Beruflich Dokumente
Kultur Dokumente
W.F.Scott
c
°1999 W.F.Scott
Department of Mathematical Sciences
King’s College
University of Aberdeen
Aberdeen
AB24 3UE
U.K.
2
Preface
This book consists largely of material written for Parts A2 and D1 of the U.K. actuarial exami-
nations (old system). It is hoped that the material given here will prove useful for much of Subjects
104 and 105 of the new examinations and certain similar examinations at universities, and that it
may also be useful as a general reference work on life assurance mathematics.
3 ASSURANCES 37
3.1 A general introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Whole life assurances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3 Commutation functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.4 The variance of the present value of benefits . . . . . . . . . . . . . . . . . . . . . . . 39
3.5 Assurances payable at the end of the year of death. . . . . . . . . . . . . . . . . . . . 41
3.6 Assurances payable at the end of the 1/m of a year of death. . . . . . . . . . . . . . 44
3.7 Temporary and deferred assurances . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.8 Pure endowments and endowment assurances . . . . . . . . . . . . . . . . . . . . . . 46
3.9 Varying assurances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.10 Valuing the benefits under with profits policies . . . . . . . . . . . . . . . . . . . . . 50
3.11 Guaranteed bonus policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.12 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.13 Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4 ANNUITIES 61
4.1 Annuities payable continuously . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2 Annuities payable annually . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.3 Temporary annuities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.4 Deferred annuities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3
4 CONTENTS
5 PREMIUMS 81
5.1 Principles of premium calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.2 Notation for premiums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.3 The variance of the present value of the profit on a policy. . . . . . . . . . . . . . . . 83
5.4 Premiums allowing for expenses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.5 Premiums for with profits policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.6 Return of premium problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.7 Annuities with guarantees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.8 Family income benefits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.10 Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6 RESERVES 95
6.1 What are reserves? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.2 Prospective reserves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.3 Net premium reserves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.4 Retrospective reserves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
6.5 Gross premium valuations and asset shares. . . . . . . . . . . . . . . . . . . . . . . . 102
6.6 The variance of L . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.7 Zillmerised reserves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.8 Full preliminary term reserves. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.9 Reserves for with-profits policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.11 Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
9 PROFIT-TESTING 151
9.1 Principles of profit-testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
9.2 Cash flow calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
9.3 The profit vector and the profit signature . . . . . . . . . . . . . . . . . . . . . . . . 155
9.4 The assessment of profits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
9.5 Some theoretical results about {σt } . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
9.6 Withdrawals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
9.7 The actual emergence of profits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
9.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
9.9 Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
APPENDICES 333
A Some notes on examination technique . . . . . . . . . . . . . . . . . . . . . . . . . . 333
B Some technical points about the tables used in examinations . . . . . . . . . . . . . 334
C Some common mistakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
D Some formulae for numerical integration . . . . . . . . . . . . . . . . . . . . . . . . . 336
SUPPLEMENT 337
8 CONTENTS
Chapter 1
lx = lα s(x) (1.1.2)
where lα is called the “radix” of the table, and is usually a large number such as 1, 000, 000. If lives
are considered from birth, we naturally have α = 0, and we have lx > 0 for x < w, lw = 0, unless
w is infinite, in which case lx → 0 as x → ∞. The function lx is assumed to be continuous. By
formulae (1.1.1) and (1.1.2) ,
lx
P r{ (α) survives to age x } = (1.1.3)
lα
Example 1.1.1. English Life Table No.12 - Males refers to the male population of England and
Wales in 1970-72. In this table α = 0 , w may be taken to be about 106, and the radix of the table is
100, 000. We observe (for example) that, since l50 = 90, 085 , the probability that a new-born child
will survive to age 50 is s(50) = l50 /l0 = 0.90085, or 90.1%.
We now consider survival from age α in terms of a random variable. Let T (or T (x) if the age x
◦
is not clear; Tx is sometimes used, but this may be confused with the symbol Tx = lx ex used in
stationary population problems) be the variable representing the future lifetime (in years, including
fractions) of (x). According to our definitions,
9
10 CHAPTER 1. NON-SELECT LIFE TABLES
t px + t qx = 1 (1.2.7)
s(x + t) lx+t
t px = = (1.2.8)
s(x) lx
and
lx+t
t qx =1− (1.2.9)
lx
That is, the (cumulative) distribution function of the variable T is
(
1 − lx+t
lx , if t≥0
F (t) = P r{T ≤ t} = t qx = (1.2.10)
0 , if t<0
We may say that “the expected number of survivors at age x + t among lx lives aged x is lx+t ”
because the chance that a given life aged x will survive to age x + t is lx+t
lx . In practice, the word
“expected” is sometimes omitted, but it is always to be understood; assuming that the lx lives are
independent, the number of survivors at age x + t will follow a binomial distribution with mean
lx · t px = lx+t
and variance
µ ¶
lx+t
lx · t px (1 − t px ) = lx+t 1 −
lx
1.2. PROBABILITIES OF DEATH AND SURVIVAL 11
(Even if the lives are not independent, the formula for the expected number of deaths is correct.)
Note that
lx+1
px = (1.2.11)
lx
lx+1 dx
qx = 1 − = (1.2.12)
lx lx
where
dx = lx − lx+1 (1.2.13)
= the (expected) number of deaths at age x last
birthday among lx lives aged x
Example 1.2.1. The following extract from English Life Table No. 12 - Males illustrates the
functions lx , dx , px and qx .
age
x lx dx px qx
(The high death rate at age 0, relative to the rates at ages 1 to 4, is quite noticeable.)
We also observe that, for all ages x3 ≥ x2 ≥ x1 ≥ α , we have the “rule of multiplication of
probabilities of survival”:
x3 −x1 px1 = x2 −x1 px1 · x3 −x2 px2 (1.2.14)
(which is merely a re-statement of equation (1.2.1). )
in agreement with our earlier definition. Further, t px is sometimes written as Sx (t). When α = 0
we sometimes write S0 (t) = S(t), so S(t) and s(t) are equal. The notation S(t) is much used in
survival analysis, in which t denotes the time lived by a patient since the start of a given treatment.
In reliability engineering the notation R(t) may be used instead of S(t).
12 CHAPTER 1. NON-SELECT LIFE TABLES
lx0
µx = − (1.3.2)
lx
Proof. we have
−[s(x + h) − s(x)] −s0+ (x)
µx = lim =
h→0+ hs(x) s(x)
where s0+ (x) denotes the R.H. derivative of s(x). But s(x) and s0+ (x) = −s(x)µx are continuous
on [α, w), so s0 (x) exists and equals −s(x)µx (see McCutcheon and Scott, An Introduction to the
Mathematics of Finance, Appendix 1. )
Hence
s0 (x) l0
µx = − =− x
s(x) lx
We remark that
d
µx = − log lx (1.3.3)
dx
Hence
Z x Z x
d
− µy dy = (log ly ) dy
α α dy
x
= [ log ly ]α
= log lx − log lα
µ ¶
lx
= log
lα
and
µ Z x ¶
s(x) = exp − µy dy (1.3.5)
α
It follows that
µ Z x+t ¶ µ Z t ¶
lx+t
t px = = exp − µy dy = exp − µx+r dr (1.3.6)
lx x 0
If µx is piecewise continuous, we may apply the above formulae over each age-range and combine
the results; the above formulae remain true, except that lx0 does not exist at the points at which
1.3. THE FORCE OF MORTALITY, µx 13
µx is not continuous (the R.H. and L.H. derivatives of lx differ at these points.) Alternatively, we
may regard the actual position as closely approximated by a model in which the force of mortality
is continuous.
We may now express the probability density function of the variable T (the future lifetime of (x))
in terms of the force of mortality. The p.d.f. f (t) of T is the derivative of the distribution function
F (t), i.e. ³ ´
d
0 1 − lx+t
lx , if t > 0
f (t) = F (t) = dt
0 , if t < 0
By the chain rule
d 0
lx+t = lx+t
dt
= −lx+t µx+t (by formula (1.3.2))
so we obtain (
lx+t µx+t
lx = t px µx+t , if t > 0
f (t) = (1.3.7)
0 , if t < 0
In view of the following general formula (connecting the d.f. and p.d.f.):
Z t
F (t) = f (r) dr
−∞
we have Z t
t qx = r px µx+r dr (1.3.8)
0
An alternative notation
Some statisticians refer to the force of mortality as the “hazard rate” or “transition intensity.”
Let f (t) and F (t) denote the p.d.f. and d.f. respectively of T = T (x) . The hazard rate at age x + t
(often considered to be “at time t years after entry to assurance (say) at age x”) is defined in terms
of f (t) and F (t) by the formula
f (t)
h(t) = (1.3.9)
1 − F (t)
which equals µx+t (by formulae (1.2.7) and (1.3.7).) If α = 0 and x = 0, we naturally have T = the
future lifetime of a new-born child, and
f (t)
µt = = hazard rate at age t years (1.3.10)
1 − F (t)
Solution.
µ Z x ¶
s(x) = exp − µy dy
µ Zαx ¶
δ−1
= exp − cδy dy
³ £α ¤x ´
= exp −c y δ α
¡ ¢ ¡ ¢
= exp −cxδ . exp cαδ
hence
¡ £ ¤¢
t px = Sx (t) = exp −c (x + t)δ − xδ
Deferred probabilities
The symbol m | indicates deferment for m years; for example,
m |n qx = P {(x) will die between ages x + m and x + m + n } (1.3.14)
By elementary probability, this equals
P r{(x) will die before age x + m + n} − P r{(x) will die before age x + m}
= m+n qx − m qx
= m px − m+n px
lx+m − lx+m+n
=
lx
That is,
lx+m − lx+m+n
m |n qx = (1.3.15)
lx
This may be remembered by the following rule: of lx lives aged x, lx+m − lx+m+n is the (expected)
number of deaths aged between ages x + m and x + m + n. If n = 1 it may be omitted, so we have
Example 1.3.2. Consider the force of mortality to be as in Example 1.3.1. Find an expression for
the chance that (x) will die between ages x + m and x + m + n.
Solution. We require:
lx+m − lx+m+n
m |n qx =
lx
s(x + m) − s(x + m + n)
=
s(x)
exp[−c(x + m)δ ] − exp[−c(x + m + n)δ ]
=
exp[−cxδ ]
by formula 1.3.7. If w < ∞ , the upper limit of the integral should be replaced by w − x.
Theorem 1.4.1. Z ∞
◦
ex = t px dt (1.4.3)
0
(with ∞ replaced by w − x if w is finite).
Proof. By integration by parts, with u = t , v = −t px ,
Z ∞ Z ∞
◦ ∞
ex = t · t px µx+t dt = [−t · t px ]0 − −t px dt
0 0
Z ∞
= t px dt
0
since it may be shown that t · t px → 0 as t → ∞ (using the fact that E(T ) < ∞. ).
We may also use the formula
Var(T ) = E(T 2 ) − [E(T )]2 (1.4.4)
where
Z ∞
E(T 2 ) = t2 · t px µx+t dt
0
Z ∞
£ ¤∞
= −t2 · t px 0 + 2t · t px dt
0
(on setting u = t2 , v = −t px )
Z ∞
=2 t · t px dt
0
16 CHAPTER 1. NON-SELECT LIFE TABLES
Example 1.4.1. Find the median future lifetime of (10), according to English Life Table No.12 -
Males.
This variable is used in many actuarial calculations. In particular, the curtate expectation of
life at age x, written ex , is the mean of K ; that is
∞
X ∞
X
ex = k · k |qx = k.k |qx (1.4.9)
k=0 k=1
1.5. THE ASSUMPTION OF A UNIFORM DISTRIBUTION OF DEATHS 17
Theorem 1.4.2.
∞
X lx+1 + lx+2 + . . .
ex = k px = (1.4.10)
lx
k=1
Proof.
dx+1 dx+2 dx+3
ex = +2 +3 + ...
lx lx lx
1
= [(dx+1 + dx+2 + dx+3 + . . .)
lx
+ (dx+2 + dx+3 + . . .)
+ (dx+3 + . . .)
+ . . .]
lx+1 + lx+2 + lx+3 + . . .
=
lx
Theorem 1.5.1. The following conditions are each equivalent to the assumption of U.D.D. between
ages x and x + 1:
t qx = t · qx (0 ≤ t ≤ 1) (1.5.3)
t px µx+t = qx (0 ≤ t < 1) (1.5.4)
18 CHAPTER 1. NON-SELECT LIFE TABLES
Example 1.5.1. In a certain non-select mortality table, there is a uniform distribution of deaths
between any two consecutive integer ages. Find formulae in terms of l30 , l31 and l32 for
Solution.
l32 l32
(i) = 1 since l30+t is linear for 0 < t < 1
l30.5 2 (l30 + l31 )
and hence
◦
Tx = lx ex (1.6.2)
We also define
Z 1
Lx = lx+t dt (1.6.3)
0
1.7. LAWS OF MORTALITY 19
Note that
Z x+1
Lx = ly dy
Zx∞ Z ∞
= ly dy − ly dy
x x+1
= Tx − Tx+1 (1.6.4)
Assume that there is a Uniform Distribution of Deaths (U.D.D) between the ages x and x + 1, i.e.
lx+t is linear for 0 ≤ t ≤ 1.
We have
Z 1
Lx = lx+t dt = 12 (lx + lx+1 ) (1.6.5)
0
Lx = lx − 21 dx (1.6.6)
Formulae (1.6.5) and (1.6.6) are sometimes used as approximations when U.D.D. does not hold.
The central death rate at age x is defined as
dx
mx = (1.6.7)
Lx
Example 1.7.1. Show that, if Gompertz’ Law holds for all ages greater than or equal to α, there
is a positive constant g such that
x
(ct −1)
t px = gc for x ≥ α, t ≥ 0 (1.7.2)
Solution.
· Z t ¸
t px = exp − µx+r dr
0
· Z t ¸
= exp −Bcx cr dr
0
· ¸
Bcx (ct − 1)
= exp −
log c
· ¸
x
(ct −1) B
= gc with g = exp −
log c
In 1860 Makeham suggested the addition of a constant term to Gompertz’ formula for µx , giv-
ing Makeham’s law:
µx = A + Bcx (x ≥ α) (1.7.3)
Example 1.7.2. Show that, if Makeham’s law holds for all ages greater than or equal to α, there
are positive constants s and g such that
x
(ct −1)
t px = st g c for x ≥ α, t ≥ 0 (1.7.4)
Solution.
· Z t ¸
t px = exp − µx+r dr
0
· Z t ¸
= exp − (A + Bcx cr )dr
0
x
(ct −1)
= exp[−At]g c (using Example 1.7.1)
· ¸
x
(ct −1) B
= st g c with g = exp − , s = e−A
log c
1.7. LAWS OF MORTALITY 21
Weibull’s law of mortality has already been mentioned in Example 1.3.1. In practice it is usu-
ally less successful than those of Gompertz and Makeham in representing human mortality. The
fitting of laws of mortality is complicated by the fact that mortality rates may be varying with time.
If a law of mortality holds, certain mortality and monetary functions may be evaluated analyt-
ically (rather than numerically), but this point is of little practical importance in the computer
age. There are also certain simplifications in the evaluation of joint-life functions (that is, functions
depending on the survival of more than one life.)
22 CHAPTER 1. NON-SELECT LIFE TABLES
Exercises
1.1 Calculate the following probabilities on the basis of English Life Table No. 12-Males.
(i) 40 p25 ,
(ii) 60 |10 q25 ,
(iii) the probability that (30) survives for at least 10 years,
(iv) the probability that (40) survives to age 65,
(v) the probability that (50) dies within 10 years,
(vi) the probability that (50) fails to reach age 70,
(vii) the probability that (60) dies between ages 80 and 85,
(viii) the probability that (60) dies within the first five years after retiring at age 65.
In addition, express (i) and (ii) in words.
1.2 On the basis of E.L.T. No. 12 - Males, find the probability that a life aged 30 will
1.3 A man aged 50 has just retired because of ill health. Up to exact age 58 he will be subject to
a constant force of mortality of 0.019803 p.a., after which his mortality will be that of E.L.T.
No. 12 - Males. Find the probability that he will
1.4 For the first 5 years after arrival in a certain country, lives are subject to a constant force of
mortality of 0.005. Thereafter lives are subject to mortality according to English Life Table
No. 12- Males with an addition of 0.039221 to the force of mortality.
(a) Show that the probability that the life will survive to age 35 is 0.97531.
(b) Find the probability that the life will survive to age 60.
(ii) What is the probability that a life aged exactly 33 who has been in the country for 3
years will die between ages 50 and 51?
(Assume that these lives will remain in the given country.)
1.7 Suppose that Gompertz’ law, µx = Bcx , holds for all x ≥ α, c being greater than 1. Assume
that µα < log c.
1.8 (Difficult) Suppose that there is a “uniform distribution of deaths” from age x to age x + 1,
i.e. lx+t is linear in t for 0 ≤ t ≤ 1.
Show that, for all 0 ≤ s ≤ t ≤ 1,
t−s
t−s qx+s = qx
1 − sqx
1.9 (Difficult) Suppose that the “Balducci hypothesis” holds from age x to age x + 1, i.e.
1−t qx+t = (1 − t)qx for 0 ≤ t ≤ 1.
(i) µx ,
◦
(ii) ex (Note finite limiting age),
(iii) 10 p70 ,
(iv) 40 |5 q35
1.11 A certain group of lives now aged 60 experience mortality according to a(55) males ultimate
with addition to the force of mortality. The addition is 0.0005 at age 60, increasing linearly to
0.0025 at age 80, at which level the addition remains constant.
Find the probability that a life aged exactly 60 dies within 20 years.
24 CHAPTER 1. NON-SELECT LIFE TABLES
Solutions
l65
1.1 (i) 40 p25 = = 0.71528
l25
l85 − l95
(ii) 60 |10 q25 = = 0.10021
l25
l40
(iii) 10 p30 = = 0.98452
l30
l65
(iv) 25 p40 = = 0.73025
l40
l50 − l60
(v) 10 q50 = = 0.12389
l50
l50 − l70
(vi) 20 q50 = = 0.39162
l50
l80 − l85
(vii) 20 |5 q60 = = 0.16173
l60
l65 − l70
(viii) 5 |5 q60 = = 0.17338
l60
In words,
(i) is “the probability that a life aged 25 will survive to age 65.”
(ii) is “the probability that a life aged 25 will survive to age 85 and die before age 95.”
l40
1.2 (i) 10 p30 = = 0.98452
l30
l30 − l50
(ii) 20 q30 = = 0.05437
l30
d49
(iii) 19 |q30 = = 0.00613
l30
l40 − l50
(iv) 10 |10 q30 = = 0.03889
l30
1.3 Let us use an asterisk to denote the mortality table of the life concerned, while lx etc. refers
to E.L.T. 12-Males. Let k = 0.019803.
∗
(i) 5 q50 = 1 − 5 p∗50 = 1 − exp(−5k) = 0.09427
1.9. SOLUTIONS 25
∗ l65
(ii) 15 p50 = 8 p∗50 7 p∗58 = exp(−8k) = 0.71188
l58
∗
(iii) 5 |5 q50 = 5 p∗50 − 10 p∗50 = 5 p∗50 − 8 p∗50 2 p∗58
l60
= exp(−5k) − exp(−8k) = 0.08540
l58
µ Z 5 ¶
1.4 (i) (a) Prob. of survival for 5 years = exp − 0.005dt = 0.97531
0
ELT
l
(b) 0.97531 exp (−0.039221 × 25) 60
ELT = 0.3051
l35
(ii) Prob. = prob. of survival for 17 years - prob. of survival for 18 years
ELT
l50
= exp(−2 × 0.005) exp (−0.039221 × 15) ELT
l35
ELT
l51
− exp(−2 × 0.005) exp (−0.039221 × 16) ELT
l35
= 0.02377
Z ∞ Z ∞ · ¸∞
◦ lx −1
1.5 (i) e0 = dx = (1 + x)−2 dx = =1
0 l0 0 1+x 0
l0 2
(ii) µx = − x = , so µ1 = 1
lx 1+x
l1 − l2 1 1
(iii) 1|
q0 = = − = 0.13889
l0 4 9
Now f (0) = 0, and f (t) is decreasing for 0 ≤ t ≤ 1, so f (1) < 0; therefore qx < µx .
Let x0 be the unique point at which this occurs. Notice that (by equation (∗) above)
d
(lx µx ) > 0 for x < x0 , and is negative for x > x0 .
dx
Therefore x0 is a maximum point of lx µx .
(t − s)qx lx 1
Therefore t−s qx+s = , since = .
1 − sqx lx+s 1 − sqx
1.9 µ ¶
lx+t lx+1 lx+t lx+1
t−s qx+s = 1− =1− − 1−
lx+s lx+s lx+s lx+t
= q
1−s x+s − (1 − q ) q
t−s x+s 1−t x+t
Now apply “Balducci” to the R.H.S. of the equation. This gives the desired result.
d a
1.10 (i) µx = − [log s(x)] =
dx w−x
(ii)
Z w−x Z w
◦
ex = t px dt = ly dy/lx
R0w x
x
(1 − y/w)a dy
=
(1 − x/w)a
= 0.4108
∴ Ans. = 0.5892
28 CHAPTER 1. NON-SELECT LIFE TABLES
Chapter 2
In the previous chapter we considered mortality rates to depend only on age: life tables of this form
are sometimes called aggregate life tables. We now consider the situation when mortality rates (or
the force of mortality ) depend on two factors: (i) age, and (ii) the time (duration) since a certain
event, known as “selection”.
One important example of “selection” is the acceptance of a proposal for life assurance at normal
rates of premium: the mortality rates of lives who have been recently accepted for life assurance
at normal rates may be expected to be lower than those of the general population. After a certain
period the difference in mortality between those who have been accepted for life assurance and the
general population of the same age decreases, but it is not correct to say that the effect of selection
“wears off” entirely (since the general population contains some lives who would never have been
accepted for life assurance at normal premium rates). This point may be confirmed by a comparison
between the mortality rates of English Life Table No.12 - Males and A1967-70 ultimate, which we
will discuss below. (The process by which life offices decide whether to accept lives for assurance,
and on what terms, is called “underwriting”.)
Another form of selection is the “self-selection” exercised by those who buy annuities: it may nor-
mally be assumed that such lives are in good health (for their age), for the purchase of an annuity
is otherwise likely to be a poor investment.
In these examples, the mortality of those selected is lower than that of the general population,
particularly in the period soon after selection. In “reverse selection”, mortality rates are higher than
those of the general population (or some other reference group). An example of reverse selection
is early retirement due to ill-health. After a certain period from the date of ill-health retirement,
the mortality of these lives may be expected to become closer to that of lives who retired in normal
health, or who are not yet retired.
29
30 CHAPTER 2. SELECT LIFE TABLES
y =x+t
We write “[x] + t” as a shorthand notation for a “life aged x at selection and duration t years since
selection”. Thus for example,
The select period. We assume that there is a period, s years, such that the mortality of those
selected at least s years ago depends only on the attained age, x + t. That is,
q[x]+t = qx+t (t ≥ s) (2.2.3)
µ[x]+t = µx+t (t ≥ s) (2.2.4)
where qx+t and µx+t refer to those who were selected at least s years ago: such people are called
ultimate lives.
For each fixed entry age x, we may define the family of random variables.
T ([x] + t) = the future lifetime of “[x] + t” (t ≥ 0)
Regarding x as fixed, we may construct a “life table” for those selected at age x by methods similar
to those of chapter 1 (with t in place of x, and α = 0). We have, for example,
(which are found by replacing “x” by “[x] + t” in the formulae of chapter 1) are true. We may omit
the square brackets enclosing x in expressions such as h p[x]+t and l[x]+t+h if t ≥ s or t + h ≥ s
respectively.
We now give formulae for l[x]+t for 0 ≤ t ≤ s. If q[x]+t is given for t = 0, 1, 2, . . . , s − 1 , we proceed
recursively, using the formulae
l[x]+t+1
= 1 − q[x]+t (t = s − 1, , s − 2, . . . , 0) (2.2.7)
l[x]+t
That is
l[x]+s−1 = l[x]+s /(1 − q[x]+s−1 )
l[x]+s−2 = l[x]+s−1 /(1 − q[x]+s−2 )
······························
select select
duration 0 duration 1 ultimate
34, 489 = l0
34, 489(1 − q0 ) = 34, 463.8 = l1
l
[0]+1
l[0] = 1−q [0]
← l[0]+1 = 1−ql[0]+1
2
← 34, 463.8(1 − q1 ) = 34, 440.4 = l2
··············· ··············· ← 34, 440.4(1 − q2 ) = 34, 418.7 = l3
..
.
32 CHAPTER 2. SELECT LIFE TABLES
Table 2.3.1
f (t + 1) − f (t − 1)
f 0 (t) +
2
to obtain
1
µ[x]+t + − log(l[x]+t+1 /l[x]+t−1 ) (2.4.3)
2
Exercises
2.2 In its premium rate basis, an office assumes a 3-year select period. Functions on this table are
indicated by an asterisk. The table is such that:
∗ ∗ ∗ ∗
q[x+7] = q[x+3]+1 = q[x]+2 = qx+1
and ultimate mortality follows A1967-70 ultimate. Assuming further that ly∗ = ly on A1967-70
∗
ultimate, calculate l[45]+t for t = 0, 1 and 2.
2.3 A certain life table has a select period of 1 year. At each integer age x , the select rate mortality
is 50% of the ultimate rate. Calculate e[60] , given that e60 = 17.5424 and q60 = 0.0142.
2.4 Explain briefly the concept of selection in relation to mortality tables. Define the term “select
period” and, using the A1967-70 table as an example, explain how a select table differs in
construction from an aggregate table such as English Life Table No. 12 - Males.
34 CHAPTER 2. SELECT LIFE TABLES
Solutions
l[50] − l[50]+1
2.1 (i) (a) q[50] = = .01601 (A)
l[50]
l[50]+2
(b) 2 p[50] = = .96411 (B)
l[50]
l[50]+2 − l53
(c) 2| q[50] = = .02410 (C)
l[50]
l53 − l56
(d) 2|3 q[50]+1 = = .09272 (D)
l[50]+1
2.3
l61 + l62 + . . .
e[60] =
l[60]
l60 l61 + l62 + . . .
= ·
l[60] l60
l60
= e60
l[60]
2.4 Mortality rates depend on the age x at “selection” [e.g. entry to assurance, purchase of an
annuity (self selection)] and duration t years since selection. Thus q[x]+t = Pr{ a life aged x + t
2.7. SOLUTIONS 35
who was selected t years ago will die within a year }, etc. The select period, s years, is such
that
q[x]+t = qx+t for t ≥ s
i.e. for lives who were selected at least s years ago (these being called “ultimate” lives) mortality
depends only on the attained aged. The l[x]+t function is constructed to be such that
l[x]+t = lx+t (t ≥ s)
ASSURANCES
In the case of with profits policies, the sum assured may be increased by additions, called bonuses,
which depend on the experience of the office and its bonus distribution policy. With profits policies
may be found in traditional or unitised form (see later discussion). One also encounters unit-
linked policies, in which the benefits are directly linked to the performance of certain assets, e.g. a
certain portfolio of equities. If the benefits (and premiums) are completely specified in money terms
in the contract, the policy is said to be without profits (or non-profit). Policies may also provide
endowment benefits, i.e. sums payable on the survival of the life (or lives) assured until a certain
date (or dates).
In this chapter we shall consider contracts issued on the life of one person. The benefits and expenses
are paid for by premiums, which we shall discuss later.
δ = log(1 + i)
Now consider a life aged x, who is subject to a certain non-select mortality table, and let T be the
future lifetime of (x). A whole life assurance is a policy providing a certain sum assured, S, on
the death of (x) at any future date. The present value of this benefit (assumed for the moment to
be payable immediately on death) is
Z = g(T ) = Sv T (3.2.2)
37
38 CHAPTER 3. ASSURANCES
The mean of this variable, E(Z), is called the mean (or expected) present value (M.P.V.) of
the whole life assurance benefit, and is
Z ∞
E(Z) = g(t)f (t) dt (3.2.3)
0
When S = 1 , we write
Z ∞
E(Z) = v t t px µx+t dt = Āx (3.2.5)
0
where A stands for “assurance” and the bar indicates that the sum assured is payable immediately
on the death of (x). The M.P.V. of the sum of £S payable immediately on the death of (x) is
therefore
S Āx
Example 3.2.1. Find the M.P.V. of a whole life assurance of £10, 000 , payable immediately on
the death of (40), according to E.L.T. No. 12-Males with interest at 4% p.a.
Solution.
Select tables
For a select life (i.e. a person who has just been selected) aged x, we replace x by [x], and for an
“[x] + t” life (i.e. a life now aged x + t who was selected t years ago) we replace x by [x] + t. Thus,
letting T = T ([x]) denote the future lifetime of “[x]”, we write
Ā[x] = M.P.V. of v T
Z ∞
= v t t p[x] µ[x]+t dt (3.2.6)
0
Ā[x]+t = M.P.V. of v T
Z ∞
= v r r p[x]+t µ[x]+t+r dr (3.2.7)
0
3.3. COMMUTATION FUNCTIONS 39
Commutation functions are numerical devices (developed by Griffith Davies and others) which allow
the calculation of certain common assurance (and annuity) values at a specified rate of interest from
a small number of columns, labelled Dx , M̄x etc. In view of their importance in the Tables for
Actuarial Examinations we shall look at them in detail.
At this point we consider only whole life assurances payable immediately on death. Define
Dx = v x l x (3.3.1)
Z 1
C̄x = v x+t lx+t µx+t dt (3.3.2)
0
∞
X
M̄x = C̄x+t (3.3.3)
t=0
Now
R∞
v x+t lx+t µx+t dt
0
Āx =
v x lx
C̄x + C̄x+1 + C̄x+2 + . . .
=
Dx
Z r+1 Z 1
x+t
(on writing v lx+t µx+t dt = v x+r+u lx+r+u µx+r+u du
r 0
by the change of variable u = t − r)
so that P∞
C̄x+t M̄x t=0
Āx =
= (3.3.4)
Dx Dx
We shall consider commutation functions again in relation to temporary and deferred assurances,
etc.
The standard deviation of Z is, of course, found by taking the square root of var(Z). If we consider
a block of n identical whole life policies on independent lives aged x, the total present value of the
assurance benefits is
Z = Z1 + Z2 + . . . + Zn
where, for i = 1, 2, . . . , n,
with T (i) = future lifetime of ith life. Since the variables {T (i) } are assumed to be independent, the
variance of Z is
n
X
Var(Z) = Var(Zi )
i=1
Xn h ∗ ¡ ¢ i
2
= S 2 Āx − Āx
i=1
h ∗ ¡ ¢ i
2
= nS 2 Āx − Āx (3.4.4)
We remark that the mean present value of a group of policies is the sum of their separate M.P.V’s
even if the lives are not independent. Formulae 3.4.4 and 3.4.5 may be generalised to cover the case
when the sums assured and/or ages of the lives are different (see exercise 3.3).
Jensen’s inequality
Let g(t) be convex, which is the case of g 00 (t) ≥ 0 for all t > 0. Jensen’s inequality shows that
By Jensen’s inequality,
E(v T ) ≥ v E(T )
but
◦
E(T ) = ex
and
E(v T ) = Āx
Example 3.4.1. Verify formula 3.4.7 for x = 50 on the basis of E.L.T. No.12-Males, 4% p.a. interest.
◦
Solution. e50 = 22.68
◦
so v e50 = 0.41085
P r{K = k} = k |qx (k = 0, 1, 2, . . . )
Z = g(K) = Sv K+1
which has mean
∞
X
E[g(K)] = Sv k+1 k |qx (3.5.1)
k=0
= SAx
42 CHAPTER 3. ASSURANCES
where
∞
X
Ax = v k+1 k |qx (3.5.2)
k=0
Cy = v y+1 dy (3.5.3)
X∞
My = Cy+k (3.5.4)
k=0
∗
where indicates at the rate of interest 2i + i2 .
Solution.
which follows from the mean value theorem for integrals. Formulae (3.3.3) and (3.5.4) now show
that
1
M̄x + (1 + i) 2 Mx (3.5.9)
3.5. ASSURANCES PAYABLE AT THE END OF THE YEAR OF DEATH. 43
and hence
1
M̄x (1 + i) 2 Mx 1
Āx = + = (1 + i) 2 Ax (3.5.10)
Dx Dx
Theorem 3.5.1. If there is a uniform distribution of deaths between the ages x + k and x + k + 1
(for k = 0, 1, 2, . . . ), we have
i
Āx = Ax (3.5.11)
δ
£P∞ ¤± P∞
Proof. It is sufficient (since Āx = k=0 C̄x+k Dx and Ax = [ k=0 Cx+k ]/ Dx ) to show that
i
C̄y = Cy for y = x + k (k = 0, 1, 2, . . . )
δ
Now, since we have U.D. of D. between ages y and y + 1,
Therefore
Z 1
C̄y = v y+t ly+t µy+t dt
0
Z 1
y
= dy v v t dt
µ0 ¶
y 1−v
= dy v
δ
µ ¶
i
= dy v y+1
δ
i
= Cy
δ
and
1 1
(1 + i) 2 + 1 + i
2
so we sometimes find the following approximations (which are suitable only when i is small):
1
Āx = (1 + i)Ax
2
1
C̄x = (1 + i)Cx (3.5.12)
2
1
M̄x = (1 + i)Mx
2
44 CHAPTER 3. ASSURANCES
Figure 3.6.1:
The mean of this variable is written as S Āx1:n , the “1” indicating that the status (x) must fail
before the status n (a fixed period of n years) for a payment to be due (and that the payment is
due when this event occurs). Thus
Z n
Āx1:n = v t t px µx+t dt (3.7.1)
0
This may be evaluated by numerical integration or by commutation functions, i.e.
n−1
X
C̄x+t
t=0 M̄x − M̄x+n
Āx1:n = = (3.7.2)
Dx Dx
3.7. TEMPORARY AND DEFERRED ASSURANCES 45
n−1
X
Ax1:n = v t+1 t |qx (3.7.3)
t=0
1 i 1
Āx1:n ; (1 + i) 2 Ax1:n ; A (3.7.5)
δ x :n
A deferred assurance provides a sum of £S (say) on the death of (x) if this occurs after a certain
period, called the deferred period (or, more correctly, the period of deferment), which we usually
write as n years. If the benefit is payable immediately on death of (x) after the deferred period has
elapsed, the present value is
½
Sv T if T > n
Z = g(T ) =
0 if T < n
if the sum assured is payable immediately on death, or
½
Sv K+1 if K ≥ n
Z = g(K) =
0 if K < n
if the sum assured is payable at the end of year of death.
The corresponding M.P.V.’s are written as
and
∞
X
n |Ax = v t+1 t| qx (3.7.10)
t=n
X∞
Cx+t
t=n
=
Dx
Mx+n
= (3.7.11)
Dx
v n n px
(since P r{T ≥ n} = n px ). This M.P.V. is written as
n Ex or Ax:n1 or Āx:n1
i.e.
n Ex = Ax:n1 = Āx:n1 = v n n px (3.8.1)
In terms of commutation functions we have the useful result that
Dx+n
n Ex = (3.8.2)
Dx
Returning to the formulae (3.7.6) and (3.7.7) , we see that the M.P.V. of a deferred assurance may be
written as the product of n Ex and Āx+n or Ax+n . That is, the value is obtained by multiplying the
M.P.V. of the assurance at the “vesting date” (when the life attains age x + n) by a pure endowment
factor.
We finally consider an endowment assurance of term n years with sum assured £S. In this
policy, the sum assured is payable if (x) dies within n years or on maturity of the contract at time
3.8. PURE ENDOWMENTS AND ENDOWMENT ASSURANCES 47
n years, whichever occurs first. It follows that an endowment assurance is a combination of a term
assurance and a pure endowment (of the same term). If we assume that the death benefit is payable
immediately on death, if death occurs within n years, the present value of the benefits is
½
Sv T if T < n
Z = g(T ) =
Sv n if T ≥ n
The mean of Z is written S Āx:n .Thus
In view of the fact that the term Dx+n occurs in formulae (3.8.4) and (3.8.6), it is not correct to
write
1 A COMMON
Āx:n ; (1 + i) 2 Ax:n
MISTAKE
and
1
(1 + i) 2 (Mx − Mx+n ) + Dx+n
Āx:n ; (3.8.8)
Dx
Select tables The adjustments to the formulae when a select table is used are straightforward: one
need only replace ‘x + t’ by ‘[x] + t’ (or ‘[x]’ if the life assured has just been selected) and dispense
with [ ] if the duration since selection is at least equal to the select period, s years, and in compound
interest functions.
For example,
C[x]+t = v x+t+1 d[x]+t
Example 3.8.1. On the basis of A1967-70 select mortality and 4% p.a. interest, calculate the mean
present value of each of the following assurance benefits for a life aged 30:
48 CHAPTER 3. ASSURANCES
(i) A whole life assurance for £10, 000, payable immediately on death;
(ii) A 20-year term assurance for £50, 000, payable at the end of the year of death;
(iii) A 20-year endowment assurance for £50, 000, with the death benefit payable immediately on
death;
(iv) A deferred temporary assurance for £100, 000, payable at the end of the year of death, if death
occurs between ages 40 and 50 exactly.
1
[Use the factor (1 + i) 2 for accelerating payments from end of year of death to the moment of death.]
Solution.
1
(i) 10, 000Ā[30] = 10, 000(1.04) 2 A[30]
= £1, 934.87
(M[30] − M50 )
(ii) 50, 000 = £1, 012.90
D[30]
1 D50
(iii) (1.04) 2 × 1, 012.90 + 50, 000 = £23, 070.56
D[30]
(M40 − M50 )
(iv) 100, 000 = £1, 360.91
D[30]
If β(t) = [t] + 1 , where [t] indicates the integer part of t, the M.P.V. is written as (I Ā)x , and it may
easily be shown that
∞
X
(t + 1)C̄x+t
t=0
(I Ā)x =
Dx
½ ¾Á
£ ¤ £ ¤
= C̄x + C̄x+1 + · · · + C̄x+1 + C̄x+2 + · · · + · · · Dx
M̄x + M̄x+1 + · · ·
=
Dx
R̄x
= (3.9.3)
Dx
where
∞
X
R̄x = M̄x+k (3.9.4)
k=0
In view of the fact that the average benefit is 12 less when β(t) = t than in the present case, we have
the approximation
1
(I¯Ā)x ; (I Ā)x − Āx (3.9.5)
2
We now consider the case when a death benefit of β(k) in year k is payable at the end of the year
of death. The present value is
where
∞
X
Rx = Mx+k (3.9.9)
k=0
50 CHAPTER 3. ASSURANCES
1
In view of the approximations C̄x ; (1 + i) 2 Cx , etc., we find that
i 1
R̄x ; Rx ; (1 + i) 2 Rx (3.9.10)
δ
and
i 1
(I Ā)x ; (IA)x ; (1 + i) 2 (IA)x (3.9.11)
δ
We may also define similar symbols for increasing temporary and endowment policies, e.g.
n−1
X
(IA)x1:n = (k + 1)v k+1 k |qx
k=0
n−1
X
(k + 1)Cx+k
k=0
=
Dx
½
= [Cx + 2Cx+1 + · · · ] − [Cx+n + 2Cx+n+1 + · · · ]
¾Á
− n [Cx+n + Cx+n+1 + · · · ] Dx
Rx − Rx+n − nMx+n
= (3.9.12)
Dx
and
Nowadays one may also encounter unitised with profit policies. These are not the same as
unit-linked policies: the maturity proceeds of the latter are linked to the performance of a unit
trust or internal fund of the life office, while for U.W.P. policies the office credits each policy with
internal units which are subject to certain guarantees.
