Sie sind auf Seite 1von 6

Mortality Laws

De Moivre (1725) [7] used a 1-parameter formula:


1
; equivalent to
x

x
(1)
x p0 = 1

De Moivre used age 86 for .


Lambert (1776) [17] used a 4-parameter formula:


 x
x
ax 2
b e c e d
(2)
x p0 =
x
x =

Mathematical Formulae
From the time of De Moivre (1725), suggestions have
been made as to the law of mortality as a mathematical formula, of which the most famous is perhaps that of Gompertz (1825). Since John Graunt
(16201674) life tables had been constructed empirically. These life tables represented mortality over
the whole human lifespan, and it was natural to
ask if the mathematical functions defined by the life
table could be described by simple laws, as had been
so successfully achieved in natural philosophy. The
choice of function that should be described by a
law of mortality (a mathematical formula depending
on age) has varied, as different authors have considered x , qx , or mx (see International Actuarial
Notation) (or something else) to be most suitable; in
modern terminology we would model the parameter
that is most natural to estimate, given the underlying
probabilistic model for the data.
It should be noted that, since qx and mx are
all 1 to first order, it makes little difference
x+

at younger ages whether the formula is used to


represent the functions x , qx , or mx (see below for
advanced ages).
Since laws of mortality are attempts to summarize
empirical observations, they have been intimately
linked with the statistical techniques of analyzing
mortality data, and, nowadays, would be described as
parametric forms for quantities appearing in statistical
models. In some cases, these models may have a
physiological basis and may attempt to model the
ageing process.
A typical modern pattern of mortality can be
divided into three periods. The first period is the
mortality of infants, namely, a rapid decrease of
mortality during the first few years of life. The
second period contains the so-called accident hump
where the deaths are mainly due to accidents, for
example, in cars or motor bicycles. The third period
is the almost geometric increase of mortality with age
(the rate of increase slowing down after age 80)
senescent mortality.
We will now describe the most important laws of
mortality that have been proposed and their features.

Babbage (1823) [2] used a 2-parameter formula,


assuming that x p0 was quadratic (rather than linear,
as De Moivre had assumed)
x p0

= 1 bx ax 2 ,

(3)

for suitable values of a, b, c, see [1].


(First) Gompertz (1825) [11] used a 2-parameter
formula:
x = Bcx
or equivalently
x = e1 +2 x ;

x p0

= ek(c

1)

(4)

The important conclusion of Gompertz was that the


force of mortality (practically the same as the rate
of mortality for one-half of a year less except for
high ages) increased in geometric progression with
age. This simple law has proved to be a remarkably
good model in different populations and in different
epochs, and many subsequent laws are modifications
of it, to account for known deviations at very young
or very old ages, or for particular features.
Young (1826) proposed a very complex mathematical expression involving x 40 [1, 35].
Littrow (1832) [19] used a polynomial of degree
higher than 2 (as an extension of Babbage).
Moser (1839) [23] used a 5-parameter formula of
the form:
1

x p0

= 1 ax 4 + bx 4 cx

17
4

dx

25
4

+ ex

33
4

(5)
where the values (a, b, c, d, e) for Brunes Tables,
used by the celebrated C. F. Gauss for the Widows
Fund at Gottingen University are given in [23].
Second Gompertz (1860) [12] was a 10-parameter
formula:
x

ln lx = bcx + ghx xdf x j k m

(xn)

(6)

Mortality Laws

This was intended to represent mortality over the


whole span of life, which was not adequately described by the first Gompertz law. Note that the derivative of the function ln l(x) is equivalent to the force
of mortality x . This formula seemed to have been in
advance of its time but was too complex for normal
practical use.
Third Gompertz (1862) [13] was a version of the
second Gompertz formula above.
First Makeham (1867) [20] was a 3-parameter
formula:
x = A + Bcx
or equivalently
x = 1 + e2 +3 x ; x p0 = ek(bx+c

1)

(7)

This extended the first Gompertz law simply by


including a term independent of age, representing
nonsenescent deaths, for example, from accidents.
Double Geometric (unknown author) [16, 18, 24]
was a 5-parameter formula:
x = A + Bcx + Mnx

(8)

Oppermann (1870) [25] proposed a 3-parameter


formula:
1

x = ax 2 + b + cx 3 ,

(9)

which only applies for young ages, x 20.


