You are on page 1of 7

John Graunt, John Arbuthnott, and the Human Sex Ratio

Author(s): R. B. CAMPBELL
Source: Human Biology , August 2001, Vol. 73, No. 4 (August 2001), pp. 605-610
Published by: Wayne State University Press

Stable URL: https://www.jstor.org/stable/41466835

REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/41466835?seq=1&cid=pdf-
reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Wayne State University Press is collaborating with JSTOR to digitize, preserve and extend
access to Human Biology

This content downloaded from


62.204.213.86 on Thu, 19 Nov 2020 10:41:09 UTC
All use subject to https://about.jstor.org/terms
John Graunt, John Arbuthnott, and the Human Sex Ratio

R. B. CAMPBELL'

Abstract John Graunt was the first person to compile data that showed an
excess of male births over female births. He also noticed spatial and temporal
variation in the sex ratio, but the variation in his data is not significant. John
Arbuthnott was the first person to demonstrate that the excess of male births
is statistically significant. He erroneously concluded that there is less varia-
tion in the sex ratio than would occur by chance, and asserted without a basis
that the sex ratio would be uniform over all time and space.

The statistical study of the human sex ratio required both the collection of data on
the human sex ratio and the development of statistical tests to determine the sig-
nificance of the data. Important dates for the collection of data include 1538 when
the Anglican Church instituted parish registers for the recording of christenings,
burials, and weddings; 1604 when weekly and annual bills of mortality (including
christenings) were published for all of London; 1629 when the sexes of those
christened were included; and 1662 when John Graunt (1662) published Natural
and Political Observations made upon the Bills of Mortality, which was the first
descriptive statistical analysis of such data (Hald 1990). The first important date
for the inferential statistical analysis of sex ratio data is 1710, when John Arbuth-
nott (1710) published "An Argument for Divine Providence, Taken from the Con-
stant Regularity Observed in the Births of Both Sexes." This work is often cited
as the first paper on inferential statistics (Bellhouse 1989; Eisenhart and Birn-
baum 1967) because of its demonstration that the excess of males over females is
not due to chance. These works by Graunt and Arbuthnott are important in the
history of statistics, and are also the earliest investigations of the human sex
ratio.

In his dedicatory epistle to John Lord Roberts, Graunt states that his pur-
pose in writing Natural and Political Observations made upon the Bills of Mor-
tality is "to have reduced several great confused Volumes into a few perspicuous
Tables, and abridged such Observations as naturally flowed from them, into a few
succinct Paragraphs." His book is recognized as one of the first examples of de-
scriptive statistics because he achieved this objective. The book discusses a wide
range of topics related to birth and mortality, and includes a life table. Of rele-

'Department of Mathematics, University of Northern Iowa, Cedar Falls, IA 50614-0506.

Human Biology, August 2001, v. 73, no. 4, pp. 605-610.


Copyright © 2001 Wayne State University Press, Detroit, Michigan 48201-1309

KEY WORDS: SEX RATIO, HISTORY OF STATISTICS

This content downloaded from


62.204.213.86 on Thu, 19 Nov 2020 10:41:09 UTC
All use subject to https://about.jstor.org/terms
606 / CAMPBELL

vanee to the sex ratio are the observations "that there be more males than fe-
males," and that "London is somewhat more apt to produce males than the coun-
try." The former is based on christening data for London from 1629 to 1661. He
does not remark on the constancy of the sex ratio across the years, but that is man-
ifest from the data. The latter is based on contrasting the London data with chris-
tening data for the town of Romsey from 1569 to 1658. He also remarks on the
variation over time of the sex ratio in Romsey.
During the years 1629-1661, 139,782 males and 130,866 females were
christened in London. These figures yield a sex ratio of 1.068, which Graunt ap-
proximates as a ratio of 14 to 13. The proportion of males is 0.5164. If one per-
forms the standard large sample test for a proportion to determine whether this
proportion is different from 0.5 [z = (p - po)/(po(l - po)/n )0 5; Weiss 1999], the z-
score is 17, which is very significant. However, this test was not available to
Graunt. In fact, the concept of statistical significance had not yet been developed.
Graunt cites providing one man for every woman after correcting for deaths in
war, drownings at sea, emigration to foreign lands, and unmarried Fellows of
Colleges as the reason for the excess male births.
During the years 1569 to 1658, 3256 males and 3083 females were chris-
tened in Romsey. These figures yield a sex ratio of 1.056, which Graunt approxi-
mates as a ratio of 16 to 15. The proportion of males is 0.5136. If one performs
the standard test for the difference between two proportions 'z = (p' - p2)/(p(l-
p)(l/rii + 1M2))° 5; Weiss 1999] for London versus Romsey, the resultant z-score is
0.44, which is not significant. Graunt does not explain the difference in sex ratios,
but conjectures that there may be other geographic variation in the sex ratio.
Graunt also remarks that there is great variation in the sex ratio between decades.
There is not significant excess variation in the Romsey data by decades, but a chi-
squared test [%2 = (n - 1 )s2/g2; Weiss 1999] on the annual data assuming the bi-
nomial distribution shows that there is excess variation at the 0.05 significance
level.

