F E D E R A L SECURITY A G E N C Y

UNITED STATES PUBLIC HEALTH SERVICE

NATIONAL OFFICE OF VITAL STATISTICS

UNITED STATES LIFE TABLES

and
ACTUARIAL TABLES
1939-1 941

BY
THOMAS N. E. GREVILLE
Actuarial M a t h e m a t i c i a n /

Prepared under the supervision o/

HALBERT L. DUNN, M. D. .
CHIEF, NATIONAL OFFICE OF VITAL STATISTICS

UNITED STATES
GOVERNMENT PRINTING OFFICE
WASHINGTON : I947

For role by the Superintendent of Documents, U. S. Government Printing Office, Washington 45, D. C. - Price 31 .P5
 PART I1
tIFE TABLES
r
This. part contains the principal life tables presented Columns 6 and 6--Stationary population (Lzand
in this volume. Life tables are giqen for whites, T,).-Suppose that a group of 100,000 individuals like
Negroes, and other-races, separately by sex, and for that assumed in columns 3 and 4 is born every year, each
both sexes combined, and also for the total population such group being iubject throughout life to b e rates of
- and for total males and total females. This makes mortality shown in column 2. If there were no migra-
altogether 12 life tables. I n addition, table 13 gives, tion and if the births were evenly distributed over the
for the sa.me 12 classes and combinations of classes, calendar year, the survivors of these births would make
-
life table palues for certain subdivisions of the first up what is called a stationary population because in
year of life. A4 these tables are based on the 1940 such a population the number of persons living in any
census of population and the deaths of t.he 3-year given age group would never change. When an indi-
period 1939-1941. vidual left the group, either by death or by growing older
and entering the next higher age group, his place would
Explanation of the columns of the life table immediately be taken by someone entering from the
. Both the descriptive titles and the conventional next loyer age group. Thus, a census taken at any
actuarial symbols appear a t the head of the columns in time in such a stationary community would always
e.ach of the tables. The description which follows gives show the same total population and the same numerical
a more detailed explanation of each column of the life distribution of that population among the various ages.
table, and may be helpful to some readers. In such a stationary population, column 3 shows the
, Column 1-Year of age (zto x+ l).-The year of age, number of persons who, each year, reach the birthday
shown in column 1, is the interval between two succes- indicated in column 1 while column 4 shows the num-
sive birthdays. For instance, “4-5” indicates the in- ber who die each year in the indicated age interval.
terval between the fourth birthday and the fifth, in Column 5 shows the number of persons in the sta-
other words, the fifth year of life. tionary population in the indicated age interval. For
Column 2-Mortality rate (1,000 p,).-This colnmn example, the figure given for white males in the year of
shows the number of deaths within 1 year after the life 4 5 4 6 is 83,962. This means that in a stationary
birthday indicated, among .1,000 persons alive on that population of white males supported by 100,000 annual
birthday. For example, the rate of mortality a.t age births and subject always to the rates of mortality
45 for white males (table 5) is 7.66 per 1,000. In other shown in column 2 , a \census taken on any date would
words, during 1939-1941, 7.66 out of every 1,000 white show 83,962 persons between 45 and 46 years old.
males who were alive on their forty-fifth birthday Column 6 shows the total number of persons in the
. died before reaching age 46. The rates of morta.lity stationary population (column 5) in the indicated age
form the basis of the life table, all the other columns interval and all subsequent age intervals. For example,
being derived from them. in the stationary population of white males referred to
Column .!&-Number living (12) .-This column shows in the last illustration, column 6 shows that there would
the number of persons who would survive to each age be at any given moment a total of 2,180,567 persons
out of a cohort of 100,000 live births, subject throughout who have passed their forty-fifth birthday. The popu-
life to the rates-of mortality shown in column 2. Thus, lation at all ages 0 and above (in other words, the total
table 5 shows that out of 100,000 white male babies population of the stationary community) would be
born alive, 95,188 will complete t.he first year of life 6,281,188.
and enter the second; 94,724 will begin the third year; Column 7’-Average future lifetime (d,) .-The average
92,098 will reach age 21; and 33,404 will live to age 75. future lifetime (also called the complete expectation of
Column &-Number dying (d,) .--This column shows life) at any age is the average number of years remain:
the number dying in each successive year of age out of ing to be lived by those surviving to that age, on the
100,000 live births. Out of 100,000 white males born basis of a given set of mortality rates. The values in
alive (table 5), 4,812 die in the first year of life, 464 in column 5 can also be interpreted in terms of a single life
- the second year, 195 in the twenty-first year, and 2,762 table cohort, without introducing the concept of the
in the seventyififth year. Each figure in column 4 is the stationary population. From this point of view, each
difference between two successive figures in column 3. figurein column 5 represents the total time (in years)
21
 -7