Thiele (1871) [31] proposed a 7-parameter formula:
x = AeBx + CeD(xE) + F Gx ,
2

(10)

which covers the whole span of life. Each term in this


formula represents one of the recognizable features
of human mortality mentioned above. The first term
represents infant mortality, declining steeply after
birth; the second term represents the accident hump;
the third term is the Gompertz law appropriate at
older ages.
Second Makeham (1890) [21] was a 4-parameter
formula:
x = A + H x + Bcx

x =

A + Bcx
1 + Dcx

+cx 1)

(11)

(PER1)

(12)

PER1 is the logistic curve, equivalently expressed


GH x
as x = A + 1+KGH
x (see HP3 below). Perks (1932)
[26] also proposed a 5-parameter formula:
x =

A + Bcx
+ 1 + Dcx

Kcx

(PER2)

(13)

The effect of the denominators is to flatten out the


exponential increase of the Gompertz term in the
numerator, noticeable at ages above about 80, which
is now a well-established feature of the mortality.
PER1 results, if each member of a population has
mortality that follows Makehams Law x = A +
Bcx but B, which is fixed for an individual member of the population, follows a gamma distribution
(see Continuous Parametric Distributions) for the
population as a whole. PER1 was used in the graduation of the table of annuitant lives in the United
Kingdom known as a (55).
Weibull (1951) [33] proposed a 2-parameter
formula:
 B+1 
x = Ax B ;

x p0

=e

x
B+1

(14)

The English Life Tables (ELT) 11 and 12 [27]


(deaths 19501952 and 19601962) were graduated
using a 7-parameter curve


b
2

mx = 1 = a + ce(xx2 ) +
(xx1 )
x+
1
+
e
2
(15)
This mathematical formula was used for English Life
Tables 11 and 12 for an age range above a certain
age; in the case of ELT12 from age 27 upwards.
The method of splines was used for ELT13 and
14 [8, 22].
Beard (1971) [3, 5] proposed a 5-parameter formula:
qx = A +

or equivalently
x = 1 + 2 x + e3 +4 x ; x p0 = ek(bx+dx

Perks (1932) [26] proposed a 4-parameter formula


(this is known as the logistic curve):-

Ec2x

Bcx
+ 1 + Dcx

(16)

This formula was used to graduate the table of


assured lives known as A194952 in the United
Kingdom.

Mortality Laws
Barnett (1974) [4] proposed a 4-parameter formula:
f (x)
qx =
1 + f (x)

Third HeligmanPollard HP3 (1980) [14] was a


9-parameter formula:


x 2
C
qx = A(x+B) + DeE loge F

where

GH x
(HP3)
(22)
1 + KGH x
which covers the whole span of life. As an example,
the Figure 1 shows the published rates of mortality qx
of the English Life Tables No. 15 (Males) (ELT15M)
from ages 0 to 109 (this represents the mortality
in England and Wales in 19901992). ELT15 was
graduated by splines, but we can see that the third
HeligmanPollard law gives a very close fit, called
ELM15M(Formula) on the figure below. The parameters A 103 ,B 103 ,C 102 ,D 104 , E, F, G 105 , H, K
are 0.59340, 10.536, 9.5294, 6.3492, 8.9761, 21.328,
2.6091, 1.1101, 1.4243. This has K = 1.4243 givGH x
ing, at higher ages, qx = 1+1.4243GH
x (a Perks/logistic
curve with the constant term being zero) and an
asymptotic value of 0.7.
Fourth HeligmanPollard HP4 (1980) [14]
was a 9-parameter formula:
k
x 2
GH x
E loge
(x+B)C
F
qx = A
+ De
+
(HP4)
1 + GH x k
(23)
which covers the whole span of life.
Forfar, McCutcheon, Wilkie FMW1 and FMW2
(1988) [9] proposed a general family, incorporating
as many parameters (r + s) as are found to be
significant:

f (x) = A H x + Bc

(17)