In this book we shall study reversionary bonuses only. There are 3 systems of adding bonuses
in common use:
(1) the simple bonus system;
(2) the compound bonus system, and
(3) the two-tier (or supercompound) bonus system.
3.10. VALUING THE BENEFITS UNDER WITH PROFITS POLICIES 51
(1) Simple bonuses. In this system, bonuses are calculated only on the basic sum assured (B.S.A.),
which we shall denote by S. Consider a whole life with profits policy issued t years ago to a life
then aged x. Suppose that bonuses vest annually in advance, let the bonuses already vested be B,
and suppose that a bonus is about to be added. the formula used to calculate new bonus addition
is
new bonus = S × rate of bonus p.a. (3.10.1)
To value the benefits we must estimate the rate of simple bonus p.a. (applying now and for the
forseeable future) as b (say). The total sum assured is thus
S + B + bS in year 1 from the present time,
S + B + 2bS in year 2 from the present time,
and so on, giving a total benefit in year k of
S + B + kbS
(3.10.2)
If the benefits are payable at the end of the year of death, the M.P.V. of the benefits is
∞
X
[S + B + (k + 1)bS] v k+1 k |qx+t
k=0
=(S + B)Ax+t + bS(IA)x+t (3.10.3)
at end of year
: (S + B)Ax+t:n−t + bS(IA)x+t:n−t
of death (3.10.5)
immediately
: (S + B)Āx+t:n−t + bS(I Ā)x+t:n−t
on death (3.10.6)
(2) Compound bonuses In this system, the formula for bonus additions is
basic sum
assured
new bonus = plus bonuses × rate of bonus p.a. (3.10.7)
already
vesting
Assuming a bonus rate of b p.a. applies now and for the forseeable future, the benefits will be
(S + B)(1 + b) in year 1
2
(S + B)(1 + b) in year 2
and so on, giving
(S + B)(1 + b)k in year k
(3.10.8)
52 CHAPTER 3. ASSURANCES
If the benefits are payable at the end of the year of death, their M.P.V. is
∞
X
(S + B)(1 + b)k+1 v k+1 k |qx+t
k=0
∞ µ
X ¶k+1
1+b
=(S + B) k |qx+t
1+i
k=0
=(S + B)A∗x+t (3.10.9)
(S + B)A∗x+t:n−t (3.10.12)
and h i
1
(S + B) (1 + i) 2 A∗ x+t
1
:n−t + A ∗ 1
x+t:n−t (3.10.13)
Notes. (1) In formula (3.10.13), we observe that the pure endowment part is as in formula
(3.10.12): only the term assurance part is “accelerated”
(2) At inception, we have t = 0 and B = 0.
Let us assume that the bonus rate p.a. in respect of the B.S.A. and previous bonus additions are
a and b respectively (now and for the forseeable future).
Let B(k) = total bonus in year k (from the present time), and let us define B(0) = B. We have
B(1) = B(1 + b) + aS
B(2) = B(1 + b)2 + aS[1 + (1 + b)]
the result is true for k = 0. Now assume formula (3.10.15) is true for B(k + 1), k being some value.
We have
It follows that, in this bonus system, the M.P.V. of a whole life assurance is
∞
X
[S + B(k + 1)]v k+1 k |qx+t
k=0
∞ ½
X · ¸ ¾
(1 + b)k+1 − 1
= S + aS + B(1 + b)k+1 k |qx+t
b
k=0
∞ n
³
X a´ ³a ´ o
= S 1− + S + B (1 + b)k+1 k |qx+t
b b
k=0
³ a ´ ³ a ´
=S 1 − Ax+t + S + B A∗x+t (3.10.16)
b b
i−b
where A∗x+t is at rate
1+b
Notes. (1) If the sum assured is payable immediately on death, the M.P.V. is approximately
³ a´ ³a ´ 1
S 1− Āx+t + S + B (1 + i) 2 A∗x+t
b b (3.10.17)
Example 3.11.1. A company issues 20-year endowment assurances each with a basic sum assured
of £1, 000 to male lives aged 45.
Guaranteed simple reversionary bonuses at the rate of 2.25% of the current sum assured vest on pay-
ment of each annual premium. Alternatively, at the outset of a policy, a life assured may elect that
compound reversionary bonuses should be added to the policy instead of simple reversionary bonuses.
The sum assured and added bonuses will be payable at maturity, or at the end of the year of
death, if earlier.
Derive an expression from which can be calculated the guaranteed rate of compound bonus the
company should offer so that the value of the benefits at the outset is the same.
Solution. Let b be the annual rate of guaranteed compound bonus. The equation of value for b is:
Exercises
3.3 Consider n lives now aged x1 , x2 , . . . , xn respectively. Let Z be the total present value of a
contract providing the sum of £Si immediately on death of (xi ), i = 1, 2, . . . , n. The n lives
are subject to the same non-select mortality table, Table B, and the interest is taken to be
fixed at rate i p.a.
(i) Show that E(Z) = S1 Āx1 + · · · + Sn Āxn on Table B with interest at rate i p.a.
(ii) Assuming further that the future lifetimes T (xi ) of the lives are independent variables,
show that
Xn
Var(Z) = Sj2 [Ā∗xj − (Āxj )2 ]
j=1
3.4 Using commutation functions or otherwise calculate the values of the following:
3.5 A life aged 50 who is subject to the mortality of the A1967-70 Select table, effects a pure
endowment policy with a term of 20 years for a sum assured of £10,000.
(i) Write down the present value of the benefits under this contract, regarded as a random
variable.
(ii) Assuming an effective rate of interest of 5% per annum, calculate the mean and the
variance of the present value of the benefits available under this contract.
3.6 What are the random variables (in terms of K = curtate future lifetime of (x)) whose means
are represented by the following symbols?
(i) n Ex
(ii) Ax1:n
(iii) Ax:n1
3.7 A life aged exactly 60 wishes to arrange for a payment to be made to a charity in 10 years’
time. If he is still alive at that date the payment will be £1000. If he dies before the payment
date, the amount given will be £500. Assuming an effective interest rate of 6% per annum and
56 CHAPTER 3. ASSURANCES
mortality according to ELT No.12-Males, calculate the standard deviation of the present value
of the liability.
3.8 (Difficult)
You are given that
Calculate 1000(IA)51 .
3.13. SOLUTIONS 57
Solutions
1
3.1 (i) (a) Ā60 = (1 + 0.04) 2 A60
1
= (1.04) 2 0.51726
= 0.52750
(b) A60 = 0.51726 from tables
∞
X
= qx v + px k−1 |qx+1 v k+1
k=1
∞
X
= qx v + px v k |qx+1 v k+1
k=0
= qx v + px vAx+1
1
(ii) A61 = (0.02 + 0.98 · 0.6) = 0.57905
1.05
1
A62 = (0.015 + 0.985 · 0.57905) = 0.55749
1.05
3.3 (i)
We have
n
X
Z= Sj v T j
j=1
so
n
X n
X
E(Z) = Sj E(v Tj ) = Sj Āxj
j=1 j=1
1
1 1
(iii) Ā30 :20 ; (1 + i) 2 A30:20 = 0.020956
1 D50
(iv) Ā30:20 = Ā30 :20 + = 0.46157
D30
1
Ā30 :20 1 M̄30 − M̄50
(v) 1
A30 :20 ; 1 = (1 + i)− 2 = 0.03286
(1 + i) 2 D30
3.5 (i) ½
10, 000v 20 if T ≥ 20
P.V. =
0 if T < 20
3.6 (i) ½
vn if K ≥ n
Z=
0 if K < n
(ii) ½
0 if K ≥ n
Z=
v K+1 if K < n
3.8
ANNUITIES
Z ∞
āx = vt t px dt (4.1.3)
0 | {z } | {z }
interest (or probability
discounting) factor of survival
of (x) for t years
To show that these definitions are equivalent, we may use integration by parts, as follows:
Z ∞
āt t px µx+t dt
0 | {z }
| {z }
Z t d d
āt = v r dr (−t px ) = (t qx )
0
dt dt
d = t px µx+t
So (ā ) = v t
dt t
61
62 CHAPTER 4. ANNUITIES
Z ∞ Z ∞
£ ¤∞
āt t px µx+t dt = āt (−t px ) 0 + v t t px dt
0 0
Z ∞
= v t t px dt
0
Note the following conversion relationship (i.e. a formula connecting assurance and annuity
values):
Āx = 1 − δāx (4.1.4)
which follows from:
Z ∞
āx =
āt t px µx+t dt
0
∞µ Z ¶
1 − vt
= t px µx+t dt
0 δ
1
= (1 − Āx ).
δ
We may also argue as follows. By the Mathematics of Finance,
Solution.
Ā∗x = 1 − (2δ)ā∗x since force of interest is 2δ
so, by formula (4.1.5),
1 − 2δ(ā∗x ) − (1 − δāx )2
Var(āT ) =
δ2
1 − 2δ(ā∗x ) − 1 − δ 2 (āx )2 + 2δāx
=
δ2
1£ ¤
= 2(āx − ā∗x ) − δ(āx )2
δ
4.2. ANNUITIES PAYABLE ANNUALLY 63
Z t+1
but (on change of variable) D̄x+t = v x+r lx+r dr for t = 0, 1, 2, . . ., so
t
Z ∞
N̄x = v x+r lx+r dr
0
äx = 1 + ax VITAL!
(4.2.3)
(since the annuitant gets extra immediate payment of 1 in case (i)).
Now
¡ ¢
äx =E äK+1
∞
X
= äk+1 k |qx (4.2.4)
k=0
But
1 − v K+1
äK+1 = (K = 0, 1, 2 . . . )
d
64 CHAPTER 4. ANNUITIES
so
· ¸
1 − v K+1 1 − E(v K+1 )
äx =E =
d d
1 − Ax
= (4.2.5)
d
We thus have the conversion relationship
Ax = 1 − däx (4.2.6)
Two Proofs
(1) Regard the annuity as sum of pure endowments due at times 0,1,2,. . .. Thus
äx = 0 Ex + 1 Ex + 2 Ex . . .
X∞
= v t t px
t=0
as required.
We obtain
v x lx + v x+1 lx+1 + . . .
äx =
v x lx
Dx + Dx+1 + . . .
=
Dx
Nx
=
Dx
Remark If we require ax , use ax = äx − 1 or
Nx+1
ax = (since Nx+1 = Dx+1 + Dx+2 + . . . )
Dx
The variance of äK+1 .
We use the formula
µ ¶
¡ ¢ 1 − v K+1
Var äK+1 = Var
d
1 ¡ K+1 ¢
= 2 Var v
d
A∗ − (Ax )2
= x 2 (4.2.9)
d
where ∗ indicates the rate of interest i∗ = 2i + i2 p.a.
Note For an annuity payable in arrear, we use the result that
¡ ¢
Var (aK ) = Var äK+1 − 1
¡ ¢
= Var äK+1 (4.2.10)
Commutation Functions.
If n is an integer, Rn
0
v x+t lx+t dt N̄x − N̄x+n
āx:n = x
= (4.3.3)
v lx Dx
66 CHAPTER 4. ANNUITIES
If payments are made annually in advance, and are limited to at most n payments, we obtain
(
äK+1 if K < n h i
äx:n = mean of = E ämin{K+1,n} (4.3.4)
än if K ≥ n
n−1
X
= äk+1 k |qx + än n px
k=0
(i.e., the sum of the values of n pure endowments each for £1)
Dx + Dx+1 + · · · + Dx+n−1
=
Dx
Hence
Nx − Nx+n
äx:n = (4.3.5)
Dx
If the payments are made annually in arrear, we obtain
(
aK if K < n h i
ax:n = mean of = E amin{K,n} (4.3.6)
an if K ≥ n
leading to
Dx+1 + Dx+2 + · · · + Dx+n Nx+1 − Nx+n+1
ax:n = = (4.3.7)
Dx Dx
Note that
(a) äx:n = 1 + ax:n−1 (4.3.8)
(b) äx:n and ä[x]:n are tabulated for certain values of x+n in the A1967-70 section of “Formulae
and Tables.”
Example 4.3.1. By expressing Ax:n and äx:n as expectations of appropriate random variables, or
otherwise, prove the conversion relationships
Proof.
1 − v min{K+1,n}
(i) By Maths. in Finance, ämin{K+1,n} = .
d
Take expected values to obtain
1 − Ax:n
äx:n =
d
which gives the required result.
(ii) Take expected values in the equation
1 − v min{T,n}
āmin{T,n} = .
δ
Select Tables The modifications needed for select tables are straightforward, and are illustrated in
Table 4.4.1 below.
68 CHAPTER 4. ANNUITIES
In terms of Commutation
Type of Annuity Symbol In terms of t p[x] Functions
∞
X N[x]+1
Immediate a[x] v t t p[x] D[x]
t=1
∞
X N[x]
Annuity-due ä[x] v t t p[x]
t=0
D[x]
n
X N[x]+1 − Nx+n+1
Temporary (n years) a[x]:n v t t p[x] (n ≥ 1)
D[x]
t=1
n−1
X
Temporary annuity-due N[x] − Nx+n
ä[x]:n v t t p[x] (n ≥ 2)
(n years’ payments) D[x]
t=0
∞
X Nx+m+1
Deferred (m years) v t t p[x] (m ≥ 1)
m |a[x] D[x]
t=m+1
∞
X
Deferred annuity-due Nx+m
m |ä[x] v t t p[x] (m ≥ 2)
(m years) D[x]
t=m
Table 4.4.1: Expression for Values of Single-Life Curtate Annuities. Life aged x.
Select Mortality Table (select period 2 years)
Example 4.5.1. Use the Euler-Maclaurin formula to deduce an approximate formula for āx in terms
of äx
4.5. ANNUITIES PAYABLE M TIMES PER ANNUM 69
³ R ´
t
Solution. Let f (t) = v t t px = exp − 0 (δ + µx+r ) dr . Hence
³ R ´
t
f 0 (t) = −(δ + µx+t ) exp − 0 (δ + µx+r ) dr . ∴ f (0) = 1 and f 0 (0) = −(δ + µx ). By E.-M.
1 1 1 1
āx + äx − − (µx + δ) or ax + − (µx + δ) (4.5.3)
2 12 2 12
In practice the final term is usually ignored, giving
1 1
āx + äx − = ax + (4.5.4)
2 2
◦
Example 4.5.2. Find an approximate formula for ex in terms of ex .
(m)
Example 4.5.3. Find an approximate formula for äx in terms of äx .
Define
(m)
äx:n = M.P.V. of £1 p.a. payable mthly in advance to (x)
for at most n years
= ä(m)
x − n |ä(m)
x (4.5.12)
D x+n (m)
= ä(m)
x − ä
Dx x+n
· µ ¶¸ µ ¶· µ ¶¸
m−1 Dx+n m−1
+ äx − − äx+n −
2m Dx 2m
µ ¶ µ ¶
m−1 Dx+n
= äx:n − × 1− (4.5.13)
2m Dx
| {z }
CARE!
DO NOT OMIT
THIS TERM
We also have µ ¶µ ¶
m−1 Dx+n
a(m)
x:n
= ax:n + 1− (4.5.14)
2m Dx
◦ (m)
ax = M.P.V. of complete annuity
1
+ a(m)
x + Āx (4.6.1)
2m
◦ (m) ◦
When m = 1 we may write ax =ax , giving
◦ 1
ax + ax + Āx (4.6.2)
2
◦
More accurate formulae for ax . If we have U.D. of D in each of the age-ranges x to x + 1, x + 1
to x + 2, etc., we have the exact formulae
µ ¶ µ ¶
◦ i−δ i−δ
ax = ax + Ax = ax + Āx (4.6.3)
δ2 iδ
Proof. The final payment is shown below: Hence (using the formula for a varying assurance)
4.7. VARYING ANNUITIES 71
final
payment
(on death of (x) ) 6
1
· · · ·
· · · ·
· · · ·
· · · · .
· · · · ..
.
· · · · ..
· · · ·
· · · · ··
· · · · · - time (years)
0 1 2 3 4
∞
X Z 1
◦ t
ax = ax + v t px rv r r px+t µx+t+r dr
t=0 | {z } 0 | {z }
pure endowment value at time t
factor to age x + t of death benefit
in year t + 1
By U.D. of D.,
∞
X Z 1
◦
ax = ax + v t t px qx+t rv r dr
t=0 0
| {z }
¯ ā −v
= (Iā) 1 =
1
δ
X∞ · ¸
t+1 ā1 − v
= ax + v t |qx (1 + i)
t=1
δ
µ ¶ µ ¶
s̄1 − 1 (1 + i) − 1
= ax + Ax = ax + Ax − 1 /δ
δ δ
µ ¶
i−δ
= ax + Ax
δ2
The second part of (4.6.3) follows from Āx = δi Ax (which holds by U.D.D.)
Z ∞ µZ t ¶ Z ∞ µZ ∞ ¶
u(t) v(r) dr dt = v(t) u(r) dr dt (∗)
0 0 0 t
¯
Solution. Let b(t) = t, t ≥ 0. This gives g(T ) = (Iā) T , and we have
¯ x = M.P.V. of annuity to (x) with payment at rate t p.a. at time t
(Iā)
Z ∞
= tv t t px dt (4.7.4)
0
where
∞
X
S̄x = N̄x+t (4.7.7)
t=0
Now suppose that benefits are paid annually so long as (x) survives, the benefit at time t years being
b(t) (t = 0, 1, 2, . . . ). The present value of the benefits (regarded as a random variable) is
K
X
g(K) = v t b(t) (4.7.8)
t=0
since the last payment is made at time K. (If b(0) > 0, this is a variable annuity-due). The M.P.V.
of the varying annuity is thus
∞
X
E[g(K)] = g(k)k |qx
k=0
1
= {g(0)dx + g(1)dx+1 + . . . }
lx
1
= {g(0)[dx + dx+1 + . . . ]
lx
+ [g(1) − g(0)][dx+1 + dx+2 + . . . ]
+[g(2) − g(1)][dx+2 + dx+3 + . . . ] + . . . }
1 © ª
= b(0)lx + vb(1)lx+1 + v 2 b(2)lx+2 + . . .
lx
∞
X
E[g(K)] = b(t) vt t px (4.7.9)
t=0 | {z } | {z } | {z }
benefit interest survival
at time t factor factor
Applications
Let b(t) = t + 1 (t = 0, 1, 2, . . . ). The M.P.V. is defined as (Iä)x , so we have
∞
X
(Iä)x = (t + 1)v t t px
t=0
" ∞
#
X
= (t + 1)Dx+t /Dx
t=0
Dx + 2Dx+1 + . . .
=
Dx
(Dx + Dx+1 + . . . ) + (Dx+1 + Dx+2 + . . . ) + . . .
=
Dx
Nx + Nx+1 + . . .
=
Dx
Sx
= (4.7.10)
Dx
where
∞
X
Sx = Nx+t (4.7.11)
t=0
∞
X X∞
t Dx+t
(Ia)x = tv t px = t
t=1 t=1
Dx
Sx+1
= (4.7.12)
Dx
Solution.
∞
X
(Ia)x = tv t t px
t=0
∞
X
= (t + 1 − 1)v t t px
t=0
∞
X ∞
X
= (t + 1)v t t px − v t t px
t=0 t=0
=(Iä)x − äx
Define
(Iä)x:n = M.P.V. of payments of t + 1 at time t
(t = 0, 1, 2, . . . , n − 1), provided that (x) is
then alive "n−1 #
n−1
X X
t
= (t + 1)v t px = (t + 1)Dx+t /Dx
t=0 t=0
Dx + 2Dx+1 + · · · + (n − 1)Dx+n−2 + nDx+n−1
=
Dx
(Dx + 2Dx+1 + . . . ) (Dx+n + 2Dx+n+1 + . . . )
= −
Dx Dx
(nDx+n + nDx+n+1 + . . . )
−
Dx
Sx − Sx+n − nNx+n
= (4.7.13)
Dx
Similarly,
Dx+1 + 2Dx+2 + · · · + nDx+n
(Ia)x:n =
Dx
(Dx+1 + 2Dx+2 + . . . ) (Dx+n+1 + 2Dx+n+2 + . . . )
= −
Dx Dx
(nDx+n+1 + nDx+n+2 + . . . )
−
Dx
Sx+1 − Sx+n+1 − nNx+n+1
=
Dx
Also,
Z n
¡ ¢
¯
Iā = tv t t px dt
x:n
0
S̄x − S̄x+n − nN̄x+n
(Iā)x:n = (4.7.14)
Dx
Note that (for example) µ ¶
Dx+m
m | (Iä)x:n = (Iä)x+m:n
Dx
(∗∗) holds for t = 0, 1, 2, 3, . . . by McCutcheon and Scott, formula 3.6.6. It may be proved for general
t ≥ 0 by letting t = n + r ( n integer, 0 ≤ r < 1) and observing that
Z n+r
(Iā)t = (Iā)n + v s (n + 1) ds
n
µ ¶
än − nv n 1 − vr
= + (n + 1)v n
δ δ
t
ä − (n + 1)v
= n+1
δ
Note. There are also conversion relationships for temporary increasing annuities, e.g.
¯ x:n + (I¯Ā)x:n
āx:n = δ(Iā)
4.8. EXERCISES 77
Exercises
4.1 The following is an extract from a life table with a select period of 1 year.
Age x l[x] lx+1 age x + 1
55 90,636 90,032 56
56 89,739 89,132 57
88,151 58
87,094 59
85,874 60
84,586 61
4.2 Given that ax = 20, ax:n = 18 and ax+n = 8, find the values of n Ex and äx:n .
4.5 (i) Find the present value of a deferred annuity of £1000 p.a. to a man aged 40. Payments
commence on his 60th birthday, if he is then alive, and continue thereafter annually for
life.
Basis: A1967-70 ultimate mortality, 4 % p.a. interest.
(ii) As above, but select mortality (at entry.)
(iii) As above, but A1967-70 ultimate mortality to exact age 60 and a(55) males ultimate
mortality above exact age 60.
4.6 According to a certain mortality table, which has a select period of 1 year and is such that
q[x] = 0.6qx for each x,
a70 at 10% interest = 5.641, and,
a71 at 10% interest = 5.449.
Solutions
4.1
l56 + vl57 + · · · + v 4 l60
ä56:5 = = 4.451
l56
l[56] + vl57 + · · · + v 4 l60
ä[56]:5 = = 4.463
l[56]
4.2
1
ax = ax:n + n Ex ax+n =⇒ n Ex =
4
äx:n = ax:n − n Ex + 1 = 18.75
i.e.
(1 + i)a45 15.719
p45 = = 1.0425 × = .9926136
ä46 16.509
Hence
q45 = 0.0073864
Then
q[45] = 0.6q45 = 0.0044318
Hence
p[45] = 1 − q[45] = 0.9955682
From (a) we then get
a[45] = 15.7658
4.4
ä[40]:30 = 16.960 directly tabulated
ä∗60 is evaluated on a(55) males mortality. The value is thus 1000 × .40873 × 12.625 =
5160.22, say £5160.
4.6 a70 = v(1 − q70 )ä71 , so 1 − q70 = 0.96218. Hence q[70] = 0.6q70 = 0.022918, and q[70] =
v(1 − q[70] )ä71 = 5.730
80 CHAPTER 4. ANNUITIES
Chapter 5
PREMIUMS
Premiums are usually calculated by the equivalence principle, which may be stated as follows:
E(Z) = 0 (5.1.1)
where
Z = present value of profit to the life office on the contract
In some cases an explicit loading for profit is included in the calculation; that is, we replace equation
(5.1.1) by
E(Z) = B (5.1.2)
where B = M.P.V. of profit on the contract. In the absence of information to the contrary, however,
we shall assume that the equivalence principle (5.1.1) is to be used, although in some cases the use
of “conservative” assumptions regarding mortality, interest and expenses means that the office has
an implicit margin of expected profit. The rate of interest is usually taken to be fixed (not random).
Example 5.1.1. Consider a whole life assurance of 1 payable immediately on the death of (x),
and suppose that premiums are payable continuously at rate P p.a. until the death of (x). Give a
formula for P in terms of annuity functions, assuming that the equivalence principle applies.
Solution. Let
Z = g(T ) = P.V. of profit to office on this contract
= P āT − v T
81
82 CHAPTER 5. PREMIUMS
E{P āT − v T } = 0
¡ ¢
∴ P.E (āT ) = E v T
∴ P āx = Āx
Āx 1
∴ P = = −δ using the conversion relationship Āx = 1 − δāx
āx āx
The equations (5.1.1) and (5.1.2) may be expressed as equations of payments, or equations of
value, i.e.
P āx = Āx
Some of the more important rules of the International Notation are the following:
1. The symbols P ( ), or P̄ ( ), or P (m) ( ) indicate the level net annual premium for the benefit
indicated in the brackets. Premiums are assumed to continue for as long as the contract can
provide benefits, i.e. n years for a n-year contract, provided of course that the life assured is
still alive. P, P̄ , P (m) indicate that payments are made annually in advance, continuously
and mthly in advance respectively.
2. If premiums are limited to, at most t years’ payments (where t < the term of contract, i.e. the
term of the benefits), we write t P ( ), t P̄ ( ), t P (m) ( ).
e.g.
3. When benefits are payable at end of year of death we may (optionally) shorten the notation
as follows:
(e.g.) P (Ax ) = Px
5.3. THE VARIANCE OF THE PRESENT VALUE OF THE PROFIT ON A POLICY. 83
Table 5.2.1 relates to a mortality table with select period 2 years, for example A1967-70, and refers
to policies in which premiums are payable annually in advance.
(We need not assume the equivalence principle holds: if it does, E[g(T )] = 0.) Note that
µ ¶
1 − vT
g(T ) = P − vT
δ
µ ¶
P P +δ
= − vT
δ δ
µ ¶2
P +δ
∴ Var[g(T )] = Var(v T )
δ
µ ¶2
P +δ £ ∗ ¡ ¢2 ¤
= Āx − Āx (5.3.1)
δ
| {z }
at rate
2i + i2
Note the idea of “collecting” all the terms involving v T before taking the variance: a similar technique
may be used for endowment assurance policies (but not generally).
Example 5.4.1. Consider n-year endowment assurance without profits with sum assured £1 issued
to a select life aged x. Expenses are e per premium payment (including the first) plus I at issue
date (so the total initial expense is I + e ). Find formulae for the level office annual premium, P 00 .
CHAPTER 5. PREMIUMS
where P 00 = level annual premium. Suppose further that e = kP 00 (i.e. renewal expenses are a
proportion k of the office premium): we have
(For example, k = 2 21 % and c = 47 12 % if expenses are 50% of the first premium and 2 12 % of all
subsequent premiums .)
Example 5.5.1. A with-profits whole-life assurance is about to be issued to a man aged 45. The
basic sum assured is £12, 000. The office assumes that at the start of each policy year there will be
bonus additions at the rate of 1% per annum compound. The basic sum assured and bonuses are
payable immediately on the man’s death. The policy has half-yearly premiums, payable for at most
20 years. Calculate the half-yearly premium on the following basis: (Note: at 3.75% per annum
interest, the value of A[45] is 0.34587).
i − 0.01
where ∗ is at rate + 3.96%
1 + 0.01
Equation of value is
(2) (2)
0.96P ä[45] :20 = 0.16P ä[45] :1 + 200 + 4033.69
| {z } | {z }
12.362 0.98753
∴ P = £361.58
Hence the half-yearly premium is £180.79
Now suppose that the premiums are returned with compound interest at rate j p.a. In practice,
such policies are often described as “ Return With Interest” (R.W.I.). If the death benefit is paid
at the end of the year of death, the term P (IA)x1:n in equation (5.6.1) must be replaced by
n−1
X n o
v k+1 k |qx P s̈k+1 j
k=0
| {z }
death benefit
in year k + 1
X · (1 + j)k+1 − 1 ¸
n−1
=P v k+1 k |qx
dj
k=0
P
= . {A∗ x1:n − Ax1:n } (5.6.2)
[j/(1 + j)]
i−j
where ∗ indicates the rate of interest 1+j
Notes. 1. If the death benefit is payable immediately on death, formula (5.6.2) should be multi-
1 1 1
plied by (1 + i) 2 /(1 + j) 2 = (1 + i∗ ) 2 .
2. If i (the rate of interest used to discount premiums and benefits in the equation of
value) equals j, some of these problems may be simplified: if there are no expenses,
mortality (within n years) may be ignored giving the compound interest equation
P än = 1000v n
5.7. ANNUITIES WITH GUARANTEES 87
This result may best be proved by reference to reserves, which we shall discuss later.
Example 5.6.1. A life office sells policies each providing a cash sum at age 65. Premiums of £1, 000
are payable annually in advance during the deferred period. On the death of the policyholder during
the period of deferment, the premiums paid are returned immediately without interest. In respect
of a life now aged 45, find the cash sum at age 65, given that the office uses the following basis:
A1967-70 select
4% p.a. interest
expenses are ignored
Let us consider an annuity of £1 p.a., payable continuously with this guarantee to a life aged x
at the issue date. Ignoring expenses, the equation of value for B is
Hence
f 0 (B) = v B − v B t px − 1
= v B B qx − 1
<0 for all B > 0
1
Now f (0) = āx > 0 and, as B → ∞, f (B) → −∞ (as āx → δ and B |āx → 0). It follows that
f (B) = 0 has a unique solution
88 CHAPTER 5. PREMIUMS
Note. The purchase price of an annuity with this guarantee may be considerably larger than for
an ordinary annuity.
If the annuity instalments are paid annually in arrear we must solve the equation
B = an + n |ax (5.7.2)
subject to the condition n − 1 < B ≤ n. (Here n is the guarantee period, which must be an integer.)
This is solved by trial and error.
Another type of guaranteed annuity is that in which the balance (if any) of the purchase price
(or a proportion of it) over the total annuity payments received is paid immediately on the death
of the annuitant. For example, if the guarantee consists of a payment of the balance of 85% of the
purchase price over the total annuity instalments received, equation (5.7.1) would be replaced by
Z n
B = āB + [0.85B − t]v t t px µx+t dt (5.7.3)
0
where n = 0.85B (the time when the annuity payments equal 85% of the purchase price).
Example 5.7.1. A man aged 65 buys a guaranteed annuity payable continuously for a purchase
price of £32, 258. The annuity payments are guaranteed to continue until the total payments reach
£20, 000. The office issuing the contract uses the following basis:
A1967-70 ultimate;
4% p.a. interest;
expenses are ignored.
Let n be the guarantee period. Give an equation for n, and prove that this equation has a unique
solution, which lies between 6 and 7 years.
Solution. Let Z be the annual rate of payment of the annuity. We have the equation
32, 258 = Z(ān + n |ā65 ) (1)
subject to nZ = 20, 000 (2)
Replace Z by 20, 000/n in (1) to obtain
32, 258
ān + n |ā65 = n = 1.6129n
20, 000
Try n = 7
1, 289.7567
f (n) = 6.12136 + × 7.772 − 11.2903
2, 144.171
<0
Try n = 6
1, 401.6093
f (n) = 5.34626 + × 8.112 − 9.6774
2, 144.171
>0
Hence 6 < n < 7
Example 5.8.1. Ten years ago a woman aged exactly 35 effected an assurance policy by level annual
premiums payable for a maximum of 25 years. The policy provided the following benefits:
(i) a whole life assurance benefit of £2, 000 payable at the end of the year of death,
(ii) a family income benefit of term 25 years, with payments of £300 per month, beginning im-
mediately on death, if this occurs within 25 years. The final payment is made in the month
ending 25 years after the issue date.
Calculate the annual premium on the basis given below:
A1967-70
6% interest
expenses are 3% of all premiums.
Mortgage protection policies are similar to those for family income benefits, although the
death benefits are payable in one sum rather than in instalments over balance of the term. The
similarity arises from the fact that the loan outstanding after the tth payment has been made is
equal to the value of the future loan instalments (see Mc Cutcheon and Scott, “An Introduction to
the Mathematics of Finance”, Table 3.8.1).
5.9. EXERCISES 91
Exercises
5.1 A life aged 40 effects a 25-year without profits endowment assurance policy with a sum assured
of £50,000 (payable at the end of the year of death or on survival to the end of the term). Level
premiums are payable annually in advance throughout the term of the policy or until earlier
death of the life assured. Calculate the level premium, P , using the following premium basis
Mortality: A1967-70 Ultimate;
Interest: 6% p.a.
Expenses: none
5.2 An office issues a large number of 25-year without-profit endowment assurances on lives aged
exactly 40. Level annual premiums are payable throughout the term, and the sum assured of
each policy is £10,000, payable at the end of the year of death or on survival to end of the
term. The office’s premium basis is:
A1967-1970 ultimate;
4% p.a. interest;
expenses are 5% of each annual premium including the first, with additional initial ex-
penses of 1% of the sum assured.
5.3 A 5-year temporary assurance, issued to a woman aged 55, has a sum assured of £50,000
in the first year, reducing by £10,000 each year. The sum assured is payable at the end of the
year of death. Level premiums, limited to at most 3 years’ payments, are payable annually in
advance. Calculate the annual premium.
Basis:
A1967-1970 select mortality
4% p.a. interest
expenses are 10% of all office premiums
5.4 A life office sells immediate annuities, using English Life Table No. 12 - Males, 4% p.a. interest
with no expenses as the premium basis. Assuming that the mortality of annuitants does follow
this table, that investments will earn 4% per annum, and that expenses are negligible, find the
probability that the office will make a profit on the sale of an annuity payable continuously to
a life aged 55.
5.5 (i) Let g(T ) be the present value of the profit to the life office, at the issue date, in respect
of an n-year without profits endowment assurance to (x) with sum assured (payable
immediately on death if this occurs within n years) and premium P per annum, payable
continuously for the term of the policy. Expenses are ignored in all calculations.
(a) Write down an expression for g(T ).
(b) Derive expressions for
(1) the mean, and
(2) the variance of g(T ).
(c) For what value of P is the mean of g(T ) equal to zero?
(ii) An office issues a block of 400 without profits endowment assurances, each for a term of
25 years, to lives aged exactly 35. The sum assured under each policy is £10, 000 and the
premium is £260 per annum, payable continuously during the term. The sum assured is
payable immediately on death, if death occurs within the term of the policy.
92 CHAPTER 5. PREMIUMS
Assuming that the office will earn 4% interest per annum, that the future lifetime of the
lives may be described statistically in terms of the A1967-70 ultimate table, and that
expenses may be ignored, find
(a) the mean present value of the profit to the office on the block of policies, and
(b) the standard deviation of the present value of this profit.
£ ¤
Ā35:25 = 0.15646 on A1967-70 ultimate 8.16%
5.6 A life office issued a certain policy to a life aged 40. The benefits under this contract are as
follows:
On death before age 60: an immediate lump sum of £1, 000
On survival to age 60: an annuity of £500 p.a., payable continuously for the
remaining lifetime of the policyholder.
Level annual premiums are payable continuously until age 60 or earlier death.
Premiums are calculated according to the following basis:
Mortality: English Life Table No.12-Males
Interest: 4% p.a.
Expenses: Nil
Calculate the annual premium.
5.10. SOLUTIONS 93
Solutions
5.3 Let the annual premium be P . M.P.V. of assurance benefits is (from first principles)
10000 © ª 1 1
5C[55] + 4C[55]+1 + 3C57 + 2C58 + C59 [= 60000A[55] :5 − 10000(IA)[55]:5 ]
D[55]
5.4 Consider £1 p.a. of annuity. The purchase price is ā55 and the office will make a profit if death
occurs before time t, where
ā55 = āt at 4%
That is,
1 − vt log [1 − δā55 ]
= ā55 , so t=
δ log v
i.e.
log Ā55 log Ā55
t= = = 16.99
log v −δ
The probability of making a profit is thus
l55+t l71.99
t q55 = 1 − =1− = 0.434 (or 43.4%)
l55 l55
(by interpolation)
(
P āT − Sv T if T < n
5.5 (i) (a) g(T ) =
P ān − Sv n if T ≥ n
94 CHAPTER 5. PREMIUMS
D60
− D35 (ä60 − 12 )
= 136.60
∴ For 400 policies, M.P.V. of profit = £54, 640
(b) For 1 policy, variance of p.v. of profit is
µ ¶2
260 £ 8.16% ¤
+ 10, 000 Ā35:25 − ( Ā35:25 )2
δ
| {z } | {z }
0.15646 0.39043
= (1.662915 × 104 )2 × 0.004024415 = 1, 112, 866
√
∴ s.d. for 400 policies = 20 × 1.662915 × 104 × 0.004024415
= £21, 098
RESERVES
The bases used to calculate “reserves” for each of these purposes may be different, and some
practical considerations are beyond our present scope.
The reserve or policy value of the contract (calculated prospectively, i.e. by reference to future
cashflows) is defined as
tV = E(L) (6.2.2)
= M.P.V. of future benefits and expenses
− M.P.V. of future premiums (6.2.3)
95
96 CHAPTER 6. RESERVES
The mortality, interest and expense assumptions used to evaluate t V are known as the reserving
basis. This may or may not agree with the premium basis. If these bases agree (or are assumed
to agree) and there are no expenses, we obtain net premium reserves, which we shall consider
in the next section. By convention, the reserve t V is calculated just before receipt of any premium
then due: the reserve just after payment of this premium is
tV + P 00 − e (6.2.5)
where P 00 is the premium paid at duration t years and e is the expense then incurred.
Certain formulae may give negative values of t V, at least for some policies and at early durations t.
A negative reserve should not normally be held because the life office is thereby treating the contract
as an asset: if the policy is discontinued there is no way in which the policyholder can be made to
pay money to the office! Similarly, the reserves stated in Statutory Returns should not be negative,
although negative reserves may be permissible in certain internal office calculations.
The general rule is therefore that, if a formula gives a negative value of t V, this should be replaced by
zero. Policies should in general be designed so that negative reserves do not arise (cf. the discussion
of family income benefit policies in Section 5.8).
By formula (6.2.4),
tV = net premium reserve
Example 6.3.1. Consider a whole life policy with sum assured £1 without profits, payable imme-
diately on the death of (x). The policy was issued t years ago by level annual premiums payable
continuously throughout life. Find a formula for the net premium reserve t V (on a given mortality
and interest basis).
6.3. NET PREMIUM RESERVES 97
L = v U − P āU
where U = future lifetime of (x + t), and P = P̄ (Āx ).
∴ t V = E(L) = Āx+t − P̄ (Āx )āx+t .
Notation. The International Notation for net premium reserves is similar to that for premiums,
with the addition of “t” to indicate the duration. If premiums are limited to (say) h years of payment,
the symbol h is placed above the duration t, as shown in Table 6.3.1. The general symbol t V may
be used for any sum assured and any reserve, whether net premium or not, but t Vx , t V̄ (Āx ), etc.,
refer to net premium reserves for a policy with a sum assured of £1. If select mortality tables are
used, we write t V[x] , etc.