"An Argument for Divine Providence, Taken from the Constant Regularity
Observed in the Births of Both Sexes" (Arbuthnott 1710) is a short paper that is
concerned only with the sex ratio as illustrated in Figure 1. Arbuthnott concludes
that "provident Nature, by the Disposal of its wise Creator, brings forth more
Males than Females, and that in almost a constant proportion." The demonstra-
tion that there are significantly more males than females is identified as the first
use of inferential statistics. His contemporaries soon demonstrated that there was
not less variation in the sex ratio than would occur by chance (Shoesmith 1987).
Arbuthnott also asserts, without justification, that the excess of males shall hap-
pen in a constant proportion "not only for 82 Years, but for Ages and Ages, and
not only at London, but all over the World." He attributes the excess of males to
Divine Providence compensating for the greater mortality that males are subject
to, and concludes that "Polygamy is contrary to the Law of Nature and Justice."
The standard large sample test for a proportion to determine whether 0.516
(the observed proportion of males) is different from 0.5 employing all 938,240

This content downloaded from


62.204.213.86 on Thu, 19 Nov 2020 10:41:09 UTC
All use subject to https://about.jstor.org/terms
Graunt, Arbuthnott, and the Human Sex Ratio / 607

Figure 1. Christenings in London, 1629-17 10. John Arbuthnott noticed that all the proportions are
greater than one-half, but misjudged the amount of variation.

christenings in London during the years 1629-1710 produces a z-score of 31,


which is very significant. However, this test was not available to Arbuthnott. In-
stead, he performed what is today called a sign test by noting that there were more
males than females christened during each of the 82 years, which would occur
with probability (1/2)82 if male and female births were equally likely (equal fre-
quency of births entails equal frequency of years in which each sex has the ex-
cess). Since this number is so small, he concluded that art rather than chance act-
ing on equal frequencies governs the sex ratio. This rejection of a hypothesis
because the observed data would be unlikely if the hypothesis were true is cited
as the first use of inferential statistics.
Arbuthnott's argument that there is less variation than would occur by
chance is not valid (Hacking 1975). He merely states that "it is very improbable
(if mere chance govern' d) that they [frequencies of the sexes] would never reach
as far as the Extremities," without quantifying the probabilities. In fact, there is
twice as much variation (as measured by the variance) in the proportion of males
as would occur by chance. Assuming a population size (number of births per
year) of n- 10,398 (the harmonic mean of the numbers of births), a probability
p = 0.517 of births being male, and a binomial distribution, the variance of the
proportion of males should be 0.0000240 [p( 1 - p)/n', which is less than half the
observed variance of 0.0000520. This yields %2= 170 with 81 degrees of freedom
(Anscombe 1981), which is very significant.

This content downloaded from


62.204.213.86 on Thu, 19 Nov 2020 10:41:09 UTC
All use subject to https://about.jstor.org/terms
608 / CAMPBELL

At this point we can summarize the observations and analyses of Graunt


and Arbuthnott. Both Graunt and Arbuthnott correctly observed that there is an
excess of males over females christened. This is well known today (for births).
Arbuthnott demonstrated that the excess was significant by a sign test, and both
Graunťs and Arbuthnott's data sets provide a significant excess by the present test
for the binomial distribution (the Romsey data, which has a sample size of 6339,
is only significant at the 0.03 level). Graunt claimed there was great temporal
variation in his Romsey data, but the excess variation is not significant for the de-
cade groupings he employed. Arbuthnott claimed there was too little variation in
his London data, when in fact there is twice as much variation as would occur by
chance. It is now known (James 1987a, 2000) that there is unexplained excess
temporal variation in the sex ratio, but it does not reflect a long-range change over
time. Graunt observed that more males were born in the city than in the country,
but the difference in his data is not significant by present statistical tests. Arbuth-
nott claimed that the sex ratio would be constant over space and time, but provid-
ed no data to support that claim. It is now known (James 1987a) that there are eth-
nic and social class (perhaps psychological) differences in the sex ratio.
Any analysis of the data from Graunt and Arbuthnott should acknowledge
that it has limited utility. It is based on christening rather than birth statistics,
which entails in addition to the time delay between birth and christening that only
persons subscribing to the Church of England are counted. Annual mortality data,
including cause of death, is available for London, but there is less detail for some
years than for others (Creighton 1965). No reliable estimates of the population
size of London are available (Harding 1990). Although the christening and mor-
tality data both suggest that the population of London increased between 1623
and 1710 (more than doubled as linear growth), the available estimates of the
population of London indicate that the population did not double, and may have
even decreased. Furthermore, the occurrence of plague and the fire of London
preclude assuming linear or exponential (or even monotone) growth. The absence
of data on population size restricts the statistical analyses that can be done. Yet
when a significant association is found, its meaning should be considered; errors
in data tend to reduce rather than cause correlation.
The first result of reanalyzing the data is that the excess temporal variation
of the proportion of males is real. Anscombe (1981) and Hald (1990) suggest that
counting or copying errors in the data may account for the excess; indeed the four
most extreme observations account for one quarter of the value of %2. But omit-
ting those four years still leaves an excess that is significant at the 0.005 level. It
has been reported that the ratio of christenings to births was significantly reduced
during the two decades prior to the Restoration in 1660 (Hald 1990). If there was
a sex bias in the reduced christenings, that could explain part of the excess varia-
tion in the proportion of males. When the two decades preceding the Restoration
are omitted, the variance in the proportion of males is reduced by only 20% to
0.00004. Hence, omitting those years does not eliminate the excess variation.
In order to gain more insight into the nature of the variation of the sex ratio,
the correlation of the proportion of males with the number of christenings and