22 . UNITED STATES. LIFE TABLES AND ACTUARIAL TABLES

lived between the indicated birthdays by those reaching by all the 97,194 survivors to the age of exactly 2 weeks
the earlier birthday among the survivors of a cohort of in the life t.able cohort which started with 100,000 white
100,000 live births. Thus, the figure 83,962 for white male births. Alternatively, it is the entire population
males in the year of life 45-46 is the total number of a t all ages beyond 2 wecks in the stationary population
years that will be lived bctween the forty-fifth and already referred to. Finally, the average future life-
’
forty-sixth birthdays by trhe 84,285 (column 3) who time of 64.59 shown in column 7 is the average number
reach their forty-fiit.11 Lirlhday out of 100,000 whitc of years lived beyond the first 2 weeks of life by the f

males born alive. The corresponding figure in column 97,194 survivors to the age of esactly 2 weeks in the
6 (2,180,567) is the total number of years that will be lifc table cohort.
lived after attaining age 45 by thc 84,285 reaching that Use of life tables in estimating and forecasting popula-
age. This number of yeak divided by t,he number of tions
persons (2,180,567 divided by 84,285) gives 25.87 a.s the
average future lifetime of white males at age 45. One of the most important applications of life tables
Care must be exercised in drawing conclusioiis from in demographic research is their use in estimating the
the figurcs in column 7. Thus, observing that the “ex- age distribution of a population on a given postcen$
pectation of life” at birtsh is alwaj7s gmater for white date. I n partic,ular cmes, this may be either a past,
persons than for Negroes, one should not conclude t.hat psesent, or future date. While an exhaustive discussion
the ‘oldest ages reached by white persons necessarily
of the subject would be beyond the scope of this vol-
exceed those attained by the‘most long-lived Negroes. ume,’ an outline of the general procedure will be given.
The difference in the average length of life is due to the Basically this consists, in the usual method of population
fact that a ‘greater proportion of Negroes die before projection, in multiplying the number enumerated a t
each age in the census by a survival rate derived from a
keaching old age. For example, t8henumber surviving
to a.ge 65 out of 100,000 born dive is far greater among life table, in order to obtain the estimated number of
survivors on the given date. It is usuallx most appro-
white;; than among Negroes; yet the average length of
priate to obtain the survival rates from the L, column’
life remaining a t age 65 is practically the same €or both
races. of the life table (column 5 of the tables on ppl26 to 49).
For example, suppose that in a certain group of white
Table IS-Subdivisions of the$& year of lije.-What
ma.les there were enumerated, in the ,1940 census, 32,000
has been said about the various columns of the life table
a t age 47 on the last birthday, and that it is desircd to
applies also, with certain obvious modifications, to tlic
estimate the number of survivors just 6 years later
3ife table values for subdivisions of the first year of life,
(that is, on April 1, 1946), on the supposition that the
given In table 13. The figures corresponding to age
mortality during the 6-year period will be approximately
“2-3 weeks” for white males may be taken as an illus-
the same as that indicated at.the ages in question by
tration. The age interval (column 1) is the. period be-
ginning with the exact age 2 weeks and cxtcnding up to the 1939-1941 life table for white males. Now the
original group of 32,000 presumably included persons
the cxactage 3 wecks: in other words, the third week of
life. The mortality rate of 1.64 in column 2 meam that a t all ages bctween exact age 47 and exact age 48, and
out of every 1,000 white m d c infants dive exactly 2 was, therefore, similar in its age composition to the
group at age 47 on the last birthday in the stationary
weeks after birth during 1939-1941 this number, on the
average, died during the following week. The number life table population, which numbered 82,568. Now,
living (97,194 in column 3) sigdles that this many since the hypothetical life table population does not I