CONVENTIONS
(1) The symbols t Vx , etc., refer to the reserves just before payment of the premium due at time
t (if a premium is then payable.) The net premium reserve just after receipt of this premium
is of course t Vx + Px , etc. Similar considerations apply if premiums are payable half-yearly,
monthly, etc.
(2) By formula (6.3.1) with t = 0, we have 0 V = 0 for all net premium reserves. In the case of
n-year endowment assurances or pure endowments, it is not always clear whether to take n V
as zero or the sum assured (S, say), i.e. whether to assume that the sum assured has or has
not already been paid. It appears that the convention n V = 0 is used in profit testing but not
elsewhere.
Type of
Notation Prospective Formula for Reserve
Policy
Whole Life
t V̄(Āx ) Āx+t − P̄ (Āx )āx+t
Assurance
n-year Term 1 1 1
t V̄(Āx:n ) Āx+t :n−t − P̄ (Āx:n )āx+t:n−t
Assurance
n-year Endowment
t V̄(Āx:n ) Āx+t:n−t − P̄ (Āx:n )āx+t:n−t
Assurance
h-Payment Years Āx+t − h P̄(Āx )āx+t:h−t t<h
h
Whole Life t V̄ (Āx )
Āx+t t≥h
Assurance
h-Payment Years
n-year h Āx+t:n−t − h P̄(Āx:n )āx+t:h−t t<h
t V̄ (Āx:n )
Endowment Āx+t:n−t t≥h
Assurance
n-year Pure 1
1
t V̄(Ax:n ) Ax+t:n−t − P̄ (Ax:n1 )āx+t:n−t
Endowment
Type of
Symbol Prospective Policy Value Retrospective Policy Value
Policy
Whole Life Dx £ ¤
t Vx Ax+t − Px äx+t Px äx:t − Ax1:t
Assurance Dx+t
Endowment Dx £ ¤
t Vx:n Ax+t:n−t − Px:n äx+t:n−t Px:n äx:t − Ax1:t
Assurance Dx+t
Temporary 1 Dx £ 1 ¤
1
t V x:n Ax+t 1
:n−t − P x:n äx+t:n−t P x:n äx:t − Ax1:t
Assurance Dx+t
Pure 1 1 Dx
t V x:n Ax+t:n−t − P x:n1 äx+t:n−t [P 1 ä ]
Endowment Dx+t x:n x:t
Example 6.3.2.
(i) What is the International Notation for the reserve of Example 6.3.1 ?
Solution.
(i) t V̄ (Āx )
(ii)
Now consider duration r + k, integer r, 0 < k < 1. The reserve r+k Vx is estimated by linear
interpolation between r Vx + Px and r+1 Vx , i.e.
v 1−k [1−k qx+r+k + (1 − 1−k qx+r+k )Ax+r+1 ] − Px v 1−k (1 − 1−k qx+r+k )äx+r+1
6.4. RETROSPECTIVE RESERVES 99
which is complicated to evaluate.] Similar formulae apply for other classes of business with premi-
ums payable annually in advance.
(m)
tV = 1, 000Āx+t − P äx+t just before payment of the premium then due,
or
P
tV + just after payment of this premium.
m
k 1
The reserve at duration t = r + m (where r is an integer multiple of m and 0 < k < 1) is usually
P
estimated by linear interpolation between r V + m and 1 V
r+ m .
NOTE One may use standard symbols, e.g.
(12) (12)
tV (Āx ) = Āx+t − P (12) (Āx )äx+t
(4) 1 (4) (4)
tV (Āx1:10 ) = Āx+t 1
:10−t − P (Āx:10 )äx+t:10−t
1 1
if t is an integer multiple of 12 and 4 respectively, but these are seldom employed.
t Vx = Ax+t − Px äx+t
The corresponding retrospective reserve is found by accumulating the funds of (say) lx identical
policies until time t, and sharing the money out among the survivors. Since mortality is assumed to
follow the table exactly, there are dx deaths in the first policy year, dx+1 in the second policy year,
and so on, with lx+t survivors at time t. Interest is assumed to be earned at the rate i p.a. used in
100 CHAPTER 6. RESERVES
the premium and the reserving bases, so the accumulated funds at time t are
On division by lx+t (the assumed number of survivors at policy duration t), we find that
retrospective
Dx
reserve at = [Px äx:t − Ax1:t ]
Dx+t
duration t years
More generally, we have
M.P.V. (at issue date) of
retrospective
Dx premiums, minus M.P.V.
reserve at = (6.4.1)
duration t years Dx+t of benefits and expenses, in
the first t years
If there are no expenses, we have
retrospective
Dx M.P.V. (at issue date) of
reserve at = premiums, less benefits in (6.4.2)
Dx+t
duration t years first t years
Some examples of formulae for retrospective reserves are given in Table 6.3.2 above.
Theorem 6.4.1. If the premium and the reserving bases agree, the prospective and the retrospective
reserves of a given policy are equal.
Proof. We illustrate the argument by means of the whole life policy discussed earlier in this section.
The retrospective and prospective reserves at policy duration t are
Dx
VR = [Px äx:t − Ax1:t ]
Dx+t
and
VP = Ax+t − Px äx+t
respectively. Hence
Dx+t
(VR − VP ) = Px äx:t − Ax1:t
Dx
Dx+t Dx+t
− Ax+t + Px äx+t
Dx Dx
= Px äx − Ax
(using the facts that äx = äx:t + t |äx and Ax = Ax1:t + t |Ax .)
But Px äx − Ax = 0, by the equation of value for the level annual premium, so we have proved the
desired result.
This argument may easily be extended to cover other policies, including situations in which there
are expenses.
6.4. RETROSPECTIVE RESERVES 101
It follows that, if the premium and reserving bases agree (or are assumed to agree in reserve calcu-
lations), one need not specify whether a prospective or retrospective reserve is required. In practice,
however, the prospective method is more often used.
Example 6.4.1. Give formulae for the following net premium reserves in terms of other monetary
functions:
(i) t Vx:n ,
(ii) t V̄ (Āx:n ),
Solution.
äx+t:n−t
(i) t Vx:n =1− (6.4.3)
äx:n
āx+t:n−t
(ii) t V̄ (Āx:n ) = 1 − (6.4.4)
āx:n
Solution.
(i)
(ii)
We begin by considering prospective gross premium reserves. The premium basis (or the premi-
ums themselves) must be specified, together with the reserving basis, which gives
If the allowance for future expenses is a proportion k of future premiums (including any due
now), the formula for the prospective reserve is
Example 6.5.1. Consider the whole life policy discussed in section 6.4 above, and suppose that
(i) the premium basis is A1967-70 ultimate, 6% p.a. interest, with no allowance for expenses; and
(ii) reserves are to be calculated prospectively by the gross premium method on the following
basis:
Find a formula for the reserve at duration t years (just before receipt of the premium then due.)
6.5. GROSS PREMIUM VALUATIONS AND ASSET SHARES. 103
Solution
The annual premium is
Dx £ ¤
tV = 0.9P 0 äx:t − Ax1:t
Dx+t
where the annuity and assurance factors, and Dx /Dx+t , are on A1967-70 ultimate at 4% p.a. inter-
est (or whatever mortality and interest basis is to be used.)
Asset shares
The asset share of a life assurance policy is a retrospective gross premium reserve calculated on the
basis of the mortality, interest and expenses actually experienced by the office. In order to illustrate
the calculations, let us consider an n-year policy issued t years ago to (x), and define
Let t V denote the asset share at duration t of the policy, before payment of the premium then due.
We clearly have 0 V = 0, and the following relation holds:
t+1 V = (t V + Pt+1 − et+1 )(1 + it+1 ) − death cost, Dt+1 0≤t≤n−1 (6.5.2)
To calculate the “death cost”, we must consider the given policy’s share of the cost of making the
asset share (at the end of the policy year) up to the sum assured, St+1 , for those policies which
became claims in the year. Since the proportion of policies in force at the start of the year which
became claims was q[x]+t , the death cost is
Notes
1. As will be stated in the next chapter, the quantity St+1 − t+1 V is called the death strain at risk
of the policy in year t + 1.
2. By substituting formula (6.5.3) in formula (6.5.2), we may obtain an expression for t+1 V in terms
of t V and other known quantities. In practice, however, one may perhaps replace Dt+1 by a
suitable approximate value and use formula (6.5.2) directly.
104 CHAPTER 6. RESERVES
3. The asset share at policy duration n years may of course differ from the maturity value, leading
to a profit or loss to the office at that time.
4. In the unlikely event that mortality , interest and expenses follow a certain valuation basis
exactly, the asset share of a policy will equal the retrospective gross premium reserve on that
basis.
5. Asset shares are particularly important for with profit policies, as their surrender and maturity
values are not fixed in advance and the asset share gives one method of deciding “fair” surrender
and maturity values. A number of other considerations are, however, also involved, and we shall
not pursue the discussion of asset shares.
the bases being perhaps different (reflecting expected future conditions and actual past conditions
respectively.) The position is further complicated by the possible existence of guaranteed surrender
values and the difficulty of estimating future conditions. In practice, a net premium valuation at
a low rate of interest may be used, either by choice or because this is required by the supervisory
authorities: such a basis (though apparently artificial, since the premium value is unrelated to
the actual premium being charged) may have the property that the corresponding reserves are
comparable with, or generally greater than, the more “scientific” reserves given above. Similar
arguments lead to the “net premium method” for with-profit policies (provided that terminal bonuses
are allowed for separately): cf. Section 6.9 below.
The reason for the use of a low rate of interest in net premium reserves is that, by Lidstone’s
theorem, net premium reserves normally increase as the rate of interest falls.
The supervisory authorities may be prepared to allow Zillmerisation (see Section 6.7 below) of
the reserves (subject to limits on I), but many offices prefer to publish unmodified net premium
reserves, in case Zillmerisation is seen as a sign of financial weakness.
The above discussion refers mainly to the calculation of reserves for solvency testing, but reserves
are also used for internal purposes: deciding upon bonus levels, alterations to policies, etc. For these
purposes the office may use other bases.
The office must also allow for random fluctuations from the mean numbers of deaths (especially
when there is a small number of policies with very large sums assured or death strains at risk) and
consider the exchange or reinsurance of large risks.
Let us consider a policy providing £S immediately on the death of a life aged x at issue, with
premiums of P p.a. payable continuously throughout life. The random variable L is defined as
L = prospective loss on contract (at duration t)
= Sv U − P āU (6.6.1)
where U is the future lifetime of the policyholder, who is now aged x + t, and expenses are ignored.
Now we cannot use the formula
Var(L) = Var(Sv U ) + Var(P āU ) wrong!
(because Sv U and P āU are not independent), so we express L in terms of v U . That is,
µ ¶
U 1 − vU
L = Sv − P
δ
µ ¶
P P
= S+ vU −
δ δ
from which we obtain
µ ¶2
P
Var(L) = S+ Var(v U )
δ
µ ¶2 h
P ¡ ¢2 i
= S+ Ā∗x+t − Āx+t (6.6.2)
δ
where ∗ indicates a rate of interest of 2i + i2 p.a. This technique may also be used for certain other
types of contracts.
We now consider a number of life assurance policies, labelled from j = 1 to j = N . The total
reserves for the group of policies is
N
X
E[Lj ] = V1 + V2 + · · · + VN (6.6.3)
j=1
where Lj and Vj are, respectively, the net liability and the reserve of the jth policy. If we further
assume that the lives assured are independent,
XN XN
Var Lj = Var(Lj ) (6.6.4)
j=1 j=1
where Var(Lj ) may be calculated by (say) formula (6.6.2). If we now suppose that all the policies
are identical (i.e. they are of the same sum assured, type and duration, and were issued at the same
date to N different lives all of the same age), the average of the net liabilities is
N
X
L̄ = Lj /N (6.6.5)
j=1
which has mean E(L) and variance Var(L)/N , where L is the net liability of a given policy. These
results are sometimes used in solvency calculations, etc. It is sometimes also assumed that N is so
N
X
large that the distribution of Lj or L̄ is approximately normal (by the central limit theorem
j=1
and related results), although this may only be accurate for very large values of N .
106 CHAPTER 6. RESERVES
Z
tV = t Vx:n − I [1 − t Vx:n ]
= (1 + I)t Vx:n − I
as required.
Notes
1. The case of a whole life policy is similarly dealt with (put n = ∞ in the above formulae.)
2. A similar argument may be used to establish formula (6.7.1) if the sum assured is payable
immediately on death and premiums are payable continuously. In this case, t V = t V̄(Āx:n ).
3. Zillmer’s formula does not in general hold for other classes of policy, nor when the premium
due at time t has been paid. (It is, however, sometimes used in practice for all policies and all
durations.)
4. When the duration t is short (e.g. t = 1) Zillmer’s formula may give a negative reserve, which
should, as a rule, be replaced by zero, i.e. a policy should not be treated as an asset to the
office. Note that formula (6.7.1) gives
Z
0V = −I
which is correct if the additional initial expense I is thought of as being disbursed before the
first premium is received: if it is not, one should write
Z
0V =0
Example 6.7.1. Ten years ago life office issued a 20-year endowment assurance without profits to
(35). The sum assured is £10,000, payable at the end of the year of death (or on survival for 20
years), and premiums are payable annually in advance. The basis for premiums and reserves is:
A1967-70 ultimate;
6% p.a. interest;
expenses are 3% of all office premiums (including the first) with additional initial expenses of
1.5% of the sum assured.
Calculate
(i) the annual premium, and
(ii) the reserve,
(a) just before receipt of the premium now due, and
108 CHAPTER 6. RESERVES
Solution.
l55
ä35:20 = ä35 − v 20 ä55
l35
= 11.9969
and
A35:20 = 1 − dä35:20
(ii) One may use Zillmer’s formula in this case (or other methods). The reserve is
Example 6.7.2. Suppose that the policyholder of example 6.7.1 were to surrender his policy (before
paying the premium now due) and the office grants a surrender value equal to the reserve, as defined
in that question. What yield per annum would the policyholder have obtained?
i s̈10
0.03 11.81
0.04 12.49
Interpolate:
Hence
Ax+1:n−1
P 00 − e = = Px+1:n−1
äx+1:n−1
Hence
This reserve is called the Full Preliminary Term reserve, which we shall write as t V F P T . We have
shown that
FPT
tV = t−1 Vx+1:n−1 (6.8.2)
In the case of a whole life policy, we may set n = ∞ to obtain
FPT
tV = t−1 Vx+1 (6.8.3)
If the sum assured is not £1, we naturally multiply these expressions by the sum assured. Similar
formulae hold if the sum assured is payable immediately on death, when premiums are payable
continuously or monthly, etc., and for certain other types of policy.
What level of initial expenses leads to a F.P.T. reserve? We observe that, by the retrospective
method,
FPT Dx £ 00 ¤
1V = (P − e)äx:1 − I − Ax1:1
Dx+1
=0
from which we obtain
P 00 = e + I + Ax1:1 (6.8.4)
This states that the first premium is exactly sufficient to pay the initial expenses, I + e, and the
cost of the first year’s life assurance cover, Ax1:1 .
The use of full preliminary term reserves was suggested by T. B. Sprague in 1870. Other actu-
aries (particularly in North America) have devised formulae for “modified full preliminary term
reserves”, in which the allowance for total initial expenses may be smaller than that of formula
(6.8.4), but we do not pursue this topic.
110 CHAPTER 6. RESERVES
(1) The net premium method. The reserve is taken to be the net premium reserve for corre-
sponding non profit policy, plus the mean present value of the bonuses already declared. In the
case of the endowment assurance policy discussed above, the reserve is given by the formulae:
WP
tV = St Vx:n + BAx+t:n−t (6.9.1)
= (S + B)Ax+t:n−t − SPx:n äx+t:n−t (6.9.2)
where all the actuarial functions are calculated on the given mortality and interest basis.
The rationale of this method is that the additional premiums for a with profits policy (relative
to the corresponding without profits policy) are considered to have “earned” the bonus B so far
declared, so an additional reserve is required to cover the value of these bonuses.
Example 6.9.1. On 1st January 1993 a life office issued a with-profit assurance endowment
policy with a term of 10 years to each of 100 male lives then aged exactly 50. The basic sum as-
sured under each contract was £20, 000, and the basic sum assured, plus any bonuses, is payable
on maturity or immediately on earlier death. Level annual premiums are payable in advance
throughout the term. Among the group of policies there were 2 deaths during 1993 and 4 deaths
during 1994. There were no lapses or surrenders during 1993 or 1994. For this group of policies,
the office has declared a compound reversionary bonus of 3 43 % per annum vesting in advance on
each 1st January from outset. The office values the policies by a net premium method, using
A1967-70 ultimate mortality and interest at 3%p.a.
Calculate the total reserves which the office held for the group of policies on 31st December 1994.
6.9. RESERVES FOR WITH-PROFITS POLICIES 111
Solution. We require 94×reserve of one policy. Using the net premium method for with profits
policies,
reserve = S[2 V(Ā50:10 )] + B Ā52:8
per policy
1
= (1.03) 2 (A52:8 − v 8 8 p52 ) + v 8 8 p52
1
= (1.03) 2 (0.06142) + 0.73329
= 0.79562
= £443, 417
(2) The bonus reserve method. As the name suggests, reserves are calculated on the assumption
that reversionary bonuses will be declared at certain annual rates. The value of the benefits
is therefore calculated by the formulae given in Section 3.10 above, according to the system of
bonus declarations used by the office, and the reserve is then found by subtracting the value of
the office premiums less projected future expenses. If, in the case of the endowment assurance
policy discussed above, it is assumed that future bonuses will be at the rate b per annum on the
compound system and a proportion k of future premiums will be absorbed in expenses, a formula
for the reserve is
WP
tV = (S + B)A∗x+t:n−t − (1 − k)P 00 äx+t:n−t (6.9.5)
where ∗ indicates the rate of interest (i − b)/(1 + b). Similar formulae may be derived for other
policies.
Since the premium valued is the actual (or office) premium for the policy, this method is some-
times referred to as the gross premium method for with profits policies.
As might be expected, the reserves calculated by the bonus reserve method depend greatly on the
assured future levels of bonus, and in the early years of a contract the method may give reserves
which are quite unrelated to the accumulation of premiums.
112 CHAPTER 6. RESERVES
(3) The asset share method. This has been described (for non-profit policies) in Section 6.5
above; the formulae are almost unchanged for with profit policies, the only difference being the
fact that the sum assured in each policy year includes the vested bonuses. Since these only affect
the death cost, the reserves given by the asset share method do not depend greatly on past bonus
declarations.
Exercises
6.1 (i) Express t Vx in terms of Ax and Ax+t . Hence, or otherwise, find the values of n Vx and
n Vx+n , given that
1 − Ax+2n = Ax+2n − Ax+n = Ax+n − Ax
(ii) Calculate 1 V40 given that P40 = .01536, p40 = .99647 and i = .05.
6.2 Consider an n-year pure endowment policy, issued to (x), with sum assured 1 and with annual
premiums payable throughout the duration of the policy. In the event of death within the n
years, all premiums paid will be returned without interest at the end of the year of death.
Obtain expressions for the reserve at duration t
Using commutation functions, show that your expressions are equal. (Assume that the premium
due at time t has not yet been paid.)
6.3 (Difficult)
6.4 Ten years ago a life office issued a 20-year endowment assurance without profits to (35). The
sum assured is £10000, payable at the end of the year of death (or on survival for 20 years),
and premiums are payable annually in advance. Using A1967-70 ultimate 6%, and ignoring
expenses, calculate
(i) Px = 0.01212
(ii) 20 Px = 0.01508
(iii) P x:101 = 0.06942
(iv) 10 Vx = 0.11430
20
Calculate 10 Vx .
6.6 A whole life assurance with sum assured £100,000 payable at the end of the year of death was
purchased by a life aged 30. The policy has annual premiums payable throughout life.
The basis for calculating reserves for this policy is as follows:
net premium method: A1967-70 ultimate, 5% p.a. interest.
Estimate the policy value at duration 28 14 by interpolation.
Solutions
6.1 (i)
äx+t däx − däx+t Ax+t − Ax
t Vx =1− = =
äx däx 1 − Ax
The given information implies that
Ax+n = 2Ax+2n − 1
and
Ax = 2Ax+n − Ax+2n = 3Ax+2n − 2
Hence
Ax+n − Ax 1 − Ax+2n 1
n Vx = = =
1 − Ax 3 − 3Ax+2n 3
and
Ax+2n − Ax+n 1 − Ax+2n 1
n Vx+n = = =
1 − Ax+n 2 − 2Ax+2n 2
(ii) Retrospective method gives
l40 P40 (1 + i) − q40
1 V40 = [P40 − vq40 ](1 + i) = = 0.01264
l41 p40
or
P (Nx − Nx+n ) = Dx+n + P (Rx − Rx+n − nMx+n ) (∗)
The expressions for the reserves are as follows:
Retrospective:
tV = [P (Nx − Nx+t ) − P (Rx − Rx+t − tMx+t )] /Dx+t (∗∗)
Prospective:
1 1
tV = n−t Ex+t + tP Ax+t :n−t + P (IA)x+t:n−t − P äx+t:n−t
= [Dx+n + tP (Mx+t − Mx+n )
+ P (Rx+t − Rx+n − [n − t]Mx+n )
− P (Nx+t − Nx+n )] /Dx+t (∗ ∗ ∗)
1 1 äx − äx:n
− =
äx:n äx äx äx:n
n Ex äx+n
=
äx äx:n
= 0.94P x:n1
= 0.235
µ ¶ µ ¶
1 1 1 1
Px:n − Px = −d − −d = − = 0.235
äx:n äx äx:n äx
so
Px:n = 0.255
P x1:n = Px:n − P x:n1 = 0.005
6.5 The prospective method does not work. By the retrospective method,
20
(reminder: 10 Vx refers to limited payments policy)
20 Dx
10 Vx − 10 Vx = [20 Px äx:10 − Px äx:10 ] (since benefits are the same
Dx+10 under both policies)
µ ¶
Dx äx:10
= (20 Px − Px )
Dx+10
P
20 x − P x
= = 0.04264
P x:101
∴ 20
10 Vx = 0.15694
6.6
28 V = 100, 000 28 V30 = 33, 101
29 V = 100, 000 29 V30 = 34, 754
Interpolate between 28 V + P and 29 V, where P is the annual premium, i.e. P = 100, 000
P30 = 724.00. This gives
3 1
28 14 V = (33, 101 + 724) + (34, 754)
4 4
= £34, 057
6.11. SOLUTIONS 117
6.7 (i) Net premium reserve = reserve calculated ignoring expenses and
assuming premium basis agrees with re-
serving basis
= t Vx = Ax+t − Px äx+t
in example given
(ii) Zillmerised reserve = (1 + I)t V − I per unit sum assured
(where t V = net prem. reserve, I = additional initial expenses per £1 sum assured)
à !
³ ´ value of future
value of future
(iv) Gross prem. reserve = − gross premiums
benefits
less expenses.
APPLICATIONS OF RESERVES
(i) the experience of the office since the policy was issued,
(iv) competition between life offices (and possibly with a market for “second hand” life policies),
and
A full discussion of all points involved is beyond our present scope. In some cases surrender values
are calculated from paid up policy values, which we discuss in the next section.
Transfer values in pension schemes correspond to surrender values of life assurance policies: a
transfer value is, however, payable only to another pension scheme and not to the scheme member
directly.
119
120 CHAPTER 7. APPLICATIONS OF RESERVES
t Vx = t Wx Ax+t (7.2.2)
where t V is the reserve or surrender value and A is an assurance factor, such as Ax+t or Āx+t:n−t .
(We note that no expenses have been allowed for at or after the date of alteration: if such an allowance
is desired, the formulae should be modified.)
In the important case of endowment assurances, the proportionate rule is sometimes used in
practice, i.e.
t
tW = S (7.2.6)
n
where S is the original sum assured, t is the total number of premiums actually paid and n is the
number originally payable.
Surrender values of endowment assurances (or other policies) are sometimes found from the P.U.S.A.
by the formula
(SV )t = (P.U.S.A.)A (7.2.7)
where A is the appropriate assurance factor at age x + t. In particular, if the proportionate rule for
endowment assurances is used, the surrender value at duration t years is sometimes taken as
t
SAx+t:n−t (7.2.8)
n
7.3. ALTERATIONS AND CONVERSIONS 121
if the sum assured is payable at the end of the year of death, if this occurs within the balance of the
term, n − t years, or
t
S Āx+t:n−t (7.2.9)
n
if the sum assured is payable immediately on death within this period.
Example 7.2.1.
(i) Four years ago a life then aged 35 effected a 25-year without profits endowment assurance by
annual premiums for a sum assured of £50,000 payable at the end of the year of death or on
survival for 25 years. He surrendered the policy just before paying the premium now due.
The life office calculated the premiums for this policy on the basis of A1967-70 select at 4%
interest with the following expense loadings:
2.5% of each premium, including the first, plus a further
initial expense of 1% of the sum assured.
The office also calculates reserves on this basis, and surrender values are equal to 95% of
reserves.
(a) Calculate the office annual premium.
(b) Calculate the surrender value of the policy.
(ii) Suppose that the policyholder of (i) had made his policy paid-up instead of surrendering it,
and that the office calculates paid-up values by the proportionate rule. Calculate the yield per
annum (to the nearest 0.1%) that he would have obtained on his entire transaction, assuming
that he survives to maturity date.
Solution.
(i) (a) Let A.P. be P 0 . We have
0.975P 0 ä[35]:25 = 50000(A[35]:25 + 0.01)
∴ P 0 = £1289.10
(b)
Reserve = 50000A39:21 − 0.975P 0 ä39:21
= £4790.00
∴ S.V. = 0.95 × reserve = £4550.50
i.e.
(1 + i)21 s̈4 = 6.20588
By trials and interpolation, i ; 1.9%.
assurance maturing at age 60. The policyholder might also request his policy to be converted to
another class of business: for example, from a whole life assurance to an endowment assurance.
The usual rule when carrying out these calculations is to equate the reserves before and after
conversion. If V1 and V2 denote the reserves before and after the conversion respectively, the rule
may be stated as follows:
V1 = V2 (7.3.1)
If there are expenses of C for the conversion itself, the formula becomes
V1 − C = V2 (7.3.2)
The bases used to calculate V1 and V2 may in practice be different, and V2 must be calculated
prospectively by the gross premium method.
To show that formulae (7.3.1) and (7.3.2) hold, we argue as follows. At the date of conversion,
the policyholder’s original policy is worth V1 on immediate surrender. (In practice, however, the
reserve V1 used in conversion calculations may exceed the surrender value, as the office is not losing
business on a conversion or alteration.) Imagine that V1 is applied as a special single premium
towards the new contract. The equation of value of the new contract is thus
The office then solves this equation for the unknown quantity (usually the new sum assured or
premium.) An example of the use of formula (7.3.1) has already been encountered in equation
(7.2.2) above.
Example 7.3.1. Ten years ago, a man now aged 40 effected a whole of life assurance for a sum
assured of £10000, payable at the end of the year of death, by level annual premiums. The premiums
were calculated using A1967-70 ultimate 4% and an allowance for expenses of 50% of the first year’s
premium and 5% of each subsequent premium.
Immediately before payment of the eleventh annual premium the man requests that the policy be
converted into an endowment assurance maturing on his sixtieth birthday, with the annual premium
remaining unaltered. Calculate his revised sum assured using a full preliminary term reserve (on
A1967-70 ultimate, 4% interest) for finding the reserve of the original policy, and
A1967-70 ultimate, 4% interest, expenses of 5% of premiums
for calculating the reserve of the new policy.
Example 7.3.2. An office issued a ten-year endowment assurance to a man aged exactly 45. The
monthly premium was £25 during the first five years increasing to £50 thereafter. The sum assured,
payable immediately on death, was calculated using A1967-70 select 4%, allowing for expenses of
£1 per month with additional initial expenses of 2 12 % the sum assured.
After five years the man requested that the premium remain at £25 per month for the next five
years, with the death benefit staying unaltered. Calculate the reduced sum payable on survival using
the premium basis and allowing for a £30 alteration charge.
D50
(using ä(12) (12)
[45] :10 = 8.154, ä 50 :5 = 4.489, = 0.8093, Ā[45]:10 = 0.6808 )
D[45]
Let S.A. on survival be reduced to S 0 . Equate reserves before and after conversion [allowing for
cost of conversion]:
Example 7.3.3. In a certain country, the Universities operate the following superannuation scheme
for academic staff, of whom there are two grades, Lecturer and Professor. Lecturers retire at age 65
with a lump sum of £30, 000, while Professors retire at age 70 with a pension of £12, 000 per annum
payable monthly in advance. Staff pay for their benefits by means of level annual premiums, payable
annually in advance until retirement or earlier death. Premiums are returned without interest on
death before retirement, and withdrawals may be ignored. The employers do not contribute to the
scheme. Death benefits are paid at end of year of death. The benefits are provided by policies issued
by a life office which uses the following basis for all calculations:
A1967-70 ultimate
124 CHAPTER 7. APPLICATIONS OF RESERVES
Solution.
(i) Let the annual premium be P . The equation of value is
P ä25:40 = 30000A25:401 + P (IA)25
1
:40
=⇒ P = £274.44
(ii)
D65 1 1
V1 = 30000 + 20P A45 :20 + P (IA)45:20 − P ä45:20
D45
= £8536
The mean of this variable (even without the assumption of independence of the lives) is
Note. Once the number of deaths θx+t , is known (at the end of the policy year), ADS ceases to be
a random variable: it is then a realisation of the variable.
The position may be generalised to cover cases in which policies considered are not identical,
provided that their premiums are due on the same date. The actual death strain is
X
ADS = (S − t+1 V) (7.4.4)
death claims
in policy
year t + 1
Hence
126 CHAPTER 7. APPLICATIONS OF RESERVES
X X X
mortality profit = (t V + P )(1 + i) − (S − t+1 V) − t+1 V (7.5.2)
all in force deaths all in force
at the at the
start of start of
year year
Theorem 7.5.1. Under the present assumptions, the mortality profit to the life office in respect of
this group of policies, i.e. the difference between the actual funds at the end of the policy year and
the funds required to pay claims and set up reserves, is equal to E.D.S. minus A.D.S.; that is,
mortality profit in policy year t + 1 = E.D.S. − A.D.S. (7.5.3)
which may be thought of as saying that the accumulated funds at the end of the year, (t V + P )(1 + i) ,
must provide:
(i) the expected cost of death claims, Sqx+t , plus
(ii) the expected cost of setting up reserves for the survivors, (1 − qx+t )t+1 V.
On summing over all policies in force at the start of the year, we find that
X X
(t V + P )(1 + i) = t+1 V + EDS (7.5.6)
all in force all in force
at the at the
start of start of
year year
Example 7.5.1. On 1st January 1988, a certain life office sold 100 10-year without profits endow-
ment assurance policies to lives then aged 50. Premiums are payable annually in advance, and the
sum assured under each contract is £30000, payable on maturity or at the end of the year of previous
death. The office calculates premiums and reserves on the following basis:
7.5. MORTALITY PROFIT AND LOSS 127
expenses: nil
All 100 policies were still in force on 1st January 1994, and 2 policyholders died during 1994.
Calculate the mortality profit or loss to the office during 1994 in respect of this business, at 31
December 1994.
Hence
Example 7.5.2. On the 1st January 1979 a life office issued a number of 20-year non-profit endow-
ment assurance policies, with annual premiums payable in advance throughout the term and sums
assured payable on maturity or at the end of the year of death, to lives then aged exactly 45. At
31st December 1993, total sums assured of £4,000,000 remained in force. During 1993 sums assured
of £50,000 became death claims, and there were no other exits in 1993.
The office calculates premiums, and maintains net premium reserves, on the basis given below.
Calculate the profit or loss from mortality for this group of policies for the year ending 31st
December 1993.
expenses: nil.
Solution.
Sums assured in force on 1.1.1993 = £4050000
Some generalisations
2. If the premium basis differs from the reserving basis, formula (7.5.3) remains true provided that
reserves are calculated by a gross premium method (either retrospective or prospective.) This
follows from the fact that equation (7.5.8) remains true (with mortality, interest and expenses
according to the reserving basis and P on the premium basis.)
Example 7.5.3. For several years an insurance company has issued a large number of special en-
dowment assurances. Each policy matures at exact age 65 and is effected by annual premiums
payable on each January 1st throughout the term. The sum assured, payable at the end of the year
of death during the term, is one half of the sum assured that will be paid if the policyholder survives
to age 65. Details of the policies in force on 31 December 1988 are as follows:
The claims in 1989 were on policies with the following total sums assured and premiums:
Assuming that at 31.12.88 the office’s funds were equal to its reserves, calculate the mortality profit
or loss in 1989, given that the following reserving basis is used:
prospective method, gross premium reserve using A1967-70 ult., 4% p.a. interest, no expenses.
Solution. Let EDS and ADS denote the expected death strain and the actual death strain in 1989.
Then X
EDS = q55 [Sj − {Sj (A56:9 + 9 E56 ) − Pj ä56:9 }]
all j
7.6. OTHER SOURCES OF PROFIT AND LOSS 129
where the summation is over all policies in force at the start of the year, i.e.
hX X X i
EDS = q55 ( Sj ) − ( Sj )(A56:9 + 9 E56 ) + ( Pj )ä56:9
X X
= q55 (694048) (since Sj = 3500000 and Pj = 250000)
all j all j
= 5859
The actual death strain is obtained by a summation of the death strains at risk over those policies
which become claims. Thus
X
ADS = [Sj − {Sj (A56:9 + 9 E56 ) − Pj ä56:9 }]
Claims
X X X
=( Sj ) − ( Sj )(A56:9 + 9 E56 ) + ( Pj )ä56:9
Claims Claims Claims
X X
= 13082 (since Sj = 50000 and Pj = 40000)
Claims Claims
Interest profits. Suppose that the office earns interest at a rate i0 p.a. in policy year t + 1, the
interest rate in the reserving basis being i p.a. Assuming as before that the funds at the end of
policy year t equal the reserves for the business under consideration, the interest profit is
X
[ (t V + P − e)](i0 − i) (7.6.1)
all in force
at the
start of policy
year t + 1
Example 7.6.1. On 1st January 1978 a special twenty-year endowment assurance policy for an
initial sum assured of £20,000 was issued to a life then aged 40 exactly. The sum assured, which is
payable at the end of the year of death or on survival to maturity date, increased to £25,000 after 10
years and to £30,000 on maturity. The annual premium was P for the first ten years, and thereafter
increased to 1.25P . The office’s basis for premiums and reserves is
A1967-70 ultimate
(i) Calculate the initial annual premium P , and the reserve for the contract at (a) 31st December
1989, and (b) 31st December 1990.
130 CHAPTER 7. APPLICATIONS OF RESERVES
(ii) Find the profit or loss from mortality for the year ending 31st December 1990 for a group of
such policies, all effected on 1st January 1978 by lives then aged 40 exactly, given that one
policyholder dies in 1990 and 120 policies remained in force at 31st December 1990. There
were no withdrawals in 1990.
(iii) The office found that it made neither profit nor loss in 1990 in respect of the policies of (ii)
above. What rate of interest did it earn in 1990?
(Expenses are negligible.)
Solution.
(i) Let initial premium be P .
D60
P [ä40:20 + 0.2510 |ä40:10 ] = 20000A40:20 + 500010 |A40:10 + 5000
D40
(a)
D60
12 V = 25000A52:8 + 5000 − 1.25P ä52:8
D52
= £14085
(b)
(ii)
∴ i0 = 0.0414, or 4.14%
Surrender profits. Assume that surrenders may take place only at the end of a policy year (just
before payment of any premium then due). The surrender profit is
X
(t+1 V − S.V.) (7.6.2)
surrenders
7.6. OTHER SOURCES OF PROFIT AND LOSS 131
Example 7.6.2. On 1st January 1977 an office issued a block of without profits endowment assur-
ance policies and a block of without profits pure endowment policies to men aged exactly 40. Each
policy was for a term of 20 years and a sum assured of £1000. Premiums are payable annually in
advance during the term of the policies and death claims are paid on 31st December each year.
On 1st January 1987 there were 1000 endowment assurance policies and 5000 pure endowment
policies in force. During the year to 31st December 1987 sixty endowment assurance policies became
claims by death and twenty pure endowment policies were terminated, without payment of any sum
assured, by the deaths of the assured lives.
On 31st December 1987 ten endowment assurance policies and five pure endowment policies were
surrendered.
Calculate the profit or loss arising during the year 1987, showing separately the contributions
from mortality and from surrender if the office uses the following basis:
Ignore expenses.
Endowment Assurances.
Reserve per policy at end of 1987 = 100011 V40:20
µ ¶
ä
= 1000 1 − 51:9
ä40:20
µ ¶
7531
= 1000 1 −
13.764
= 452.85
∴ Death strain at risk per policy in force at start of 1987 is 1000 − 452.85 = 547.15
∴ E.D.S. = 1000q50 × 547.15 = 26202
A.D.S. = 60 × 547.15 = 32829
132 CHAPTER 7. APPLICATIONS OF RESERVES
S.V.= net premium reserve at 6%. It is easier to find the reserve prospectively.
1000v 20 ll60 279.25
Define P 0 = net premium at 6% p.a. interest = 40
= = 23.52
ä40:20 11.871
l60
∴ Reserve = 1000v 9 − P 0 ä51:9 at 6% p.a. interest
l51
= 546.87 − 165.32 = 381.55
Summary
Exercises
7.1 Explain briefly how a life office calculates the new sum assured or annual premium on the
alteration or conversion of an existing policy.
7.2 A life office issued a whole life without profits policy to a woman aged 30, with sum assured of
£110,000. Premiums were payable annually in advance throughout life, and the sum assured
was payable at the end of the year of death. Immediately before payment of the 15th premium,
the policyholder requests that the policy be converted to a without profits endowment assurance
for the same sum assured, payable on maturity at age 60 or at the end of the year of death, if
before age 60. The office calculates premiums and maintains reserves on the basis of A1967-70
ultimate mortality and 4% per annum interest, with expenses of 4% of each premium. Expenses
of alteration may be ignored.
Find the revised annual premium.
7.3 On 1 January 1972 an office issued a large number of 40-year endowment assurance policies,
each with sum assured £1000 (payable at the end of the year of death, if this occurs within the
term) to a group of lives all then aged 25. On 31 December 1988 there were 8567 policies still
in force. During 1989 there were 13 deaths among the policyholders.
Find the actual death strain, the expected death strain and the mortality profit or loss for the
business in 1989, using the following basis for all calculations:
7.4 On 1 January 1991, a life office issued a number of identical special endowment assurances to
lives then aged 45. Each policy had a term of 15 years and level annual premiums were payable
throughout the term. On survival to the end of the term, a sum assured of £5000 is payable.
On death before the end of the term, a sum assured of £1000 is payable at the end of the year
of death and, in addition, all premiums paid are returned without interest.
The premium basis for these policies was as follows:
(i) Show that the annual premium for each of these policies is £242.27.
(ii) Calculate the reserve, on the premium basis, on 1 January 1996 for one of these policies,
assuming that the premium then due is unpaid.
(iii) On 1 January 1995, a total of 10,000 of these policies were still in force. In 1995, 47
of these lives died, ten policies were surrendered, and the life office earned 4% interest
on its investments. Expenses were negligible in 1995. The surrenders all took place at
the end of 1995, and the office gave surrender values equal to 95% of the reserve on the
premium basis.