This formula was used in the United Kingdom to graduate the experience of insured lives over the period
19671970, and this table is known as A 196770.
Wilkie (1976) [6, 34] proposed a family, incorporating as many parameters as are significant
qx =

f (x)
1 + f (x)

where
f (x) = exp


i x

i1

(18)

i=1

graduated ages.
First and Second HeligmanPollard HP1 and
HP2 (1980) [14] are two 8-parameter formulae that
cover the whole span of life:
qx =

f (x)
1 + f (x)

where
f (x) = A(x+B) + DeE
C

+ GH x

loge

x
F

(HP1)

qx = g(x) = A(x+B) + DeE


C

GH x
+
1 + GH x

loge

(HP2)


x 2

(19)
x = GM r,s (x) where GMr,s (x) =

qx =

r+s
i=r+1

i x

ir1

i x i1

i=1


(FMW1)

(24)

r,s

g(x)
where at higher ages
1 + g(x)
GH x
1 + 2GH x

+e

(20)

Because the first two terms are very small at higher


ages, HP1 and HP2 are practically the same. At
GH x
higher ages both are qx = 1+GH
x . The three terms in
the formula describe infant mortality (high mortality
in the first few years of life, particularly the first year
of life), the accident hump (centered around age 20),
and senescent mortality, as did Thieles law over a
100 years ago.
A variation is qx =

(HP2a)

(21)

qx =

GM (x)
= LGMr,s (x) (FMW2)
1 + GMr,s (x)
(25)

The formula for x incorporates (First) Gompertz


and First and Second Makeham. A mathematical
formula approach was used in the United Kingdom
to graduate the 80 series and 92 series of standard
tables for life insurance companies [6, 9, 22]. For
UK assured lives (two-year select period) and life
office pensioners (normal retirals, no select period)
in the 92 series, (life office deaths 19911994)
the mathematical formula x = GM(2, 3) was fitted
to x .

Mortality Laws
12

10

ln{1 00 000q(x)}

ELT15M
ELT15M (Formula)
ELT (limit)

0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
10
0
10
5
11
0
11
5
12
0
12
5

0
2
Age

Figure 1

Mortality ELTM, ELTM(FIT) and ELTM(limit)

Mortality at the Highest Ages


Whereas from ages 30 to 80 the yearly rate of
increase in mortality rates is almost constant (10%
11.5% a year in round terms), the rate of increase
slows down markedly above age 80 [2830]. The
GH x
term 1+KGH
x in the Third HeligmanPollard formula
is similar to a Perks (see above) term (with A = 0);
fitting HP3 to ELT15M gives an asymptotic value
of (1/K) = 0.7 for qx (see above). The asymptotic
value of x then approximates to 1.2 { ln(10.7)}.
At age 110, qx = 0.55, so the asymptotic value has
nearly (but not quite) been reached.
Since the rate of growth of x and qx
declines above age 80, perhaps the functional form
GM 1,2 (x)
for either x or qx may be preferable to
1+KGM 1,2 (x)
GM 1,3 (x), but this has not been tried.

Mortality Projections for Pensioners and


Annuitants [particularly CMI Report
No. 17]
Human mortality has changed dramatically during the
few centuries in which laws of mortality have been
pursued, notably during the twentieth century [6, 10,
15, 32]. Of interest in themselves, such changes
are of great financial importance to life insurers

(see Life Insurance). Generally we have seen mortality falling and longevity increasing, which makes
pensions and annuities more expensive, so actuaries have attempted to project such future trends in
order to price and reserve for annuity contracts. This
subject cannot be described in detail here, but a study
of the laws of mortality that have been found to apply
over time, and how their parameters have changed,
gives useful information. Projections of annuitant and
pensioner mortality have been made in the United
Kingdom since the 1920s. The projections have been
based on a double entry table where the axes of the
table are age and calendar year.

Future Mortality
Ever since Gompertz (1825), an exponential term has
appeared in some form in most laws of mortality; it
is convenient to focus on the parameter H (which is
around 1.10) in the Third HeligmanPollard formula.
This appears to be significant as it plausibly relates
to the natural aging of the human body, as originally
argued by Gompertz. Modern medicine does not
appear to have made a difference to H . In fact, H
appears to be increasing with time as Table 1 shows.
Eliminating all infant mortality (the first term
in the Third HeligmanPollard formula) and all

Mortality Laws
[6]

Table 1

Mortality table and date


English life tables No. 2 males 1841
English life tables No. 15 males 1991

Optimum
value of
H in HP3

[7]