‘
change with the passage of time either in its total
would still be alive oxactly 2 wecks after birth out of tpe
. 1ife.table cohort of 100,000 live births, on the assump- number or in its age composition, the survivors 6 years
later of this group of 82,568 would be merely the number
tion tXat the mortality rates shown in column 2 have
at age 53 in -the life table population, which is 76,953.
prevailed during the first 2 weeks of life. The number
Therefore, the survival rate to be applied to the group.
dying (159 in column 4) means that out of the 97,194
alive exactly 2 weeks after birth this number would die
of 32,000 is 76,953 divided by 82,568, which is .93200;
a n d . the estimated, number of survivors is 32,000
during the hllowiqg week. The figure 1,861 in column
multiplied by .93200, which gives 29,824. I n algebraic
5 indicates that during the third week of life the sur-
vivors of the life table cohort of 100,000 white male terms, the Lz+Bpersons aged x+6 in 1946 are the sur-
births have lived a total of 1,861 person-years of’life. vivors of the L, persons aged 5 in 1940. Therefore, the
. Or, alternatively, this figure is the number of infants
survival rate to be applied to the-population at age x
aged 2-3 wecks in a stationary population of white is Lz+aIL.z.
If migration during the 6 years is thought to have
males supported by 100,000 annual births and subject
been a signiGcant factor, it is of course .necessary to
always to the mortality rates shown for white males in
column 2 of this table m d of table 5. The figure 1 For a detailed discussion of the subject. see Eatimotca OJ Future Population ojfhe

6,277,446 in column 6 represents the total number of UnifcdStates, 1 0 4 0 , W (prepared by Warren S. Thonipson and P. K. Whelpton. and
lssued by the National Resources Planning’Board), Government Printing Ofice,
person-years of life lived beyond the first2 weeks of life Washington, D. O., 1943. !

i
 *
\

LIFE TABLES 23 .
obtain some information or to make some assumption figure 11. Perhaps the most obvious characteristic of
as to the number and age composition of the net mi- this distribution is that it is bimodal: that is, it has two
grants each year, and to adjust the number of survivors modes or maxima, one in the year of age 0-1 and an-
accordingly. other ib the year of age 75-76. The mode at age 0-1
I
Estimation of the populations at ages under 6 on is the higher, more deaths occurring in this than in any
April 1 , 1946, would require a knowledge of the number other single year of age. It is also clear th&t the fre-
of births during each of the 6 years. For example, sup- quency distribution is decidedly skewed toward the
pose 51,000 white males entered’ the group through left: that is, the frequencies rise very gradually from the
birth during the year April 1, 1943, to April 1, 1944. “trough” at age 10 to the ‘“peak” at age 75, and then
On April 1, 1946, the survivors of lhese births would be drop off sharply above age 75. T h e arithmetic mean of
between exact ages 2. and 3. Now, in the life table the distribution is the average age at death in the hypo-
population, the number of births during m y yew is the thetical cohort, or in other words, the average duration
radix of the life table-in this case, 100,000-while the of life. Its value in this case is 62.81 years (column 7
number. of survivors on a date just 2 years d t e r the end of the life tabl;) :It is clear that the value of the arith-
of the year in which the births occurred would be metic mean is very much influenced by the large num-
merely the number at age 2 in the life table population ber of deaths in the first year of life. If the deaths
(or 94,592 in column 5). Therefore, the survival rate to occurring in the first year were excluded from the dis-
be applied to the 51,000 births is 94,592 divided by tribution, the average age a t death of the remaining
100,000 which is .94592; and the estimated number of 95,188 individuals would be one plus the average future
survivors is 51,000 multiplied by .94592, which gives lifetime at age 1: that is, 65.98 years. This represents a t