Calculate the profit or loss to the office in 1995 from this block of business in respect
of
(a) mortality , and
134 CHAPTER 7. APPLICATIONS OF RESERVES
(b) surrenders.
7.5 A life office has issued a 5-year decreasing temporary assurance to a life aged 60. The sum
assured, payable at the end of the year of death, is £100,000 if the life dies in the first year,
£90,000 if the life dies in the second year, and so on, decreasing by £10,000 each year. Level
annual premiums are payable throughout the term of the policy.
The premium and reserve basis for this policy is as follows:
mortality: A1967-70 Ultimate
interest: 4% p.a.
expenses: Nil
Solutions
7.1 The office calculates V1 , the reserve for the original policy at the date of conversion. The office
then sets up the equation
V1 − C = V2
where
C = cost of alteration (if any)
and
V2 = (prospective) reserve of new policy.
7.2 Let
110000
P = original A.P. = P30
1 − 0.04
7.3 On 31/12/88 the duration is 17 years. The reserve at the end of 1989 (per unit sum assured)
therefore equals
18 V[25]:40 = 1 − ä43:22 /ä[25]:40 = .28397
Hence in 1989 we have, per policy in force at the start of the year, a death strain at risk of
(1000-283.97)=£716.03.
The expected death strain per policy is 716.03 × q42 = 1.31105.
For the block of business, therefore, we have:
(ii)
D60 1 1
5V = 1000A50:10 + 4000 + 5P A50 :10 + P (IA)50:10 − P ä50:10
D50
= £1, 349
(iii) (a)
D.S.A.R. in 1995 in respect of 1 policy = (1000 + 5P ) − 5 V
= £861.97
EDS = 10, 000q49 × DSAR = £36, 714
ADS = 47 × DSAR = £40, 513
∴ mortality profit = EDS − ADS = −£3, 798 (loss of £3,798)
(b)
Surrender profit per policy = 0.05 × reserve
∴ Total surrender profit = 10 × 0.05 × 5 V
= £675.
which gives
P = £1, 347.03
(ii)
(t V + P )(1.04) = q60+t (100, 000 − 10, 000t) + (1 − q60+t )t+1 V
(since the sum assured on death in year t + 1 is 100, 000 − 10, 000t).
Chapter 8
EXTRA RISKS
8.1 Introduction
It is usual in practice for a life office to accept the majority of proposers for life assurance policies on
“normal” (or “standard”) terms, the remainder being either declined or accepted on special terms,
i.e. they are accepted but with higher premiums or reduced benefits. Those accepted on special
terms are considered to be subject to an extra mortality risk due to a health impairment, a dangerous
occupation or a dangerous spare-time activity (such as motor racing.) The assessment of extra risks
is usually carried out by experienced underwriters using a “numerical rating system”; we shall not
go into this, but shall assume that the addition to the normal rates or force of mortality is known.
Some offices make extensive use of “rating up”: that is impaired lives are treated as normal
lives a number of years older than actual age. (Females are sometimes treated as males 5, say, years
younger, although this rule is very inaccurate at certain young ages.) Smokers are sometimes treated
as non-smokers 6 years older.
Hence
· Z t ¸
∗ ∗
p
t [x] = exp − µ [x]+r dr
0
· Z t ¸
= exp − (µ[x]+r + k) dr
0
= e−kt t p[x] (8.2.2)
137
138 CHAPTER 8. EXTRA RISKS
Also,
∗
t E[x] = v t t p∗[x]
= e−(δ+k)t t p[x]
0
= e−δ t t p[x] (8.2.3)
where
δ0 = δ + k (8.2.4)
0
and hence i0 = eδ − 1 = eδ+k − 1.
Since an annuity factor is a sum or integral of pure endowment factors, we have (for example)
ä∗[x] = ä[x]
on normal mortality but with force of interest δ 0 = δ + k (8.2.5) This rule does not
apply to assurances and premiums, but one may sometimes use conversion relationships to express
these in terms of annuity factors, as in the following example.
Example 8.2.1. A certain life office uses the following basis for calculating premiums for assurance
policies on lives accepted at normal rates:
A1967-70 ultimate
4% interest
A certain proposer, aged 50, for a 15-year endowment assurance without profits by annual premiums
is considered by the office to be subject to the mortality of the office’s normal table with an addition
of 0.019048 to the force of mortality for the next 15 years. The sum assured is £10,000, payable
at the end of the year of death, or on survival for 15 years. Calculate the addition to the normal
annual premium.
Note that
∗ 1
P50:15 = − d by a conversion relationship
ä∗50:15
1
= − d since δ.04 + 0.019048 = δ.06
ä50:15 0.06
In some problems we must evaluate assurance factors directly, without the use of conversion rela-
tionships; e.g.
Hence
Remark. To find the median future lifetime of the impaired life, we solve the equation
∗
t qx = 0.5
∗ lx+t e−kt
i.e. t px = = 0.5
lx
Z t
In some cases ξ(r) dr may be calculated analytically, as in the next example.
0
Example 8.3.1. A certain impaired life aged 75 experiences mortality according to a(55) males
ultimate with an addition to the force of mortality. The addition is 0.005 at age 75, increasing
linearly to 0.020 at age 90.
(i) Find a formula for the probability that this life will be alive at age 75 + t years (0 ≤ t ≤ 15).
(ii) Estimate, by approximate integration using the trapezoidal rule (not repeated), the single
premium for a temporary assurance of £50,000 payable immediately on death of (75), if this
occurs within 15 years, on the following basis:
Solution.
(i)
µ∗75+t = µ75+t + 0.005 + 0.001t (0 ≤ t ≤ 15)
· Z t ¸
∴ t p∗75 = t p75 exp − (0.005 + 0.001r) dr
0
£ ¤
= t p75 exp −(0.005t + 0.0005t2 )
8.4 Rating up
In practice, impaired lives are often considered to experience the mortality of a “normal” life r years
older (where r depends on the severity of the impairment.) This allows the life office to base all its
premiums on one set of tables.
8.4. RATING UP 141
A practical method of finding “r” is as follows. Suppose that the normal mortality table follows
Gompertz’ law, so there are constants B, c such that
µx+t = Bcx+t (0 ≤ t ≤ n)
(This is usually a fairly accurate representation of mortality for all ages over about 30.) Suppose
that the impairment is such that
for some constant, θ. (This formula is usually suitable for “medical” impairments; for accident risks
it is generally better to add a constant to the normal force of mortality.) Hence
µ∗x+t = (1 + θ)Bcx+t
= Bc(x+r)+t (0 ≤ t ≤ n)
where cr = 1 + θ, i.e.
log(1 + θ)
r= (8.4.2)
log c
In practice c is usually about 1.08, so we have the approximate rule
log(1 + θ)
r= (8.4.3)
log 1.08
log 1.4
r; = 4.37
log 1.08
For practical purposes this value of r would be rounded to 4 or 5 years. This rule is not precise,
but it may be sufficiently accurate in practice since the value of θ may in general be estimated only
rather approximately.
Example 8.4.1. A certain life office’s premium basis for lives accepted at normal rates is as follows:
Expenses: none
A certain impaired life, aged 50, is considered to be subject to the office’s normal mortality rates
for a life 10 years older. This life requires a contract providing the following benefits:
(i) £30, 000 at the end of the year of death, if within 10 years;
The impaired life finds that the level annual premium for this policy is the same for a policy providing
identical benefits for a life his age accepted at normal rates. Find S.
142 CHAPTER 8. EXTRA RISKS
Solution.
30, 000A50:10 + S D60
D50
normal premium =
ä50:10
µ Á ¶
D60
= 30, 000P50:10 + S ä50:10
D50
µ Á ¶
D70
“impaired life” = 30, 000P60:10 +S ä60:10
D60
8.5 Debts
Instead of paying higher premiums, a proposer accepted on special terms may elect to pay the normal
premiums with reduced death benefits. Such reductions in the death benefits are called “debts” or
“liens”, and may be either
We first give an example involving a level debt. (Note that a debt is a reduction in the death benefit
only: the maturity benefit is not reduced.)
Example 8.5.1. A life office calculates annual premiums for without profits endowment assurances
using the following basis:
Interest: 4%
(i) Calculate the annual premium payable for a 25 year endowment assurance taken out by a male
life aged exactly 45 for a sum assured of £50,000. The death benefit is payable at the end of
the year of death.
(ii) A man aged exactly 45 effects a policy identical to that in (i) above but rated up 7 years.
(a) Calculate the level extra premium, payable throughout the term of the policy.
(b) Alternatively, the life office offers to charge the standard premium but to impose a level
debt. Calculate the amount of the debt.
8.5. DEBTS 143
Solution.
(i) Let A.P. be P .
0.95P ä[45]:25 = 50, 000A[45]:25 + 0.45P
∴ P = £1, 492.41
(ii) (a) Consider a “normal” (select) life aged 52. Let P 0 be premium.
0.95P 0 ä[52]:25 = 50, 000A[52]:25 + 0.45P 0
∴ P 0 = 1, 752.32
∴ Extra premium = £259.91
(b) Let death benefit be reduced by D. The equation of value is
1
0.95P ä[52]:25 = 50, 000A[52]:25 − DA[52] :25 + 0.45P
∴ D = £12, 649
Now suppose that we are dealing with an n-year endowment assurance, and that the debt diminishes
in arithmetical progression to zero in the final policy year; that is, the debts run as follows:
(n − 1)D in year 1
(n − 2)D in year 2
..................
D in year n − 1
0 in year n
D{nA∗ [x]
1 ∗ 1
:n − (IA) [x]:n } (8.5.1)
where ∗ indicates impaired lives’ mortality; the death benefits are taken to be payable at the end of
the year of death (if these benefits are payable immediately on death, a bar should be placed over
the A symbols.) The calculations are illustrated in the next example.
Example 8.5.2. A life office issues 20-year without-profit endowment assurance policies providing
a sum assured of £8,000 on maturity or at the end of the year of earlier death. Level monthly
premiums are payable in advance throughout the term or until earlier death.
A certain impaired life aged 45 is considered by the office to have the mortality rates of a “nor-
mal” (select) life 5 years older.
Assuming that the ordinary premium appropriate to this life’s actual age is payable, calculate the
initial amount of the debt (i.e. the deduction from the sum assured) such that the debt reduces
uniformly each year to nil in the last policy year. Basis:
Interest: 4%
144 CHAPTER 8. EXTRA RISKS
P
Solution. Let be “normal” monthly premium. The equation of value for P is
12
(12) (12)
0.95P ä[45]:20 − 0.25P ä[45]:1 = 8000A[45]:20
8000 × [0.48051]
P =n h ³ ´i h ³
D[45]+1
´io
.95 ä[45]:20 − 11
24 1 − DD[45]
65
− 0.25 1 − 11
24 1− D[45]
3844.08
=
{0.95[13.507 − 0.45833 × 0.62253] − 0.25[1 − 0.45833 × 0.04014]}
3, 844.08
=
12.5606 − 0.24541
= £312.14 (£26.01 per month)
½ · µ ¶¸
D70
D= 8000 × 0.49664 − 312.14 × 0.95 × ä[50]:20 − 0.45833 1 −
D[50]
· µ ¶¸¾
D[50]+1
+0.25 × 312.14 1 − 0.45833 1 −
D[50]
. ½ 20(M ¾
[50] − M70 ) − (R[50] − R70 − 20M70 )
D[50]
3973.12 − 296.53[13.087 − 0.45833 × 0.66887] + 78.035[0.98111]
= 1
4581.322 [20 × 758.289 − 9360.22]
259.898
= = 205.09
1.2672
∴ Initial Debt is 19D = £3, 896.71
8.6. EXERCISES 145
Exercises
8.1 An explorer aged exactly 57 has just made a proposal to a life office for a whole life assurance
with a sum assured of £10,000 payable at the end of the year of death. For lives accepted at
normal rate, level annual premiums are payable until death under this policy.
The explorer is about to undertake a hazardous expedition which will last three years. The life
office estimates that during these three years the explorer will experience a constant addition of
0.02871 to the normal force of mortality, but after three years will experience normal mortality.
The life office quotes a level extra premium payable for the first three years.
Calculate this level extra premium on the following basis:
8.2 An impaired life aged exactly 55 wishes to effect a without profit endowment assurance for a
sum assured of £1,000 payable at the end of 10 years or at the end of the year of earlier death.
Level annual premiums are payable throughout the term of the policy.
Special terms are offered on the assumption that the life will experience mortality which can
be represented by:
(a) for the first five years, a constant addition of 0.009569 to the normal force of mortality,
and
(b) for the remaining five years, the mortality of a life 8 years older.
The life office quotes a level extra premium payable throughout the term. Calculate this level
extra premium. Basis:
8.3 A group of impaired lives now aged 40 experience mortality according to A1967-70 ultimate
with an addition to the force of mortality. The addition is 0.0005 at age 40, increasing linearly
to 0.0025 at age 60, at which level the addition remains constant.
Find the probability that an impaired life aged exactly 40
8.4 (i) Can you envisage circumstances under which an office could offer an impaired life who
wishes to pay the “normal” premiums a level debt but not a diminishing debt?
Hint Consider a life who is very severely impaired, and think of an approximate rela-
tionship between
(a) the level debt and
(b) the initial debt when debts decrease linearly to zero.
146 CHAPTER 8. EXTRA RISKS
(ii) A certain proposer, aged 40, for a 20-year endowment assurance by annual premiums
with sum assured £40,000 without profits (payable at the end of the year of death) is
considered by your office to be subject to the mortality of the normal table with an
addition of 10 years to the age. Your office’s basis for calculating premiums for “normal”
lives is
A1967-70 ultimate
4% p.a. interest
expenses of 6% of all premiums.
(a) The proposer asks to pay the annual premium for a “normal” life aged 40, and
suggests that the office should make the normal death benefits subject to a debt
which decreases in arithmetical progression to zero in the final year. Calculate the
initial debt.
(b) Suppose that the proposer also asks your office to quote a level debt. Calculate the
level debt.
8.5 The mortality of a certain impaired life aged exactly 60 may be represented as follows:
(1) at ages up to exact age 65, there is a constant addition of 0.009569 to the “normal” force
of mortality, which is that of A1967-70 select; and
(2) at ages between exact age 65 and exact age 70, mortality follows that of a “normal”
(ultimate) life 4 years older.
Suppose that this life requires a 10-year endowment assurance without profits with sum assured
£10, 000 payable on maturity or at the end of the year of earlier death. The life wishes, however,
to pay the “normal” level annual premium for his age. Calculate the level debt which would
be deducted from the normal death benefit, using the basis given below:
mortality: as given above
interest: 4% p.a.
expenses: 2 12 % of all premiums, with an additional initial expense of
1% of the sum assured (before deducting any debt.)
8.7. SOLUTIONS 147
Solutions
P ; 377.41
Let E be the extra premium charged for 3 years. Then, the equation of value will be (* denotes
impaired mortality):
(P + E)ä∗57:3 + P 3 |ä∗57 = 10, 000A∗57
where
ä∗57:3 3% = ä57:3 6% = 2.804
∗ l60
3 |ä57 = 1.06−3 ä60 at 3% = 11.076
l57
A∗57 = A∗ 57
1 ∗
:3 + 3 E57 A60 at 3%
= A∗57:3 − 3 E57
∗ ∗
+ 3 E57 A60 at 3%
l60
= 1 − d3% ä57:3 6% − 1.06−3 (1 − A60 at 3%)
l57
= 0.59572
Hence
10, 000A∗57 − P (ä∗57:3 + 3 |ä∗57 )
E= = 256.3207
ä57:3
at 4%
ä∗55:10 = ä∗55:5 + [5 E55
∗
]ä∗60:5
l60
= ä55:5 at 5% + v 5 at 5% [ä68:5 at 4%]
l55
= 7.677
(i)
½ Z 20 ¾
∗
20 p40 = exp − (µ40+t + 0.0005 + 0.0001t) dt
0
( · ¸20 )
1 2
= 20 p40 exp − 0.0005t + (0.0001t )
2 0
l60
= × exp {−(0.01 + 0.02)}
l40
= 0.895579 × 0.9704455 = 0.86911
∴ ans = 0.1309
(ii)
∗
30 p40 = 0.86911 × 10 p∗60
l70
= 0.86911 × × exp[−10 × 0.0025]
l60
= 0.66656
∴ Ans = 0.3334
(iii) 0.3334−0.1309=0.2025
8.4 (i) The debt cannot, of course, exceed the “normal” sum assured on death. The level debt
must be (very roughly) about one-half of the initial debt (since the average dim. debt is
one-half of the initial debt), so we may have cases in which:
402.119
∴D=
0.230719
= £1, 743
150 CHAPTER 8. EXTRA RISKS
Chapter 9
PROFIT-TESTING
(1) The sale of a without-profits policy is considered by the life office as an “investment” on
the part of the shareholders and/or with-profits policyholders. (In a mutual life office there
are no shareholders.) Similarly, the sale of a with-profits policy may be considered to be an
“investment” on the part of the office’s shareholders, though the position is complicated by the
fact that future profits are shared between the with-profits policyholders and the shareholders.
(2) The office calculates the expected net cash flow, and the expected profit, in each policy year in
two stages.
Stage A. The net cash flow in each policy year, and the profit after allowing for appropriate
reserves, of a given contract is calculated on the assumption that the policy is still in force at
the start of that year. The resulting net cash flow is called the in force net cash flow, and the
vector of profits (indexed by the policy year, t) is called the profit vector, which is written as
{(PROt )}.
Stage B. The estimated net cash flow per policy sold in policy year t is called the initial
net cash flow for that year, and is found by multiplying the “in force” net cash flow by the
probability that the policy is still in force at the start of the tth policy year, t−1 px . (We assume
here that the age at entry to assurance is x, and that a non-select mortality table is used.) The
corresponding vector of profits (after allowing for reserves) is called the profit signature, which
is written as {σt }.
(3) All cash flows are considered at the end of the policy year, so premiums less expenses are
accumulated to the end of the appropriate policy year at the rate of interest which the office
assumes it will earn on the life funds.
(4) Random fluctuations in mortality are ignored; that is, each policy is considered as if it were one
of a large number of identical contracts whose experience exactly follows the mortality table
used by the office in its projections.
151
152 CHAPTER 9. PROFIT-TESTING
Example 9.2.1. An endowment assurance policy, with sum assured £5000, term five years and level
annual premiums, is issued to a life aged 55.
Solution.
(i) Let the annual premium be P . Then
4
X
P ä55:5 = 5000A55:5 + 250 + 42(1.05)t−1 t p55 v t
t=1
(where all functions are on A1967-70 ultimate, 6%) , from which it follows that
5000 × 0.75171 + 250 + 152.62
P =
4.386
= £948.74
154 CHAPTER 9. PROFIT-TESTING
where column (8) gives the probability that the policy will be in force at the start of year t.
9.3. THE PROFIT VECTOR AND THE PROFIT SIGNATURE 155
Calculation of (PRO)t
We observe that
(PRO)t = t−1 V(1 + i) + (CF)t − px+t−1 .t V (9.3.4)
| {z } | {z } | {z }
accum. of “in force” money needed
reserves net cash for reserves
until end flow in at end of
of year year t the year
(since the chance that a given policy in force at the start of the policy year remains in force at the
end is px+t−1 .) This formula is sometimes written in the form
where
Example 9.3.1. Consider the policy described in example 9.2.1. Assuming that the premium and
experience basis are as described in that question and that the reserves held are as follows:
0 V=0
156 CHAPTER 9. PROFIT-TESTING
1 V=919
2 V=1,876
3 V=2,873
4 V=3,914
5 V=0
where column (8) is the probability that the policy will be in force at the start of year t.
9.4. THE ASSESSMENT OF PROFITS 157
Example 9.3.2. A life office issues a 3-year without profits endowment assurance policy to a life
aged 62. The sum assured of £1, 500 is payable on maturity or at end of the year of death, if within
3 years, and there are level annual premiums of £472.50 payable in advance.
The office uses the following “experience” basis:
renewal expenses: £5 at the beginning of the second and third policy years.
Solution. We first work out the “in force” net cash flows, (CF)t .
(1) One could work out the Internal Rate of Return (or yield) by solving the equation
n
X
v t σt = 0 (9.4.1)
t=1
n.p.v. of profits
, both at some rate of interest, im , say
n.p.v. of premiums
Pn
v t σt
= Pn−1t=1 , at rate im (9.4.3)
t
t=0 Pt+1t px v
Example 9.4.1. A life office issues a three-year non-profit endowment assurance policy to a man
aged 30. The sum assured is £60,000 on maturity or at the end of the year of earlier death. Level
premiums of £19,000 are payable annually in advance.
The office maintains reserves as follows:
reserve at end
policy year
of policy year
1 £19,000
2 £38,000
The office expects that its life funds will earn interest at 7% p.a. over the next 3 years.
The office expects expenses to be as follows:
initial expenses: 10% of the first year’s premium,
renewal expenses: 2% of later premiums.
Mortality is expected to follow A1967-70 ultimate.
Calculate
(i) the profit signature;
(ii) the net present value at the issue date of the profit to the office, using a risk discount rate of
10% p.a.
9.5. SOME THEORETICAL RESULTS ABOUT {σT } 159
Solution.
(i) We work out the net cash flows, (CF)t , per policy in force at the start of the policy year.
(1) (2) (CF)t
t (Pt − et ) (Pt − et )(1 + i) death + mat. costs = (1) − (2)
1 17,100 18,297 39.22 18,258
2 18,620 19,923 40.26 19,833
3 18,620 19,923 60,000 −40,077
−729
2, 238 × 0.999346 = 2, 237
583 × 0.998676 = 582
(ii)
−729v + 2, 237v 2 + 582v 3 at 10% = £1, 623
Theorem 9.5.2. If the premium basis and experience basis agree, the internal rate of return earned
by the office on a sale of a policy is equal to the rate i p.a. earned by the office’s life funds.
160 CHAPTER 9. PROFIT-TESTING
Now we have
n
X n
X
t
v σt = v t [t−1 px (CF)t + t−1 px .t−1 V(1 + i) − t px .t V]
t=1 t=1
But
n
X
v t t−1 px (CF)t = 0
t=1
as this is merely a restatement of the equation of value for calculating premiums. Hence
n
X n
X n
X
v t σt = v t t−1 px (1 + i)t−1 V − v t t px .t V
t=1 t=1 t=1
n−1
X n
X
= v r r px .r V − v t t px .t V
r=1 t=1
= 0, using the fact that 0 V = n V = 0
Corollary. If the life office wishes to obtain an I.R.R. on the sale of a policy in excess of the rate
of interest it believes it will earn on the life funds, premiums must be higher than those calculated
on the “experience” basis.
Example 9.5.1. What is the I.R.R. associated with the profit signature of example 9.3.1?
Solution. Since the premium and reserving bases agree, the I.R.R. must be i = 6% p.a.
9.6 Withdrawals
So far we have ignored the possibility of surrender. We shall now assume that surrenders may occur,
but only at the end of a policy year (just before payment of the premium then due). Let wt denote
the chance that a policy will be surrendered at the end of year t (t = 1, 2, . . . , n − 1). We may
assume that wn = 0, since the policyholder will receive the maturity benefits (if any) at that time.
Let (SV)t denote surrender value at time t. We observe that (PRO)t is as before, except that
there are additional profits due to the surrender of
Hence
Example 9.6.1. Consider the policy of Example 9.3.1, but now assume that at each of the durations
1,2,3,4 years, 5% of the surviving policyholders will surrender, and that the S.V. is 98% of the office’s
reserve. Find the new profit signature.
t (PRO)0t 0
t−1 px σt0
1 −211.87 1 −211.87
2 31.71 .99156 × .95 29.87
Hence we have:
3 54.99 .98222 × .952 48.75
4 79.13 .97191 × .953 65.93
5 102.97 .96054 × .954 80.56
Notes
1. The I.R.R. allowing for withdrawals may be found by solving the equation
n
X
v t σt0 = 0 (9.6.5)
t=1
2. Even if (PRO)0t = (PRO)t for t = 1, 2, . . . , n, {σt0 } is not the same as {σt } because the factors
{t−1 p0x } are not the same as {t−1 px }.
3. Unlike mortality rates, which are fairly predictable by actuaries, withdrawal rates depend to a
great extent on economic and commercial factors which are less easy to forecast accurately.
162 CHAPTER 9. PROFIT-TESTING
(1) If the policy is no longer in force at the beginning of policy year t, the profit emerging in that
year is zero;
(2) If the policy is still in force at the beginning of the year, the profit is
These formulae may be easily modified if the policy is altered, as in the following example.
Example 9.7.1. A life office issues a 5-year with-profits endowment assurance policy to a life aged
exactly 60. The policy has a basic sum assured of £10,000 payable at the end of the year of death
or at the maturity date. Level premiums are payable annually is advance throughout the term of
the policy. Simple reversionary bonuses vest at the start of each year, including the first.
The premium is calculated according to the following basis:
(ii) The office holds net premium reserves using a rate of interest of 3% per annum and A1967-70
ultimate mortality .
Calculate the profit signature for this policy, assuming that the office will earn interest at
7% per annum on its assets, mortality follows the A1967-70 ultimate table, and expenses and
bonuses will follow the premium basis.
9.7. THE ACTUAL EMERGENCE OF PROFITS 163
(iii) Immediately before the fourth premium was due (before the fourth bonus declaration) the
policy was made paid-up, with no entitlement to further bonuses. The paid up sum assured
was 60% of the benefits immediately before the alteration, including declared bonuses.
The policyholder survived to the maturity date. Interest was earned on the life funds was at
6% per annum over the period of the contract, and bonuses in the first three years followed
the premium assumptions. Expenses followed the premium assumptions up to the alteration
date, and no expenses were incurred after the policy was made paid-up.
For each of the five years of the policy term, calculate the actual year-end profit earned on
the policy.
Solution.
(i) Let P be the annual premium. M.P.V. of premiums less expenses is
M.P.V. of benefits is
· ¸
R[60] − R65 − 5M65 + 5D65
10, 000A[60]:5 + 400(IA)[60]:5 = 10, 000 × 0.82565 + 400
D[60]
= 9867.88
Hence
P = £2, 627.02
Exercises
9.1 (i) In the context of profit-testing, explain the difference between the “profit vector” and
the “profit signature”.
(ii) A certain life office sells assurance policies with term 3 years to lives aged 70. For each
policy, the profit vector is estimated to be (−50, 30, 30). Given that the mortality of the
policyholder is expected to follow A1967-70 ultimate, calculate
(a) the profit signature per policy sold;
(b) the net present value of the profit to the office on the basis of a risk discount rate of
8% per annum.
9.2 A life office issues a 5-year guaranteed bonus endowment assurance policy to a life aged 60,
with basic sum assured £30,000. The sum assured, with attaching bonuses, is payable at the
end of the year of death or at maturity. Level premiums are payable annually in advance.
The office holds net premium reserves, using the basis A1967-70 ult at 3% p.a. Interest on
premiums and reserves is expected to be earned at an effective rate of 8% p.a. Bonuses will be
declared annually at a rate of 3% of the basic sum assured.
Bonuses vest at the start of each policy year. Expenses of 40% of the first year’s premiums and
5% of subsequent years’ premiums will be incurred. Mortality is expected to follow A1967-70
ultimate. Withdrawals may be ignored.
(i) For each policy year, calculate the total sum assured.
(ii) For each duration t = 1, 2, 3, 4 years, calculate the reserve (t V) immediately before pay-
ment of the premium then due, given the following values on A1967-70 ultimate, 3% p.a.
interest:
(IA)60:5 = 4.1853
(IA)61:4 = 3.4681
(IA)62:3 = 2.6973
(IA)63:2 = 1.8672
(IA)64:1 = v = 0.97087
(iii) (Difficult.) Calculate the annual premium (P ) required for the shareholders/with profits
policyholders to achieve an internal rate of return of 12% p.a. on the sale of this contract.
9.3 Ten years ago a life office issued a large block of 10-year without profits endowment assurances
to lives then aged 30. Each policy was effected by annual premiums, and had a sum assured
of £40,000, payable on survival or at the end of the year of death. The office’s premium basis
was
A1967-70 ultimate
5% interest
expenses of 2% of all premiums with additional initial expenses of 0.5% of the sum assured.
It was found that mortality , interest and expenses followed these assumptions, but there were
surrenders just before payment of the premiums due at durations 1,2 and 3 years. At each of
these times, 3% of the surviving policyholders surrendered their contracts, and were given a
surrender value equal to the office’s reserve, which was calculated on the premium basis, minus
a surrender penalty of £40, which the office transferred to the surplus account. By using a
profit-testing approach, or otherwise, calculate the surplus accruing to the office at the end of
each of the first three policy years, per policy sold.
Hint. Note that (PRO)t = 0, so profits arise only from surrenders.
166 CHAPTER 9. PROFIT-TESTING
9.4 Your office is considering the issue of 3-year annual-premium endowment assurance policies
without profits to lives aged 62. In respect of a policy with sum assured £10, 000, payable at
the end of the year of death (if within 3 years) or on maturity, calculate the net present value
of the profit signature on the following assumptions:
premium basis: mortality: A1967-70 ultimate
interest: 6% p.a.
expenses: 3% of all premiums.
reserve basis: net premium method using A1967-70 ultimate,
4% p.a. interest
rate of interest to
be earned in life fund: 8% p.a.
expenses: 3% of office premiums
mortality: A1967-70 ultimate
risk discount rate: 10% p.a.
9.5 A life office issues 3-year term assurance policy to a man aged exactly 59. The sum assured
is £15, 000, payable at the end of the year of death. Level premiums are payable annually in
advance. Expenses are expected to be as follows:
initial expenses: £10
renewal expenses: £2 incurred at the beginning of the 2nd and each subsequent policy year.
It is assumed that interest of 7% per annum will be earned on the life funds, and that mortality
follows the A1967-70 ultimate table. The risk discount rate used by the office is 15% per annum.
The office calculates the annual premium by requiring that the net present value of the expected
profit on each policy is equal to 20% of one office premium. Calculate the office premium on
each of the following reserving bases:
(i) The office holds zero reserves at each year-end.
(ii) The office sets up a reserve at each year-end (except the last) equal to 80% of one office
premium.
9.9. SOLUTIONS 167
Solutions
9.1 (i) The profit vector refers to expected profits per policy in force at start of year, whilst the
profit signature refers to profits per policy sold. Relationship is
σt = (PRO)t × prob. that policy is in force at time t − 1
(b) N.P.V. = σ1 v + σ2 v 2 + σ3 v 3 at 8%
9.2 Note: In (ii), we need the increasing assurance factors (A65+t:5−t and ä65+t:5−t are directly
tabulated)
where
P1 = net premium on A67-70 ult. 3%
30, 000A60:5 + 900(IA)60:5
=
ä60:5
30, 000 × 0.86682 + 900 × 4.1853
= = £6511.67
4.572
(1)prs. less exp., with interest (2)death costs (3)survival costs (1)−(2)−(3)
t (P − et )(1.08) q59+t Dt p59+t St (CF)t
1 0.648P 445.96 0 0.648P − 445.96
2 1.026P 509.23 0 1.026P − 509.23
3 1.026P 580.42 0 1.026P − 580.42
4 1.026P 660.39 0 1.026P − 660.39
5 1.026P 34500 1.026P − 34, 500
| {z }
since sum paid on
death or survival
We now work out (PRO)t :
i.e.
(0.648P − 6707)v
+(1.026P − 6388)(0.985568)v 2
+(1.026P − 6057)(0.969785)v 3
+(1.026P − 5717)(0.952572)v 4
+(1.026P − 5358)(0.93385)v 5
=0
(Rough check: right order of magnitude:- P =' P1 , and 5P ' 34, 500 (maturity benefit)).
9.3 Profits arise only in respect of surrenders at times 1,2,3. The profit at time t years (t = 1, 2, 3)
per policy in force at start of year is
(PRO)0t = (1 − q30+t−1 )(0.03) (40) = 1.2p30+t−1
| {z }
surrender
penalty
= 3, 146.38
µ ¶
5.138
2 V = 10, 000 1 −
6.519
= 6, 472.66, 3V =0
Net cash flows (ignoring reserves)
t (CF)t
1 0.97P (1.08) − 10, 000q62 = 3, 090.91
2 0.97P (1.08) − 10, 000q63 = 3, 071.86
3 0.97P (1.08) − 10, 000 = −6, 731.59
STATIONARY POPULATIONS
171
172 CHAPTER 10. STATIONARY POPULATIONS
Define
Z ∞
Tx = lx+t dt
Z0 ∞
= ly dy (on setting y = x + t)
x
R∞
◦ x
ly dy
ex =
lx
Tx
=
lx
and hence
◦
Tx = lx ex (10.1.4)
A stationary population is an idealised large population such that (in an old-fashioned mathe-
matical notation) the number of lives in the population between ages x and x + dx is always equal
to
klx dx
where k is called the scaling factor and lx is according to some life table.
By integrating over all ages greater than or equal to x, we can see that
Z ∞
Total population aged ≥ x = kly dy
x
= k.Tx (10.1.5)
(It is nearly always advisable to draw a diagram in stationary population questions.) In any time-
interval of length t years, the number of lives attaining a specified exact age x is klx .t, so the rate at
which lives attain age x is klx per annum, i.e. klx lives ‘flow’ continuously over exact age x each year.
Note that (if lives are considered at all ages) there will be kl0 births each year. Also, the num-
ber of deaths each year between ages x1 and x2 (x1 < x2 ) is
The number of lives in the population between any two ages x1 and x2 (x1 < x2 ) is
Example 10.1.1. The membership of a certain learned society is stationary at 11,500. Members
enter only at exact age 50. They are subject to the mortality of English Life Table No.12 - Males,
and there are no withdrawals.
Solution
One knows that
k.T50 = 11, 500
So
◦
kl50 e50 = 11, 500
Example 10.1.2. A large company has for many years maintained a staff in a stationary condition
by recruiting 500 annual entrants at exact age 20, uniformly over the year. If the staff retire at age
60, there are no withdrawals, and English Life No. 12 – Males mortality is experienced, find:
(a) the size of the staff,
(b) the number of staff who retire each year,
(c) the number of pensioners.
Solution
(a) We have
kl20 = 500.
= kl60
µ ¶
500
= l60
l20
= 410
= kT60
µ ¶
500 ◦
= l60 e60
l20
= 6, 172.
174 CHAPTER 10. STATIONARY POPULATIONS
◦
ex:n = average lifetime lived by (x) between ages x and x + n
= E(T ∗ )
where the random variable T ∗ is defined as follows:
(
∗ T if (x) dies within n years
T =
n if (x) lives for n years.
The variable T ∗ has p.d.f. t px µx+t for 0 < t < n, and there is discrete probability n px that T ∗ = n.
Hence
Z n
∗
E(T ) = t.t px µx+t dt + n.n px
0
Z n
n
= [−t.t px ]0 + t px dt + n.n px (by integration by parts)
0
Z n
= t px dt
0
Hence
◦ Tx − Tx+n
ex:n =
lx
◦ lx+n ◦
=ex − ex+n (10.1.9)
lx
◦
(Notice that āx:n =ex:n if i = 0 .)
Similarly,
Z 10 Z 11
◦ lt+1 lx
e1:10 = dt = dx
0 l 1 1 l1
Z 11 · ¸
log x
= 1− dx
1 9 log 10
1 x=11
= 10 − [x log x − x]x=1
9 log 10
11 log 11 − 10
= 10 −
9 log 10
= 9.2097.
The average age at death of those who die between ages x and x + n may be worked out as
follows. We have
so that
◦
ex:n = E(T ∗ |T ∗ < n)n q x + n.n px
Hence the average age at death of those who die between ages x and x + n is
x + E(T ∗ |T ∗ < n)
◦
ex:n −n.n px
=x +
n qx
Tx − Tx+n − nlx+n
=x + (10.1.11)
lx − lx+n
Note that this result refers to a randomly-chosen life aged x: we need not assume that there is a
stationary population.
Note that
R1
0
lx+t µx+t dt
mx = R1 ' µx+ 12 (10.2.2)
0
lx+t dt
δ.F = B − C (10.4.1)
iF = B − C (10.4.2)
Note: In case (b), F denotes the fund at the start of the year.
10.4. STATIONARY FUNDS 177
Example 10.4.1. Each year for many years a life office has issued 10,000 temporary assurance
policies each with a term of ten years and a sum assured of £5,000 to lives aged 25 exactly.
One third of those who survive to age 35 then effect a without profits whole life policy for the
same sum assured as the term policy, and one quarter effect a 25-year without profits endowment
assurance for twice that sum assured. All premiums are payable annually in advance and death
claims are paid at the end of the year of death. Policies are issued uniformly throughout the year.
The office calculates premiums on A1967-70 ultimate 4%, ignoring expenses. If the office’s ex-
perience follows this basis, calculate the size of the fund held for these contracts.
Solution
The term assurance premium = 5, 000 M25 −M35
N25 −N35 = £3.29.
Whole life premium = 5, 000P35 = £56.85.
Endowment assurance premium = 10, 000P35:25 = £245.30.
Annual premium income is
· ¸
l35
10, 000 1 + e25 − (1 + e35 ) × 3.29
l25
l35 1
+ 10, 000 × (1 + e35 ) × 56.85
l25 3
· ¸
l35 1 l60
+ 10, 000 × 1 + e35 − (1 + e60 ) × 245.30
l25 4 l35
= 328, 003 + 7, 634, 065 + 14, 769, 057
= £22, 731, 125.
where i = 0.04.
18992591
Therefore Fund = δ = £484, 249, 000.
178 CHAPTER 10. STATIONARY POPULATIONS
Exercises
10.1 The staff of a large company is maintained as a stationary population by 500 new entrants
each year at exact age 20. One third of those reaching age 30 leave immediately. Of the
remainder, 41 of those attaining age 60 retire immediately and the survivors retire at age 65.
The only other decrement is death.
Calculate
(i) the number of staff,
(ii) the number of deaths in service each year.
Basis: English Life Table No. 12 - Males
10.2 A certain country’s school system provides education for all children between the ages of 5
and 16 exactly. The country’s population is stationary, there being 100,000 births per year,
uniformly distributed over the year. The population of the country is subject to the mortality
of English Life Table No. 12 - Males.
(i) Find the number of pupils at any given time.
(ii) The country’s teacher training colleges are such that a constant flow of new entrants to
the profession is maintained. Teachers are recruited uniformly over the year, and the ratio of
pupils to teachers is 20 to 1. All teachers enter the profession at age 21 and retire at age 60,
there being no withdrawals. Find the annual number of new teachers recruited.
10.3 For many years a company has recruited, uniformly over each year, 200 employees on their
20th birthdays and a fixed number of additional employees on their 25th birthdays. Mor-
tality has followed English Life Table No. 12 - Males. Employees may retire on their 60th
or 65th birthdays, and one third of employees reaching their 60th birthdays retire on that
date. Employees leave the company only through death or retirement, and the total number
of employees is 10,000.
10.6 In a certain country, all civil servants are recruited on their 25th birthdays. (Birthdays are
assumed to occur uniformly over the calendar year). Of those who reach exact age 40, 10%
obtain employment in private companies and leave the civil service immediately. All remaining
civil servants retire when they reach age 60 or die in service before this age. Civil servants
have for many years experienced the mortality of English Life Table No. 12 - Males, and will
do so for the indefinite future. The population of civil servants has been stationary for many
years, and civil servants are recruited at the rate of 3,000 each year.
