 ,p, + "'

= f (2k-'I) P,·
k=I
1

Again, if there is a limiting age, we replace oo with w - x. Given the two formulas above, we
can easily obtain Var(Kx).

In the exam, you may also be asked to calculate E(K, A n) = E[min(K,, n)]. This is called the

n-year temporary curtate expectation of life at age x, and is denoted by e·"". It can be shown that

,,
e,,, = 2: ,p,,
k=I

that is, instead of summing to infinity, we just sum ton.

(

t-
There are two other equations that you need to know. First, you need to know that e, and ex+ 1 are
related to each other as follows:

ex= p,(I + ex+J).

Formulas of this form are called recursion formulas. We will further discuss recursion formulas
in Chapters 3 and 4.

Second, assuming UDO holds, we have T, = Kx + U, where U follows a uniform distribution

over the interval [O, !]. Taking expectation on both sides, we have the following relation:

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

Chapter 2: Life Tables

. 1
e:r =e.f+
2.
The following is a summary of the key equations for the moments of Kx.

« .
:'-'i'/////////////,"./'//////////////,'l'//////////////,'l'//////////////,'l'//////////////,/

i%
•% Moments of Kx
?,:
%'.
L kPx·
ro '/,'
ex= (2.9)
:% kol %:
:% ro %;
E(K;)= L(2k-l),Px· (2.10)
:% %:
=I. ,p,
kol

:% (2.11) %:
:% k=l %'
;% ex= Px( 1 + ex+i). (2.12) %:
;% • 1
:% Under UDO, e, =ex+- (2. I 3) %,
2 %:

Example 2.11

You are given the following excerpt of a life table:

x 95 96 97 98
Ix 400 300 100 0
Calculate the following:
(a) e9s
(b) Var(K9s)
(c) e,,,,
(d) ; 95 , assuming UDO
(e) e96. using the recursion formula

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

 (c2-26)
________...,"'--"",,......____________________________________________________________________
Chapter 2: Life
__
Tables

-Solution

(b) We have
3

E(Ki,)=

I I I
= __'!£_ + 3 x -2L+ 5 x-2!L
/ 95 I,, 195
= 300 + 3 x 100 + 5 x___Q_
400 400 400
= 1.5.
Hence, Var(K95 ) = 1.5 - 12 = 0.5.

(c) e,,i
.
=I ,p,, =
k=I
1.. = 300
I,, 400
= 0.75.

(d) Assuming UDD, e,, = e 95 + 0.5 = I+ 0.5 = 1.5.

(e) Using the recursion formula, e 95 =p 95 (1 + e96 ) = 0.75(1 + e96). Therefore,

e 96 = 1/0.75 - 1 = 0.3333.

[END]

• 2. 5 Useful Shortcuts

Constant Force of Mortality for All Ages

Very often, you are given that µx =µfar all x 2'. 0. In this case, we can easily find that

-µI
1Px = e , F(t)=
x 1-e-µ' , fx(t) = µe-µ1.

From the density function, you can tell that in this case T., follows an exponential distribution
with parameterµ. By using the properties of an exponential distribution, we have

e, =E(Tx)= IIµ, Var(Tx) =I/Ji- for all x.

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

Chapter 2: Life Tables

These shortcuts can save you a lot of time on doing integration. For instance, had you known
these shortcuts, you could complete Example 2.9 in a blink!

De Moivre's Law

De Moivre's law refers to the situation when

Ix = W- X for 0 ,,; x < w,

or equivalently
1
µ, =--.
w-x

De Moivre's law implies that the age at death random variable (To) is uniformly distributed over
the interval [O, w). It also implies that the future lifetime random variable (Tx) is uniformly

distributed over the interval [O, w- x), that is, for 0,,; I < w - x,

I I
F,(t)=-, f,(t)=-, µx+I = (
w-x w-x w-x-

By using the properties of uniform distributions, we can immediately obtain

Var(Tx) = (w-x)'
12

The useful shortcuts are summarized in the following table.

.
e, Var(T,)
Assumption µx+r 1Px Fx(t) fx(t)

Constant force
for all ages
µ e-µr l -e-111 µe
-µI
IIµ 11µ-

I t I I w-x (w-x) 2
De Moivre's law 1--- -- -- --
w-x-t w-x w-x w-x 2 12

r' Acrex 2013 I Johnny Li and Andrew Ng I SoA Exam MLC

 ((228) Chapter 2: Life Tables

--------...

Example 2.12

You are given:

Ix= 100-x, O:Sx< 100.

Calculate the following:

(a) 2sp2s
(b) q15

(c) µso
(d) eso

-Solution
First of all, note that 1., = I 00- x for 0 x < I 00 means mortality follows De Moivre's law with
OJ= 100.

l 25 2
(a) ,,p,, = -100-25 3
(b) q2s =I - l26 / i2s =I - 74/75 = 1175.
Alternatively, you can obtain the answer by using the fact that T25 is uniformly distributed
over the interval [O, 75). It immediately follows that the probability that (25) dies within one
year is 1/75.
I (
(c) µ,,=100-50 °· 02 · \

100-50 -25
(d) eso -
0 _

- .
2

[ END]

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