48,242. ‘In algebraic terms, the survival rate to be difference of more than 3 years in the value.
applied to the births of the nth year preceding the date The median of the distribution (that is, the value
of the estimate is L,,J?o.
I n the original example of the 32,000 enumerated at
which has the same number of elements on either side
of it) is the median length of life, or probable lifetime,
I
sage ‘47, suppose it had been desired to estimate the another possible measure of longevity to whick refer-
number of survivors 6 months later, on October 1, 1946. ence was made in part I.3 Since the distribution of I

These individuals would then be at ages ranging from ages at death in a life table cohort is always character-
exact age 53j4 to exact age 54:h. Now the number of ized by a greater’ dispersion below the median value
persons between these ages in the life table population than above it, the median always exceeds the arith-
is approximately Z, (column 3) : that is, the number of metic mean. I n the particular case under considera- , I
survivors to age 54 out of the life table cohort of 100,000 tion, the median is 68.67 ypars, which exceeds the mean
live births, as indicated in column 3: I n this particular value by 2.69 years.
case, the figure is 76,380. Therefore, the survival rate I n part I the longevity of differqt subdivisions of
to be applied is 1541L47, or 76,380 divided by 82,568, the population was compared also by means of a third
which is .92506; and the estimated number of survivors criterion, the number of persons-surviving to specified
is 32,000 multiplied by .92506; which gives 29,602. ages in the hypothetical cohort. Which of these is
If the population data are given in’5-year age groups, the better measure of longevity is a question that
or can be combined into such groups, it is possible to cannot be answered categoricdy. The answer per-
shorten the arihhmetic wi,th very little loss of accuracy haps depends primarily on the purpose such a measure
by using an average survival rate for each 5-year age is intended to serve. Certainly no one figure can con-
group as a whole. Thus, the survival rate over a 6-year tain within itself all the information which is provided
period for the age group x to s+4 would be by the complete frequency distribution.
I n view of the pronounced skewness of the distribu-
(TI+B- TZ+IJ+(Tz- TZ+3. Other situations which may
arise can be dealt with along similar links. tion, it may belfelt that the arithmetic average is not
sufficiently rcpresentative. The layman, in inquiring
The life table as ;frequency distribution I what is the ‘(life expectancy” of a newborn infant,
The agegat death in the hypothetical life table cohort probably has in the back of his mind the idea of an
(as show; in colunm 4 of the life tables on pp. 26 to 49). age to which the infant has a reasonably good chance
cpnstitute a frequency distribution. I n the following of surviving. If he is told that the infant’s “expecta-
I discussion, the case of the life table for white males tionrof life” is 62.81 years, he may be surprised to be
(table‘5) will be taken as an illustration, but the remarks told later that more than 62 percent of whjte male
to be made will apply equally to all the.life tables, ex- infants alixe at birth outlive their expccthtion of life
cept for some difference in the ages and numerical while less than 38 percent die’before reaching that age.
values quoted. The fiequency distribution based on The alternative statement that 68.67 years is the prob-
the white males life table is exhibited graphically in able lifetime, the age to which the infant has a fifty-fifty
chance of surviving, is probably a more satisfactory
P of a life table is the number of births with which thc lifctahle cohort
T ~ rodf
begins, or, in algebraic terms, the value of 1.. In the tables on pp. 26 to 49, this is
answer to the. layman’s question.
&own In column 3 opposite the year of age 0-1, and is always 100,000. ‘see P. a.
 ? *'
rn

24 UNITED STATES LIFE TABLES AND ACTUARIAL TABLES

~L-FREQUENCY
FIGURE OF AGESAT DEATH
DISTRIBUTION I N A COHORT.STARTING W I T H 100,000 LIVEBIRTHS,
OF W H I T E MALES:U N I T E D STATES, 1939-1941
BASEDON THE MORTALITY .