(i) How many civil servants are in service at any given date?
(ii) The government of the country has just decided to reduce the size of the civil service by
10%. This is to be done by immediately lowering the retirement age of civil servants. Find
the new retirement age (to the nearest month), assuming that there are no other changes.
10.7 In a mortality table with a one-year select period, the following relationships apply at age x:
1
l[x] = (lx + lx+1 )
2
t q[x] = t.q[x] (0 ≤ t ≤ 1)
Solutions
1
(ii) Number of teachers = 20 × 1, 066, 409 = 53, 320.
Let annual number of recruits be c.
Then the equation is
c
(T21 − T60 ) = 53, 320
l21
Hence c = 1, 431 teachers per annum.
10.3 kl20 = 200 employees enter at age 20 per annum. Let cl25 be the number who enter at age 25
per annum.
Then the number µof employees is ¶ µ ¶
1 2 1 2
k T20 − T60 − T65 + c T25 − T60 − T65 = 10, 000
3 3 3 3
where k = 200
l20 .
Solving this for c gives
= 1, 000
Number of staff = k(T20 − T21 ) + 0.8k(T21 − T35 ) + 0.4k(T35 − T40 )
1000
= (T20 − 0.2T21 − 0.4T35 − 0.4T40 )
l20
= 14, 060
10.7 (i)
lx+1 lx+1 2px
p[x] = = 1 =
l[x] 2 (lx + lx+1 ) 1 + px
(ii)
lx+1 + lx+2 + ...
e[x] =
l[x]
µ ¶
lx p[x] 2
= · ex = · ex = ex
l[x] px 1 + px
(iii) Z ∞
◦ 1
e[x] = [ l[x]+t dt]
l[x] 0
Z 1 Z ∞
1
= [ l[x]+t dt +
lx+t dt]
l[x] 0 1
1 1 ◦
= [ (l[x] + lx+1 ) + lx+1 . ex+1 ]
l[x] 2
(iv)
R1
l[x]+t µ[x]+t dt
0
m[x] = R1
l
0 [x]+t
dt
l[x] − lx+1
= 1
2 (l[x] + lx+1 )
1 − p[x]
= 1
2 (1 + p[x] )
2(1 − px )
= (using part (i))
1 + 3px
10.8 (i) Let P 0 be the office annual premium per £1 sum assured.
We have
0.95P 0 ā55:10 = A55:10 + 0.25P 0 ā55:1
Hence P 0 = 0.100663.
The total sums assured in force at any given time are
5, 000, 000
(T55 − T65 )
l55
5, 000, 000
(£5,000,000 “flows across” age 55, so the scaling factor is .)
l55
µ ¶
◦ l65 ◦
Total sums assured = 5, 000, 000 e55 − e65
l55
= £45, 618, 854.
10.6. SOLUTIONS 183
l65
(ii) 5, 000, 000 = £3, 985, 870
l55
JOINT-LIFE FUNCTIONS
Remark
If (x) and (y) are a married couple, or business partners, the assumption of independence of T1 , T2
is questionable: consider, for example, a car accident in which both lives are killed. But in practice
independence is normally assumed.
Define
I II
t pxy = P r{both (x), (y) will survive for t years at least}
= t pIx · t pII
y (by independence of T1 , T2 ) (11.1.1)
and
I II
t q xy = P r{T ≤ t}
= the distribution function of T (11.1.2)
I II
=1− t pxy (11.1.3)
t pxy = t px · t py
185
186 CHAPTER 11. JOINT-LIFE FUNCTIONS
lxy = lx ly
t q xy = 1 − t pxy
= 1 − t px .t py
= 1 − (1 − t q x )(1 − t q y )
= t q x + t q y − t q x .t q y (11.1.4)
h qxy
µxy = lim
h→0+ h
h q ³ q ´ ³ q ´ i
h x h qy h x h y
= lim + − · h
h→0+ h h h h
= µx + µy (11.1.5)
h qx q
(since h → µx , hhy → µy as h → 0+ ).
Example 11.1.1.
calculate the probability that, of two lives aged x and y, only one will survive for 10 years.
Solution
lx+10:y
10 px = = 0.96
lx:y
lx:y+10
10 py = = 0.92
lxy
Theorem
Z t
t pxy = exp(− µx+r:y+r dr) (11.1.6)
0
f (t) = p.d.f. of T
= F 0 (t)
d
= [1 − t pxy ] (for t > 0)
dt
· µ Z t ¶¸
d
=− exp − µx+r:y+r dr (from (11.1.6)).
dt 0
· Z t ¸
= exp − µx+r:y+r dr µx+t:y+t (using Chain Rule)
0
(
t pxy µx+t:y+t (t > 0)
Thus f (t) = (11.1.7)
0 (t < 0)
where K1 = integer part of T1 , K2 = integer part of T2 . This variable is discrete, with probabilities
which leads to
l[x]+r+t:[y]+u+t
t p[x]+r:[y]+u =
l[x]+r:[y]+u
We also define ³ q ´
h [x]+r:[y]+u
µ[x]+r:[y]+u = lim = µ[x]+r + µ[y]+u
h→0+ h
The p.d.f. of T = future lifetime of ([x] + r, [y] + u) is
(
t p[x]+r:[y]+u .µ[x]+r+t:[y]+u+t (t ≥ 0)
f (t) =
0 (t < 0).
t pxyz = P r{(x), (y) and (z) all survive for at least t years}
= t px .t py .t pz
lx+t:y+t:z+t
=
lxyz
11.3. EXTENSIONS TO MORE THAN 2 LIVES 189
where lxyz = lx ly lz for all x, y, z ≥ α1 , α2 , α3 (α1 , α2 , α3 being the youngest ages in tables I, II, III).
Other functions follow similarly; in particular
h qxyz
µxyz = lim+ = µx + µy + µz
h→0 h
and T = min{T1 , T2 , T3 } has p.d.f.
(
t pxyz µx+t:y+t:z+t = t pxyz (µx+t + µy+t + µz+t ) (t ≥ 0)
f (t) =
0 (t < 0)
Example 11.3.2. The probability that exactly one of three lives aged x will survive for n years is
27 times the probability that all three will die within n years. Find the probabilities that
(a) at least two will survive n years
(b) at least one will die within n years
Solution
(a) 3n px (1 −n px )2 = 27(1 −n px )3
thus n px = 9(1 −n px )
hence n px = 0.9, so n qx = 0.1.
190 CHAPTER 11. JOINT-LIFE FUNCTIONS
Hence Z ∞
◦
exy = t.t pxy .µx+t:y+t dt
0
We also have
◦ 1
exy 'exy −
2
The variance of T is calculated as follows:
1
(the proof is similar to single-life case; note that v = 1+i ). Note also that
∞
X
axy = v t .t pxy = äxy − 1
t=1
äxy:n = äxy −n pxy v n äx+n:y+n
Woolhouse’s approximation for annuities payable mthly hold for joint lives, i.e.
m−1
ä(m)
xy ' äxy −
2m
(m) m−1
äxy:n ' äxy:n − (1 − n pxy v n )
2m
The Euler–MacLaurin formula gives
1 1
āxy ' äxy − ' axy + .
2 2
tables are
1
Dxy = v 2 (x+y) lxy (11.5.3)
X∞
Nxy = Dx+t:y+t (11.5.4)
t=0
and
∞
X lx+t:y+t
äxy = vt
t=0
lxy
Nxy
= .
Dxy
Note: In A67 − 70 we only find Dxx and Nxx (i.e. equal ages only.)
(b) Assurances
Āxy = 1 − δāxy .
We also have
∞
X
Axy = E[v K+1 ] = v t+1 t |qxy (11.5.6)
t=0
So Axy = 1 − d.äxy
Example 11.5.1. Show how one can evaluate Ā 1 on the a(55) tables (male/female.)
z}|{
xy :n
Solution. Method 1:
1
Ā 1 ' (1 + i) 2 A 1
z}|{ z}|{
xy :n xy :n
1
= (1 + i) 2 [Axy:n − v n ·n pxy ] (as in single-life case).
½ ¾
1
n lx+n ly+n
= (1 + i) 2 1 − däxy:n − v ·
lx ly
lx+n ly+n
Now use äxy:n = äxy − v n · äx+n:y+n
lx ly
and finally use äxy = axy + 1.
Method 2:
Ā 1 = Āxy:n − v n ·n pxy
z}|{
xy :n
lx+n ly+n
= (1 − δāxy:n ) − v n · ·
lx ly
· ¸
1 lx+n ly+n 1
' 1 − δ(axy + ) − v n · · 1 − δ(ax+n:y+n + ) .
2 lx ly 2
(c) Premiums
We employ equations of value in the same way as for single-life policies. If there are no expenses we
obtain net premiums, e.g., in the International Actuarial Notation,
Āxy 1
P̄ (Āxy ) = = −δ
āxy āxy
If there are expenses, calculate office premiums as for single-life policies by setting up an equation
of value. The general symbols P and P 0 may be used for any net or gross premium.
194 CHAPTER 11. JOINT-LIFE FUNCTIONS
(d) Reserves
Calculated as for single-life policies, e.g.,
Exam tables
(i) A67-70 gives:
Example 11.5.2. Two lives, aged 50 and 60, effect a 10-year temporary joint life assurance with
sum assured £60,000. The sum assured is payable immediately on the first death, provided that
this occurs within 10 years of the issue date. The policy has annual premiums, payable during the
joint lifetime of both lives for at most 10 years.
1
(i) Express Āz }| as an integral. Estimate the value of the integral by the following form of
{
50 : 60:10
Simpson’s rule: Z 10
10
f (t) dt = · [f (0) + 4f (5) + f (10)]
0 6
(ii) Hence, or otherwise, estimate the values of A50:60:10 and ä50:60:10 , and calculate the annual
premium for the policy.
Basis: A1967 − 70 ultimate mortality
4% p.a. interest
No expenses.
11.6. LAST SURVIVOR PROBABILITIES (TWO LIVES ONLY) 195
Solution R 10
1
(i) Āz }| { = 0 v t .t p50:60 (µ50+t + µ60+t ) dt.
50 : 60:10
Simpson’s rule gives an approximate answer of
Āz 1 = 0.2238.
}| {
50 : 60:10
(ii)
l60 l70
A50:60:10 = Az 1
}| { + v 10 ·
50 : 60:10 l50 l60
1 l70
' (1.04)− 2 .Āz 1
}| { + v 10
50 : 60:10 l50
= 0.70792.
1 − A50:60:10
ä50:60:10 = = 7.594.
d
The premium P p.a. is
Āz 1
}| {
50 : 60:10
P = 60000 = £1, 768.24
ä50:60:10
t pxy = t px + t py − t pxy
and hence
t q xy = t q x + t q y − t q xy
These formulae are true even if (x), (y) are not independent. If they are independent, we may write
t pxy = t px · t py
giving
t pxy = t px + t py − t px · t py (11.6.1)
and
t q xy = t qx · t qy . (11.6.2)
Note that t q xy is the distribution function of the variable
Tmax = max{T1 , T2 }.
d d d d
f (t) = (t q ) = (t q x ) + (t q y ) − (t q xy )
dt xy dt dt dt
196 CHAPTER 11. JOINT-LIFE FUNCTIONS
where
This gives
t px µx+t + t py µy+t − t px .t py (µx+t + µy+t )
µxy (t) =
t pxy
Note: these formulae hold even without the assumption of independence of the lives.
(b) Assurances
Taking expected values on each side of the relationship
one obtains
Similarly, when the sum assured is payable at the end of the year of death, we have
and hence
Axy = Ax + Ay − Axy (11.7.3)
(which is still true even when the lives are not independent.)
Note also the conversion relationships
Axy = Ax + Ay − Axy
= (1 − däx ) + (1 − däy ) − (1 − däxy )
= 1 − däxy
Solution
A40:40 = 1 − dä40:40
= 1 − d[2ä40 − ä40:40 ]
= 1 − d[2 × 18.894 − 17.052]
= 0.20246
(c) Premiums.
We may use an equation of value to calculate net or gross premiums. For example, using the
International Actuarial Notation, we have
Pxy = net annual premium payable annually in advance for a policy providing
£1 at the end of the year of the death of the second to die of (x), (y)
Axy
=
äxy
Using the conversion relationships given above, the following equations can be derived:
Āxy 1
P̄ (Āxy ) = = −δ
āxy āxy
and
1
Pxy = − d.
äxy
198 CHAPTER 11. JOINT-LIFE FUNCTIONS
We find that
£ ¤
t V1 = S Ax+t:y+t − Pxy äx+t:y+t
t V2 = S [Ax+t − Pxy äx+t ]
t V3 = S [Ay+t − Pxy äy+t ]
Theorem When the first life dies, the prospective reserve will increase (from t V1 to t V2 or t V3 ).
Proof
£ ¤
t V1 = S (1 − däx+t:y+t ) − Pxy äx+t:y+t
£ ¤
= S 1 − (Pxy + d)äx+t:y+t
< S [1 − (Pxy + d)äx+t ] (as äx+t < äx+t:y+t )
= S [(1 − däx+t ) − Pxy äx+t ]
= S [Ax+t − Pxy äx+t ]
= t V 2 ( and similarly for t V3 )
If it is not known whether one or both of (x), (y) survive at time t (and in practice this may be
the case since one death does not result in a claim or a change in the premium), one may value the
policy as a ‘weighted average’ of t V 1 , t V 2 and t V 3 , with weights proportional to the probabilities
t pxy , t px (1 − t py ) and t py (1 − t px ) respectively. Hence the weights are
t pxy t px (1 − t py ) t py (1 − t px )
, , respectively
t pxy t pxy t pxy
Finally, it can be shown that the retrospective reserve is equal to the weighted average reserve,
i.e.
1
V = S Pxy äxy:t − A 1 .
v t .t pxy z}|{
xy :t
11.8. RESERVES FOR LAST SURVIVOR ASSURANCES 199
Example 11.8.1. A last-survivor policy provided £10, 000 immediately on the death of the sec-
ond to die of a man aged 65 and his wife aged 60. Premiums were payable monthly in advance so
long as at least one of the couple survived. The office which issued the policy used the following basis:
(12) 11
ä m ' ä m f − = 7.228
65:60
f
65:60 24
(12) 11
ä m ' ä m − = 7.942
65 65 24
(12) 11
ä f ' ä f − = 9.836
60 60 24
(12) 11 11
Hence ä m ' ä m f − = 11.0075 − = 10.549
65:60
f
65:60 24 24
Am f
= 1 − dä m f
= 0.18463
65:60 65:60
1
10, 000 × (1.08) 2 × 0.18463
0
Hence P = = £186.55
0.975 × 10.549
Therefore monthly premium = £15.55 .
1 (12)
(ii) Reserve = 10000(1 + i) 2 Af60 − 0.975P 0 a f
60
= 2, 476.96 − 1, 773.82
(12) 11
(using Af60 = 1 − d.äf60 , and a f = af + )
60 60 24
Hence Reserve = £694.15.
200 CHAPTER 11. JOINT-LIFE FUNCTIONS
Exercises
11.1 Given that n px = 0.3, n py = 0.4, n pz = 0.6, find the probability that, of the lives (x), (y) and
(z),
(a) none will survive n years
(b) exactly one will survive n years
(c) at least one will survive n years.
11.2 Prove that
(IA)xy = äxy − d.(Iä)xy
11.3 Express in terms of px , py , and pz the probabilities that, of three lives (x), (y) and (z),
(a) all three will survive one year
(b) at least one will survive one year
(c) exactly two will survive one year
(d) at least two will survive one year
11.4 Derive the formula
∞
X
exy = E(K) = t pxy where K = integer part of T (T = min{T1 , T2 })
t=1
Solutions
11.5
Ā75:75 = 2Ā75 − Ā75:75
1 1
' 2(1.04) 2 A75 − (1.04) 2 [1 − d.ä75:75 ]
= 0.64676.
202 CHAPTER 11. JOINT-LIFE FUNCTIONS
CONTINGENT ASSURANCES
Let
T1 = future lifetime of (x) with pdf f1 (t1 ) = t1 px µx+t1 (t1 > 0)
T2 = future lifetime of (y) with pdf f2 (t2 ) = t2 py µy+t2 (t2 > 0)
Then
t q1 = P r{T1 ≤ t and T1 ≤ T2 }
xy
ZZ
= f1 (t1 )f2 (t2 ) dt1 dt2
0<t1 ≤t
t1 ≤t2
Z t ·Z ∞ ¸
= t1 px µx+t1 · t2 py µy+t2 dt2 dt1
0 t1
Z t
= t1 px µx+t1 · t1 py dt1
0
Z t
= r pxy µx+r dr (12.1.1)
0
205
206 CHAPTER 12. CONTINGENT ASSURANCES
t q1 + tq 1 = t q xy (12.1.3)
xy xy
One may also define deferred contingent probabilities, and we find that
t |q 1 = t pxy .q 1
x:y x+t:y+t
and t |∞ q 1 = t pxy .∞ q 1
x:y x+t:y+t
Z t
t q2 = (1 − r py )r px µx+r dr
xy 0
= t qx − t q1
xy
t qx = t q1 + t q2 (12.1.4)
xy xy
Example 12.2.1. Using A1967-70 mortality and 4% interest estimate, by approximate integration,
the value of
Ā 1
50:60
[Assume l108 = 0, so that the integral is over a range of 48 years. Break this into 4 sub-intervals and
use Simpson’s rule over each.]
Solution
Z ∞
Ā 1 = v t t p50 µ50+t t p60 dt (at 4% interest)
50:60 0
Z 48
= f (t) dt
0
(since there is a 50% chance that the first “x” of (x, x) will be the first to die).
We may therefore evaluate contingent assurances on 2 equal ages from tables of Āxx . As expected
by general reasoning, we also have
Axy = A 1 + A 1 (12.2.4)
xy xy
Example 12.2.2. (a) Estimate, by approximate integration, the value of a contingent assurance of
£100,000 payable immediately on the death of a female aged 60, provided a male aged 50 is still alive
and provided her death occurs within 15 years. The mortality of the female follows a(55) ultimate
(females), and that of the male follows A1967-70 ultimate. The interest rate is 7 21 % per annum and
expenses are ignored.
Note. A very accurate approximation to the value is not expected.
(b) It has been suggested that a policy providing the above benefit should be issued by annual
premiums ceasing on the death of the female life or after 15 years, whichever is earlier. State with
reasons whether you agree with this suggestion. If you do not, suggest a more suitable premium-
paying term, other than issuing the contract by a single premium.
208 CHAPTER 12. CONTINGENT ASSURANCES
Solution
(b) No: premiums must also cease on death of (50), otherwise they may still be payable with no
prospect of claim: negative prospective reserve. Restrict premium-paying period so that premiums
cease on first death or after 15 years, if earlier.
Hence
We may also encounter temporary contingent assurances payable on the second death, e.g. the
m.p.v. of £1 payable immediately on death of (x) within n years, provided that (x) dies after (y),
equals
Z n
v t t px µx+t (1 − t py ) dt
0
= A1 − A1
x:n xy:n
Example 12.2.3. Using the fact that Ā1 + Ā 1 = Āxy , find on A1967 − 70 ultimate at 4% interest
xy xy
the values of (i) Ā 1 , (ii) Ā 2 and (iii) Ā 1 .
60:60 60:60 60:60:10
12.2. CONTINGENT ASSURANCES 209
Solution.
(i) Ā 1 = 0.5Ā60:60
60:60
1
' 0.5(1.04) 2 [1 − dä60:60 ]
' 0.31491
Ā 2 = 0.21259
60:60
1 1
(iii) Ā 1 = Āz }| {
60:60:10 2 60 : 60:10
· ¸
1 1 D70:70
' (1.04) A60:60 −
2 A70:70
2 D60:60
= 0.15513
These are the same as Ā1 , etc., but with the benefit payable at the end of the year of death.
xy
An exact expression is
∞
X
A1 = v t+1 t pxy q 1 (12.2.7)
xy x+t:y+t
t=0
Note the following results, similar to those given above for A 1 , etc.:
xy
Axy = A 1 + A 1
xy xy
Ax = A1 + A2
xy xy
1
A1 = Axx
xx 2
1
A2 = Axx
xx 2
Z = present value of £1 payable on death of (x), if this occurs before the death of (y)
(
1
v T1 if T1 ≤ T2 where v = 1+i
=
0 if T1 > T2
E(Z) = Ā1
xy
210 CHAPTER 12. CONTINGENT ASSURANCES
The variance of Z is
E(Z 2 ) − [E(Z)]2
where
· 2 T ¸
(v ) 1 if T1 ≤ T2
E(Z 2 ) = E
0 if T1 > T2
· ∗ T ¸
(v ) 1 if T1 ≤ T2 1
=E where v ∗ = with i∗ = i2 + 2i
0 if T1 > T2 1 + i∗
= Ā∗1 at rate of interest i2 + 2i (or force of interest 2δ)
xy
Hence
Var(Z) = Ā∗1 − (Ā1 )2 (12.2.9)
xy xy
Example 12.3.1. Two lives (A and B) are both aged 30. Calculate, on the basis of A1967 − 70
ultimate mortality and 4% p.a. interest, the annual premium, payable during the lifetime of A, to
provide an insurance of £1000 payable at the end of the year of death of A, provided A dies after
B.
Find the policy value (on the premium basis) after 10 years (before the premium then due is
paid) if
Solution
A2 = A30 − A 1
30:30 30:30
1 1
= A30 − A30:30 = (1 − dä30 ) − (1 − dä30:30 )
2 2
1 1
= − d(ä30 − ä30:30 )
2 2
Hence
A2 · µ ¶¸
30:30 1 1 ä30:30
premium = · 1000 = 1000 −d 1−
ä30 2ä30 2 ä30
= 1000.(0.00327) = 3.27 = P, say.
Lapse options
One should avoid having a negative prospective reserve at any duration, as this may lead to a lapse
option against the office. In particular, one should ensure that premiums cease as soon as there is
no possibility of future benefits. (See Example 12.2.2 above.)
Definitions
A reversion is a contract providing a sum of money payable on the death of a certain life, (y).
An absolute reversion is a reversion in which the sum is paid under all circumstances, whilst in a
contingent reversion the sum is paid only if certain other lives are (or are not) then alive.
Example 12.4.1. J. Brown(50) will receive £1,000,000 on the death of his mother (aged 80),
provided that he is then alive. He wishes to sell his “interest”. What is it worth?
Solution
Suppose that a purchaser uses a certain rate of interest, i p.a., and assumes that a certain mortality
table applies to both J. Brown and to his mother. (Note that the purchaser might assume different
tables for the lives). The interest is contingent since it requires (50) to be alive when (80) dies.
Hence
In practice, the purchaser will want to be sure of getting the money when (80) dies, and will
“plug the gap” with an insurance policy (which pays out on the death of (80) if she dies second.)
Suppose that the life office issuing the insurance policy uses the same mortality and interest basis
as the purchaser, and also ignores expenses.
Hence purchase price = 1, 000, 000Ā80 − cost of premium for insurance policy
h i
= 1, 000, 000 Ā80 − Ā 2
50:80
= 1, 000, 000Ā 1 (as found before)
50:80
Note:- The life office issuing the policy would examine the health (and any other risk factors) of
(50) very carefully.
212 CHAPTER 12. CONTINGENT ASSURANCES
More complicated functions may be defined and evaluated, but we do not pursue this topic.
12.6. EXERCISES 213
Exercises
12.2 Adams (aged 40) and Brown (aged 50) are two business partners. Adams wishes to provide
for the sum of £80,000 to be paid immediately on Brown’s death if Brown predeceases him
within ten years, and effects a policy providing this benefit by single premium. The life office
issuing the contract employs the following basis:
Mortality (both lives) : A1967 − 70 ultimate
Interest : 6%
Expenses: 2% of the single premium.
Using Simpson’s rule, or otherwise, estimate the single premium payable by Adams.
12.3 Estimate the value of ∞ q1 on the basis of A1967 − 70 ultimate mortality.
74:84
(Assume l108 = 0, so that the integral is over a range of 24 years. Break this into 3 sub-intervals
and use Simpson’s Rule over each.)
12.4 (i) Express A2 in terms of äx , äxx and the rate of interest.
xx
(ii) Smith and Jones are both aged 60. A life office has been asked to issue a special joint-life
assurance policy providing £10,000 at the end of the year of death of the first to die of these
two lives. In addition, if Smith is the second to die, a further £5,000 will be payable at the
end of the year of his death. The policy is to have annual premiums payable during the joint
lifetime of Smith and Jones.
(iii) Write down (but do NOT evaluate) formulae for the reserve at duration 10 years (imme-
diately before payment of the premium then due) on the premium basis, if
(a) both Smith and Jones are alive; and
(b) Jones has died but Smith is alive.
12.5 A policy providing the sum of £100,000 immediately on the death of (x) if she dies before (y)
is to be issued by a life office to a group of trustees.
(i) Ignoring expenses, write down an expression for the single premium in terms of an integral.
(ii) The trustees suggest that level annual premiums should be payable in advance until the
death of the last survivor of (x) and (y).
214 CHAPTER 12. CONTINGENT ASSURANCES
(b) Would you advise the life office to issue the policy with premiums payable as suggested?
Give reasons for your answer.
12.6 Estimate, by the trapezoidal rule or another suitable rule for approximate integration, the
single premium for a temporary contingent assurance of £50,000 payable immediately on the
death of Mrs Smith (aged 60), provided that this event occurs within 5 years and that her
husband (aged 50) is alive at the date of her death. Mrs Smith is subject to the mortality of
a(55) ultimate (females) and Mr Smith is subject to the mortality of A1967 − 70 ultimate.
An interest rate of 7.5% p.a. is to be used, and allowance is to be made for expenses of 6% of
the single premium.
(Note A very accurate answer is not expected.)
12.7 Define the following functions in words, and give an expression for each of them in terms of
an integral.
(i) ∞ q1
xy
(ii) Ā2
xy
(iii) Ā1
x:y:n
12.8 The chief of a certain tribe holds that office until age 50 or earlier death, and may be succeeded
only by a person aged from 36 to 45. (A person aged exactly 36 is eligible, but a person aged
exactly 45 is not.) The customs of the tribe require that a chief’s successor be his oldest
eligible brother; if there is no eligible brother, the position of chief is given to someone from
outside the previous chief’s family, who are then permanently debarred from becoming chief.
The present chief is aged exactly 47 and has two brothers, aged exactly 37 and 33 respectively.
The chief and his brothers may be regarded as independent lives subject to the mortality of
a given table.
Obtain an expression, in terms of quantities of the form n px , n q1 only, for the probability
xy
that (33) will become chief.
12.9 Your life office has been asked to quote a single premium for a contingent assurance policy
providing £300,000 immediately on the death of a woman now aged 80 within 15 years,
provided that at the date of her death a man now aged 60 has died. Your office uses the
following basis:
(i) Assuming that the two lives are independent, write down a formula for the single premium
in terms of an integral.
(ii) State a suitable non-repeated rule of approximate integration for evaluating this integral.
(You are NOT required to carry out the evaluation.)
(iii) Would you subject the male life to stringent underwriting procedures? Give brief reasons
for your answer.
12.7. SOLUTIONS 215
Solutions
12.3 Z ∞
∞ q1 = t p74 µ74+t .t p84 dt
74:84 0
Z 24
= f (t) dt
0
where f (t) = t p74 µ74+t .t p84
4
' [f (0) + 4f (4) + 2f (8) + 4f (12) + 2f (16) + 4f (20) + f (24)]
3
= 0.2907.
12.4
1 1
(i) A2 = Axx = Ax − Axx
xx 2 2
1
(or use A2 = Ax − A1 = Ax − Axx );
xx xx 2
1
= (1 − däx ) − (1 − däxx )
µ ¶2
1 i 1
= − (äx − äxx )
2 1+i 2
Putting x = 60 gives ¸ ·
0.04
0.95P · (9.943) − 100 =10, 000 1 − · 9.943
1.04
0.04
+ 5, 000[0.5 − (12.551 − 0.5 × 9.943)]
1.04
=6175.77 + 1042.40 = 7218.17
7318.17
Hence P = = £774.75
9.44585
216 CHAPTER 12. CONTINGENT ASSURANCES
Hence
50000 × 0.041726
P ' = £2, 219
0.94
12.7 (i)
∞ q1 = P r{(x) will die before (y)}
xy
Z ∞
= t pxy µx+t dt
0
(ii)
Ā2 = m.p.v. of a contingent assurance of £1
x:y
payable immediately on death of (x),
provided this occurs after the death of (y).
Z ∞
= v t t px µx+t (1 − t py ) dt
0
(iii)
Ā 1 = m.p.v. of a temporary contingent assurance of £1 payable
x:y:n
12.8 Let chief = (y) = (47), brothers (x1 ) = (37), (x2 ) = (33).
(x2 ) becomes chief if and only if
(1) he survives for 3 years and (y) “retires” then, (x1 ) having died.
(2) (x1 ) becomes chief (by succeeding (y) on his death or “retirement”) and dies after time 3
and before time 12, leaving (x2 ) alive.
Probability of event (1) is
3 py (1 − 3 px1 )3 px2
12.9 R 15
300, 000 0
vt t p f µ f (1 − t p m ) dt
80 80+t 60
(i) annual premium =
0.90
(ii) The three-eighths rule would be suitable, since it avoids evaluation of the integrand when
t = 7 12 .
(iii) Yes. If (60) dies soon, there is a high chance that (80) will die before age 95 and so give
rise to a claim. The office must check:
(1) the health of (60);
(2) whether he has any occupational or other risks, e.g. participation in a dangerous sport.
Note. The sum assured is large enough to justify the costs of a medical examination.
218 CHAPTER 12. CONTINGENT ASSURANCES
Chapter 13
REVERSIONARY ANNUITIES
āx|y = m.p.v. of Z
ZZ
= (au − āt )(u py µy+u )(t px µx+t ) du dt
u>t
Z ∞ ·Z ∞ ¸
= (āu − āt )u py µy+u du t px µx+t dt (13.1.1)
0 t
Theorem
āx|y = āy − āxy (13.1.2)
Proof It is easiest to proceed indirectly, as follows. Observe that
(
(āU − āT ) + āT if U > T
Z + āmin(T,U ) = = aU
āU if U ≤ T
219
220 CHAPTER 13. REVERSIONARY ANNUITIES
Proof
Z ∞
āx|y = v t t py (1 − t px ) dt
0
Z ∞
= f (t)g 0 (t) dt
0
where
f (t) = 1 − t px = tRq x (which is such that f 0 (t) = t px µx+t ) and
∞
g(t) = −t |āy = − t v r r py dr (which is such that g 0 (t) = v t t py )
Using integration by parts
Z ∞
āx|y = [f (t)g(t)]∞ 0 − f 0 (t)g(t) dt
0
Z ∞
∞
= [(−t |āy )t q x ]0 + t px µx+t (t |āy ) dt
0
Z ∞
t
= t px µx+t v t py āy+t dt.
0
Solution
ā m f ' a f − am f
65|69 69 65:69
ax|y = ay − axy
By Woolhouse’s formula
(m) m−1 m−1
ax|y ' (ay + ) − (axy + )
2m 2m
= ay − axy
= ax|y .
X∞
(m) 1 t t
ax|y = v m P r{(x) has died but (y) is alive at time }
t=1
m m
∞
1 X t
= v m (1 − mt px ). mt py
m t=1
= a(m)
y − a(m)
xy (as before).
Suppose now that a mthly reversionary annuity (of £1 p.a.) begins immediately on the death of
1
(x). Since the payments begin on average 2m year earlier than in the case discussed previously, the
m.p.v. is approximately
1 (m)
(1 + i) 2m ax|y (13.2.1)
An exact formula is
Z ∞
(m)
v t t px µx+t·t py äy+t dt
0
Z ∞ µ ¶
1
' v t t pxy µx+t āy+t + dt
0 2m
1
=āx|y + Ā 1
2m xy
1
'ax|y + Ā 1 (13.2.2)
2m xy
In practice this gives results similar to (13.2.1).
where
Hence the value of widow’s pension for each member retiring at age 65 (marital status unknown)
is
h65 ā m f (13.3.1)
65|65−d
where
(12)f
ä65−d+t = m.p.v. of widow’s annuity to the widow of a
man dying at age 65 + t, if widow is
assumed to be d years younger than husband
Example 13.3.1. A life office sells “personal pensions” policies under which the benefits for men
on retirement at age 65 consist of:
(a) a member’s pension (payable monthly in advance for 5 years certain and for life thereafter),
and (for men married at age 65 only)
(b) a spouse’s pension (payable monthly in advance, beginning immediately on the death of the
member) of half the member’s pension. The possibility of divorce of men aged over 65 may be
ignored, and post-retirement marriages do not give rise to spouse’s pension.
The member’s contributions are invested in certain unitised with-profits funds, and it is assumed
that, in respect of a certain Mr Brown’s policy, the fund available to purchase pension at age 65 will
be £100,000.
Suppose that the life office uses the following basis to calculate the amount of pension which may
be purchased at retirement in respect of a given fund:
mortality of males: a(55) males ultimate
mortality of females: a(55) females ultimate
interest: 8% per annum
expenses: 1% of the fund (at age 65)
13.3. WIDOW’S (OR SPOUSE’S) PENSION ON DEATH AFTER RETIREMENT 223
Men who are not married at age 65 need not buy spouse’s pension, but married men must buy
this. Calculate Mr Brown’s expected monthly pension if
(ii) he is assumed to be married at age 65, and his wife is 3 years younger.
Solution. (i) Let P be annual pension, payable monthly in advance.
(12) (12)
0.99 × 100, 000 = P (ä5 +5 |ä65 ) on males’ table
· ¸
i l70 11
= P ( (12) a5 + v 5 ä70 − )
d l65 24
= P (4.1637 + 0.68058 × 0.86621 × 6.8097)
= 8.1782P
Hence P = £12, 105, so monthly pension = £1, 008.75
Notes
(1) There may be rules to try to exclude benefits for widows of “deathbed marriages”.
(2) There may be problems if the man was married more than once (and his ex-wives are still
alive). In the U.K. the last wife receives all the pension.
(3) If a wife is very much younger than her husband, there may be an “actuarial reduction” (see
later).
(4) Widow’s pensions might cease on remarriage, but this rule is no longer common.
(5) h65+t is sometimes assumed to be “piecewise continuous”, e.g.
0.85 for 0 ≤ t < 1
0.87 for 1 ≤ t < 2
h65+t =
0.89 for 2 ≤ t < 3
0.90 for t ≥ 3.
1
m (12) 1 (12)
' 0.85v 2 q65 ä f
+ 0.87v 1 2 1 |q m ä f
65
65−d+ 21 65−d+1 21
1 (12)
+ 0.89v 2 2 2 |q m ä f
65
65−d+2 21
" #
3 12 (12) 4 12 (12)
+ 0.9 v ·3 |q m äf +v ·4 |q m äf + ···
65 66
65−d+3 21 65−d+4 12
(assuming that the man dies on average half-way through each year of age).
224 CHAPTER 13. REVERSIONARY ANNUITIES
This gives
(12)
ä f
x−10
R = actuarial reduction factor = (12)
äf
y
13.5. EXERCISES 225
Exercises
13.1 (i) Define the following symbols in words, and give a formula in terms of an integral for each
of them:
(a) A 1
xy
(b) A 2
xy
(c) ay|x
Obtain an expression for the present value of this annuity in terms of single and joint-life
annuity factors, life table and compound interest functions. Assume that the same (non-
select) table of mortality is appropriate for the two lives.
226 CHAPTER 13. REVERSIONARY ANNUITIES
13.5 A single-premium policy provides the following benefits to a husband and wife each aged 40.
(1) An annuity of £5,000 per annum, payable continuously, commencing on the husband’s
death within 25 years, or on his survival for 25 years, and continuing so long as either husband
or wife is alive.
(2) A return of half the single premium without interest immediately on the death of the
husband within 25 years, provided that his wife has already died.
Solutions
13.1 (i) (a) The m.p.v. of £1 payable immediately on the death of (y), if this occurs before that
of (x). Z ∞
A 1= v t t pxy µy+t dt
xy 0
(c) The m.p.v. of a reversionary annuity of £1 p.a. payable continuously to (x) after the
death of (y). Z ∞
ay|x = v t t py µy+t · t px · ax+t dt
0
13.2
Benefit = 10, 000Axy + 20, 000Axy + 1000(āx|y + āy|x )
= 10, 000(Āxy + 2Ax + 2Ay − 2Axy ) + 1000(āx + āy − 2āxy )
= 10, 000(2Ax + 2Ay − Axy ) + 1000(āx + āy − 2āxy )
= 10, 000(2 − 2δāx + 2 − 2δāy − 1 + δāxy ) + 1000(āx + āy − 2āxy )
= (1000 − 20, 000δ)(āx + āy ) + (10, 000δ − 2000)āxy + 30, 000
= 10000[ā f − ā m f + ā m − ā m f ]
60 64:60 64 64:60
' 10000[a f + a m − 2a m f ]
60 64 64:60
= 42230
lm lf
ä m f = ä m f − v 20 84
· 80
· ä m f = 8.5709
64:60:20 64:60 lm lf 84:80
64 60
42230
Hence P = = £5, 284 .
0.95 × 8.5709 − 0.15
13.4 (a) value of benefit = 3000axy .
(b) value of benefit = 2000(ax − axy )
228 CHAPTER 13. REVERSIONARY ANNUITIES
=ā m f + v 25 25 p m ā m
40|40 40 65
Z
P 25 t
Benefit (2) = v t p40 µ40+t (1 − t p40 ) dt
2 0
P
= [A 1 −A1 ]
2 40:25 40:40:25
P 1 1
= [A 1 − Az }| { ]
2 40:25 2 40 : 40:25
· µ ¶¸
P 1 D65 1 D65:65
= (1.04) 2 A40:25 − − 1 − d.ä40:40:25 −
2 D40 2 D40:40
P
= (0.094993 − 0.5 × 0.17511) = 0.003719P
2
13.6. SOLUTIONS 229
231
232 CHAPTER 14. PROFIT TESTING FOR UNIT-LINKED POLICIES
The cost of allocation in year t is the money actually used by the office to buy units, that is
(1 − λ)at Pt .
Hence Pt − (1 − λ)at Pt may be transferred to the sterling reserves as a deduction for expenses.
Define
Define
F0 = 0
Then
Ft = [Ft−1 + (1 − λ)at Pt ] (1 + iu ) − ct (14.2.1)
Suppose that the fund management charge is a proportion m of the unit fund; we have
and thus
Ft = (1 − m) [Ft−1 + (1 − λ)at Pt ] (1 + iu ) (14.2.3)
Note. In some cases ct may be a fixed sum rather than a proportion of the fund.
Example 14.2.1. A life office issues a large block of 3-year unit-linked endowment assurances under
which 80% of the first year’s premium and 101% of subsequent premiums are invested in units at
the offer price. The bid price of the units is 95% of the offer price. The units are subject to an
annual management charge of 0.75% of the bid value of the fund at the end of each policy year.
The annual premium is £1,000 and unit prices are assumed to grow at 9% per annum.
Calculate the bid value of the units at the end of each year, according to the office’s projections.