This content downloaded from


62.204.213.86 on Thu, 19 Nov 2020 10:41:09 UTC
All use subject to https://about.jstor.org/terms
Graunt, Arbuthnott, and the Human Sex Ratio / 609

number of deaths for the 82 years of Arbuthnotťs data, and with specific causes of
death (plague, fever, smallpox, measles, griping of the gut) for the years 1661 to
1686, was computed. Minitab was used for the calculations, and the adjusted co-
efficient of determination (which corrects for random associations) is used as the
measure of association. The strongest association is r1- 0.106 between the pro-
portion of males and the number of christenings. Furthermore, the coefficient of
determination between number of births and proportion of males when the two
decades prior to the Restoration are omitted is still r2 = 0.08, which buttresses the
notion that the association between the sex ratio and the birth rate is real. This as-
sociation can account for one-fifth of the excess variation in the sex ratio. The as-

sociation between the proportion of males and the number of deaths is only r2 =
0.025. Furthermore, there is a positive correlation (r = 0.343) between the number
of christenings and number of deaths, which is contrary to Graunťs conclusion
"that the more sickly the year is, the less fertile of births." There is only one sig-
nificant association between the proportion of males and a disease, which is r2 =
0.072 for measles.

A cross-correlation between the proportion of males and the number of


births was performed with the result that it peaked (r = -0.401) when the number
of births was measured two years before the proportion of males. This is not a
large difference; the correlation at a shift of 0 years is -0.346. But it is consistent
with time delay inherent in the unexplained increase in the sex ratio during and
after times of war (James 1987a), if war is interpreted as a cause of variation in
the sex ratio.

The first studies of the human sex ratio by Graunt and Arbuthnott are of in-
terest primarily for historical reasons. It merits mention that both are available on
the Web ( Natural and Political Observations made upon the Bills of Mortality at
http://www.ac.wwu.edu/~stephan/Graunt/bills.html, and "An Argument for Di-
vine Providence, Taken from the Constant Regularity Observed in the Births
of Both Sexes" at http://panoramix.univ-parisl.fr/CHPE/Textes/Arbuthnot/
arbuth.html). They cannot contribute to our understanding of genetic causes (Ed-
wards 1962) or the influence of hormone levels (James 1987b) on the sex ratio.
But they do at least evidence the relative stability of the sex ratio over several
centuries.

Received 3 April 2000; revision received 20 February 2001.

Literature Cited

Anscombe, F J. 1981. Computing in Statistical Science through APL. New York, NY: Springer- Verlag.
Arbuthnott, J. 1710. An argument for divine providence, taken from the constant regularity observed
in the births of both sexes. Philosophical Transactions 27:186-190.
Bellhouse, D.R. 1989. A manuscript written by John Arbuthnot. International Statistical Review
57:249-259.

This content downloaded from


62.204.213.86 on Thu, 19 Nov 2020 10:41:09 UTC
All use subject to https://about.jstor.org/terms
610 / CAMPBELL

Creighton, С. 1965. A History of Epidemics in Britain. Vols. 1 and 2. 2nd ed. New York, NY: Barnes
and Noble.
Edwards, A.W.F. 1962. Genetics and the human sex ratio. Advances in Genetics 11:239-272.
Eisenhart, С. and A. Birnbaum. 1967. Anniversaries in 1966-67 of interest to statisticians. The Ameri-
can Statistician 21:22-29.

Graunt, J. 1662. Natural and Political Observations made upon the Bills of Mortality. London, UK:
Martyn.
Hacking, I. 1975. The Emergence of Probability. London, UK: Cambridge University Press.
Hald, A. 1990. A History of Probability and Statistics and Their Applications before 1750. New York,
NY: John Wiley.
Harding, V. 1990. The population of London 1550-1700: A review of the published evidence. London
Journal 15:111-128.

James, W. H. 1987a. The human sex ratio, part 1: A review of the literature. Human Biology
59:721-752.

James, W. H. 1987b. The human sex ratio, part 2: A hypothesis and a program of research. Human Bi-
ology 59:873-900.
James, W. H. 2000. The variation of the probability of a son within and across couples. Human Re-
production 15:1184-1188.
Shoesmith, E. 1987. The continental controversy over Arbuthnot's argument for divine providence.
Historia Mathematica 14:133-146.

Weiss, N. A. 1999. Introductory Statistics. 5th ed. Reading, MA: Addison- Wesley.

This content downloaded from


62.204.213.86 on Thu, 19 Nov 2020 10:41:09 UTC
All use subject to https://about.jstor.org/terms

You might also like