4mJ

.-E
h
a
2 3mo
2,
b
a
@u
0
2000
P
E

1.000

0
0 IO 20 40 .so 60 70 80 90 100
Age at death in years

On the other hand, the objection may be made that mortality' a t any age, or in the ages at death in the life
the probable lifetime is not SufEiciently sensitive. to table cohort.
changes in the ages at death of the members of @e life Use of the life table in studying there productive capac-
table cohort. In fact, its value is not 'affected by any ity of populations
change in which the age at death of a0 individual is
Another important application of life tables in demo-
not actually shifted from one side to the other of the
graphic research is their use in conjunction with fer- .
probable lifetime its'elf. If, for example, the deaths of
tility rates in investigating the inherent capacity of a
the 4,812 dying before age 1 in the white males life
population to reproduce itself. This is studied, for the
table were equally spread over all the years of age
most part, by means of certain specific measures devised
between birth and- age 68, many of these individuals
for that purpose, the most important of which are the
would live much longer; yet the value of the probable
gross and net reproduction rates and the true rate of
- lifetime would be unchanged. However, the effect of
natural increase: While life table survival rates are
transferring deaths from one age to another in the
an important component in the-calculation of these
hypothetical life table cohort is not entirely relevant,
measures, they involve other considerations of a highly
since the mortality rates in the life table were not ob-
technical-nature, which are outside tho scope of this
tained by observing a single cohort over a period of
volume.
time, but rather by observing many cohorts simul-
taneously, a different one at each age. Therefore, the Mathematical notation employed
important thing is the effect of a specified change in the One of the mathematical symbols used in the head-
rate of mortality at a particular age, without reference ings of table 13 represents a departure from the stand-
- to any offsetting change elsewhere. Any chmge in ard notation in use by actuaries. This is the symbol
the'mortality rate at any age less than the probable See Robcrt It. Kuczynskl, The BaZance a/ Births and Deafhr, 2 vols., The Mac-
4

Co., New York, 19% Ferlility and Reproduction, Falcon Press, NPWYork, - 1
lifetime (unaccompanied by other changes) will ' alter millan
1932; The Alenduremmf of Populntion Growth, Oxford University Prcss, Now York,
the value of the probable lifetime. However, changes 1%?.70D. ; V. Qlasr, .Populalion Polieie8 and Mouemenls in Europe (Appendix), Oxrord
in mortality rates at ages greater than the probable University Prcss, I,ondon, 1040.
8 Soc Louis 1. Dublin and Alfrcd J . Lotka. Length o/Li/c, The Ronald Press Co.,
lifetime will have no .effect whatever on its value. New York, 1936; On the True Rate nl Nafural Increase, Journal nf the Americnu Statls-
Similar remarks apply to the third criterion suggested, tical Asaocintion, \-ol. 20,No. 151. pp. 305-339, Scplcmher 192.5; AIrrcd J. Lotka. The
Qeoyraphic Distribution @jIntrinsic Natural Increase in f k c United Slafea, and an Eram-
the number of survivors to a designated age. The inotion OJ the Relation Bctmeen Scueral Alrasures OJ Ne1 Reprodudiofy, ibid., vol. 31,
value of the average duration of life, on the contrary, is No. 101, pp. 273-204, June 1936; Some Recent &&a in Population Analyais, ibid.,
vol. 33, No. 201, lip. 164-li8, March 1038. Sce also Glass' book citcd In the preceding
affected in some measure by any change in the rate of lootnolo.