14.3. THE STERLING FUND (OR STERLING RESERVES) 233
Solution
(1) (2) (3) (4) (5)
Cost of Fund brought Fund at end
Allocation forward from start of year
t Pt (1 − λ)at Ft−1 + Pt (1 − λ)at before F.M.C. F.M.C. Ft
Define
The “initial” expected net cash flow in the sterling fund in year t; that is the expected net cash
flow per policy sold, is given by
Case 2
Suppose now that the office wishes to maintain sterling reserves of t V at the end of year t
(t = 1, 2, ..., n − 1). It is assumed that 0 V = n V = 0.
The calculations are very similar to those for conventional profit-testing.
The profit vector is
(P RO)t = (SCF )t + is ·t−1 V − (IR)t (14.3.5)
where
Therefore
σt =t−1 px · (P RO)t
Solution
(1) (2) (3) (4) (5) (6)
Premium less cost Accumulation Death Guarantee In Force
of allocation Expenses of Sterling Costs FMC cash flow
t Pt − Pt (1 − λ)at et Fund in year (DG)t ct (SCF )t
1 44.28 1 44.28
2 35.67 0.997 35.56
3 44.92 0.994 44.65
(2) The shareholders may value the net profits at a certain rate of interest, j per annum. This
rate is called the Risk Discount Rate, and may reflect uncertainties in {σt }, with j normally higher
than is .
The net present value of the profits is thus
n
X
N P V (j) = v t σt at rate j.
t=1
Example 14.4.1. Find the net present value of the profit signature in example 4.3.1 at a risk
discount rate of 10%.
Solution
3
X
n.p.v. of profit = v t σt at 10%
t=1
= 44.58v + 35.56v 2 + 44.65v 3
= 103.46.
Note
X is the amount required at time 1, and hence the negative cash flows are discounted to time 1 (not
the start of the policy).
Hence the shareholders may take a profit at time 1 of
There will now be no need for capital injections (from the shareholders to the Sterling Reserves)
at times 2, 3 and 4. The shareholders may still take a profit of £0.62 at time 5.
The zeroised profit signature {σt0 } is thus
11.76
0
(σt0 ) =
0
0
0.62
Notes
14.6. WITHDRAWALS 237
(2) We may calculate the revised profit vector using the equation
σt0
(P RO)0t =
t−1 px
Example 14.5.1. Using the profit signature (14.5.1), and supposing that qx+t−1 = 0.01 for t =
1, 2, 3, 4, 5, calculate the sterling reserves in each year that are implied by zeroisation of the profit
signature.
Solution
(1) (2) (3) (4) (5) (6)
Original profit Zeroised profit Remainder of Accum. Prob. that
signature signature net-cash flow of (3) policy is in force
t σt σt0 σt − σt0 at is t px tV
(4)
(6) =
(5)
Observe that column (4) gives the sterling fund per policy sold, and the probability of being in
force at time t is t px . Thus
Funds needed at time t per policy sold = t V × P r{ policy is still in force at time t}.
14.6 Withdrawals
So far we have ignored the possibility of surrender. Now assume that the surrender of a policy may
occur only at the end of a policy year (just before payment of the premium then due).
Define
Assume that wn = 0 as the policyholder will receive the maturity benefit at that time.
The chance that a policy in force at the start of year t is surrendered at the end of year t is
therefore,
px+t−1 · wt
Define
(SV )t = the surrender value at time t
which is usually equal to Ft , the bid value of the policy’s units, in some cases minus a surrender
penalty of (say) £10 or 1% of Ft .
238 CHAPTER 14. PROFIT TESTING FOR UNIT-LINKED POLICIES
(P RO)0t = (P RO)t .
The profit signature, σt , must also be adjusted to allow for surrenders; we have
where
0
t−1 px = P r{ policy is in force at time t − 1, allowing for withdrawals}
(
t−1 px (1 − w1 )(1 − w2 )...(1 − wt−1 ) if t ≥ 2
=
1 if t = 1
Note
Even if (SV )t = Ft , and hence (P RO)0t = (P RO)t , σt0 will still differ from σt as the probability of
the policy still being in force in each year will be different, when t > 1.
The profit signature allowing for withdrawals, {σt0 }, may be zeroised in the same way as before.
When calculating the reserves, t V , implied by zeroisation, the probability of the policy being in force
at time t should be changed from t px to
t px (1 − w1 )(1 − w2 )...(1 − wt )
Example 14.6.1. A life office issues a large number of 3-year unit-linked endowment policies to men
aged 65, under each of which level annual premiums of £1,000 are paid. 80% of the first premium
and 105% of each subsequent premium is invested in units at the offer price. There is a bid/offer
spread in unit values, the bid price being 95% of the offer price.
A fund management charge of 0.5% of the bid value of the policyholder’s fund is deducted at the
end of each policy year, before payment of any benefits then due.
The death benefit, which is payable at the end of the year of death, is £3,000 or the bid value
of the units if greater. The maturity value is equal to the bid value of the units.
The office incurs expenses of £100 at the start of the first year and £20 at the start of each of
the second and third years.
Mortality is assumed to follow A1967-70 ultimate. It is assumed that, at the end of each of the
first two policy years, 2% of the surviving policyholders withdraw. The withdrawal benefit is 98%
of the bid value of the units, after deducting the management charge.
14.6. WITHDRAWALS 239
(a) Assuming that the growth in the unit value is 7% p.a. and that the office holds unit reserves
equal to the bid value of units and zero Sterling Reserves at the end of each year, calculate the
profit emerging at the end of each policy year per policy sold. Sterling Reserves are assumed to earn
interest at 6% p.a.
(b) Calculate the revised profit emerging at the end of each year if the office takes a smaller
profit in year 1 in order to ensure that the profit emerging in the second and third policy years is
zero.
Solution
(a) We first work out the unit fund, Ft , per policy sold, assuming it remains in force.
Policy Cost of Funds brought Funds on end Unit fund
Year allocation forward from start before deduction F.M.C. at end of year
t Pt (1 − λ)at Pt (1 − λ)at + Ft−1 of charge ct Ft
X = 35.10v + 2.66v 2 at is = 6%
= 35.48.
Exercises
14.1 A life office issues a three-year unit-linked endowment policy to a life aged exactly 60. The
annual premium is £2,000, payable at the start of each year. The allocation proportion is
90% in year 1 and 97% thereafter. At the end of year of death during the term, the policy
pays the higher of £10,000 and the bid value of units allocated to the policy, after deduction
of the fund management charge. A bonus of 2% of the (bid) value of the unit fund is payable
at maturity. The life office makes the following assumptions in projecting future cash flows:
The office’s mortality basis is A1967-70 ultimate, and Sterling Reserves earn interest at 5%
per annum.
Calculate
(i) the profit signature per policy sold, ignoring any need to maintain Sterling
Reserves at the end of each year, and
14.7. EXERCISES 241
(ii) the profit signature per policy sold if Sterling reserves are zeroised.
(c) The office now wishes to make an allowance for surrenders. It assumes that, at the end
of the first and the second policy years, 3% of the surviving policyholders will surrender (just
before payment of the second and third annual premiums respectively.) Surrender values are
equal to the value of the policyholder’s units (after deduction of fund management charges),
with a surrender penalty of £10. Calculate
(i) the revised profit signature per policy sold, ignoring any need to maintain Sterling Reserves
at the end of each year,
(ii) the revised profit signature per policy sold if the Sterling Reserves are zeroised, and
(iii) the net present value, at a risk discount rate of 15% per annum, of the revised profit
signature per policy sold, assuming that the Sterling Reserves are zeroised.
14.3 (a) If a profit test for a unit-linked policy reveals negative cash flows in the second or any
subsequent policy year, it is customary to eliminate these negative values by setting up sterling
reserves at the end of each year.
Describe briefly the technique (“zeroisation”) by which these reserves are calculated.
(b) An office issues a 3-year unit-linked policy with a yearly premium of £500. The death
benefit, payable at the end of the year of death, is £1,000 or the bid value of units if greater.
The maturity value is the bid value of the units at maturity.
95% of each premium is invested in units at the offer price. The bid price of units is 95% of
the offer price. Management charges of 41 % of the bid value of the units are deducted at the
end of each year (before payment of death and maturity claims).
The office expects to incur expenses of £75 at the start of the first year and £25 at the start
of each subsequent year.
Using a profit testing analysis, calculate for a life aged 60 at entry
(i) the expected profit in each of the 3 years per policy in force at the beginning of the year,
(ii) the net present value at the issue date of the expected profit from one policy assuming a
risk discount rate of 10% per annum.
Assume that the unit fund grows at 8% per annum (before deduction of management charges),
that sterling reserves need not be maintained at the end of each year, and that the possibility
of surrender may be ignored. The mortality of policyholders follows A1967-70 ultimate and
sterling reserves earn interest at 6% per annum during each policy year.
14.4 If a profit test for a unit-linked policy reveals negative cash flows in the second or any sub-
sequent policy year, it is customary to eliminate these negative values by setting up sterling
reserves at the end of each year.
Calculate the sterling reserves required at the end of each policy year, per policy then in force,
for a 3- year policy for which the profit signature (with no allowance for sterling reserves at
the end of each year) is (250, - 100, - 50), given that the rate of mortality is 0.01 per annum
at each age and sterling reserves earn interest at 8% p.a.
14.5 An office issues a unit-linked endowment assurance with annual premium £400 and term five
years to a life aged 60 who is subject to A1967-70 ultimate mortality.
The sum assured, payable at the end of year of death or at the maturity date, is the bid value
of the units held, subject to a guaranteed minimum death benefit of £2,000. The allocation
proportion is 70% for the first annual premium and 98% for all subsequent annual premiums.
For units the bid/offer spread is 5% and the annual rate of management charge is 0.75%.
In determining the sterling reserves necessary for the policy the office makes the following
assumptions:
242 CHAPTER 14. PROFIT TESTING FOR UNIT-LINKED POLICIES
Solutions
14.1 (i)
(1)
t Pt (1 − λ)at Pt (1 − λ)at + Ft−1 (1) × 1.1 F.M.C. Ft
1 1692 1692 1861.20 37.22 1823.98
2 1823.60 3647.58 4012.34 80.25 3932.09
3 1823.60 5755.69 6331.26 126.63 6204.63
(ii)
(2) (3) Maturity
t Pt − Pt (1 − λ)at et [(2) − (3)] × 1.04 (DG)t FMC Cost (SCF )t
1 308 300 8.32 118.00 37.22 0 -72.46
2 176.40 50 131.46 97.17 80.25 0 114.54
3 176.40 50 131.46 67.37 126.63 121.89 68.83
(Maturity cost in year 3 = 0.02(6204.63) × p62 .)
(ii) Let X be the amount withheld at time 1 to cover negative cash flows at times 2 and 3.
X = 108.77v + 3.12v 2 at 5% interest
= 106.42
84.70
0
Hence σt =
0
10.10
(c) (i)
Notice that Ft − (SV )t = 10 for all t.
Hence (P RO)0t = (P RO)t + 10wt p64+t
244 CHAPTER 14. PROFIT TESTING FOR UNIT-LINKED POLICIES
9.50
(iii) NPV = 88.53v + 9.50v 4 at 15% interest = £82.41.
14.3 (a) Zeroisation is the process whereby the profit in the first year is reduced to pay for any
future negative cash flows (as explained in text).
(b) (i)
(1)
t Pt (1 − λ)at Pt (1 − λ)at + Ft−1 (1) × 1.08 FMC Ft
1 451.25 451.25 487.35 1.22 486.13
2 451.25 937.38 1012.37 2.53 1009.84
3 451.25 1461.09 1577.98 3.94 1574.04
(2) (3)
t Pt − Pt (1 − λ)at et [(2) − (3)] × 1.06 (DG)t FMC (SCF )t = (P RO)t
1 48.75 75 -27.82 7.42 1.22 -34.02
2 48.75 25 25.17 0 2.53 27.70
3 48.75 25 25.17 0 3.94 29.11
(ii)
t (P RO)t t−1 p60 σt
1 -34.02 1 -34.02
2 27.70 0.98557 27.30
3 29.11 0.96979 28.23
3
X
Hence net present value = σt v t at 10% interest
t=1
= −34.02v + 27.30v 2 + 28.23v 3
= £12.84.
14.4
250
σt = −100
−50
14.8. SOLUTIONS 245
Let X be the sum retained in year 1 to cover the later negative cash flows. Then
X = 100v + 50v 2 at 8% interest
= £135.46
114.54
Hence the zeroised profit signature is σt0 = 0
0
(1) Accumulation
t σt σt0 σt − σt0 of (1) t px tV
1 250 114.54 135.46 135.46 0.99 136.83
2 -100 0 -100 46.30 0.980 47.24
3 -50 0 -50 0 0.970 0
14.5 (a) (i)
(1)
t Pt (1 − λ)at Pt (1 − λ)at + Ft−1 (1) × 1.07 FMC Ft
1 266.00 266.00 284.62 2.13 282.49
2 372.40 654.89 700.73 5.26 695.47
3 372.40 1067.87 1142.62 8.57 1134.05
4 372.40 1506.45 1611.90 12.09 1599.81
5 372.40 1972.21 2110.26 15.83 2094.43
(ii)
(2) (3)
t Pt − Pt (1 − λ)at et [(2) − (3)] × 1.04 (DG)t FMC (SCF )t
1 134 125 9.36 24.79 2.13 -13.30
2 27.60 21.40 6.45 20.89 5.26 -9.18
3 27.60 22.90 4.89 15.37 8.57 -1.91
4 27.60 24.50 3.22 7.87 12.09 7.44
5 27.60 26.22 1.44 0 15.83 17.27
So thezeroisedprofit signature is
−23.71
0
σt0 =
0
7.09
16.13
246 CHAPTER 14. PROFIT TESTING FOR UNIT-LINKED POLICIES
(4) Accumulation of
t σt − σt0 (4) at 4% interest t px tV
1 10.41 10.41 0.98557 10.56
2 -9.05 1.78 0.96979 1.84
3 -1.85 0 0.95257 0
4 0 - 0
5 0 - 0
(c) (i)
(iii)
t (SCF )t tV (1.06)t−1 V p59+t · t V (P RO)t t−1 p60 σt
1 -12.94 10.56 0 10.41 -23.35 1 -23.35
2 -7.76 1.84 11.19 1.81 1.62 0.98557 1.60
3 1.11 0 1.95 0 3.06 0.96979 2.97
4 12.71 0 0 0 12.71 0.95257 12.11
5 21.65 0 0 0 21.65 0.93385 20.22
This uses
(P RO)t = (SCF )t + (1 + is )t−1 V − px+t−1 .t V
MULTIPLE-DECREMENT
TABLES
15.1 Introduction
Consider a body of lives subject to two “modes of decrement”, α and β. For example, consider
a group of bachelor employees of a large company, from which men can leave by either mode α,
marriage, or by mode β, leaving the company (mortality being ignored).
Take a bachelor employee aged x, and let
and
Define
247
248 CHAPTER 15. MULTIPLE-DECREMENT TABLES
and
We may proceed through the development of “life tables”, by merely putting “a” in front of the
various functions. For example, the function, (al)x , x ≥ x0 , is constructed to be such that
(al)x+t
t (ap)x = , (x ≥ x0 , t ≥ 0)
(al)x
When t = 1 it may be omitted, giving
(ad)x (al)x − (al)x+1
(aq)x = =
(al)x (al)x
and
(ad)x+t
t |(aq)x =
(al)x
Also, define the “force of exit” from the double-decrement table (by whichever mode occurs first)
by
h (aq)x
(aµ)x = lim+
h→0 h
It follows as for ordinary life tables that
· Z t ¸
t (ap)x = exp − (aµ)x+r dr
0
The probability density function of T is
(
t (ap)x (aµ)x+t , t≥0
0 , t<0
α
t px = P r{(x) does not exit by mode α before time t,
whether exit by mode β happens or not}
( = P r{(x) does not marry within t years } in our example.)
β
t px = P r{(x) does not exit by mode β before time
t, whether exit by mode α happens or not}
( = P r{(x) does not leave service within t years} in our example.)
Note that
α
t px = P r{T1 ≥ t} and t pβx = pr {T2 ≥ t}.
We also have
α
t qx = P r{(x) exits by mode α before time t,
whether exit by mode β happens or not}
= P r{T1 ≤ t}
= 1 − t pα
x.
Similarly
β
t qx = P r{T2 ≤ t}
= 1 − t pβx .
The functions, t q α β
x , t q x are called the independent rates (or probabilities) of exit within t years
at age x in their respective single-decrement tables.
Hence we may construct the function (al)x from lxα , lxβ . In particular
β
(al)x+t lα lx+t
t (ap)x = = x+t · for all x ≥ x0 , t ≥ 0
(al)x lxα lxβ
250 CHAPTER 15. MULTIPLE-DECREMENT TABLES
Put x = x0 , y = x0 + t to obtain
where k is a constant. If we choose the radix (al)x0 to be equal to lxα0 .lxβ0 , then
t (aq)x = 1 − t (ap)x
= 1 − (1 − t q α β
x )(1 − t q x ), assuming independence of T1 and T2 ,
= t qα β α β
x + t q x − t q x .t q x
If t = 1 ,
(aq)x = qxα + qxβ − qxα qxβ (15.3.2)
Also
h (aq)x
(aµ)x = lim
h→0+ h
· α β α β
¸
h q x + h q x − h q x .h ax
= lim+
h→0 h
α β
µ α
¶µ β
¶
h qx h qx h qx h qx
= lim + lim + lim h
h→0+ h h→0+ h h→0+ h h
= µα β
x + µx (similar to joint-life argument).
In terms of our example, this is the probability that (x) gets married within t years while still
an employee of the company.
We have
α
t (aq)x = P r{T1 < t and T1 < T2 }
f1 (t1 ) = t1 pα α
x µx+t1 , t1 > 0,
β β
f2 (t2 ) = t2 px µx+t2 , t2 > 0, respectively .
So
15.4. DEPENDENT RATES OF EXIT 251
ZZ
α
t (aq)x = f1 (t1 ).f2 (t2 ) dt1 dt2
t1 <t2
t1 <t
Z tZ ∞
α α β β
= t1 px µx+t1 .t2 px µx+t2 dt2 dt1
0 t1
Z t µZ ∞ ¶
α α β β
= t1 px µx+t1 t2 px µx+t2 dt2 dt1
0 t1
Z t
α α β
= t1 px µx+t1 .t1 px dt1
0
Thus
Z t
α α α β
t (aq)x = r px .µx+r .r px dr (15.4.1)
0
Z t
α
= r (ap)x µx+r dr
0
When t = 1, Z 1
(aq)α
x = α α β
t px µx+t .t px dt.
0
(aq)α β
x , (aq)x are called the dependent rates (or probabilities) of exit at age x by mode α, β respec-
tively. (The word ‘dependent’ indicates that (aq)α x depends not only on mode α but also on mode β).
µα
x = 0.05
Z 2
β
(b) 2 (aq)α
30 =
α α
t p30 µ30+t .t p30 dt
0
Z 2 µ ¶
5−t
= 0.05e−0.05t dt
0 5
Z 2 Z 2
−0.05t
= 0.05 e dt − 0.01 t.e−0.05t dt
0 0
2
= (1 − e−0.1 ) + e−0.1 − 4(1 − e−0.1 ) (using integration by parts)
5
= 0.07645.
Hence
Z 1
(aq)α
x =
α α β
t px .µx+t .t px dt
0
Z 1
= qxα (1 − t q βx ) dt (using U.D. of D. for mode α)
0
Z 1
= qxα (1 − t.qxβ ) dt (using U.D. of D. for mode β)
0
· ¸t=1
t2 β
= qxα
t − qx
2 t=0
1
= qxα (1 − qxβ ).
2
Note
It is often assumed that U.D. of D. holds approximately in each single-decrement table, so the above
results may be used as approximations.
Example 15.4.2. In a double-decrement table with 2 causes of decrement, death (d) and withdrawal
(w), the central rates of decrement at a certain age x are
mdx = 0.01
mwx = 0.2
Show that if the central rate of withdrawal was doubled, the dependent rate of mortality would be
reduced by approximately 0.00076.
15.5. PRACTICAL CONSTRUCTION OF MULTIPLE-DECREMENT TABLES 253
mdx
qxd ' = 0.009950
1 + 21 mdx
mw x
qxw ' = 0.18182
1 + 21 mwx
1
Hence (aq)dx ' qxd (1 − .qxw ) = 0.00905
2
Now suppose mw
x is increased to 0.4.
Then
0.4
qxw ' = 0.3333.
1 + 0.2
Hence
1
(aq)dx ' qxd (1 − × 0.3333) = 0.00829
2
i.e. a reduction of 0.00076.
This can be solved (rejecting any solution outside the range 0 and 1) to find qxα , and hence qxβ
may be found from (i) or (ii).
(ad)α
x =the expected number of exits by mode α
between ages x and x + 1 in a
multiple-decrement table with
(al)x lives at age x
=(al)x (aq)α
x
(aq)x = (aq)α β
x + (aq)x
254 CHAPTER 15. MULTIPLE-DECREMENT TABLES
Also,
α
t |(aq)x = P r{(x) leaves by mode α in the
multiple-decrement table between
ages x + t and x + t + 1}
= t (ap)x (aq)α
x+t
(ad)α
x+t
=
(al)x
Procedure
There are 3 basic steps in the construction of a multiple-decrement table.
Step 1
Choose a radix (al)x0 where x0 = the youngest age considered.
(For example, (al)x0 may be 10,000 or 100,000).
Step 2
Calculate qxα , qxβ (x = x0 , x0 + 1, ...) and hence evaluate (aq)α β
x , (aq)x (x = x0 , x0 + 1, ...), assuming
U.D. of D.
Step 3
Calculate
(ad)α α
x0 = (al)x0 (aq)x0
and
d = death
i = permanent disability
15.5. PRACTICAL CONSTRUCTION OF MULTIPLE-DECREMENT TABLES 255
The independent rates of mortality are in accordance with A67-70 ult., and there is an indepen-
dent rate of disablement of 0.01 at each age x (60 ≤ x ≤ 62).
Construct a multiple-decrement table for ages x = 60, 61, 62 (and include the number of survivors
at age 63.)
Solution
Use U.D. of D. to find
1 i
(aq)dx = qxd (1 − ·q )
2 x
and
1 d
(aq)ix = qxi (1 − ·q )
2 x
Choose a radix of (al)x0 = 100, 000 (for example)
Example 15.5.2. Using the multiple-decrement table constructed in example 15.5.1, calculate
(a) The probability that a person aged exactly 60 is alive and not disabled
at age 63.
(b) The probability that a person aged 60 becomes disabled within 3 years.
Solution
(a)
(al)63
3 (ap)60 =
(al)60
92428
= = 0.92428
100000
(b)
2. Financial calculations can be carried out using multiple-decrement tables, as will be shown in the
next chapter.
256 CHAPTER 15. MULTIPLE-DECREMENT TABLES
Proof
Assume that µα β
x and µx are continuous.
Then
α
h (aq)x
(aµ)α
x = lim+
h→0 h
Rh
0 t
(ap)x µα
x+t dt
= lim+
h→0 h
d
Rh α
dh [ 0 t (ap)x µx+t dt]
(aµ)α
x = lim d
dh (h)
h→0+
α
h (ap)x µx+h
= lim+
h→0 1
= µα
x (as lim h (ap)x = 1)
h→0+
h (aq)x
(aµ)x = lim+
h→0 h
· α β
¸
h (aq) x h (aq)x
= lim +
h→0+ h h
= (aµ)α β
x + (aµ)x
= µα β
x + µx
(ad)α
x
(am)α
x =
(aL)x
Rl
(al)x+t (aµ)αx+t dt
= 0 R1
0
(al)x+t dt
R1
(al)x+t µαx+t dt
= 0R 1 (from (15.6.1))
0
(al)x+t dt
' µα
x+ 1 2
R1 α
lx+t µα
x+t dt
mα
x = 0
R1 α
l
0 x+t
dt
' µα
x+ 1 2
(am)α α
x ' mx
Note
We may also construct tables using select rates of decrement, using
for example.
and Z 1
(aq)α
x =
α α β γ
t px µx+t .t px .t px dt
0
Theorem
Under U.D. of D. of each mode of decrement in its single-decrement table
1 β 1 β γ
(aq)α α γ
x = qx [1 − (qx + qx ) + qx · qx ]
2 3
Proof
Z 1
(aq)α
x = α α β γ
t px µx+t .t px .t px dt
0
Z 1
= qxα (1 − t q βx )(1 − t q γx ) dt (using U.D. of D. on α)
0
Z 1
= qxα (1 − t.qxβ )(1 − t.qxγ ) dt (using U.D. of D. on β and γ)
0
Z 1 £ ¤
= qxα 1 − t(qxβ + qxγ ) + t2 qxβ .qxγ dt
0
· ¸t=1
t2 β t3 β γ
= qxα γ
t − (qx + qx ) + qx .qx
2 3
· ¸ t=0
1 1
= qxα 1 − (qxβ + qxγ ) + qxβ .qxγ .
2 3
Note also that the identity of the forces remains true, and
(aµ)x = µα β γ
x + µx + µx .
258 CHAPTER 15. MULTIPLE-DECREMENT TABLES
In the practical construction of a multiple-decrement table, we find the values of (al)x , (ad)α
x,
(ad)βx , (ad)γx and use
(al)x+1 = (al)x − [(ad)α β γ
x + (ad)x + (ad)x ].
Example 15.7.1. The members of a large company’s manual workforce are subject to three modes
of decrement, death, withdrawal and promotion to supervisor. It is known that these workers’
independent rates of mortality are those of English Life Table No. 12 - Males, the independent
withdrawal rate is 0.04 at each age, and their independent promotion rate is 0.02 at age 50 and 0.03
at age 51.
(a) Draw up a service table for manual workers from age 50 to age 51 with a radix of 100,000
at age 50, including the value of (al)52 .
(b) Calculate the probability that a life aged exactly 50 will gain promotion within 2 years.
Solution
50 0.00728 0.04 0.02 0.00706 0.03946 0.01953 100,000 706 3946 1953
51 0.00823 0.04 0.03 0.00795 0.03924 0.02928 93,395 742 3665 2735
52 86,253
This table assumes U.D. of D., i.e.
1 1
(aq)dx = qxd [1 − (qxw + qxp ) + qxw .qxp ].
2 3
with similar formulae for withdrawal and promotion rates.
(b)
p (ad)p50 + (ad)p51
2 (aq)50 =
(al)50
1953 + 2735
= = 0.04688.
100, 000
cannot be used.
To deal with this “abnormal” mode of decrement, we argue that from age x to x + k, mode β
operates by itself. Then, at age x + k, there is a chance qxα of exit by mode α, so
(aq)α
x = P r{(x) survives to age x + k under mode β only}
× P r{(x) leaves at age x + k by mode α, given
survival until then}
= k pβx .qxα
= qxα (1 − k q βx )
(aq)α α β
x = qx (1 − k.qx ) (15.8.1)
To obtain (aq)βx , we may use
So
1 β
(aq)α α
x = qx (1 − ·q )
2 x
and
1 α
(aq)βx = qxβ (1 − ·q )
2 x
2. If k = 0
(aq)α α
x = qx
and
(aq)βx = qxβ (1 − qxα )
3. If k = 1, (i.e. we consider exits by mode α to occur just before reaching age x + 1)
(aq)α α β
x = qx (1 − qx )
and
(aq)βx = qxβ .
Example 15.8.1. Military personnel attend a training camp for an intensive course which involves
3 weeks of continual exercises.
Each soldier remains in the camp for exactly 3 weeks, provided he is not hospitalised because
260 CHAPTER 15. MULTIPLE-DECREMENT TABLES
of injury or “failed” by one of the instructors and sent back to his base. The independent weekly
Hospitalised
Week through injury Being failed
rates of decrement are as follows:
1 0.078 0.132
2 0.102 0.092
3 0.058 0.043
(i) Of a group of 1,000 soldiers who start the course, calculate the number who will successfully
complete the course, the number who will be hospitalised and the number who will be failed.
Assume a uniform distribution over each week of hospitalisation and failure. Anyone sent to
hospital or failed leaves the camp immediately and does not return. There are no other modes of
decrement.
(ii) Some time later the course is altered so that in the first week the instructors test the sol-
diers only at the start of the sixth day, and the independent weekly rate of being failed in week 1
alters to 0.160. During weeks 2 and 3, testing takes the form of continuous assessment as previously,
with no change to the independent weekly rates of being failed.
Calculate the revised numbers of soldiers successfully completing the course, being hospitalised
and being failed, assuming nothing else changes.
Solution
(i) Consider “age” as time (in weeks) since entry to camp. Assume U.D. of D., which gives
1
(aq)hx = qxh (1 − qxf )
2
1
and (aq)x = qx (1 − · qxh )
f f
2
where h = hospitalised, f = failed.
5
(ii) Here, k = 7 (test is at end of 5th day), so
2 f
(aq)xh = qxh (1 − ·q )
7 x
and
5 h
(aq)fx = qxf (1 − ·q )
7 x
for x = 0 only.
q0f is now 0.16, but all the other values of qxf and qxh are unchanged.
x qxh qxf (aq)hx (aq)fx (al)x (ad)hx (ad)fx
0 0.078 0.16 0.07443 0.15109 1000 74 151
1 0.102 0.092 0.09731 0.08731 775 75 68
2 0.058 0.043 0.05675 0.04175 632 37 26
3 570
Hence now the number who complete the course = 570.
15.8. “ABNORMAL” INCIDENCE OF DECREMENT 261
Application to Profit-Testing
Let α = withdrawal, and β = mortality. It is assumed that withdrawals occur only at the end of a
policy year. That is, mode α only operates at time k = 1.
In chapter 14, the following results were used:
(ad)βx = (al)+ β
x (aq)x
1
where (aq)βx = qxβ (1 − qxγ )
2
Similar calculations apply to “abnormal” exits at the end of the year.
262 CHAPTER 15. MULTIPLE-DECREMENT TABLES
Example 15.8.2. Among the employees of a certain firm retirement may take place at or after age
57, but is compulsory at age 60. The independent rates of mortality of the employees are those of
A1967-70 ultimate. 20% of those attaining age 57 retire at once. Leaving aside these retirements,
the central rates of retirement are as follows:
Central Rate
Age of Retirement
57 0.09
58 0.08
59 0.05
There are no withdrawals or ill-health retirements after age 50.
Construct a service table for employees from age 57 to age 60, with a radix of 100,000 employees
attaining age 57.
Solution
Create a special mode of decrement, r0 , to refer to retirement at exact age x. Then,
0
(ad)r57 = 20, 000,
leaving (al)+
57 lives who are subject to ‘normal’ retirements and death, both of which are approxi-
mately “U.D. of D.”
So construct a table of (aq)dx and (aq)rx (ignoring mode r0 ).
mrx
x qxd qxr ' 1+ 21 mrx
(aq)dx (aq)rx
(ad)rx = (al)+ r
x (aq)x .
0
x (al)x (ad)rx (al)+
x (ad)rx (ad)dx
Exercises
15.2 For a certain group of married women, for whom remarriage is not permitted, the dependent q-
type rate of widowhood at each integer age x from 70 to 72 inclusive is twice the corresponding
dependent q-type rate of mortality. The independent rates of mortality of wives follow a(55)
ultimate (females).
(i) Using a radix of 100,000 and assuming a uniform distribution of each mode of decrement in
its associated single-decrement table, construct a double-decrement table for married women
from age 70 to age 72 inclusive, giving also the value of (al)73 , the number of wives at age 73.
(ii) Find the probabilities that a wife now aged 70 will
(a) be alive and married at age 73; and
(b) be widowed within 3 years.
15.3 In a certain country, widowed and divorced men are subject to the following independent
q-type rates of decrement:
mortality: English Life Table No. 12 - Males
remarriage: rates depend on the age at, and the duration since, the end of former marriage;
the following table is an extract from these rates:
exact
age at end duration duration
of former 0 1
marriage year year
50 0.050 0.025
51 0.045 0.023
52 0.042 0.020
Calculate the probability that a man aged exactly 50 whose marriage has just ended will
remarry within 2 years.
15.4 A large industrial company recruits a constant number of school leavers aged exactly 18 years
on 1 July each year.
Upon joining, workers undergo training for one year. Of those who complete this period of
training, ten per cent fail a final test of competence and are dismissed. Employees may also
leave service voluntarily at any time. The central rate of voluntary withdrawal from service
is 0.15 for trainees and 0.10 at each age for fully trained employees.
The occupation is hazardous and all workers, including trainees, are exposed to the risk of
injury. The independent q-type rate of injury is 0.051219 at age 18 and 0.050030 at ages 19
and above. An employee who is injured is transferred to alternative work with a subsidiary
company, at a relocation cost of £1,000.
The mortality of all employees follows English Life Tables No.12 - Males. The number of
employees attaining age 21 each year is 500.
(i) Construct a service table covering the first 3 years of employment with the original com-
pany, distinguishing between those about to take the final test of competence and those who
pass it. [Regard failing the test as a special mode of decrement ]
(ii) How many people are recruited on each 1st July?
264 CHAPTER 15. MULTIPLE-DECREMENT TABLES
Solutions
µα
x+t = c, so
2
Z t
α2
t px = exp[− µα
x+t dt] = e
2 −ct
for 0 ≤ t ≤ 1
0
Z 1
α1 α1
(aq)α
x =
1 α2
t px µx+t .t px dt
0
Z t
= qxα1 e−ct dt
0
· ¸t=1
1 −ct
= qxα1 − e
c t=0
(1 − e−c )
= qxα1
· c α2 ¸
α1 −qx
= qx as c = − log(1 − qxα2 )
log(1 − qxα2 )
(aq)α α1
x = (aq)x − (aq)x
2
· ¸
−qxα2
= [qxα1 + qxα2 − qxα1 qxα2 ] − qxα1 (from above)
log(1 − qxα2 )
1 d 1 w
(aq)w w d d
x = qx (1 − qx ) = 2qx (1 − qx ) = 2(aq)x
2 2
d
2qx
Hence qxw = 1+ 21 qx
d
d d
qx (1− 21 qx )
Therefore (aq)dx = 1+ 21 qxd , and (aq)w d
x = 2(aq)x .
r = remarriage
r
Notice that q51 = q[50]+1 , i.e. duration 1 year from age 50 at end of former marriage.
x qxd qxr (aq)dx (aq)rx (al)x (ad)dx (ad)rx
50 0.00728 0.050 0.00710 0.04982 100,000 710 4,982
51 0.00823 0.025 0.00813 0.02490 94,308 767 2,348
91,193
r (ad)r50 + (ad)r51
2 (aq)50 =
(al)50
= 0.0733.
Use
mw x
qxw '
1 + 21 mw
x
1 1
(aq)dx = qxd [1 − (qxw + qxi ) + qxw qxi ]
2 3
for deaths, injuries and withdrawals, and treat the test as an “abnormal” mode of exit.
x qxd qxi qxw (aq)dx (aq)ix (aq)w
x
18 0.00112 0.05122 0.13953 0.001016 0.04762 0.13588
19 0.00117 0.05003 0.09524 0.001087 0.04762 0.09280
20 0.00119 0.05003 0.09524 0.001105 0.04762 0.09280
Define
(al)19 = number of people before test (at age 19 exactly)
(al)+
19 = number of people who pass test.
with
(al)+
19 = 0.9(al)19
FINANCIAL CALCULATIONS
USING
MULTIPLE-DECREMENT
TABLES
16.1 Principles
One may value cash flows, calculate premiums and find reserves using the same ideas as are used
when there is just one mode of decrement (death).
Notation is best considered from first principles, and there are no commutation functions (except
for pensions, which are covered later).
The procedure is generally as follows:
Stage 1
Construct a multiple-decrement table.
Stage 2
Write down an equation of value of the form
m.p.v. of premiums = m.p.v. of benefits + m.p.v. of expenses
(This assumes that the expected present value of the profits to the office is zero. If it is not, add
a suitable profit term to the equation.)
Then solve the equation of value for the unknown quantity (for example, the annual premium or
the sum assured payable on death)
Note
It should normally be assumed that exits occur, on average, in the middle of each policy year, and
that premiums cease if a life exits by any mode of decrement.
267
268CHAPTER 16. FINANCIAL CALCULATIONS USING MULTIPLE-DECREMENT TABLES
F (t) = P r{T ≤ t}
α
= t (aq)x
Z t
α α β
= r px µx+r .r px dr
0
Since
α β
∞ (aq)x + ∞ (aq)x = 1,
we have
α
lim F (t) = lim t (aq)x < 1
t→∞ t→∞
A defective variable T may have a probability density function, f (t) = F 0 (t). In the above
example, (
α α β
t px µx+t .t px , t > 0
f (t) =
0, t<0
With defective variables, we still have results such as
Z ∞
E[g(T )] = g(t)f (t) dt
−∞
For example, consider a benefit of £1 payable immediately on exit by mode α to a life now aged
x. The m.p.v. of this benefit is
Z ∞
E[v T ] = v t .t p α α β
x µx+t .t px dt
0
Notice the similarities to the joint-life and contingent assurance functions in chapters 11 and 12.
(A) Integrals
The theory of ‘defective’ variables is used to value the m.p.v. of benefits on exit by a given mode of
decrement. We also consider the m.p.v. of premium payments.
(i) Consider a double-decrement table with exits by mode α and mode β, and suppose that there is
a benefit of £S payable immediately on the exit of (x) by mode α within n years.
The present value of this benefit, as a random variable, is
(
Sv T if T < n
Z=
0 if T > n.
(ii) Consider premiums of £P per annum, payable continuously for at most n years while (x) remains
a member of the group under consideration. The mean present value is
Z n Z n
(al)x+t
P vt · dt = P v t .t (ap)x dt
0 (al)x 0
Z n
=P v t .t pα β
x .t px dt (16.3.2)
0
Example 16.3.1. Suppose there are 2 modes of decrement, death (d) and withdrawal (w), and
there is a constant force of withdrawal of k per annum.
Calculate the value of premiums of £P per annum payable continuously for at most n years while
(x) remains a member of the group.
Solution
µw
x+t = k for all t.
Hence t pw
x =e
−kt
for all t.
Z n
m.p.v. of premiums = P v t .t pdx .t pw
x dt
0
Z n
=P e−δt .t pdx .e−kt dt
0
Z n
=P e−(δ+k)t .t pdx dt
0
= P āx:n at force of interest δ 0 = δ + k
(B) Sums
Sums can be used to value death and other benefits and premium payments either exactly (if financial
transactions occur at the end of policy years) or by approximating integrals (16.3.1) and (16.3.2) if
benefits are payable immediately or premiums are paid continuously.
(i) Consider a benefit of £S payable at the end of the year of exit of (x) by mode α within n years.
The m.p.v. of this benefit is
· α α ¸
(ad)α
x 2 (ad)x+1 n (ad)x+n−1
S v +v + ··· + v .
(al)x (al)x (al)x
If the benefit is payable immediately, then exits can be assumed to occur, on average, mid-way
through year. The m.p.v. of the benefit is
· α α α ¸
1 (ad) 3 (ad)x+1 1 (ad)x+n−1
x
S v2 + v2 + · · · + v n− 2
(al)x (al)x (al)x
(ii) Consider an annual premium of £P per annum, payable in advance while (x) is still a member
of the group, for at most n years. The m.p.v. is
· ¸
(al)x+1 2 (al)x+2 n−1 (al)x+n−1
P 1+v +v + ··· + v
(al)x (al)x (al)x
270CHAPTER 16. FINANCIAL CALCULATIONS USING MULTIPLE-DECREMENT TABLES
If the premiums are payable continuously, expression (16.3.2) can be approximated to a sum.