1.099
1.113

[8]

accidents, (the second term) reducing qx to 1/3 of its


ELT15M(Formula) value, and running linearly into
0.9 times, the ELT15M(Formula) value of qx over the
ages between 60 and 120, but keeping the proportion
constant at 0.9 above age 120 will create a mortality
table denoted by ELT(limit) but will only put the
expectation of life for males at birth up from 76
(on ELT15M) to 83, on ELT(limit). The age of the
oldest person to die in the world (currently believed
to be Madame Calment of France, who died at age
122 years 164 days) may, however, not increase much
above 122 over the foreseeable future. On ELT(limit),
out of 100 million births per annum and a stationary
population of 8.5 billion there is less than 1 person
alive at age 124. This suggests that until we can find
some way of influencing the ageing process there is
a limit of, say, 85 years to the expectation of life for
male lives at birth and a limit to which the oldest
person in the world can live of, say, 125. ELT(limit)
is shown in Figure 1.

[9]

[10]

[11]

[12]

[13]
[14]

[15]

[16]

References
[17]
The references denoted by H. & S. refer to Haberman S. and
Sibbett T.A. editors (1995) History of Actuarial Science, 10
Volumes, Pickering and Chatto, London.
[1]
[2]

[3]

[4]

[5]

Adler, M.N. (1867). Memoir of the late Benjamin


Gompertz, Journal of the Institute of Actuaries 13, 120.
Babbage, C. (1823). On the Tables of Single and
Annual Assurance Premiums, Journal of the Institute of
Actuaries 6, 185.
Beard, R.E. (1971). Some aspects of theories of mortality, cause of death analysis, forecasting and stochastic
processes, in Biological Aspects of Demography, W.
Brass, ed., Taylor & Francis, London.
CMI Bureau (1974). Considerations affecting the preparation of standard tables of mortality, Journal of the
Institute of Actuaries 101, 135216. The formula on
page 141 has become known as Barnetts formula.
Continuous Mortality Investigation Committee (1956).
Mortality of assured lives, Journal of the Institute of
Actuaries 82, 384.

[18]

[19]

[20]

[21]

[22]

[23]

Continuous Mortality Investigation Reports (CMIRs),


Nos. 120 (19732001). Published jointly by the Faculty of Actuaries and Institute of Actuaries (obtainable
from the Institute of Actuaries), London.
De Moivre, A. (1725). in Annuities on Lives, see H. &
S. London, 3125.
Eilbeck, C. & McCutcheon, J.J. (1981). Some remarks
on splines, Transactions of the Faculty of Actuaries 37,
421.
Forfar, D.O., McCutcheon, J.J. & Wilkie, A.D. (1988).
On graduation by mathematical formula, Transactions
of the Faculty of Actuaries 41, 96245; Journal of the
Institute of Actuaries 115, 1149.
Forfar, D.O. & Smith, D.M. (1987). The changing face
of English Life Tables, Transactions of the Faculty of
Actuaries 40, 98.
Gompertz, B. (1825). On the nature of the function
expressive of the law of human mortality, and on a new
mode of deterrmining the value of life contingencies,
Philosophical Transactions of Royal Society, (Series A)
115, 513583; see H. & S. 2, 121191.
Gompertz, B. (1860). On one uniform law of mortality
from birth to extreme old age and on the law of sickness,
presented to International statistical congress 1860 and
reproduced in 1871, Journal of the Institute of Actuaries
16, 329344.
Gompertz, B. (1862). Philosophical Transactions of
Royal Society (Series A) 152, 511559.
Heligman, L. & Pollard, J.H. (1980). The age pattern
of mortality, Journal of the Institute of Actuaries 107,
4980.
Hustead, E.C. (1989). 100 Years of Mortality, Society of
Actuaries, Schaumburg, Illinois (obtainable from Society
of Actuaries).
King, A.E. & Reid, A.R. (1936). Some notes on the
double geometric formula, Transactions of the Faculty
of Actuaries 15, 229234.
Lambert, J.H. (1776). Dottrina degli azzardi, Gaeta and
Fontana, Milan.
Lidstone, G.J. (1935). The practical calculation of annuities for any number of joint lives on a mortality table
following the double geometric law, Journal of the Institute of Actuaries 66, 413423.
Littrow, J.J. (1852). Uber Lebensversicherungen und
Andere Versorgungsanstalten, F. Beckschen Universitats-Buchhandlung, Vienna, p. 52.
Makeham, W.M. (1867). On the law of mortality Journal
of the Institute of Actuaries 8, 301310; Journal of the
Institute of Actuaries 13, 325358; see H. & S. Vol. 2,
pp. 324334 and Vol. 8, pp. 73108, London.
Makeham, W.M. (1890). On the further development of
Gompertzs law, Journal of the Institute of Actuaries 28,
152159, 185192, 316332.
McCutcheon, J.J. (1984). Spline Graduations with Variable Knots, in Proceedings of the 22nd International
Congress of Actuaries, Vol. 4, Sydney.
Moser, L. (1839). Die Gesetze der Lebensdauer, Verlag
von Veit und Comp, Berlin, p. 315.