The m.p.v. is now " #
1 (al)x+ 1 3 (al)x+ 3 1 (al)x+n− 1
P v2 2
+ v2 2
+ · · · + v n− 2 2
Note
(al)x+ 12 , (al)x+ 23 , etc. must often be found by interpolation.
Example 16.3.2. A multiple-decrement table referring to mortality (d) and withdrawal (w) from
a life assurance contract is
x (al)x (ad)w
x (ad)dx
60 10,000 64 126
61 9,810 61 131
62 9,618 48 134
63 9,436
Suppose a 3-year term assurance is issued to a life aged 60, providing £20,000 at the end of the
year of death. If expenses consist of 5% of each premium, calculate the annual premium, P , payable
in advance while a policyholder, for at most 3 years. Interest is at 4% per annum.
Solution
20, 000
m.p.v. of benefits = [v.(ad)60 + v 2 (ad)61 + v 3 (ad)62 ]
(al)60
= 722.79
P
m.p.v. of premiums = [(al)60 + v(al)61 + v 2 (al)62 ]
(al)60
= 2.8325P.
Hence P solves
0.95 × (2.8325P ) = 722.79
Therefore P = £268.61 per annum.
Example 16.3.3. A life office issues policies to lives aged under 60 providing the following benefits:
(i) on becoming permanently disabled before age 60, an annuity of £2,000 per annum payable weekly
for life and £20,000 immediately on death, and
(ii) immediately on death before age 60 while not permanently disabled, £20,000.
Calculate the office annual premium, payable weekly and ceasing on death, on permanent disability
or on reaching age 60, for a life aged 58 if the office uses the following basis:
16.3. EVALUATION OF MEAN PRESENT VALUES 271
Mortality: the independent rates of mortality of those not permanently disabled are those of
A1967-70 ultimate; the permanently disabled are subject to the mortality of English Life Table
No.12 - Males with the age rated up by 6 12 years;
Permanent disability: a constant independent rate of 0.006;
Interest: 4% per annum;
Expenses: 2 12 % of all office premiums, plus £50 at the issue date.
Solution
i h i
3 (ad)
59
+ v2 2, 000āih
59 1 + 20, 000Ā
ih
59 1
(al)58 2 2
· d d
¸
1 (ad) 3 (ad)
58 59
(ii) 20, 000 v 2 +v 2 = 468.34
(ad)58 (al)58
For benefit (i)
āih
58 1 = ā65 on E.L.T. 12 - Males = 8.918
2
Āih
58 1 = Ā65 on E.L.T. 12 - Males = 0.65023
2
Similarly āih
59 1 = ā66 = 8.587
2
Āih
59 1 = Ā66 = 0.66323.
2
1
(al)58 12 ' [(al)58 + (al)59 ] = 99119
2
(al)59 12 ' 97310
When these are added together to give (aµ)x+t , this will often be the force of mortality for a
given life table.
Example 16.4.1. A national newspaper recently advertised “free insurance for subscribers”, whereby
a benefit of £10,000 would be paid immediately on
accidental death (mode α) within n years. Assume µα x+t = 0.0005 for t ≥ 0 (for all x) and mortality
due to all causes follows A1967-70 ult.
Assuming a given rate of interest and ignoring expenses, calculate the annual premium payable
continuously by the newspaper to provide this “free” policy.
Solution
Let the annual premium be P .
Z n
m.p.v. of benefits = 10000 v t .t pα α β
x µx+t .t px dt
0
Z n
= 10000 × 0.0005 v t .t pA67−70ult
x dt
0
= 5āx:n on A67-70 ult.
m.p.v. of premiums = P ax:n
Note:
Observe that P does not depend on the mortality table (for all causes), nor the rate of interest, nor
the term of n years. Also notice that £0.42 is a net premium, whereas an office premium (allowing
for expenses) would be larger.
One may wish to calculate premiums for policies which pay out only on death from the “addi-
tional” cause, in which case the formulae in Section 16.3 may be used. If the death benefit is paid
on any cause of death, the problem can be treated as an “extra risk” question, as covered in the
Actuarial Subject A2.
Example 16.5.1. A certain life office’s premium basis for policies accepted at normal rates is:
16.5. EXTRA RISKS TREATED AS AN ADDITIONAL MODE OF DECREMENT 273
A1967-70 select,
4% interest,
expenses are ignored.
A proposer, aged 45, for temporary assurance ceasing at age 65 is subject to an extra occupational
hazard which is considered to be equivalent to an addition of 0.009569 to the force of mortality at
all ages. The sum assured, which is payable immediately on death, is £10,000.
(a) Calculate the level annual premium, payable throughout the term of the policy.
(b) The proposer requests that, in the event of death occurring as a result of the special occupa-
tional hazard, the sum assured should be doubled, and offers to pay an additional single premium
at the outset for this extra cover.
Calculate this single premium.
Solution
(a) Consider α = normal mortality, β = extra occupational mortality
Value of benefits is
Z 20
β β
10000 v t t pα α
[45] .t p[45] (µ[45]+t + µ[45]+t ) dt
0
Z 20
= 10, 000 v t e−kt .t p[45] (µ[45]+t + k) dt (where k = 0.009569)
0
∗
= 10, 000Ā 1 where * indicates normal plus extra mortality
[45]:20
1
= 10, 000(1.04) 2 A∗ 1
[45]:20
1
= 10, 000(1.04) [(1 − d.ä∗[45]:20 ) − A∗
2
1 ]
[45]:20
Now use the rule (as in extra risks) that the rate of interest (in annuity and pure endowment
functions) may be altered to allow for the addition to the force of mortality.
2102.87
So annual premium =
ä∗[45]:20
2102.87
=
ä[45]:20 0.05
= £167.78
(b) The m.p.v. of this benefit is found by considering exit by mode β only. This gives a single
274CHAPTER 16. FINANCIAL CALCULATIONS USING MULTIPLE-DECREMENT TABLES
premium of
Z 20
β β
10, 000 v t .t pα
[45] .t p[45] .µ[45]+t dt
0
Z 20
= 10, 000 e−δt .e−kt .t p[45] .k dt (k = 0.009569)
0
= 10, 000kā∗[45]:20
= 10, 000kā[45]:20 0.05
· ¸
1 l65 20
' 95.69 ä[45]:20 0.05 − (1 − v0.05 )
2 l[45]
' 95.69 × 12.1899 = £1, 166.45
Example 16.6.1. Manual workers between the ages of 50 and 60 are subject to 2 modes of decre-
ment, d = death, and p = promotion to foreman.
Foremen are subject to the mortality of another table, T 0 , with functions lx0 , etc...
Find the probabilities that
(a) A manual worker aged 50 will be alive and a foreman at age 52.
(b) A manual worker aged 50 will be promoted to foreman but then die before age 52.
Solution
(a) The exact formula is
Z 2 µ 0
¶
d p p l52
t p50 .t p50 µ50+t 0 dt
0 l50+t
(the expression in brackets giving survival as a foreman to age 52), which
approximates to
(ad)p50 l520
(ad)p51 l52
0
· 0 + 0
(al)50 l50 1 (al)50 l51 1
2 2
Exercises
16.1 The following is an extract from a multiple-decrement table referring to mortality and with-
drawal from certain life assurance contracts, these modes being referred to as ‘d’ and ‘w’
respectively.
age,
x (al)x (ad)w
x (ad)dx
50 15,490 24 51
51 15,415 21 58
52 15,336 15 60
(i) What is the probability that a policyholder aged 51 will withdraw before attaining age 53?
(ii) Suppose that a single-premium three-year term assurance contract is to be issued to a life
aged 50, subject to the mortality and withdrawal rates shown in the above table. The sum
assured is £50,000, payable immediately on death, and the benefit on withdrawal is zero. On
the basis of rate of interest of 6% p.a., and allowing for expenses of 10% of the single premium,
calculate the single premium for the above policy.
(iii) Suppose now that the office issuing the policy of (ii) above wishes to introduce surrender
values for these three-year contracts. The surrender value is to be equal to 0.5 per cent of
the single premium for each week between the date of withdrawal and the end of the term.
Assuming that surrenders in each policy year take place on average half-way through the year,
write down an equation of value from which you could calculate the revised single premium.
(Do NOT proceed to the evaluation of this premium.)
16.2 Employees of a certain company are given the opportunity of early retirement immediately
after they complete a 3-year overseas assignment.
Those who undertake this assignment effect a 3-year policy providing the following benefits
payable at the end of the year of claim:
(a) on death or ill-health retirement during the term, the sum of £6,000;
(b) on survival as an employee of the company to the end of the term, a lump sum;
(c) on withdrawal from the company during the second year an amount equal to 1 41 times the
annual premium, and on withdrawal from the company during the third year an amount equal
to 2 12 times the annual premium. No benefit is payable on withdrawal from the company in
the first year.
The following multiple-decrement table is applicable to employees going overseas at exact age
47.
(d = death, w = withdrawal, i = ill-health retirement)
x (al)x (ad)dx (ad)ix (ad)w
x
at ages 20 and 21 is 0.05 per annum, and withdrawals are uniformly spread over the year
of age (in the single-decrement table for withdrawals). At exact age 22, apprentices join the
permanent staff, who retire at exact age 60. The force of withdrawal of permanent staff is
0.0094787 per annum. Both apprentices and permanent staff experience mortality according
to A1967-70 ultimate. The employer pays an immediate benefit of £2,000 on withdrawal and
£8,000 on death in service to permanent staff. The death benefit (but not the withdrawal
benefit) is paid to apprentices.
(a) Construct a double-decrement table for apprentices between ages 20 and 22 with a radix
of 100,000 at age 20.
(b) Calculate, using an interest rate of 5% per annum, the value of the benefits in respect of
a new employee aged 20.
16.4 A life office sells policies to lives aged 63 which provide a benefit of £50 per week, ceasing at
age 65 or earlier death, to those becoming permanently disabled before age 65. The following
basis is used to calculate the single premium for this contract:
Independent rates of
mortality of policyholders
(not disabled): A1967-70 select (at entry)
Force of disablement: 0.01 p.a.
Mortality of disabled lives: A1967-70 ultimate, rated up by 7 years
Rate of interest: 4% p.a.
Expenses are ignored.
Weekly benefits may be taken as payable continuously.
Evaluate the single premium for this contract.
16.5 In a certain country, married men are subject to the following independent q-type rates of
decrement:
Mortality: A1967-70 ultimate
Widowhood: wives of married men are subject to A1967-70 ultimate
Divorce: 0.02 per annum at all ages
Widowed and divorced men are subject to the following independent q-type rates of decrement:
Mortality: English Life Table No. 12 - Males;
Remarriage: rates depend on the age at, and the duration since, the end of the former
marriage; the following table is an extract from these rates:
age at end duration duration
of former 0 1
Marriage year year
50 21 0.050 0.025
51 12 0.045 0.023
52 12 0.042 0.020
53 12 0.040 0.018
Calculate the probability that a married man aged 50, whose wife is also aged 50, will attain
age 52 as a widower who has not remarried between ages 50 and 52.
16.8. SOLUTIONS 277
Solutions
16.1 (i)
w (ad)w w
51 + (ad)52
2 (aq)51 =
(al)51
= 0.002335
Hence P = £554.29.
(iii) Let P 0 be the premium allowing for surrender values. There are on average 52.18 weeks
in 1 year. So, the surrender value in year 1 is on average 52.18 × 0.005 × 2.5P 0 .
The average surrender value in year 2 is 52.18 × 0.005 × 1.5P 0 , and so on.
The equation of value is
m.p.v. benefits:
(a)
6, 000
[v((ad)d47 + (ad)i47 ) + v 2 ((ad)d48 + (ad)i48 ) + v 3 ((ad)d49 + (ad)i49 )]
(al)47
= 1987.92
(al)50
(b) X.v 3 = 0.27563(X)
(al)47
2000 w 2
(c) (al)47 [1.25(ad)48 .v + 2.5(ad)w 3
49 .v ]
= 279.07
So X solves
3996.99 = 2266.99 + 0.27563(X)
Hence X = £6276.53
16.3 (a)
w mw 20 0.05 w
q20 ' = = 0.04878 = q21
1 + 12 mw20 1.025
278CHAPTER 16. FINANCIAL CALCULATIONS USING MULTIPLE-DECREMENT TABLES
where
16.8. SOLUTIONS 279
1
' [ā70:2 + ā71:1 ]
2
1
= [1.84904 + 0.96010] = 1.4046
2
1
' (0.96010) = 0.48005
2
Hence single premium = £47.08.
16.5 Denote
α = mortality of married men
β = mortality of wives
γ = divorce.
Use formula of the form
1 β 1 β γ
(aq)α α γ
x = qx [1 − (qx + qx ) + qx qx ]
2 3
So, P r{ married man aged 50 becomes widower between ages 51 and 52}
= 1 |(aq)β50 = (ap)50 (aq)β51 = 0.005154
1
P r{ widower aged 50 attains age 52 without remarrying}
2
l52 1
= (1 − 0.050)(1 − × 0.025) (using E.L.T. 12-Males)
l50 12 2
= 0.92674. (i)
1
P r{ widower aged 51 attains age 52 without remarrying}
2
l52 1
= (1 − × 0.045) = 0.97322. (ii)
l51 12 2
MULTIPLE-STATE MODELS
281
282 CHAPTER 17. MULTIPLE-STATE MODELS
pij (y, y + h)
µij (y) = lim+ (i 6= j)
h→0 h
Also,
X
µi (y) = µij (y)
i6=j
Note
If all these forces are constant, this is a homogeneous chain. If the forces of transition vary with
age, it is called inhomogeneous.
Kolmogorov’s Forward Equations (for the general case in which there are n states) give the
following system for each fixed x:
d
pij (x, x + t) = −µj (x + t)pij (x, x + t)
dt X
+ µνj (x + t)piν (x, x + t) (1 ≤ i, j ≤ n, t ≥ 0) (17.2.1)
ν6=j
d
p12 (x, x + t) = p11 (x, x + t)µ12 (x + t) (17.2.2)
dt
and
d
p13 (x, x + t) = p11 (x, x + t)µ13 (x + t) (17.2.3)
dt
Note also that
d d d
p11 (x, x + t) = − p12 (x, x + t) − p13 (x, x + t)
dt dt dt
17.3. LIFE TABLES AS STOCHASTIC PROCESSES 283
This is a straightforward first-order differential equation. Using the initial condition p11 (x, x) = 1,
(17.2.5) can be solved to give the unique solution
Z t
p11 (x, x + t) = exp[− µ1 (x + r) dr] (17.2.6)
0
Returning to the differential equation (17.2.2) and solving with the initial condition p12 (x, x) = 0
gives Z t
p12 (x, x + t) = p11 (x, x + r)µ12 (x + r) dr
0
Again this corresponds to a “traditional” formula in chapter 15, i.e.
Z t
α α
t (aq)x = r (ap)x µx+r dr
0
Proceeding as in section 17.2, but with only one mode of decrement, we find that Kolmogorov’s
Forward Equations may be solved to give
· Z t ¸
p11 (x, x + t) = exp − µ12 (x + r) dr
0
Remark
In the stochastic processes approach, it is assumed that the Chapman–Kolmogorov equations hold.
That is,
Xn
pij (x, z) = pik (x, y)pkj (y, z) (17.3.1)
k=1
and
p22 (x1 , x2 ) = 1 for all x1 ≤ x2
Equation (17.3.1) with i = 1 and j = 1 is just a re-statement of the following axiom of the
traditional life table:
z−x px =y−x px .z−y py
The various forces of transition have been graduated using Permanent Health Insurance data.
Suppose one wishes to calculate probabilities such as
The existence of the parameter z complicates the multiple-state model: Kolmogorov’s Forward
Equations have to be modified, as the forces of recovery and mortality of sick lives depend on
17.4. SICKNESS MODELS 285
duration of sickness as well as age. This leads to “semi-Markov” processes which are more difficult
to handle than the cases so far discussed.
A further complication is caused by the need (in practical applications) to consider “deferred
periods”. For example, one may wish to calculate
P r{ a healthy life aged x will be sick at time t and has been sick for
at least d weeks (d is the deferred period)}
A possible simplification
One possible way to simplify this model is to initially ignore mortality and to consider only forward
transitions between “healthy” and “sick” states. Define T1 , T2 , T3 , ... to be the times (in years) spent
as
1. a healthy life (who has not been sick since age x),
2. a sick life (who has been sick exactly once),
3. a healthy life (who has been sick exactly once),
and so on.
A pictorial representation of this is as follows:
In this set-up, one can work out
P r{ aXhealthy life aged x will be sick at age x + t}
= P r{ a life in state 1 at age x will be in state j at time t}
j=2,4,6,...
Each term may be evaluated, using the joint density function f (t1 , t2 , ..., tj ) of the variable
(T1 , T2 , ..., Tj ). That is,
Disadvantages:
1. There can be difficulties in evaluating the multiple integrals for large values of j.
286 CHAPTER 17. MULTIPLE-STATE MODELS
In practice, the Manchester Unity System is often used to calculate sickness functions; this will
be covered in Chapter 18.
Chapter 18
SICKNESS FUNCTIONS
zx(h) = P r{(x) is entitled to sickness benefit for the time period until age x + h}
If benefit is paid at the rate of £1 per week, this amounts to £52.18 per year on average. Suppose
that benefit is payable in advance over intervals of length h years, where h = n1 , i.e. each year is
divided into n intervals. The expected cash to be paid in sickness benefit between ages x and x + 1
for a life now aged x is
n−1
X (h)
jh px (52.18h)zx+jh
j=0
sx is called the annual rate of sickness at age x: it is the expected number of weeks of sickness
between ages x and x + 1 for a life now aged x.
We also define
287
288 CHAPTER 18. SICKNESS FUNCTIONS
and hence
sx ' 12 p zx (18.1.5)
x
“Formulae and Tables for Actuarial Examinations” makes use of the Manchester Unity Expe-
rience 1893-97, Occupational Groups AHJ. Fuller details of this experience can be found in the
Appendix to this chapter.
Sickness by Duration
Benefits may depend on the duration of sickness. Define
z̄xm/n = P r{(x) is “sick” and has been sick for more than m weeks
but not more than m + n weeks.}
m/n m/n
Similar modifications are used to define zx and sx . For example,
Note that
z̄x13 = z̄x0/13 = P r{(x) is “sick” and has been sick for less than 13 weeks}
Also,
z̄x104/all = P r{(x) is sick with duration of sickness greater than 104 weeks, or 2 years}.
Furthermore,
13/13 26/26 52/52 104/all
zx = zxall = zx13 + zx + zx + zx + zx
If the sickness benefit falls with duration of sickness, there may be a temptation for people to
temporarily “recover” and then to start claiming benefit again at the higher initial rate. To prevent
this, an off-period is specified so that if 2 spells of sickness are separated by less than the off-period,
the later spell is treated, for benefit purposes, as a continuation of the first spell. In the Manchester
Unity experience, the off-period is 1 year. If benefits do not fall as duration of sickness increases
there is no need for an off-period rule.
Letting n → ∞ gives
Z T
m.p.v. of sickness benefit = 52.18 v t .t px z̄x+t dt (18.2.1)
0
T
X −1
1
m.p.v. of sickness benefit ' v t+ 2 .t+ 12 p (52.18z̄x+t+ 12 )
x
t=0
T
X −1
1
= v t+ 2 .t+ 12 px zx+t (18.2.2)
t=0
Notes
1. If sickness benefits continue throughout life (although this is unusual as it is not easy to define
“sickness” among the very old), we may let T → ∞, giving
Z ∞
m.p.v. of sickness benefit = 52.18 v t .t px z̄x+t dt (18.2.3)
0
∞
X 1
' v t+ 2 .t+ 12 px zx+t (18.2.4)
t=0
2. If the sickness benefit is payable for sickness of duration greater than m weeks but less than m + n
m/n m/n
weeks, replace z̄x and zx by z̄x and zx , respectively.
The M.P.V. of a sickness benefit may be evaluated by direct evaluation of a sum, by approximate
integration or by commutation functions.
Commutation functions
Define
Z 1
Hx = 52.18 v x+t lx+t z̄x+t dt (18.2.5)
0
290 CHAPTER 18. SICKNESS FUNCTIONS
so that the m.p.v. of a sickness benefit of £1 per week for life to (x) is approximately
∞
X Dx+t+ 1 2
zx+t (from (18.2.4))
t=0
Dx
P∞
t=0 Hx+t
=
Dx
Kx
= (18.2.9)
Dx
Note
13/13 26/26 52/52 104/all
Kx = Kxall = Kx13 + Kx + Kx + Kx + Kx , where Hx13 and Kx13 , etc., are defined by
13
replacing zx by zx , etc., in the formulae for Hx and Kx .
Commutation functions are available in “Formulae and Tables” on the following basis only:
• 4% interest
Kx+n
Dx
rather than
Kx
Dx
This is used in connection with waiting periods. If there is a waiting period of 6 months for sickness
benefits, replace Kx by Kx+ 12 (interpolate in the Tables.)
18.2. VALUING SICKNESS BENEFITS 291
Example 18.2.1. A life office is proposing to issue 3-year sickness benefit policies to lives aged 30.
The benefits are £50 per week during sickness within the next three years. There is no waiting
period and the off-period is as in the Tables provided. Find the single premium on each of the
following bases:
mortality: English Life Table No.12 - Males
interest: (i) 4% p.a.,
(ii) 5% p.a.
sickness: Manchester Unity 1893-97 (AHJ)
expenses: none.
Solution
(i)
K30 − K33
m.p.v. of benefits = 50 ·
D30
· ¸
1784760 − 1706624
= 50
29372
= £133.01
If sickness benefit is payable for sickness lasting more than m weeks but less than m + n weeks,
m/n m/n
replace Kx and Hx by Kx and Hx respectively.
For example,
13/all 13/all
Kx − K65
= m.p.v. of a sickness benefit of £1 per week
Dx
payable up to age 65 on sickness lasting more than 13 weeks
13/all
Since Kx is not given directly in the Tables, one must use
13/all 13/13 26/26 52/52 104/all
Kx = Kx + Kx + Kx + Kx
Example 18.2.2. A friendly society issued a policy providing the following benefits to a man aged
exactly 25 at entry:
(a) on death at any time before age 60, the sum of £4,000 payable immediately;
(b) on survival to age 60, an annuity of £8 per week payable weekly in advance for as long as he
survives;
(c) on sickness, an income benefit to be payable during sickness of £32 per week for the first 6
months reducing to £16 per week for the next 18 months and to £8 per week thereafter. Sickness
292 CHAPTER 18. SICKNESS FUNCTIONS
The society uses the following basis to calculate premiums. Find the monthly premium.
to
" 104/all 104/all
#
Kx+2 − K65
10 (18.3.2)
Dx
since the benefit cannot be received in the first 2 years.
This point is discussed in the Appendix to this Chapter, and it is concluded that neither formula is
exactly right, so the use of adjusted version (18.3.2) is usually optional.
The only exception (in which (18.3.2) is accurate and (18.3.1) is not) occurs when the term of
the policy is very short. For example when x = 63, formula (18.3.2) gives zero, which is correct.
18.3. VARIOUS OTHER POINTS 293
Example 18.3.1. A friendly society issues sickness insurance policies which provide income during
periods of sickness as follows:
(a) £100 per week while a sickness has duration in excess of 13 weeks but less than 1 year;
(b) £75 per week while a sickness has duration in excess of 1 year but less than 2 years;
(c) £50 per week while a sickness has duration in excess of 2 years.
All benefits cease at age 65. Premiums are payable weekly until age 65 and are waived when
sickness benefit is being paid. There is no waiting period.
Calculate the weekly premium for a life aged 38 at entry. Basis:
mortality: E.L.T. No.12 (Males)
sickness: Manchester Unity Sickness Experience 1893-97, Occupation Group
AHJ.
interest: 4% p.a.
expenses: 50% of all premiums payable in the first year, plus 10% of all
premiums payable after the first year.
Solution
Let the weekly premium be P .
m.p.v. of (premiums-expenses) is
0.9 × 52.18P ā38:27 − 0.4 × 52.18P a38:1
= 713.98P (i)
m.p.v. of benefits and premiums waived is
13/39 13/39 52/52 52/52 104/all 104/all
K38 − K65 K38 − K65 K38 − K65
100[ ] + 75[ ] + 50[ ]
D38 D38 D38
13/all 13/all
K38 − K65
+ P[ ]
D38
= 784.80 + 267.63 + 543.00 + P × 22.276
= 1595.43 + 22.276P (ii)
Setting (i) = (ii) gives
P (713.98 − 22.276) = 1595.43
Hence P = £2.31
(c) Reserves
These are calculated prospectively or retrospectively (usually the former) in a way similar to that
used for life policies.
If the premium and reserve bases agree, the prospective and retrospective reserves are equal. If
the bases differ, one must specify how to find the reserve.
294 CHAPTER 18. SICKNESS FUNCTIONS
Example 18.3.2. Consider a policy issued to (x) providing a sickness benefit of £10 per week
ceasing at age 65.
(a) Give a formula for the weekly premium, P , ceasing at age 65 (ignoring expenses).
(b) Calculate the reserve at duration t years by
(i) the prospective method, and
(ii) the retrospective method,
on the premium basis.
Solution h i
(a) 52.18P āx:65−x = 10 KxD −K65
h i x
Kx+t −K65
(b) (i) t V = 10 − 52.18P āx+t:65−x−t
h Dx+t ³ ´i
(ii) t V = DDx+t
x
52.18P āx:t − 10 Kx −KDx
x+t
Note that (i) and (ii) are equal, by an argument similar to that used for life policies.
Exercises
Estimate Kx .
(ii) An office offers an optional waiver of premium benefit on sickness of any duration in respect
of a 25-year with or without profits endowment assurance policy with weekly premiums payable
for 25 years or until earlier death. There is a waiting period of 12 months for the waiver of
premium benefit, and, during the second year of the policy, only half the premium (including
the extra premium for the waiver benefit) is waived. The sum assured under the endowment
policy is payable immediately on death, or on survival until the end of the term.
Using the basis given below, calculate the percentage by which the normal weekly premium
(i.e. the premium for a policy without the waiver benefit) for a life aged exactly 30 at entry
should be increased in order to provide the waiver benefit.
18.4 A certain friendly society recruits only married men aged under 55. The society provides the
following benefits:
(i) immediately on the death of the member at any age, £2,000,
(ii) during the first 10 years of membership, immediately on the death of the member’s wife
before her husband, £500,
(iii) on survival of the member to age 65, an annuity of £10 per week for life, and
(iv) during any spell of sickness of the member before age 65, £10 per week reducing to £6
per week after 3 months’ sickness.
The basis for all calculations is:
mortality of members: E.L.T. No.12 - Males
mortality of wives: E.L.T. No. 12 - Males with an age-deduction of 5 years
interest: 4% per annum
sickness: Manchester Unity 1893-97 AHJ
expenses: 5% of all contributions, including those waived during sickness
Wives of members are taken as being of the same age as their husbands; there is no benefit
on the death of wives of marriages taking place after entry to the society, and the possibility
of divorce is to be ignored. There is no waiting period and the off-period is as in the Tables
provided. The possibility of withdrawal is ignored.
(a) Calculate the weekly contribution rate, waived during sickness and ceasing at age 65 or
the previous death of the member, for an entrant aged 25.
(b) Calculate the reserve for a member aged 35 who joined at age 25.
Note Use Simpson’s rule for approximate integration, i.e.
Z b
b−a a+b
f (t) dt ' [f (a) + 4f ( ) + f (b)]
a 6 2
Solutions
18.1 (i)
Kx = Hx + Kx+1
' Dx+ 12 .zx + Kx+1
1
' (Dx + Dx+1 )zx + Kx+1
2
= 581, 464
(ii) Let weekly premium for the basic policy be P , and let k.P be the extra premium for the
waiver benefit.
The equation of value is ·1 ¸
2 (K31 − K32 ) + (K32 − K55 )
52.18 × 0.85kP ā30:25 = P (1 + k)
D30
Note
One does not need to calculate the actual premium P in this example.
18.2 Let weekly premium be P .
The equation of value is Z 30
D65 (12)
52.18P.ā35:30 = 5000 ä65 + v t .t p35 µ35+t (52.18P s̄t̄| ) dt
D35 0
h i
26/all 26/all
+ P K35 − K65 /Dx (i)
Z 30
Note that v t .t p35 µ35+t (52.18P s̄t̄| ) dt
0
Z 30 µ ¶
1 − vt
= 52.18P t p35 µ35+t dt
0 δ
52.18P h i
= 30 q35 − Ā 1
δ 35:30
(iii) 10 × 52.18 · D
D25 ā65 = 693.28
65
³ ´ ³ 13 13 ´
K25 −K65
(iv) (6 + P ) K25D−K25
65
+ 4 D25 = 254.23 + 31.853P
Hence the equation of value is
52.18 × 0.95P ā25:40 = 377.64 + 4.319 + 693.28 + 254.23 + 31.853P
Use,
1
(I¯Ā) 1 ' (I Ā) 1 − Ā 1
35:25 35:25 2 35:25
13
and find K 35 21 and K35 1 by linear interpolation.
2
300 CHAPTER 18. SICKNESS FUNCTIONS
and so
P = £1.86 per week
(b)
5V = (1000 + 5 × 52.18P )Ā1 + (52.18 × 1.86)(I¯Ā)1
40:20 40:20
µ ¶ µ 13 13
¶
D60 K40 − K60 K40 − K60
+ 2000 + (12.5 + P ) + 12.5
D40 D40 D40
− 0.95 × 52.18P ā40:20 (Prospectively)
2. Friendly societies existed largely to protect wage-earners in times of adversity, there being
no State benefits before 1908 (unless one counts the Workmen’s Compensation Acts of 1897,
which covered injuries at work.) There was (and to some extent still is) a wide variety of
societies ranging in size and financial stability, and not all societies provided sickness benefits.
From 1911 to 1946, sickness benefits for wage-earners were provided by Approved Societies
(friendly societies or life offices approved under the National Insurance Act, 1911). A part
of their income was provided by a Government grant, and they had to use a valuation basis
specified by the Government Actuary. Societies with favourable experience were able to pro-
vide additional (in practice, usually dental and ophthalmic) benefits. Since 1946 the State has
provided welfare benefits (including sickness benefits) directly, either in return for national
insurance contributions or under a means test. The friendly societies have generally declined
in importance; many have disappeared, and others concentrate on social activities. A certain
number (particularly of centralized societies) continue, and indeed prosper, in modern con-
ditions: under the Friendly Societies Act 1992, they are able to conduct a greater variety of
business.
3. The Manchester Unity 1893-97 investigation was carried out by Alfred W. Watson, and pub-
lished in 1903. A calendar year system was used, lives being classified according to age nearest
birthday at the start of the calendar year. The unadjusted sickness rates are of the form
These rates were graduated by an adjusted-average formula to produce the published rates, zx .
(The statistical basis of the graduation of sickness rates is more complex than for mortality
rates, and we do not attempt a discussion.) The rates were also subdivided by periods of
13/13 26/26 52/52 104/all
sickness, giving zx13 , zx , zx , zx , zx . The off-period assumed in the investigation
was 1 year, this being the actual off-period for most of the lodges.
next. The experience of occupation group AHJ is given in Tables for Actuarial Examinations.
Since the middle classes and well-to-do did not (in general) join friendly societies providing
sickness benefits, this represents the experience of wage-earners in 1893-97, excluding those in
the more hazardous occupations. (In some cases, however, sickness rates were high because
the jobs demanded a high level of physical fitness.)
5. Mortality investigations were also conducted, with separate tables for 3 geographical areas
and for urban and rural areas, but these have long been out of date. It is usual to combine
the M.U. sickness rates with a more modern mortality table (e.g. English Life Table No. 12-
Males, as used in Tables for Actuarial Examinations.)
6. The sickness rates of M.U. (whole society) were found to be remarkably similar to those for
employed men in the national insurance scheme in 1953-58.
7. Strictly speaking, sickness rates should be “select”, i.e. they should depend on duration of
membership as well as on attained age, since new entrants are not normally accepted unless
they are reasonably healthy. Another point is illustrated by the example of a new entrant aged
x who is to receive £20 per week for the first 26 weeks of sickness, £10 for the second 26 weeks
and £5 per week for the remainder of sickness, subject to a waiting period of 6 months. The
value of the benefits might be adjusted from
26 26/26 52/all
20Kx+ 1 + 10K
x+ 1
+ 5Kx+ 1
2 2 2
(i)
Dx
to
26 26/26 52/all
20Kx+ 1 + 10Kx+1 + 5Kx+1.5
2
(ii)
Dx
to allow for the fact that no “26/26” benefit can be received before age x + 1 and no “52/all”
benefit can be received before age x + 1.5. But formula (ii) is not quite right; consider, for
example, the “26/26” benefits, which are now valued as
26/26 26/26
10[Dx+1.5 zx+1 + Dx+2.5 zx+2 + ...]
Dx
26/26
Now zx+1 is based on an experience (M.U.) in which the exposed to risk included some recent
entrants who could not claim “26/26” benefit because they had not been eligible for sickness
26/26
benefit for more than 6 months. That is, zx+1 should be slightly adjusted upwards relative
26/26
to that of M.U. Similarly, zx+2 should be slightly increased, so formula (ii) understates the
expected value of the benefits. It is therefore better to use the formula
26/26 26/26
10[0.5Dx+0.5 zx + Dx+1.5 zx+1 + ...]
Dx
since the “extra” first term allows approximately for the fact that the later terms are too low.
This argument leads us to use sickness rates without deferment (except for the waiting period),
as in formula (i), except perhaps when benefits cease soon after the entry age.
In practice, the uncertain definition of “sickness” and other factors are likely to be of much
greater importance in the estimation of sickness rates. But when comparing the actual weeks
of sickness claim at the later durations in a given experience with that expected on the basis of
Manchester Unity (say), one must bear in mind any differences in the proportions of members
whose durations of membership are so short that they cannot claim. Thus, for example, an
experience consisting mainly of recent entrants may be expected to have low rates of sickness
at the later durations.
Chapter 19
PENSION FUNDS
Remarks
1. A “non-contributory” pension scheme is one in which the employees do not contribute (e.g. U.K.
civil service).
2. An “insured” pension scheme is one in which benefits are secured by contracts with a life office.
303
304 CHAPTER 19. PENSION FUNDS
Notes
1. The rate of interest, i per annum, used in valuing the benefits and contributions is normally gross
(free of tax). i is called the valuation interest rate.
2. Expenses are usually ignored, as they are paid for separately by the employer.
3. A service table is required. This is a multiple-decrement table with various modes of decrement:
death, ill-health retirement, etc.
4. If benefits or contributions depend on the employee’s salary, a salary scale is required to estimate
future salaries from current salaries.
5. Sometimes one may require to find the employer’s contribution rate by setting the reserve equal
to zero at entry to the scheme of a new entrant, or for a group of new entrants.
Notes
1. “Age” retirements are the retirements at or above the minimum normal retirement age (NRA) of
the scheme. “Age” retirements may be concentrated at an exact age (for example 60 or 65) or may
be spread uniformly between ages 60 and 65, or both.
2. Members often have to retire by a certain age (often 65). In the “Formulae and Tables”, “age”
retirements occur only from age 60 onwards, and all members must retire at age 65 at the latest.
3. To calculate the value of benefits one may require separate mortality tables for “age” retirement
pensioners and ill-health retirement pensioners. We may also have to deal with benefits for people
with “deferred” pensions in the scheme who no longer work for the company, so a mortality table
for them is sometimes required (see later.)
Also define
Z x+1 Z 1
sx = s̄y dy = s̄x+t dt (19.4.2)
x 0
19.4. SALARY SCALES 305
1
sx ' s̄x+ 12 ' [s̄x + s̄x+1 ] (19.4.3)
2
“Formulae and Tables” gives values of sx (not s̄x ), so we find s̄x by linear interpolation:
1
s̄x ' [sx−1 + sx ] ' sx− 12 (19.4.4)
2
Note
To estimate s̄65 , one must use s̄64 12 ' s64 and s̄63 12 ' s63 and then employ linear extrapolation: see
Example 19.4.1 below.
The salary to be earned by (x) between ages x + t and x + t + 1, given that he is an active
member during this age-interval, is estimated as:
sx+t
(SAL) (19.4.5)
s̄x
s̄x+t
(SAL) = assumed salary rate per annum
s̄x
at exact age x + t (19.4.6)
Adjustments
1. If SAL refers to the earnings received in the past year (i.e. between ages x − 1 and x), adjust the
denominator from s̄x to sx−1 .
2. If SAL refers to the expected earnings in the coming year (i.e. between ages x and x + 1), adjust
the denominator from s̄x to sx .
Example 19.4.1. (a) Consider a person now aged exactly 25 whose annual salary rate is currently
£9,192. Estimate
(i) his annual salary rate at exact age 53,
(ii) his earnings between exact ages 64 and 65,
(iii) the average amount earned by him each year between exact ages 60 and 65,
(iv) his annual salary rate at exact age 65.
If he dies at age 57 last birthday, determine the average values of
(v) his annual salary rate at the moment of death,
(vi) the total amount earned over his last year of life.
(b) Calculate revised answers for (i) to (vi) assuming that the person aged exactly 25 is expected
to earn £9,192 in the coming year.
Assume that salaries are revised continuously, and use the pension table in “Formulae and Tables
for Actuarial Exams” with 4% p.a. interest.
Solution
(a) (i) 9192
s̄25 · s̄53 '
9192
s24 1 s52 2
1 = 9192 × 4.77
1.80 = £24, 359
2
9192 9192
(ii) s̄25 s64' s24 1 s64 = £27, 576
£ 2
s60 +s61 +s62 +s63 +s64
¤
(iii) 9192
s̄25 5 ' £27, 198.
306 CHAPTER 19. PENSION FUNDS
(b) Adjusting the denominator from s̄25 to s25 results in multiplying each answer by a factor of
s̄25 s24 12
' ' 0.962567.
s25 s25
This gives answers of:
(i) £23, 447 (ii) £26, 544 (iii) £26, 180
(iv) £26, 618 (v) £24, 922 (vi) £24, 775
Estimation of s̄x and sx
s̄x is usually estimated by a product of 2 factors, one to allow for future inflationary increases and
one to take account of “career progression”. The inflation factor is usually of the form (1 + e)x ,
where e is the assumed annual rate of future salary escalation.
It is often assumed that
where i is the valuation interest rate and future salary inflation is assumed to be e per annum.
(a) The mean present value of the future contributions (of employee, employer or both) at rate k%
of salary for a member age x with current salary rate of £SAL per annum is
Z
k SAL 65−x t lx+t
· v s̄x+t dt
100 s̄x 0 lx
This is usually approximated to
64−x
k SAL X t+ 1 lx+t+ 12
· v 2 sx+t (using s̄x+ 21 ' sx ) (19.5.1)
100 s̄x t=0 lx
R1
D̄x = 0 Dx+t dt ' Dx+ 21 ' 12 [Dx + Dx+1 ]
s
D̄x = sx D̄x
and P
s 64−x
N̄x = t=0 s D̄x+t
One can now evaluate (19.5.1) by means of commutation functions, giving a mean present value
of
1
64−x x+t+ 2
k SAL X v .lx+t+ 12
· x
sx+t
100 s̄x t=0 v lx
64−x
k SAL X Dx+t+ 12
= sx+t
100 s̄x t=0 Dx
64−x
k SAL X s
' D̄x+t
100 s̄x Dx t=0
k SAL s
= · N̄ x .
100 s̄x Dx
k SAL s
· N̄ x (19.5.2)
100 s̄x Dx
Notes
1. If SAL refers to the past year, change s̄x to sx−1 in the denominator.
2. If SAL refers to the coming year, change s̄x to sx in the denominator.
Example 19.5.1. Consider a life aged 35 with current salary rate £10,000 who contributes 5% of
his salary to a pension scheme. Using the Examination Tables, calculate the mean present value of
the employee’s future contributions.
Solution
m.p.v. of contributions is
(b) Suppose contributions are independent of salary and that these are a fixed annual sum of £F,
308 CHAPTER 19. PENSION FUNDS
t=0
lx
64−x
X Dx+t+ 12
=F
t=0
Dx
64−x
X
1
'F · D̄x+t
Dx t=0
µ ¶
N̄x
=F (19.5.3)
Dx
P64−x
where N̄x = t=0 D̄x+t .
Example 19.5.2. If a life aged 35 contributes £500 each year to his pension scheme, calculate the
value of his future contributions.
Solution
N̄35
m.p.v. = 500 · = £7, 096.
D35
Suppose now that contributions are k% of the “pensionable salary”, where pensionable salary
is defined as “salary - A”, A being a fixed amount. Assuming that the current salary rate, SAL,
already exceeds A, the m.p.v. of these contributions is
µ ¶ µ ¶
k SAL s N̄x k N̄x
· − ·A (19.5.4)
100 s̄x Dx 100 Dx
Notes
1. Years of service normally include fractions counting pro rata.
2. The benefits on the date of age retirement or ill-health retirement are valued by multiplying
the annual pension by an appropriate annuity function. For example, the benefit on age retirement
at age 65 is
annual pension × ār65
where r indicates age retirement mortality rates.
Similarly, āix refers to a life retiring at age x due to ill-health. Ill-health retirement mortality is
usually heavier than that for age retirements.
3. Although the symbol ār65 is usually used, the benefit may be payable in various ways, e.g.
monthly in advance, and is possibly subject to a guarantee that at least 5 years’ payments will be
made. In this example,
(12) (12)
ār65 = ä5 + 5 |ä65
on a suitable mortality table.
Ill-health retirements
The m.p.v. of the benefit is
64−x
X 1 ix+t 1
v t+ 2 (n + t + )P āix+t+ 1
t=0
lx 2 2
It follows that
64−x
X 1 ia
R̄xia = (t + )Cx+t
t=0
2
Proof.
R̄xia = M̄xia + M̄x+1
ia ia
+ ... + M̄64
1
= ( Cxia + Cx+1ia
+ ... + C64ia
)
2
1 ia ia ia
+ ( Cx+1 + Cx+2 + ... + C64 )
2
+ ...
1 ia
+ C64
2
1 ia 1 ia 1
= Cx + 1 Cx+1 + ... + (64 − x + )C64
2 2 2
64−x
X 1 ia
= (t + )Cx+t .
t=0
2
64−x
1 X ia
= nP · C
Dx t=0 x+t
µ ia ¶
Mx
= nP
Dx
The value of the F.S.P. is
64−x
X 1
v x+t+ 2 ix+t 1
P x
(t + )āix+t+ 1
t=0
v lx 2 2
64−x
1 X 1 ia
=P· (t + )Cx+t
Dx t=0 2
µ ia ¶
R̄x
=P
Dx
Hence the value of all ill-health pension benefits is
· ¸
n.Mxia + R̄xia
P (19.7.2)
Dx
Age retirements
The valuation of benefits caused by age retirements is very similar to that for ill-health retirements,
but with a final term corresponding to age retirement at exact age 65. The m.p.v. of age retirement
benefits is
64−x
X µ ¶
t+ 12 rx+t 1 r65
v (n + t + )P ārx+t+ 1 + v 65−x (n + 65 − x)P ār65 (19.7.3)
t=0
lx 2 2 lx
19.8. AVERAGE SALARY SCHEMES 311
Again, this may be separated into the P.S.P. and the F.S.P. terms.
Define the commutations ( 1
ra
v x+ 2 rx ārx+ 1 , x < 65
Cx = 2
v 65 r65 ār65 , x = 65
P 65−x ra
Mxra = t=0 Cx+t (note the summation is up to “65 − x”)
µ ra ¶
Mx
= nP ·
Dx
The value of the F.S.P. is
"64−x #
1 X 1 x+t+ 1 r 65 r
P· (t + )v 2r
x+t āx+t+ 1 + (65 − x)v r65 ā65
Dx t=0 2 2
"64−x #
1 X 1 ra ra
=P· (t + )Cx+t + (65 − x)C65
Dx t=0 2
µ ra ¶
R̄x
=P
Dx
Hence the value of all age retirement pension benefits is
· ¸
n.Mxra + R̄xra
P .
Dx
(i+r)a
So if one defines Mx = Mxia + Mxra , and similarly for other commutation functions, then the
value of a fixed pension of £P per annum payable for any retirement is
" #
(i+r)a (i+r)a
n.Mx + R̄x
P (19.7.4)
Dx
for a member aged x with n years’ past service.
Thus the benefits can be separated into the Past Service Pension and the Future Service Pension.
Value of P.S.P.
1
This is just the value of a fixed pension of 60 (T.P.S.) so, using the functions defined in the
previous section, Ã !
(i+r)a
T.P.S. Mx
m.p.v. of past service benefits = (19.8.1)
60 Dx
Value of F.S.P.
1
Consider ill-health retirement first. The salary to be earned between ages x + t and x + t + 2 is
estimated as
1 sx+t
(SAL)
2 s̄x
The m.p.v. of benefits is therefore
64−x
1 X t+ 1 ix+t SAL 1
v 2 (sx + sx+1 + ... + sx+t−1 + sx+t )āix+t+ 1
60 t=0 lx s̄x 2 2
64−x
SAL X 1 ia
= (sx + sx+1 + ... + sx+t−1 + sx+t )Cx+t (19.8.2)
60s̄x Dx t=0 2
has to be added.
On defining the commutation functions
P64−x
s ra
M̄xra = sx M̄xra and s R̄xra = t=0 s M̄ x+t
and using a similar argument to that used for ill-health retirements, the value of the F.S.P. is
found to be
SAL s ra
· R̄x
60s̄x Dx
Thus the value of all benefits for an average salary scheme is
à ! à !
(i+r)a (i+r)a
T P S Mx SAL s R̄x
+ (19.8.3)
60 Dx 60s̄x Dx
Example 19.8.1. A pension scheme provides each member who retires (whether for “age” or “ill-
1
health” reasons) with an annual pension of 60 th of his average annual income over a member’s
entire service, for each year of service. Fractions of years of service are included when calculating
the amount of pension payable.
If contributions are paid entirely by the employer, calculate the appropriate contribution rate
(as a percentage of salary) for a new entrant aged 20.
Solution
Let k% be the contribution rate. Then k must solve
µ ¶ Ã !
(i+r)a
k SAL s N̄20 SAL s R̄20
· =
100 s̄20 D20 60s̄20 D20
s (i+r)a
R̄20
Hence, k = = 5.25%.
0.6s N̄20
Ill-health retirements
The m.p.v. of benefits for a member aged x with n years past service is
64−x
X SAL 1 1 ix+t
zx+t+ 12 (n + t + )v t+ 2 āi 1 (19.9.1)
t=0
80s̄x 2 lx x+t+ 2
314 CHAPTER 19. PENSION FUNDS
64−x
X
z
Mxia = z
C ia
x+t ,
t=0
1 z ia
z
M̄xia = z M ia
x − C ,
2 x
and
64−x
X 64−x
X
ia 1
z
R̄xia = z
M̄ x+t = (t + )z C ia
x+t
t=0 t=0
2
(note similarity to definitions in section 19.7)
Age retirements
Again this is similar to the fixed pension case, but with salary factors inserted. The new commutation
functions are
(
z ra zx+ 21 Cxra if x < 65,
Cx = ra
z65 C65 if x = 65,
65−x
X
z
Mxra = z
C ra
x+t (note the upper limit of 65 − x)
t=0
1 z ra
z
M̄xra = z M ra
x − C
2 x
64−x
X
z ra
R̄xra = z
M̄ x+t
t=0
We find that the m.p.v. of benefits due to age retirements is
"64−x #
SAL X 1
(n + t + )z Cx+t
ra
+ (n + 65 − x)z C65
ra
80s̄x Dx t=0
2
SAL n.z Mxra SAL z R̄xra
= + (19.9.3)
80s̄x Dx 80s̄x Dx
Thus the value of all pension benefits for a final salary scheme is
SAL h z (i+r)a z (i+r)a i
n. M x + R̄x (19.9.4)
80s̄x Dx
19.9. FINAL SALARY SCHEMES 315
Example 19.9.1. Three members of a pension scheme whose age nearest birthday is 45 have the
following annual rates of salary and exact numbers of years of past service:
A : £15,000 20 years
B : £12,000 10 years
C : £14,000 5 years
For these members find the present value of a pension, payable on age-retirement or on ill-health
1
retirement, of 100 th of the average salary in the final 3 years before retirement for each year of service,
including fractions. Give separately the values of the past-service and future-service benefits.
Solution
P3
1 i=1 ni (SAL)i z (i+r)a
Value of P.S.P. = · M 45
100 s̄45 D45
1 490000(z M45 ia
+ z M ra45 )
' 1
100 (s
2 44 + s 45 )D 45
= 29, 071
P3
1 (SAL)i z (i+r)a
Value of F.S.P. = · i=1 · R̄45
100 s̄45 D45
1 41000 ia ra
' (z R̄45 + z R̄45 )
100 12 (s44 + s45 )D45
= 42, 929.
Remarks
1. If retirement is possible within m years, the factor zx should be adjusted to allow for the actual
past salary progression. Also, if someone retires with less than m years’ service, the final salary
must be redefined as the average over a shorter period. (In calculations these points are sometimes
ignored.)
2. If “final salary” means the salary rate at the date of retirement, then we change zx to s̄x . Also,
(i+r)a (i+r)a
if m = 1, zx = sx−1 . But one must not change z R̄x to s R̄x , as the latter refers to average
salary schemes.
3. If fractions of a year are not included when calculating the pension value, it is usually sufficiently
accurate to adjust the n years’ past service to n − 12 . If there is no past service, consider the member
as having “− 12 ” year past service. So one must substitute
1
(n − ).z Mx(i+r)a for n.z Mx(i+r)a
2
in formula (19.9.4) (even when n = 0).
4. Benefits may depend on a “pensionable salary” equal to salary minus some fixed sum, say £A. If
so, then the m.p.v. of the pension benefit in a final salary scheme is now
à !
SAL ³ z (i+r)a z (i+r)a ´ A n.Mx
(i+r)a (i+r)a
− R̄x
n. Mx + R̄x −
80s̄x Dx 80 Dx
5. It may be the case that the number of years of service are restricted to, say, 40 for pension
purposes. This will affect lives joining at ages under 25, assuming a latest retirement age of 65.
There are 2 cases to consider:
316 CHAPTER 19. PENSION FUNDS
(a) If n < 40, then leave the P.S.P. unchanged but restrict the F.S.P. . In the final salary scheme
mentioned above, the value of the F.S.P. is altered to
(b) If n ≥ 40, then restrict the past service to 40 years and make the F.S.P. zero. Hence the
value of pension benefits is
40 SAL z (i+r)a
· Mx (19.9.6)
80 s̄x Dx
1
Example 19.9.2. An executive pension scheme provides a pension of 45 of final salary for each
2
year of scheme service, with a maximum of 3 of final salary, upon retirement due to age between
the ages of 60 and 65.
Final salary is defined as salary in the 3 years prior to retirement.
A director, now aged 47 exactly has 14 years of past service and expects to earn £80,000 over
the coming year.
Using the symbols defined in the Formulae and Tables for Actuarial Examinations, what is the
expected present value of the future service pension on age retirement for this member?
and so on, changing the annuity factor in all commutation functions (both “age” or “ill-health”) to
1.
These functions are tabulated (on the same basis as in “Formulae and Tables”) in the Supplement
to this book.
Using nearly the same arguments as before, the m.p.v. of the lump sum on retirement is
SAL i+r
3· (n.z M i+r
x + z R̄x ) (19.10.1)
80s̄x Dx
If the annual pension is a fixed sum of £P per year of service, then a lump sum of 3P on
retirement has m.p.v. µ ¶
n.Mxi+r + R̄xi+r
3P (19.10.2)
Dx
19.11. DEATH AND WITHDRAWAL BENEFITS 317
Withdrawals
Similar formulae may be developed (replacing the ‘d’ with ‘w’) but withdrawal benefits are usually
in the form of a return of contributions or a deferred pension. (We shall discuss the formulae for
valuing these benefits later.)
Note
The commutation functions Mxd , Mxw , etc. are tabulated in the Supplement to this book. Note that
Mxd corresponds to j Mxd when j = 0 (page v of the Supplement) and Mxw corresponds to j Mxw when
j = 0 (page vi of Supplement).
318 CHAPTER 19. PENSION FUNDS
Example 19.11.1. A company has a pension scheme which provides a pension upon retirement
1
of 80 th of the final salary per year of service. In addition the sum of £10,000 is paid on death in
service of a member. If all contributions are paid by the employer, find the contribution rate as a
percentage of salary required for a new entrant aged 40 with a salary rate of £10,000.
Use the basis of “Formulae and Tables” (with the supplement); final salary is the average annual
salary in the 3 years prior to retirement.
Solution
Let k be the contribution rate per cent. Then k solves
j w 1j w
M̄xw = j M x − C
2 x
sj w
M̄xw = sx j M̄ x
64−x
X
sj 1 w
R̄xw = (1 + j)−(x+t+ 2 )sj M̄ x+t
t=0
Similar functions for returns of contributions on death are defined by substituting ‘d’ for ‘w’.
Suppose that the employees contribute at rate k% of salary. Then the m.p.v. of the return of
contributions on withdrawal for a member aged x is
j
Mxw k SAL sj w
(T P C)(1 + j)−x · + · · R̄x (19.12.1)
Dx 100 s̄x Dx
Proof. See the Appendix to this Chapter.
Note
w w
The values of j M x , sj
R̄x , etc. are given for j = 0.03 and j = 0 in the Supplement to this book.
Mxw k SAL s w
(T P C) + · R̄x (19.12.3)
Dx 100 s̄x Dx
Mxw R̄w
(T P C) +A· x (19.12.4)
Dx Dx
If employee’s contributions are returned on death (with or without interest), formulae (19.12.1)
to (19.12.4) are changed by just substituting ‘d’ for ‘w’.
Furthermore, if contributions are returned on either death or withdrawal, one may adjust formula
(19.12.1) to
j
Mxd+w k SAL sj d+w
(T P C)(1 + j)−x + · · R̄x (19.12.5)
Dx 100 s̄x Dx
d w
where j Mxd+w = j M x + j M x
d w
and sj R̄xd+w = sj R̄x + sj R̄x , as expected.
(Similar remarks apply to formulae (19.12.2) to (19.12.4).)
Example 19.12.1. You are actuary to a pension scheme which provides the following benefits:
1
(a) a pension of 80 × final salary for each year of service (including fractions pro rata) on retire-
ment for “age” or “ill-health” reasons.
(b) a lump sum on retirement of 3 times the annual pension.
Your actuarial basis is that given in the “Formulae and Tables”, and you calculate reserves prospec-
tively, ignoring expenses. Final salary is the average annual salary in the 3 years before retirement.
Members pay contributions at the rate of 2% of salary. The total contribution rate is assessed
for each member separately, and is the proportion of salary which will pay for the benefits, when
the member joins, i.e. the prospective reserve at the entry date is zero.
1. Find the employer’s contribution rate for a member joining at age 20.
2. Find the reserve held for a member aged 45, with salary at the rate of £20,000 per annum,
who joined the employer at age 20.
3. Your pension scheme accepts transfer values from other pension schemes. Suppose that a life
aged 45 joins the scheme, bringing a transfer value of £10,000. The member’s current salary rate is
£15,000 p.a., and he asks for his transfer value to be used to give him “added years” of service, i.e.
to credit him with n years of past service. Find n, given that, in the event of death in service, the
transfer value will be returned with compound interest at 3% p.a.
Note
There are no withdrawals over age 45 in the Tables, so we need not specify what happens to the
transfer value in that event.
320 CHAPTER 19. PENSION FUNDS
Solution
1. Let k be total contribution rate per cent for a new member aged 20. Then
k SAL s SAL z (i+r)a i+r
· N̄20 = [ R̄20 + 3.z R̄20 ]
100 s̄20 D20 80s̄20 D20
Thus
" #
z (i+r)a i+r
100 R̄20 + 3.z R̄20
k= s N̄
= 7.443.
80 20
If spouse’s pension ceases on remarriage, a double-decrement table (for death and remarriage)
should be constructed and used in place of the mortality table for spouses.
The size of spouse’s pension may depend on the member’s salary at or near the date of death
and the number of years of service.
Case (ii) : R∞
(annual widow’s pension). 0 v t .t pm
65 µ m h65+t ā f dt (19.13.3)
65+t 65−d+t
Example 19.13.1. A pension fund provides the following benefits for widows of male members:
(a) on death of the member in service, a widow’s pension of annual amount equal to one-third
of the member’s salary rate at the time of his death; and
1
(b) on death of the member after age or ill-health retirement, a widow’s pension of 120 th of the
member’s average salary in the 3 years immediately preceding retirement for each year of service,
fractions of a year counting pro rata.
Develop formulae for valuing the widow’s benefits for a male member aged x with n years’ past
service (including fractions). You may assume that all age-retirements take place at exact age 60,
between ages 60 and 65, or at exact age 65, and that ill-health retirements may take place at any age
between 35 and 60. Widows’ pensions on death after retirement are payable to any widow (not just
to a widow who was married to the member when he retired.) Widows’ pensions are payable monthly
in advance and do not cease on remarriage. You may assume that a service table has already been
constructed, and that the member’s current salary rate per annum is SAL. Commutation functions
need not be developed.
Solution
(a) Define s̄x , sx , hx as previously. Let
1
zx = (sx−3 + sx−2 + sx−1 ) as before.
3
Let
where
Z ∞
(12)
aIH
x+t+ 21 = v r .r pih ih
x+t+ 1 µx+t+ 1 +r hx+t+ 2 +r äy+t+ 1 +r dr
1
2 2 2
0
(“ih” indicates that a mortality table for men retiring due to ill-health is employed).
For age retirements, there are terms for retirement (i) between ages 60 and 65, (ii) at exact ages
60 and 65.
(i) value =
64−x µ ¶
SAL X t+ 1 rx+t zx+t+ 12 1
v 2 (n + t + )aN H
1
120 t=60−x lx s̄x 2 x+t+ 2
where aN H
x+t+ 1
is as for aIH
x+t+ 1
, but with a mortality table for age retirements.
2 2
µ 0 ¶
SAL 60−x r60 z60
(ii) value = v (n + 60 − x)aN
60
H
120 lx s̄x
µ 0 ¶
SAL 65−x r65 z65
+ v (n + 65 − x)aN
65
H
120 lx s̄x
(rx0 refers to retirements at exact ages 60, 65; note that l65 = r65
0
).
where ā65 is the annuity factor, with the pension possibly payable monthly, etc.
To evaluate (19.14.1) by commutation functions, one needs to define suitable ‘new’ commutation
functions. That is, the expression (19.14.1) is equal to
64−x
SAL X 1
(n + t + )z Cx+t
wa
60s̄x Dx t=0 2
where
" 1 #
wx zx+ 12 (1 + j)65−x− 2
z
Cxwa = ˆl65 a65 v 65
ˆl 1
x+ 2
19.14. PRESERVED PENSIONS ON LEAVING SERVICE 323
We now define
64−x
X
z
Mxwa = z
C wa
x+t ,
t=0
1 z wa
z
M̄xwa = z M wa
x − C
2 x
and
64−x
X
z wa
R̄xwa = z
M̄ x+t
t=0
(Note the similarities with the definitions used in previous sections.) The present value of the
deferred pension is
SAL £ z wa z wa ¤
n. Mx + R̄x (19.14.2)
60s̄x Dx
Note The notation used here is not standard, but is an example of “do-it-yourself” commutation
functions.
324 CHAPTER 19. PENSION FUNDS
Exercises
19.1 Contributions to a pension scheme by employees are made at a rate of 5% of salary when aged
under 35, 6% between ages 35 and 45, and 7 12 % when aged 45 or over. Calculate the present
value of the future contributions payable by a member aged exactly 30 who in the past year
has received a total salary of £12,718.
19.2 A company pension scheme provides the following benefits for all members:
(1) a pension on retirement (on grounds of ill-health or of age) of one-eightieth of final pen-
sionable salary for each year of service (including fractions),
(2) a lump sum on retirement of 3 times the annual pension,
(3) on death in service, a lump sum of £30,000,
(4) on withdrawal from service, a return of the employee’s contributions, accumulated at 3%
per annum compound.
Final pensionable salary is defined as the average annual salary in the three years immediately
before retirement. All members who reach age 65 retire immediately.
Employees contribute to the scheme at the rate of 3% of salary, payable continuously. Salaries
are revised continuously. The employer’s contribution rate is assessed for each member sep-
arately, and is such that the prospective reserve for each new entrant is zero. Expenses are
ignored.
(i) (a) Derive a formula, in terms of suitable commutation functions, for valuing benefit (1)
above in respect of a new entrant aged x with annual salary rate SAL. (You need not define
the service table functions.)
(b) In respect of a new entrant aged x with annual salary rate SAL, give formulae for valuing
benefits (2), (3) and (4) above, using suitable commutation functions. (You need not derive
the formulae.)
(c) Hence find a formula for the employer’s contribution rate for a new member aged x and a
starting salary rate of £10,000 p.a.
(ii)(a) Using the basis given in the pension fund section of the Formulae and Tables (and the
supplement), find the value of each of the benefits (1), (2), (3) and (4) for a new entrant aged
45 with salary rate £10,000 per annum.
(b) Hence or otherwise find the employer’s contribution rate for this new member.
19.3 The pension scheme of a certain company provides an annual pension on retirement (for
‘age’ or ‘ill- health’ reasons) of amount equal to one per cent of the member’s total earnings
throughout his service. The pension is payable weekly. In addition, in the event of a member
dying in service there is payable at the time of death a lump sum of £30,000. There is no
benefit on withdrawal.
The company pays a constant percentage of all the members’ salaries into the pension fund.
The percentage is that which will exactly cover the cost of benefits for a new entrant to the
fund at age 30 with an initial salary rate of £10,000 per annum. Contributions are payable
continuously, and the employees do not contribute to the scheme. Expenses are negligible.
(a) Calculate the contribution rate paid by the company, assuming the last retirement age is
65.
(b) A valuation of the fund is to be conducted. For each active member of the scheme there
is recorded
19.15. EXERCISES 325
(i) the age nearest birthday (which is regarded as the member’s exact age) at the valuation
date,
(ii) the annual salary rate at the valuation date, and
(iii) the total past earnings in service (prior to the valuation date.)
For each age, the totals of (ii) and (iii) are recorded and the following is an extract from the
data.
Assuming that the basis of the Tables provided is appropriate, find the liability at the valuation
date for the benefits payable to the members aged 25, and determine whether the future
contributions payable in respect of these members are more or less than sufficient to cover the
benefits.
19.4 A pension scheme provides each member who retires (for any reason) with annual pension
1
equal to 60 × final salary per year of service. Final salary is the average income over the last
3 years of service, and fractions of a year of service are not included when calculating the
pension.
Assuming that equal contributions are payable by the member and his employer, that in the
event of death in service a benefit is payable equal to the return without interest of both the
member’s and the employer’s contributions, and that in the event of withdrawal from service
a return without interest is made of the member’s contributions, calculate the appropriate
contribution rate payable by both the member and his employer in respect of a new entrant
aged 40.
19.5 You are consulting actuary to a small pension scheme, which has just been established. You
have decided to use the pension fund tables in Formulae and Tables for Actuarial Examinations
as the basis for all calculations. The scheme provides the following benefits to employees:
1
(i) on retirement (for ill-health reasons or otherwise), a pension of 80 th of annual pension-
able salary, averaged over the previous three years, for each year of future service including
fractions;
(ii) on withdrawal or death in service, a return of the member’s contributions, accumulated
at 3% per annum compound interest.
Employees contribute 2% of salary to the scheme. Pensionable salary is defined as salary less
£4000. Salaries are revised continuously, and contributions are made continuously.
Details of the current membership are as follows.
member age current salary rate (£)
1 45 30,000
2 45 20,000
3 35 10,000
4 35 10,000
(a) The employer has decided to contribute the proportion of total salaries which, together
with the employees’ contributions, will exactly pay for the benefits. Calculate the employer’s
contribution rate.
(b) A new employee, aged 35 and with current salary rate £8,000, is about to be hired.
Calculate the surplus or deficiency in the pension fund after this new member joins.
326 CHAPTER 19. PENSION FUNDS
19.6 It is desired to set up a pension scheme for the group of employees described below. For each
member there is recorded his exact age, his exact length of service with the company, and his
annual rate of salary.
1
The scheme will provide pensions of 60 th of “pensionable salary” for each year of service
(fractions of a year being included) and on death or withdrawal from service a return will be
made of the member’s contributions with 3% compound interest. All members will contribute
at the same rate and the employer will contribute the same amount as each member. The
contribution rate will be such as to provide exactly the benefits for future service. The basis
of “Formulae and Tables” is to be used.
(a) Assuming that pensionable salary is the average annual earnings over the three-year period
ending on the retirement date, calculate the contribution rate payable by each member.
(b) Calculate also the total liability for past service benefits. The employer wishes to meet
this liability by paying additional contributions proportional to future salary payments. At
what rate should these additional contributions be made?
(c) Immediately after the scheme is set up as described above the 45 year old member with
5 years of service withdraws. Is the position of the fund improved or worsened by this with-
drawal?
19.7 The pension scheme of a large company provides the following benefits, among others:
(1) on death in service of married members: a spouse’s pension of one-third of the member’s
annual salary at the date of death;
(2) on death after normal retirement of the member at age 65 or after ill-health retirement
at an earlier age: a spouse’s pension of 1% of annual salary at the date of retirement for
each year of scheme membership (including fractions of a year); no pension is payable if the
marriage has taken place after the member’s retirement date.
Assume that a service table has been constructed, and that the proportion of members mar-
ried at exact y is hy . Assume further that spouses are the same age as members, and that
a unisex mortality table is used for all calculations, with the age rated up by 5 years on
ill-health retirement. You are given the age nearest birthday, current salary rate and years
of past service (including fractions) of each member. Spouse’s pension is payable monthly in
advance, beginning immediately on the death of the member.
(i) Using the rate of interest i per annum, find formulae for the mean present value of each
of the benefits (1) and (2) above for a member aged x (x < 65). You are NOT required to
construct commutation functions.
(ii) Suppose that the scheme’s rules are to be changed so that benefit (2) is to be payable to
any surviving spouse. Show how to modify the formulae of (i) to accommodate this change.
19.16. SOLUTIONS 327
Solutions
19.1
12, 718
{0.05s N̄30 + 0.01s N̄35 + 0.015s N̄45 }
s29 × D30
= 12, 718 × 1.30162
= £16, 554
z i+r
3.SAL R̄x
(b)(2): 80s̄x · Dx
Mxd
(3) 30000 · Dx
SAL sj w
(4) 0.03 · s̄x Dx R̄x where j = 0.03
(c) Let employer’s contribution rate be k for this member. Then
SAL s
(k + 0.03) N̄ x = above benefits.
s̄x Dx
So
· ¸
1 1 z (i+r)a 3 z i+r d sj w
k + 0.03 = R̄ x + R̄ + 3s̄ M
x x + 0.03. R̄
s N̄ x 80 80 x x
" #
s (i+r)a
s̄30 R̄30 d
Hence k = s 0.01 + 3.M30
N̄30 s̄30
= 0.0535 = 5.35%
328 CHAPTER 19. PENSION FUNDS
P P return of
x SAL (n × SAL) P.S.P. F.S.P. contributions (per 1% of salary)
25 43,000 163,100 7,776 76,783 2,223.7
35 62,600 752,300 55,421 126,780 1,720.5
45 68,200 1,044,000 103,230 119,012 1,022.7
166,427 322,575 4,966.9
The value of the contributions
P is, per 1% of contributions,
SAL s N̄x
(1)
100s̄x Dx
x Value of (1)
25 8,791.2
35 11,919.1
45 9,724.0
30,434.3
Let members’ contribution be k%. The equation of value is thus
2k × 30, 434.3 = 322, 575 + 4, 966.9k
Hence k = 5.77%.
(b) The past service liability is £166,427. Let the contribution rate per cent needed to pay
for this be p. Then p solves,
p × 30, 434.3 = 166, 427
Hence p = 5.468%.
(c) The reserve for the withdrawn member is
13, 000 z (i+r)a 5 × 13000 z (i+r)a 13000 sj d+w
· R̄45 + M 45 + 0.0577 × · R̄45
60s̄45 D45 60s̄45 D45 s̄45 D45
13000 s
− (2 × 0.0577 + 0.05468) · · N̄45
s̄45 D45
= 30, 238 − 31, 526
= −1, 288
Benefit (2):
On normal retirement the value is
SAL r65 h (12) 1
i
· v 65−x .s̄65 (n + 65 − x) × h65 a65|65 (1 + i) 24
100s̄x lx
to
Z ∞
(12)
v r .r px+t+ 12 +5 µx+t+ 12 +5+r hx+t+ 12 +r äx+t+ 1 +r dr
2
0
19.16. SOLUTIONS 331
j w 1j w
M̄xw = j M x − C
2 x
sj
M̄xw = sx .j M̄xw
64−x
X 1
sj
R̄xw = (1 + j)−(x+t+ 2 ) .sj M̄x+t
w
t=0
Each candidate should spend some time going over his or her notes, the textbook and (especially)
past examination papers, making a note of subjects and formulae of particular importance. For this
purpose a set of postcards/computer cards is useful: each topic may be summarised on one postcard,
enabling the candidate to learn the most important facts and then have a “revision aid”. (He or she
should ask a friend to choose a card at random, to give the title and to ask for a description of the
topic.)
(i) technical aspects, such as the correct use of actuarial tables and commutation functions, and
(ii) arithmetic.
It is sometimes necessary to omit (ii) or even (i) under extreme time-pressure, but one should
leave space to return to the question later if time permits. On the other hand, it is satisfying to give
a complete answer, and this should be attempted if time is not short. One should keep the left-hand
pages of the examination book free for rough working and arithmetic.
On first seeing the examination paper, one should tick the questions about which one is most
confident - especially bookwork. Leave the less familiar questions until later. One should ensure that
an approximate time-scheme (say, 15 minutes per 10-mark question) is adhered to, within reasonable
limits. Routine arithmetic (or even looking up the tables) may sometimes be left till near the end,
when one is too tired to think out new ideas but can still do (or check) arithmetic. As in musical
performances, it is essential to get off to a good start, by being in the right frame of mind and
picking the most suitable questions to attempt first.
333
Appendix B
2. In the A1967-70 section of “Formulae and Tables” commutation functions are given only at 4%
interest. For other interest rates one most use the more limited tables provided, e.g.
lx+n
äx:n at 6% = äx − v n äx+n
lx
lx+n Dx+n
(There is no point writing v n as since Dx and Dx+n are not given at 6% in the tables.)
lx Dx
3. Remember to make use of the functions maturing at ages 60, 65, etc. This saves time.
4. If only limited tables are available, one must proceed directly, using suitable approximations. For
1
example, suppose that one is asked to evaluate Ā50.5 :2.5 at 10% interest on A1967-70 ultimate.
This may be done by the trapezoidal rule as follows:
Z 2.5 · ¸
2.5 l53 µ53
v t t p50.5 µ50.5+t dt + µ50.5 + v 2.5
0 2 l50.5
5. Remember that for a(55) there are two tables - male and female (each with a select period of 1
year.) Assume that males and females are subject to the table of the appropriate sex (unless the
question states otherwise.)
6. Be careful (especially in A1967-70, but also in a(55)) that you distinguish between “select” and
“ultimate” tables and use the correct rate of interest.
334
Appendix C
WRONG CORRECT
1 1
1. Āx:n ; (1 + i) 2 Ax:n Āx:n ; (1 + i) 2 Ax1:n + Ax:n1
1
(1 + i) 2 (Mx − Mx+n ) + Dx+n
=
Dx
i 1 1
2. āx + ax āx ; ax + 2 or äx − 2
δ
µ ¶µ ¶
i m−1 Dx+n
3. ä(m)
x:n
; ax:n ä(m)
x:n
; äx:n − 1−
d(m) 2m Dx
Z t · ¸t Z t · ¸t
£ ¤t cr+1 cr
5. c dr = cr 0 or
r r
c dr = (Note that cr = er log c )
0 r+1 0 0 log c 0
lx0 lx0
9. µx = µx = −
lx lx
µZ t ¶ µ Z t ¶
10. t px = exp µx+r dr t px = exp − µx+r dr
0 0
335
Appendix D
1. Mid-point rule
Z b µ ¶
a+b
f (x)dx + (b − a)f
a 2
2. Trapezoidal rule
Z b µ ¶
b−a
f (x)dx + [f (a) + f (b)]
a 2
3. Simpson’s rule
Z b µ ¶· µ ¶ ¸
b−a a+b
f (x)dx + f (a) + 4f + f (b)
a 6 2
4. Three-eighths rule
Z b µ ¶· µ ¶ µ ¶ ¸
b−a 2a + b a + 2b
f (x)dx + f (a) + 3f + 3f + f (b)
a 8 3 3
336
336
age
i i
x Cxi Mxi Rx z
Cxi z
Mxi z
Rx
20 0.0 188.8 6835.5 0.0 904.4 33753.7 TABLES FOR
21 0.0 188.8 6646.7 0.0 904.4 32849.3 LUMP SUM
22 0.0 188.8 6457.9 0.0 904.4 31944.9 BENEFITS ON
23 0.0 188.8 6269.1 0.0 904.4 31040.5 RETIREMENT
24 0.0 188.8 6080.3 0.0 904.4 30136.1
337
SUPPLEMENT
age
r r
x Cxr Mxr Rx z
Cxr z
Mxr z
Rx
20 0.0 1524.0 65671.0 0.0 8048.7 347094.1 TABLES FOR
21 0.0 1524.0 64147.0 0.0 8048.7 339045.4 LUMP SUM
22 0.0 1524.0 62623.1 0.0 8048.7 330996.8 BENEFITS ON
23 0.0 1524.0 61099.1 0.0 8048.7 322948.1 RETIREMENT
24 0.0 1524.0 59575.2 0.0 8048.7 314899.5
338
SUPPLEMENT
Functions for valuing (at interest rate i per annum) refunds on death of contributions with
compound interest at interest rate j per annum
i = .04
j = 0.03
j d d
x Md j
Rx sj
Rx
P64 x j d P64 −(y+.5) j d P 64 d
= y=x Cy = y=x (1 + j) . M y = y=x (1 + j)−(y+.5) .sj M y
20 3108.18 31393.46 87740.98
25 2878.57 23700.42 76568.15
30 2700.59 17508.94 63315.76
35 2524.59 12500.67 49478.97
40 2338.12 8479.14 36324.46
45 2120.85 5291.34 24305.71
50 1804.15 2861.89 13941.80
55 1333.41 1177.45 6005.02
60 655.64 243.17 1286.33
61 490.11 147.36 784.29
62 350.45 79.12 423.29
63 225.80 33.70 181.20
64 108.85 8.09 43.67
Notes.
(1) j Cxd = (1 + j)x+.5 .(1 + i)−(x+.5) .dx
P64−x d
(2) j Mxd = t=0 j C x+t
d d 1 d
(3) j M x = j M x − 2 · jCx
sj d j d
(4) M x = sx · M x
339
SUPPLEMENT
Functions for valuing (at interest rate i per annum) refunds on withdrawal of contributions with
compound interest at interest rate j per annum
i = .04
j = 0.03
j w w
x Mw j
Rx sj
Rx
P64 x j w P64 −(y+.5) j w P64 w
= y=x Cy = y=x (1 + j) . M y = y=x (1 + j)−(y+.5) .sj M y
20 43547.50 119414.16 207941.08
25 19045.16 42216.97 98759.32
30 8041.95 13393.17 38299.39
35 3006.98 3281.50 10746.75
36 2399.53 2334.90 7831.22
37 1875.23 1608.25 5520.49
38 1426.97 1063.28 3732.97
39 1049.20 666.53 2391.95
40 735.82 388.85 1425.63
41 482.26 204.89 767.03
42 284.70 92.43 353.18
43 140.12 31.95 124.58
44 46.19 6.20 24.67
Notes.
(1) j Cxw = (1 + j)x+.5 .(1 + i)−(x+.5) .wx
P64−x w
(2) j Mxw = t=0 j C x+t
w w w
(3) j M x = j M x − 21 · j C x
w w
(4) sj M x = sx · j M x
340
SUPPLEMENT
Functions for valuing (at interest rate i per annum) refunds on death of contributions with
compound interest at interest rate j per annum
i = .04
j=0
j d d
x Md j
Rx sj
Rx
P64 x j d P64 −(y+.5) j d P 64 d
= y=x Cy = y=x (1 + j) . M y = y=x (1 + j)−(y+.5) .sj M y
20 792.85 18575.75 56093.05
25 674.16 14932.38 50789.94
30 594.98 11769.88 44005.80
35 527.62 8965.83 36247.43
40 466.07 6483.42 28117.45
45 404.38 4303.49 19889.59
50 326.91 2467.10 12048.69
55 227.54 1071.57 5469.33
60 104.08 232.42 1229.62
61 76.40 142.17 756.75
62 53.72 77.11 412.57
63 34.07 33.21 178.59
64 16.18 8.09 43.67
Notes.
(1) j Cxd = (1 + j)x+.5 .(1 + i)−(x+.5) .dx
P64−x d
(2) j Mxd = t=0 j C x+t
d d 1 d
(3) j M x = j M x − 2 · jCx
sj d j d
(4) M x = sx · M x
341
SUPPLEMENT
Functions for valuing (at interest rate i per annum) refunds on withdrawal of contributions with
compound interest at interest rate j per annum
i = .04
j=0
j w w
x Mw j
Rx sj
Rx
P64 x j w P64 −(y+.5) j w P64 w
= y=x Cy = y=x (1 + j) . M y = y=x (1 + j)−(y+.5) .sj M y
20 20579.08 103387.17 180950.89
25 7847.29 37164.60 87256.75
30 2916.96 12076.18 34605.99
35 971.53 3051.10 10002.54
36 758.81 2185.93 7337.82
37 580.57 1516.24 5208.20
38 432.61 1009.65 3546.59
39 311.55 637.57 2288.96
40 214.05 374.77 1374.41
41 137.46 199.01 745.20
42 79.53 90.52 345.94
43 38.36 31.58 123.14
44 12.40 6.20 24.67
Notes.
(1) j Cxw = (1 + j)x+.5 .(1 + i)−(x+.5) .wx
P64−x w
(2) j Mxw = t=0 j C x+t
w w w
(3) j M x = j M x − 21 · j C x
w w
(4) sj M x = sx · j M x
342