6
[24]

[25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

[33]

[34]

[35]

Mortality Laws
Olifiers, E. (1937). Graduation by double geometric
laws supplementary to Makehams basic curve, with
an application to the graduation of the A 192429
(ultimate) table, Journal of the Institute of Actuaries 68,
526534.
Opperman, L. (1870). Insurance record 1870, 4243,
4647; Journal of the Institute of Actuaries 16, (1872);
H. & S. 2, 335353.
Perks, W. (1932). On some experiments in the graduation of mortality statistics, Journal of the Institute of
Actuaries 63, 1240.
Registrar General (1951). Registrar Generals Decennial
Supplement for England and Wales-Life Tables, Her
Majestys Stationery Office.
Thatcher, A.R. (1990). Some results on the Gompertz
and Heligman and Pollard laws of mortality, Journal of
the Institute of Actuaries 117, 135142.
Thatcher, A.R. (1999). The Long Term Pattern of Adult
Mortality and the highest attained age, Journal of the
Royal Statistical Society, Series A 162, 544.
Thatcher, A.R., Kannisto, V. & Vaupel, J.W. (1997).
The Force of Mortality From age 80 to 120, Odense
University Press, Denmark.
Thiele, P.N. (1871). On a mathematical formula to
express the rate of mortality throughout the whole of
life, Journal of the Institute of Actuaries 16, 313329.
Vallin, J. Mesle, F. (2001). Tables de mortalite francaises
pour les XIX e et XX e si`ecles et projections pour le XXI e
si`ecle, Institut national des e tudes demographiques.
Weibull, W. (1951). A statistical distribution function
of wide applicability, Journal of Applied Mechanics 18,
293297.
Wilkie, A.D. (1976). in Transactions of the 20th International Congress of Actuaries in Tokyo, Vol. 2,
pp. 761781.
Young, T. (1826). A formula for expressing the decrement of human life, Philosophical Transactions of the
Royal Society 116, 281303.

Further Reading
Benjamin, B. & Pollard, J.H. (1993). The Analysis of Mortality and other Actuarial Statistics (obtainable from the
Institute of Actuaries).
Benjamin, B. & Soliman, A.S. (1993). Mortality on the Move,
City University Print Unit (obtainable from Institute of
Actuaries).
Gavrilov, L.A. & Gavrilova, N.S. (1991). The Biology of Life
Span: A Quantitative Approach, Harwood Academic Publishers, Chur, Switzerland, English translation of Russian
edition published in 1986.
Hartmann, M. (1987). Past and recent attempts to model
mortality at all ages, Journal of Official Statistics Sweden
3(1), 1936.
Horiuchi, S. & Coale, A.J. (1990). Age pattern for older
women: an analysis using the age-specific rate of mortality change with age, Mathematical Population Studies
2(4), 245267.
Nature (2000). 405, 744, 789.
Science (1990). 250, 634; (1992). 258, 461; (1998). 280, 855;
(2001). 291, 1491; (2001). 292, 1654; (2002). 296, 1029.
Tableau, E., van den Berg Jeths, A. & Heathcote, C. (2001).
Forecasting Mortality in Developed Countries, Kluwer
Academic Publishers, Netherlands.
Various authors (2002). North American Actuarial Journal 6(3),
1103.
Wunsch, G., Mouchart, M. & Duchene, J. (2002). The Life
Table, Modelling Survival and Death, Kluwer Academic
Publishers, Netherlands.

(See also Decrement Analysis; Early Mortality


Tables; Life Table)
DAVID O. FORFAR